1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper@@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -fno-pretty-templates @gol
193 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
194 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
195 -fno-default-inline -fvisibility-inlines-hidden @gol
196 -fvisibility-ms-compat @gol
197 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
198 -Wnon-virtual-dtor -Wreorder @gol
199 -Weffc++ -Wstrict-null-sentinel @gol
200 -Wno-non-template-friend -Wold-style-cast @gol
201 -Woverloaded-virtual -Wno-pmf-conversions @gol
204 @item Objective-C and Objective-C++ Language Options
205 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
206 Objective-C and Objective-C++ Dialects}.
207 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
208 -fgnu-runtime -fnext-runtime @gol
209 -fno-nil-receivers @gol
210 -fobjc-call-cxx-cdtors @gol
211 -fobjc-direct-dispatch @gol
212 -fobjc-exceptions @gol
214 -freplace-objc-classes @gol
217 -Wassign-intercept @gol
218 -Wno-protocol -Wselector @gol
219 -Wstrict-selector-match @gol
220 -Wundeclared-selector}
222 @item Language Independent Options
223 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
224 @gccoptlist{-fmessage-length=@var{n} @gol
225 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
226 -fdiagnostics-show-option}
228 @item Warning Options
229 @xref{Warning Options,,Options to Request or Suppress Warnings}.
230 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
231 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
232 -Wno-attributes -Wno-builtin-macro-redefined @gol
233 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
234 -Wchar-subscripts -Wclobbered -Wcomment @gol
235 -Wconversion -Wcoverage-mismatch -Wcpp -Wno-deprecated @gol
236 -Wno-deprecated-declarations -Wdisabled-optimization @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas @gol
263 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
264 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
265 -Wunused-but-set-parameter -Wunused-but-set-variable -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-nrv -fdump-tree-vect @gol
299 -fdump-tree-sink @gol
300 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
304 -ftree-vectorizer-verbose=@var{n} @gol
305 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
306 -fdump-final-insns=@var{file} @gol
307 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
308 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
309 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
310 -fenable-icf-debug @gol
311 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
312 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
313 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
314 -ftest-coverage -ftime-report -fvar-tracking @gol
315 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
316 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
317 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
318 -gvms -gxcoff -gxcoff+ @gol
319 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
320 -fdebug-prefix-map=@var{old}=@var{new} @gol
321 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
322 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
323 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
324 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
325 -print-prog-name=@var{program} -print-search-dirs -Q @gol
326 -print-sysroot -print-sysroot-headers-suffix @gol
327 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
329 @item Optimization Options
330 @xref{Optimize Options,,Options that Control Optimization}.
332 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
333 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
334 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
335 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
336 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
337 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
338 -fdata-sections -fdce -fdce @gol
339 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
340 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
341 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
342 -fforward-propagate -ffunction-sections @gol
343 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
344 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
345 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
346 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
347 -fipa-profile -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
348 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
349 -fira-region=@var{region} -fira-coalesce @gol
350 -fira-loop-pressure -fno-ira-share-save-slots @gol
351 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
352 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
353 -floop-block -floop-interchange -floop-strip-mine -fgraphite-identity @gol
354 -floop-parallelize-all -flto -flto-compression-level -flto-report -fltrans @gol
355 -fltrans-output-list -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
356 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
357 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
358 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
359 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
360 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
361 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
362 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
363 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
364 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
365 -fprofile-generate=@var{path} @gol
366 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
367 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
368 -freorder-blocks-and-partition -freorder-functions @gol
369 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
370 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
371 -fsched-spec-load -fsched-spec-load-dangerous @gol
372 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
373 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
374 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
375 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
376 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
377 -fselective-scheduling -fselective-scheduling2 @gol
378 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
379 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
380 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
381 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
382 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
383 -ftree-copyrename -ftree-dce @gol
384 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
385 -ftree-phiprop -ftree-loop-distribution @gol
386 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
387 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
388 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
389 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
390 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
391 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
392 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
393 -fwhole-program -fwhopr -fwpa -fuse-linker-plugin @gol
394 --param @var{name}=@var{value}
395 -O -O0 -O1 -O2 -O3 -Os}
397 @item Preprocessor Options
398 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
399 @gccoptlist{-A@var{question}=@var{answer} @gol
400 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
401 -C -dD -dI -dM -dN @gol
402 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
403 -idirafter @var{dir} @gol
404 -include @var{file} -imacros @var{file} @gol
405 -iprefix @var{file} -iwithprefix @var{dir} @gol
406 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
407 -imultilib @var{dir} -isysroot @var{dir} @gol
408 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
409 -P -fworking-directory -remap @gol
410 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
411 -Xpreprocessor @var{option}}
413 @item Assembler Option
414 @xref{Assembler Options,,Passing Options to the Assembler}.
415 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
418 @xref{Link Options,,Options for Linking}.
419 @gccoptlist{@var{object-file-name} -l@var{library} @gol
420 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
421 -s -static -static-libgcc -static-libstdc++ -shared @gol
422 -shared-libgcc -symbolic @gol
423 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
426 @item Directory Options
427 @xref{Directory Options,,Options for Directory Search}.
428 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir}}
429 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-
432 @item Machine Dependent Options
433 @xref{Submodel Options,,Hardware Models and Configurations}.
434 @c This list is ordered alphanumerically by subsection name.
435 @c Try and put the significant identifier (CPU or system) first,
436 @c so users have a clue at guessing where the ones they want will be.
439 @gccoptlist{-EB -EL @gol
440 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
441 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
444 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
445 -mabi=@var{name} @gol
446 -mapcs-stack-check -mno-apcs-stack-check @gol
447 -mapcs-float -mno-apcs-float @gol
448 -mapcs-reentrant -mno-apcs-reentrant @gol
449 -msched-prolog -mno-sched-prolog @gol
450 -mlittle-endian -mbig-endian -mwords-little-endian @gol
451 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
452 -mfp16-format=@var{name}
453 -mthumb-interwork -mno-thumb-interwork @gol
454 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
455 -mstructure-size-boundary=@var{n} @gol
456 -mabort-on-noreturn @gol
457 -mlong-calls -mno-long-calls @gol
458 -msingle-pic-base -mno-single-pic-base @gol
459 -mpic-register=@var{reg} @gol
460 -mnop-fun-dllimport @gol
461 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
462 -mpoke-function-name @gol
464 -mtpcs-frame -mtpcs-leaf-frame @gol
465 -mcaller-super-interworking -mcallee-super-interworking @gol
467 -mword-relocations @gol
468 -mfix-cortex-m3-ldrd}
471 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
472 -mcall-prologues -mtiny-stack -mint8}
474 @emph{Blackfin Options}
475 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
476 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
477 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
478 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
479 -mno-id-shared-library -mshared-library-id=@var{n} @gol
480 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
481 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
482 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
486 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
487 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
488 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
489 -mstack-align -mdata-align -mconst-align @gol
490 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
491 -melf -maout -melinux -mlinux -sim -sim2 @gol
492 -mmul-bug-workaround -mno-mul-bug-workaround}
495 @gccoptlist{-mmac -mpush-args}
497 @emph{Darwin Options}
498 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
499 -arch_only -bind_at_load -bundle -bundle_loader @gol
500 -client_name -compatibility_version -current_version @gol
502 -dependency-file -dylib_file -dylinker_install_name @gol
503 -dynamic -dynamiclib -exported_symbols_list @gol
504 -filelist -flat_namespace -force_cpusubtype_ALL @gol
505 -force_flat_namespace -headerpad_max_install_names @gol
507 -image_base -init -install_name -keep_private_externs @gol
508 -multi_module -multiply_defined -multiply_defined_unused @gol
509 -noall_load -no_dead_strip_inits_and_terms @gol
510 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
511 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
512 -private_bundle -read_only_relocs -sectalign @gol
513 -sectobjectsymbols -whyload -seg1addr @gol
514 -sectcreate -sectobjectsymbols -sectorder @gol
515 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
516 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
517 -segprot -segs_read_only_addr -segs_read_write_addr @gol
518 -single_module -static -sub_library -sub_umbrella @gol
519 -twolevel_namespace -umbrella -undefined @gol
520 -unexported_symbols_list -weak_reference_mismatches @gol
521 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
522 -mkernel -mone-byte-bool}
524 @emph{DEC Alpha Options}
525 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
526 -mieee -mieee-with-inexact -mieee-conformant @gol
527 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
528 -mtrap-precision=@var{mode} -mbuild-constants @gol
529 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
530 -mbwx -mmax -mfix -mcix @gol
531 -mfloat-vax -mfloat-ieee @gol
532 -mexplicit-relocs -msmall-data -mlarge-data @gol
533 -msmall-text -mlarge-text @gol
534 -mmemory-latency=@var{time}}
536 @emph{DEC Alpha/VMS Options}
537 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
540 @gccoptlist{-msmall-model -mno-lsim}
543 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
544 -mhard-float -msoft-float @gol
545 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
546 -mdouble -mno-double @gol
547 -mmedia -mno-media -mmuladd -mno-muladd @gol
548 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
549 -mlinked-fp -mlong-calls -malign-labels @gol
550 -mlibrary-pic -macc-4 -macc-8 @gol
551 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
552 -moptimize-membar -mno-optimize-membar @gol
553 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
554 -mvliw-branch -mno-vliw-branch @gol
555 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
556 -mno-nested-cond-exec -mtomcat-stats @gol
560 @emph{GNU/Linux Options}
561 @gccoptlist{-muclibc}
563 @emph{H8/300 Options}
564 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
567 @gccoptlist{-march=@var{architecture-type} @gol
568 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
569 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
570 -mfixed-range=@var{register-range} @gol
571 -mjump-in-delay -mlinker-opt -mlong-calls @gol
572 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
573 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
574 -mno-jump-in-delay -mno-long-load-store @gol
575 -mno-portable-runtime -mno-soft-float @gol
576 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
577 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
578 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
579 -munix=@var{unix-std} -nolibdld -static -threads}
581 @emph{i386 and x86-64 Options}
582 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
583 -mfpmath=@var{unit} @gol
584 -masm=@var{dialect} -mno-fancy-math-387 @gol
585 -mno-fp-ret-in-387 -msoft-float @gol
586 -mno-wide-multiply -mrtd -malign-double @gol
587 -mpreferred-stack-boundary=@var{num}
588 -mincoming-stack-boundary=@var{num}
589 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
590 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
591 -maes -mpclmul -mfused-madd @gol
592 -msse4a -m3dnow -mpopcnt -mabm -mfma4 -mxop -mlwp @gol
593 -mthreads -mno-align-stringops -minline-all-stringops @gol
594 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
595 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
596 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
597 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
598 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
599 -mcmodel=@var{code-model} -mabi=@var{name} @gol
600 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
604 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
605 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
606 -mconstant-gp -mauto-pic -mfused-madd @gol
607 -minline-float-divide-min-latency @gol
608 -minline-float-divide-max-throughput @gol
609 -mno-inline-float-divide @gol
610 -minline-int-divide-min-latency @gol
611 -minline-int-divide-max-throughput @gol
612 -mno-inline-int-divide @gol
613 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
614 -mno-inline-sqrt @gol
615 -mdwarf2-asm -mearly-stop-bits @gol
616 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
617 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
618 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
619 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
620 -msched-spec-ldc -msched-spec-control-ldc @gol
621 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
622 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
623 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
624 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
626 @emph{IA-64/VMS Options}
627 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
630 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
631 -msign-extend-enabled -muser-enabled}
633 @emph{M32R/D Options}
634 @gccoptlist{-m32r2 -m32rx -m32r @gol
636 -malign-loops -mno-align-loops @gol
637 -missue-rate=@var{number} @gol
638 -mbranch-cost=@var{number} @gol
639 -mmodel=@var{code-size-model-type} @gol
640 -msdata=@var{sdata-type} @gol
641 -mno-flush-func -mflush-func=@var{name} @gol
642 -mno-flush-trap -mflush-trap=@var{number} @gol
646 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
648 @emph{M680x0 Options}
649 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
650 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
651 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
652 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
653 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
654 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
655 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
656 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
659 @emph{M68hc1x Options}
660 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
661 -mauto-incdec -minmax -mlong-calls -mshort @gol
662 -msoft-reg-count=@var{count}}
665 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
666 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
667 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
668 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
669 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
672 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
673 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
674 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
675 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
679 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
680 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
681 -mips64 -mips64r2 @gol
682 -mips16 -mno-mips16 -mflip-mips16 @gol
683 -minterlink-mips16 -mno-interlink-mips16 @gol
684 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
685 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
686 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
687 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
688 -mfpu=@var{fpu-type} @gol
689 -msmartmips -mno-smartmips @gol
690 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
691 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
692 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
693 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
694 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
695 -membedded-data -mno-embedded-data @gol
696 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
697 -mcode-readable=@var{setting} @gol
698 -msplit-addresses -mno-split-addresses @gol
699 -mexplicit-relocs -mno-explicit-relocs @gol
700 -mcheck-zero-division -mno-check-zero-division @gol
701 -mdivide-traps -mdivide-breaks @gol
702 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
703 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
704 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
705 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
706 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
707 -mflush-func=@var{func} -mno-flush-func @gol
708 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
709 -mfp-exceptions -mno-fp-exceptions @gol
710 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
711 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
714 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
715 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
716 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
717 -mno-base-addresses -msingle-exit -mno-single-exit}
719 @emph{MN10300 Options}
720 @gccoptlist{-mmult-bug -mno-mult-bug @gol
721 -mam33 -mno-am33 @gol
722 -mam33-2 -mno-am33-2 @gol
723 -mreturn-pointer-on-d0 @gol
726 @emph{PDP-11 Options}
727 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
728 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
729 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
730 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
731 -mbranch-expensive -mbranch-cheap @gol
732 -msplit -mno-split -munix-asm -mdec-asm}
734 @emph{picoChip Options}
735 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
736 -msymbol-as-address -mno-inefficient-warnings}
738 @emph{PowerPC Options}
739 See RS/6000 and PowerPC Options.
741 @emph{RS/6000 and PowerPC Options}
742 @gccoptlist{-mcpu=@var{cpu-type} @gol
743 -mtune=@var{cpu-type} @gol
744 -mpower -mno-power -mpower2 -mno-power2 @gol
745 -mpowerpc -mpowerpc64 -mno-powerpc @gol
746 -maltivec -mno-altivec @gol
747 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
748 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
749 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
750 -mfprnd -mno-fprnd @gol
751 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
752 -mnew-mnemonics -mold-mnemonics @gol
753 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
754 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
755 -malign-power -malign-natural @gol
756 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
757 -msingle-float -mdouble-float -msimple-fpu @gol
758 -mstring -mno-string -mupdate -mno-update @gol
759 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
760 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
761 -mstrict-align -mno-strict-align -mrelocatable @gol
762 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
763 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
764 -mdynamic-no-pic -maltivec -mswdiv @gol
765 -mprioritize-restricted-insns=@var{priority} @gol
766 -msched-costly-dep=@var{dependence_type} @gol
767 -minsert-sched-nops=@var{scheme} @gol
768 -mcall-sysv -mcall-netbsd @gol
769 -maix-struct-return -msvr4-struct-return @gol
770 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
771 -misel -mno-isel @gol
772 -misel=yes -misel=no @gol
774 -mspe=yes -mspe=no @gol
776 -mgen-cell-microcode -mwarn-cell-microcode @gol
777 -mvrsave -mno-vrsave @gol
778 -mmulhw -mno-mulhw @gol
779 -mdlmzb -mno-dlmzb @gol
780 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
781 -mprototype -mno-prototype @gol
782 -msim -mmvme -mads -myellowknife -memb -msdata @gol
783 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
786 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
788 -mbig-endian-data -mlittle-endian-data @gol
791 -mas100-syntax -mno-as100-syntax@gol
793 -mmax-constant-size=@gol
795 -msave-acc-in-interrupts}
797 @emph{S/390 and zSeries Options}
798 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
799 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
800 -mlong-double-64 -mlong-double-128 @gol
801 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
802 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
803 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
804 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
805 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
808 @gccoptlist{-meb -mel @gol
812 -mscore5 -mscore5u -mscore7 -mscore7d}
815 @gccoptlist{-m1 -m2 -m2e @gol
816 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
818 -m4-nofpu -m4-single-only -m4-single -m4 @gol
819 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
820 -m5-64media -m5-64media-nofpu @gol
821 -m5-32media -m5-32media-nofpu @gol
822 -m5-compact -m5-compact-nofpu @gol
823 -mb -ml -mdalign -mrelax @gol
824 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
825 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
826 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
827 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
828 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
829 -maccumulate-outgoing-args -minvalid-symbols}
832 @gccoptlist{-mcpu=@var{cpu-type} @gol
833 -mtune=@var{cpu-type} @gol
834 -mcmodel=@var{code-model} @gol
835 -m32 -m64 -mapp-regs -mno-app-regs @gol
836 -mfaster-structs -mno-faster-structs @gol
837 -mfpu -mno-fpu -mhard-float -msoft-float @gol
838 -mhard-quad-float -msoft-quad-float @gol
839 -mimpure-text -mno-impure-text -mlittle-endian @gol
840 -mstack-bias -mno-stack-bias @gol
841 -munaligned-doubles -mno-unaligned-doubles @gol
842 -mv8plus -mno-v8plus -mvis -mno-vis
843 -threads -pthreads -pthread}
846 @gccoptlist{-mwarn-reloc -merror-reloc @gol
847 -msafe-dma -munsafe-dma @gol
849 -msmall-mem -mlarge-mem -mstdmain @gol
850 -mfixed-range=@var{register-range} @gol
852 -maddress-space-conversion -mno-address-space-conversion @gol
853 -mcache-size=@var{cache-size} @gol
854 -matomic-updates -mno-atomic-updates}
856 @emph{System V Options}
857 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
860 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
861 -mprolog-function -mno-prolog-function -mspace @gol
862 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
863 -mapp-regs -mno-app-regs @gol
864 -mdisable-callt -mno-disable-callt @gol
870 @gccoptlist{-mg -mgnu -munix}
872 @emph{VxWorks Options}
873 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
874 -Xbind-lazy -Xbind-now}
876 @emph{x86-64 Options}
877 See i386 and x86-64 Options.
879 @emph{i386 and x86-64 Windows Options}
880 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
881 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows
882 -fno-set-stack-executable}
884 @emph{Xstormy16 Options}
887 @emph{Xtensa Options}
888 @gccoptlist{-mconst16 -mno-const16 @gol
889 -mfused-madd -mno-fused-madd @gol
890 -mserialize-volatile -mno-serialize-volatile @gol
891 -mtext-section-literals -mno-text-section-literals @gol
892 -mtarget-align -mno-target-align @gol
893 -mlongcalls -mno-longcalls}
895 @emph{zSeries Options}
896 See S/390 and zSeries Options.
898 @item Code Generation Options
899 @xref{Code Gen Options,,Options for Code Generation Conventions}.
900 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
901 -ffixed-@var{reg} -fexceptions @gol
902 -fnon-call-exceptions -funwind-tables @gol
903 -fasynchronous-unwind-tables @gol
904 -finhibit-size-directive -finstrument-functions @gol
905 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
906 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
907 -fno-common -fno-ident @gol
908 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
909 -fno-jump-tables @gol
910 -frecord-gcc-switches @gol
911 -freg-struct-return -fshort-enums @gol
912 -fshort-double -fshort-wchar @gol
913 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
914 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
915 -fno-stack-limit @gol
916 -fleading-underscore -ftls-model=@var{model} @gol
917 -ftrapv -fwrapv -fbounds-check @gol
922 * Overall Options:: Controlling the kind of output:
923 an executable, object files, assembler files,
924 or preprocessed source.
925 * C Dialect Options:: Controlling the variant of C language compiled.
926 * C++ Dialect Options:: Variations on C++.
927 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
929 * Language Independent Options:: Controlling how diagnostics should be
931 * Warning Options:: How picky should the compiler be?
932 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
933 * Optimize Options:: How much optimization?
934 * Preprocessor Options:: Controlling header files and macro definitions.
935 Also, getting dependency information for Make.
936 * Assembler Options:: Passing options to the assembler.
937 * Link Options:: Specifying libraries and so on.
938 * Directory Options:: Where to find header files and libraries.
939 Where to find the compiler executable files.
940 * Spec Files:: How to pass switches to sub-processes.
941 * Target Options:: Running a cross-compiler, or an old version of GCC.
944 @node Overall Options
945 @section Options Controlling the Kind of Output
947 Compilation can involve up to four stages: preprocessing, compilation
948 proper, assembly and linking, always in that order. GCC is capable of
949 preprocessing and compiling several files either into several
950 assembler input files, or into one assembler input file; then each
951 assembler input file produces an object file, and linking combines all
952 the object files (those newly compiled, and those specified as input)
953 into an executable file.
955 @cindex file name suffix
956 For any given input file, the file name suffix determines what kind of
961 C source code which must be preprocessed.
964 C source code which should not be preprocessed.
967 C++ source code which should not be preprocessed.
970 Objective-C source code. Note that you must link with the @file{libobjc}
971 library to make an Objective-C program work.
974 Objective-C source code which should not be preprocessed.
978 Objective-C++ source code. Note that you must link with the @file{libobjc}
979 library to make an Objective-C++ program work. Note that @samp{.M} refers
980 to a literal capital M@.
983 Objective-C++ source code which should not be preprocessed.
986 C, C++, Objective-C or Objective-C++ header file to be turned into a
991 @itemx @var{file}.cxx
992 @itemx @var{file}.cpp
993 @itemx @var{file}.CPP
994 @itemx @var{file}.c++
996 C++ source code which must be preprocessed. Note that in @samp{.cxx},
997 the last two letters must both be literally @samp{x}. Likewise,
998 @samp{.C} refers to a literal capital C@.
1002 Objective-C++ source code which must be preprocessed.
1004 @item @var{file}.mii
1005 Objective-C++ source code which should not be preprocessed.
1009 @itemx @var{file}.hp
1010 @itemx @var{file}.hxx
1011 @itemx @var{file}.hpp
1012 @itemx @var{file}.HPP
1013 @itemx @var{file}.h++
1014 @itemx @var{file}.tcc
1015 C++ header file to be turned into a precompiled header.
1018 @itemx @var{file}.for
1019 @itemx @var{file}.ftn
1020 Fixed form Fortran source code which should not be preprocessed.
1023 @itemx @var{file}.FOR
1024 @itemx @var{file}.fpp
1025 @itemx @var{file}.FPP
1026 @itemx @var{file}.FTN
1027 Fixed form Fortran source code which must be preprocessed (with the traditional
1030 @item @var{file}.f90
1031 @itemx @var{file}.f95
1032 @itemx @var{file}.f03
1033 @itemx @var{file}.f08
1034 Free form Fortran source code which should not be preprocessed.
1036 @item @var{file}.F90
1037 @itemx @var{file}.F95
1038 @itemx @var{file}.F03
1039 @itemx @var{file}.F08
1040 Free form Fortran source code which must be preprocessed (with the
1041 traditional preprocessor).
1043 @c FIXME: Descriptions of Java file types.
1049 @item @var{file}.ads
1050 Ada source code file which contains a library unit declaration (a
1051 declaration of a package, subprogram, or generic, or a generic
1052 instantiation), or a library unit renaming declaration (a package,
1053 generic, or subprogram renaming declaration). Such files are also
1056 @item @var{file}.adb
1057 Ada source code file containing a library unit body (a subprogram or
1058 package body). Such files are also called @dfn{bodies}.
1060 @c GCC also knows about some suffixes for languages not yet included:
1071 @itemx @var{file}.sx
1072 Assembler code which must be preprocessed.
1075 An object file to be fed straight into linking.
1076 Any file name with no recognized suffix is treated this way.
1080 You can specify the input language explicitly with the @option{-x} option:
1083 @item -x @var{language}
1084 Specify explicitly the @var{language} for the following input files
1085 (rather than letting the compiler choose a default based on the file
1086 name suffix). This option applies to all following input files until
1087 the next @option{-x} option. Possible values for @var{language} are:
1089 c c-header c-cpp-output
1090 c++ c++-header c++-cpp-output
1091 objective-c objective-c-header objective-c-cpp-output
1092 objective-c++ objective-c++-header objective-c++-cpp-output
1093 assembler assembler-with-cpp
1095 f77 f77-cpp-input f95 f95-cpp-input
1100 Turn off any specification of a language, so that subsequent files are
1101 handled according to their file name suffixes (as they are if @option{-x}
1102 has not been used at all).
1104 @item -pass-exit-codes
1105 @opindex pass-exit-codes
1106 Normally the @command{gcc} program will exit with the code of 1 if any
1107 phase of the compiler returns a non-success return code. If you specify
1108 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1109 numerically highest error produced by any phase that returned an error
1110 indication. The C, C++, and Fortran frontends return 4, if an internal
1111 compiler error is encountered.
1114 If you only want some of the stages of compilation, you can use
1115 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1116 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1117 @command{gcc} is to stop. Note that some combinations (for example,
1118 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1123 Compile or assemble the source files, but do not link. The linking
1124 stage simply is not done. The ultimate output is in the form of an
1125 object file for each source file.
1127 By default, the object file name for a source file is made by replacing
1128 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1130 Unrecognized input files, not requiring compilation or assembly, are
1135 Stop after the stage of compilation proper; do not assemble. The output
1136 is in the form of an assembler code file for each non-assembler input
1139 By default, the assembler file name for a source file is made by
1140 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1142 Input files that don't require compilation are ignored.
1146 Stop after the preprocessing stage; do not run the compiler proper. The
1147 output is in the form of preprocessed source code, which is sent to the
1150 Input files which don't require preprocessing are ignored.
1152 @cindex output file option
1155 Place output in file @var{file}. This applies regardless to whatever
1156 sort of output is being produced, whether it be an executable file,
1157 an object file, an assembler file or preprocessed C code.
1159 If @option{-o} is not specified, the default is to put an executable
1160 file in @file{a.out}, the object file for
1161 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1162 assembler file in @file{@var{source}.s}, a precompiled header file in
1163 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1168 Print (on standard error output) the commands executed to run the stages
1169 of compilation. Also print the version number of the compiler driver
1170 program and of the preprocessor and the compiler proper.
1174 Like @option{-v} except the commands are not executed and all command
1175 arguments are quoted. This is useful for shell scripts to capture the
1176 driver-generated command lines.
1180 Use pipes rather than temporary files for communication between the
1181 various stages of compilation. This fails to work on some systems where
1182 the assembler is unable to read from a pipe; but the GNU assembler has
1187 If you are compiling multiple source files, this option tells the driver
1188 to pass all the source files to the compiler at once (for those
1189 languages for which the compiler can handle this). This will allow
1190 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1191 language for which this is supported is C@. If you pass source files for
1192 multiple languages to the driver, using this option, the driver will invoke
1193 the compiler(s) that support IMA once each, passing each compiler all the
1194 source files appropriate for it. For those languages that do not support
1195 IMA this option will be ignored, and the compiler will be invoked once for
1196 each source file in that language. If you use this option in conjunction
1197 with @option{-save-temps}, the compiler will generate multiple
1199 (one for each source file), but only one (combined) @file{.o} or
1204 Print (on the standard output) a description of the command line options
1205 understood by @command{gcc}. If the @option{-v} option is also specified
1206 then @option{--help} will also be passed on to the various processes
1207 invoked by @command{gcc}, so that they can display the command line options
1208 they accept. If the @option{-Wextra} option has also been specified
1209 (prior to the @option{--help} option), then command line options which
1210 have no documentation associated with them will also be displayed.
1213 @opindex target-help
1214 Print (on the standard output) a description of target-specific command
1215 line options for each tool. For some targets extra target-specific
1216 information may also be printed.
1218 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1219 Print (on the standard output) a description of the command line
1220 options understood by the compiler that fit into all specified classes
1221 and qualifiers. These are the supported classes:
1224 @item @samp{optimizers}
1225 This will display all of the optimization options supported by the
1228 @item @samp{warnings}
1229 This will display all of the options controlling warning messages
1230 produced by the compiler.
1233 This will display target-specific options. Unlike the
1234 @option{--target-help} option however, target-specific options of the
1235 linker and assembler will not be displayed. This is because those
1236 tools do not currently support the extended @option{--help=} syntax.
1239 This will display the values recognized by the @option{--param}
1242 @item @var{language}
1243 This will display the options supported for @var{language}, where
1244 @var{language} is the name of one of the languages supported in this
1248 This will display the options that are common to all languages.
1251 These are the supported qualifiers:
1254 @item @samp{undocumented}
1255 Display only those options which are undocumented.
1258 Display options which take an argument that appears after an equal
1259 sign in the same continuous piece of text, such as:
1260 @samp{--help=target}.
1262 @item @samp{separate}
1263 Display options which take an argument that appears as a separate word
1264 following the original option, such as: @samp{-o output-file}.
1267 Thus for example to display all the undocumented target-specific
1268 switches supported by the compiler the following can be used:
1271 --help=target,undocumented
1274 The sense of a qualifier can be inverted by prefixing it with the
1275 @samp{^} character, so for example to display all binary warning
1276 options (i.e., ones that are either on or off and that do not take an
1277 argument), which have a description the following can be used:
1280 --help=warnings,^joined,^undocumented
1283 The argument to @option{--help=} should not consist solely of inverted
1286 Combining several classes is possible, although this usually
1287 restricts the output by so much that there is nothing to display. One
1288 case where it does work however is when one of the classes is
1289 @var{target}. So for example to display all the target-specific
1290 optimization options the following can be used:
1293 --help=target,optimizers
1296 The @option{--help=} option can be repeated on the command line. Each
1297 successive use will display its requested class of options, skipping
1298 those that have already been displayed.
1300 If the @option{-Q} option appears on the command line before the
1301 @option{--help=} option, then the descriptive text displayed by
1302 @option{--help=} is changed. Instead of describing the displayed
1303 options, an indication is given as to whether the option is enabled,
1304 disabled or set to a specific value (assuming that the compiler
1305 knows this at the point where the @option{--help=} option is used).
1307 Here is a truncated example from the ARM port of @command{gcc}:
1310 % gcc -Q -mabi=2 --help=target -c
1311 The following options are target specific:
1313 -mabort-on-noreturn [disabled]
1317 The output is sensitive to the effects of previous command line
1318 options, so for example it is possible to find out which optimizations
1319 are enabled at @option{-O2} by using:
1322 -Q -O2 --help=optimizers
1325 Alternatively you can discover which binary optimizations are enabled
1326 by @option{-O3} by using:
1329 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1330 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1331 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1334 @item -no-canonical-prefixes
1335 @opindex no-canonical-prefixes
1336 Do not expand any symbolic links, resolve references to @samp{/../}
1337 or @samp{/./}, or make the path absolute when generating a relative
1342 Display the version number and copyrights of the invoked GCC@.
1346 Invoke all subcommands under a wrapper program. It takes a single
1347 comma separated list as an argument, which will be used to invoke
1351 gcc -c t.c -wrapper gdb,--args
1354 This will invoke all subprograms of gcc under "gdb --args",
1355 thus cc1 invocation will be "gdb --args cc1 ...".
1357 @item -fplugin=@var{name}.so
1358 Load the plugin code in file @var{name}.so, assumed to be a
1359 shared object to be dlopen'd by the compiler. The base name of
1360 the shared object file is used to identify the plugin for the
1361 purposes of argument parsing (See
1362 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1363 Each plugin should define the callback functions specified in the
1366 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1367 Define an argument called @var{key} with a value of @var{value}
1368 for the plugin called @var{name}.
1370 @include @value{srcdir}/../libiberty/at-file.texi
1374 @section Compiling C++ Programs
1376 @cindex suffixes for C++ source
1377 @cindex C++ source file suffixes
1378 C++ source files conventionally use one of the suffixes @samp{.C},
1379 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1380 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1381 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1382 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1383 files with these names and compiles them as C++ programs even if you
1384 call the compiler the same way as for compiling C programs (usually
1385 with the name @command{gcc}).
1389 However, the use of @command{gcc} does not add the C++ library.
1390 @command{g++} is a program that calls GCC and treats @samp{.c},
1391 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1392 files unless @option{-x} is used, and automatically specifies linking
1393 against the C++ library. This program is also useful when
1394 precompiling a C header file with a @samp{.h} extension for use in C++
1395 compilations. On many systems, @command{g++} is also installed with
1396 the name @command{c++}.
1398 @cindex invoking @command{g++}
1399 When you compile C++ programs, you may specify many of the same
1400 command-line options that you use for compiling programs in any
1401 language; or command-line options meaningful for C and related
1402 languages; or options that are meaningful only for C++ programs.
1403 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1404 explanations of options for languages related to C@.
1405 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1406 explanations of options that are meaningful only for C++ programs.
1408 @node C Dialect Options
1409 @section Options Controlling C Dialect
1410 @cindex dialect options
1411 @cindex language dialect options
1412 @cindex options, dialect
1414 The following options control the dialect of C (or languages derived
1415 from C, such as C++, Objective-C and Objective-C++) that the compiler
1419 @cindex ANSI support
1423 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1424 equivalent to @samp{-std=c++98}.
1426 This turns off certain features of GCC that are incompatible with ISO
1427 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1428 such as the @code{asm} and @code{typeof} keywords, and
1429 predefined macros such as @code{unix} and @code{vax} that identify the
1430 type of system you are using. It also enables the undesirable and
1431 rarely used ISO trigraph feature. For the C compiler,
1432 it disables recognition of C++ style @samp{//} comments as well as
1433 the @code{inline} keyword.
1435 The alternate keywords @code{__asm__}, @code{__extension__},
1436 @code{__inline__} and @code{__typeof__} continue to work despite
1437 @option{-ansi}. You would not want to use them in an ISO C program, of
1438 course, but it is useful to put them in header files that might be included
1439 in compilations done with @option{-ansi}. Alternate predefined macros
1440 such as @code{__unix__} and @code{__vax__} are also available, with or
1441 without @option{-ansi}.
1443 The @option{-ansi} option does not cause non-ISO programs to be
1444 rejected gratuitously. For that, @option{-pedantic} is required in
1445 addition to @option{-ansi}. @xref{Warning Options}.
1447 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1448 option is used. Some header files may notice this macro and refrain
1449 from declaring certain functions or defining certain macros that the
1450 ISO standard doesn't call for; this is to avoid interfering with any
1451 programs that might use these names for other things.
1453 Functions that would normally be built in but do not have semantics
1454 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1455 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1456 built-in functions provided by GCC}, for details of the functions
1461 Determine the language standard. @xref{Standards,,Language Standards
1462 Supported by GCC}, for details of these standard versions. This option
1463 is currently only supported when compiling C or C++.
1465 The compiler can accept several base standards, such as @samp{c90} or
1466 @samp{c++98}, and GNU dialects of those standards, such as
1467 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1468 compiler will accept all programs following that standard and those
1469 using GNU extensions that do not contradict it. For example,
1470 @samp{-std=c90} turns off certain features of GCC that are
1471 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1472 keywords, but not other GNU extensions that do not have a meaning in
1473 ISO C90, such as omitting the middle term of a @code{?:}
1474 expression. On the other hand, by specifying a GNU dialect of a
1475 standard, all features the compiler support are enabled, even when
1476 those features change the meaning of the base standard and some
1477 strict-conforming programs may be rejected. The particular standard
1478 is used by @option{-pedantic} to identify which features are GNU
1479 extensions given that version of the standard. For example
1480 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1481 comments, while @samp{-std=gnu99 -pedantic} would not.
1483 A value for this option must be provided; possible values are
1489 Support all ISO C90 programs (certain GNU extensions that conflict
1490 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1492 @item iso9899:199409
1493 ISO C90 as modified in amendment 1.
1499 ISO C99. Note that this standard is not yet fully supported; see
1500 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1501 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1504 ISO C1X, the draft of the next revision of the ISO C standard.
1505 Support is limited and experimental and features enabled by this
1506 option may be changed or removed if changed in or removed from the
1511 GNU dialect of ISO C90 (including some C99 features). This
1512 is the default for C code.
1516 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1517 this will become the default. The name @samp{gnu9x} is deprecated.
1520 GNU dialect of ISO C1X. Support is limited and experimental and
1521 features enabled by this option may be changed or removed if changed
1522 in or removed from the standard draft.
1525 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1529 GNU dialect of @option{-std=c++98}. This is the default for
1533 The working draft of the upcoming ISO C++0x standard. This option
1534 enables experimental features that are likely to be included in
1535 C++0x. The working draft is constantly changing, and any feature that is
1536 enabled by this flag may be removed from future versions of GCC if it is
1537 not part of the C++0x standard.
1540 GNU dialect of @option{-std=c++0x}. This option enables
1541 experimental features that may be removed in future versions of GCC.
1544 @item -fgnu89-inline
1545 @opindex fgnu89-inline
1546 The option @option{-fgnu89-inline} tells GCC to use the traditional
1547 GNU semantics for @code{inline} functions when in C99 mode.
1548 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1549 is accepted and ignored by GCC versions 4.1.3 up to but not including
1550 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1551 C99 mode. Using this option is roughly equivalent to adding the
1552 @code{gnu_inline} function attribute to all inline functions
1553 (@pxref{Function Attributes}).
1555 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1556 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1557 specifies the default behavior). This option was first supported in
1558 GCC 4.3. This option is not supported in @option{-std=c90} or
1559 @option{-std=gnu90} mode.
1561 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1562 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1563 in effect for @code{inline} functions. @xref{Common Predefined
1564 Macros,,,cpp,The C Preprocessor}.
1566 @item -aux-info @var{filename}
1568 Output to the given filename prototyped declarations for all functions
1569 declared and/or defined in a translation unit, including those in header
1570 files. This option is silently ignored in any language other than C@.
1572 Besides declarations, the file indicates, in comments, the origin of
1573 each declaration (source file and line), whether the declaration was
1574 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1575 @samp{O} for old, respectively, in the first character after the line
1576 number and the colon), and whether it came from a declaration or a
1577 definition (@samp{C} or @samp{F}, respectively, in the following
1578 character). In the case of function definitions, a K&R-style list of
1579 arguments followed by their declarations is also provided, inside
1580 comments, after the declaration.
1584 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1585 keyword, so that code can use these words as identifiers. You can use
1586 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1587 instead. @option{-ansi} implies @option{-fno-asm}.
1589 In C++, this switch only affects the @code{typeof} keyword, since
1590 @code{asm} and @code{inline} are standard keywords. You may want to
1591 use the @option{-fno-gnu-keywords} flag instead, which has the same
1592 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1593 switch only affects the @code{asm} and @code{typeof} keywords, since
1594 @code{inline} is a standard keyword in ISO C99.
1597 @itemx -fno-builtin-@var{function}
1598 @opindex fno-builtin
1599 @cindex built-in functions
1600 Don't recognize built-in functions that do not begin with
1601 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1602 functions provided by GCC}, for details of the functions affected,
1603 including those which are not built-in functions when @option{-ansi} or
1604 @option{-std} options for strict ISO C conformance are used because they
1605 do not have an ISO standard meaning.
1607 GCC normally generates special code to handle certain built-in functions
1608 more efficiently; for instance, calls to @code{alloca} may become single
1609 instructions that adjust the stack directly, and calls to @code{memcpy}
1610 may become inline copy loops. The resulting code is often both smaller
1611 and faster, but since the function calls no longer appear as such, you
1612 cannot set a breakpoint on those calls, nor can you change the behavior
1613 of the functions by linking with a different library. In addition,
1614 when a function is recognized as a built-in function, GCC may use
1615 information about that function to warn about problems with calls to
1616 that function, or to generate more efficient code, even if the
1617 resulting code still contains calls to that function. For example,
1618 warnings are given with @option{-Wformat} for bad calls to
1619 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1620 known not to modify global memory.
1622 With the @option{-fno-builtin-@var{function}} option
1623 only the built-in function @var{function} is
1624 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1625 function is named that is not built-in in this version of GCC, this
1626 option is ignored. There is no corresponding
1627 @option{-fbuiltin-@var{function}} option; if you wish to enable
1628 built-in functions selectively when using @option{-fno-builtin} or
1629 @option{-ffreestanding}, you may define macros such as:
1632 #define abs(n) __builtin_abs ((n))
1633 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1638 @cindex hosted environment
1640 Assert that compilation takes place in a hosted environment. This implies
1641 @option{-fbuiltin}. A hosted environment is one in which the
1642 entire standard library is available, and in which @code{main} has a return
1643 type of @code{int}. Examples are nearly everything except a kernel.
1644 This is equivalent to @option{-fno-freestanding}.
1646 @item -ffreestanding
1647 @opindex ffreestanding
1648 @cindex hosted environment
1650 Assert that compilation takes place in a freestanding environment. This
1651 implies @option{-fno-builtin}. A freestanding environment
1652 is one in which the standard library may not exist, and program startup may
1653 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1654 This is equivalent to @option{-fno-hosted}.
1656 @xref{Standards,,Language Standards Supported by GCC}, for details of
1657 freestanding and hosted environments.
1661 @cindex openmp parallel
1662 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1663 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1664 compiler generates parallel code according to the OpenMP Application
1665 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1666 implies @option{-pthread}, and thus is only supported on targets that
1667 have support for @option{-pthread}.
1669 @item -fms-extensions
1670 @opindex fms-extensions
1671 Accept some non-standard constructs used in Microsoft header files.
1673 Some cases of unnamed fields in structures and unions are only
1674 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1675 fields within structs/unions}, for details.
1679 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1680 options for strict ISO C conformance) implies @option{-trigraphs}.
1682 @item -no-integrated-cpp
1683 @opindex no-integrated-cpp
1684 Performs a compilation in two passes: preprocessing and compiling. This
1685 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1686 @option{-B} option. The user supplied compilation step can then add in
1687 an additional preprocessing step after normal preprocessing but before
1688 compiling. The default is to use the integrated cpp (internal cpp)
1690 The semantics of this option will change if "cc1", "cc1plus", and
1691 "cc1obj" are merged.
1693 @cindex traditional C language
1694 @cindex C language, traditional
1696 @itemx -traditional-cpp
1697 @opindex traditional-cpp
1698 @opindex traditional
1699 Formerly, these options caused GCC to attempt to emulate a pre-standard
1700 C compiler. They are now only supported with the @option{-E} switch.
1701 The preprocessor continues to support a pre-standard mode. See the GNU
1702 CPP manual for details.
1704 @item -fcond-mismatch
1705 @opindex fcond-mismatch
1706 Allow conditional expressions with mismatched types in the second and
1707 third arguments. The value of such an expression is void. This option
1708 is not supported for C++.
1710 @item -flax-vector-conversions
1711 @opindex flax-vector-conversions
1712 Allow implicit conversions between vectors with differing numbers of
1713 elements and/or incompatible element types. This option should not be
1716 @item -funsigned-char
1717 @opindex funsigned-char
1718 Let the type @code{char} be unsigned, like @code{unsigned char}.
1720 Each kind of machine has a default for what @code{char} should
1721 be. It is either like @code{unsigned char} by default or like
1722 @code{signed char} by default.
1724 Ideally, a portable program should always use @code{signed char} or
1725 @code{unsigned char} when it depends on the signedness of an object.
1726 But many programs have been written to use plain @code{char} and
1727 expect it to be signed, or expect it to be unsigned, depending on the
1728 machines they were written for. This option, and its inverse, let you
1729 make such a program work with the opposite default.
1731 The type @code{char} is always a distinct type from each of
1732 @code{signed char} or @code{unsigned char}, even though its behavior
1733 is always just like one of those two.
1736 @opindex fsigned-char
1737 Let the type @code{char} be signed, like @code{signed char}.
1739 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1740 the negative form of @option{-funsigned-char}. Likewise, the option
1741 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1743 @item -fsigned-bitfields
1744 @itemx -funsigned-bitfields
1745 @itemx -fno-signed-bitfields
1746 @itemx -fno-unsigned-bitfields
1747 @opindex fsigned-bitfields
1748 @opindex funsigned-bitfields
1749 @opindex fno-signed-bitfields
1750 @opindex fno-unsigned-bitfields
1751 These options control whether a bit-field is signed or unsigned, when the
1752 declaration does not use either @code{signed} or @code{unsigned}. By
1753 default, such a bit-field is signed, because this is consistent: the
1754 basic integer types such as @code{int} are signed types.
1757 @node C++ Dialect Options
1758 @section Options Controlling C++ Dialect
1760 @cindex compiler options, C++
1761 @cindex C++ options, command line
1762 @cindex options, C++
1763 This section describes the command-line options that are only meaningful
1764 for C++ programs; but you can also use most of the GNU compiler options
1765 regardless of what language your program is in. For example, you
1766 might compile a file @code{firstClass.C} like this:
1769 g++ -g -frepo -O -c firstClass.C
1773 In this example, only @option{-frepo} is an option meant
1774 only for C++ programs; you can use the other options with any
1775 language supported by GCC@.
1777 Here is a list of options that are @emph{only} for compiling C++ programs:
1781 @item -fabi-version=@var{n}
1782 @opindex fabi-version
1783 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1784 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1785 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1786 the version that conforms most closely to the C++ ABI specification.
1787 Therefore, the ABI obtained using version 0 will change as ABI bugs
1790 The default is version 2.
1792 Version 3 corrects an error in mangling a constant address as a
1795 Version 4 implements a standard mangling for vector types.
1797 See also @option{-Wabi}.
1799 @item -fno-access-control
1800 @opindex fno-access-control
1801 Turn off all access checking. This switch is mainly useful for working
1802 around bugs in the access control code.
1806 Check that the pointer returned by @code{operator new} is non-null
1807 before attempting to modify the storage allocated. This check is
1808 normally unnecessary because the C++ standard specifies that
1809 @code{operator new} will only return @code{0} if it is declared
1810 @samp{throw()}, in which case the compiler will always check the
1811 return value even without this option. In all other cases, when
1812 @code{operator new} has a non-empty exception specification, memory
1813 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1814 @samp{new (nothrow)}.
1816 @item -fconserve-space
1817 @opindex fconserve-space
1818 Put uninitialized or runtime-initialized global variables into the
1819 common segment, as C does. This saves space in the executable at the
1820 cost of not diagnosing duplicate definitions. If you compile with this
1821 flag and your program mysteriously crashes after @code{main()} has
1822 completed, you may have an object that is being destroyed twice because
1823 two definitions were merged.
1825 This option is no longer useful on most targets, now that support has
1826 been added for putting variables into BSS without making them common.
1828 @item -fno-deduce-init-list
1829 @opindex fno-deduce-init-list
1830 Disable deduction of a template type parameter as
1831 std::initializer_list from a brace-enclosed initializer list, i.e.
1834 template <class T> auto forward(T t) -> decltype (realfn (t))
1841 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1845 This option is present because this deduction is an extension to the
1846 current specification in the C++0x working draft, and there was
1847 some concern about potential overload resolution problems.
1849 @item -ffriend-injection
1850 @opindex ffriend-injection
1851 Inject friend functions into the enclosing namespace, so that they are
1852 visible outside the scope of the class in which they are declared.
1853 Friend functions were documented to work this way in the old Annotated
1854 C++ Reference Manual, and versions of G++ before 4.1 always worked
1855 that way. However, in ISO C++ a friend function which is not declared
1856 in an enclosing scope can only be found using argument dependent
1857 lookup. This option causes friends to be injected as they were in
1860 This option is for compatibility, and may be removed in a future
1863 @item -fno-elide-constructors
1864 @opindex fno-elide-constructors
1865 The C++ standard allows an implementation to omit creating a temporary
1866 which is only used to initialize another object of the same type.
1867 Specifying this option disables that optimization, and forces G++ to
1868 call the copy constructor in all cases.
1870 @item -fno-enforce-eh-specs
1871 @opindex fno-enforce-eh-specs
1872 Don't generate code to check for violation of exception specifications
1873 at runtime. This option violates the C++ standard, but may be useful
1874 for reducing code size in production builds, much like defining
1875 @samp{NDEBUG}. This does not give user code permission to throw
1876 exceptions in violation of the exception specifications; the compiler
1877 will still optimize based on the specifications, so throwing an
1878 unexpected exception will result in undefined behavior.
1881 @itemx -fno-for-scope
1883 @opindex fno-for-scope
1884 If @option{-ffor-scope} is specified, the scope of variables declared in
1885 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1886 as specified by the C++ standard.
1887 If @option{-fno-for-scope} is specified, the scope of variables declared in
1888 a @i{for-init-statement} extends to the end of the enclosing scope,
1889 as was the case in old versions of G++, and other (traditional)
1890 implementations of C++.
1892 The default if neither flag is given to follow the standard,
1893 but to allow and give a warning for old-style code that would
1894 otherwise be invalid, or have different behavior.
1896 @item -fno-gnu-keywords
1897 @opindex fno-gnu-keywords
1898 Do not recognize @code{typeof} as a keyword, so that code can use this
1899 word as an identifier. You can use the keyword @code{__typeof__} instead.
1900 @option{-ansi} implies @option{-fno-gnu-keywords}.
1902 @item -fno-implicit-templates
1903 @opindex fno-implicit-templates
1904 Never emit code for non-inline templates which are instantiated
1905 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1906 @xref{Template Instantiation}, for more information.
1908 @item -fno-implicit-inline-templates
1909 @opindex fno-implicit-inline-templates
1910 Don't emit code for implicit instantiations of inline templates, either.
1911 The default is to handle inlines differently so that compiles with and
1912 without optimization will need the same set of explicit instantiations.
1914 @item -fno-implement-inlines
1915 @opindex fno-implement-inlines
1916 To save space, do not emit out-of-line copies of inline functions
1917 controlled by @samp{#pragma implementation}. This will cause linker
1918 errors if these functions are not inlined everywhere they are called.
1920 @item -fms-extensions
1921 @opindex fms-extensions
1922 Disable pedantic warnings about constructs used in MFC, such as implicit
1923 int and getting a pointer to member function via non-standard syntax.
1925 @item -fno-nonansi-builtins
1926 @opindex fno-nonansi-builtins
1927 Disable built-in declarations of functions that are not mandated by
1928 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1929 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1931 @item -fno-operator-names
1932 @opindex fno-operator-names
1933 Do not treat the operator name keywords @code{and}, @code{bitand},
1934 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1935 synonyms as keywords.
1937 @item -fno-optional-diags
1938 @opindex fno-optional-diags
1939 Disable diagnostics that the standard says a compiler does not need to
1940 issue. Currently, the only such diagnostic issued by G++ is the one for
1941 a name having multiple meanings within a class.
1944 @opindex fpermissive
1945 Downgrade some diagnostics about nonconformant code from errors to
1946 warnings. Thus, using @option{-fpermissive} will allow some
1947 nonconforming code to compile.
1949 @item -fno-pretty-templates
1950 @opindex fno-pretty-templates
1951 When an error message refers to a specialization of a function
1952 template, the compiler will normally print the signature of the
1953 template followed by the template arguments and any typedefs or
1954 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1955 rather than @code{void f(int)}) so that it's clear which template is
1956 involved. When an error message refers to a specialization of a class
1957 template, the compiler will omit any template arguments which match
1958 the default template arguments for that template. If either of these
1959 behaviors make it harder to understand the error message rather than
1960 easier, using @option{-fno-pretty-templates} will disable them.
1964 Enable automatic template instantiation at link time. This option also
1965 implies @option{-fno-implicit-templates}. @xref{Template
1966 Instantiation}, for more information.
1970 Disable generation of information about every class with virtual
1971 functions for use by the C++ runtime type identification features
1972 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1973 of the language, you can save some space by using this flag. Note that
1974 exception handling uses the same information, but it will generate it as
1975 needed. The @samp{dynamic_cast} operator can still be used for casts that
1976 do not require runtime type information, i.e.@: casts to @code{void *} or to
1977 unambiguous base classes.
1981 Emit statistics about front-end processing at the end of the compilation.
1982 This information is generally only useful to the G++ development team.
1984 @item -ftemplate-depth=@var{n}
1985 @opindex ftemplate-depth
1986 Set the maximum instantiation depth for template classes to @var{n}.
1987 A limit on the template instantiation depth is needed to detect
1988 endless recursions during template class instantiation. ANSI/ISO C++
1989 conforming programs must not rely on a maximum depth greater than 17
1990 (changed to 1024 in C++0x).
1992 @item -fno-threadsafe-statics
1993 @opindex fno-threadsafe-statics
1994 Do not emit the extra code to use the routines specified in the C++
1995 ABI for thread-safe initialization of local statics. You can use this
1996 option to reduce code size slightly in code that doesn't need to be
1999 @item -fuse-cxa-atexit
2000 @opindex fuse-cxa-atexit
2001 Register destructors for objects with static storage duration with the
2002 @code{__cxa_atexit} function rather than the @code{atexit} function.
2003 This option is required for fully standards-compliant handling of static
2004 destructors, but will only work if your C library supports
2005 @code{__cxa_atexit}.
2007 @item -fno-use-cxa-get-exception-ptr
2008 @opindex fno-use-cxa-get-exception-ptr
2009 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2010 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2011 if the runtime routine is not available.
2013 @item -fvisibility-inlines-hidden
2014 @opindex fvisibility-inlines-hidden
2015 This switch declares that the user does not attempt to compare
2016 pointers to inline methods where the addresses of the two functions
2017 were taken in different shared objects.
2019 The effect of this is that GCC may, effectively, mark inline methods with
2020 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2021 appear in the export table of a DSO and do not require a PLT indirection
2022 when used within the DSO@. Enabling this option can have a dramatic effect
2023 on load and link times of a DSO as it massively reduces the size of the
2024 dynamic export table when the library makes heavy use of templates.
2026 The behavior of this switch is not quite the same as marking the
2027 methods as hidden directly, because it does not affect static variables
2028 local to the function or cause the compiler to deduce that
2029 the function is defined in only one shared object.
2031 You may mark a method as having a visibility explicitly to negate the
2032 effect of the switch for that method. For example, if you do want to
2033 compare pointers to a particular inline method, you might mark it as
2034 having default visibility. Marking the enclosing class with explicit
2035 visibility will have no effect.
2037 Explicitly instantiated inline methods are unaffected by this option
2038 as their linkage might otherwise cross a shared library boundary.
2039 @xref{Template Instantiation}.
2041 @item -fvisibility-ms-compat
2042 @opindex fvisibility-ms-compat
2043 This flag attempts to use visibility settings to make GCC's C++
2044 linkage model compatible with that of Microsoft Visual Studio.
2046 The flag makes these changes to GCC's linkage model:
2050 It sets the default visibility to @code{hidden}, like
2051 @option{-fvisibility=hidden}.
2054 Types, but not their members, are not hidden by default.
2057 The One Definition Rule is relaxed for types without explicit
2058 visibility specifications which are defined in more than one different
2059 shared object: those declarations are permitted if they would have
2060 been permitted when this option was not used.
2063 In new code it is better to use @option{-fvisibility=hidden} and
2064 export those classes which are intended to be externally visible.
2065 Unfortunately it is possible for code to rely, perhaps accidentally,
2066 on the Visual Studio behavior.
2068 Among the consequences of these changes are that static data members
2069 of the same type with the same name but defined in different shared
2070 objects will be different, so changing one will not change the other;
2071 and that pointers to function members defined in different shared
2072 objects may not compare equal. When this flag is given, it is a
2073 violation of the ODR to define types with the same name differently.
2077 Do not use weak symbol support, even if it is provided by the linker.
2078 By default, G++ will use weak symbols if they are available. This
2079 option exists only for testing, and should not be used by end-users;
2080 it will result in inferior code and has no benefits. This option may
2081 be removed in a future release of G++.
2085 Do not search for header files in the standard directories specific to
2086 C++, but do still search the other standard directories. (This option
2087 is used when building the C++ library.)
2090 In addition, these optimization, warning, and code generation options
2091 have meanings only for C++ programs:
2094 @item -fno-default-inline
2095 @opindex fno-default-inline
2096 Do not assume @samp{inline} for functions defined inside a class scope.
2097 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2098 functions will have linkage like inline functions; they just won't be
2101 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2104 Warn when G++ generates code that is probably not compatible with the
2105 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2106 all such cases, there are probably some cases that are not warned about,
2107 even though G++ is generating incompatible code. There may also be
2108 cases where warnings are emitted even though the code that is generated
2111 You should rewrite your code to avoid these warnings if you are
2112 concerned about the fact that code generated by G++ may not be binary
2113 compatible with code generated by other compilers.
2115 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2120 A template with a non-type template parameter of reference type is
2121 mangled incorrectly:
2124 template <int &> struct S @{@};
2128 This is fixed in @option{-fabi-version=3}.
2131 SIMD vector types declared using @code{__attribute ((vector_size))} are
2132 mangled in a non-standard way that does not allow for overloading of
2133 functions taking vectors of different sizes.
2135 The mangling is changed in @option{-fabi-version=4}.
2138 The known incompatibilities in @option{-fabi-version=1} include:
2143 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2144 pack data into the same byte as a base class. For example:
2147 struct A @{ virtual void f(); int f1 : 1; @};
2148 struct B : public A @{ int f2 : 1; @};
2152 In this case, G++ will place @code{B::f2} into the same byte
2153 as@code{A::f1}; other compilers will not. You can avoid this problem
2154 by explicitly padding @code{A} so that its size is a multiple of the
2155 byte size on your platform; that will cause G++ and other compilers to
2156 layout @code{B} identically.
2159 Incorrect handling of tail-padding for virtual bases. G++ does not use
2160 tail padding when laying out virtual bases. For example:
2163 struct A @{ virtual void f(); char c1; @};
2164 struct B @{ B(); char c2; @};
2165 struct C : public A, public virtual B @{@};
2169 In this case, G++ will not place @code{B} into the tail-padding for
2170 @code{A}; other compilers will. You can avoid this problem by
2171 explicitly padding @code{A} so that its size is a multiple of its
2172 alignment (ignoring virtual base classes); that will cause G++ and other
2173 compilers to layout @code{C} identically.
2176 Incorrect handling of bit-fields with declared widths greater than that
2177 of their underlying types, when the bit-fields appear in a union. For
2181 union U @{ int i : 4096; @};
2185 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2186 union too small by the number of bits in an @code{int}.
2189 Empty classes can be placed at incorrect offsets. For example:
2199 struct C : public B, public A @{@};
2203 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2204 it should be placed at offset zero. G++ mistakenly believes that the
2205 @code{A} data member of @code{B} is already at offset zero.
2208 Names of template functions whose types involve @code{typename} or
2209 template template parameters can be mangled incorrectly.
2212 template <typename Q>
2213 void f(typename Q::X) @{@}
2215 template <template <typename> class Q>
2216 void f(typename Q<int>::X) @{@}
2220 Instantiations of these templates may be mangled incorrectly.
2224 It also warns psABI related changes. The known psABI changes at this
2230 For SYSV/x86-64, when passing union with long double, it is changed to
2231 pass in memory as specified in psABI. For example:
2241 @code{union U} will always be passed in memory.
2245 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2246 @opindex Wctor-dtor-privacy
2247 @opindex Wno-ctor-dtor-privacy
2248 Warn when a class seems unusable because all the constructors or
2249 destructors in that class are private, and it has neither friends nor
2250 public static member functions.
2252 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2253 @opindex Wnon-virtual-dtor
2254 @opindex Wno-non-virtual-dtor
2255 Warn when a class has virtual functions and accessible non-virtual
2256 destructor, in which case it would be possible but unsafe to delete
2257 an instance of a derived class through a pointer to the base class.
2258 This warning is also enabled if -Weffc++ is specified.
2260 @item -Wreorder @r{(C++ and Objective-C++ only)}
2262 @opindex Wno-reorder
2263 @cindex reordering, warning
2264 @cindex warning for reordering of member initializers
2265 Warn when the order of member initializers given in the code does not
2266 match the order in which they must be executed. For instance:
2272 A(): j (0), i (1) @{ @}
2276 The compiler will rearrange the member initializers for @samp{i}
2277 and @samp{j} to match the declaration order of the members, emitting
2278 a warning to that effect. This warning is enabled by @option{-Wall}.
2281 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2284 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2287 Warn about violations of the following style guidelines from Scott Meyers'
2288 @cite{Effective C++} book:
2292 Item 11: Define a copy constructor and an assignment operator for classes
2293 with dynamically allocated memory.
2296 Item 12: Prefer initialization to assignment in constructors.
2299 Item 14: Make destructors virtual in base classes.
2302 Item 15: Have @code{operator=} return a reference to @code{*this}.
2305 Item 23: Don't try to return a reference when you must return an object.
2309 Also warn about violations of the following style guidelines from
2310 Scott Meyers' @cite{More Effective C++} book:
2314 Item 6: Distinguish between prefix and postfix forms of increment and
2315 decrement operators.
2318 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2322 When selecting this option, be aware that the standard library
2323 headers do not obey all of these guidelines; use @samp{grep -v}
2324 to filter out those warnings.
2326 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2327 @opindex Wstrict-null-sentinel
2328 @opindex Wno-strict-null-sentinel
2329 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2330 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2331 to @code{__null}. Although it is a null pointer constant not a null pointer,
2332 it is guaranteed to be of the same size as a pointer. But this use is
2333 not portable across different compilers.
2335 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2336 @opindex Wno-non-template-friend
2337 @opindex Wnon-template-friend
2338 Disable warnings when non-templatized friend functions are declared
2339 within a template. Since the advent of explicit template specification
2340 support in G++, if the name of the friend is an unqualified-id (i.e.,
2341 @samp{friend foo(int)}), the C++ language specification demands that the
2342 friend declare or define an ordinary, nontemplate function. (Section
2343 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2344 could be interpreted as a particular specialization of a templatized
2345 function. Because this non-conforming behavior is no longer the default
2346 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2347 check existing code for potential trouble spots and is on by default.
2348 This new compiler behavior can be turned off with
2349 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2350 but disables the helpful warning.
2352 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2353 @opindex Wold-style-cast
2354 @opindex Wno-old-style-cast
2355 Warn if an old-style (C-style) cast to a non-void type is used within
2356 a C++ program. The new-style casts (@samp{dynamic_cast},
2357 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2358 less vulnerable to unintended effects and much easier to search for.
2360 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2361 @opindex Woverloaded-virtual
2362 @opindex Wno-overloaded-virtual
2363 @cindex overloaded virtual fn, warning
2364 @cindex warning for overloaded virtual fn
2365 Warn when a function declaration hides virtual functions from a
2366 base class. For example, in:
2373 struct B: public A @{
2378 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2386 will fail to compile.
2388 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2389 @opindex Wno-pmf-conversions
2390 @opindex Wpmf-conversions
2391 Disable the diagnostic for converting a bound pointer to member function
2394 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2395 @opindex Wsign-promo
2396 @opindex Wno-sign-promo
2397 Warn when overload resolution chooses a promotion from unsigned or
2398 enumerated type to a signed type, over a conversion to an unsigned type of
2399 the same size. Previous versions of G++ would try to preserve
2400 unsignedness, but the standard mandates the current behavior.
2405 A& operator = (int);
2415 In this example, G++ will synthesize a default @samp{A& operator =
2416 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2419 @node Objective-C and Objective-C++ Dialect Options
2420 @section Options Controlling Objective-C and Objective-C++ Dialects
2422 @cindex compiler options, Objective-C and Objective-C++
2423 @cindex Objective-C and Objective-C++ options, command line
2424 @cindex options, Objective-C and Objective-C++
2425 (NOTE: This manual does not describe the Objective-C and Objective-C++
2426 languages themselves. See @xref{Standards,,Language Standards
2427 Supported by GCC}, for references.)
2429 This section describes the command-line options that are only meaningful
2430 for Objective-C and Objective-C++ programs, but you can also use most of
2431 the language-independent GNU compiler options.
2432 For example, you might compile a file @code{some_class.m} like this:
2435 gcc -g -fgnu-runtime -O -c some_class.m
2439 In this example, @option{-fgnu-runtime} is an option meant only for
2440 Objective-C and Objective-C++ programs; you can use the other options with
2441 any language supported by GCC@.
2443 Note that since Objective-C is an extension of the C language, Objective-C
2444 compilations may also use options specific to the C front-end (e.g.,
2445 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2446 C++-specific options (e.g., @option{-Wabi}).
2448 Here is a list of options that are @emph{only} for compiling Objective-C
2449 and Objective-C++ programs:
2452 @item -fconstant-string-class=@var{class-name}
2453 @opindex fconstant-string-class
2454 Use @var{class-name} as the name of the class to instantiate for each
2455 literal string specified with the syntax @code{@@"@dots{}"}. The default
2456 class name is @code{NXConstantString} if the GNU runtime is being used, and
2457 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2458 @option{-fconstant-cfstrings} option, if also present, will override the
2459 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2460 to be laid out as constant CoreFoundation strings.
2463 @opindex fgnu-runtime
2464 Generate object code compatible with the standard GNU Objective-C
2465 runtime. This is the default for most types of systems.
2467 @item -fnext-runtime
2468 @opindex fnext-runtime
2469 Generate output compatible with the NeXT runtime. This is the default
2470 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2471 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2474 @item -fno-nil-receivers
2475 @opindex fno-nil-receivers
2476 Assume that all Objective-C message dispatches (e.g.,
2477 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2478 is not @code{nil}. This allows for more efficient entry points in the runtime
2479 to be used. Currently, this option is only available in conjunction with
2480 the NeXT runtime on Mac OS X 10.3 and later.
2482 @item -fobjc-call-cxx-cdtors
2483 @opindex fobjc-call-cxx-cdtors
2484 For each Objective-C class, check if any of its instance variables is a
2485 C++ object with a non-trivial default constructor. If so, synthesize a
2486 special @code{- (id) .cxx_construct} instance method that will run
2487 non-trivial default constructors on any such instance variables, in order,
2488 and then return @code{self}. Similarly, check if any instance variable
2489 is a C++ object with a non-trivial destructor, and if so, synthesize a
2490 special @code{- (void) .cxx_destruct} method that will run
2491 all such default destructors, in reverse order.
2493 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2494 thusly generated will only operate on instance variables declared in the
2495 current Objective-C class, and not those inherited from superclasses. It
2496 is the responsibility of the Objective-C runtime to invoke all such methods
2497 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2498 will be invoked by the runtime immediately after a new object
2499 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2500 be invoked immediately before the runtime deallocates an object instance.
2502 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2503 support for invoking the @code{- (id) .cxx_construct} and
2504 @code{- (void) .cxx_destruct} methods.
2506 @item -fobjc-direct-dispatch
2507 @opindex fobjc-direct-dispatch
2508 Allow fast jumps to the message dispatcher. On Darwin this is
2509 accomplished via the comm page.
2511 @item -fobjc-exceptions
2512 @opindex fobjc-exceptions
2513 Enable syntactic support for structured exception handling in Objective-C,
2514 similar to what is offered by C++ and Java. This option is
2515 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2524 @@catch (AnObjCClass *exc) @{
2531 @@catch (AnotherClass *exc) @{
2534 @@catch (id allOthers) @{
2544 The @code{@@throw} statement may appear anywhere in an Objective-C or
2545 Objective-C++ program; when used inside of a @code{@@catch} block, the
2546 @code{@@throw} may appear without an argument (as shown above), in which case
2547 the object caught by the @code{@@catch} will be rethrown.
2549 Note that only (pointers to) Objective-C objects may be thrown and
2550 caught using this scheme. When an object is thrown, it will be caught
2551 by the nearest @code{@@catch} clause capable of handling objects of that type,
2552 analogously to how @code{catch} blocks work in C++ and Java. A
2553 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2554 any and all Objective-C exceptions not caught by previous @code{@@catch}
2557 The @code{@@finally} clause, if present, will be executed upon exit from the
2558 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2559 regardless of whether any exceptions are thrown, caught or rethrown
2560 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2561 of the @code{finally} clause in Java.
2563 There are several caveats to using the new exception mechanism:
2567 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2568 idioms provided by the @code{NSException} class, the new
2569 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2570 systems, due to additional functionality needed in the (NeXT) Objective-C
2574 As mentioned above, the new exceptions do not support handling
2575 types other than Objective-C objects. Furthermore, when used from
2576 Objective-C++, the Objective-C exception model does not interoperate with C++
2577 exceptions at this time. This means you cannot @code{@@throw} an exception
2578 from Objective-C and @code{catch} it in C++, or vice versa
2579 (i.e., @code{throw @dots{} @@catch}).
2582 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2583 blocks for thread-safe execution:
2586 @@synchronized (ObjCClass *guard) @{
2591 Upon entering the @code{@@synchronized} block, a thread of execution shall
2592 first check whether a lock has been placed on the corresponding @code{guard}
2593 object by another thread. If it has, the current thread shall wait until
2594 the other thread relinquishes its lock. Once @code{guard} becomes available,
2595 the current thread will place its own lock on it, execute the code contained in
2596 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2597 making @code{guard} available to other threads).
2599 Unlike Java, Objective-C does not allow for entire methods to be marked
2600 @code{@@synchronized}. Note that throwing exceptions out of
2601 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2602 to be unlocked properly.
2606 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2608 @item -freplace-objc-classes
2609 @opindex freplace-objc-classes
2610 Emit a special marker instructing @command{ld(1)} not to statically link in
2611 the resulting object file, and allow @command{dyld(1)} to load it in at
2612 run time instead. This is used in conjunction with the Fix-and-Continue
2613 debugging mode, where the object file in question may be recompiled and
2614 dynamically reloaded in the course of program execution, without the need
2615 to restart the program itself. Currently, Fix-and-Continue functionality
2616 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2621 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2622 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2623 compile time) with static class references that get initialized at load time,
2624 which improves run-time performance. Specifying the @option{-fzero-link} flag
2625 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2626 to be retained. This is useful in Zero-Link debugging mode, since it allows
2627 for individual class implementations to be modified during program execution.
2631 Dump interface declarations for all classes seen in the source file to a
2632 file named @file{@var{sourcename}.decl}.
2634 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2635 @opindex Wassign-intercept
2636 @opindex Wno-assign-intercept
2637 Warn whenever an Objective-C assignment is being intercepted by the
2640 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2641 @opindex Wno-protocol
2643 If a class is declared to implement a protocol, a warning is issued for
2644 every method in the protocol that is not implemented by the class. The
2645 default behavior is to issue a warning for every method not explicitly
2646 implemented in the class, even if a method implementation is inherited
2647 from the superclass. If you use the @option{-Wno-protocol} option, then
2648 methods inherited from the superclass are considered to be implemented,
2649 and no warning is issued for them.
2651 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2653 @opindex Wno-selector
2654 Warn if multiple methods of different types for the same selector are
2655 found during compilation. The check is performed on the list of methods
2656 in the final stage of compilation. Additionally, a check is performed
2657 for each selector appearing in a @code{@@selector(@dots{})}
2658 expression, and a corresponding method for that selector has been found
2659 during compilation. Because these checks scan the method table only at
2660 the end of compilation, these warnings are not produced if the final
2661 stage of compilation is not reached, for example because an error is
2662 found during compilation, or because the @option{-fsyntax-only} option is
2665 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2666 @opindex Wstrict-selector-match
2667 @opindex Wno-strict-selector-match
2668 Warn if multiple methods with differing argument and/or return types are
2669 found for a given selector when attempting to send a message using this
2670 selector to a receiver of type @code{id} or @code{Class}. When this flag
2671 is off (which is the default behavior), the compiler will omit such warnings
2672 if any differences found are confined to types which share the same size
2675 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2676 @opindex Wundeclared-selector
2677 @opindex Wno-undeclared-selector
2678 Warn if a @code{@@selector(@dots{})} expression referring to an
2679 undeclared selector is found. A selector is considered undeclared if no
2680 method with that name has been declared before the
2681 @code{@@selector(@dots{})} expression, either explicitly in an
2682 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2683 an @code{@@implementation} section. This option always performs its
2684 checks as soon as a @code{@@selector(@dots{})} expression is found,
2685 while @option{-Wselector} only performs its checks in the final stage of
2686 compilation. This also enforces the coding style convention
2687 that methods and selectors must be declared before being used.
2689 @item -print-objc-runtime-info
2690 @opindex print-objc-runtime-info
2691 Generate C header describing the largest structure that is passed by
2696 @node Language Independent Options
2697 @section Options to Control Diagnostic Messages Formatting
2698 @cindex options to control diagnostics formatting
2699 @cindex diagnostic messages
2700 @cindex message formatting
2702 Traditionally, diagnostic messages have been formatted irrespective of
2703 the output device's aspect (e.g.@: its width, @dots{}). The options described
2704 below can be used to control the diagnostic messages formatting
2705 algorithm, e.g.@: how many characters per line, how often source location
2706 information should be reported. Right now, only the C++ front end can
2707 honor these options. However it is expected, in the near future, that
2708 the remaining front ends would be able to digest them correctly.
2711 @item -fmessage-length=@var{n}
2712 @opindex fmessage-length
2713 Try to format error messages so that they fit on lines of about @var{n}
2714 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2715 the front ends supported by GCC@. If @var{n} is zero, then no
2716 line-wrapping will be done; each error message will appear on a single
2719 @opindex fdiagnostics-show-location
2720 @item -fdiagnostics-show-location=once
2721 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2722 reporter to emit @emph{once} source location information; that is, in
2723 case the message is too long to fit on a single physical line and has to
2724 be wrapped, the source location won't be emitted (as prefix) again,
2725 over and over, in subsequent continuation lines. This is the default
2728 @item -fdiagnostics-show-location=every-line
2729 Only meaningful in line-wrapping mode. Instructs the diagnostic
2730 messages reporter to emit the same source location information (as
2731 prefix) for physical lines that result from the process of breaking
2732 a message which is too long to fit on a single line.
2734 @item -fdiagnostics-show-option
2735 @opindex fdiagnostics-show-option
2736 This option instructs the diagnostic machinery to add text to each
2737 diagnostic emitted, which indicates which command line option directly
2738 controls that diagnostic, when such an option is known to the
2739 diagnostic machinery.
2741 @item -Wcoverage-mismatch
2742 @opindex Wcoverage-mismatch
2743 Warn if feedback profiles do not match when using the
2744 @option{-fprofile-use} option.
2745 If a source file was changed between @option{-fprofile-gen} and
2746 @option{-fprofile-use}, the files with the profile feedback can fail
2747 to match the source file and GCC can not use the profile feedback
2748 information. By default, GCC emits an error message in this case.
2749 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2750 error. GCC does not use appropriate feedback profiles, so using this
2751 option can result in poorly optimized code. This option is useful
2752 only in the case of very minor changes such as bug fixes to an
2757 @node Warning Options
2758 @section Options to Request or Suppress Warnings
2759 @cindex options to control warnings
2760 @cindex warning messages
2761 @cindex messages, warning
2762 @cindex suppressing warnings
2764 Warnings are diagnostic messages that report constructions which
2765 are not inherently erroneous but which are risky or suggest there
2766 may have been an error.
2768 The following language-independent options do not enable specific
2769 warnings but control the kinds of diagnostics produced by GCC.
2772 @cindex syntax checking
2774 @opindex fsyntax-only
2775 Check the code for syntax errors, but don't do anything beyond that.
2779 Inhibit all warning messages.
2784 Make all warnings into errors.
2789 Make the specified warning into an error. The specifier for a warning
2790 is appended, for example @option{-Werror=switch} turns the warnings
2791 controlled by @option{-Wswitch} into errors. This switch takes a
2792 negative form, to be used to negate @option{-Werror} for specific
2793 warnings, for example @option{-Wno-error=switch} makes
2794 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2795 is in effect. You can use the @option{-fdiagnostics-show-option}
2796 option to have each controllable warning amended with the option which
2797 controls it, to determine what to use with this option.
2799 Note that specifying @option{-Werror=}@var{foo} automatically implies
2800 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2803 @item -Wfatal-errors
2804 @opindex Wfatal-errors
2805 @opindex Wno-fatal-errors
2806 This option causes the compiler to abort compilation on the first error
2807 occurred rather than trying to keep going and printing further error
2812 You can request many specific warnings with options beginning
2813 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2814 implicit declarations. Each of these specific warning options also
2815 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2816 example, @option{-Wno-implicit}. This manual lists only one of the
2817 two forms, whichever is not the default. For further,
2818 language-specific options also refer to @ref{C++ Dialect Options} and
2819 @ref{Objective-C and Objective-C++ Dialect Options}.
2821 When an unrecognized warning label is requested (e.g.,
2822 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2823 that the option is not recognized. However, if the @samp{-Wno-} form
2824 is used, the behavior is slightly different: No diagnostic will be
2825 produced for @option{-Wno-unknown-warning} unless other diagnostics
2826 are being produced. This allows the use of new @option{-Wno-} options
2827 with old compilers, but if something goes wrong, the compiler will
2828 warn that an unrecognized option was used.
2833 Issue all the warnings demanded by strict ISO C and ISO C++;
2834 reject all programs that use forbidden extensions, and some other
2835 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2836 version of the ISO C standard specified by any @option{-std} option used.
2838 Valid ISO C and ISO C++ programs should compile properly with or without
2839 this option (though a rare few will require @option{-ansi} or a
2840 @option{-std} option specifying the required version of ISO C)@. However,
2841 without this option, certain GNU extensions and traditional C and C++
2842 features are supported as well. With this option, they are rejected.
2844 @option{-pedantic} does not cause warning messages for use of the
2845 alternate keywords whose names begin and end with @samp{__}. Pedantic
2846 warnings are also disabled in the expression that follows
2847 @code{__extension__}. However, only system header files should use
2848 these escape routes; application programs should avoid them.
2849 @xref{Alternate Keywords}.
2851 Some users try to use @option{-pedantic} to check programs for strict ISO
2852 C conformance. They soon find that it does not do quite what they want:
2853 it finds some non-ISO practices, but not all---only those for which
2854 ISO C @emph{requires} a diagnostic, and some others for which
2855 diagnostics have been added.
2857 A feature to report any failure to conform to ISO C might be useful in
2858 some instances, but would require considerable additional work and would
2859 be quite different from @option{-pedantic}. We don't have plans to
2860 support such a feature in the near future.
2862 Where the standard specified with @option{-std} represents a GNU
2863 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2864 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2865 extended dialect is based. Warnings from @option{-pedantic} are given
2866 where they are required by the base standard. (It would not make sense
2867 for such warnings to be given only for features not in the specified GNU
2868 C dialect, since by definition the GNU dialects of C include all
2869 features the compiler supports with the given option, and there would be
2870 nothing to warn about.)
2872 @item -pedantic-errors
2873 @opindex pedantic-errors
2874 Like @option{-pedantic}, except that errors are produced rather than
2880 This enables all the warnings about constructions that some users
2881 consider questionable, and that are easy to avoid (or modify to
2882 prevent the warning), even in conjunction with macros. This also
2883 enables some language-specific warnings described in @ref{C++ Dialect
2884 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2886 @option{-Wall} turns on the following warning flags:
2888 @gccoptlist{-Waddress @gol
2889 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2891 -Wchar-subscripts @gol
2892 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2894 -Wimplicit-function-declaration @gol
2897 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2898 -Wmissing-braces @gol
2904 -Wsequence-point @gol
2905 -Wsign-compare @r{(only in C++)} @gol
2906 -Wstrict-aliasing @gol
2907 -Wstrict-overflow=1 @gol
2910 -Wuninitialized @gol
2911 -Wunknown-pragmas @gol
2912 -Wunused-function @gol
2915 -Wunused-variable @gol
2916 -Wvolatile-register-var @gol
2919 Note that some warning flags are not implied by @option{-Wall}. Some of
2920 them warn about constructions that users generally do not consider
2921 questionable, but which occasionally you might wish to check for;
2922 others warn about constructions that are necessary or hard to avoid in
2923 some cases, and there is no simple way to modify the code to suppress
2924 the warning. Some of them are enabled by @option{-Wextra} but many of
2925 them must be enabled individually.
2931 This enables some extra warning flags that are not enabled by
2932 @option{-Wall}. (This option used to be called @option{-W}. The older
2933 name is still supported, but the newer name is more descriptive.)
2935 @gccoptlist{-Wclobbered @gol
2937 -Wignored-qualifiers @gol
2938 -Wmissing-field-initializers @gol
2939 -Wmissing-parameter-type @r{(C only)} @gol
2940 -Wold-style-declaration @r{(C only)} @gol
2941 -Woverride-init @gol
2944 -Wuninitialized @gol
2945 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2946 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2949 The option @option{-Wextra} also prints warning messages for the
2955 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2956 @samp{>}, or @samp{>=}.
2959 (C++ only) An enumerator and a non-enumerator both appear in a
2960 conditional expression.
2963 (C++ only) Ambiguous virtual bases.
2966 (C++ only) Subscripting an array which has been declared @samp{register}.
2969 (C++ only) Taking the address of a variable which has been declared
2973 (C++ only) A base class is not initialized in a derived class' copy
2978 @item -Wchar-subscripts
2979 @opindex Wchar-subscripts
2980 @opindex Wno-char-subscripts
2981 Warn if an array subscript has type @code{char}. This is a common cause
2982 of error, as programmers often forget that this type is signed on some
2984 This warning is enabled by @option{-Wall}.
2988 @opindex Wno-comment
2989 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2990 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2991 This warning is enabled by @option{-Wall}.
2994 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
2996 Suppress warning messages emitted by @code{#warning} directives.
3001 @opindex ffreestanding
3002 @opindex fno-builtin
3003 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3004 the arguments supplied have types appropriate to the format string
3005 specified, and that the conversions specified in the format string make
3006 sense. This includes standard functions, and others specified by format
3007 attributes (@pxref{Function Attributes}), in the @code{printf},
3008 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3009 not in the C standard) families (or other target-specific families).
3010 Which functions are checked without format attributes having been
3011 specified depends on the standard version selected, and such checks of
3012 functions without the attribute specified are disabled by
3013 @option{-ffreestanding} or @option{-fno-builtin}.
3015 The formats are checked against the format features supported by GNU
3016 libc version 2.2. These include all ISO C90 and C99 features, as well
3017 as features from the Single Unix Specification and some BSD and GNU
3018 extensions. Other library implementations may not support all these
3019 features; GCC does not support warning about features that go beyond a
3020 particular library's limitations. However, if @option{-pedantic} is used
3021 with @option{-Wformat}, warnings will be given about format features not
3022 in the selected standard version (but not for @code{strfmon} formats,
3023 since those are not in any version of the C standard). @xref{C Dialect
3024 Options,,Options Controlling C Dialect}.
3026 Since @option{-Wformat} also checks for null format arguments for
3027 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3029 @option{-Wformat} is included in @option{-Wall}. For more control over some
3030 aspects of format checking, the options @option{-Wformat-y2k},
3031 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3032 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3033 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3036 @opindex Wformat-y2k
3037 @opindex Wno-format-y2k
3038 If @option{-Wformat} is specified, also warn about @code{strftime}
3039 formats which may yield only a two-digit year.
3041 @item -Wno-format-contains-nul
3042 @opindex Wno-format-contains-nul
3043 @opindex Wformat-contains-nul
3044 If @option{-Wformat} is specified, do not warn about format strings that
3047 @item -Wno-format-extra-args
3048 @opindex Wno-format-extra-args
3049 @opindex Wformat-extra-args
3050 If @option{-Wformat} is specified, do not warn about excess arguments to a
3051 @code{printf} or @code{scanf} format function. The C standard specifies
3052 that such arguments are ignored.
3054 Where the unused arguments lie between used arguments that are
3055 specified with @samp{$} operand number specifications, normally
3056 warnings are still given, since the implementation could not know what
3057 type to pass to @code{va_arg} to skip the unused arguments. However,
3058 in the case of @code{scanf} formats, this option will suppress the
3059 warning if the unused arguments are all pointers, since the Single
3060 Unix Specification says that such unused arguments are allowed.
3062 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3063 @opindex Wno-format-zero-length
3064 @opindex Wformat-zero-length
3065 If @option{-Wformat} is specified, do not warn about zero-length formats.
3066 The C standard specifies that zero-length formats are allowed.
3068 @item -Wformat-nonliteral
3069 @opindex Wformat-nonliteral
3070 @opindex Wno-format-nonliteral
3071 If @option{-Wformat} is specified, also warn if the format string is not a
3072 string literal and so cannot be checked, unless the format function
3073 takes its format arguments as a @code{va_list}.
3075 @item -Wformat-security
3076 @opindex Wformat-security
3077 @opindex Wno-format-security
3078 If @option{-Wformat} is specified, also warn about uses of format
3079 functions that represent possible security problems. At present, this
3080 warns about calls to @code{printf} and @code{scanf} functions where the
3081 format string is not a string literal and there are no format arguments,
3082 as in @code{printf (foo);}. This may be a security hole if the format
3083 string came from untrusted input and contains @samp{%n}. (This is
3084 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3085 in future warnings may be added to @option{-Wformat-security} that are not
3086 included in @option{-Wformat-nonliteral}.)
3090 @opindex Wno-format=2
3091 Enable @option{-Wformat} plus format checks not included in
3092 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3093 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3095 @item -Wnonnull @r{(C and Objective-C only)}
3097 @opindex Wno-nonnull
3098 Warn about passing a null pointer for arguments marked as
3099 requiring a non-null value by the @code{nonnull} function attribute.
3101 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3102 can be disabled with the @option{-Wno-nonnull} option.
3104 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3106 @opindex Wno-init-self
3107 Warn about uninitialized variables which are initialized with themselves.
3108 Note this option can only be used with the @option{-Wuninitialized} option.
3110 For example, GCC will warn about @code{i} being uninitialized in the
3111 following snippet only when @option{-Winit-self} has been specified:
3122 @item -Wimplicit-int @r{(C and Objective-C only)}
3123 @opindex Wimplicit-int
3124 @opindex Wno-implicit-int
3125 Warn when a declaration does not specify a type.
3126 This warning is enabled by @option{-Wall}.
3128 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3129 @opindex Wimplicit-function-declaration
3130 @opindex Wno-implicit-function-declaration
3131 Give a warning whenever a function is used before being declared. In
3132 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3133 enabled by default and it is made into an error by
3134 @option{-pedantic-errors}. This warning is also enabled by
3139 @opindex Wno-implicit
3140 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3141 This warning is enabled by @option{-Wall}.
3143 @item -Wignored-qualifiers @r{(C and C++ only)}
3144 @opindex Wignored-qualifiers
3145 @opindex Wno-ignored-qualifiers
3146 Warn if the return type of a function has a type qualifier
3147 such as @code{const}. For ISO C such a type qualifier has no effect,
3148 since the value returned by a function is not an lvalue.
3149 For C++, the warning is only emitted for scalar types or @code{void}.
3150 ISO C prohibits qualified @code{void} return types on function
3151 definitions, so such return types always receive a warning
3152 even without this option.
3154 This warning is also enabled by @option{-Wextra}.
3159 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3160 a function with external linkage, returning int, taking either zero
3161 arguments, two, or three arguments of appropriate types. This warning
3162 is enabled by default in C++ and is enabled by either @option{-Wall}
3163 or @option{-pedantic}.
3165 @item -Wmissing-braces
3166 @opindex Wmissing-braces
3167 @opindex Wno-missing-braces
3168 Warn if an aggregate or union initializer is not fully bracketed. In
3169 the following example, the initializer for @samp{a} is not fully
3170 bracketed, but that for @samp{b} is fully bracketed.
3173 int a[2][2] = @{ 0, 1, 2, 3 @};
3174 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3177 This warning is enabled by @option{-Wall}.
3179 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3180 @opindex Wmissing-include-dirs
3181 @opindex Wno-missing-include-dirs
3182 Warn if a user-supplied include directory does not exist.
3185 @opindex Wparentheses
3186 @opindex Wno-parentheses
3187 Warn if parentheses are omitted in certain contexts, such
3188 as when there is an assignment in a context where a truth value
3189 is expected, or when operators are nested whose precedence people
3190 often get confused about.
3192 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3193 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3194 interpretation from that of ordinary mathematical notation.
3196 Also warn about constructions where there may be confusion to which
3197 @code{if} statement an @code{else} branch belongs. Here is an example of
3212 In C/C++, every @code{else} branch belongs to the innermost possible
3213 @code{if} statement, which in this example is @code{if (b)}. This is
3214 often not what the programmer expected, as illustrated in the above
3215 example by indentation the programmer chose. When there is the
3216 potential for this confusion, GCC will issue a warning when this flag
3217 is specified. To eliminate the warning, add explicit braces around
3218 the innermost @code{if} statement so there is no way the @code{else}
3219 could belong to the enclosing @code{if}. The resulting code would
3236 This warning is enabled by @option{-Wall}.
3238 @item -Wsequence-point
3239 @opindex Wsequence-point
3240 @opindex Wno-sequence-point
3241 Warn about code that may have undefined semantics because of violations
3242 of sequence point rules in the C and C++ standards.
3244 The C and C++ standards defines the order in which expressions in a C/C++
3245 program are evaluated in terms of @dfn{sequence points}, which represent
3246 a partial ordering between the execution of parts of the program: those
3247 executed before the sequence point, and those executed after it. These
3248 occur after the evaluation of a full expression (one which is not part
3249 of a larger expression), after the evaluation of the first operand of a
3250 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3251 function is called (but after the evaluation of its arguments and the
3252 expression denoting the called function), and in certain other places.
3253 Other than as expressed by the sequence point rules, the order of
3254 evaluation of subexpressions of an expression is not specified. All
3255 these rules describe only a partial order rather than a total order,
3256 since, for example, if two functions are called within one expression
3257 with no sequence point between them, the order in which the functions
3258 are called is not specified. However, the standards committee have
3259 ruled that function calls do not overlap.
3261 It is not specified when between sequence points modifications to the
3262 values of objects take effect. Programs whose behavior depends on this
3263 have undefined behavior; the C and C++ standards specify that ``Between
3264 the previous and next sequence point an object shall have its stored
3265 value modified at most once by the evaluation of an expression.
3266 Furthermore, the prior value shall be read only to determine the value
3267 to be stored.''. If a program breaks these rules, the results on any
3268 particular implementation are entirely unpredictable.
3270 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3271 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3272 diagnosed by this option, and it may give an occasional false positive
3273 result, but in general it has been found fairly effective at detecting
3274 this sort of problem in programs.
3276 The standard is worded confusingly, therefore there is some debate
3277 over the precise meaning of the sequence point rules in subtle cases.
3278 Links to discussions of the problem, including proposed formal
3279 definitions, may be found on the GCC readings page, at
3280 @w{@uref{http://gcc.gnu.org/readings.html}}.
3282 This warning is enabled by @option{-Wall} for C and C++.
3285 @opindex Wreturn-type
3286 @opindex Wno-return-type
3287 Warn whenever a function is defined with a return-type that defaults
3288 to @code{int}. Also warn about any @code{return} statement with no
3289 return-value in a function whose return-type is not @code{void}
3290 (falling off the end of the function body is considered returning
3291 without a value), and about a @code{return} statement with an
3292 expression in a function whose return-type is @code{void}.
3294 For C++, a function without return type always produces a diagnostic
3295 message, even when @option{-Wno-return-type} is specified. The only
3296 exceptions are @samp{main} and functions defined in system headers.
3298 This warning is enabled by @option{-Wall}.
3303 Warn whenever a @code{switch} statement has an index of enumerated type
3304 and lacks a @code{case} for one or more of the named codes of that
3305 enumeration. (The presence of a @code{default} label prevents this
3306 warning.) @code{case} labels outside the enumeration range also
3307 provoke warnings when this option is used (even if there is a
3308 @code{default} label).
3309 This warning is enabled by @option{-Wall}.
3311 @item -Wswitch-default
3312 @opindex Wswitch-default
3313 @opindex Wno-switch-default
3314 Warn whenever a @code{switch} statement does not have a @code{default}
3318 @opindex Wswitch-enum
3319 @opindex Wno-switch-enum
3320 Warn whenever a @code{switch} statement has an index of enumerated type
3321 and lacks a @code{case} for one or more of the named codes of that
3322 enumeration. @code{case} labels outside the enumeration range also
3323 provoke warnings when this option is used. The only difference
3324 between @option{-Wswitch} and this option is that this option gives a
3325 warning about an omitted enumeration code even if there is a
3326 @code{default} label.
3328 @item -Wsync-nand @r{(C and C++ only)}
3330 @opindex Wno-sync-nand
3331 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3332 built-in functions are used. These functions changed semantics in GCC 4.4.
3336 @opindex Wno-trigraphs
3337 Warn if any trigraphs are encountered that might change the meaning of
3338 the program (trigraphs within comments are not warned about).
3339 This warning is enabled by @option{-Wall}.
3341 @item -Wunused-but-set-parameter
3342 @opindex Wunused-but-set-parameter
3343 @opindex Wno-unused-but-set-parameter
3344 Warn whenever a function parameter is assigned to, but otherwise unused
3345 (aside from its declaration).
3347 To suppress this warning use the @samp{unused} attribute
3348 (@pxref{Variable Attributes}).
3350 This warning is also enabled by @option{-Wunused} together with
3353 @item -Wunused-but-set-variable
3354 @opindex Wunused-but-set-variable
3355 @opindex Wno-unused-but-set-variable
3356 Warn whenever a local variable is assigned to, but otherwise unused
3357 (aside from its declaration).
3358 This warning is enabled by @option{-Wall}.
3360 To suppress this warning use the @samp{unused} attribute
3361 (@pxref{Variable Attributes}).
3363 This warning is also enabled by @option{-Wunused}, which is enabled
3366 @item -Wunused-function
3367 @opindex Wunused-function
3368 @opindex Wno-unused-function
3369 Warn whenever a static function is declared but not defined or a
3370 non-inline static function is unused.
3371 This warning is enabled by @option{-Wall}.
3373 @item -Wunused-label
3374 @opindex Wunused-label
3375 @opindex Wno-unused-label
3376 Warn whenever a label is declared but not used.
3377 This warning is enabled by @option{-Wall}.
3379 To suppress this warning use the @samp{unused} attribute
3380 (@pxref{Variable Attributes}).
3382 @item -Wunused-parameter
3383 @opindex Wunused-parameter
3384 @opindex Wno-unused-parameter
3385 Warn whenever a function parameter is unused aside from its declaration.
3387 To suppress this warning use the @samp{unused} attribute
3388 (@pxref{Variable Attributes}).
3390 @item -Wno-unused-result
3391 @opindex Wunused-result
3392 @opindex Wno-unused-result
3393 Do not warn if a caller of a function marked with attribute
3394 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3395 its return value. The default is @option{-Wunused-result}.
3397 @item -Wunused-variable
3398 @opindex Wunused-variable
3399 @opindex Wno-unused-variable
3400 Warn whenever a local variable or non-constant static variable is unused
3401 aside from its declaration.
3402 This warning is enabled by @option{-Wall}.
3404 To suppress this warning use the @samp{unused} attribute
3405 (@pxref{Variable Attributes}).
3407 @item -Wunused-value
3408 @opindex Wunused-value
3409 @opindex Wno-unused-value
3410 Warn whenever a statement computes a result that is explicitly not
3411 used. To suppress this warning cast the unused expression to
3412 @samp{void}. This includes an expression-statement or the left-hand
3413 side of a comma expression that contains no side effects. For example,
3414 an expression such as @samp{x[i,j]} will cause a warning, while
3415 @samp{x[(void)i,j]} will not.
3417 This warning is enabled by @option{-Wall}.
3422 All the above @option{-Wunused} options combined.
3424 In order to get a warning about an unused function parameter, you must
3425 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3426 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3428 @item -Wuninitialized
3429 @opindex Wuninitialized
3430 @opindex Wno-uninitialized
3431 Warn if an automatic variable is used without first being initialized
3432 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3433 warn if a non-static reference or non-static @samp{const} member
3434 appears in a class without constructors.
3436 If you want to warn about code which uses the uninitialized value of the
3437 variable in its own initializer, use the @option{-Winit-self} option.
3439 These warnings occur for individual uninitialized or clobbered
3440 elements of structure, union or array variables as well as for
3441 variables which are uninitialized or clobbered as a whole. They do
3442 not occur for variables or elements declared @code{volatile}. Because
3443 these warnings depend on optimization, the exact variables or elements
3444 for which there are warnings will depend on the precise optimization
3445 options and version of GCC used.
3447 Note that there may be no warning about a variable that is used only
3448 to compute a value that itself is never used, because such
3449 computations may be deleted by data flow analysis before the warnings
3452 These warnings are made optional because GCC is not smart
3453 enough to see all the reasons why the code might be correct
3454 despite appearing to have an error. Here is one example of how
3475 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3476 always initialized, but GCC doesn't know this. Here is
3477 another common case:
3482 if (change_y) save_y = y, y = new_y;
3484 if (change_y) y = save_y;
3489 This has no bug because @code{save_y} is used only if it is set.
3491 @cindex @code{longjmp} warnings
3492 This option also warns when a non-volatile automatic variable might be
3493 changed by a call to @code{longjmp}. These warnings as well are possible
3494 only in optimizing compilation.
3496 The compiler sees only the calls to @code{setjmp}. It cannot know
3497 where @code{longjmp} will be called; in fact, a signal handler could
3498 call it at any point in the code. As a result, you may get a warning
3499 even when there is in fact no problem because @code{longjmp} cannot
3500 in fact be called at the place which would cause a problem.
3502 Some spurious warnings can be avoided if you declare all the functions
3503 you use that never return as @code{noreturn}. @xref{Function
3506 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3508 @item -Wunknown-pragmas
3509 @opindex Wunknown-pragmas
3510 @opindex Wno-unknown-pragmas
3511 @cindex warning for unknown pragmas
3512 @cindex unknown pragmas, warning
3513 @cindex pragmas, warning of unknown
3514 Warn when a #pragma directive is encountered which is not understood by
3515 GCC@. If this command line option is used, warnings will even be issued
3516 for unknown pragmas in system header files. This is not the case if
3517 the warnings were only enabled by the @option{-Wall} command line option.
3520 @opindex Wno-pragmas
3522 Do not warn about misuses of pragmas, such as incorrect parameters,
3523 invalid syntax, or conflicts between pragmas. See also
3524 @samp{-Wunknown-pragmas}.
3526 @item -Wstrict-aliasing
3527 @opindex Wstrict-aliasing
3528 @opindex Wno-strict-aliasing
3529 This option is only active when @option{-fstrict-aliasing} is active.
3530 It warns about code which might break the strict aliasing rules that the
3531 compiler is using for optimization. The warning does not catch all
3532 cases, but does attempt to catch the more common pitfalls. It is
3533 included in @option{-Wall}.
3534 It is equivalent to @option{-Wstrict-aliasing=3}
3536 @item -Wstrict-aliasing=n
3537 @opindex Wstrict-aliasing=n
3538 @opindex Wno-strict-aliasing=n
3539 This option is only active when @option{-fstrict-aliasing} is active.
3540 It warns about code which might break the strict aliasing rules that the
3541 compiler is using for optimization.
3542 Higher levels correspond to higher accuracy (fewer false positives).
3543 Higher levels also correspond to more effort, similar to the way -O works.
3544 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3547 Level 1: Most aggressive, quick, least accurate.
3548 Possibly useful when higher levels
3549 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3550 false negatives. However, it has many false positives.
3551 Warns for all pointer conversions between possibly incompatible types,
3552 even if never dereferenced. Runs in the frontend only.
3554 Level 2: Aggressive, quick, not too precise.
3555 May still have many false positives (not as many as level 1 though),
3556 and few false negatives (but possibly more than level 1).
3557 Unlike level 1, it only warns when an address is taken. Warns about
3558 incomplete types. Runs in the frontend only.
3560 Level 3 (default for @option{-Wstrict-aliasing}):
3561 Should have very few false positives and few false
3562 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3563 Takes care of the common pun+dereference pattern in the frontend:
3564 @code{*(int*)&some_float}.
3565 If optimization is enabled, it also runs in the backend, where it deals
3566 with multiple statement cases using flow-sensitive points-to information.
3567 Only warns when the converted pointer is dereferenced.
3568 Does not warn about incomplete types.
3570 @item -Wstrict-overflow
3571 @itemx -Wstrict-overflow=@var{n}
3572 @opindex Wstrict-overflow
3573 @opindex Wno-strict-overflow
3574 This option is only active when @option{-fstrict-overflow} is active.
3575 It warns about cases where the compiler optimizes based on the
3576 assumption that signed overflow does not occur. Note that it does not
3577 warn about all cases where the code might overflow: it only warns
3578 about cases where the compiler implements some optimization. Thus
3579 this warning depends on the optimization level.
3581 An optimization which assumes that signed overflow does not occur is
3582 perfectly safe if the values of the variables involved are such that
3583 overflow never does, in fact, occur. Therefore this warning can
3584 easily give a false positive: a warning about code which is not
3585 actually a problem. To help focus on important issues, several
3586 warning levels are defined. No warnings are issued for the use of
3587 undefined signed overflow when estimating how many iterations a loop
3588 will require, in particular when determining whether a loop will be
3592 @item -Wstrict-overflow=1
3593 Warn about cases which are both questionable and easy to avoid. For
3594 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3595 compiler will simplify this to @code{1}. This level of
3596 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3597 are not, and must be explicitly requested.
3599 @item -Wstrict-overflow=2
3600 Also warn about other cases where a comparison is simplified to a
3601 constant. For example: @code{abs (x) >= 0}. This can only be
3602 simplified when @option{-fstrict-overflow} is in effect, because
3603 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3604 zero. @option{-Wstrict-overflow} (with no level) is the same as
3605 @option{-Wstrict-overflow=2}.
3607 @item -Wstrict-overflow=3
3608 Also warn about other cases where a comparison is simplified. For
3609 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3611 @item -Wstrict-overflow=4
3612 Also warn about other simplifications not covered by the above cases.
3613 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3615 @item -Wstrict-overflow=5
3616 Also warn about cases where the compiler reduces the magnitude of a
3617 constant involved in a comparison. For example: @code{x + 2 > y} will
3618 be simplified to @code{x + 1 >= y}. This is reported only at the
3619 highest warning level because this simplification applies to many
3620 comparisons, so this warning level will give a very large number of
3624 @item -Warray-bounds
3625 @opindex Wno-array-bounds
3626 @opindex Warray-bounds
3627 This option is only active when @option{-ftree-vrp} is active
3628 (default for -O2 and above). It warns about subscripts to arrays
3629 that are always out of bounds. This warning is enabled by @option{-Wall}.
3631 @item -Wno-div-by-zero
3632 @opindex Wno-div-by-zero
3633 @opindex Wdiv-by-zero
3634 Do not warn about compile-time integer division by zero. Floating point
3635 division by zero is not warned about, as it can be a legitimate way of
3636 obtaining infinities and NaNs.
3638 @item -Wsystem-headers
3639 @opindex Wsystem-headers
3640 @opindex Wno-system-headers
3641 @cindex warnings from system headers
3642 @cindex system headers, warnings from
3643 Print warning messages for constructs found in system header files.
3644 Warnings from system headers are normally suppressed, on the assumption
3645 that they usually do not indicate real problems and would only make the
3646 compiler output harder to read. Using this command line option tells
3647 GCC to emit warnings from system headers as if they occurred in user
3648 code. However, note that using @option{-Wall} in conjunction with this
3649 option will @emph{not} warn about unknown pragmas in system
3650 headers---for that, @option{-Wunknown-pragmas} must also be used.
3653 @opindex Wfloat-equal
3654 @opindex Wno-float-equal
3655 Warn if floating point values are used in equality comparisons.
3657 The idea behind this is that sometimes it is convenient (for the
3658 programmer) to consider floating-point values as approximations to
3659 infinitely precise real numbers. If you are doing this, then you need
3660 to compute (by analyzing the code, or in some other way) the maximum or
3661 likely maximum error that the computation introduces, and allow for it
3662 when performing comparisons (and when producing output, but that's a
3663 different problem). In particular, instead of testing for equality, you
3664 would check to see whether the two values have ranges that overlap; and
3665 this is done with the relational operators, so equality comparisons are
3668 @item -Wtraditional @r{(C and Objective-C only)}
3669 @opindex Wtraditional
3670 @opindex Wno-traditional
3671 Warn about certain constructs that behave differently in traditional and
3672 ISO C@. Also warn about ISO C constructs that have no traditional C
3673 equivalent, and/or problematic constructs which should be avoided.
3677 Macro parameters that appear within string literals in the macro body.
3678 In traditional C macro replacement takes place within string literals,
3679 but does not in ISO C@.
3682 In traditional C, some preprocessor directives did not exist.
3683 Traditional preprocessors would only consider a line to be a directive
3684 if the @samp{#} appeared in column 1 on the line. Therefore
3685 @option{-Wtraditional} warns about directives that traditional C
3686 understands but would ignore because the @samp{#} does not appear as the
3687 first character on the line. It also suggests you hide directives like
3688 @samp{#pragma} not understood by traditional C by indenting them. Some
3689 traditional implementations would not recognize @samp{#elif}, so it
3690 suggests avoiding it altogether.
3693 A function-like macro that appears without arguments.
3696 The unary plus operator.
3699 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3700 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3701 constants.) Note, these suffixes appear in macros defined in the system
3702 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3703 Use of these macros in user code might normally lead to spurious
3704 warnings, however GCC's integrated preprocessor has enough context to
3705 avoid warning in these cases.
3708 A function declared external in one block and then used after the end of
3712 A @code{switch} statement has an operand of type @code{long}.
3715 A non-@code{static} function declaration follows a @code{static} one.
3716 This construct is not accepted by some traditional C compilers.
3719 The ISO type of an integer constant has a different width or
3720 signedness from its traditional type. This warning is only issued if
3721 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3722 typically represent bit patterns, are not warned about.
3725 Usage of ISO string concatenation is detected.
3728 Initialization of automatic aggregates.
3731 Identifier conflicts with labels. Traditional C lacks a separate
3732 namespace for labels.
3735 Initialization of unions. If the initializer is zero, the warning is
3736 omitted. This is done under the assumption that the zero initializer in
3737 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3738 initializer warnings and relies on default initialization to zero in the
3742 Conversions by prototypes between fixed/floating point values and vice
3743 versa. The absence of these prototypes when compiling with traditional
3744 C would cause serious problems. This is a subset of the possible
3745 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3748 Use of ISO C style function definitions. This warning intentionally is
3749 @emph{not} issued for prototype declarations or variadic functions
3750 because these ISO C features will appear in your code when using
3751 libiberty's traditional C compatibility macros, @code{PARAMS} and
3752 @code{VPARAMS}. This warning is also bypassed for nested functions
3753 because that feature is already a GCC extension and thus not relevant to
3754 traditional C compatibility.
3757 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3758 @opindex Wtraditional-conversion
3759 @opindex Wno-traditional-conversion
3760 Warn if a prototype causes a type conversion that is different from what
3761 would happen to the same argument in the absence of a prototype. This
3762 includes conversions of fixed point to floating and vice versa, and
3763 conversions changing the width or signedness of a fixed point argument
3764 except when the same as the default promotion.
3766 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3767 @opindex Wdeclaration-after-statement
3768 @opindex Wno-declaration-after-statement
3769 Warn when a declaration is found after a statement in a block. This
3770 construct, known from C++, was introduced with ISO C99 and is by default
3771 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3772 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3777 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3779 @item -Wno-endif-labels
3780 @opindex Wno-endif-labels
3781 @opindex Wendif-labels
3782 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3787 Warn whenever a local variable shadows another local variable, parameter or
3788 global variable or whenever a built-in function is shadowed.
3790 @item -Wlarger-than=@var{len}
3791 @opindex Wlarger-than=@var{len}
3792 @opindex Wlarger-than-@var{len}
3793 Warn whenever an object of larger than @var{len} bytes is defined.
3795 @item -Wframe-larger-than=@var{len}
3796 @opindex Wframe-larger-than
3797 Warn if the size of a function frame is larger than @var{len} bytes.
3798 The computation done to determine the stack frame size is approximate
3799 and not conservative.
3800 The actual requirements may be somewhat greater than @var{len}
3801 even if you do not get a warning. In addition, any space allocated
3802 via @code{alloca}, variable-length arrays, or related constructs
3803 is not included by the compiler when determining
3804 whether or not to issue a warning.
3806 @item -Wunsafe-loop-optimizations
3807 @opindex Wunsafe-loop-optimizations
3808 @opindex Wno-unsafe-loop-optimizations
3809 Warn if the loop cannot be optimized because the compiler could not
3810 assume anything on the bounds of the loop indices. With
3811 @option{-funsafe-loop-optimizations} warn if the compiler made
3814 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3815 @opindex Wno-pedantic-ms-format
3816 @opindex Wpedantic-ms-format
3817 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3818 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3819 depending on the MS runtime, when you are using the options @option{-Wformat}
3820 and @option{-pedantic} without gnu-extensions.
3822 @item -Wpointer-arith
3823 @opindex Wpointer-arith
3824 @opindex Wno-pointer-arith
3825 Warn about anything that depends on the ``size of'' a function type or
3826 of @code{void}. GNU C assigns these types a size of 1, for
3827 convenience in calculations with @code{void *} pointers and pointers
3828 to functions. In C++, warn also when an arithmetic operation involves
3829 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3832 @opindex Wtype-limits
3833 @opindex Wno-type-limits
3834 Warn if a comparison is always true or always false due to the limited
3835 range of the data type, but do not warn for constant expressions. For
3836 example, warn if an unsigned variable is compared against zero with
3837 @samp{<} or @samp{>=}. This warning is also enabled by
3840 @item -Wbad-function-cast @r{(C and Objective-C only)}
3841 @opindex Wbad-function-cast
3842 @opindex Wno-bad-function-cast
3843 Warn whenever a function call is cast to a non-matching type.
3844 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3846 @item -Wc++-compat @r{(C and Objective-C only)}
3847 Warn about ISO C constructs that are outside of the common subset of
3848 ISO C and ISO C++, e.g.@: request for implicit conversion from
3849 @code{void *} to a pointer to non-@code{void} type.
3851 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3852 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3853 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3854 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3858 @opindex Wno-cast-qual
3859 Warn whenever a pointer is cast so as to remove a type qualifier from
3860 the target type. For example, warn if a @code{const char *} is cast
3861 to an ordinary @code{char *}.
3863 Also warn when making a cast which introduces a type qualifier in an
3864 unsafe way. For example, casting @code{char **} to @code{const char **}
3865 is unsafe, as in this example:
3868 /* p is char ** value. */
3869 const char **q = (const char **) p;
3870 /* Assignment of readonly string to const char * is OK. */
3872 /* Now char** pointer points to read-only memory. */
3877 @opindex Wcast-align
3878 @opindex Wno-cast-align
3879 Warn whenever a pointer is cast such that the required alignment of the
3880 target is increased. For example, warn if a @code{char *} is cast to
3881 an @code{int *} on machines where integers can only be accessed at
3882 two- or four-byte boundaries.
3884 @item -Wwrite-strings
3885 @opindex Wwrite-strings
3886 @opindex Wno-write-strings
3887 When compiling C, give string constants the type @code{const
3888 char[@var{length}]} so that copying the address of one into a
3889 non-@code{const} @code{char *} pointer will get a warning. These
3890 warnings will help you find at compile time code that can try to write
3891 into a string constant, but only if you have been very careful about
3892 using @code{const} in declarations and prototypes. Otherwise, it will
3893 just be a nuisance. This is why we did not make @option{-Wall} request
3896 When compiling C++, warn about the deprecated conversion from string
3897 literals to @code{char *}. This warning is enabled by default for C++
3902 @opindex Wno-clobbered
3903 Warn for variables that might be changed by @samp{longjmp} or
3904 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3907 @opindex Wconversion
3908 @opindex Wno-conversion
3909 Warn for implicit conversions that may alter a value. This includes
3910 conversions between real and integer, like @code{abs (x)} when
3911 @code{x} is @code{double}; conversions between signed and unsigned,
3912 like @code{unsigned ui = -1}; and conversions to smaller types, like
3913 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3914 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3915 changed by the conversion like in @code{abs (2.0)}. Warnings about
3916 conversions between signed and unsigned integers can be disabled by
3917 using @option{-Wno-sign-conversion}.
3919 For C++, also warn for confusing overload resolution for user-defined
3920 conversions; and conversions that will never use a type conversion
3921 operator: conversions to @code{void}, the same type, a base class or a
3922 reference to them. Warnings about conversions between signed and
3923 unsigned integers are disabled by default in C++ unless
3924 @option{-Wsign-conversion} is explicitly enabled.
3926 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
3927 @opindex Wconversion-null
3928 @opindex Wno-conversion-null
3929 Do not warn for conversions between @code{NULL} and non-pointer
3930 types. @option{-Wconversion-null} is enabled by default.
3933 @opindex Wempty-body
3934 @opindex Wno-empty-body
3935 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3936 while} statement. This warning is also enabled by @option{-Wextra}.
3938 @item -Wenum-compare
3939 @opindex Wenum-compare
3940 @opindex Wno-enum-compare
3941 Warn about a comparison between values of different enum types. In C++
3942 this warning is enabled by default. In C this warning is enabled by
3945 @item -Wjump-misses-init @r{(C, Objective-C only)}
3946 @opindex Wjump-misses-init
3947 @opindex Wno-jump-misses-init
3948 Warn if a @code{goto} statement or a @code{switch} statement jumps
3949 forward across the initialization of a variable, or jumps backward to a
3950 label after the variable has been initialized. This only warns about
3951 variables which are initialized when they are declared. This warning is
3952 only supported for C and Objective C; in C++ this sort of branch is an
3955 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
3956 can be disabled with the @option{-Wno-jump-misses-init} option.
3958 @item -Wsign-compare
3959 @opindex Wsign-compare
3960 @opindex Wno-sign-compare
3961 @cindex warning for comparison of signed and unsigned values
3962 @cindex comparison of signed and unsigned values, warning
3963 @cindex signed and unsigned values, comparison warning
3964 Warn when a comparison between signed and unsigned values could produce
3965 an incorrect result when the signed value is converted to unsigned.
3966 This warning is also enabled by @option{-Wextra}; to get the other warnings
3967 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3969 @item -Wsign-conversion
3970 @opindex Wsign-conversion
3971 @opindex Wno-sign-conversion
3972 Warn for implicit conversions that may change the sign of an integer
3973 value, like assigning a signed integer expression to an unsigned
3974 integer variable. An explicit cast silences the warning. In C, this
3975 option is enabled also by @option{-Wconversion}.
3979 @opindex Wno-address
3980 Warn about suspicious uses of memory addresses. These include using
3981 the address of a function in a conditional expression, such as
3982 @code{void func(void); if (func)}, and comparisons against the memory
3983 address of a string literal, such as @code{if (x == "abc")}. Such
3984 uses typically indicate a programmer error: the address of a function
3985 always evaluates to true, so their use in a conditional usually
3986 indicate that the programmer forgot the parentheses in a function
3987 call; and comparisons against string literals result in unspecified
3988 behavior and are not portable in C, so they usually indicate that the
3989 programmer intended to use @code{strcmp}. This warning is enabled by
3993 @opindex Wlogical-op
3994 @opindex Wno-logical-op
3995 Warn about suspicious uses of logical operators in expressions.
3996 This includes using logical operators in contexts where a
3997 bit-wise operator is likely to be expected.
3999 @item -Waggregate-return
4000 @opindex Waggregate-return
4001 @opindex Wno-aggregate-return
4002 Warn if any functions that return structures or unions are defined or
4003 called. (In languages where you can return an array, this also elicits
4006 @item -Wno-attributes
4007 @opindex Wno-attributes
4008 @opindex Wattributes
4009 Do not warn if an unexpected @code{__attribute__} is used, such as
4010 unrecognized attributes, function attributes applied to variables,
4011 etc. This will not stop errors for incorrect use of supported
4014 @item -Wno-builtin-macro-redefined
4015 @opindex Wno-builtin-macro-redefined
4016 @opindex Wbuiltin-macro-redefined
4017 Do not warn if certain built-in macros are redefined. This suppresses
4018 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4019 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4021 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4022 @opindex Wstrict-prototypes
4023 @opindex Wno-strict-prototypes
4024 Warn if a function is declared or defined without specifying the
4025 argument types. (An old-style function definition is permitted without
4026 a warning if preceded by a declaration which specifies the argument
4029 @item -Wold-style-declaration @r{(C and Objective-C only)}
4030 @opindex Wold-style-declaration
4031 @opindex Wno-old-style-declaration
4032 Warn for obsolescent usages, according to the C Standard, in a
4033 declaration. For example, warn if storage-class specifiers like
4034 @code{static} are not the first things in a declaration. This warning
4035 is also enabled by @option{-Wextra}.
4037 @item -Wold-style-definition @r{(C and Objective-C only)}
4038 @opindex Wold-style-definition
4039 @opindex Wno-old-style-definition
4040 Warn if an old-style function definition is used. A warning is given
4041 even if there is a previous prototype.
4043 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4044 @opindex Wmissing-parameter-type
4045 @opindex Wno-missing-parameter-type
4046 A function parameter is declared without a type specifier in K&R-style
4053 This warning is also enabled by @option{-Wextra}.
4055 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4056 @opindex Wmissing-prototypes
4057 @opindex Wno-missing-prototypes
4058 Warn if a global function is defined without a previous prototype
4059 declaration. This warning is issued even if the definition itself
4060 provides a prototype. The aim is to detect global functions that fail
4061 to be declared in header files.
4063 @item -Wmissing-declarations
4064 @opindex Wmissing-declarations
4065 @opindex Wno-missing-declarations
4066 Warn if a global function is defined without a previous declaration.
4067 Do so even if the definition itself provides a prototype.
4068 Use this option to detect global functions that are not declared in
4069 header files. In C++, no warnings are issued for function templates,
4070 or for inline functions, or for functions in anonymous namespaces.
4072 @item -Wmissing-field-initializers
4073 @opindex Wmissing-field-initializers
4074 @opindex Wno-missing-field-initializers
4078 Warn if a structure's initializer has some fields missing. For
4079 example, the following code would cause such a warning, because
4080 @code{x.h} is implicitly zero:
4083 struct s @{ int f, g, h; @};
4084 struct s x = @{ 3, 4 @};
4087 This option does not warn about designated initializers, so the following
4088 modification would not trigger a warning:
4091 struct s @{ int f, g, h; @};
4092 struct s x = @{ .f = 3, .g = 4 @};
4095 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4096 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4098 @item -Wmissing-noreturn
4099 @opindex Wmissing-noreturn
4100 @opindex Wno-missing-noreturn
4101 Warn about functions which might be candidates for attribute @code{noreturn}.
4102 Note these are only possible candidates, not absolute ones. Care should
4103 be taken to manually verify functions actually do not ever return before
4104 adding the @code{noreturn} attribute, otherwise subtle code generation
4105 bugs could be introduced. You will not get a warning for @code{main} in
4106 hosted C environments.
4108 @item -Wmissing-format-attribute
4109 @opindex Wmissing-format-attribute
4110 @opindex Wno-missing-format-attribute
4113 Warn about function pointers which might be candidates for @code{format}
4114 attributes. Note these are only possible candidates, not absolute ones.
4115 GCC will guess that function pointers with @code{format} attributes that
4116 are used in assignment, initialization, parameter passing or return
4117 statements should have a corresponding @code{format} attribute in the
4118 resulting type. I.e.@: the left-hand side of the assignment or
4119 initialization, the type of the parameter variable, or the return type
4120 of the containing function respectively should also have a @code{format}
4121 attribute to avoid the warning.
4123 GCC will also warn about function definitions which might be
4124 candidates for @code{format} attributes. Again, these are only
4125 possible candidates. GCC will guess that @code{format} attributes
4126 might be appropriate for any function that calls a function like
4127 @code{vprintf} or @code{vscanf}, but this might not always be the
4128 case, and some functions for which @code{format} attributes are
4129 appropriate may not be detected.
4131 @item -Wno-multichar
4132 @opindex Wno-multichar
4134 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4135 Usually they indicate a typo in the user's code, as they have
4136 implementation-defined values, and should not be used in portable code.
4138 @item -Wnormalized=<none|id|nfc|nfkc>
4139 @opindex Wnormalized=
4142 @cindex character set, input normalization
4143 In ISO C and ISO C++, two identifiers are different if they are
4144 different sequences of characters. However, sometimes when characters
4145 outside the basic ASCII character set are used, you can have two
4146 different character sequences that look the same. To avoid confusion,
4147 the ISO 10646 standard sets out some @dfn{normalization rules} which
4148 when applied ensure that two sequences that look the same are turned into
4149 the same sequence. GCC can warn you if you are using identifiers which
4150 have not been normalized; this option controls that warning.
4152 There are four levels of warning that GCC supports. The default is
4153 @option{-Wnormalized=nfc}, which warns about any identifier which is
4154 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4155 recommended form for most uses.
4157 Unfortunately, there are some characters which ISO C and ISO C++ allow
4158 in identifiers that when turned into NFC aren't allowable as
4159 identifiers. That is, there's no way to use these symbols in portable
4160 ISO C or C++ and have all your identifiers in NFC@.
4161 @option{-Wnormalized=id} suppresses the warning for these characters.
4162 It is hoped that future versions of the standards involved will correct
4163 this, which is why this option is not the default.
4165 You can switch the warning off for all characters by writing
4166 @option{-Wnormalized=none}. You would only want to do this if you
4167 were using some other normalization scheme (like ``D''), because
4168 otherwise you can easily create bugs that are literally impossible to see.
4170 Some characters in ISO 10646 have distinct meanings but look identical
4171 in some fonts or display methodologies, especially once formatting has
4172 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4173 LETTER N'', will display just like a regular @code{n} which has been
4174 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4175 normalization scheme to convert all these into a standard form as
4176 well, and GCC will warn if your code is not in NFKC if you use
4177 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4178 about every identifier that contains the letter O because it might be
4179 confused with the digit 0, and so is not the default, but may be
4180 useful as a local coding convention if the programming environment is
4181 unable to be fixed to display these characters distinctly.
4183 @item -Wno-deprecated
4184 @opindex Wno-deprecated
4185 @opindex Wdeprecated
4186 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4188 @item -Wno-deprecated-declarations
4189 @opindex Wno-deprecated-declarations
4190 @opindex Wdeprecated-declarations
4191 Do not warn about uses of functions (@pxref{Function Attributes}),
4192 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4193 Attributes}) marked as deprecated by using the @code{deprecated}
4197 @opindex Wno-overflow
4199 Do not warn about compile-time overflow in constant expressions.
4201 @item -Woverride-init @r{(C and Objective-C only)}
4202 @opindex Woverride-init
4203 @opindex Wno-override-init
4207 Warn if an initialized field without side effects is overridden when
4208 using designated initializers (@pxref{Designated Inits, , Designated
4211 This warning is included in @option{-Wextra}. To get other
4212 @option{-Wextra} warnings without this one, use @samp{-Wextra
4213 -Wno-override-init}.
4218 Warn if a structure is given the packed attribute, but the packed
4219 attribute has no effect on the layout or size of the structure.
4220 Such structures may be mis-aligned for little benefit. For
4221 instance, in this code, the variable @code{f.x} in @code{struct bar}
4222 will be misaligned even though @code{struct bar} does not itself
4223 have the packed attribute:
4230 @} __attribute__((packed));
4238 @item -Wpacked-bitfield-compat
4239 @opindex Wpacked-bitfield-compat
4240 @opindex Wno-packed-bitfield-compat
4241 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4242 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4243 the change can lead to differences in the structure layout. GCC
4244 informs you when the offset of such a field has changed in GCC 4.4.
4245 For example there is no longer a 4-bit padding between field @code{a}
4246 and @code{b} in this structure:
4253 @} __attribute__ ((packed));
4256 This warning is enabled by default. Use
4257 @option{-Wno-packed-bitfield-compat} to disable this warning.
4262 Warn if padding is included in a structure, either to align an element
4263 of the structure or to align the whole structure. Sometimes when this
4264 happens it is possible to rearrange the fields of the structure to
4265 reduce the padding and so make the structure smaller.
4267 @item -Wredundant-decls
4268 @opindex Wredundant-decls
4269 @opindex Wno-redundant-decls
4270 Warn if anything is declared more than once in the same scope, even in
4271 cases where multiple declaration is valid and changes nothing.
4273 @item -Wnested-externs @r{(C and Objective-C only)}
4274 @opindex Wnested-externs
4275 @opindex Wno-nested-externs
4276 Warn if an @code{extern} declaration is encountered within a function.
4281 Warn if a function can not be inlined and it was declared as inline.
4282 Even with this option, the compiler will not warn about failures to
4283 inline functions declared in system headers.
4285 The compiler uses a variety of heuristics to determine whether or not
4286 to inline a function. For example, the compiler takes into account
4287 the size of the function being inlined and the amount of inlining
4288 that has already been done in the current function. Therefore,
4289 seemingly insignificant changes in the source program can cause the
4290 warnings produced by @option{-Winline} to appear or disappear.
4292 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4293 @opindex Wno-invalid-offsetof
4294 @opindex Winvalid-offsetof
4295 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4296 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4297 to a non-POD type is undefined. In existing C++ implementations,
4298 however, @samp{offsetof} typically gives meaningful results even when
4299 applied to certain kinds of non-POD types. (Such as a simple
4300 @samp{struct} that fails to be a POD type only by virtue of having a
4301 constructor.) This flag is for users who are aware that they are
4302 writing nonportable code and who have deliberately chosen to ignore the
4305 The restrictions on @samp{offsetof} may be relaxed in a future version
4306 of the C++ standard.
4308 @item -Wno-int-to-pointer-cast
4309 @opindex Wno-int-to-pointer-cast
4310 @opindex Wint-to-pointer-cast
4311 Suppress warnings from casts to pointer type of an integer of a
4312 different size. In C++, casting to a pointer type of smaller size is
4313 an error. @option{Wint-to-pointer-cast} is enabled by default.
4316 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4317 @opindex Wno-pointer-to-int-cast
4318 @opindex Wpointer-to-int-cast
4319 Suppress warnings from casts from a pointer to an integer type of a
4323 @opindex Winvalid-pch
4324 @opindex Wno-invalid-pch
4325 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4326 the search path but can't be used.
4330 @opindex Wno-long-long
4331 Warn if @samp{long long} type is used. This is enabled by either
4332 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4333 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4335 @item -Wvariadic-macros
4336 @opindex Wvariadic-macros
4337 @opindex Wno-variadic-macros
4338 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4339 alternate syntax when in pedantic ISO C99 mode. This is default.
4340 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4345 Warn if variable length array is used in the code.
4346 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4347 the variable length array.
4349 @item -Wvolatile-register-var
4350 @opindex Wvolatile-register-var
4351 @opindex Wno-volatile-register-var
4352 Warn if a register variable is declared volatile. The volatile
4353 modifier does not inhibit all optimizations that may eliminate reads
4354 and/or writes to register variables. This warning is enabled by
4357 @item -Wdisabled-optimization
4358 @opindex Wdisabled-optimization
4359 @opindex Wno-disabled-optimization
4360 Warn if a requested optimization pass is disabled. This warning does
4361 not generally indicate that there is anything wrong with your code; it
4362 merely indicates that GCC's optimizers were unable to handle the code
4363 effectively. Often, the problem is that your code is too big or too
4364 complex; GCC will refuse to optimize programs when the optimization
4365 itself is likely to take inordinate amounts of time.
4367 @item -Wpointer-sign @r{(C and Objective-C only)}
4368 @opindex Wpointer-sign
4369 @opindex Wno-pointer-sign
4370 Warn for pointer argument passing or assignment with different signedness.
4371 This option is only supported for C and Objective-C@. It is implied by
4372 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4373 @option{-Wno-pointer-sign}.
4375 @item -Wstack-protector
4376 @opindex Wstack-protector
4377 @opindex Wno-stack-protector
4378 This option is only active when @option{-fstack-protector} is active. It
4379 warns about functions that will not be protected against stack smashing.
4382 @opindex Wno-mudflap
4383 Suppress warnings about constructs that cannot be instrumented by
4386 @item -Woverlength-strings
4387 @opindex Woverlength-strings
4388 @opindex Wno-overlength-strings
4389 Warn about string constants which are longer than the ``minimum
4390 maximum'' length specified in the C standard. Modern compilers
4391 generally allow string constants which are much longer than the
4392 standard's minimum limit, but very portable programs should avoid
4393 using longer strings.
4395 The limit applies @emph{after} string constant concatenation, and does
4396 not count the trailing NUL@. In C90, the limit was 509 characters; in
4397 C99, it was raised to 4095. C++98 does not specify a normative
4398 minimum maximum, so we do not diagnose overlength strings in C++@.
4400 This option is implied by @option{-pedantic}, and can be disabled with
4401 @option{-Wno-overlength-strings}.
4403 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4404 @opindex Wunsuffixed-float-constants
4406 GCC will issue a warning for any floating constant that does not have
4407 a suffix. When used together with @option{-Wsystem-headers} it will
4408 warn about such constants in system header files. This can be useful
4409 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4410 from the decimal floating-point extension to C99.
4413 @node Debugging Options
4414 @section Options for Debugging Your Program or GCC
4415 @cindex options, debugging
4416 @cindex debugging information options
4418 GCC has various special options that are used for debugging
4419 either your program or GCC:
4424 Produce debugging information in the operating system's native format
4425 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4428 On most systems that use stabs format, @option{-g} enables use of extra
4429 debugging information that only GDB can use; this extra information
4430 makes debugging work better in GDB but will probably make other debuggers
4432 refuse to read the program. If you want to control for certain whether
4433 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4434 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4436 GCC allows you to use @option{-g} with
4437 @option{-O}. The shortcuts taken by optimized code may occasionally
4438 produce surprising results: some variables you declared may not exist
4439 at all; flow of control may briefly move where you did not expect it;
4440 some statements may not be executed because they compute constant
4441 results or their values were already at hand; some statements may
4442 execute in different places because they were moved out of loops.
4444 Nevertheless it proves possible to debug optimized output. This makes
4445 it reasonable to use the optimizer for programs that might have bugs.
4447 The following options are useful when GCC is generated with the
4448 capability for more than one debugging format.
4452 Produce debugging information for use by GDB@. This means to use the
4453 most expressive format available (DWARF 2, stabs, or the native format
4454 if neither of those are supported), including GDB extensions if at all
4459 Produce debugging information in stabs format (if that is supported),
4460 without GDB extensions. This is the format used by DBX on most BSD
4461 systems. On MIPS, Alpha and System V Release 4 systems this option
4462 produces stabs debugging output which is not understood by DBX or SDB@.
4463 On System V Release 4 systems this option requires the GNU assembler.
4465 @item -feliminate-unused-debug-symbols
4466 @opindex feliminate-unused-debug-symbols
4467 Produce debugging information in stabs format (if that is supported),
4468 for only symbols that are actually used.
4470 @item -femit-class-debug-always
4471 Instead of emitting debugging information for a C++ class in only one
4472 object file, emit it in all object files using the class. This option
4473 should be used only with debuggers that are unable to handle the way GCC
4474 normally emits debugging information for classes because using this
4475 option will increase the size of debugging information by as much as a
4480 Produce debugging information in stabs format (if that is supported),
4481 using GNU extensions understood only by the GNU debugger (GDB)@. The
4482 use of these extensions is likely to make other debuggers crash or
4483 refuse to read the program.
4487 Produce debugging information in COFF format (if that is supported).
4488 This is the format used by SDB on most System V systems prior to
4493 Produce debugging information in XCOFF format (if that is supported).
4494 This is the format used by the DBX debugger on IBM RS/6000 systems.
4498 Produce debugging information in XCOFF format (if that is supported),
4499 using GNU extensions understood only by the GNU debugger (GDB)@. The
4500 use of these extensions is likely to make other debuggers crash or
4501 refuse to read the program, and may cause assemblers other than the GNU
4502 assembler (GAS) to fail with an error.
4504 @item -gdwarf-@var{version}
4505 @opindex gdwarf-@var{version}
4506 Produce debugging information in DWARF format (if that is
4507 supported). This is the format used by DBX on IRIX 6. The value
4508 of @var{version} may be either 2, 3 or 4; the default version is 2.
4510 Note that with DWARF version 2 some ports require, and will always
4511 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4513 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4514 for maximum benefit.
4516 @item -gstrict-dwarf
4517 @opindex gstrict-dwarf
4518 Disallow using extensions of later DWARF standard version than selected
4519 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4520 DWARF extensions from later standard versions is allowed.
4522 @item -gno-strict-dwarf
4523 @opindex gno-strict-dwarf
4524 Allow using extensions of later DWARF standard version than selected with
4525 @option{-gdwarf-@var{version}}.
4529 Produce debugging information in VMS debug format (if that is
4530 supported). This is the format used by DEBUG on VMS systems.
4533 @itemx -ggdb@var{level}
4534 @itemx -gstabs@var{level}
4535 @itemx -gcoff@var{level}
4536 @itemx -gxcoff@var{level}
4537 @itemx -gvms@var{level}
4538 Request debugging information and also use @var{level} to specify how
4539 much information. The default level is 2.
4541 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4544 Level 1 produces minimal information, enough for making backtraces in
4545 parts of the program that you don't plan to debug. This includes
4546 descriptions of functions and external variables, but no information
4547 about local variables and no line numbers.
4549 Level 3 includes extra information, such as all the macro definitions
4550 present in the program. Some debuggers support macro expansion when
4551 you use @option{-g3}.
4553 @option{-gdwarf-2} does not accept a concatenated debug level, because
4554 GCC used to support an option @option{-gdwarf} that meant to generate
4555 debug information in version 1 of the DWARF format (which is very
4556 different from version 2), and it would have been too confusing. That
4557 debug format is long obsolete, but the option cannot be changed now.
4558 Instead use an additional @option{-g@var{level}} option to change the
4559 debug level for DWARF.
4563 Turn off generation of debug info, if leaving out this option would have
4564 generated it, or turn it on at level 2 otherwise. The position of this
4565 argument in the command line does not matter, it takes effect after all
4566 other options are processed, and it does so only once, no matter how
4567 many times it is given. This is mainly intended to be used with
4568 @option{-fcompare-debug}.
4570 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4571 @opindex fdump-final-insns
4572 Dump the final internal representation (RTL) to @var{file}. If the
4573 optional argument is omitted (or if @var{file} is @code{.}), the name
4574 of the dump file will be determined by appending @code{.gkd} to the
4575 compilation output file name.
4577 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4578 @opindex fcompare-debug
4579 @opindex fno-compare-debug
4580 If no error occurs during compilation, run the compiler a second time,
4581 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4582 passed to the second compilation. Dump the final internal
4583 representation in both compilations, and print an error if they differ.
4585 If the equal sign is omitted, the default @option{-gtoggle} is used.
4587 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4588 and nonzero, implicitly enables @option{-fcompare-debug}. If
4589 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4590 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4593 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4594 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4595 of the final representation and the second compilation, preventing even
4596 @env{GCC_COMPARE_DEBUG} from taking effect.
4598 To verify full coverage during @option{-fcompare-debug} testing, set
4599 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4600 which GCC will reject as an invalid option in any actual compilation
4601 (rather than preprocessing, assembly or linking). To get just a
4602 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4603 not overridden} will do.
4605 @item -fcompare-debug-second
4606 @opindex fcompare-debug-second
4607 This option is implicitly passed to the compiler for the second
4608 compilation requested by @option{-fcompare-debug}, along with options to
4609 silence warnings, and omitting other options that would cause
4610 side-effect compiler outputs to files or to the standard output. Dump
4611 files and preserved temporary files are renamed so as to contain the
4612 @code{.gk} additional extension during the second compilation, to avoid
4613 overwriting those generated by the first.
4615 When this option is passed to the compiler driver, it causes the
4616 @emph{first} compilation to be skipped, which makes it useful for little
4617 other than debugging the compiler proper.
4619 @item -feliminate-dwarf2-dups
4620 @opindex feliminate-dwarf2-dups
4621 Compress DWARF2 debugging information by eliminating duplicated
4622 information about each symbol. This option only makes sense when
4623 generating DWARF2 debugging information with @option{-gdwarf-2}.
4625 @item -femit-struct-debug-baseonly
4626 Emit debug information for struct-like types
4627 only when the base name of the compilation source file
4628 matches the base name of file in which the struct was defined.
4630 This option substantially reduces the size of debugging information,
4631 but at significant potential loss in type information to the debugger.
4632 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4633 See @option{-femit-struct-debug-detailed} for more detailed control.
4635 This option works only with DWARF 2.
4637 @item -femit-struct-debug-reduced
4638 Emit debug information for struct-like types
4639 only when the base name of the compilation source file
4640 matches the base name of file in which the type was defined,
4641 unless the struct is a template or defined in a system header.
4643 This option significantly reduces the size of debugging information,
4644 with some potential loss in type information to the debugger.
4645 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4646 See @option{-femit-struct-debug-detailed} for more detailed control.
4648 This option works only with DWARF 2.
4650 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4651 Specify the struct-like types
4652 for which the compiler will generate debug information.
4653 The intent is to reduce duplicate struct debug information
4654 between different object files within the same program.
4656 This option is a detailed version of
4657 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4658 which will serve for most needs.
4660 A specification has the syntax
4661 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4663 The optional first word limits the specification to
4664 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4665 A struct type is used directly when it is the type of a variable, member.
4666 Indirect uses arise through pointers to structs.
4667 That is, when use of an incomplete struct would be legal, the use is indirect.
4669 @samp{struct one direct; struct two * indirect;}.
4671 The optional second word limits the specification to
4672 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4673 Generic structs are a bit complicated to explain.
4674 For C++, these are non-explicit specializations of template classes,
4675 or non-template classes within the above.
4676 Other programming languages have generics,
4677 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4679 The third word specifies the source files for those
4680 structs for which the compiler will emit debug information.
4681 The values @samp{none} and @samp{any} have the normal meaning.
4682 The value @samp{base} means that
4683 the base of name of the file in which the type declaration appears
4684 must match the base of the name of the main compilation file.
4685 In practice, this means that
4686 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4687 but types declared in other header will not.
4688 The value @samp{sys} means those types satisfying @samp{base}
4689 or declared in system or compiler headers.
4691 You may need to experiment to determine the best settings for your application.
4693 The default is @samp{-femit-struct-debug-detailed=all}.
4695 This option works only with DWARF 2.
4697 @item -fenable-icf-debug
4698 @opindex fenable-icf-debug
4699 Generate additional debug information to support identical code folding (ICF).
4700 This option only works with DWARF version 2 or higher.
4702 @item -fno-merge-debug-strings
4703 @opindex fmerge-debug-strings
4704 @opindex fno-merge-debug-strings
4705 Direct the linker to not merge together strings in the debugging
4706 information which are identical in different object files. Merging is
4707 not supported by all assemblers or linkers. Merging decreases the size
4708 of the debug information in the output file at the cost of increasing
4709 link processing time. Merging is enabled by default.
4711 @item -fdebug-prefix-map=@var{old}=@var{new}
4712 @opindex fdebug-prefix-map
4713 When compiling files in directory @file{@var{old}}, record debugging
4714 information describing them as in @file{@var{new}} instead.
4716 @item -fno-dwarf2-cfi-asm
4717 @opindex fdwarf2-cfi-asm
4718 @opindex fno-dwarf2-cfi-asm
4719 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4720 instead of using GAS @code{.cfi_*} directives.
4722 @cindex @command{prof}
4725 Generate extra code to write profile information suitable for the
4726 analysis program @command{prof}. You must use this option when compiling
4727 the source files you want data about, and you must also use it when
4730 @cindex @command{gprof}
4733 Generate extra code to write profile information suitable for the
4734 analysis program @command{gprof}. You must use this option when compiling
4735 the source files you want data about, and you must also use it when
4740 Makes the compiler print out each function name as it is compiled, and
4741 print some statistics about each pass when it finishes.
4744 @opindex ftime-report
4745 Makes the compiler print some statistics about the time consumed by each
4746 pass when it finishes.
4749 @opindex fmem-report
4750 Makes the compiler print some statistics about permanent memory
4751 allocation when it finishes.
4753 @item -fpre-ipa-mem-report
4754 @opindex fpre-ipa-mem-report
4755 @item -fpost-ipa-mem-report
4756 @opindex fpost-ipa-mem-report
4757 Makes the compiler print some statistics about permanent memory
4758 allocation before or after interprocedural optimization.
4760 @item -fprofile-arcs
4761 @opindex fprofile-arcs
4762 Add code so that program flow @dfn{arcs} are instrumented. During
4763 execution the program records how many times each branch and call is
4764 executed and how many times it is taken or returns. When the compiled
4765 program exits it saves this data to a file called
4766 @file{@var{auxname}.gcda} for each source file. The data may be used for
4767 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4768 test coverage analysis (@option{-ftest-coverage}). Each object file's
4769 @var{auxname} is generated from the name of the output file, if
4770 explicitly specified and it is not the final executable, otherwise it is
4771 the basename of the source file. In both cases any suffix is removed
4772 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4773 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4774 @xref{Cross-profiling}.
4776 @cindex @command{gcov}
4780 This option is used to compile and link code instrumented for coverage
4781 analysis. The option is a synonym for @option{-fprofile-arcs}
4782 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4783 linking). See the documentation for those options for more details.
4788 Compile the source files with @option{-fprofile-arcs} plus optimization
4789 and code generation options. For test coverage analysis, use the
4790 additional @option{-ftest-coverage} option. You do not need to profile
4791 every source file in a program.
4794 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4795 (the latter implies the former).
4798 Run the program on a representative workload to generate the arc profile
4799 information. This may be repeated any number of times. You can run
4800 concurrent instances of your program, and provided that the file system
4801 supports locking, the data files will be correctly updated. Also
4802 @code{fork} calls are detected and correctly handled (double counting
4806 For profile-directed optimizations, compile the source files again with
4807 the same optimization and code generation options plus
4808 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4809 Control Optimization}).
4812 For test coverage analysis, use @command{gcov} to produce human readable
4813 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4814 @command{gcov} documentation for further information.
4818 With @option{-fprofile-arcs}, for each function of your program GCC
4819 creates a program flow graph, then finds a spanning tree for the graph.
4820 Only arcs that are not on the spanning tree have to be instrumented: the
4821 compiler adds code to count the number of times that these arcs are
4822 executed. When an arc is the only exit or only entrance to a block, the
4823 instrumentation code can be added to the block; otherwise, a new basic
4824 block must be created to hold the instrumentation code.
4827 @item -ftest-coverage
4828 @opindex ftest-coverage
4829 Produce a notes file that the @command{gcov} code-coverage utility
4830 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4831 show program coverage. Each source file's note file is called
4832 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4833 above for a description of @var{auxname} and instructions on how to
4834 generate test coverage data. Coverage data will match the source files
4835 more closely, if you do not optimize.
4837 @item -fdbg-cnt-list
4838 @opindex fdbg-cnt-list
4839 Print the name and the counter upperbound for all debug counters.
4841 @item -fdbg-cnt=@var{counter-value-list}
4843 Set the internal debug counter upperbound. @var{counter-value-list}
4844 is a comma-separated list of @var{name}:@var{value} pairs
4845 which sets the upperbound of each debug counter @var{name} to @var{value}.
4846 All debug counters have the initial upperbound of @var{UINT_MAX},
4847 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4848 e.g. With -fdbg-cnt=dce:10,tail_call:0
4849 dbg_cnt(dce) will return true only for first 10 invocations
4850 and dbg_cnt(tail_call) will return false always.
4852 @item -d@var{letters}
4853 @itemx -fdump-rtl-@var{pass}
4855 Says to make debugging dumps during compilation at times specified by
4856 @var{letters}. This is used for debugging the RTL-based passes of the
4857 compiler. The file names for most of the dumps are made by appending
4858 a pass number and a word to the @var{dumpname}, and the files are
4859 created in the directory of the output file. @var{dumpname} is
4860 generated from the name of the output file, if explicitly specified
4861 and it is not an executable, otherwise it is the basename of the
4862 source file. These switches may have different effects when
4863 @option{-E} is used for preprocessing.
4865 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4866 @option{-d} option @var{letters}. Here are the possible
4867 letters for use in @var{pass} and @var{letters}, and their meanings:
4871 @item -fdump-rtl-alignments
4872 @opindex fdump-rtl-alignments
4873 Dump after branch alignments have been computed.
4875 @item -fdump-rtl-asmcons
4876 @opindex fdump-rtl-asmcons
4877 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4879 @item -fdump-rtl-auto_inc_dec
4880 @opindex fdump-rtl-auto_inc_dec
4881 Dump after auto-inc-dec discovery. This pass is only run on
4882 architectures that have auto inc or auto dec instructions.
4884 @item -fdump-rtl-barriers
4885 @opindex fdump-rtl-barriers
4886 Dump after cleaning up the barrier instructions.
4888 @item -fdump-rtl-bbpart
4889 @opindex fdump-rtl-bbpart
4890 Dump after partitioning hot and cold basic blocks.
4892 @item -fdump-rtl-bbro
4893 @opindex fdump-rtl-bbro
4894 Dump after block reordering.
4896 @item -fdump-rtl-btl1
4897 @itemx -fdump-rtl-btl2
4898 @opindex fdump-rtl-btl2
4899 @opindex fdump-rtl-btl2
4900 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4901 after the two branch
4902 target load optimization passes.
4904 @item -fdump-rtl-bypass
4905 @opindex fdump-rtl-bypass
4906 Dump after jump bypassing and control flow optimizations.
4908 @item -fdump-rtl-combine
4909 @opindex fdump-rtl-combine
4910 Dump after the RTL instruction combination pass.
4912 @item -fdump-rtl-compgotos
4913 @opindex fdump-rtl-compgotos
4914 Dump after duplicating the computed gotos.
4916 @item -fdump-rtl-ce1
4917 @itemx -fdump-rtl-ce2
4918 @itemx -fdump-rtl-ce3
4919 @opindex fdump-rtl-ce1
4920 @opindex fdump-rtl-ce2
4921 @opindex fdump-rtl-ce3
4922 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4923 @option{-fdump-rtl-ce3} enable dumping after the three
4924 if conversion passes.
4926 @itemx -fdump-rtl-cprop_hardreg
4927 @opindex fdump-rtl-cprop_hardreg
4928 Dump after hard register copy propagation.
4930 @itemx -fdump-rtl-csa
4931 @opindex fdump-rtl-csa
4932 Dump after combining stack adjustments.
4934 @item -fdump-rtl-cse1
4935 @itemx -fdump-rtl-cse2
4936 @opindex fdump-rtl-cse1
4937 @opindex fdump-rtl-cse2
4938 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4939 the two common sub-expression elimination passes.
4941 @itemx -fdump-rtl-dce
4942 @opindex fdump-rtl-dce
4943 Dump after the standalone dead code elimination passes.
4945 @itemx -fdump-rtl-dbr
4946 @opindex fdump-rtl-dbr
4947 Dump after delayed branch scheduling.
4949 @item -fdump-rtl-dce1
4950 @itemx -fdump-rtl-dce2
4951 @opindex fdump-rtl-dce1
4952 @opindex fdump-rtl-dce2
4953 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4954 the two dead store elimination passes.
4957 @opindex fdump-rtl-eh
4958 Dump after finalization of EH handling code.
4960 @item -fdump-rtl-eh_ranges
4961 @opindex fdump-rtl-eh_ranges
4962 Dump after conversion of EH handling range regions.
4964 @item -fdump-rtl-expand
4965 @opindex fdump-rtl-expand
4966 Dump after RTL generation.
4968 @item -fdump-rtl-fwprop1
4969 @itemx -fdump-rtl-fwprop2
4970 @opindex fdump-rtl-fwprop1
4971 @opindex fdump-rtl-fwprop2
4972 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4973 dumping after the two forward propagation passes.
4975 @item -fdump-rtl-gcse1
4976 @itemx -fdump-rtl-gcse2
4977 @opindex fdump-rtl-gcse1
4978 @opindex fdump-rtl-gcse2
4979 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4980 after global common subexpression elimination.
4982 @item -fdump-rtl-init-regs
4983 @opindex fdump-rtl-init-regs
4984 Dump after the initialization of the registers.
4986 @item -fdump-rtl-initvals
4987 @opindex fdump-rtl-initvals
4988 Dump after the computation of the initial value sets.
4990 @itemx -fdump-rtl-into_cfglayout
4991 @opindex fdump-rtl-into_cfglayout
4992 Dump after converting to cfglayout mode.
4994 @item -fdump-rtl-ira
4995 @opindex fdump-rtl-ira
4996 Dump after iterated register allocation.
4998 @item -fdump-rtl-jump
4999 @opindex fdump-rtl-jump
5000 Dump after the second jump optimization.
5002 @item -fdump-rtl-loop2
5003 @opindex fdump-rtl-loop2
5004 @option{-fdump-rtl-loop2} enables dumping after the rtl
5005 loop optimization passes.
5007 @item -fdump-rtl-mach
5008 @opindex fdump-rtl-mach
5009 Dump after performing the machine dependent reorganization pass, if that
5012 @item -fdump-rtl-mode_sw
5013 @opindex fdump-rtl-mode_sw
5014 Dump after removing redundant mode switches.
5016 @item -fdump-rtl-rnreg
5017 @opindex fdump-rtl-rnreg
5018 Dump after register renumbering.
5020 @itemx -fdump-rtl-outof_cfglayout
5021 @opindex fdump-rtl-outof_cfglayout
5022 Dump after converting from cfglayout mode.
5024 @item -fdump-rtl-peephole2
5025 @opindex fdump-rtl-peephole2
5026 Dump after the peephole pass.
5028 @item -fdump-rtl-postreload
5029 @opindex fdump-rtl-postreload
5030 Dump after post-reload optimizations.
5032 @itemx -fdump-rtl-pro_and_epilogue
5033 @opindex fdump-rtl-pro_and_epilogue
5034 Dump after generating the function pro and epilogues.
5036 @item -fdump-rtl-regmove
5037 @opindex fdump-rtl-regmove
5038 Dump after the register move pass.
5040 @item -fdump-rtl-sched1
5041 @itemx -fdump-rtl-sched2
5042 @opindex fdump-rtl-sched1
5043 @opindex fdump-rtl-sched2
5044 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5045 after the basic block scheduling passes.
5047 @item -fdump-rtl-see
5048 @opindex fdump-rtl-see
5049 Dump after sign extension elimination.
5051 @item -fdump-rtl-seqabstr
5052 @opindex fdump-rtl-seqabstr
5053 Dump after common sequence discovery.
5055 @item -fdump-rtl-shorten
5056 @opindex fdump-rtl-shorten
5057 Dump after shortening branches.
5059 @item -fdump-rtl-sibling
5060 @opindex fdump-rtl-sibling
5061 Dump after sibling call optimizations.
5063 @item -fdump-rtl-split1
5064 @itemx -fdump-rtl-split2
5065 @itemx -fdump-rtl-split3
5066 @itemx -fdump-rtl-split4
5067 @itemx -fdump-rtl-split5
5068 @opindex fdump-rtl-split1
5069 @opindex fdump-rtl-split2
5070 @opindex fdump-rtl-split3
5071 @opindex fdump-rtl-split4
5072 @opindex fdump-rtl-split5
5073 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5074 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5075 @option{-fdump-rtl-split5} enable dumping after five rounds of
5076 instruction splitting.
5078 @item -fdump-rtl-sms
5079 @opindex fdump-rtl-sms
5080 Dump after modulo scheduling. This pass is only run on some
5083 @item -fdump-rtl-stack
5084 @opindex fdump-rtl-stack
5085 Dump after conversion from GCC's "flat register file" registers to the
5086 x87's stack-like registers. This pass is only run on x86 variants.
5088 @item -fdump-rtl-subreg1
5089 @itemx -fdump-rtl-subreg2
5090 @opindex fdump-rtl-subreg1
5091 @opindex fdump-rtl-subreg2
5092 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5093 the two subreg expansion passes.
5095 @item -fdump-rtl-unshare
5096 @opindex fdump-rtl-unshare
5097 Dump after all rtl has been unshared.
5099 @item -fdump-rtl-vartrack
5100 @opindex fdump-rtl-vartrack
5101 Dump after variable tracking.
5103 @item -fdump-rtl-vregs
5104 @opindex fdump-rtl-vregs
5105 Dump after converting virtual registers to hard registers.
5107 @item -fdump-rtl-web
5108 @opindex fdump-rtl-web
5109 Dump after live range splitting.
5111 @item -fdump-rtl-regclass
5112 @itemx -fdump-rtl-subregs_of_mode_init
5113 @itemx -fdump-rtl-subregs_of_mode_finish
5114 @itemx -fdump-rtl-dfinit
5115 @itemx -fdump-rtl-dfinish
5116 @opindex fdump-rtl-regclass
5117 @opindex fdump-rtl-subregs_of_mode_init
5118 @opindex fdump-rtl-subregs_of_mode_finish
5119 @opindex fdump-rtl-dfinit
5120 @opindex fdump-rtl-dfinish
5121 These dumps are defined but always produce empty files.
5123 @item -fdump-rtl-all
5124 @opindex fdump-rtl-all
5125 Produce all the dumps listed above.
5129 Annotate the assembler output with miscellaneous debugging information.
5133 Dump all macro definitions, at the end of preprocessing, in addition to
5138 Produce a core dump whenever an error occurs.
5142 Print statistics on memory usage, at the end of the run, to
5147 Annotate the assembler output with a comment indicating which
5148 pattern and alternative was used. The length of each instruction is
5153 Dump the RTL in the assembler output as a comment before each instruction.
5154 Also turns on @option{-dp} annotation.
5158 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5159 dump a representation of the control flow graph suitable for viewing with VCG
5160 to @file{@var{file}.@var{pass}.vcg}.
5164 Just generate RTL for a function instead of compiling it. Usually used
5165 with @option{-fdump-rtl-expand}.
5169 Dump debugging information during parsing, to standard error.
5173 @opindex fdump-noaddr
5174 When doing debugging dumps, suppress address output. This makes it more
5175 feasible to use diff on debugging dumps for compiler invocations with
5176 different compiler binaries and/or different
5177 text / bss / data / heap / stack / dso start locations.
5179 @item -fdump-unnumbered
5180 @opindex fdump-unnumbered
5181 When doing debugging dumps, suppress instruction numbers and address output.
5182 This makes it more feasible to use diff on debugging dumps for compiler
5183 invocations with different options, in particular with and without
5186 @item -fdump-unnumbered-links
5187 @opindex fdump-unnumbered-links
5188 When doing debugging dumps (see @option{-d} option above), suppress
5189 instruction numbers for the links to the previous and next instructions
5192 @item -fdump-translation-unit @r{(C++ only)}
5193 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5194 @opindex fdump-translation-unit
5195 Dump a representation of the tree structure for the entire translation
5196 unit to a file. The file name is made by appending @file{.tu} to the
5197 source file name, and the file is created in the same directory as the
5198 output file. If the @samp{-@var{options}} form is used, @var{options}
5199 controls the details of the dump as described for the
5200 @option{-fdump-tree} options.
5202 @item -fdump-class-hierarchy @r{(C++ only)}
5203 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5204 @opindex fdump-class-hierarchy
5205 Dump a representation of each class's hierarchy and virtual function
5206 table layout to a file. The file name is made by appending
5207 @file{.class} to the source file name, and the file is created in the
5208 same directory as the output file. If the @samp{-@var{options}} form
5209 is used, @var{options} controls the details of the dump as described
5210 for the @option{-fdump-tree} options.
5212 @item -fdump-ipa-@var{switch}
5214 Control the dumping at various stages of inter-procedural analysis
5215 language tree to a file. The file name is generated by appending a
5216 switch specific suffix to the source file name, and the file is created
5217 in the same directory as the output file. The following dumps are
5222 Enables all inter-procedural analysis dumps.
5225 Dumps information about call-graph optimization, unused function removal,
5226 and inlining decisions.
5229 Dump after function inlining.
5233 @item -fdump-statistics-@var{option}
5234 @opindex fdump-statistics
5235 Enable and control dumping of pass statistics in a separate file. The
5236 file name is generated by appending a suffix ending in
5237 @samp{.statistics} to the source file name, and the file is created in
5238 the same directory as the output file. If the @samp{-@var{option}}
5239 form is used, @samp{-stats} will cause counters to be summed over the
5240 whole compilation unit while @samp{-details} will dump every event as
5241 the passes generate them. The default with no option is to sum
5242 counters for each function compiled.
5244 @item -fdump-tree-@var{switch}
5245 @itemx -fdump-tree-@var{switch}-@var{options}
5247 Control the dumping at various stages of processing the intermediate
5248 language tree to a file. The file name is generated by appending a
5249 switch specific suffix to the source file name, and the file is
5250 created in the same directory as the output file. If the
5251 @samp{-@var{options}} form is used, @var{options} is a list of
5252 @samp{-} separated options that control the details of the dump. Not
5253 all options are applicable to all dumps, those which are not
5254 meaningful will be ignored. The following options are available
5258 Print the address of each node. Usually this is not meaningful as it
5259 changes according to the environment and source file. Its primary use
5260 is for tying up a dump file with a debug environment.
5262 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5263 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5264 use working backward from mangled names in the assembly file.
5266 Inhibit dumping of members of a scope or body of a function merely
5267 because that scope has been reached. Only dump such items when they
5268 are directly reachable by some other path. When dumping pretty-printed
5269 trees, this option inhibits dumping the bodies of control structures.
5271 Print a raw representation of the tree. By default, trees are
5272 pretty-printed into a C-like representation.
5274 Enable more detailed dumps (not honored by every dump option).
5276 Enable dumping various statistics about the pass (not honored by every dump
5279 Enable showing basic block boundaries (disabled in raw dumps).
5281 Enable showing virtual operands for every statement.
5283 Enable showing line numbers for statements.
5285 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5287 Enable showing the tree dump for each statement.
5289 Enable showing the EH region number holding each statement.
5291 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5292 and @option{lineno}.
5295 The following tree dumps are possible:
5299 @opindex fdump-tree-original
5300 Dump before any tree based optimization, to @file{@var{file}.original}.
5303 @opindex fdump-tree-optimized
5304 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5307 @opindex fdump-tree-gimple
5308 Dump each function before and after the gimplification pass to a file. The
5309 file name is made by appending @file{.gimple} to the source file name.
5312 @opindex fdump-tree-cfg
5313 Dump the control flow graph of each function to a file. The file name is
5314 made by appending @file{.cfg} to the source file name.
5317 @opindex fdump-tree-vcg
5318 Dump the control flow graph of each function to a file in VCG format. The
5319 file name is made by appending @file{.vcg} to the source file name. Note
5320 that if the file contains more than one function, the generated file cannot
5321 be used directly by VCG@. You will need to cut and paste each function's
5322 graph into its own separate file first.
5325 @opindex fdump-tree-ch
5326 Dump each function after copying loop headers. The file name is made by
5327 appending @file{.ch} to the source file name.
5330 @opindex fdump-tree-ssa
5331 Dump SSA related information to a file. The file name is made by appending
5332 @file{.ssa} to the source file name.
5335 @opindex fdump-tree-alias
5336 Dump aliasing information for each function. The file name is made by
5337 appending @file{.alias} to the source file name.
5340 @opindex fdump-tree-ccp
5341 Dump each function after CCP@. The file name is made by appending
5342 @file{.ccp} to the source file name.
5345 @opindex fdump-tree-storeccp
5346 Dump each function after STORE-CCP@. The file name is made by appending
5347 @file{.storeccp} to the source file name.
5350 @opindex fdump-tree-pre
5351 Dump trees after partial redundancy elimination. The file name is made
5352 by appending @file{.pre} to the source file name.
5355 @opindex fdump-tree-fre
5356 Dump trees after full redundancy elimination. The file name is made
5357 by appending @file{.fre} to the source file name.
5360 @opindex fdump-tree-copyprop
5361 Dump trees after copy propagation. The file name is made
5362 by appending @file{.copyprop} to the source file name.
5364 @item store_copyprop
5365 @opindex fdump-tree-store_copyprop
5366 Dump trees after store copy-propagation. The file name is made
5367 by appending @file{.store_copyprop} to the source file name.
5370 @opindex fdump-tree-dce
5371 Dump each function after dead code elimination. The file name is made by
5372 appending @file{.dce} to the source file name.
5375 @opindex fdump-tree-mudflap
5376 Dump each function after adding mudflap instrumentation. The file name is
5377 made by appending @file{.mudflap} to the source file name.
5380 @opindex fdump-tree-sra
5381 Dump each function after performing scalar replacement of aggregates. The
5382 file name is made by appending @file{.sra} to the source file name.
5385 @opindex fdump-tree-sink
5386 Dump each function after performing code sinking. The file name is made
5387 by appending @file{.sink} to the source file name.
5390 @opindex fdump-tree-dom
5391 Dump each function after applying dominator tree optimizations. The file
5392 name is made by appending @file{.dom} to the source file name.
5395 @opindex fdump-tree-dse
5396 Dump each function after applying dead store elimination. The file
5397 name is made by appending @file{.dse} to the source file name.
5400 @opindex fdump-tree-phiopt
5401 Dump each function after optimizing PHI nodes into straightline code. The file
5402 name is made by appending @file{.phiopt} to the source file name.
5405 @opindex fdump-tree-forwprop
5406 Dump each function after forward propagating single use variables. The file
5407 name is made by appending @file{.forwprop} to the source file name.
5410 @opindex fdump-tree-copyrename
5411 Dump each function after applying the copy rename optimization. The file
5412 name is made by appending @file{.copyrename} to the source file name.
5415 @opindex fdump-tree-nrv
5416 Dump each function after applying the named return value optimization on
5417 generic trees. The file name is made by appending @file{.nrv} to the source
5421 @opindex fdump-tree-vect
5422 Dump each function after applying vectorization of loops. The file name is
5423 made by appending @file{.vect} to the source file name.
5426 @opindex fdump-tree-slp
5427 Dump each function after applying vectorization of basic blocks. The file name
5428 is made by appending @file{.slp} to the source file name.
5431 @opindex fdump-tree-vrp
5432 Dump each function after Value Range Propagation (VRP). The file name
5433 is made by appending @file{.vrp} to the source file name.
5436 @opindex fdump-tree-all
5437 Enable all the available tree dumps with the flags provided in this option.
5440 @item -ftree-vectorizer-verbose=@var{n}
5441 @opindex ftree-vectorizer-verbose
5442 This option controls the amount of debugging output the vectorizer prints.
5443 This information is written to standard error, unless
5444 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5445 in which case it is output to the usual dump listing file, @file{.vect}.
5446 For @var{n}=0 no diagnostic information is reported.
5447 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5448 and the total number of loops that got vectorized.
5449 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5450 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5451 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5452 level that @option{-fdump-tree-vect-stats} uses.
5453 Higher verbosity levels mean either more information dumped for each
5454 reported loop, or same amount of information reported for more loops:
5455 if @var{n}=3, vectorizer cost model information is reported.
5456 If @var{n}=4, alignment related information is added to the reports.
5457 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5458 memory access-patterns) is added to the reports.
5459 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5460 that did not pass the first analysis phase (i.e., may not be countable, or
5461 may have complicated control-flow).
5462 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5463 If @var{n}=8, SLP related information is added to the reports.
5464 For @var{n}=9, all the information the vectorizer generates during its
5465 analysis and transformation is reported. This is the same verbosity level
5466 that @option{-fdump-tree-vect-details} uses.
5468 @item -frandom-seed=@var{string}
5469 @opindex frandom-seed
5470 This option provides a seed that GCC uses when it would otherwise use
5471 random numbers. It is used to generate certain symbol names
5472 that have to be different in every compiled file. It is also used to
5473 place unique stamps in coverage data files and the object files that
5474 produce them. You can use the @option{-frandom-seed} option to produce
5475 reproducibly identical object files.
5477 The @var{string} should be different for every file you compile.
5479 @item -fsched-verbose=@var{n}
5480 @opindex fsched-verbose
5481 On targets that use instruction scheduling, this option controls the
5482 amount of debugging output the scheduler prints. This information is
5483 written to standard error, unless @option{-fdump-rtl-sched1} or
5484 @option{-fdump-rtl-sched2} is specified, in which case it is output
5485 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5486 respectively. However for @var{n} greater than nine, the output is
5487 always printed to standard error.
5489 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5490 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5491 For @var{n} greater than one, it also output basic block probabilities,
5492 detailed ready list information and unit/insn info. For @var{n} greater
5493 than two, it includes RTL at abort point, control-flow and regions info.
5494 And for @var{n} over four, @option{-fsched-verbose} also includes
5498 @itemx -save-temps=cwd
5500 Store the usual ``temporary'' intermediate files permanently; place them
5501 in the current directory and name them based on the source file. Thus,
5502 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5503 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5504 preprocessed @file{foo.i} output file even though the compiler now
5505 normally uses an integrated preprocessor.
5507 When used in combination with the @option{-x} command line option,
5508 @option{-save-temps} is sensible enough to avoid over writing an
5509 input source file with the same extension as an intermediate file.
5510 The corresponding intermediate file may be obtained by renaming the
5511 source file before using @option{-save-temps}.
5513 If you invoke GCC in parallel, compiling several different source
5514 files that share a common base name in different subdirectories or the
5515 same source file compiled for multiple output destinations, it is
5516 likely that the different parallel compilers will interfere with each
5517 other, and overwrite the temporary files. For instance:
5520 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5521 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5524 may result in @file{foo.i} and @file{foo.o} being written to
5525 simultaneously by both compilers.
5527 @item -save-temps=obj
5528 @opindex save-temps=obj
5529 Store the usual ``temporary'' intermediate files permanently. If the
5530 @option{-o} option is used, the temporary files are based on the
5531 object file. If the @option{-o} option is not used, the
5532 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5537 gcc -save-temps=obj -c foo.c
5538 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5539 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5542 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5543 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5544 @file{dir2/yfoobar.o}.
5546 @item -time@r{[}=@var{file}@r{]}
5548 Report the CPU time taken by each subprocess in the compilation
5549 sequence. For C source files, this is the compiler proper and assembler
5550 (plus the linker if linking is done).
5552 Without the specification of an output file, the output looks like this:
5559 The first number on each line is the ``user time'', that is time spent
5560 executing the program itself. The second number is ``system time'',
5561 time spent executing operating system routines on behalf of the program.
5562 Both numbers are in seconds.
5564 With the specification of an output file, the output is appended to the
5565 named file, and it looks like this:
5568 0.12 0.01 cc1 @var{options}
5569 0.00 0.01 as @var{options}
5572 The ``user time'' and the ``system time'' are moved before the program
5573 name, and the options passed to the program are displayed, so that one
5574 can later tell what file was being compiled, and with which options.
5576 @item -fvar-tracking
5577 @opindex fvar-tracking
5578 Run variable tracking pass. It computes where variables are stored at each
5579 position in code. Better debugging information is then generated
5580 (if the debugging information format supports this information).
5582 It is enabled by default when compiling with optimization (@option{-Os},
5583 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5584 the debug info format supports it.
5586 @item -fvar-tracking-assignments
5587 @opindex fvar-tracking-assignments
5588 @opindex fno-var-tracking-assignments
5589 Annotate assignments to user variables early in the compilation and
5590 attempt to carry the annotations over throughout the compilation all the
5591 way to the end, in an attempt to improve debug information while
5592 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5594 It can be enabled even if var-tracking is disabled, in which case
5595 annotations will be created and maintained, but discarded at the end.
5597 @item -fvar-tracking-assignments-toggle
5598 @opindex fvar-tracking-assignments-toggle
5599 @opindex fno-var-tracking-assignments-toggle
5600 Toggle @option{-fvar-tracking-assignments}, in the same way that
5601 @option{-gtoggle} toggles @option{-g}.
5603 @item -print-file-name=@var{library}
5604 @opindex print-file-name
5605 Print the full absolute name of the library file @var{library} that
5606 would be used when linking---and don't do anything else. With this
5607 option, GCC does not compile or link anything; it just prints the
5610 @item -print-multi-directory
5611 @opindex print-multi-directory
5612 Print the directory name corresponding to the multilib selected by any
5613 other switches present in the command line. This directory is supposed
5614 to exist in @env{GCC_EXEC_PREFIX}.
5616 @item -print-multi-lib
5617 @opindex print-multi-lib
5618 Print the mapping from multilib directory names to compiler switches
5619 that enable them. The directory name is separated from the switches by
5620 @samp{;}, and each switch starts with an @samp{@@} instead of the
5621 @samp{-}, without spaces between multiple switches. This is supposed to
5622 ease shell-processing.
5624 @item -print-multi-os-directory
5625 @opindex print-multi-os-directory
5626 Print the path to OS libraries for the selected
5627 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5628 present in the @file{lib} subdirectory and no multilibs are used, this is
5629 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5630 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5631 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5632 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5634 @item -print-prog-name=@var{program}
5635 @opindex print-prog-name
5636 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5638 @item -print-libgcc-file-name
5639 @opindex print-libgcc-file-name
5640 Same as @option{-print-file-name=libgcc.a}.
5642 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5643 but you do want to link with @file{libgcc.a}. You can do
5646 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5649 @item -print-search-dirs
5650 @opindex print-search-dirs
5651 Print the name of the configured installation directory and a list of
5652 program and library directories @command{gcc} will search---and don't do anything else.
5654 This is useful when @command{gcc} prints the error message
5655 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5656 To resolve this you either need to put @file{cpp0} and the other compiler
5657 components where @command{gcc} expects to find them, or you can set the environment
5658 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5659 Don't forget the trailing @samp{/}.
5660 @xref{Environment Variables}.
5662 @item -print-sysroot
5663 @opindex print-sysroot
5664 Print the target sysroot directory that will be used during
5665 compilation. This is the target sysroot specified either at configure
5666 time or using the @option{--sysroot} option, possibly with an extra
5667 suffix that depends on compilation options. If no target sysroot is
5668 specified, the option prints nothing.
5670 @item -print-sysroot-headers-suffix
5671 @opindex print-sysroot-headers-suffix
5672 Print the suffix added to the target sysroot when searching for
5673 headers, or give an error if the compiler is not configured with such
5674 a suffix---and don't do anything else.
5677 @opindex dumpmachine
5678 Print the compiler's target machine (for example,
5679 @samp{i686-pc-linux-gnu})---and don't do anything else.
5682 @opindex dumpversion
5683 Print the compiler version (for example, @samp{3.0})---and don't do
5688 Print the compiler's built-in specs---and don't do anything else. (This
5689 is used when GCC itself is being built.) @xref{Spec Files}.
5691 @item -feliminate-unused-debug-types
5692 @opindex feliminate-unused-debug-types
5693 Normally, when producing DWARF2 output, GCC will emit debugging
5694 information for all types declared in a compilation
5695 unit, regardless of whether or not they are actually used
5696 in that compilation unit. Sometimes this is useful, such as
5697 if, in the debugger, you want to cast a value to a type that is
5698 not actually used in your program (but is declared). More often,
5699 however, this results in a significant amount of wasted space.
5700 With this option, GCC will avoid producing debug symbol output
5701 for types that are nowhere used in the source file being compiled.
5704 @node Optimize Options
5705 @section Options That Control Optimization
5706 @cindex optimize options
5707 @cindex options, optimization
5709 These options control various sorts of optimizations.
5711 Without any optimization option, the compiler's goal is to reduce the
5712 cost of compilation and to make debugging produce the expected
5713 results. Statements are independent: if you stop the program with a
5714 breakpoint between statements, you can then assign a new value to any
5715 variable or change the program counter to any other statement in the
5716 function and get exactly the results you would expect from the source
5719 Turning on optimization flags makes the compiler attempt to improve
5720 the performance and/or code size at the expense of compilation time
5721 and possibly the ability to debug the program.
5723 The compiler performs optimization based on the knowledge it has of the
5724 program. Compiling multiple files at once to a single output file mode allows
5725 the compiler to use information gained from all of the files when compiling
5728 Not all optimizations are controlled directly by a flag. Only
5729 optimizations that have a flag are listed in this section.
5731 Most optimizations are only enabled if an @option{-O} level is set on
5732 the command line. Otherwise they are disabled, even if individual
5733 optimization flags are specified.
5735 Depending on the target and how GCC was configured, a slightly different
5736 set of optimizations may be enabled at each @option{-O} level than
5737 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5738 to find out the exact set of optimizations that are enabled at each level.
5739 @xref{Overall Options}, for examples.
5746 Optimize. Optimizing compilation takes somewhat more time, and a lot
5747 more memory for a large function.
5749 With @option{-O}, the compiler tries to reduce code size and execution
5750 time, without performing any optimizations that take a great deal of
5753 @option{-O} turns on the following optimization flags:
5756 -fcprop-registers @gol
5759 -fdelayed-branch @gol
5761 -fguess-branch-probability @gol
5762 -fif-conversion2 @gol
5763 -fif-conversion @gol
5764 -fipa-pure-const @gol
5766 -fipa-reference @gol
5768 -fsplit-wide-types @gol
5769 -ftree-builtin-call-dce @gol
5772 -ftree-copyrename @gol
5774 -ftree-dominator-opts @gol
5776 -ftree-forwprop @gol
5784 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5785 where doing so does not interfere with debugging.
5789 Optimize even more. GCC performs nearly all supported optimizations
5790 that do not involve a space-speed tradeoff.
5791 As compared to @option{-O}, this option increases both compilation time
5792 and the performance of the generated code.
5794 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5795 also turns on the following optimization flags:
5796 @gccoptlist{-fthread-jumps @gol
5797 -falign-functions -falign-jumps @gol
5798 -falign-loops -falign-labels @gol
5801 -fcse-follow-jumps -fcse-skip-blocks @gol
5802 -fdelete-null-pointer-checks @gol
5803 -fexpensive-optimizations @gol
5804 -fgcse -fgcse-lm @gol
5805 -finline-small-functions @gol
5806 -findirect-inlining @gol
5808 -foptimize-sibling-calls @gol
5811 -freorder-blocks -freorder-functions @gol
5812 -frerun-cse-after-loop @gol
5813 -fsched-interblock -fsched-spec @gol
5814 -fschedule-insns -fschedule-insns2 @gol
5815 -fstrict-aliasing -fstrict-overflow @gol
5816 -ftree-switch-conversion @gol
5820 Please note the warning under @option{-fgcse} about
5821 invoking @option{-O2} on programs that use computed gotos.
5825 Optimize yet more. @option{-O3} turns on all optimizations specified
5826 by @option{-O2} and also turns on the @option{-finline-functions},
5827 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5828 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5832 Reduce compilation time and make debugging produce the expected
5833 results. This is the default.
5837 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5838 do not typically increase code size. It also performs further
5839 optimizations designed to reduce code size.
5841 @option{-Os} disables the following optimization flags:
5842 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5843 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5844 -fprefetch-loop-arrays -ftree-vect-loop-version}
5846 If you use multiple @option{-O} options, with or without level numbers,
5847 the last such option is the one that is effective.
5850 Options of the form @option{-f@var{flag}} specify machine-independent
5851 flags. Most flags have both positive and negative forms; the negative
5852 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5853 below, only one of the forms is listed---the one you typically will
5854 use. You can figure out the other form by either removing @samp{no-}
5857 The following options control specific optimizations. They are either
5858 activated by @option{-O} options or are related to ones that are. You
5859 can use the following flags in the rare cases when ``fine-tuning'' of
5860 optimizations to be performed is desired.
5863 @item -fno-default-inline
5864 @opindex fno-default-inline
5865 Do not make member functions inline by default merely because they are
5866 defined inside the class scope (C++ only). Otherwise, when you specify
5867 @w{@option{-O}}, member functions defined inside class scope are compiled
5868 inline by default; i.e., you don't need to add @samp{inline} in front of
5869 the member function name.
5871 @item -fno-defer-pop
5872 @opindex fno-defer-pop
5873 Always pop the arguments to each function call as soon as that function
5874 returns. For machines which must pop arguments after a function call,
5875 the compiler normally lets arguments accumulate on the stack for several
5876 function calls and pops them all at once.
5878 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5880 @item -fforward-propagate
5881 @opindex fforward-propagate
5882 Perform a forward propagation pass on RTL@. The pass tries to combine two
5883 instructions and checks if the result can be simplified. If loop unrolling
5884 is active, two passes are performed and the second is scheduled after
5887 This option is enabled by default at optimization levels @option{-O},
5888 @option{-O2}, @option{-O3}, @option{-Os}.
5890 @item -fomit-frame-pointer
5891 @opindex fomit-frame-pointer
5892 Don't keep the frame pointer in a register for functions that
5893 don't need one. This avoids the instructions to save, set up and
5894 restore frame pointers; it also makes an extra register available
5895 in many functions. @strong{It also makes debugging impossible on
5898 On some machines, such as the VAX, this flag has no effect, because
5899 the standard calling sequence automatically handles the frame pointer
5900 and nothing is saved by pretending it doesn't exist. The
5901 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5902 whether a target machine supports this flag. @xref{Registers,,Register
5903 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5905 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5907 @item -foptimize-sibling-calls
5908 @opindex foptimize-sibling-calls
5909 Optimize sibling and tail recursive calls.
5911 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5915 Don't pay attention to the @code{inline} keyword. Normally this option
5916 is used to keep the compiler from expanding any functions inline.
5917 Note that if you are not optimizing, no functions can be expanded inline.
5919 @item -finline-small-functions
5920 @opindex finline-small-functions
5921 Integrate functions into their callers when their body is smaller than expected
5922 function call code (so overall size of program gets smaller). The compiler
5923 heuristically decides which functions are simple enough to be worth integrating
5926 Enabled at level @option{-O2}.
5928 @item -findirect-inlining
5929 @opindex findirect-inlining
5930 Inline also indirect calls that are discovered to be known at compile
5931 time thanks to previous inlining. This option has any effect only
5932 when inlining itself is turned on by the @option{-finline-functions}
5933 or @option{-finline-small-functions} options.
5935 Enabled at level @option{-O2}.
5937 @item -finline-functions
5938 @opindex finline-functions
5939 Integrate all simple functions into their callers. The compiler
5940 heuristically decides which functions are simple enough to be worth
5941 integrating in this way.
5943 If all calls to a given function are integrated, and the function is
5944 declared @code{static}, then the function is normally not output as
5945 assembler code in its own right.
5947 Enabled at level @option{-O3}.
5949 @item -finline-functions-called-once
5950 @opindex finline-functions-called-once
5951 Consider all @code{static} functions called once for inlining into their
5952 caller even if they are not marked @code{inline}. If a call to a given
5953 function is integrated, then the function is not output as assembler code
5956 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5958 @item -fearly-inlining
5959 @opindex fearly-inlining
5960 Inline functions marked by @code{always_inline} and functions whose body seems
5961 smaller than the function call overhead early before doing
5962 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5963 makes profiling significantly cheaper and usually inlining faster on programs
5964 having large chains of nested wrapper functions.
5970 Perform interprocedural scalar replacement of aggregates, removal of
5971 unused parameters and replacement of parameters passed by reference
5972 by parameters passed by value.
5974 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
5976 @item -finline-limit=@var{n}
5977 @opindex finline-limit
5978 By default, GCC limits the size of functions that can be inlined. This flag
5979 allows coarse control of this limit. @var{n} is the size of functions that
5980 can be inlined in number of pseudo instructions.
5982 Inlining is actually controlled by a number of parameters, which may be
5983 specified individually by using @option{--param @var{name}=@var{value}}.
5984 The @option{-finline-limit=@var{n}} option sets some of these parameters
5988 @item max-inline-insns-single
5989 is set to @var{n}/2.
5990 @item max-inline-insns-auto
5991 is set to @var{n}/2.
5994 See below for a documentation of the individual
5995 parameters controlling inlining and for the defaults of these parameters.
5997 @emph{Note:} there may be no value to @option{-finline-limit} that results
5998 in default behavior.
6000 @emph{Note:} pseudo instruction represents, in this particular context, an
6001 abstract measurement of function's size. In no way does it represent a count
6002 of assembly instructions and as such its exact meaning might change from one
6003 release to an another.
6005 @item -fkeep-inline-functions
6006 @opindex fkeep-inline-functions
6007 In C, emit @code{static} functions that are declared @code{inline}
6008 into the object file, even if the function has been inlined into all
6009 of its callers. This switch does not affect functions using the
6010 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6011 inline functions into the object file.
6013 @item -fkeep-static-consts
6014 @opindex fkeep-static-consts
6015 Emit variables declared @code{static const} when optimization isn't turned
6016 on, even if the variables aren't referenced.
6018 GCC enables this option by default. If you want to force the compiler to
6019 check if the variable was referenced, regardless of whether or not
6020 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6022 @item -fmerge-constants
6023 @opindex fmerge-constants
6024 Attempt to merge identical constants (string constants and floating point
6025 constants) across compilation units.
6027 This option is the default for optimized compilation if the assembler and
6028 linker support it. Use @option{-fno-merge-constants} to inhibit this
6031 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6033 @item -fmerge-all-constants
6034 @opindex fmerge-all-constants
6035 Attempt to merge identical constants and identical variables.
6037 This option implies @option{-fmerge-constants}. In addition to
6038 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6039 arrays or initialized constant variables with integral or floating point
6040 types. Languages like C or C++ require each variable, including multiple
6041 instances of the same variable in recursive calls, to have distinct locations,
6042 so using this option will result in non-conforming
6045 @item -fmodulo-sched
6046 @opindex fmodulo-sched
6047 Perform swing modulo scheduling immediately before the first scheduling
6048 pass. This pass looks at innermost loops and reorders their
6049 instructions by overlapping different iterations.
6051 @item -fmodulo-sched-allow-regmoves
6052 @opindex fmodulo-sched-allow-regmoves
6053 Perform more aggressive SMS based modulo scheduling with register moves
6054 allowed. By setting this flag certain anti-dependences edges will be
6055 deleted which will trigger the generation of reg-moves based on the
6056 life-range analysis. This option is effective only with
6057 @option{-fmodulo-sched} enabled.
6059 @item -fno-branch-count-reg
6060 @opindex fno-branch-count-reg
6061 Do not use ``decrement and branch'' instructions on a count register,
6062 but instead generate a sequence of instructions that decrement a
6063 register, compare it against zero, then branch based upon the result.
6064 This option is only meaningful on architectures that support such
6065 instructions, which include x86, PowerPC, IA-64 and S/390.
6067 The default is @option{-fbranch-count-reg}.
6069 @item -fno-function-cse
6070 @opindex fno-function-cse
6071 Do not put function addresses in registers; make each instruction that
6072 calls a constant function contain the function's address explicitly.
6074 This option results in less efficient code, but some strange hacks
6075 that alter the assembler output may be confused by the optimizations
6076 performed when this option is not used.
6078 The default is @option{-ffunction-cse}
6080 @item -fno-zero-initialized-in-bss
6081 @opindex fno-zero-initialized-in-bss
6082 If the target supports a BSS section, GCC by default puts variables that
6083 are initialized to zero into BSS@. This can save space in the resulting
6086 This option turns off this behavior because some programs explicitly
6087 rely on variables going to the data section. E.g., so that the
6088 resulting executable can find the beginning of that section and/or make
6089 assumptions based on that.
6091 The default is @option{-fzero-initialized-in-bss}.
6093 @item -fmudflap -fmudflapth -fmudflapir
6097 @cindex bounds checking
6099 For front-ends that support it (C and C++), instrument all risky
6100 pointer/array dereferencing operations, some standard library
6101 string/heap functions, and some other associated constructs with
6102 range/validity tests. Modules so instrumented should be immune to
6103 buffer overflows, invalid heap use, and some other classes of C/C++
6104 programming errors. The instrumentation relies on a separate runtime
6105 library (@file{libmudflap}), which will be linked into a program if
6106 @option{-fmudflap} is given at link time. Run-time behavior of the
6107 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6108 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6111 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6112 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6113 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6114 instrumentation should ignore pointer reads. This produces less
6115 instrumentation (and therefore faster execution) and still provides
6116 some protection against outright memory corrupting writes, but allows
6117 erroneously read data to propagate within a program.
6119 @item -fthread-jumps
6120 @opindex fthread-jumps
6121 Perform optimizations where we check to see if a jump branches to a
6122 location where another comparison subsumed by the first is found. If
6123 so, the first branch is redirected to either the destination of the
6124 second branch or a point immediately following it, depending on whether
6125 the condition is known to be true or false.
6127 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6129 @item -fsplit-wide-types
6130 @opindex fsplit-wide-types
6131 When using a type that occupies multiple registers, such as @code{long
6132 long} on a 32-bit system, split the registers apart and allocate them
6133 independently. This normally generates better code for those types,
6134 but may make debugging more difficult.
6136 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6139 @item -fcse-follow-jumps
6140 @opindex fcse-follow-jumps
6141 In common subexpression elimination (CSE), scan through jump instructions
6142 when the target of the jump is not reached by any other path. For
6143 example, when CSE encounters an @code{if} statement with an
6144 @code{else} clause, CSE will follow the jump when the condition
6147 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6149 @item -fcse-skip-blocks
6150 @opindex fcse-skip-blocks
6151 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6152 follow jumps which conditionally skip over blocks. When CSE
6153 encounters a simple @code{if} statement with no else clause,
6154 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6155 body of the @code{if}.
6157 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6159 @item -frerun-cse-after-loop
6160 @opindex frerun-cse-after-loop
6161 Re-run common subexpression elimination after loop optimizations has been
6164 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6168 Perform a global common subexpression elimination pass.
6169 This pass also performs global constant and copy propagation.
6171 @emph{Note:} When compiling a program using computed gotos, a GCC
6172 extension, you may get better runtime performance if you disable
6173 the global common subexpression elimination pass by adding
6174 @option{-fno-gcse} to the command line.
6176 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6180 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6181 attempt to move loads which are only killed by stores into themselves. This
6182 allows a loop containing a load/store sequence to be changed to a load outside
6183 the loop, and a copy/store within the loop.
6185 Enabled by default when gcse is enabled.
6189 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6190 global common subexpression elimination. This pass will attempt to move
6191 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6192 loops containing a load/store sequence can be changed to a load before
6193 the loop and a store after the loop.
6195 Not enabled at any optimization level.
6199 When @option{-fgcse-las} is enabled, the global common subexpression
6200 elimination pass eliminates redundant loads that come after stores to the
6201 same memory location (both partial and full redundancies).
6203 Not enabled at any optimization level.
6205 @item -fgcse-after-reload
6206 @opindex fgcse-after-reload
6207 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6208 pass is performed after reload. The purpose of this pass is to cleanup
6211 @item -funsafe-loop-optimizations
6212 @opindex funsafe-loop-optimizations
6213 If given, the loop optimizer will assume that loop indices do not
6214 overflow, and that the loops with nontrivial exit condition are not
6215 infinite. This enables a wider range of loop optimizations even if
6216 the loop optimizer itself cannot prove that these assumptions are valid.
6217 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6218 if it finds this kind of loop.
6220 @item -fcrossjumping
6221 @opindex fcrossjumping
6222 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6223 resulting code may or may not perform better than without cross-jumping.
6225 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6227 @item -fauto-inc-dec
6228 @opindex fauto-inc-dec
6229 Combine increments or decrements of addresses with memory accesses.
6230 This pass is always skipped on architectures that do not have
6231 instructions to support this. Enabled by default at @option{-O} and
6232 higher on architectures that support this.
6236 Perform dead code elimination (DCE) on RTL@.
6237 Enabled by default at @option{-O} and higher.
6241 Perform dead store elimination (DSE) on RTL@.
6242 Enabled by default at @option{-O} and higher.
6244 @item -fif-conversion
6245 @opindex fif-conversion
6246 Attempt to transform conditional jumps into branch-less equivalents. This
6247 include use of conditional moves, min, max, set flags and abs instructions, and
6248 some tricks doable by standard arithmetics. The use of conditional execution
6249 on chips where it is available is controlled by @code{if-conversion2}.
6251 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6253 @item -fif-conversion2
6254 @opindex fif-conversion2
6255 Use conditional execution (where available) to transform conditional jumps into
6256 branch-less equivalents.
6258 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6260 @item -fdelete-null-pointer-checks
6261 @opindex fdelete-null-pointer-checks
6262 Assume that programs cannot safely dereference null pointers, and that
6263 no code or data element resides there. This enables simple constant
6264 folding optimizations at all optimization levels. In addition, other
6265 optimization passes in GCC use this flag to control global dataflow
6266 analyses that eliminate useless checks for null pointers; these assume
6267 that if a pointer is checked after it has already been dereferenced,
6270 Note however that in some environments this assumption is not true.
6271 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6272 for programs which depend on that behavior.
6274 Some targets, especially embedded ones, disable this option at all levels.
6275 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6276 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6277 are enabled independently at different optimization levels.
6279 @item -fexpensive-optimizations
6280 @opindex fexpensive-optimizations
6281 Perform a number of minor optimizations that are relatively expensive.
6283 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6285 @item -foptimize-register-move
6287 @opindex foptimize-register-move
6289 Attempt to reassign register numbers in move instructions and as
6290 operands of other simple instructions in order to maximize the amount of
6291 register tying. This is especially helpful on machines with two-operand
6294 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6297 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6299 @item -fira-algorithm=@var{algorithm}
6300 Use specified coloring algorithm for the integrated register
6301 allocator. The @var{algorithm} argument should be @code{priority} or
6302 @code{CB}. The first algorithm specifies Chow's priority coloring,
6303 the second one specifies Chaitin-Briggs coloring. The second
6304 algorithm can be unimplemented for some architectures. If it is
6305 implemented, it is the default because Chaitin-Briggs coloring as a
6306 rule generates a better code.
6308 @item -fira-region=@var{region}
6309 Use specified regions for the integrated register allocator. The
6310 @var{region} argument should be one of @code{all}, @code{mixed}, or
6311 @code{one}. The first value means using all loops as register
6312 allocation regions, the second value which is the default means using
6313 all loops except for loops with small register pressure as the
6314 regions, and third one means using all function as a single region.
6315 The first value can give best result for machines with small size and
6316 irregular register set, the third one results in faster and generates
6317 decent code and the smallest size code, and the default value usually
6318 give the best results in most cases and for most architectures.
6320 @item -fira-coalesce
6321 @opindex fira-coalesce
6322 Do optimistic register coalescing. This option might be profitable for
6323 architectures with big regular register files.
6325 @item -fira-loop-pressure
6326 @opindex fira-loop-pressure
6327 Use IRA to evaluate register pressure in loops for decision to move
6328 loop invariants. Usage of this option usually results in generation
6329 of faster and smaller code on machines with big register files (>= 32
6330 registers) but it can slow compiler down.
6332 This option is enabled at level @option{-O3} for some targets.
6334 @item -fno-ira-share-save-slots
6335 @opindex fno-ira-share-save-slots
6336 Switch off sharing stack slots used for saving call used hard
6337 registers living through a call. Each hard register will get a
6338 separate stack slot and as a result function stack frame will be
6341 @item -fno-ira-share-spill-slots
6342 @opindex fno-ira-share-spill-slots
6343 Switch off sharing stack slots allocated for pseudo-registers. Each
6344 pseudo-register which did not get a hard register will get a separate
6345 stack slot and as a result function stack frame will be bigger.
6347 @item -fira-verbose=@var{n}
6348 @opindex fira-verbose
6349 Set up how verbose dump file for the integrated register allocator
6350 will be. Default value is 5. If the value is greater or equal to 10,
6351 the dump file will be stderr as if the value were @var{n} minus 10.
6353 @item -fdelayed-branch
6354 @opindex fdelayed-branch
6355 If supported for the target machine, attempt to reorder instructions
6356 to exploit instruction slots available after delayed branch
6359 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6361 @item -fschedule-insns
6362 @opindex fschedule-insns
6363 If supported for the target machine, attempt to reorder instructions to
6364 eliminate execution stalls due to required data being unavailable. This
6365 helps machines that have slow floating point or memory load instructions
6366 by allowing other instructions to be issued until the result of the load
6367 or floating point instruction is required.
6369 Enabled at levels @option{-O2}, @option{-O3}.
6371 @item -fschedule-insns2
6372 @opindex fschedule-insns2
6373 Similar to @option{-fschedule-insns}, but requests an additional pass of
6374 instruction scheduling after register allocation has been done. This is
6375 especially useful on machines with a relatively small number of
6376 registers and where memory load instructions take more than one cycle.
6378 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6380 @item -fno-sched-interblock
6381 @opindex fno-sched-interblock
6382 Don't schedule instructions across basic blocks. This is normally
6383 enabled by default when scheduling before register allocation, i.e.@:
6384 with @option{-fschedule-insns} or at @option{-O2} or higher.
6386 @item -fno-sched-spec
6387 @opindex fno-sched-spec
6388 Don't allow speculative motion of non-load instructions. This is normally
6389 enabled by default when scheduling before register allocation, i.e.@:
6390 with @option{-fschedule-insns} or at @option{-O2} or higher.
6392 @item -fsched-pressure
6393 @opindex fsched-pressure
6394 Enable register pressure sensitive insn scheduling before the register
6395 allocation. This only makes sense when scheduling before register
6396 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6397 @option{-O2} or higher. Usage of this option can improve the
6398 generated code and decrease its size by preventing register pressure
6399 increase above the number of available hard registers and as a
6400 consequence register spills in the register allocation.
6402 @item -fsched-spec-load
6403 @opindex fsched-spec-load
6404 Allow speculative motion of some load instructions. This only makes
6405 sense when scheduling before register allocation, i.e.@: with
6406 @option{-fschedule-insns} or at @option{-O2} or higher.
6408 @item -fsched-spec-load-dangerous
6409 @opindex fsched-spec-load-dangerous
6410 Allow speculative motion of more load instructions. This only makes
6411 sense when scheduling before register allocation, i.e.@: with
6412 @option{-fschedule-insns} or at @option{-O2} or higher.
6414 @item -fsched-stalled-insns
6415 @itemx -fsched-stalled-insns=@var{n}
6416 @opindex fsched-stalled-insns
6417 Define how many insns (if any) can be moved prematurely from the queue
6418 of stalled insns into the ready list, during the second scheduling pass.
6419 @option{-fno-sched-stalled-insns} means that no insns will be moved
6420 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6421 on how many queued insns can be moved prematurely.
6422 @option{-fsched-stalled-insns} without a value is equivalent to
6423 @option{-fsched-stalled-insns=1}.
6425 @item -fsched-stalled-insns-dep
6426 @itemx -fsched-stalled-insns-dep=@var{n}
6427 @opindex fsched-stalled-insns-dep
6428 Define how many insn groups (cycles) will be examined for a dependency
6429 on a stalled insn that is candidate for premature removal from the queue
6430 of stalled insns. This has an effect only during the second scheduling pass,
6431 and only if @option{-fsched-stalled-insns} is used.
6432 @option{-fno-sched-stalled-insns-dep} is equivalent to
6433 @option{-fsched-stalled-insns-dep=0}.
6434 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6435 @option{-fsched-stalled-insns-dep=1}.
6437 @item -fsched2-use-superblocks
6438 @opindex fsched2-use-superblocks
6439 When scheduling after register allocation, do use superblock scheduling
6440 algorithm. Superblock scheduling allows motion across basic block boundaries
6441 resulting on faster schedules. This option is experimental, as not all machine
6442 descriptions used by GCC model the CPU closely enough to avoid unreliable
6443 results from the algorithm.
6445 This only makes sense when scheduling after register allocation, i.e.@: with
6446 @option{-fschedule-insns2} or at @option{-O2} or higher.
6448 @item -fsched-group-heuristic
6449 @opindex fsched-group-heuristic
6450 Enable the group heuristic in the scheduler. This heuristic favors
6451 the instruction that belongs to a schedule group. This is enabled
6452 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6453 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6455 @item -fsched-critical-path-heuristic
6456 @opindex fsched-critical-path-heuristic
6457 Enable the critical-path heuristic in the scheduler. This heuristic favors
6458 instructions on the critical path. This is enabled by default when
6459 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6460 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6462 @item -fsched-spec-insn-heuristic
6463 @opindex fsched-spec-insn-heuristic
6464 Enable the speculative instruction heuristic in the scheduler. This
6465 heuristic favors speculative instructions with greater dependency weakness.
6466 This is enabled by default when scheduling is enabled, i.e.@:
6467 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6468 or at @option{-O2} or higher.
6470 @item -fsched-rank-heuristic
6471 @opindex fsched-rank-heuristic
6472 Enable the rank heuristic in the scheduler. This heuristic favors
6473 the instruction belonging to a basic block with greater size or frequency.
6474 This is enabled by default when scheduling is enabled, i.e.@:
6475 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6476 at @option{-O2} or higher.
6478 @item -fsched-last-insn-heuristic
6479 @opindex fsched-last-insn-heuristic
6480 Enable the last-instruction heuristic in the scheduler. This heuristic
6481 favors the instruction that is less dependent on the last instruction
6482 scheduled. This is enabled by default when scheduling is enabled,
6483 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6484 at @option{-O2} or higher.
6486 @item -fsched-dep-count-heuristic
6487 @opindex fsched-dep-count-heuristic
6488 Enable the dependent-count heuristic in the scheduler. This heuristic
6489 favors the instruction that has more instructions depending on it.
6490 This is enabled by default when scheduling is enabled, i.e.@:
6491 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6492 at @option{-O2} or higher.
6494 @item -freschedule-modulo-scheduled-loops
6495 @opindex freschedule-modulo-scheduled-loops
6496 The modulo scheduling comes before the traditional scheduling, if a loop
6497 was modulo scheduled we may want to prevent the later scheduling passes
6498 from changing its schedule, we use this option to control that.
6500 @item -fselective-scheduling
6501 @opindex fselective-scheduling
6502 Schedule instructions using selective scheduling algorithm. Selective
6503 scheduling runs instead of the first scheduler pass.
6505 @item -fselective-scheduling2
6506 @opindex fselective-scheduling2
6507 Schedule instructions using selective scheduling algorithm. Selective
6508 scheduling runs instead of the second scheduler pass.
6510 @item -fsel-sched-pipelining
6511 @opindex fsel-sched-pipelining
6512 Enable software pipelining of innermost loops during selective scheduling.
6513 This option has no effect until one of @option{-fselective-scheduling} or
6514 @option{-fselective-scheduling2} is turned on.
6516 @item -fsel-sched-pipelining-outer-loops
6517 @opindex fsel-sched-pipelining-outer-loops
6518 When pipelining loops during selective scheduling, also pipeline outer loops.
6519 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6521 @item -fcaller-saves
6522 @opindex fcaller-saves
6523 Enable values to be allocated in registers that will be clobbered by
6524 function calls, by emitting extra instructions to save and restore the
6525 registers around such calls. Such allocation is done only when it
6526 seems to result in better code than would otherwise be produced.
6528 This option is always enabled by default on certain machines, usually
6529 those which have no call-preserved registers to use instead.
6531 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6533 @item -fconserve-stack
6534 @opindex fconserve-stack
6535 Attempt to minimize stack usage. The compiler will attempt to use less
6536 stack space, even if that makes the program slower. This option
6537 implies setting the @option{large-stack-frame} parameter to 100
6538 and the @option{large-stack-frame-growth} parameter to 400.
6540 @item -ftree-reassoc
6541 @opindex ftree-reassoc
6542 Perform reassociation on trees. This flag is enabled by default
6543 at @option{-O} and higher.
6547 Perform partial redundancy elimination (PRE) on trees. This flag is
6548 enabled by default at @option{-O2} and @option{-O3}.
6550 @item -ftree-forwprop
6551 @opindex ftree-forwprop
6552 Perform forward propagation on trees. This flag is enabled by default
6553 at @option{-O} and higher.
6557 Perform full redundancy elimination (FRE) on trees. The difference
6558 between FRE and PRE is that FRE only considers expressions
6559 that are computed on all paths leading to the redundant computation.
6560 This analysis is faster than PRE, though it exposes fewer redundancies.
6561 This flag is enabled by default at @option{-O} and higher.
6563 @item -ftree-phiprop
6564 @opindex ftree-phiprop
6565 Perform hoisting of loads from conditional pointers on trees. This
6566 pass is enabled by default at @option{-O} and higher.
6568 @item -ftree-copy-prop
6569 @opindex ftree-copy-prop
6570 Perform copy propagation on trees. This pass eliminates unnecessary
6571 copy operations. This flag is enabled by default at @option{-O} and
6574 @item -fipa-pure-const
6575 @opindex fipa-pure-const
6576 Discover which functions are pure or constant.
6577 Enabled by default at @option{-O} and higher.
6579 @item -fipa-reference
6580 @opindex fipa-reference
6581 Discover which static variables do not escape cannot escape the
6583 Enabled by default at @option{-O} and higher.
6585 @item -fipa-struct-reorg
6586 @opindex fipa-struct-reorg
6587 Perform structure reorganization optimization, that change C-like structures
6588 layout in order to better utilize spatial locality. This transformation is
6589 affective for programs containing arrays of structures. Available in two
6590 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6591 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6592 to provide the safety of this transformation. It works only in whole program
6593 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6594 enabled. Structures considered @samp{cold} by this transformation are not
6595 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6597 With this flag, the program debug info reflects a new structure layout.
6601 Perform interprocedural pointer analysis and interprocedural modification
6602 and reference analysis. This option can cause excessive memory and
6603 compile-time usage on large compilation units. It is not enabled by
6604 default at any optimization level.
6607 @opindex fipa-profile
6608 Perform interprocedural profile propagation. The functions called only from
6609 cold functions are marked as cold. Also functions executed once (such as
6610 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6611 functions and loop less parts of functions executed once are then optimized for
6613 Enabled by default at @option{-O} and higher.
6617 Perform interprocedural constant propagation.
6618 This optimization analyzes the program to determine when values passed
6619 to functions are constants and then optimizes accordingly.
6620 This optimization can substantially increase performance
6621 if the application has constants passed to functions.
6622 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6624 @item -fipa-cp-clone
6625 @opindex fipa-cp-clone
6626 Perform function cloning to make interprocedural constant propagation stronger.
6627 When enabled, interprocedural constant propagation will perform function cloning
6628 when externally visible function can be called with constant arguments.
6629 Because this optimization can create multiple copies of functions,
6630 it may significantly increase code size
6631 (see @option{--param ipcp-unit-growth=@var{value}}).
6632 This flag is enabled by default at @option{-O3}.
6634 @item -fipa-matrix-reorg
6635 @opindex fipa-matrix-reorg
6636 Perform matrix flattening and transposing.
6637 Matrix flattening tries to replace an @math{m}-dimensional matrix
6638 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6639 This reduces the level of indirection needed for accessing the elements
6640 of the matrix. The second optimization is matrix transposing that
6641 attempts to change the order of the matrix's dimensions in order to
6642 improve cache locality.
6643 Both optimizations need the @option{-fwhole-program} flag.
6644 Transposing is enabled only if profiling information is available.
6648 Perform forward store motion on trees. This flag is
6649 enabled by default at @option{-O} and higher.
6653 Perform sparse conditional constant propagation (CCP) on trees. This
6654 pass only operates on local scalar variables and is enabled by default
6655 at @option{-O} and higher.
6657 @item -ftree-switch-conversion
6658 Perform conversion of simple initializations in a switch to
6659 initializations from a scalar array. This flag is enabled by default
6660 at @option{-O2} and higher.
6664 Perform dead code elimination (DCE) on trees. This flag is enabled by
6665 default at @option{-O} and higher.
6667 @item -ftree-builtin-call-dce
6668 @opindex ftree-builtin-call-dce
6669 Perform conditional dead code elimination (DCE) for calls to builtin functions
6670 that may set @code{errno} but are otherwise side-effect free. This flag is
6671 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6674 @item -ftree-dominator-opts
6675 @opindex ftree-dominator-opts
6676 Perform a variety of simple scalar cleanups (constant/copy
6677 propagation, redundancy elimination, range propagation and expression
6678 simplification) based on a dominator tree traversal. This also
6679 performs jump threading (to reduce jumps to jumps). This flag is
6680 enabled by default at @option{-O} and higher.
6684 Perform dead store elimination (DSE) on trees. A dead store is a store into
6685 a memory location which will later be overwritten by another store without
6686 any intervening loads. In this case the earlier store can be deleted. This
6687 flag is enabled by default at @option{-O} and higher.
6691 Perform loop header copying on trees. This is beneficial since it increases
6692 effectiveness of code motion optimizations. It also saves one jump. This flag
6693 is enabled by default at @option{-O} and higher. It is not enabled
6694 for @option{-Os}, since it usually increases code size.
6696 @item -ftree-loop-optimize
6697 @opindex ftree-loop-optimize
6698 Perform loop optimizations on trees. This flag is enabled by default
6699 at @option{-O} and higher.
6701 @item -ftree-loop-linear
6702 @opindex ftree-loop-linear
6703 Perform linear loop transformations on tree. This flag can improve cache
6704 performance and allow further loop optimizations to take place.
6706 @item -floop-interchange
6707 Perform loop interchange transformations on loops. Interchanging two
6708 nested loops switches the inner and outer loops. For example, given a
6713 A(J, I) = A(J, I) * C
6717 loop interchange will transform the loop as if the user had written:
6721 A(J, I) = A(J, I) * C
6725 which can be beneficial when @code{N} is larger than the caches,
6726 because in Fortran, the elements of an array are stored in memory
6727 contiguously by column, and the original loop iterates over rows,
6728 potentially creating at each access a cache miss. This optimization
6729 applies to all the languages supported by GCC and is not limited to
6730 Fortran. To use this code transformation, GCC has to be configured
6731 with @option{--with-ppl} and @option{--with-cloog} to enable the
6732 Graphite loop transformation infrastructure.
6734 @item -floop-strip-mine
6735 Perform loop strip mining transformations on loops. Strip mining
6736 splits a loop into two nested loops. The outer loop has strides
6737 equal to the strip size and the inner loop has strides of the
6738 original loop within a strip. The strip length can be changed
6739 using the @option{loop-block-tile-size} parameter. For example,
6746 loop strip mining will transform the loop as if the user had written:
6749 DO I = II, min (II + 50, N)
6754 This optimization applies to all the languages supported by GCC and is
6755 not limited to Fortran. To use this code transformation, GCC has to
6756 be configured with @option{--with-ppl} and @option{--with-cloog} to
6757 enable the Graphite loop transformation infrastructure.
6760 Perform loop blocking transformations on loops. Blocking strip mines
6761 each loop in the loop nest such that the memory accesses of the
6762 element loops fit inside caches. The strip length can be changed
6763 using the @option{loop-block-tile-size} parameter. For example, given
6768 A(J, I) = B(I) + C(J)
6772 loop blocking will transform the loop as if the user had written:
6776 DO I = II, min (II + 50, N)
6777 DO J = JJ, min (JJ + 50, M)
6778 A(J, I) = B(I) + C(J)
6784 which can be beneficial when @code{M} is larger than the caches,
6785 because the innermost loop will iterate over a smaller amount of data
6786 that can be kept in the caches. This optimization applies to all the
6787 languages supported by GCC and is not limited to Fortran. To use this
6788 code transformation, GCC has to be configured with @option{--with-ppl}
6789 and @option{--with-cloog} to enable the Graphite loop transformation
6792 @item -fgraphite-identity
6793 @opindex fgraphite-identity
6794 Enable the identity transformation for graphite. For every SCoP we generate
6795 the polyhedral representation and transform it back to gimple. Using
6796 @option{-fgraphite-identity} we can check the costs or benefits of the
6797 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6798 are also performed by the code generator CLooG, like index splitting and
6799 dead code elimination in loops.
6801 @item -floop-parallelize-all
6802 Use the Graphite data dependence analysis to identify loops that can
6803 be parallelized. Parallelize all the loops that can be analyzed to
6804 not contain loop carried dependences without checking that it is
6805 profitable to parallelize the loops.
6807 @item -fcheck-data-deps
6808 @opindex fcheck-data-deps
6809 Compare the results of several data dependence analyzers. This option
6810 is used for debugging the data dependence analyzers.
6812 @item -ftree-loop-distribution
6813 Perform loop distribution. This flag can improve cache performance on
6814 big loop bodies and allow further loop optimizations, like
6815 parallelization or vectorization, to take place. For example, the loop
6832 @item -ftree-loop-im
6833 @opindex ftree-loop-im
6834 Perform loop invariant motion on trees. This pass moves only invariants that
6835 would be hard to handle at RTL level (function calls, operations that expand to
6836 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6837 operands of conditions that are invariant out of the loop, so that we can use
6838 just trivial invariantness analysis in loop unswitching. The pass also includes
6841 @item -ftree-loop-ivcanon
6842 @opindex ftree-loop-ivcanon
6843 Create a canonical counter for number of iterations in the loop for that
6844 determining number of iterations requires complicated analysis. Later
6845 optimizations then may determine the number easily. Useful especially
6846 in connection with unrolling.
6850 Perform induction variable optimizations (strength reduction, induction
6851 variable merging and induction variable elimination) on trees.
6853 @item -ftree-parallelize-loops=n
6854 @opindex ftree-parallelize-loops
6855 Parallelize loops, i.e., split their iteration space to run in n threads.
6856 This is only possible for loops whose iterations are independent
6857 and can be arbitrarily reordered. The optimization is only
6858 profitable on multiprocessor machines, for loops that are CPU-intensive,
6859 rather than constrained e.g.@: by memory bandwidth. This option
6860 implies @option{-pthread}, and thus is only supported on targets
6861 that have support for @option{-pthread}.
6865 Perform function-local points-to analysis on trees. This flag is
6866 enabled by default at @option{-O} and higher.
6870 Perform scalar replacement of aggregates. This pass replaces structure
6871 references with scalars to prevent committing structures to memory too
6872 early. This flag is enabled by default at @option{-O} and higher.
6874 @item -ftree-copyrename
6875 @opindex ftree-copyrename
6876 Perform copy renaming on trees. This pass attempts to rename compiler
6877 temporaries to other variables at copy locations, usually resulting in
6878 variable names which more closely resemble the original variables. This flag
6879 is enabled by default at @option{-O} and higher.
6883 Perform temporary expression replacement during the SSA->normal phase. Single
6884 use/single def temporaries are replaced at their use location with their
6885 defining expression. This results in non-GIMPLE code, but gives the expanders
6886 much more complex trees to work on resulting in better RTL generation. This is
6887 enabled by default at @option{-O} and higher.
6889 @item -ftree-vectorize
6890 @opindex ftree-vectorize
6891 Perform loop vectorization on trees. This flag is enabled by default at
6894 @item -ftree-slp-vectorize
6895 @opindex ftree-slp-vectorize
6896 Perform basic block vectorization on trees. This flag is enabled by default at
6897 @option{-O3} and when @option{-ftree-vectorize} is enabled.
6899 @item -ftree-vect-loop-version
6900 @opindex ftree-vect-loop-version
6901 Perform loop versioning when doing loop vectorization on trees. When a loop
6902 appears to be vectorizable except that data alignment or data dependence cannot
6903 be determined at compile time then vectorized and non-vectorized versions of
6904 the loop are generated along with runtime checks for alignment or dependence
6905 to control which version is executed. This option is enabled by default
6906 except at level @option{-Os} where it is disabled.
6908 @item -fvect-cost-model
6909 @opindex fvect-cost-model
6910 Enable cost model for vectorization.
6914 Perform Value Range Propagation on trees. This is similar to the
6915 constant propagation pass, but instead of values, ranges of values are
6916 propagated. This allows the optimizers to remove unnecessary range
6917 checks like array bound checks and null pointer checks. This is
6918 enabled by default at @option{-O2} and higher. Null pointer check
6919 elimination is only done if @option{-fdelete-null-pointer-checks} is
6924 Perform tail duplication to enlarge superblock size. This transformation
6925 simplifies the control flow of the function allowing other optimizations to do
6928 @item -funroll-loops
6929 @opindex funroll-loops
6930 Unroll loops whose number of iterations can be determined at compile
6931 time or upon entry to the loop. @option{-funroll-loops} implies
6932 @option{-frerun-cse-after-loop}. This option makes code larger,
6933 and may or may not make it run faster.
6935 @item -funroll-all-loops
6936 @opindex funroll-all-loops
6937 Unroll all loops, even if their number of iterations is uncertain when
6938 the loop is entered. This usually makes programs run more slowly.
6939 @option{-funroll-all-loops} implies the same options as
6940 @option{-funroll-loops},
6942 @item -fsplit-ivs-in-unroller
6943 @opindex fsplit-ivs-in-unroller
6944 Enables expressing of values of induction variables in later iterations
6945 of the unrolled loop using the value in the first iteration. This breaks
6946 long dependency chains, thus improving efficiency of the scheduling passes.
6948 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6949 same effect. However in cases the loop body is more complicated than
6950 a single basic block, this is not reliable. It also does not work at all
6951 on some of the architectures due to restrictions in the CSE pass.
6953 This optimization is enabled by default.
6955 @item -fvariable-expansion-in-unroller
6956 @opindex fvariable-expansion-in-unroller
6957 With this option, the compiler will create multiple copies of some
6958 local variables when unrolling a loop which can result in superior code.
6960 @item -fpredictive-commoning
6961 @opindex fpredictive-commoning
6962 Perform predictive commoning optimization, i.e., reusing computations
6963 (especially memory loads and stores) performed in previous
6964 iterations of loops.
6966 This option is enabled at level @option{-O3}.
6968 @item -fprefetch-loop-arrays
6969 @opindex fprefetch-loop-arrays
6970 If supported by the target machine, generate instructions to prefetch
6971 memory to improve the performance of loops that access large arrays.
6973 This option may generate better or worse code; results are highly
6974 dependent on the structure of loops within the source code.
6976 Disabled at level @option{-Os}.
6979 @itemx -fno-peephole2
6980 @opindex fno-peephole
6981 @opindex fno-peephole2
6982 Disable any machine-specific peephole optimizations. The difference
6983 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6984 are implemented in the compiler; some targets use one, some use the
6985 other, a few use both.
6987 @option{-fpeephole} is enabled by default.
6988 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6990 @item -fno-guess-branch-probability
6991 @opindex fno-guess-branch-probability
6992 Do not guess branch probabilities using heuristics.
6994 GCC will use heuristics to guess branch probabilities if they are
6995 not provided by profiling feedback (@option{-fprofile-arcs}). These
6996 heuristics are based on the control flow graph. If some branch probabilities
6997 are specified by @samp{__builtin_expect}, then the heuristics will be
6998 used to guess branch probabilities for the rest of the control flow graph,
6999 taking the @samp{__builtin_expect} info into account. The interactions
7000 between the heuristics and @samp{__builtin_expect} can be complex, and in
7001 some cases, it may be useful to disable the heuristics so that the effects
7002 of @samp{__builtin_expect} are easier to understand.
7004 The default is @option{-fguess-branch-probability} at levels
7005 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7007 @item -freorder-blocks
7008 @opindex freorder-blocks
7009 Reorder basic blocks in the compiled function in order to reduce number of
7010 taken branches and improve code locality.
7012 Enabled at levels @option{-O2}, @option{-O3}.
7014 @item -freorder-blocks-and-partition
7015 @opindex freorder-blocks-and-partition
7016 In addition to reordering basic blocks in the compiled function, in order
7017 to reduce number of taken branches, partitions hot and cold basic blocks
7018 into separate sections of the assembly and .o files, to improve
7019 paging and cache locality performance.
7021 This optimization is automatically turned off in the presence of
7022 exception handling, for linkonce sections, for functions with a user-defined
7023 section attribute and on any architecture that does not support named
7026 @item -freorder-functions
7027 @opindex freorder-functions
7028 Reorder functions in the object file in order to
7029 improve code locality. This is implemented by using special
7030 subsections @code{.text.hot} for most frequently executed functions and
7031 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7032 the linker so object file format must support named sections and linker must
7033 place them in a reasonable way.
7035 Also profile feedback must be available in to make this option effective. See
7036 @option{-fprofile-arcs} for details.
7038 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7040 @item -fstrict-aliasing
7041 @opindex fstrict-aliasing
7042 Allow the compiler to assume the strictest aliasing rules applicable to
7043 the language being compiled. For C (and C++), this activates
7044 optimizations based on the type of expressions. In particular, an
7045 object of one type is assumed never to reside at the same address as an
7046 object of a different type, unless the types are almost the same. For
7047 example, an @code{unsigned int} can alias an @code{int}, but not a
7048 @code{void*} or a @code{double}. A character type may alias any other
7051 @anchor{Type-punning}Pay special attention to code like this:
7064 The practice of reading from a different union member than the one most
7065 recently written to (called ``type-punning'') is common. Even with
7066 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7067 is accessed through the union type. So, the code above will work as
7068 expected. @xref{Structures unions enumerations and bit-fields
7069 implementation}. However, this code might not:
7080 Similarly, access by taking the address, casting the resulting pointer
7081 and dereferencing the result has undefined behavior, even if the cast
7082 uses a union type, e.g.:
7086 return ((union a_union *) &d)->i;
7090 The @option{-fstrict-aliasing} option is enabled at levels
7091 @option{-O2}, @option{-O3}, @option{-Os}.
7093 @item -fstrict-overflow
7094 @opindex fstrict-overflow
7095 Allow the compiler to assume strict signed overflow rules, depending
7096 on the language being compiled. For C (and C++) this means that
7097 overflow when doing arithmetic with signed numbers is undefined, which
7098 means that the compiler may assume that it will not happen. This
7099 permits various optimizations. For example, the compiler will assume
7100 that an expression like @code{i + 10 > i} will always be true for
7101 signed @code{i}. This assumption is only valid if signed overflow is
7102 undefined, as the expression is false if @code{i + 10} overflows when
7103 using twos complement arithmetic. When this option is in effect any
7104 attempt to determine whether an operation on signed numbers will
7105 overflow must be written carefully to not actually involve overflow.
7107 This option also allows the compiler to assume strict pointer
7108 semantics: given a pointer to an object, if adding an offset to that
7109 pointer does not produce a pointer to the same object, the addition is
7110 undefined. This permits the compiler to conclude that @code{p + u >
7111 p} is always true for a pointer @code{p} and unsigned integer
7112 @code{u}. This assumption is only valid because pointer wraparound is
7113 undefined, as the expression is false if @code{p + u} overflows using
7114 twos complement arithmetic.
7116 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7117 that integer signed overflow is fully defined: it wraps. When
7118 @option{-fwrapv} is used, there is no difference between
7119 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7120 integers. With @option{-fwrapv} certain types of overflow are
7121 permitted. For example, if the compiler gets an overflow when doing
7122 arithmetic on constants, the overflowed value can still be used with
7123 @option{-fwrapv}, but not otherwise.
7125 The @option{-fstrict-overflow} option is enabled at levels
7126 @option{-O2}, @option{-O3}, @option{-Os}.
7128 @item -falign-functions
7129 @itemx -falign-functions=@var{n}
7130 @opindex falign-functions
7131 Align the start of functions to the next power-of-two greater than
7132 @var{n}, skipping up to @var{n} bytes. For instance,
7133 @option{-falign-functions=32} aligns functions to the next 32-byte
7134 boundary, but @option{-falign-functions=24} would align to the next
7135 32-byte boundary only if this can be done by skipping 23 bytes or less.
7137 @option{-fno-align-functions} and @option{-falign-functions=1} are
7138 equivalent and mean that functions will not be aligned.
7140 Some assemblers only support this flag when @var{n} is a power of two;
7141 in that case, it is rounded up.
7143 If @var{n} is not specified or is zero, use a machine-dependent default.
7145 Enabled at levels @option{-O2}, @option{-O3}.
7147 @item -falign-labels
7148 @itemx -falign-labels=@var{n}
7149 @opindex falign-labels
7150 Align all branch targets to a power-of-two boundary, skipping up to
7151 @var{n} bytes like @option{-falign-functions}. This option can easily
7152 make code slower, because it must insert dummy operations for when the
7153 branch target is reached in the usual flow of the code.
7155 @option{-fno-align-labels} and @option{-falign-labels=1} are
7156 equivalent and mean that labels will not be aligned.
7158 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7159 are greater than this value, then their values are used instead.
7161 If @var{n} is not specified or is zero, use a machine-dependent default
7162 which is very likely to be @samp{1}, meaning no alignment.
7164 Enabled at levels @option{-O2}, @option{-O3}.
7167 @itemx -falign-loops=@var{n}
7168 @opindex falign-loops
7169 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7170 like @option{-falign-functions}. The hope is that the loop will be
7171 executed many times, which will make up for any execution of the dummy
7174 @option{-fno-align-loops} and @option{-falign-loops=1} are
7175 equivalent and mean that loops will not be aligned.
7177 If @var{n} is not specified or is zero, use a machine-dependent default.
7179 Enabled at levels @option{-O2}, @option{-O3}.
7182 @itemx -falign-jumps=@var{n}
7183 @opindex falign-jumps
7184 Align branch targets to a power-of-two boundary, for branch targets
7185 where the targets can only be reached by jumping, skipping up to @var{n}
7186 bytes like @option{-falign-functions}. In this case, no dummy operations
7189 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7190 equivalent and mean that loops will not be aligned.
7192 If @var{n} is not specified or is zero, use a machine-dependent default.
7194 Enabled at levels @option{-O2}, @option{-O3}.
7196 @item -funit-at-a-time
7197 @opindex funit-at-a-time
7198 This option is left for compatibility reasons. @option{-funit-at-a-time}
7199 has no effect, while @option{-fno-unit-at-a-time} implies
7200 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7204 @item -fno-toplevel-reorder
7205 @opindex fno-toplevel-reorder
7206 Do not reorder top-level functions, variables, and @code{asm}
7207 statements. Output them in the same order that they appear in the
7208 input file. When this option is used, unreferenced static variables
7209 will not be removed. This option is intended to support existing code
7210 which relies on a particular ordering. For new code, it is better to
7213 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7214 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7219 Constructs webs as commonly used for register allocation purposes and assign
7220 each web individual pseudo register. This allows the register allocation pass
7221 to operate on pseudos directly, but also strengthens several other optimization
7222 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7223 however, make debugging impossible, since variables will no longer stay in a
7226 Enabled by default with @option{-funroll-loops}.
7228 @item -fwhole-program
7229 @opindex fwhole-program
7230 Assume that the current compilation unit represents the whole program being
7231 compiled. All public functions and variables with the exception of @code{main}
7232 and those merged by attribute @code{externally_visible} become static functions
7233 and in effect are optimized more aggressively by interprocedural optimizers.
7234 While this option is equivalent to proper use of the @code{static} keyword for
7235 programs consisting of a single file, in combination with option
7236 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7237 compile many smaller scale programs since the functions and variables become
7238 local for the whole combined compilation unit, not for the single source file
7241 This option implies @option{-fwhole-file} for Fortran programs.
7245 This option runs the standard link-time optimizer. When invoked
7246 with source code, it generates GIMPLE (one of GCC's internal
7247 representations) and writes it to special ELF sections in the object
7248 file. When the object files are linked together, all the function
7249 bodies are read from these ELF sections and instantiated as if they
7250 had been part of the same translation unit.
7252 To use the link-timer optimizer, @option{-flto} needs to be specified at
7253 compile time and during the final link. For example,
7256 gcc -c -O2 -flto foo.c
7257 gcc -c -O2 -flto bar.c
7258 gcc -o myprog -flto -O2 foo.o bar.o
7261 The first two invocations to GCC will save a bytecode representation
7262 of GIMPLE into special ELF sections inside @file{foo.o} and
7263 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7264 @file{foo.o} and @file{bar.o}, merge the two files into a single
7265 internal image, and compile the result as usual. Since both
7266 @file{foo.o} and @file{bar.o} are merged into a single image, this
7267 causes all the inter-procedural analyses and optimizations in GCC to
7268 work across the two files as if they were a single one. This means,
7269 for example, that the inliner will be able to inline functions in
7270 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7272 Another (simpler) way to enable link-time optimization is,
7275 gcc -o myprog -flto -O2 foo.c bar.c
7278 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7279 merge them together into a single GIMPLE representation and optimize
7280 them as usual to produce @file{myprog}.
7282 The only important thing to keep in mind is that to enable link-time
7283 optimizations the @option{-flto} flag needs to be passed to both the
7284 compile and the link commands.
7286 Note that when a file is compiled with @option{-flto}, the generated
7287 object file will be larger than a regular object file because it will
7288 contain GIMPLE bytecodes and the usual final code. This means that
7289 object files with LTO information can be linked as a normal object
7290 file. So, in the previous example, if the final link is done with
7293 gcc -o myprog foo.o bar.o
7296 The only difference will be that no inter-procedural optimizations
7297 will be applied to produce @file{myprog}. The two object files
7298 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7301 Additionally, the optimization flags used to compile individual files
7302 are not necessarily related to those used at link-time. For instance,
7305 gcc -c -O0 -flto foo.c
7306 gcc -c -O0 -flto bar.c
7307 gcc -o myprog -flto -O3 foo.o bar.o
7310 This will produce individual object files with unoptimized assembler
7311 code, but the resulting binary @file{myprog} will be optimized at
7312 @option{-O3}. Now, if the final binary is generated without
7313 @option{-flto}, then @file{myprog} will not be optimized.
7315 When producing the final binary with @option{-flto}, GCC will only
7316 apply link-time optimizations to those files that contain bytecode.
7317 Therefore, you can mix and match object files and libraries with
7318 GIMPLE bytecodes and final object code. GCC will automatically select
7319 which files to optimize in LTO mode and which files to link without
7322 There are some code generation flags that GCC will preserve when
7323 generating bytecodes, as they need to be used during the final link
7324 stage. Currently, the following options are saved into the GIMPLE
7325 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7326 @option{-m} target flags.
7328 At link time, these options are read-in and reapplied. Note that the
7329 current implementation makes no attempt at recognizing conflicting
7330 values for these options. If two or more files have a conflicting
7331 value (e.g., one file is compiled with @option{-fPIC} and another
7332 isn't), the compiler will simply use the last value read from the
7333 bytecode files. It is recommended, then, that all the files
7334 participating in the same link be compiled with the same options.
7336 Another feature of LTO is that it is possible to apply interprocedural
7337 optimizations on files written in different languages. This requires
7338 some support in the language front end. Currently, the C, C++ and
7339 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7340 something like this should work
7345 gfortran -c -flto baz.f90
7346 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7349 Notice that the final link is done with @command{g++} to get the C++
7350 runtime libraries and @option{-lgfortran} is added to get the Fortran
7351 runtime libraries. In general, when mixing languages in LTO mode, you
7352 should use the same link command used when mixing languages in a
7353 regular (non-LTO) compilation. This means that if your build process
7354 was mixing languages before, all you need to add is @option{-flto} to
7355 all the compile and link commands.
7357 If LTO encounters objects with C linkage declared with incompatible
7358 types in separate translation units to be linked together (undefined
7359 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7360 issued. The behavior is still undefined at runtime.
7362 If object files containing GIMPLE bytecode are stored in a library
7363 archive, say @file{libfoo.a}, it is possible to extract and use them
7364 in an LTO link if you are using @command{gold} as the linker (which,
7365 in turn requires GCC to be configured with @option{--enable-gold}).
7366 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7370 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7373 With the linker plugin enabled, @command{gold} will extract the needed
7374 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7375 to make them part of the aggregated GIMPLE image to be optimized.
7377 If you are not using @command{gold} and/or do not specify
7378 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7379 will be extracted and linked as usual, but they will not participate
7380 in the LTO optimization process.
7382 Link time optimizations do not require the presence of the whole
7383 program to operate. If the program does not require any symbols to
7384 be exported, it is possible to combine @option{-flto} and
7385 @option{-fwhopr} with @option{-fwhole-program} to allow the
7386 interprocedural optimizers to use more aggressive assumptions which
7387 may lead to improved optimization opportunities.
7389 Regarding portability: the current implementation of LTO makes no
7390 attempt at generating bytecode that can be ported between different
7391 types of hosts. The bytecode files are versioned and there is a
7392 strict version check, so bytecode files generated in one version of
7393 GCC will not work with an older/newer version of GCC.
7395 Link time optimization does not play well with generating debugging
7396 information. Combining @option{-flto} or @option{-fwhopr} with
7397 @option{-g} is experimental.
7399 This option is disabled by default.
7403 This option is identical in functionality to @option{-flto} but it
7404 differs in how the final link stage is executed. Instead of loading
7405 all the function bodies in memory, the callgraph is analyzed and
7406 optimization decisions are made (whole program analysis or WPA). Once
7407 optimization decisions are made, the callgraph is partitioned and the
7408 different sections are compiled separately (local transformations or
7409 LTRANS)@. This process allows optimizations on very large programs
7410 that otherwise would not fit in memory. This option enables
7411 @option{-fwpa} and @option{-fltrans} automatically.
7413 Disabled by default.
7415 This option is experimental.
7419 This is an internal option used by GCC when compiling with
7420 @option{-fwhopr}. You should never need to use it.
7422 This option runs the link-time optimizer in the whole-program-analysis
7423 (WPA) mode, which reads in summary information from all inputs and
7424 performs a whole-program analysis based on summary information only.
7425 It generates object files for subsequent runs of the link-time
7426 optimizer where individual object files are optimized using both
7427 summary information from the WPA mode and the actual function bodies.
7428 It then drives the LTRANS phase.
7430 Disabled by default.
7434 This is an internal option used by GCC when compiling with
7435 @option{-fwhopr}. You should never need to use it.
7437 This option runs the link-time optimizer in the local-transformation (LTRANS)
7438 mode, which reads in output from a previous run of the LTO in WPA mode.
7439 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7441 Disabled by default.
7443 @item -fltrans-output-list=@var{file}
7444 @opindex fltrans-output-list
7445 This is an internal option used by GCC when compiling with
7446 @option{-fwhopr}. You should never need to use it.
7448 This option specifies a file to which the names of LTRANS output files are
7449 written. This option is only meaningful in conjunction with @option{-fwpa}.
7451 Disabled by default.
7453 @item -flto-compression-level=@var{n}
7454 This option specifies the level of compression used for intermediate
7455 language written to LTO object files, and is only meaningful in
7456 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7457 values are 0 (no compression) to 9 (maximum compression). Values
7458 outside this range are clamped to either 0 or 9. If the option is not
7459 given, a default balanced compression setting is used.
7462 Prints a report with internal details on the workings of the link-time
7463 optimizer. The contents of this report vary from version to version,
7464 it is meant to be useful to GCC developers when processing object
7465 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7467 Disabled by default.
7469 @item -fuse-linker-plugin
7470 Enables the extraction of objects with GIMPLE bytecode information
7471 from library archives. This option relies on features available only
7472 in @command{gold}, so to use this you must configure GCC with
7473 @option{--enable-gold}. See @option{-flto} for a description on the
7474 effect of this flag and how to use it.
7476 Disabled by default.
7478 @item -fcprop-registers
7479 @opindex fcprop-registers
7480 After register allocation and post-register allocation instruction splitting,
7481 we perform a copy-propagation pass to try to reduce scheduling dependencies
7482 and occasionally eliminate the copy.
7484 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7486 @item -fprofile-correction
7487 @opindex fprofile-correction
7488 Profiles collected using an instrumented binary for multi-threaded programs may
7489 be inconsistent due to missed counter updates. When this option is specified,
7490 GCC will use heuristics to correct or smooth out such inconsistencies. By
7491 default, GCC will emit an error message when an inconsistent profile is detected.
7493 @item -fprofile-dir=@var{path}
7494 @opindex fprofile-dir
7496 Set the directory to search the profile data files in to @var{path}.
7497 This option affects only the profile data generated by
7498 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7499 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7500 and its related options.
7501 By default, GCC will use the current directory as @var{path}
7502 thus the profile data file will appear in the same directory as the object file.
7504 @item -fprofile-generate
7505 @itemx -fprofile-generate=@var{path}
7506 @opindex fprofile-generate
7508 Enable options usually used for instrumenting application to produce
7509 profile useful for later recompilation with profile feedback based
7510 optimization. You must use @option{-fprofile-generate} both when
7511 compiling and when linking your program.
7513 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7515 If @var{path} is specified, GCC will look at the @var{path} to find
7516 the profile feedback data files. See @option{-fprofile-dir}.
7519 @itemx -fprofile-use=@var{path}
7520 @opindex fprofile-use
7521 Enable profile feedback directed optimizations, and optimizations
7522 generally profitable only with profile feedback available.
7524 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7525 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7527 By default, GCC emits an error message if the feedback profiles do not
7528 match the source code. This error can be turned into a warning by using
7529 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7532 If @var{path} is specified, GCC will look at the @var{path} to find
7533 the profile feedback data files. See @option{-fprofile-dir}.
7536 The following options control compiler behavior regarding floating
7537 point arithmetic. These options trade off between speed and
7538 correctness. All must be specifically enabled.
7542 @opindex ffloat-store
7543 Do not store floating point variables in registers, and inhibit other
7544 options that might change whether a floating point value is taken from a
7547 @cindex floating point precision
7548 This option prevents undesirable excess precision on machines such as
7549 the 68000 where the floating registers (of the 68881) keep more
7550 precision than a @code{double} is supposed to have. Similarly for the
7551 x86 architecture. For most programs, the excess precision does only
7552 good, but a few programs rely on the precise definition of IEEE floating
7553 point. Use @option{-ffloat-store} for such programs, after modifying
7554 them to store all pertinent intermediate computations into variables.
7556 @item -fexcess-precision=@var{style}
7557 @opindex fexcess-precision
7558 This option allows further control over excess precision on machines
7559 where floating-point registers have more precision than the IEEE
7560 @code{float} and @code{double} types and the processor does not
7561 support operations rounding to those types. By default,
7562 @option{-fexcess-precision=fast} is in effect; this means that
7563 operations are carried out in the precision of the registers and that
7564 it is unpredictable when rounding to the types specified in the source
7565 code takes place. When compiling C, if
7566 @option{-fexcess-precision=standard} is specified then excess
7567 precision will follow the rules specified in ISO C99; in particular,
7568 both casts and assignments cause values to be rounded to their
7569 semantic types (whereas @option{-ffloat-store} only affects
7570 assignments). This option is enabled by default for C if a strict
7571 conformance option such as @option{-std=c99} is used.
7574 @option{-fexcess-precision=standard} is not implemented for languages
7575 other than C, and has no effect if
7576 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7577 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7578 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7579 semantics apply without excess precision, and in the latter, rounding
7584 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7585 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7586 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7588 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7590 This option is not turned on by any @option{-O} option since
7591 it can result in incorrect output for programs which depend on
7592 an exact implementation of IEEE or ISO rules/specifications for
7593 math functions. It may, however, yield faster code for programs
7594 that do not require the guarantees of these specifications.
7596 @item -fno-math-errno
7597 @opindex fno-math-errno
7598 Do not set ERRNO after calling math functions that are executed
7599 with a single instruction, e.g., sqrt. A program that relies on
7600 IEEE exceptions for math error handling may want to use this flag
7601 for speed while maintaining IEEE arithmetic compatibility.
7603 This option is not turned on by any @option{-O} option since
7604 it can result in incorrect output for programs which depend on
7605 an exact implementation of IEEE or ISO rules/specifications for
7606 math functions. It may, however, yield faster code for programs
7607 that do not require the guarantees of these specifications.
7609 The default is @option{-fmath-errno}.
7611 On Darwin systems, the math library never sets @code{errno}. There is
7612 therefore no reason for the compiler to consider the possibility that
7613 it might, and @option{-fno-math-errno} is the default.
7615 @item -funsafe-math-optimizations
7616 @opindex funsafe-math-optimizations
7618 Allow optimizations for floating-point arithmetic that (a) assume
7619 that arguments and results are valid and (b) may violate IEEE or
7620 ANSI standards. When used at link-time, it may include libraries
7621 or startup files that change the default FPU control word or other
7622 similar optimizations.
7624 This option is not turned on by any @option{-O} option since
7625 it can result in incorrect output for programs which depend on
7626 an exact implementation of IEEE or ISO rules/specifications for
7627 math functions. It may, however, yield faster code for programs
7628 that do not require the guarantees of these specifications.
7629 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7630 @option{-fassociative-math} and @option{-freciprocal-math}.
7632 The default is @option{-fno-unsafe-math-optimizations}.
7634 @item -fassociative-math
7635 @opindex fassociative-math
7637 Allow re-association of operands in series of floating-point operations.
7638 This violates the ISO C and C++ language standard by possibly changing
7639 computation result. NOTE: re-ordering may change the sign of zero as
7640 well as ignore NaNs and inhibit or create underflow or overflow (and
7641 thus cannot be used on a code which relies on rounding behavior like
7642 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7643 and thus may not be used when ordered comparisons are required.
7644 This option requires that both @option{-fno-signed-zeros} and
7645 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7646 much sense with @option{-frounding-math}. For Fortran the option
7647 is automatically enabled when both @option{-fno-signed-zeros} and
7648 @option{-fno-trapping-math} are in effect.
7650 The default is @option{-fno-associative-math}.
7652 @item -freciprocal-math
7653 @opindex freciprocal-math
7655 Allow the reciprocal of a value to be used instead of dividing by
7656 the value if this enables optimizations. For example @code{x / y}
7657 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7658 is subject to common subexpression elimination. Note that this loses
7659 precision and increases the number of flops operating on the value.
7661 The default is @option{-fno-reciprocal-math}.
7663 @item -ffinite-math-only
7664 @opindex ffinite-math-only
7665 Allow optimizations for floating-point arithmetic that assume
7666 that arguments and results are not NaNs or +-Infs.
7668 This option is not turned on by any @option{-O} option since
7669 it can result in incorrect output for programs which depend on
7670 an exact implementation of IEEE or ISO rules/specifications for
7671 math functions. It may, however, yield faster code for programs
7672 that do not require the guarantees of these specifications.
7674 The default is @option{-fno-finite-math-only}.
7676 @item -fno-signed-zeros
7677 @opindex fno-signed-zeros
7678 Allow optimizations for floating point arithmetic that ignore the
7679 signedness of zero. IEEE arithmetic specifies the behavior of
7680 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7681 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7682 This option implies that the sign of a zero result isn't significant.
7684 The default is @option{-fsigned-zeros}.
7686 @item -fno-trapping-math
7687 @opindex fno-trapping-math
7688 Compile code assuming that floating-point operations cannot generate
7689 user-visible traps. These traps include division by zero, overflow,
7690 underflow, inexact result and invalid operation. This option requires
7691 that @option{-fno-signaling-nans} be in effect. Setting this option may
7692 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7694 This option should never be turned on by any @option{-O} option since
7695 it can result in incorrect output for programs which depend on
7696 an exact implementation of IEEE or ISO rules/specifications for
7699 The default is @option{-ftrapping-math}.
7701 @item -frounding-math
7702 @opindex frounding-math
7703 Disable transformations and optimizations that assume default floating
7704 point rounding behavior. This is round-to-zero for all floating point
7705 to integer conversions, and round-to-nearest for all other arithmetic
7706 truncations. This option should be specified for programs that change
7707 the FP rounding mode dynamically, or that may be executed with a
7708 non-default rounding mode. This option disables constant folding of
7709 floating point expressions at compile-time (which may be affected by
7710 rounding mode) and arithmetic transformations that are unsafe in the
7711 presence of sign-dependent rounding modes.
7713 The default is @option{-fno-rounding-math}.
7715 This option is experimental and does not currently guarantee to
7716 disable all GCC optimizations that are affected by rounding mode.
7717 Future versions of GCC may provide finer control of this setting
7718 using C99's @code{FENV_ACCESS} pragma. This command line option
7719 will be used to specify the default state for @code{FENV_ACCESS}.
7721 @item -fsignaling-nans
7722 @opindex fsignaling-nans
7723 Compile code assuming that IEEE signaling NaNs may generate user-visible
7724 traps during floating-point operations. Setting this option disables
7725 optimizations that may change the number of exceptions visible with
7726 signaling NaNs. This option implies @option{-ftrapping-math}.
7728 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7731 The default is @option{-fno-signaling-nans}.
7733 This option is experimental and does not currently guarantee to
7734 disable all GCC optimizations that affect signaling NaN behavior.
7736 @item -fsingle-precision-constant
7737 @opindex fsingle-precision-constant
7738 Treat floating point constant as single precision constant instead of
7739 implicitly converting it to double precision constant.
7741 @item -fcx-limited-range
7742 @opindex fcx-limited-range
7743 When enabled, this option states that a range reduction step is not
7744 needed when performing complex division. Also, there is no checking
7745 whether the result of a complex multiplication or division is @code{NaN
7746 + I*NaN}, with an attempt to rescue the situation in that case. The
7747 default is @option{-fno-cx-limited-range}, but is enabled by
7748 @option{-ffast-math}.
7750 This option controls the default setting of the ISO C99
7751 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7754 @item -fcx-fortran-rules
7755 @opindex fcx-fortran-rules
7756 Complex multiplication and division follow Fortran rules. Range
7757 reduction is done as part of complex division, but there is no checking
7758 whether the result of a complex multiplication or division is @code{NaN
7759 + I*NaN}, with an attempt to rescue the situation in that case.
7761 The default is @option{-fno-cx-fortran-rules}.
7765 The following options control optimizations that may improve
7766 performance, but are not enabled by any @option{-O} options. This
7767 section includes experimental options that may produce broken code.
7770 @item -fbranch-probabilities
7771 @opindex fbranch-probabilities
7772 After running a program compiled with @option{-fprofile-arcs}
7773 (@pxref{Debugging Options,, Options for Debugging Your Program or
7774 @command{gcc}}), you can compile it a second time using
7775 @option{-fbranch-probabilities}, to improve optimizations based on
7776 the number of times each branch was taken. When the program
7777 compiled with @option{-fprofile-arcs} exits it saves arc execution
7778 counts to a file called @file{@var{sourcename}.gcda} for each source
7779 file. The information in this data file is very dependent on the
7780 structure of the generated code, so you must use the same source code
7781 and the same optimization options for both compilations.
7783 With @option{-fbranch-probabilities}, GCC puts a
7784 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7785 These can be used to improve optimization. Currently, they are only
7786 used in one place: in @file{reorg.c}, instead of guessing which path a
7787 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7788 exactly determine which path is taken more often.
7790 @item -fprofile-values
7791 @opindex fprofile-values
7792 If combined with @option{-fprofile-arcs}, it adds code so that some
7793 data about values of expressions in the program is gathered.
7795 With @option{-fbranch-probabilities}, it reads back the data gathered
7796 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7797 notes to instructions for their later usage in optimizations.
7799 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7803 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7804 a code to gather information about values of expressions.
7806 With @option{-fbranch-probabilities}, it reads back the data gathered
7807 and actually performs the optimizations based on them.
7808 Currently the optimizations include specialization of division operation
7809 using the knowledge about the value of the denominator.
7811 @item -frename-registers
7812 @opindex frename-registers
7813 Attempt to avoid false dependencies in scheduled code by making use
7814 of registers left over after register allocation. This optimization
7815 will most benefit processors with lots of registers. Depending on the
7816 debug information format adopted by the target, however, it can
7817 make debugging impossible, since variables will no longer stay in
7818 a ``home register''.
7820 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
7824 Perform tail duplication to enlarge superblock size. This transformation
7825 simplifies the control flow of the function allowing other optimizations to do
7828 Enabled with @option{-fprofile-use}.
7830 @item -funroll-loops
7831 @opindex funroll-loops
7832 Unroll loops whose number of iterations can be determined at compile time or
7833 upon entry to the loop. @option{-funroll-loops} implies
7834 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7835 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7836 small constant number of iterations). This option makes code larger, and may
7837 or may not make it run faster.
7839 Enabled with @option{-fprofile-use}.
7841 @item -funroll-all-loops
7842 @opindex funroll-all-loops
7843 Unroll all loops, even if their number of iterations is uncertain when
7844 the loop is entered. This usually makes programs run more slowly.
7845 @option{-funroll-all-loops} implies the same options as
7846 @option{-funroll-loops}.
7849 @opindex fpeel-loops
7850 Peels the loops for that there is enough information that they do not
7851 roll much (from profile feedback). It also turns on complete loop peeling
7852 (i.e.@: complete removal of loops with small constant number of iterations).
7854 Enabled with @option{-fprofile-use}.
7856 @item -fmove-loop-invariants
7857 @opindex fmove-loop-invariants
7858 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7859 at level @option{-O1}
7861 @item -funswitch-loops
7862 @opindex funswitch-loops
7863 Move branches with loop invariant conditions out of the loop, with duplicates
7864 of the loop on both branches (modified according to result of the condition).
7866 @item -ffunction-sections
7867 @itemx -fdata-sections
7868 @opindex ffunction-sections
7869 @opindex fdata-sections
7870 Place each function or data item into its own section in the output
7871 file if the target supports arbitrary sections. The name of the
7872 function or the name of the data item determines the section's name
7875 Use these options on systems where the linker can perform optimizations
7876 to improve locality of reference in the instruction space. Most systems
7877 using the ELF object format and SPARC processors running Solaris 2 have
7878 linkers with such optimizations. AIX may have these optimizations in
7881 Only use these options when there are significant benefits from doing
7882 so. When you specify these options, the assembler and linker will
7883 create larger object and executable files and will also be slower.
7884 You will not be able to use @code{gprof} on all systems if you
7885 specify this option and you may have problems with debugging if
7886 you specify both this option and @option{-g}.
7888 @item -fbranch-target-load-optimize
7889 @opindex fbranch-target-load-optimize
7890 Perform branch target register load optimization before prologue / epilogue
7892 The use of target registers can typically be exposed only during reload,
7893 thus hoisting loads out of loops and doing inter-block scheduling needs
7894 a separate optimization pass.
7896 @item -fbranch-target-load-optimize2
7897 @opindex fbranch-target-load-optimize2
7898 Perform branch target register load optimization after prologue / epilogue
7901 @item -fbtr-bb-exclusive
7902 @opindex fbtr-bb-exclusive
7903 When performing branch target register load optimization, don't reuse
7904 branch target registers in within any basic block.
7906 @item -fstack-protector
7907 @opindex fstack-protector
7908 Emit extra code to check for buffer overflows, such as stack smashing
7909 attacks. This is done by adding a guard variable to functions with
7910 vulnerable objects. This includes functions that call alloca, and
7911 functions with buffers larger than 8 bytes. The guards are initialized
7912 when a function is entered and then checked when the function exits.
7913 If a guard check fails, an error message is printed and the program exits.
7915 @item -fstack-protector-all
7916 @opindex fstack-protector-all
7917 Like @option{-fstack-protector} except that all functions are protected.
7919 @item -fsection-anchors
7920 @opindex fsection-anchors
7921 Try to reduce the number of symbolic address calculations by using
7922 shared ``anchor'' symbols to address nearby objects. This transformation
7923 can help to reduce the number of GOT entries and GOT accesses on some
7926 For example, the implementation of the following function @code{foo}:
7930 int foo (void) @{ return a + b + c; @}
7933 would usually calculate the addresses of all three variables, but if you
7934 compile it with @option{-fsection-anchors}, it will access the variables
7935 from a common anchor point instead. The effect is similar to the
7936 following pseudocode (which isn't valid C):
7941 register int *xr = &x;
7942 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7946 Not all targets support this option.
7948 @item --param @var{name}=@var{value}
7950 In some places, GCC uses various constants to control the amount of
7951 optimization that is done. For example, GCC will not inline functions
7952 that contain more that a certain number of instructions. You can
7953 control some of these constants on the command-line using the
7954 @option{--param} option.
7956 The names of specific parameters, and the meaning of the values, are
7957 tied to the internals of the compiler, and are subject to change
7958 without notice in future releases.
7960 In each case, the @var{value} is an integer. The allowable choices for
7961 @var{name} are given in the following table:
7964 @item struct-reorg-cold-struct-ratio
7965 The threshold ratio (as a percentage) between a structure frequency
7966 and the frequency of the hottest structure in the program. This parameter
7967 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7968 We say that if the ratio of a structure frequency, calculated by profiling,
7969 to the hottest structure frequency in the program is less than this
7970 parameter, then structure reorganization is not applied to this structure.
7973 @item predictable-branch-outcome
7974 When branch is predicted to be taken with probability lower than this threshold
7975 (in percent), then it is considered well predictable. The default is 10.
7977 @item max-crossjump-edges
7978 The maximum number of incoming edges to consider for crossjumping.
7979 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7980 the number of edges incoming to each block. Increasing values mean
7981 more aggressive optimization, making the compile time increase with
7982 probably small improvement in executable size.
7984 @item min-crossjump-insns
7985 The minimum number of instructions which must be matched at the end
7986 of two blocks before crossjumping will be performed on them. This
7987 value is ignored in the case where all instructions in the block being
7988 crossjumped from are matched. The default value is 5.
7990 @item max-grow-copy-bb-insns
7991 The maximum code size expansion factor when copying basic blocks
7992 instead of jumping. The expansion is relative to a jump instruction.
7993 The default value is 8.
7995 @item max-goto-duplication-insns
7996 The maximum number of instructions to duplicate to a block that jumps
7997 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7998 passes, GCC factors computed gotos early in the compilation process,
7999 and unfactors them as late as possible. Only computed jumps at the
8000 end of a basic blocks with no more than max-goto-duplication-insns are
8001 unfactored. The default value is 8.
8003 @item max-delay-slot-insn-search
8004 The maximum number of instructions to consider when looking for an
8005 instruction to fill a delay slot. If more than this arbitrary number of
8006 instructions is searched, the time savings from filling the delay slot
8007 will be minimal so stop searching. Increasing values mean more
8008 aggressive optimization, making the compile time increase with probably
8009 small improvement in executable run time.
8011 @item max-delay-slot-live-search
8012 When trying to fill delay slots, the maximum number of instructions to
8013 consider when searching for a block with valid live register
8014 information. Increasing this arbitrarily chosen value means more
8015 aggressive optimization, increasing the compile time. This parameter
8016 should be removed when the delay slot code is rewritten to maintain the
8019 @item max-gcse-memory
8020 The approximate maximum amount of memory that will be allocated in
8021 order to perform the global common subexpression elimination
8022 optimization. If more memory than specified is required, the
8023 optimization will not be done.
8025 @item max-pending-list-length
8026 The maximum number of pending dependencies scheduling will allow
8027 before flushing the current state and starting over. Large functions
8028 with few branches or calls can create excessively large lists which
8029 needlessly consume memory and resources.
8031 @item max-inline-insns-single
8032 Several parameters control the tree inliner used in gcc.
8033 This number sets the maximum number of instructions (counted in GCC's
8034 internal representation) in a single function that the tree inliner
8035 will consider for inlining. This only affects functions declared
8036 inline and methods implemented in a class declaration (C++).
8037 The default value is 300.
8039 @item max-inline-insns-auto
8040 When you use @option{-finline-functions} (included in @option{-O3}),
8041 a lot of functions that would otherwise not be considered for inlining
8042 by the compiler will be investigated. To those functions, a different
8043 (more restrictive) limit compared to functions declared inline can
8045 The default value is 50.
8047 @item large-function-insns
8048 The limit specifying really large functions. For functions larger than this
8049 limit after inlining, inlining is constrained by
8050 @option{--param large-function-growth}. This parameter is useful primarily
8051 to avoid extreme compilation time caused by non-linear algorithms used by the
8053 The default value is 2700.
8055 @item large-function-growth
8056 Specifies maximal growth of large function caused by inlining in percents.
8057 The default value is 100 which limits large function growth to 2.0 times
8060 @item large-unit-insns
8061 The limit specifying large translation unit. Growth caused by inlining of
8062 units larger than this limit is limited by @option{--param inline-unit-growth}.
8063 For small units this might be too tight (consider unit consisting of function A
8064 that is inline and B that just calls A three time. If B is small relative to
8065 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8066 large units consisting of small inlineable functions however the overall unit
8067 growth limit is needed to avoid exponential explosion of code size. Thus for
8068 smaller units, the size is increased to @option{--param large-unit-insns}
8069 before applying @option{--param inline-unit-growth}. The default is 10000
8071 @item inline-unit-growth
8072 Specifies maximal overall growth of the compilation unit caused by inlining.
8073 The default value is 30 which limits unit growth to 1.3 times the original
8076 @item ipcp-unit-growth
8077 Specifies maximal overall growth of the compilation unit caused by
8078 interprocedural constant propagation. The default value is 10 which limits
8079 unit growth to 1.1 times the original size.
8081 @item large-stack-frame
8082 The limit specifying large stack frames. While inlining the algorithm is trying
8083 to not grow past this limit too much. Default value is 256 bytes.
8085 @item large-stack-frame-growth
8086 Specifies maximal growth of large stack frames caused by inlining in percents.
8087 The default value is 1000 which limits large stack frame growth to 11 times
8090 @item max-inline-insns-recursive
8091 @itemx max-inline-insns-recursive-auto
8092 Specifies maximum number of instructions out-of-line copy of self recursive inline
8093 function can grow into by performing recursive inlining.
8095 For functions declared inline @option{--param max-inline-insns-recursive} is
8096 taken into account. For function not declared inline, recursive inlining
8097 happens only when @option{-finline-functions} (included in @option{-O3}) is
8098 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8099 default value is 450.
8101 @item max-inline-recursive-depth
8102 @itemx max-inline-recursive-depth-auto
8103 Specifies maximum recursion depth used by the recursive inlining.
8105 For functions declared inline @option{--param max-inline-recursive-depth} is
8106 taken into account. For function not declared inline, recursive inlining
8107 happens only when @option{-finline-functions} (included in @option{-O3}) is
8108 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8111 @item min-inline-recursive-probability
8112 Recursive inlining is profitable only for function having deep recursion
8113 in average and can hurt for function having little recursion depth by
8114 increasing the prologue size or complexity of function body to other
8117 When profile feedback is available (see @option{-fprofile-generate}) the actual
8118 recursion depth can be guessed from probability that function will recurse via
8119 given call expression. This parameter limits inlining only to call expression
8120 whose probability exceeds given threshold (in percents). The default value is
8123 @item early-inlining-insns
8124 Specify growth that early inliner can make. In effect it increases amount of
8125 inlining for code having large abstraction penalty. The default value is 8.
8127 @item max-early-inliner-iterations
8128 @itemx max-early-inliner-iterations
8129 Limit of iterations of early inliner. This basically bounds number of nested
8130 indirect calls early inliner can resolve. Deeper chains are still handled by
8133 @item min-vect-loop-bound
8134 The minimum number of iterations under which a loop will not get vectorized
8135 when @option{-ftree-vectorize} is used. The number of iterations after
8136 vectorization needs to be greater than the value specified by this option
8137 to allow vectorization. The default value is 0.
8139 @item max-unrolled-insns
8140 The maximum number of instructions that a loop should have if that loop
8141 is unrolled, and if the loop is unrolled, it determines how many times
8142 the loop code is unrolled.
8144 @item max-average-unrolled-insns
8145 The maximum number of instructions biased by probabilities of their execution
8146 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8147 it determines how many times the loop code is unrolled.
8149 @item max-unroll-times
8150 The maximum number of unrollings of a single loop.
8152 @item max-peeled-insns
8153 The maximum number of instructions that a loop should have if that loop
8154 is peeled, and if the loop is peeled, it determines how many times
8155 the loop code is peeled.
8157 @item max-peel-times
8158 The maximum number of peelings of a single loop.
8160 @item max-completely-peeled-insns
8161 The maximum number of insns of a completely peeled loop.
8163 @item max-completely-peel-times
8164 The maximum number of iterations of a loop to be suitable for complete peeling.
8166 @item max-completely-peel-loop-nest-depth
8167 The maximum depth of a loop nest suitable for complete peeling.
8169 @item max-unswitch-insns
8170 The maximum number of insns of an unswitched loop.
8172 @item max-unswitch-level
8173 The maximum number of branches unswitched in a single loop.
8176 The minimum cost of an expensive expression in the loop invariant motion.
8178 @item iv-consider-all-candidates-bound
8179 Bound on number of candidates for induction variables below that
8180 all candidates are considered for each use in induction variable
8181 optimizations. Only the most relevant candidates are considered
8182 if there are more candidates, to avoid quadratic time complexity.
8184 @item iv-max-considered-uses
8185 The induction variable optimizations give up on loops that contain more
8186 induction variable uses.
8188 @item iv-always-prune-cand-set-bound
8189 If number of candidates in the set is smaller than this value,
8190 we always try to remove unnecessary ivs from the set during its
8191 optimization when a new iv is added to the set.
8193 @item scev-max-expr-size
8194 Bound on size of expressions used in the scalar evolutions analyzer.
8195 Large expressions slow the analyzer.
8197 @item omega-max-vars
8198 The maximum number of variables in an Omega constraint system.
8199 The default value is 128.
8201 @item omega-max-geqs
8202 The maximum number of inequalities in an Omega constraint system.
8203 The default value is 256.
8206 The maximum number of equalities in an Omega constraint system.
8207 The default value is 128.
8209 @item omega-max-wild-cards
8210 The maximum number of wildcard variables that the Omega solver will
8211 be able to insert. The default value is 18.
8213 @item omega-hash-table-size
8214 The size of the hash table in the Omega solver. The default value is
8217 @item omega-max-keys
8218 The maximal number of keys used by the Omega solver. The default
8221 @item omega-eliminate-redundant-constraints
8222 When set to 1, use expensive methods to eliminate all redundant
8223 constraints. The default value is 0.
8225 @item vect-max-version-for-alignment-checks
8226 The maximum number of runtime checks that can be performed when
8227 doing loop versioning for alignment in the vectorizer. See option
8228 ftree-vect-loop-version for more information.
8230 @item vect-max-version-for-alias-checks
8231 The maximum number of runtime checks that can be performed when
8232 doing loop versioning for alias in the vectorizer. See option
8233 ftree-vect-loop-version for more information.
8235 @item max-iterations-to-track
8237 The maximum number of iterations of a loop the brute force algorithm
8238 for analysis of # of iterations of the loop tries to evaluate.
8240 @item hot-bb-count-fraction
8241 Select fraction of the maximal count of repetitions of basic block in program
8242 given basic block needs to have to be considered hot.
8244 @item hot-bb-frequency-fraction
8245 Select fraction of the maximal frequency of executions of basic block in
8246 function given basic block needs to have to be considered hot
8248 @item max-predicted-iterations
8249 The maximum number of loop iterations we predict statically. This is useful
8250 in cases where function contain single loop with known bound and other loop
8251 with unknown. We predict the known number of iterations correctly, while
8252 the unknown number of iterations average to roughly 10. This means that the
8253 loop without bounds would appear artificially cold relative to the other one.
8255 @item align-threshold
8257 Select fraction of the maximal frequency of executions of basic block in
8258 function given basic block will get aligned.
8260 @item align-loop-iterations
8262 A loop expected to iterate at lest the selected number of iterations will get
8265 @item tracer-dynamic-coverage
8266 @itemx tracer-dynamic-coverage-feedback
8268 This value is used to limit superblock formation once the given percentage of
8269 executed instructions is covered. This limits unnecessary code size
8272 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8273 feedback is available. The real profiles (as opposed to statically estimated
8274 ones) are much less balanced allowing the threshold to be larger value.
8276 @item tracer-max-code-growth
8277 Stop tail duplication once code growth has reached given percentage. This is
8278 rather hokey argument, as most of the duplicates will be eliminated later in
8279 cross jumping, so it may be set to much higher values than is the desired code
8282 @item tracer-min-branch-ratio
8284 Stop reverse growth when the reverse probability of best edge is less than this
8285 threshold (in percent).
8287 @item tracer-min-branch-ratio
8288 @itemx tracer-min-branch-ratio-feedback
8290 Stop forward growth if the best edge do have probability lower than this
8293 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8294 compilation for profile feedback and one for compilation without. The value
8295 for compilation with profile feedback needs to be more conservative (higher) in
8296 order to make tracer effective.
8298 @item max-cse-path-length
8300 Maximum number of basic blocks on path that cse considers. The default is 10.
8303 The maximum instructions CSE process before flushing. The default is 1000.
8305 @item ggc-min-expand
8307 GCC uses a garbage collector to manage its own memory allocation. This
8308 parameter specifies the minimum percentage by which the garbage
8309 collector's heap should be allowed to expand between collections.
8310 Tuning this may improve compilation speed; it has no effect on code
8313 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8314 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8315 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8316 GCC is not able to calculate RAM on a particular platform, the lower
8317 bound of 30% is used. Setting this parameter and
8318 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8319 every opportunity. This is extremely slow, but can be useful for
8322 @item ggc-min-heapsize
8324 Minimum size of the garbage collector's heap before it begins bothering
8325 to collect garbage. The first collection occurs after the heap expands
8326 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8327 tuning this may improve compilation speed, and has no effect on code
8330 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8331 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8332 with a lower bound of 4096 (four megabytes) and an upper bound of
8333 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8334 particular platform, the lower bound is used. Setting this parameter
8335 very large effectively disables garbage collection. Setting this
8336 parameter and @option{ggc-min-expand} to zero causes a full collection
8337 to occur at every opportunity.
8339 @item max-reload-search-insns
8340 The maximum number of instruction reload should look backward for equivalent
8341 register. Increasing values mean more aggressive optimization, making the
8342 compile time increase with probably slightly better performance. The default
8345 @item max-cselib-memory-locations
8346 The maximum number of memory locations cselib should take into account.
8347 Increasing values mean more aggressive optimization, making the compile time
8348 increase with probably slightly better performance. The default value is 500.
8350 @item reorder-blocks-duplicate
8351 @itemx reorder-blocks-duplicate-feedback
8353 Used by basic block reordering pass to decide whether to use unconditional
8354 branch or duplicate the code on its destination. Code is duplicated when its
8355 estimated size is smaller than this value multiplied by the estimated size of
8356 unconditional jump in the hot spots of the program.
8358 The @option{reorder-block-duplicate-feedback} is used only when profile
8359 feedback is available and may be set to higher values than
8360 @option{reorder-block-duplicate} since information about the hot spots is more
8363 @item max-sched-ready-insns
8364 The maximum number of instructions ready to be issued the scheduler should
8365 consider at any given time during the first scheduling pass. Increasing
8366 values mean more thorough searches, making the compilation time increase
8367 with probably little benefit. The default value is 100.
8369 @item max-sched-region-blocks
8370 The maximum number of blocks in a region to be considered for
8371 interblock scheduling. The default value is 10.
8373 @item max-pipeline-region-blocks
8374 The maximum number of blocks in a region to be considered for
8375 pipelining in the selective scheduler. The default value is 15.
8377 @item max-sched-region-insns
8378 The maximum number of insns in a region to be considered for
8379 interblock scheduling. The default value is 100.
8381 @item max-pipeline-region-insns
8382 The maximum number of insns in a region to be considered for
8383 pipelining in the selective scheduler. The default value is 200.
8386 The minimum probability (in percents) of reaching a source block
8387 for interblock speculative scheduling. The default value is 40.
8389 @item max-sched-extend-regions-iters
8390 The maximum number of iterations through CFG to extend regions.
8391 0 - disable region extension,
8392 N - do at most N iterations.
8393 The default value is 0.
8395 @item max-sched-insn-conflict-delay
8396 The maximum conflict delay for an insn to be considered for speculative motion.
8397 The default value is 3.
8399 @item sched-spec-prob-cutoff
8400 The minimal probability of speculation success (in percents), so that
8401 speculative insn will be scheduled.
8402 The default value is 40.
8404 @item sched-mem-true-dep-cost
8405 Minimal distance (in CPU cycles) between store and load targeting same
8406 memory locations. The default value is 1.
8408 @item selsched-max-lookahead
8409 The maximum size of the lookahead window of selective scheduling. It is a
8410 depth of search for available instructions.
8411 The default value is 50.
8413 @item selsched-max-sched-times
8414 The maximum number of times that an instruction will be scheduled during
8415 selective scheduling. This is the limit on the number of iterations
8416 through which the instruction may be pipelined. The default value is 2.
8418 @item selsched-max-insns-to-rename
8419 The maximum number of best instructions in the ready list that are considered
8420 for renaming in the selective scheduler. The default value is 2.
8422 @item max-last-value-rtl
8423 The maximum size measured as number of RTLs that can be recorded in an expression
8424 in combiner for a pseudo register as last known value of that register. The default
8427 @item integer-share-limit
8428 Small integer constants can use a shared data structure, reducing the
8429 compiler's memory usage and increasing its speed. This sets the maximum
8430 value of a shared integer constant. The default value is 256.
8432 @item min-virtual-mappings
8433 Specifies the minimum number of virtual mappings in the incremental
8434 SSA updater that should be registered to trigger the virtual mappings
8435 heuristic defined by virtual-mappings-ratio. The default value is
8438 @item virtual-mappings-ratio
8439 If the number of virtual mappings is virtual-mappings-ratio bigger
8440 than the number of virtual symbols to be updated, then the incremental
8441 SSA updater switches to a full update for those symbols. The default
8444 @item ssp-buffer-size
8445 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8446 protection when @option{-fstack-protection} is used.
8448 @item max-jump-thread-duplication-stmts
8449 Maximum number of statements allowed in a block that needs to be
8450 duplicated when threading jumps.
8452 @item max-fields-for-field-sensitive
8453 Maximum number of fields in a structure we will treat in
8454 a field sensitive manner during pointer analysis. The default is zero
8455 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8457 @item prefetch-latency
8458 Estimate on average number of instructions that are executed before
8459 prefetch finishes. The distance we prefetch ahead is proportional
8460 to this constant. Increasing this number may also lead to less
8461 streams being prefetched (see @option{simultaneous-prefetches}).
8463 @item simultaneous-prefetches
8464 Maximum number of prefetches that can run at the same time.
8466 @item l1-cache-line-size
8467 The size of cache line in L1 cache, in bytes.
8470 The size of L1 cache, in kilobytes.
8473 The size of L2 cache, in kilobytes.
8475 @item min-insn-to-prefetch-ratio
8476 The minimum ratio between the number of instructions and the
8477 number of prefetches to enable prefetching in a loop with an
8480 @item prefetch-min-insn-to-mem-ratio
8481 The minimum ratio between the number of instructions and the
8482 number of memory references to enable prefetching in a loop.
8484 @item use-canonical-types
8485 Whether the compiler should use the ``canonical'' type system. By
8486 default, this should always be 1, which uses a more efficient internal
8487 mechanism for comparing types in C++ and Objective-C++. However, if
8488 bugs in the canonical type system are causing compilation failures,
8489 set this value to 0 to disable canonical types.
8491 @item switch-conversion-max-branch-ratio
8492 Switch initialization conversion will refuse to create arrays that are
8493 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8494 branches in the switch.
8496 @item max-partial-antic-length
8497 Maximum length of the partial antic set computed during the tree
8498 partial redundancy elimination optimization (@option{-ftree-pre}) when
8499 optimizing at @option{-O3} and above. For some sorts of source code
8500 the enhanced partial redundancy elimination optimization can run away,
8501 consuming all of the memory available on the host machine. This
8502 parameter sets a limit on the length of the sets that are computed,
8503 which prevents the runaway behavior. Setting a value of 0 for
8504 this parameter will allow an unlimited set length.
8506 @item sccvn-max-scc-size
8507 Maximum size of a strongly connected component (SCC) during SCCVN
8508 processing. If this limit is hit, SCCVN processing for the whole
8509 function will not be done and optimizations depending on it will
8510 be disabled. The default maximum SCC size is 10000.
8512 @item ira-max-loops-num
8513 IRA uses a regional register allocation by default. If a function
8514 contains loops more than number given by the parameter, only at most
8515 given number of the most frequently executed loops will form regions
8516 for the regional register allocation. The default value of the
8519 @item ira-max-conflict-table-size
8520 Although IRA uses a sophisticated algorithm of compression conflict
8521 table, the table can be still big for huge functions. If the conflict
8522 table for a function could be more than size in MB given by the
8523 parameter, the conflict table is not built and faster, simpler, and
8524 lower quality register allocation algorithm will be used. The
8525 algorithm do not use pseudo-register conflicts. The default value of
8526 the parameter is 2000.
8528 @item ira-loop-reserved-regs
8529 IRA can be used to evaluate more accurate register pressure in loops
8530 for decision to move loop invariants (see @option{-O3}). The number
8531 of available registers reserved for some other purposes is described
8532 by this parameter. The default value of the parameter is 2 which is
8533 minimal number of registers needed for execution of typical
8534 instruction. This value is the best found from numerous experiments.
8536 @item loop-invariant-max-bbs-in-loop
8537 Loop invariant motion can be very expensive, both in compile time and
8538 in amount of needed compile time memory, with very large loops. Loops
8539 with more basic blocks than this parameter won't have loop invariant
8540 motion optimization performed on them. The default value of the
8541 parameter is 1000 for -O1 and 10000 for -O2 and above.
8543 @item max-vartrack-size
8544 Sets a maximum number of hash table slots to use during variable
8545 tracking dataflow analysis of any function. If this limit is exceeded
8546 with variable tracking at assignments enabled, analysis for that
8547 function is retried without it, after removing all debug insns from
8548 the function. If the limit is exceeded even without debug insns, var
8549 tracking analysis is completely disabled for the function. Setting
8550 the parameter to zero makes it unlimited.
8552 @item min-nondebug-insn-uid
8553 Use uids starting at this parameter for nondebug insns. The range below
8554 the parameter is reserved exclusively for debug insns created by
8555 @option{-fvar-tracking-assignments}, but debug insns may get
8556 (non-overlapping) uids above it if the reserved range is exhausted.
8558 @item ipa-sra-ptr-growth-factor
8559 IPA-SRA will replace a pointer to an aggregate with one or more new
8560 parameters only when their cumulative size is less or equal to
8561 @option{ipa-sra-ptr-growth-factor} times the size of the original
8564 @item graphite-max-nb-scop-params
8565 To avoid exponential effects in the Graphite loop transforms, the
8566 number of parameters in a Static Control Part (SCoP) is bounded. The
8567 default value is 10 parameters. A variable whose value is unknown at
8568 compile time and defined outside a SCoP is a parameter of the SCoP.
8570 @item graphite-max-bbs-per-function
8571 To avoid exponential effects in the detection of SCoPs, the size of
8572 the functions analyzed by Graphite is bounded. The default value is
8575 @item loop-block-tile-size
8576 Loop blocking or strip mining transforms, enabled with
8577 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8578 loop in the loop nest by a given number of iterations. The strip
8579 length can be changed using the @option{loop-block-tile-size}
8580 parameter. The default value is 51 iterations.
8585 @node Preprocessor Options
8586 @section Options Controlling the Preprocessor
8587 @cindex preprocessor options
8588 @cindex options, preprocessor
8590 These options control the C preprocessor, which is run on each C source
8591 file before actual compilation.
8593 If you use the @option{-E} option, nothing is done except preprocessing.
8594 Some of these options make sense only together with @option{-E} because
8595 they cause the preprocessor output to be unsuitable for actual
8599 @item -Wp,@var{option}
8601 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8602 and pass @var{option} directly through to the preprocessor. If
8603 @var{option} contains commas, it is split into multiple options at the
8604 commas. However, many options are modified, translated or interpreted
8605 by the compiler driver before being passed to the preprocessor, and
8606 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8607 interface is undocumented and subject to change, so whenever possible
8608 you should avoid using @option{-Wp} and let the driver handle the
8611 @item -Xpreprocessor @var{option}
8612 @opindex Xpreprocessor
8613 Pass @var{option} as an option to the preprocessor. You can use this to
8614 supply system-specific preprocessor options which GCC does not know how to
8617 If you want to pass an option that takes an argument, you must use
8618 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8621 @include cppopts.texi
8623 @node Assembler Options
8624 @section Passing Options to the Assembler
8626 @c prevent bad page break with this line
8627 You can pass options to the assembler.
8630 @item -Wa,@var{option}
8632 Pass @var{option} as an option to the assembler. If @var{option}
8633 contains commas, it is split into multiple options at the commas.
8635 @item -Xassembler @var{option}
8637 Pass @var{option} as an option to the assembler. You can use this to
8638 supply system-specific assembler options which GCC does not know how to
8641 If you want to pass an option that takes an argument, you must use
8642 @option{-Xassembler} twice, once for the option and once for the argument.
8647 @section Options for Linking
8648 @cindex link options
8649 @cindex options, linking
8651 These options come into play when the compiler links object files into
8652 an executable output file. They are meaningless if the compiler is
8653 not doing a link step.
8657 @item @var{object-file-name}
8658 A file name that does not end in a special recognized suffix is
8659 considered to name an object file or library. (Object files are
8660 distinguished from libraries by the linker according to the file
8661 contents.) If linking is done, these object files are used as input
8670 If any of these options is used, then the linker is not run, and
8671 object file names should not be used as arguments. @xref{Overall
8675 @item -l@var{library}
8676 @itemx -l @var{library}
8678 Search the library named @var{library} when linking. (The second
8679 alternative with the library as a separate argument is only for
8680 POSIX compliance and is not recommended.)
8682 It makes a difference where in the command you write this option; the
8683 linker searches and processes libraries and object files in the order they
8684 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8685 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8686 to functions in @samp{z}, those functions may not be loaded.
8688 The linker searches a standard list of directories for the library,
8689 which is actually a file named @file{lib@var{library}.a}. The linker
8690 then uses this file as if it had been specified precisely by name.
8692 The directories searched include several standard system directories
8693 plus any that you specify with @option{-L}.
8695 Normally the files found this way are library files---archive files
8696 whose members are object files. The linker handles an archive file by
8697 scanning through it for members which define symbols that have so far
8698 been referenced but not defined. But if the file that is found is an
8699 ordinary object file, it is linked in the usual fashion. The only
8700 difference between using an @option{-l} option and specifying a file name
8701 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8702 and searches several directories.
8706 You need this special case of the @option{-l} option in order to
8707 link an Objective-C or Objective-C++ program.
8710 @opindex nostartfiles
8711 Do not use the standard system startup files when linking.
8712 The standard system libraries are used normally, unless @option{-nostdlib}
8713 or @option{-nodefaultlibs} is used.
8715 @item -nodefaultlibs
8716 @opindex nodefaultlibs
8717 Do not use the standard system libraries when linking.
8718 Only the libraries you specify will be passed to the linker, options
8719 specifying linkage of the system libraries, such as @code{-static-libgcc}
8720 or @code{-shared-libgcc}, will be ignored.
8721 The standard startup files are used normally, unless @option{-nostartfiles}
8722 is used. The compiler may generate calls to @code{memcmp},
8723 @code{memset}, @code{memcpy} and @code{memmove}.
8724 These entries are usually resolved by entries in
8725 libc. These entry points should be supplied through some other
8726 mechanism when this option is specified.
8730 Do not use the standard system startup files or libraries when linking.
8731 No startup files and only the libraries you specify will be passed to
8732 the linker, options specifying linkage of the system libraries, such as
8733 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8734 The compiler may generate calls to @code{memcmp}, @code{memset},
8735 @code{memcpy} and @code{memmove}.
8736 These entries are usually resolved by entries in
8737 libc. These entry points should be supplied through some other
8738 mechanism when this option is specified.
8740 @cindex @option{-lgcc}, use with @option{-nostdlib}
8741 @cindex @option{-nostdlib} and unresolved references
8742 @cindex unresolved references and @option{-nostdlib}
8743 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8744 @cindex @option{-nodefaultlibs} and unresolved references
8745 @cindex unresolved references and @option{-nodefaultlibs}
8746 One of the standard libraries bypassed by @option{-nostdlib} and
8747 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8748 that GCC uses to overcome shortcomings of particular machines, or special
8749 needs for some languages.
8750 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8751 Collection (GCC) Internals},
8752 for more discussion of @file{libgcc.a}.)
8753 In most cases, you need @file{libgcc.a} even when you want to avoid
8754 other standard libraries. In other words, when you specify @option{-nostdlib}
8755 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8756 This ensures that you have no unresolved references to internal GCC
8757 library subroutines. (For example, @samp{__main}, used to ensure C++
8758 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8759 GNU Compiler Collection (GCC) Internals}.)
8763 Produce a position independent executable on targets which support it.
8764 For predictable results, you must also specify the same set of options
8765 that were used to generate code (@option{-fpie}, @option{-fPIE},
8766 or model suboptions) when you specify this option.
8770 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8771 that support it. This instructs the linker to add all symbols, not
8772 only used ones, to the dynamic symbol table. This option is needed
8773 for some uses of @code{dlopen} or to allow obtaining backtraces
8774 from within a program.
8778 Remove all symbol table and relocation information from the executable.
8782 On systems that support dynamic linking, this prevents linking with the shared
8783 libraries. On other systems, this option has no effect.
8787 Produce a shared object which can then be linked with other objects to
8788 form an executable. Not all systems support this option. For predictable
8789 results, you must also specify the same set of options that were used to
8790 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8791 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8792 needs to build supplementary stub code for constructors to work. On
8793 multi-libbed systems, @samp{gcc -shared} must select the correct support
8794 libraries to link against. Failing to supply the correct flags may lead
8795 to subtle defects. Supplying them in cases where they are not necessary
8798 @item -shared-libgcc
8799 @itemx -static-libgcc
8800 @opindex shared-libgcc
8801 @opindex static-libgcc
8802 On systems that provide @file{libgcc} as a shared library, these options
8803 force the use of either the shared or static version respectively.
8804 If no shared version of @file{libgcc} was built when the compiler was
8805 configured, these options have no effect.
8807 There are several situations in which an application should use the
8808 shared @file{libgcc} instead of the static version. The most common
8809 of these is when the application wishes to throw and catch exceptions
8810 across different shared libraries. In that case, each of the libraries
8811 as well as the application itself should use the shared @file{libgcc}.
8813 Therefore, the G++ and GCJ drivers automatically add
8814 @option{-shared-libgcc} whenever you build a shared library or a main
8815 executable, because C++ and Java programs typically use exceptions, so
8816 this is the right thing to do.
8818 If, instead, you use the GCC driver to create shared libraries, you may
8819 find that they will not always be linked with the shared @file{libgcc}.
8820 If GCC finds, at its configuration time, that you have a non-GNU linker
8821 or a GNU linker that does not support option @option{--eh-frame-hdr},
8822 it will link the shared version of @file{libgcc} into shared libraries
8823 by default. Otherwise, it will take advantage of the linker and optimize
8824 away the linking with the shared version of @file{libgcc}, linking with
8825 the static version of libgcc by default. This allows exceptions to
8826 propagate through such shared libraries, without incurring relocation
8827 costs at library load time.
8829 However, if a library or main executable is supposed to throw or catch
8830 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8831 for the languages used in the program, or using the option
8832 @option{-shared-libgcc}, such that it is linked with the shared
8835 @item -static-libstdc++
8836 When the @command{g++} program is used to link a C++ program, it will
8837 normally automatically link against @option{libstdc++}. If
8838 @file{libstdc++} is available as a shared library, and the
8839 @option{-static} option is not used, then this will link against the
8840 shared version of @file{libstdc++}. That is normally fine. However, it
8841 is sometimes useful to freeze the version of @file{libstdc++} used by
8842 the program without going all the way to a fully static link. The
8843 @option{-static-libstdc++} option directs the @command{g++} driver to
8844 link @file{libstdc++} statically, without necessarily linking other
8845 libraries statically.
8849 Bind references to global symbols when building a shared object. Warn
8850 about any unresolved references (unless overridden by the link editor
8851 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8854 @item -T @var{script}
8856 @cindex linker script
8857 Use @var{script} as the linker script. This option is supported by most
8858 systems using the GNU linker. On some targets, such as bare-board
8859 targets without an operating system, the @option{-T} option may be required
8860 when linking to avoid references to undefined symbols.
8862 @item -Xlinker @var{option}
8864 Pass @var{option} as an option to the linker. You can use this to
8865 supply system-specific linker options which GCC does not know how to
8868 If you want to pass an option that takes a separate argument, you must use
8869 @option{-Xlinker} twice, once for the option and once for the argument.
8870 For example, to pass @option{-assert definitions}, you must write
8871 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8872 @option{-Xlinker "-assert definitions"}, because this passes the entire
8873 string as a single argument, which is not what the linker expects.
8875 When using the GNU linker, it is usually more convenient to pass
8876 arguments to linker options using the @option{@var{option}=@var{value}}
8877 syntax than as separate arguments. For example, you can specify
8878 @samp{-Xlinker -Map=output.map} rather than
8879 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8880 this syntax for command-line options.
8882 @item -Wl,@var{option}
8884 Pass @var{option} as an option to the linker. If @var{option} contains
8885 commas, it is split into multiple options at the commas. You can use this
8886 syntax to pass an argument to the option.
8887 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8888 linker. When using the GNU linker, you can also get the same effect with
8889 @samp{-Wl,-Map=output.map}.
8891 @item -u @var{symbol}
8893 Pretend the symbol @var{symbol} is undefined, to force linking of
8894 library modules to define it. You can use @option{-u} multiple times with
8895 different symbols to force loading of additional library modules.
8898 @node Directory Options
8899 @section Options for Directory Search
8900 @cindex directory options
8901 @cindex options, directory search
8904 These options specify directories to search for header files, for
8905 libraries and for parts of the compiler:
8910 Add the directory @var{dir} to the head of the list of directories to be
8911 searched for header files. This can be used to override a system header
8912 file, substituting your own version, since these directories are
8913 searched before the system header file directories. However, you should
8914 not use this option to add directories that contain vendor-supplied
8915 system header files (use @option{-isystem} for that). If you use more than
8916 one @option{-I} option, the directories are scanned in left-to-right
8917 order; the standard system directories come after.
8919 If a standard system include directory, or a directory specified with
8920 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8921 option will be ignored. The directory will still be searched but as a
8922 system directory at its normal position in the system include chain.
8923 This is to ensure that GCC's procedure to fix buggy system headers and
8924 the ordering for the include_next directive are not inadvertently changed.
8925 If you really need to change the search order for system directories,
8926 use the @option{-nostdinc} and/or @option{-isystem} options.
8928 @item -iplugindir=@var{dir}
8929 Set the directory to search for plugins which are passed
8930 by @option{-fplugin=@var{name}} instead of
8931 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
8932 to be used by the user, but only passed by the driver.
8934 @item -iquote@var{dir}
8936 Add the directory @var{dir} to the head of the list of directories to
8937 be searched for header files only for the case of @samp{#include
8938 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8939 otherwise just like @option{-I}.
8943 Add directory @var{dir} to the list of directories to be searched
8946 @item -B@var{prefix}
8948 This option specifies where to find the executables, libraries,
8949 include files, and data files of the compiler itself.
8951 The compiler driver program runs one or more of the subprograms
8952 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8953 @var{prefix} as a prefix for each program it tries to run, both with and
8954 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8956 For each subprogram to be run, the compiler driver first tries the
8957 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8958 was not specified, the driver tries two standard prefixes, which are
8959 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8960 those results in a file name that is found, the unmodified program
8961 name is searched for using the directories specified in your
8962 @env{PATH} environment variable.
8964 The compiler will check to see if the path provided by the @option{-B}
8965 refers to a directory, and if necessary it will add a directory
8966 separator character at the end of the path.
8968 @option{-B} prefixes that effectively specify directory names also apply
8969 to libraries in the linker, because the compiler translates these
8970 options into @option{-L} options for the linker. They also apply to
8971 includes files in the preprocessor, because the compiler translates these
8972 options into @option{-isystem} options for the preprocessor. In this case,
8973 the compiler appends @samp{include} to the prefix.
8975 The run-time support file @file{libgcc.a} can also be searched for using
8976 the @option{-B} prefix, if needed. If it is not found there, the two
8977 standard prefixes above are tried, and that is all. The file is left
8978 out of the link if it is not found by those means.
8980 Another way to specify a prefix much like the @option{-B} prefix is to use
8981 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8984 As a special kludge, if the path provided by @option{-B} is
8985 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8986 9, then it will be replaced by @file{[dir/]include}. This is to help
8987 with boot-strapping the compiler.
8989 @item -specs=@var{file}
8991 Process @var{file} after the compiler reads in the standard @file{specs}
8992 file, in order to override the defaults that the @file{gcc} driver
8993 program uses when determining what switches to pass to @file{cc1},
8994 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8995 @option{-specs=@var{file}} can be specified on the command line, and they
8996 are processed in order, from left to right.
8998 @item --sysroot=@var{dir}
9000 Use @var{dir} as the logical root directory for headers and libraries.
9001 For example, if the compiler would normally search for headers in
9002 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9003 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9005 If you use both this option and the @option{-isysroot} option, then
9006 the @option{--sysroot} option will apply to libraries, but the
9007 @option{-isysroot} option will apply to header files.
9009 The GNU linker (beginning with version 2.16) has the necessary support
9010 for this option. If your linker does not support this option, the
9011 header file aspect of @option{--sysroot} will still work, but the
9012 library aspect will not.
9016 This option has been deprecated. Please use @option{-iquote} instead for
9017 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9018 Any directories you specify with @option{-I} options before the @option{-I-}
9019 option are searched only for the case of @samp{#include "@var{file}"};
9020 they are not searched for @samp{#include <@var{file}>}.
9022 If additional directories are specified with @option{-I} options after
9023 the @option{-I-}, these directories are searched for all @samp{#include}
9024 directives. (Ordinarily @emph{all} @option{-I} directories are used
9027 In addition, the @option{-I-} option inhibits the use of the current
9028 directory (where the current input file came from) as the first search
9029 directory for @samp{#include "@var{file}"}. There is no way to
9030 override this effect of @option{-I-}. With @option{-I.} you can specify
9031 searching the directory which was current when the compiler was
9032 invoked. That is not exactly the same as what the preprocessor does
9033 by default, but it is often satisfactory.
9035 @option{-I-} does not inhibit the use of the standard system directories
9036 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9043 @section Specifying subprocesses and the switches to pass to them
9046 @command{gcc} is a driver program. It performs its job by invoking a
9047 sequence of other programs to do the work of compiling, assembling and
9048 linking. GCC interprets its command-line parameters and uses these to
9049 deduce which programs it should invoke, and which command-line options
9050 it ought to place on their command lines. This behavior is controlled
9051 by @dfn{spec strings}. In most cases there is one spec string for each
9052 program that GCC can invoke, but a few programs have multiple spec
9053 strings to control their behavior. The spec strings built into GCC can
9054 be overridden by using the @option{-specs=} command-line switch to specify
9057 @dfn{Spec files} are plaintext files that are used to construct spec
9058 strings. They consist of a sequence of directives separated by blank
9059 lines. The type of directive is determined by the first non-whitespace
9060 character on the line and it can be one of the following:
9063 @item %@var{command}
9064 Issues a @var{command} to the spec file processor. The commands that can
9068 @item %include <@var{file}>
9070 Search for @var{file} and insert its text at the current point in the
9073 @item %include_noerr <@var{file}>
9074 @cindex %include_noerr
9075 Just like @samp{%include}, but do not generate an error message if the include
9076 file cannot be found.
9078 @item %rename @var{old_name} @var{new_name}
9080 Rename the spec string @var{old_name} to @var{new_name}.
9084 @item *[@var{spec_name}]:
9085 This tells the compiler to create, override or delete the named spec
9086 string. All lines after this directive up to the next directive or
9087 blank line are considered to be the text for the spec string. If this
9088 results in an empty string then the spec will be deleted. (Or, if the
9089 spec did not exist, then nothing will happened.) Otherwise, if the spec
9090 does not currently exist a new spec will be created. If the spec does
9091 exist then its contents will be overridden by the text of this
9092 directive, unless the first character of that text is the @samp{+}
9093 character, in which case the text will be appended to the spec.
9095 @item [@var{suffix}]:
9096 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9097 and up to the next directive or blank line are considered to make up the
9098 spec string for the indicated suffix. When the compiler encounters an
9099 input file with the named suffix, it will processes the spec string in
9100 order to work out how to compile that file. For example:
9107 This says that any input file whose name ends in @samp{.ZZ} should be
9108 passed to the program @samp{z-compile}, which should be invoked with the
9109 command-line switch @option{-input} and with the result of performing the
9110 @samp{%i} substitution. (See below.)
9112 As an alternative to providing a spec string, the text that follows a
9113 suffix directive can be one of the following:
9116 @item @@@var{language}
9117 This says that the suffix is an alias for a known @var{language}. This is
9118 similar to using the @option{-x} command-line switch to GCC to specify a
9119 language explicitly. For example:
9126 Says that .ZZ files are, in fact, C++ source files.
9129 This causes an error messages saying:
9132 @var{name} compiler not installed on this system.
9136 GCC already has an extensive list of suffixes built into it.
9137 This directive will add an entry to the end of the list of suffixes, but
9138 since the list is searched from the end backwards, it is effectively
9139 possible to override earlier entries using this technique.
9143 GCC has the following spec strings built into it. Spec files can
9144 override these strings or create their own. Note that individual
9145 targets can also add their own spec strings to this list.
9148 asm Options to pass to the assembler
9149 asm_final Options to pass to the assembler post-processor
9150 cpp Options to pass to the C preprocessor
9151 cc1 Options to pass to the C compiler
9152 cc1plus Options to pass to the C++ compiler
9153 endfile Object files to include at the end of the link
9154 link Options to pass to the linker
9155 lib Libraries to include on the command line to the linker
9156 libgcc Decides which GCC support library to pass to the linker
9157 linker Sets the name of the linker
9158 predefines Defines to be passed to the C preprocessor
9159 signed_char Defines to pass to CPP to say whether @code{char} is signed
9161 startfile Object files to include at the start of the link
9164 Here is a small example of a spec file:
9170 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9173 This example renames the spec called @samp{lib} to @samp{old_lib} and
9174 then overrides the previous definition of @samp{lib} with a new one.
9175 The new definition adds in some extra command-line options before
9176 including the text of the old definition.
9178 @dfn{Spec strings} are a list of command-line options to be passed to their
9179 corresponding program. In addition, the spec strings can contain
9180 @samp{%}-prefixed sequences to substitute variable text or to
9181 conditionally insert text into the command line. Using these constructs
9182 it is possible to generate quite complex command lines.
9184 Here is a table of all defined @samp{%}-sequences for spec
9185 strings. Note that spaces are not generated automatically around the
9186 results of expanding these sequences. Therefore you can concatenate them
9187 together or combine them with constant text in a single argument.
9191 Substitute one @samp{%} into the program name or argument.
9194 Substitute the name of the input file being processed.
9197 Substitute the basename of the input file being processed.
9198 This is the substring up to (and not including) the last period
9199 and not including the directory.
9202 This is the same as @samp{%b}, but include the file suffix (text after
9206 Marks the argument containing or following the @samp{%d} as a
9207 temporary file name, so that that file will be deleted if GCC exits
9208 successfully. Unlike @samp{%g}, this contributes no text to the
9211 @item %g@var{suffix}
9212 Substitute a file name that has suffix @var{suffix} and is chosen
9213 once per compilation, and mark the argument in the same way as
9214 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9215 name is now chosen in a way that is hard to predict even when previously
9216 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9217 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9218 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9219 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9220 was simply substituted with a file name chosen once per compilation,
9221 without regard to any appended suffix (which was therefore treated
9222 just like ordinary text), making such attacks more likely to succeed.
9224 @item %u@var{suffix}
9225 Like @samp{%g}, but generates a new temporary file name even if
9226 @samp{%u@var{suffix}} was already seen.
9228 @item %U@var{suffix}
9229 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9230 new one if there is no such last file name. In the absence of any
9231 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9232 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9233 would involve the generation of two distinct file names, one
9234 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9235 simply substituted with a file name chosen for the previous @samp{%u},
9236 without regard to any appended suffix.
9238 @item %j@var{suffix}
9239 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9240 writable, and if save-temps is off; otherwise, substitute the name
9241 of a temporary file, just like @samp{%u}. This temporary file is not
9242 meant for communication between processes, but rather as a junk
9245 @item %|@var{suffix}
9246 @itemx %m@var{suffix}
9247 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9248 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9249 all. These are the two most common ways to instruct a program that it
9250 should read from standard input or write to standard output. If you
9251 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9252 construct: see for example @file{f/lang-specs.h}.
9254 @item %.@var{SUFFIX}
9255 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9256 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9257 terminated by the next space or %.
9260 Marks the argument containing or following the @samp{%w} as the
9261 designated output file of this compilation. This puts the argument
9262 into the sequence of arguments that @samp{%o} will substitute later.
9265 Substitutes the names of all the output files, with spaces
9266 automatically placed around them. You should write spaces
9267 around the @samp{%o} as well or the results are undefined.
9268 @samp{%o} is for use in the specs for running the linker.
9269 Input files whose names have no recognized suffix are not compiled
9270 at all, but they are included among the output files, so they will
9274 Substitutes the suffix for object files. Note that this is
9275 handled specially when it immediately follows @samp{%g, %u, or %U},
9276 because of the need for those to form complete file names. The
9277 handling is such that @samp{%O} is treated exactly as if it had already
9278 been substituted, except that @samp{%g, %u, and %U} do not currently
9279 support additional @var{suffix} characters following @samp{%O} as they would
9280 following, for example, @samp{.o}.
9283 Substitutes the standard macro predefinitions for the
9284 current target machine. Use this when running @code{cpp}.
9287 Like @samp{%p}, but puts @samp{__} before and after the name of each
9288 predefined macro, except for macros that start with @samp{__} or with
9289 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9293 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9294 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9295 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9296 and @option{-imultilib} as necessary.
9299 Current argument is the name of a library or startup file of some sort.
9300 Search for that file in a standard list of directories and substitute
9301 the full name found. The current working directory is included in the
9302 list of directories scanned.
9305 Current argument is the name of a linker script. Search for that file
9306 in the current list of directories to scan for libraries. If the file
9307 is located insert a @option{--script} option into the command line
9308 followed by the full path name found. If the file is not found then
9309 generate an error message. Note: the current working directory is not
9313 Print @var{str} as an error message. @var{str} is terminated by a newline.
9314 Use this when inconsistent options are detected.
9317 Substitute the contents of spec string @var{name} at this point.
9320 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9322 @item %x@{@var{option}@}
9323 Accumulate an option for @samp{%X}.
9326 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9330 Output the accumulated assembler options specified by @option{-Wa}.
9333 Output the accumulated preprocessor options specified by @option{-Wp}.
9336 Process the @code{asm} spec. This is used to compute the
9337 switches to be passed to the assembler.
9340 Process the @code{asm_final} spec. This is a spec string for
9341 passing switches to an assembler post-processor, if such a program is
9345 Process the @code{link} spec. This is the spec for computing the
9346 command line passed to the linker. Typically it will make use of the
9347 @samp{%L %G %S %D and %E} sequences.
9350 Dump out a @option{-L} option for each directory that GCC believes might
9351 contain startup files. If the target supports multilibs then the
9352 current multilib directory will be prepended to each of these paths.
9355 Process the @code{lib} spec. This is a spec string for deciding which
9356 libraries should be included on the command line to the linker.
9359 Process the @code{libgcc} spec. This is a spec string for deciding
9360 which GCC support library should be included on the command line to the linker.
9363 Process the @code{startfile} spec. This is a spec for deciding which
9364 object files should be the first ones passed to the linker. Typically
9365 this might be a file named @file{crt0.o}.
9368 Process the @code{endfile} spec. This is a spec string that specifies
9369 the last object files that will be passed to the linker.
9372 Process the @code{cpp} spec. This is used to construct the arguments
9373 to be passed to the C preprocessor.
9376 Process the @code{cc1} spec. This is used to construct the options to be
9377 passed to the actual C compiler (@samp{cc1}).
9380 Process the @code{cc1plus} spec. This is used to construct the options to be
9381 passed to the actual C++ compiler (@samp{cc1plus}).
9384 Substitute the variable part of a matched option. See below.
9385 Note that each comma in the substituted string is replaced by
9389 Remove all occurrences of @code{-S} from the command line. Note---this
9390 command is position dependent. @samp{%} commands in the spec string
9391 before this one will see @code{-S}, @samp{%} commands in the spec string
9392 after this one will not.
9394 @item %:@var{function}(@var{args})
9395 Call the named function @var{function}, passing it @var{args}.
9396 @var{args} is first processed as a nested spec string, then split
9397 into an argument vector in the usual fashion. The function returns
9398 a string which is processed as if it had appeared literally as part
9399 of the current spec.
9401 The following built-in spec functions are provided:
9405 The @code{getenv} spec function takes two arguments: an environment
9406 variable name and a string. If the environment variable is not
9407 defined, a fatal error is issued. Otherwise, the return value is the
9408 value of the environment variable concatenated with the string. For
9409 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9412 %:getenv(TOPDIR /include)
9415 expands to @file{/path/to/top/include}.
9417 @item @code{if-exists}
9418 The @code{if-exists} spec function takes one argument, an absolute
9419 pathname to a file. If the file exists, @code{if-exists} returns the
9420 pathname. Here is a small example of its usage:
9424 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9427 @item @code{if-exists-else}
9428 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9429 spec function, except that it takes two arguments. The first argument is
9430 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9431 returns the pathname. If it does not exist, it returns the second argument.
9432 This way, @code{if-exists-else} can be used to select one file or another,
9433 based on the existence of the first. Here is a small example of its usage:
9437 crt0%O%s %:if-exists(crti%O%s) \
9438 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9441 @item @code{replace-outfile}
9442 The @code{replace-outfile} spec function takes two arguments. It looks for the
9443 first argument in the outfiles array and replaces it with the second argument. Here
9444 is a small example of its usage:
9447 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9450 @item @code{print-asm-header}
9451 The @code{print-asm-header} function takes no arguments and simply
9452 prints a banner like:
9458 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9461 It is used to separate compiler options from assembler options
9462 in the @option{--target-help} output.
9466 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9467 If that switch was not specified, this substitutes nothing. Note that
9468 the leading dash is omitted when specifying this option, and it is
9469 automatically inserted if the substitution is performed. Thus the spec
9470 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9471 and would output the command line option @option{-foo}.
9473 @item %W@{@code{S}@}
9474 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9477 @item %@{@code{S}*@}
9478 Substitutes all the switches specified to GCC whose names start
9479 with @code{-S}, but which also take an argument. This is used for
9480 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9481 GCC considers @option{-o foo} as being
9482 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9483 text, including the space. Thus two arguments would be generated.
9485 @item %@{@code{S}*&@code{T}*@}
9486 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9487 (the order of @code{S} and @code{T} in the spec is not significant).
9488 There can be any number of ampersand-separated variables; for each the
9489 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9491 @item %@{@code{S}:@code{X}@}
9492 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9494 @item %@{!@code{S}:@code{X}@}
9495 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9497 @item %@{@code{S}*:@code{X}@}
9498 Substitutes @code{X} if one or more switches whose names start with
9499 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9500 once, no matter how many such switches appeared. However, if @code{%*}
9501 appears somewhere in @code{X}, then @code{X} will be substituted once
9502 for each matching switch, with the @code{%*} replaced by the part of
9503 that switch that matched the @code{*}.
9505 @item %@{.@code{S}:@code{X}@}
9506 Substitutes @code{X}, if processing a file with suffix @code{S}.
9508 @item %@{!.@code{S}:@code{X}@}
9509 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9511 @item %@{,@code{S}:@code{X}@}
9512 Substitutes @code{X}, if processing a file for language @code{S}.
9514 @item %@{!,@code{S}:@code{X}@}
9515 Substitutes @code{X}, if not processing a file for language @code{S}.
9517 @item %@{@code{S}|@code{P}:@code{X}@}
9518 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9519 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9520 @code{*} sequences as well, although they have a stronger binding than
9521 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9522 alternatives must be starred, and only the first matching alternative
9525 For example, a spec string like this:
9528 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9531 will output the following command-line options from the following input
9532 command-line options:
9537 -d fred.c -foo -baz -boggle
9538 -d jim.d -bar -baz -boggle
9541 @item %@{S:X; T:Y; :D@}
9543 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9544 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9545 be as many clauses as you need. This may be combined with @code{.},
9546 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9551 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9552 construct may contain other nested @samp{%} constructs or spaces, or
9553 even newlines. They are processed as usual, as described above.
9554 Trailing white space in @code{X} is ignored. White space may also
9555 appear anywhere on the left side of the colon in these constructs,
9556 except between @code{.} or @code{*} and the corresponding word.
9558 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9559 handled specifically in these constructs. If another value of
9560 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9561 @option{-W} switch is found later in the command line, the earlier
9562 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9563 just one letter, which passes all matching options.
9565 The character @samp{|} at the beginning of the predicate text is used to
9566 indicate that a command should be piped to the following command, but
9567 only if @option{-pipe} is specified.
9569 It is built into GCC which switches take arguments and which do not.
9570 (You might think it would be useful to generalize this to allow each
9571 compiler's spec to say which switches take arguments. But this cannot
9572 be done in a consistent fashion. GCC cannot even decide which input
9573 files have been specified without knowing which switches take arguments,
9574 and it must know which input files to compile in order to tell which
9577 GCC also knows implicitly that arguments starting in @option{-l} are to be
9578 treated as compiler output files, and passed to the linker in their
9579 proper position among the other output files.
9581 @c man begin OPTIONS
9583 @node Target Options
9584 @section Specifying Target Machine and Compiler Version
9585 @cindex target options
9586 @cindex cross compiling
9587 @cindex specifying machine version
9588 @cindex specifying compiler version and target machine
9589 @cindex compiler version, specifying
9590 @cindex target machine, specifying
9592 The usual way to run GCC is to run the executable called @file{gcc}, or
9593 @file{<machine>-gcc} when cross-compiling, or
9594 @file{<machine>-gcc-<version>} to run a version other than the one that
9597 @node Submodel Options
9598 @section Hardware Models and Configurations
9599 @cindex submodel options
9600 @cindex specifying hardware config
9601 @cindex hardware models and configurations, specifying
9602 @cindex machine dependent options
9604 Each target machine types can have its own
9605 special options, starting with @samp{-m}, to choose among various
9606 hardware models or configurations---for example, 68010 vs 68020,
9607 floating coprocessor or none. A single installed version of the
9608 compiler can compile for any model or configuration, according to the
9611 Some configurations of the compiler also support additional special
9612 options, usually for compatibility with other compilers on the same
9615 @c This list is ordered alphanumerically by subsection name.
9616 @c It should be the same order and spelling as these options are listed
9617 @c in Machine Dependent Options
9623 * Blackfin Options::
9627 * DEC Alpha Options::
9628 * DEC Alpha/VMS Options::
9631 * GNU/Linux Options::
9634 * i386 and x86-64 Options::
9635 * i386 and x86-64 Windows Options::
9637 * IA-64/VMS Options::
9649 * picoChip Options::
9651 * RS/6000 and PowerPC Options::
9653 * S/390 and zSeries Options::
9658 * System V Options::
9663 * Xstormy16 Options::
9669 @subsection ARC Options
9672 These options are defined for ARC implementations:
9677 Compile code for little endian mode. This is the default.
9681 Compile code for big endian mode.
9684 @opindex mmangle-cpu
9685 Prepend the name of the cpu to all public symbol names.
9686 In multiple-processor systems, there are many ARC variants with different
9687 instruction and register set characteristics. This flag prevents code
9688 compiled for one cpu to be linked with code compiled for another.
9689 No facility exists for handling variants that are ``almost identical''.
9690 This is an all or nothing option.
9692 @item -mcpu=@var{cpu}
9694 Compile code for ARC variant @var{cpu}.
9695 Which variants are supported depend on the configuration.
9696 All variants support @option{-mcpu=base}, this is the default.
9698 @item -mtext=@var{text-section}
9699 @itemx -mdata=@var{data-section}
9700 @itemx -mrodata=@var{readonly-data-section}
9704 Put functions, data, and readonly data in @var{text-section},
9705 @var{data-section}, and @var{readonly-data-section} respectively
9706 by default. This can be overridden with the @code{section} attribute.
9707 @xref{Variable Attributes}.
9709 @item -mfix-cortex-m3-ldrd
9710 @opindex mfix-cortex-m3-ldrd
9711 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9712 with overlapping destination and base registers are used. This option avoids
9713 generating these instructions. This option is enabled by default when
9714 @option{-mcpu=cortex-m3} is specified.
9719 @subsection ARM Options
9722 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9726 @item -mabi=@var{name}
9728 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9729 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9732 @opindex mapcs-frame
9733 Generate a stack frame that is compliant with the ARM Procedure Call
9734 Standard for all functions, even if this is not strictly necessary for
9735 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9736 with this option will cause the stack frames not to be generated for
9737 leaf functions. The default is @option{-mno-apcs-frame}.
9741 This is a synonym for @option{-mapcs-frame}.
9744 @c not currently implemented
9745 @item -mapcs-stack-check
9746 @opindex mapcs-stack-check
9747 Generate code to check the amount of stack space available upon entry to
9748 every function (that actually uses some stack space). If there is
9749 insufficient space available then either the function
9750 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9751 called, depending upon the amount of stack space required. The run time
9752 system is required to provide these functions. The default is
9753 @option{-mno-apcs-stack-check}, since this produces smaller code.
9755 @c not currently implemented
9757 @opindex mapcs-float
9758 Pass floating point arguments using the float point registers. This is
9759 one of the variants of the APCS@. This option is recommended if the
9760 target hardware has a floating point unit or if a lot of floating point
9761 arithmetic is going to be performed by the code. The default is
9762 @option{-mno-apcs-float}, since integer only code is slightly increased in
9763 size if @option{-mapcs-float} is used.
9765 @c not currently implemented
9766 @item -mapcs-reentrant
9767 @opindex mapcs-reentrant
9768 Generate reentrant, position independent code. The default is
9769 @option{-mno-apcs-reentrant}.
9772 @item -mthumb-interwork
9773 @opindex mthumb-interwork
9774 Generate code which supports calling between the ARM and Thumb
9775 instruction sets. Without this option the two instruction sets cannot
9776 be reliably used inside one program. The default is
9777 @option{-mno-thumb-interwork}, since slightly larger code is generated
9778 when @option{-mthumb-interwork} is specified.
9780 @item -mno-sched-prolog
9781 @opindex mno-sched-prolog
9782 Prevent the reordering of instructions in the function prolog, or the
9783 merging of those instruction with the instructions in the function's
9784 body. This means that all functions will start with a recognizable set
9785 of instructions (or in fact one of a choice from a small set of
9786 different function prologues), and this information can be used to
9787 locate the start if functions inside an executable piece of code. The
9788 default is @option{-msched-prolog}.
9790 @item -mfloat-abi=@var{name}
9792 Specifies which floating-point ABI to use. Permissible values
9793 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9795 Specifying @samp{soft} causes GCC to generate output containing
9796 library calls for floating-point operations.
9797 @samp{softfp} allows the generation of code using hardware floating-point
9798 instructions, but still uses the soft-float calling conventions.
9799 @samp{hard} allows generation of floating-point instructions
9800 and uses FPU-specific calling conventions.
9802 The default depends on the specific target configuration. Note that
9803 the hard-float and soft-float ABIs are not link-compatible; you must
9804 compile your entire program with the same ABI, and link with a
9805 compatible set of libraries.
9808 @opindex mhard-float
9809 Equivalent to @option{-mfloat-abi=hard}.
9812 @opindex msoft-float
9813 Equivalent to @option{-mfloat-abi=soft}.
9815 @item -mlittle-endian
9816 @opindex mlittle-endian
9817 Generate code for a processor running in little-endian mode. This is
9818 the default for all standard configurations.
9821 @opindex mbig-endian
9822 Generate code for a processor running in big-endian mode; the default is
9823 to compile code for a little-endian processor.
9825 @item -mwords-little-endian
9826 @opindex mwords-little-endian
9827 This option only applies when generating code for big-endian processors.
9828 Generate code for a little-endian word order but a big-endian byte
9829 order. That is, a byte order of the form @samp{32107654}. Note: this
9830 option should only be used if you require compatibility with code for
9831 big-endian ARM processors generated by versions of the compiler prior to
9834 @item -mcpu=@var{name}
9836 This specifies the name of the target ARM processor. GCC uses this name
9837 to determine what kind of instructions it can emit when generating
9838 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9839 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9840 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9841 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9842 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9844 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9845 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9846 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9847 @samp{strongarm1110},
9848 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9849 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9850 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9851 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9852 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9853 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9854 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9855 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9},
9856 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9859 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9861 @item -mtune=@var{name}
9863 This option is very similar to the @option{-mcpu=} option, except that
9864 instead of specifying the actual target processor type, and hence
9865 restricting which instructions can be used, it specifies that GCC should
9866 tune the performance of the code as if the target were of the type
9867 specified in this option, but still choosing the instructions that it
9868 will generate based on the cpu specified by a @option{-mcpu=} option.
9869 For some ARM implementations better performance can be obtained by using
9872 @item -march=@var{name}
9874 This specifies the name of the target ARM architecture. GCC uses this
9875 name to determine what kind of instructions it can emit when generating
9876 assembly code. This option can be used in conjunction with or instead
9877 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9878 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9879 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9880 @samp{armv6}, @samp{armv6j},
9881 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9882 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9883 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9885 @item -mfpu=@var{name}
9886 @itemx -mfpe=@var{number}
9887 @itemx -mfp=@var{number}
9891 This specifies what floating point hardware (or hardware emulation) is
9892 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9893 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
9894 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
9895 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
9896 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
9897 @option{-mfp} and @option{-mfpe} are synonyms for
9898 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
9901 If @option{-msoft-float} is specified this specifies the format of
9902 floating point values.
9904 @item -mfp16-format=@var{name}
9905 @opindex mfp16-format
9906 Specify the format of the @code{__fp16} half-precision floating-point type.
9907 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9908 the default is @samp{none}, in which case the @code{__fp16} type is not
9909 defined. @xref{Half-Precision}, for more information.
9911 @item -mstructure-size-boundary=@var{n}
9912 @opindex mstructure-size-boundary
9913 The size of all structures and unions will be rounded up to a multiple
9914 of the number of bits set by this option. Permissible values are 8, 32
9915 and 64. The default value varies for different toolchains. For the COFF
9916 targeted toolchain the default value is 8. A value of 64 is only allowed
9917 if the underlying ABI supports it.
9919 Specifying the larger number can produce faster, more efficient code, but
9920 can also increase the size of the program. Different values are potentially
9921 incompatible. Code compiled with one value cannot necessarily expect to
9922 work with code or libraries compiled with another value, if they exchange
9923 information using structures or unions.
9925 @item -mabort-on-noreturn
9926 @opindex mabort-on-noreturn
9927 Generate a call to the function @code{abort} at the end of a
9928 @code{noreturn} function. It will be executed if the function tries to
9932 @itemx -mno-long-calls
9933 @opindex mlong-calls
9934 @opindex mno-long-calls
9935 Tells the compiler to perform function calls by first loading the
9936 address of the function into a register and then performing a subroutine
9937 call on this register. This switch is needed if the target function
9938 will lie outside of the 64 megabyte addressing range of the offset based
9939 version of subroutine call instruction.
9941 Even if this switch is enabled, not all function calls will be turned
9942 into long calls. The heuristic is that static functions, functions
9943 which have the @samp{short-call} attribute, functions that are inside
9944 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9945 definitions have already been compiled within the current compilation
9946 unit, will not be turned into long calls. The exception to this rule is
9947 that weak function definitions, functions with the @samp{long-call}
9948 attribute or the @samp{section} attribute, and functions that are within
9949 the scope of a @samp{#pragma long_calls} directive, will always be
9950 turned into long calls.
9952 This feature is not enabled by default. Specifying
9953 @option{-mno-long-calls} will restore the default behavior, as will
9954 placing the function calls within the scope of a @samp{#pragma
9955 long_calls_off} directive. Note these switches have no effect on how
9956 the compiler generates code to handle function calls via function
9959 @item -msingle-pic-base
9960 @opindex msingle-pic-base
9961 Treat the register used for PIC addressing as read-only, rather than
9962 loading it in the prologue for each function. The run-time system is
9963 responsible for initializing this register with an appropriate value
9964 before execution begins.
9966 @item -mpic-register=@var{reg}
9967 @opindex mpic-register
9968 Specify the register to be used for PIC addressing. The default is R10
9969 unless stack-checking is enabled, when R9 is used.
9971 @item -mcirrus-fix-invalid-insns
9972 @opindex mcirrus-fix-invalid-insns
9973 @opindex mno-cirrus-fix-invalid-insns
9974 Insert NOPs into the instruction stream to in order to work around
9975 problems with invalid Maverick instruction combinations. This option
9976 is only valid if the @option{-mcpu=ep9312} option has been used to
9977 enable generation of instructions for the Cirrus Maverick floating
9978 point co-processor. This option is not enabled by default, since the
9979 problem is only present in older Maverick implementations. The default
9980 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9983 @item -mpoke-function-name
9984 @opindex mpoke-function-name
9985 Write the name of each function into the text section, directly
9986 preceding the function prologue. The generated code is similar to this:
9990 .ascii "arm_poke_function_name", 0
9993 .word 0xff000000 + (t1 - t0)
9994 arm_poke_function_name
9996 stmfd sp!, @{fp, ip, lr, pc@}
10000 When performing a stack backtrace, code can inspect the value of
10001 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10002 location @code{pc - 12} and the top 8 bits are set, then we know that
10003 there is a function name embedded immediately preceding this location
10004 and has length @code{((pc[-3]) & 0xff000000)}.
10008 Generate code for the Thumb instruction set. The default is to
10009 use the 32-bit ARM instruction set.
10010 This option automatically enables either 16-bit Thumb-1 or
10011 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10012 and @option{-march=@var{name}} options. This option is not passed to the
10013 assembler. If you want to force assembler files to be interpreted as Thumb code,
10014 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10015 option directly to the assembler by prefixing it with @option{-Wa}.
10018 @opindex mtpcs-frame
10019 Generate a stack frame that is compliant with the Thumb Procedure Call
10020 Standard for all non-leaf functions. (A leaf function is one that does
10021 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10023 @item -mtpcs-leaf-frame
10024 @opindex mtpcs-leaf-frame
10025 Generate a stack frame that is compliant with the Thumb Procedure Call
10026 Standard for all leaf functions. (A leaf function is one that does
10027 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10029 @item -mcallee-super-interworking
10030 @opindex mcallee-super-interworking
10031 Gives all externally visible functions in the file being compiled an ARM
10032 instruction set header which switches to Thumb mode before executing the
10033 rest of the function. This allows these functions to be called from
10034 non-interworking code. This option is not valid in AAPCS configurations
10035 because interworking is enabled by default.
10037 @item -mcaller-super-interworking
10038 @opindex mcaller-super-interworking
10039 Allows calls via function pointers (including virtual functions) to
10040 execute correctly regardless of whether the target code has been
10041 compiled for interworking or not. There is a small overhead in the cost
10042 of executing a function pointer if this option is enabled. This option
10043 is not valid in AAPCS configurations because interworking is enabled
10046 @item -mtp=@var{name}
10048 Specify the access model for the thread local storage pointer. The valid
10049 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10050 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10051 (supported in the arm6k architecture), and @option{auto}, which uses the
10052 best available method for the selected processor. The default setting is
10055 @item -mword-relocations
10056 @opindex mword-relocations
10057 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10058 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10059 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10065 @subsection AVR Options
10066 @cindex AVR Options
10068 These options are defined for AVR implementations:
10071 @item -mmcu=@var{mcu}
10073 Specify ATMEL AVR instruction set or MCU type.
10075 Instruction set avr1 is for the minimal AVR core, not supported by the C
10076 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10077 attiny11, attiny12, attiny15, attiny28).
10079 Instruction set avr2 (default) is for the classic AVR core with up to
10080 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10081 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10082 at90c8534, at90s8535).
10084 Instruction set avr3 is for the classic AVR core with up to 128K program
10085 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10087 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10088 memory space (MCU types: atmega8, atmega83, atmega85).
10090 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10091 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10092 atmega64, atmega128, at43usb355, at94k).
10094 @item -mno-interrupts
10095 @opindex mno-interrupts
10096 Generated code is not compatible with hardware interrupts.
10097 Code size will be smaller.
10099 @item -mcall-prologues
10100 @opindex mcall-prologues
10101 Functions prologues/epilogues expanded as call to appropriate
10102 subroutines. Code size will be smaller.
10105 @opindex mtiny-stack
10106 Change only the low 8 bits of the stack pointer.
10110 Assume int to be 8 bit integer. This affects the sizes of all types: A
10111 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10112 and long long will be 4 bytes. Please note that this option does not
10113 comply to the C standards, but it will provide you with smaller code
10117 @node Blackfin Options
10118 @subsection Blackfin Options
10119 @cindex Blackfin Options
10122 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10124 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10125 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10126 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10127 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10128 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10129 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10130 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10132 The optional @var{sirevision} specifies the silicon revision of the target
10133 Blackfin processor. Any workarounds available for the targeted silicon revision
10134 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10135 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10136 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10137 hexadecimal digits representing the major and minor numbers in the silicon
10138 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10139 is not defined. If @var{sirevision} is @samp{any}, the
10140 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10141 If this optional @var{sirevision} is not used, GCC assumes the latest known
10142 silicon revision of the targeted Blackfin processor.
10144 Support for @samp{bf561} is incomplete. For @samp{bf561},
10145 Only the processor macro is defined.
10146 Without this option, @samp{bf532} is used as the processor by default.
10147 The corresponding predefined processor macros for @var{cpu} is to
10148 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10149 provided by libgloss to be linked in if @option{-msim} is not given.
10153 Specifies that the program will be run on the simulator. This causes
10154 the simulator BSP provided by libgloss to be linked in. This option
10155 has effect only for @samp{bfin-elf} toolchain.
10156 Certain other options, such as @option{-mid-shared-library} and
10157 @option{-mfdpic}, imply @option{-msim}.
10159 @item -momit-leaf-frame-pointer
10160 @opindex momit-leaf-frame-pointer
10161 Don't keep the frame pointer in a register for leaf functions. This
10162 avoids the instructions to save, set up and restore frame pointers and
10163 makes an extra register available in leaf functions. The option
10164 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10165 which might make debugging harder.
10167 @item -mspecld-anomaly
10168 @opindex mspecld-anomaly
10169 When enabled, the compiler will ensure that the generated code does not
10170 contain speculative loads after jump instructions. If this option is used,
10171 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10173 @item -mno-specld-anomaly
10174 @opindex mno-specld-anomaly
10175 Don't generate extra code to prevent speculative loads from occurring.
10177 @item -mcsync-anomaly
10178 @opindex mcsync-anomaly
10179 When enabled, the compiler will ensure that the generated code does not
10180 contain CSYNC or SSYNC instructions too soon after conditional branches.
10181 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10183 @item -mno-csync-anomaly
10184 @opindex mno-csync-anomaly
10185 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10186 occurring too soon after a conditional branch.
10190 When enabled, the compiler is free to take advantage of the knowledge that
10191 the entire program fits into the low 64k of memory.
10194 @opindex mno-low-64k
10195 Assume that the program is arbitrarily large. This is the default.
10197 @item -mstack-check-l1
10198 @opindex mstack-check-l1
10199 Do stack checking using information placed into L1 scratchpad memory by the
10202 @item -mid-shared-library
10203 @opindex mid-shared-library
10204 Generate code that supports shared libraries via the library ID method.
10205 This allows for execute in place and shared libraries in an environment
10206 without virtual memory management. This option implies @option{-fPIC}.
10207 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10209 @item -mno-id-shared-library
10210 @opindex mno-id-shared-library
10211 Generate code that doesn't assume ID based shared libraries are being used.
10212 This is the default.
10214 @item -mleaf-id-shared-library
10215 @opindex mleaf-id-shared-library
10216 Generate code that supports shared libraries via the library ID method,
10217 but assumes that this library or executable won't link against any other
10218 ID shared libraries. That allows the compiler to use faster code for jumps
10221 @item -mno-leaf-id-shared-library
10222 @opindex mno-leaf-id-shared-library
10223 Do not assume that the code being compiled won't link against any ID shared
10224 libraries. Slower code will be generated for jump and call insns.
10226 @item -mshared-library-id=n
10227 @opindex mshared-library-id
10228 Specified the identification number of the ID based shared library being
10229 compiled. Specifying a value of 0 will generate more compact code, specifying
10230 other values will force the allocation of that number to the current
10231 library but is no more space or time efficient than omitting this option.
10235 Generate code that allows the data segment to be located in a different
10236 area of memory from the text segment. This allows for execute in place in
10237 an environment without virtual memory management by eliminating relocations
10238 against the text section.
10240 @item -mno-sep-data
10241 @opindex mno-sep-data
10242 Generate code that assumes that the data segment follows the text segment.
10243 This is the default.
10246 @itemx -mno-long-calls
10247 @opindex mlong-calls
10248 @opindex mno-long-calls
10249 Tells the compiler to perform function calls by first loading the
10250 address of the function into a register and then performing a subroutine
10251 call on this register. This switch is needed if the target function
10252 will lie outside of the 24 bit addressing range of the offset based
10253 version of subroutine call instruction.
10255 This feature is not enabled by default. Specifying
10256 @option{-mno-long-calls} will restore the default behavior. Note these
10257 switches have no effect on how the compiler generates code to handle
10258 function calls via function pointers.
10262 Link with the fast floating-point library. This library relaxes some of
10263 the IEEE floating-point standard's rules for checking inputs against
10264 Not-a-Number (NAN), in the interest of performance.
10267 @opindex minline-plt
10268 Enable inlining of PLT entries in function calls to functions that are
10269 not known to bind locally. It has no effect without @option{-mfdpic}.
10272 @opindex mmulticore
10273 Build standalone application for multicore Blackfin processor. Proper
10274 start files and link scripts will be used to support multicore.
10275 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10276 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10277 @option{-mcorea} or @option{-mcoreb}. If it's used without
10278 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10279 programming model is used. In this model, the main function of Core B
10280 should be named as coreb_main. If it's used with @option{-mcorea} or
10281 @option{-mcoreb}, one application per core programming model is used.
10282 If this option is not used, single core application programming
10287 Build standalone application for Core A of BF561 when using
10288 one application per core programming model. Proper start files
10289 and link scripts will be used to support Core A. This option
10290 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10294 Build standalone application for Core B of BF561 when using
10295 one application per core programming model. Proper start files
10296 and link scripts will be used to support Core B. This option
10297 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10298 should be used instead of main. It must be used with
10299 @option{-mmulticore}.
10303 Build standalone application for SDRAM. Proper start files and
10304 link scripts will be used to put the application into SDRAM.
10305 Loader should initialize SDRAM before loading the application
10306 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10310 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10311 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10312 are enabled; for standalone applications the default is off.
10316 @subsection CRIS Options
10317 @cindex CRIS Options
10319 These options are defined specifically for the CRIS ports.
10322 @item -march=@var{architecture-type}
10323 @itemx -mcpu=@var{architecture-type}
10326 Generate code for the specified architecture. The choices for
10327 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10328 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10329 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10332 @item -mtune=@var{architecture-type}
10334 Tune to @var{architecture-type} everything applicable about the generated
10335 code, except for the ABI and the set of available instructions. The
10336 choices for @var{architecture-type} are the same as for
10337 @option{-march=@var{architecture-type}}.
10339 @item -mmax-stack-frame=@var{n}
10340 @opindex mmax-stack-frame
10341 Warn when the stack frame of a function exceeds @var{n} bytes.
10347 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10348 @option{-march=v3} and @option{-march=v8} respectively.
10350 @item -mmul-bug-workaround
10351 @itemx -mno-mul-bug-workaround
10352 @opindex mmul-bug-workaround
10353 @opindex mno-mul-bug-workaround
10354 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10355 models where it applies. This option is active by default.
10359 Enable CRIS-specific verbose debug-related information in the assembly
10360 code. This option also has the effect to turn off the @samp{#NO_APP}
10361 formatted-code indicator to the assembler at the beginning of the
10366 Do not use condition-code results from previous instruction; always emit
10367 compare and test instructions before use of condition codes.
10369 @item -mno-side-effects
10370 @opindex mno-side-effects
10371 Do not emit instructions with side-effects in addressing modes other than
10374 @item -mstack-align
10375 @itemx -mno-stack-align
10376 @itemx -mdata-align
10377 @itemx -mno-data-align
10378 @itemx -mconst-align
10379 @itemx -mno-const-align
10380 @opindex mstack-align
10381 @opindex mno-stack-align
10382 @opindex mdata-align
10383 @opindex mno-data-align
10384 @opindex mconst-align
10385 @opindex mno-const-align
10386 These options (no-options) arranges (eliminate arrangements) for the
10387 stack-frame, individual data and constants to be aligned for the maximum
10388 single data access size for the chosen CPU model. The default is to
10389 arrange for 32-bit alignment. ABI details such as structure layout are
10390 not affected by these options.
10398 Similar to the stack- data- and const-align options above, these options
10399 arrange for stack-frame, writable data and constants to all be 32-bit,
10400 16-bit or 8-bit aligned. The default is 32-bit alignment.
10402 @item -mno-prologue-epilogue
10403 @itemx -mprologue-epilogue
10404 @opindex mno-prologue-epilogue
10405 @opindex mprologue-epilogue
10406 With @option{-mno-prologue-epilogue}, the normal function prologue and
10407 epilogue that sets up the stack-frame are omitted and no return
10408 instructions or return sequences are generated in the code. Use this
10409 option only together with visual inspection of the compiled code: no
10410 warnings or errors are generated when call-saved registers must be saved,
10411 or storage for local variable needs to be allocated.
10415 @opindex mno-gotplt
10417 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10418 instruction sequences that load addresses for functions from the PLT part
10419 of the GOT rather than (traditional on other architectures) calls to the
10420 PLT@. The default is @option{-mgotplt}.
10424 Legacy no-op option only recognized with the cris-axis-elf and
10425 cris-axis-linux-gnu targets.
10429 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10433 This option, recognized for the cris-axis-elf arranges
10434 to link with input-output functions from a simulator library. Code,
10435 initialized data and zero-initialized data are allocated consecutively.
10439 Like @option{-sim}, but pass linker options to locate initialized data at
10440 0x40000000 and zero-initialized data at 0x80000000.
10444 @subsection CRX Options
10445 @cindex CRX Options
10447 These options are defined specifically for the CRX ports.
10453 Enable the use of multiply-accumulate instructions. Disabled by default.
10456 @opindex mpush-args
10457 Push instructions will be used to pass outgoing arguments when functions
10458 are called. Enabled by default.
10461 @node Darwin Options
10462 @subsection Darwin Options
10463 @cindex Darwin options
10465 These options are defined for all architectures running the Darwin operating
10468 FSF GCC on Darwin does not create ``fat'' object files; it will create
10469 an object file for the single architecture that it was built to
10470 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10471 @option{-arch} options are used; it does so by running the compiler or
10472 linker multiple times and joining the results together with
10475 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10476 @samp{i686}) is determined by the flags that specify the ISA
10477 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10478 @option{-force_cpusubtype_ALL} option can be used to override this.
10480 The Darwin tools vary in their behavior when presented with an ISA
10481 mismatch. The assembler, @file{as}, will only permit instructions to
10482 be used that are valid for the subtype of the file it is generating,
10483 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10484 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10485 and print an error if asked to create a shared library with a less
10486 restrictive subtype than its input files (for instance, trying to put
10487 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10488 for executables, @file{ld}, will quietly give the executable the most
10489 restrictive subtype of any of its input files.
10494 Add the framework directory @var{dir} to the head of the list of
10495 directories to be searched for header files. These directories are
10496 interleaved with those specified by @option{-I} options and are
10497 scanned in a left-to-right order.
10499 A framework directory is a directory with frameworks in it. A
10500 framework is a directory with a @samp{"Headers"} and/or
10501 @samp{"PrivateHeaders"} directory contained directly in it that ends
10502 in @samp{".framework"}. The name of a framework is the name of this
10503 directory excluding the @samp{".framework"}. Headers associated with
10504 the framework are found in one of those two directories, with
10505 @samp{"Headers"} being searched first. A subframework is a framework
10506 directory that is in a framework's @samp{"Frameworks"} directory.
10507 Includes of subframework headers can only appear in a header of a
10508 framework that contains the subframework, or in a sibling subframework
10509 header. Two subframeworks are siblings if they occur in the same
10510 framework. A subframework should not have the same name as a
10511 framework, a warning will be issued if this is violated. Currently a
10512 subframework cannot have subframeworks, in the future, the mechanism
10513 may be extended to support this. The standard frameworks can be found
10514 in @samp{"/System/Library/Frameworks"} and
10515 @samp{"/Library/Frameworks"}. An example include looks like
10516 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10517 the name of the framework and header.h is found in the
10518 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10520 @item -iframework@var{dir}
10521 @opindex iframework
10522 Like @option{-F} except the directory is a treated as a system
10523 directory. The main difference between this @option{-iframework} and
10524 @option{-F} is that with @option{-iframework} the compiler does not
10525 warn about constructs contained within header files found via
10526 @var{dir}. This option is valid only for the C family of languages.
10530 Emit debugging information for symbols that are used. For STABS
10531 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10532 This is by default ON@.
10536 Emit debugging information for all symbols and types.
10538 @item -mmacosx-version-min=@var{version}
10539 The earliest version of MacOS X that this executable will run on
10540 is @var{version}. Typical values of @var{version} include @code{10.1},
10541 @code{10.2}, and @code{10.3.9}.
10543 If the compiler was built to use the system's headers by default,
10544 then the default for this option is the system version on which the
10545 compiler is running, otherwise the default is to make choices which
10546 are compatible with as many systems and code bases as possible.
10550 Enable kernel development mode. The @option{-mkernel} option sets
10551 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10552 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10553 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10554 applicable. This mode also sets @option{-mno-altivec},
10555 @option{-msoft-float}, @option{-fno-builtin} and
10556 @option{-mlong-branch} for PowerPC targets.
10558 @item -mone-byte-bool
10559 @opindex mone-byte-bool
10560 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10561 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10562 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10563 option has no effect on x86.
10565 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10566 to generate code that is not binary compatible with code generated
10567 without that switch. Using this switch may require recompiling all
10568 other modules in a program, including system libraries. Use this
10569 switch to conform to a non-default data model.
10571 @item -mfix-and-continue
10572 @itemx -ffix-and-continue
10573 @itemx -findirect-data
10574 @opindex mfix-and-continue
10575 @opindex ffix-and-continue
10576 @opindex findirect-data
10577 Generate code suitable for fast turn around development. Needed to
10578 enable gdb to dynamically load @code{.o} files into already running
10579 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10580 are provided for backwards compatibility.
10584 Loads all members of static archive libraries.
10585 See man ld(1) for more information.
10587 @item -arch_errors_fatal
10588 @opindex arch_errors_fatal
10589 Cause the errors having to do with files that have the wrong architecture
10592 @item -bind_at_load
10593 @opindex bind_at_load
10594 Causes the output file to be marked such that the dynamic linker will
10595 bind all undefined references when the file is loaded or launched.
10599 Produce a Mach-o bundle format file.
10600 See man ld(1) for more information.
10602 @item -bundle_loader @var{executable}
10603 @opindex bundle_loader
10604 This option specifies the @var{executable} that will be loading the build
10605 output file being linked. See man ld(1) for more information.
10608 @opindex dynamiclib
10609 When passed this option, GCC will produce a dynamic library instead of
10610 an executable when linking, using the Darwin @file{libtool} command.
10612 @item -force_cpusubtype_ALL
10613 @opindex force_cpusubtype_ALL
10614 This causes GCC's output file to have the @var{ALL} subtype, instead of
10615 one controlled by the @option{-mcpu} or @option{-march} option.
10617 @item -allowable_client @var{client_name}
10618 @itemx -client_name
10619 @itemx -compatibility_version
10620 @itemx -current_version
10622 @itemx -dependency-file
10624 @itemx -dylinker_install_name
10626 @itemx -exported_symbols_list
10628 @itemx -flat_namespace
10629 @itemx -force_flat_namespace
10630 @itemx -headerpad_max_install_names
10633 @itemx -install_name
10634 @itemx -keep_private_externs
10635 @itemx -multi_module
10636 @itemx -multiply_defined
10637 @itemx -multiply_defined_unused
10639 @itemx -no_dead_strip_inits_and_terms
10640 @itemx -nofixprebinding
10641 @itemx -nomultidefs
10643 @itemx -noseglinkedit
10644 @itemx -pagezero_size
10646 @itemx -prebind_all_twolevel_modules
10647 @itemx -private_bundle
10648 @itemx -read_only_relocs
10650 @itemx -sectobjectsymbols
10654 @itemx -sectobjectsymbols
10657 @itemx -segs_read_only_addr
10658 @itemx -segs_read_write_addr
10659 @itemx -seg_addr_table
10660 @itemx -seg_addr_table_filename
10661 @itemx -seglinkedit
10663 @itemx -segs_read_only_addr
10664 @itemx -segs_read_write_addr
10665 @itemx -single_module
10667 @itemx -sub_library
10668 @itemx -sub_umbrella
10669 @itemx -twolevel_namespace
10672 @itemx -unexported_symbols_list
10673 @itemx -weak_reference_mismatches
10674 @itemx -whatsloaded
10675 @opindex allowable_client
10676 @opindex client_name
10677 @opindex compatibility_version
10678 @opindex current_version
10679 @opindex dead_strip
10680 @opindex dependency-file
10681 @opindex dylib_file
10682 @opindex dylinker_install_name
10684 @opindex exported_symbols_list
10686 @opindex flat_namespace
10687 @opindex force_flat_namespace
10688 @opindex headerpad_max_install_names
10689 @opindex image_base
10691 @opindex install_name
10692 @opindex keep_private_externs
10693 @opindex multi_module
10694 @opindex multiply_defined
10695 @opindex multiply_defined_unused
10696 @opindex noall_load
10697 @opindex no_dead_strip_inits_and_terms
10698 @opindex nofixprebinding
10699 @opindex nomultidefs
10701 @opindex noseglinkedit
10702 @opindex pagezero_size
10704 @opindex prebind_all_twolevel_modules
10705 @opindex private_bundle
10706 @opindex read_only_relocs
10708 @opindex sectobjectsymbols
10711 @opindex sectcreate
10712 @opindex sectobjectsymbols
10715 @opindex segs_read_only_addr
10716 @opindex segs_read_write_addr
10717 @opindex seg_addr_table
10718 @opindex seg_addr_table_filename
10719 @opindex seglinkedit
10721 @opindex segs_read_only_addr
10722 @opindex segs_read_write_addr
10723 @opindex single_module
10725 @opindex sub_library
10726 @opindex sub_umbrella
10727 @opindex twolevel_namespace
10730 @opindex unexported_symbols_list
10731 @opindex weak_reference_mismatches
10732 @opindex whatsloaded
10733 These options are passed to the Darwin linker. The Darwin linker man page
10734 describes them in detail.
10737 @node DEC Alpha Options
10738 @subsection DEC Alpha Options
10740 These @samp{-m} options are defined for the DEC Alpha implementations:
10743 @item -mno-soft-float
10744 @itemx -msoft-float
10745 @opindex mno-soft-float
10746 @opindex msoft-float
10747 Use (do not use) the hardware floating-point instructions for
10748 floating-point operations. When @option{-msoft-float} is specified,
10749 functions in @file{libgcc.a} will be used to perform floating-point
10750 operations. Unless they are replaced by routines that emulate the
10751 floating-point operations, or compiled in such a way as to call such
10752 emulations routines, these routines will issue floating-point
10753 operations. If you are compiling for an Alpha without floating-point
10754 operations, you must ensure that the library is built so as not to call
10757 Note that Alpha implementations without floating-point operations are
10758 required to have floating-point registers.
10761 @itemx -mno-fp-regs
10763 @opindex mno-fp-regs
10764 Generate code that uses (does not use) the floating-point register set.
10765 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10766 register set is not used, floating point operands are passed in integer
10767 registers as if they were integers and floating-point results are passed
10768 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10769 so any function with a floating-point argument or return value called by code
10770 compiled with @option{-mno-fp-regs} must also be compiled with that
10773 A typical use of this option is building a kernel that does not use,
10774 and hence need not save and restore, any floating-point registers.
10778 The Alpha architecture implements floating-point hardware optimized for
10779 maximum performance. It is mostly compliant with the IEEE floating
10780 point standard. However, for full compliance, software assistance is
10781 required. This option generates code fully IEEE compliant code
10782 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10783 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10784 defined during compilation. The resulting code is less efficient but is
10785 able to correctly support denormalized numbers and exceptional IEEE
10786 values such as not-a-number and plus/minus infinity. Other Alpha
10787 compilers call this option @option{-ieee_with_no_inexact}.
10789 @item -mieee-with-inexact
10790 @opindex mieee-with-inexact
10791 This is like @option{-mieee} except the generated code also maintains
10792 the IEEE @var{inexact-flag}. Turning on this option causes the
10793 generated code to implement fully-compliant IEEE math. In addition to
10794 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10795 macro. On some Alpha implementations the resulting code may execute
10796 significantly slower than the code generated by default. Since there is
10797 very little code that depends on the @var{inexact-flag}, you should
10798 normally not specify this option. Other Alpha compilers call this
10799 option @option{-ieee_with_inexact}.
10801 @item -mfp-trap-mode=@var{trap-mode}
10802 @opindex mfp-trap-mode
10803 This option controls what floating-point related traps are enabled.
10804 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10805 The trap mode can be set to one of four values:
10809 This is the default (normal) setting. The only traps that are enabled
10810 are the ones that cannot be disabled in software (e.g., division by zero
10814 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10818 Like @samp{u}, but the instructions are marked to be safe for software
10819 completion (see Alpha architecture manual for details).
10822 Like @samp{su}, but inexact traps are enabled as well.
10825 @item -mfp-rounding-mode=@var{rounding-mode}
10826 @opindex mfp-rounding-mode
10827 Selects the IEEE rounding mode. Other Alpha compilers call this option
10828 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10833 Normal IEEE rounding mode. Floating point numbers are rounded towards
10834 the nearest machine number or towards the even machine number in case
10838 Round towards minus infinity.
10841 Chopped rounding mode. Floating point numbers are rounded towards zero.
10844 Dynamic rounding mode. A field in the floating point control register
10845 (@var{fpcr}, see Alpha architecture reference manual) controls the
10846 rounding mode in effect. The C library initializes this register for
10847 rounding towards plus infinity. Thus, unless your program modifies the
10848 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10851 @item -mtrap-precision=@var{trap-precision}
10852 @opindex mtrap-precision
10853 In the Alpha architecture, floating point traps are imprecise. This
10854 means without software assistance it is impossible to recover from a
10855 floating trap and program execution normally needs to be terminated.
10856 GCC can generate code that can assist operating system trap handlers
10857 in determining the exact location that caused a floating point trap.
10858 Depending on the requirements of an application, different levels of
10859 precisions can be selected:
10863 Program precision. This option is the default and means a trap handler
10864 can only identify which program caused a floating point exception.
10867 Function precision. The trap handler can determine the function that
10868 caused a floating point exception.
10871 Instruction precision. The trap handler can determine the exact
10872 instruction that caused a floating point exception.
10875 Other Alpha compilers provide the equivalent options called
10876 @option{-scope_safe} and @option{-resumption_safe}.
10878 @item -mieee-conformant
10879 @opindex mieee-conformant
10880 This option marks the generated code as IEEE conformant. You must not
10881 use this option unless you also specify @option{-mtrap-precision=i} and either
10882 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10883 is to emit the line @samp{.eflag 48} in the function prologue of the
10884 generated assembly file. Under DEC Unix, this has the effect that
10885 IEEE-conformant math library routines will be linked in.
10887 @item -mbuild-constants
10888 @opindex mbuild-constants
10889 Normally GCC examines a 32- or 64-bit integer constant to
10890 see if it can construct it from smaller constants in two or three
10891 instructions. If it cannot, it will output the constant as a literal and
10892 generate code to load it from the data segment at runtime.
10894 Use this option to require GCC to construct @emph{all} integer constants
10895 using code, even if it takes more instructions (the maximum is six).
10897 You would typically use this option to build a shared library dynamic
10898 loader. Itself a shared library, it must relocate itself in memory
10899 before it can find the variables and constants in its own data segment.
10905 Select whether to generate code to be assembled by the vendor-supplied
10906 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10924 Indicate whether GCC should generate code to use the optional BWX,
10925 CIX, FIX and MAX instruction sets. The default is to use the instruction
10926 sets supported by the CPU type specified via @option{-mcpu=} option or that
10927 of the CPU on which GCC was built if none was specified.
10930 @itemx -mfloat-ieee
10931 @opindex mfloat-vax
10932 @opindex mfloat-ieee
10933 Generate code that uses (does not use) VAX F and G floating point
10934 arithmetic instead of IEEE single and double precision.
10936 @item -mexplicit-relocs
10937 @itemx -mno-explicit-relocs
10938 @opindex mexplicit-relocs
10939 @opindex mno-explicit-relocs
10940 Older Alpha assemblers provided no way to generate symbol relocations
10941 except via assembler macros. Use of these macros does not allow
10942 optimal instruction scheduling. GNU binutils as of version 2.12
10943 supports a new syntax that allows the compiler to explicitly mark
10944 which relocations should apply to which instructions. This option
10945 is mostly useful for debugging, as GCC detects the capabilities of
10946 the assembler when it is built and sets the default accordingly.
10949 @itemx -mlarge-data
10950 @opindex msmall-data
10951 @opindex mlarge-data
10952 When @option{-mexplicit-relocs} is in effect, static data is
10953 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10954 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10955 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10956 16-bit relocations off of the @code{$gp} register. This limits the
10957 size of the small data area to 64KB, but allows the variables to be
10958 directly accessed via a single instruction.
10960 The default is @option{-mlarge-data}. With this option the data area
10961 is limited to just below 2GB@. Programs that require more than 2GB of
10962 data must use @code{malloc} or @code{mmap} to allocate the data in the
10963 heap instead of in the program's data segment.
10965 When generating code for shared libraries, @option{-fpic} implies
10966 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10969 @itemx -mlarge-text
10970 @opindex msmall-text
10971 @opindex mlarge-text
10972 When @option{-msmall-text} is used, the compiler assumes that the
10973 code of the entire program (or shared library) fits in 4MB, and is
10974 thus reachable with a branch instruction. When @option{-msmall-data}
10975 is used, the compiler can assume that all local symbols share the
10976 same @code{$gp} value, and thus reduce the number of instructions
10977 required for a function call from 4 to 1.
10979 The default is @option{-mlarge-text}.
10981 @item -mcpu=@var{cpu_type}
10983 Set the instruction set and instruction scheduling parameters for
10984 machine type @var{cpu_type}. You can specify either the @samp{EV}
10985 style name or the corresponding chip number. GCC supports scheduling
10986 parameters for the EV4, EV5 and EV6 family of processors and will
10987 choose the default values for the instruction set from the processor
10988 you specify. If you do not specify a processor type, GCC will default
10989 to the processor on which the compiler was built.
10991 Supported values for @var{cpu_type} are
10997 Schedules as an EV4 and has no instruction set extensions.
11001 Schedules as an EV5 and has no instruction set extensions.
11005 Schedules as an EV5 and supports the BWX extension.
11010 Schedules as an EV5 and supports the BWX and MAX extensions.
11014 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11018 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11021 Native Linux/GNU toolchains also support the value @samp{native},
11022 which selects the best architecture option for the host processor.
11023 @option{-mcpu=native} has no effect if GCC does not recognize
11026 @item -mtune=@var{cpu_type}
11028 Set only the instruction scheduling parameters for machine type
11029 @var{cpu_type}. The instruction set is not changed.
11031 Native Linux/GNU toolchains also support the value @samp{native},
11032 which selects the best architecture option for the host processor.
11033 @option{-mtune=native} has no effect if GCC does not recognize
11036 @item -mmemory-latency=@var{time}
11037 @opindex mmemory-latency
11038 Sets the latency the scheduler should assume for typical memory
11039 references as seen by the application. This number is highly
11040 dependent on the memory access patterns used by the application
11041 and the size of the external cache on the machine.
11043 Valid options for @var{time} are
11047 A decimal number representing clock cycles.
11053 The compiler contains estimates of the number of clock cycles for
11054 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11055 (also called Dcache, Scache, and Bcache), as well as to main memory.
11056 Note that L3 is only valid for EV5.
11061 @node DEC Alpha/VMS Options
11062 @subsection DEC Alpha/VMS Options
11064 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11067 @item -mvms-return-codes
11068 @opindex mvms-return-codes
11069 Return VMS condition codes from main. The default is to return POSIX
11070 style condition (e.g.@: error) codes.
11072 @item -mdebug-main=@var{prefix}
11073 @opindex mdebug-main=@var{prefix}
11074 Flag the first routine whose name starts with @var{prefix} as the main
11075 routine for the debugger.
11079 Default to 64bit memory allocation routines.
11083 @subsection FR30 Options
11084 @cindex FR30 Options
11086 These options are defined specifically for the FR30 port.
11090 @item -msmall-model
11091 @opindex msmall-model
11092 Use the small address space model. This can produce smaller code, but
11093 it does assume that all symbolic values and addresses will fit into a
11098 Assume that run-time support has been provided and so there is no need
11099 to include the simulator library (@file{libsim.a}) on the linker
11105 @subsection FRV Options
11106 @cindex FRV Options
11112 Only use the first 32 general purpose registers.
11117 Use all 64 general purpose registers.
11122 Use only the first 32 floating point registers.
11127 Use all 64 floating point registers
11130 @opindex mhard-float
11132 Use hardware instructions for floating point operations.
11135 @opindex msoft-float
11137 Use library routines for floating point operations.
11142 Dynamically allocate condition code registers.
11147 Do not try to dynamically allocate condition code registers, only
11148 use @code{icc0} and @code{fcc0}.
11153 Change ABI to use double word insns.
11158 Do not use double word instructions.
11163 Use floating point double instructions.
11166 @opindex mno-double
11168 Do not use floating point double instructions.
11173 Use media instructions.
11178 Do not use media instructions.
11183 Use multiply and add/subtract instructions.
11186 @opindex mno-muladd
11188 Do not use multiply and add/subtract instructions.
11193 Select the FDPIC ABI, that uses function descriptors to represent
11194 pointers to functions. Without any PIC/PIE-related options, it
11195 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11196 assumes GOT entries and small data are within a 12-bit range from the
11197 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11198 are computed with 32 bits.
11199 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11202 @opindex minline-plt
11204 Enable inlining of PLT entries in function calls to functions that are
11205 not known to bind locally. It has no effect without @option{-mfdpic}.
11206 It's enabled by default if optimizing for speed and compiling for
11207 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11208 optimization option such as @option{-O3} or above is present in the
11214 Assume a large TLS segment when generating thread-local code.
11219 Do not assume a large TLS segment when generating thread-local code.
11224 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11225 that is known to be in read-only sections. It's enabled by default,
11226 except for @option{-fpic} or @option{-fpie}: even though it may help
11227 make the global offset table smaller, it trades 1 instruction for 4.
11228 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11229 one of which may be shared by multiple symbols, and it avoids the need
11230 for a GOT entry for the referenced symbol, so it's more likely to be a
11231 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11233 @item -multilib-library-pic
11234 @opindex multilib-library-pic
11236 Link with the (library, not FD) pic libraries. It's implied by
11237 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11238 @option{-fpic} without @option{-mfdpic}. You should never have to use
11242 @opindex mlinked-fp
11244 Follow the EABI requirement of always creating a frame pointer whenever
11245 a stack frame is allocated. This option is enabled by default and can
11246 be disabled with @option{-mno-linked-fp}.
11249 @opindex mlong-calls
11251 Use indirect addressing to call functions outside the current
11252 compilation unit. This allows the functions to be placed anywhere
11253 within the 32-bit address space.
11255 @item -malign-labels
11256 @opindex malign-labels
11258 Try to align labels to an 8-byte boundary by inserting nops into the
11259 previous packet. This option only has an effect when VLIW packing
11260 is enabled. It doesn't create new packets; it merely adds nops to
11263 @item -mlibrary-pic
11264 @opindex mlibrary-pic
11266 Generate position-independent EABI code.
11271 Use only the first four media accumulator registers.
11276 Use all eight media accumulator registers.
11281 Pack VLIW instructions.
11286 Do not pack VLIW instructions.
11289 @opindex mno-eflags
11291 Do not mark ABI switches in e_flags.
11294 @opindex mcond-move
11296 Enable the use of conditional-move instructions (default).
11298 This switch is mainly for debugging the compiler and will likely be removed
11299 in a future version.
11301 @item -mno-cond-move
11302 @opindex mno-cond-move
11304 Disable the use of conditional-move instructions.
11306 This switch is mainly for debugging the compiler and will likely be removed
11307 in a future version.
11312 Enable the use of conditional set instructions (default).
11314 This switch is mainly for debugging the compiler and will likely be removed
11315 in a future version.
11320 Disable the use of conditional set instructions.
11322 This switch is mainly for debugging the compiler and will likely be removed
11323 in a future version.
11326 @opindex mcond-exec
11328 Enable the use of conditional execution (default).
11330 This switch is mainly for debugging the compiler and will likely be removed
11331 in a future version.
11333 @item -mno-cond-exec
11334 @opindex mno-cond-exec
11336 Disable the use of conditional execution.
11338 This switch is mainly for debugging the compiler and will likely be removed
11339 in a future version.
11341 @item -mvliw-branch
11342 @opindex mvliw-branch
11344 Run a pass to pack branches into VLIW instructions (default).
11346 This switch is mainly for debugging the compiler and will likely be removed
11347 in a future version.
11349 @item -mno-vliw-branch
11350 @opindex mno-vliw-branch
11352 Do not run a pass to pack branches into VLIW instructions.
11354 This switch is mainly for debugging the compiler and will likely be removed
11355 in a future version.
11357 @item -mmulti-cond-exec
11358 @opindex mmulti-cond-exec
11360 Enable optimization of @code{&&} and @code{||} in conditional execution
11363 This switch is mainly for debugging the compiler and will likely be removed
11364 in a future version.
11366 @item -mno-multi-cond-exec
11367 @opindex mno-multi-cond-exec
11369 Disable optimization of @code{&&} and @code{||} in conditional execution.
11371 This switch is mainly for debugging the compiler and will likely be removed
11372 in a future version.
11374 @item -mnested-cond-exec
11375 @opindex mnested-cond-exec
11377 Enable nested conditional execution optimizations (default).
11379 This switch is mainly for debugging the compiler and will likely be removed
11380 in a future version.
11382 @item -mno-nested-cond-exec
11383 @opindex mno-nested-cond-exec
11385 Disable nested conditional execution optimizations.
11387 This switch is mainly for debugging the compiler and will likely be removed
11388 in a future version.
11390 @item -moptimize-membar
11391 @opindex moptimize-membar
11393 This switch removes redundant @code{membar} instructions from the
11394 compiler generated code. It is enabled by default.
11396 @item -mno-optimize-membar
11397 @opindex mno-optimize-membar
11399 This switch disables the automatic removal of redundant @code{membar}
11400 instructions from the generated code.
11402 @item -mtomcat-stats
11403 @opindex mtomcat-stats
11405 Cause gas to print out tomcat statistics.
11407 @item -mcpu=@var{cpu}
11410 Select the processor type for which to generate code. Possible values are
11411 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11412 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11416 @node GNU/Linux Options
11417 @subsection GNU/Linux Options
11419 These @samp{-m} options are defined for GNU/Linux targets:
11424 Use the GNU C library instead of uClibc. This is the default except
11425 on @samp{*-*-linux-*uclibc*} targets.
11429 Use uClibc instead of the GNU C library. This is the default on
11430 @samp{*-*-linux-*uclibc*} targets.
11433 @node H8/300 Options
11434 @subsection H8/300 Options
11436 These @samp{-m} options are defined for the H8/300 implementations:
11441 Shorten some address references at link time, when possible; uses the
11442 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11443 ld, Using ld}, for a fuller description.
11447 Generate code for the H8/300H@.
11451 Generate code for the H8S@.
11455 Generate code for the H8S and H8/300H in the normal mode. This switch
11456 must be used either with @option{-mh} or @option{-ms}.
11460 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11464 Make @code{int} data 32 bits by default.
11467 @opindex malign-300
11468 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11469 The default for the H8/300H and H8S is to align longs and floats on 4
11471 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11472 This option has no effect on the H8/300.
11476 @subsection HPPA Options
11477 @cindex HPPA Options
11479 These @samp{-m} options are defined for the HPPA family of computers:
11482 @item -march=@var{architecture-type}
11484 Generate code for the specified architecture. The choices for
11485 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11486 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11487 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11488 architecture option for your machine. Code compiled for lower numbered
11489 architectures will run on higher numbered architectures, but not the
11492 @item -mpa-risc-1-0
11493 @itemx -mpa-risc-1-1
11494 @itemx -mpa-risc-2-0
11495 @opindex mpa-risc-1-0
11496 @opindex mpa-risc-1-1
11497 @opindex mpa-risc-2-0
11498 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11501 @opindex mbig-switch
11502 Generate code suitable for big switch tables. Use this option only if
11503 the assembler/linker complain about out of range branches within a switch
11506 @item -mjump-in-delay
11507 @opindex mjump-in-delay
11508 Fill delay slots of function calls with unconditional jump instructions
11509 by modifying the return pointer for the function call to be the target
11510 of the conditional jump.
11512 @item -mdisable-fpregs
11513 @opindex mdisable-fpregs
11514 Prevent floating point registers from being used in any manner. This is
11515 necessary for compiling kernels which perform lazy context switching of
11516 floating point registers. If you use this option and attempt to perform
11517 floating point operations, the compiler will abort.
11519 @item -mdisable-indexing
11520 @opindex mdisable-indexing
11521 Prevent the compiler from using indexing address modes. This avoids some
11522 rather obscure problems when compiling MIG generated code under MACH@.
11524 @item -mno-space-regs
11525 @opindex mno-space-regs
11526 Generate code that assumes the target has no space registers. This allows
11527 GCC to generate faster indirect calls and use unscaled index address modes.
11529 Such code is suitable for level 0 PA systems and kernels.
11531 @item -mfast-indirect-calls
11532 @opindex mfast-indirect-calls
11533 Generate code that assumes calls never cross space boundaries. This
11534 allows GCC to emit code which performs faster indirect calls.
11536 This option will not work in the presence of shared libraries or nested
11539 @item -mfixed-range=@var{register-range}
11540 @opindex mfixed-range
11541 Generate code treating the given register range as fixed registers.
11542 A fixed register is one that the register allocator can not use. This is
11543 useful when compiling kernel code. A register range is specified as
11544 two registers separated by a dash. Multiple register ranges can be
11545 specified separated by a comma.
11547 @item -mlong-load-store
11548 @opindex mlong-load-store
11549 Generate 3-instruction load and store sequences as sometimes required by
11550 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11553 @item -mportable-runtime
11554 @opindex mportable-runtime
11555 Use the portable calling conventions proposed by HP for ELF systems.
11559 Enable the use of assembler directives only GAS understands.
11561 @item -mschedule=@var{cpu-type}
11563 Schedule code according to the constraints for the machine type
11564 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11565 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11566 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11567 proper scheduling option for your machine. The default scheduling is
11571 @opindex mlinker-opt
11572 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11573 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11574 linkers in which they give bogus error messages when linking some programs.
11577 @opindex msoft-float
11578 Generate output containing library calls for floating point.
11579 @strong{Warning:} the requisite libraries are not available for all HPPA
11580 targets. Normally the facilities of the machine's usual C compiler are
11581 used, but this cannot be done directly in cross-compilation. You must make
11582 your own arrangements to provide suitable library functions for
11585 @option{-msoft-float} changes the calling convention in the output file;
11586 therefore, it is only useful if you compile @emph{all} of a program with
11587 this option. In particular, you need to compile @file{libgcc.a}, the
11588 library that comes with GCC, with @option{-msoft-float} in order for
11593 Generate the predefine, @code{_SIO}, for server IO@. The default is
11594 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11595 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11596 options are available under HP-UX and HI-UX@.
11600 Use GNU ld specific options. This passes @option{-shared} to ld when
11601 building a shared library. It is the default when GCC is configured,
11602 explicitly or implicitly, with the GNU linker. This option does not
11603 have any affect on which ld is called, it only changes what parameters
11604 are passed to that ld. The ld that is called is determined by the
11605 @option{--with-ld} configure option, GCC's program search path, and
11606 finally by the user's @env{PATH}. The linker used by GCC can be printed
11607 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11608 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11612 Use HP ld specific options. This passes @option{-b} to ld when building
11613 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11614 links. It is the default when GCC is configured, explicitly or
11615 implicitly, with the HP linker. This option does not have any affect on
11616 which ld is called, it only changes what parameters are passed to that
11617 ld. The ld that is called is determined by the @option{--with-ld}
11618 configure option, GCC's program search path, and finally by the user's
11619 @env{PATH}. The linker used by GCC can be printed using @samp{which
11620 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11621 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11624 @opindex mno-long-calls
11625 Generate code that uses long call sequences. This ensures that a call
11626 is always able to reach linker generated stubs. The default is to generate
11627 long calls only when the distance from the call site to the beginning
11628 of the function or translation unit, as the case may be, exceeds a
11629 predefined limit set by the branch type being used. The limits for
11630 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11631 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11634 Distances are measured from the beginning of functions when using the
11635 @option{-ffunction-sections} option, or when using the @option{-mgas}
11636 and @option{-mno-portable-runtime} options together under HP-UX with
11639 It is normally not desirable to use this option as it will degrade
11640 performance. However, it may be useful in large applications,
11641 particularly when partial linking is used to build the application.
11643 The types of long calls used depends on the capabilities of the
11644 assembler and linker, and the type of code being generated. The
11645 impact on systems that support long absolute calls, and long pic
11646 symbol-difference or pc-relative calls should be relatively small.
11647 However, an indirect call is used on 32-bit ELF systems in pic code
11648 and it is quite long.
11650 @item -munix=@var{unix-std}
11652 Generate compiler predefines and select a startfile for the specified
11653 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11654 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11655 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11656 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11657 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11660 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11661 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11662 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11663 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11664 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11665 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11667 It is @emph{important} to note that this option changes the interfaces
11668 for various library routines. It also affects the operational behavior
11669 of the C library. Thus, @emph{extreme} care is needed in using this
11672 Library code that is intended to operate with more than one UNIX
11673 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11674 as appropriate. Most GNU software doesn't provide this capability.
11678 Suppress the generation of link options to search libdld.sl when the
11679 @option{-static} option is specified on HP-UX 10 and later.
11683 The HP-UX implementation of setlocale in libc has a dependency on
11684 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11685 when the @option{-static} option is specified, special link options
11686 are needed to resolve this dependency.
11688 On HP-UX 10 and later, the GCC driver adds the necessary options to
11689 link with libdld.sl when the @option{-static} option is specified.
11690 This causes the resulting binary to be dynamic. On the 64-bit port,
11691 the linkers generate dynamic binaries by default in any case. The
11692 @option{-nolibdld} option can be used to prevent the GCC driver from
11693 adding these link options.
11697 Add support for multithreading with the @dfn{dce thread} library
11698 under HP-UX@. This option sets flags for both the preprocessor and
11702 @node i386 and x86-64 Options
11703 @subsection Intel 386 and AMD x86-64 Options
11704 @cindex i386 Options
11705 @cindex x86-64 Options
11706 @cindex Intel 386 Options
11707 @cindex AMD x86-64 Options
11709 These @samp{-m} options are defined for the i386 and x86-64 family of
11713 @item -mtune=@var{cpu-type}
11715 Tune to @var{cpu-type} everything applicable about the generated code, except
11716 for the ABI and the set of available instructions. The choices for
11717 @var{cpu-type} are:
11720 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11721 If you know the CPU on which your code will run, then you should use
11722 the corresponding @option{-mtune} option instead of
11723 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11724 of your application will have, then you should use this option.
11726 As new processors are deployed in the marketplace, the behavior of this
11727 option will change. Therefore, if you upgrade to a newer version of
11728 GCC, the code generated option will change to reflect the processors
11729 that were most common when that version of GCC was released.
11731 There is no @option{-march=generic} option because @option{-march}
11732 indicates the instruction set the compiler can use, and there is no
11733 generic instruction set applicable to all processors. In contrast,
11734 @option{-mtune} indicates the processor (or, in this case, collection of
11735 processors) for which the code is optimized.
11737 This selects the CPU to tune for at compilation time by determining
11738 the processor type of the compiling machine. Using @option{-mtune=native}
11739 will produce code optimized for the local machine under the constraints
11740 of the selected instruction set. Using @option{-march=native} will
11741 enable all instruction subsets supported by the local machine (hence
11742 the result might not run on different machines).
11744 Original Intel's i386 CPU@.
11746 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11747 @item i586, pentium
11748 Intel Pentium CPU with no MMX support.
11750 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11752 Intel PentiumPro CPU@.
11754 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11755 instruction set will be used, so the code will run on all i686 family chips.
11757 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11758 @item pentium3, pentium3m
11759 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11762 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11763 support. Used by Centrino notebooks.
11764 @item pentium4, pentium4m
11765 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11767 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11770 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11771 SSE2 and SSE3 instruction set support.
11773 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11774 instruction set support.
11776 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11777 instruction set support.
11779 AMD K6 CPU with MMX instruction set support.
11781 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
11782 @item athlon, athlon-tbird
11783 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
11785 @item athlon-4, athlon-xp, athlon-mp
11786 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
11787 instruction set support.
11788 @item k8, opteron, athlon64, athlon-fx
11789 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11790 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
11791 @item k8-sse3, opteron-sse3, athlon64-sse3
11792 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11793 @item amdfam10, barcelona
11794 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11795 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
11796 instruction set extensions.)
11798 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11801 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
11802 instruction set support.
11804 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
11805 implemented for this chip.)
11807 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11808 implemented for this chip.)
11810 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
11813 While picking a specific @var{cpu-type} will schedule things appropriately
11814 for that particular chip, the compiler will not generate any code that
11815 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11818 @item -march=@var{cpu-type}
11820 Generate instructions for the machine type @var{cpu-type}. The choices
11821 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11822 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11824 @item -mcpu=@var{cpu-type}
11826 A deprecated synonym for @option{-mtune}.
11828 @item -mfpmath=@var{unit}
11830 Generate floating point arithmetics for selected unit @var{unit}. The choices
11831 for @var{unit} are:
11835 Use the standard 387 floating point coprocessor present majority of chips and
11836 emulated otherwise. Code compiled with this option will run almost everywhere.
11837 The temporary results are computed in 80bit precision instead of precision
11838 specified by the type resulting in slightly different results compared to most
11839 of other chips. See @option{-ffloat-store} for more detailed description.
11841 This is the default choice for i386 compiler.
11844 Use scalar floating point instructions present in the SSE instruction set.
11845 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11846 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11847 instruction set supports only single precision arithmetics, thus the double and
11848 extended precision arithmetics is still done using 387. Later version, present
11849 only in Pentium4 and the future AMD x86-64 chips supports double precision
11852 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11853 or @option{-msse2} switches to enable SSE extensions and make this option
11854 effective. For the x86-64 compiler, these extensions are enabled by default.
11856 The resulting code should be considerably faster in the majority of cases and avoid
11857 the numerical instability problems of 387 code, but may break some existing
11858 code that expects temporaries to be 80bit.
11860 This is the default choice for the x86-64 compiler.
11865 Attempt to utilize both instruction sets at once. This effectively double the
11866 amount of available registers and on chips with separate execution units for
11867 387 and SSE the execution resources too. Use this option with care, as it is
11868 still experimental, because the GCC register allocator does not model separate
11869 functional units well resulting in instable performance.
11872 @item -masm=@var{dialect}
11873 @opindex masm=@var{dialect}
11874 Output asm instructions using selected @var{dialect}. Supported
11875 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11876 not support @samp{intel}.
11879 @itemx -mno-ieee-fp
11881 @opindex mno-ieee-fp
11882 Control whether or not the compiler uses IEEE floating point
11883 comparisons. These handle correctly the case where the result of a
11884 comparison is unordered.
11887 @opindex msoft-float
11888 Generate output containing library calls for floating point.
11889 @strong{Warning:} the requisite libraries are not part of GCC@.
11890 Normally the facilities of the machine's usual C compiler are used, but
11891 this can't be done directly in cross-compilation. You must make your
11892 own arrangements to provide suitable library functions for
11895 On machines where a function returns floating point results in the 80387
11896 register stack, some floating point opcodes may be emitted even if
11897 @option{-msoft-float} is used.
11899 @item -mno-fp-ret-in-387
11900 @opindex mno-fp-ret-in-387
11901 Do not use the FPU registers for return values of functions.
11903 The usual calling convention has functions return values of types
11904 @code{float} and @code{double} in an FPU register, even if there
11905 is no FPU@. The idea is that the operating system should emulate
11908 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11909 in ordinary CPU registers instead.
11911 @item -mno-fancy-math-387
11912 @opindex mno-fancy-math-387
11913 Some 387 emulators do not support the @code{sin}, @code{cos} and
11914 @code{sqrt} instructions for the 387. Specify this option to avoid
11915 generating those instructions. This option is the default on FreeBSD,
11916 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11917 indicates that the target cpu will always have an FPU and so the
11918 instruction will not need emulation. As of revision 2.6.1, these
11919 instructions are not generated unless you also use the
11920 @option{-funsafe-math-optimizations} switch.
11922 @item -malign-double
11923 @itemx -mno-align-double
11924 @opindex malign-double
11925 @opindex mno-align-double
11926 Control whether GCC aligns @code{double}, @code{long double}, and
11927 @code{long long} variables on a two word boundary or a one word
11928 boundary. Aligning @code{double} variables on a two word boundary will
11929 produce code that runs somewhat faster on a @samp{Pentium} at the
11930 expense of more memory.
11932 On x86-64, @option{-malign-double} is enabled by default.
11934 @strong{Warning:} if you use the @option{-malign-double} switch,
11935 structures containing the above types will be aligned differently than
11936 the published application binary interface specifications for the 386
11937 and will not be binary compatible with structures in code compiled
11938 without that switch.
11940 @item -m96bit-long-double
11941 @itemx -m128bit-long-double
11942 @opindex m96bit-long-double
11943 @opindex m128bit-long-double
11944 These switches control the size of @code{long double} type. The i386
11945 application binary interface specifies the size to be 96 bits,
11946 so @option{-m96bit-long-double} is the default in 32 bit mode.
11948 Modern architectures (Pentium and newer) would prefer @code{long double}
11949 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11950 conforming to the ABI, this would not be possible. So specifying a
11951 @option{-m128bit-long-double} will align @code{long double}
11952 to a 16 byte boundary by padding the @code{long double} with an additional
11955 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11956 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11958 Notice that neither of these options enable any extra precision over the x87
11959 standard of 80 bits for a @code{long double}.
11961 @strong{Warning:} if you override the default value for your target ABI, the
11962 structures and arrays containing @code{long double} variables will change
11963 their size as well as function calling convention for function taking
11964 @code{long double} will be modified. Hence they will not be binary
11965 compatible with arrays or structures in code compiled without that switch.
11967 @item -mlarge-data-threshold=@var{number}
11968 @opindex mlarge-data-threshold=@var{number}
11969 When @option{-mcmodel=medium} is specified, the data greater than
11970 @var{threshold} are placed in large data section. This value must be the
11971 same across all object linked into the binary and defaults to 65535.
11975 Use a different function-calling convention, in which functions that
11976 take a fixed number of arguments return with the @code{ret} @var{num}
11977 instruction, which pops their arguments while returning. This saves one
11978 instruction in the caller since there is no need to pop the arguments
11981 You can specify that an individual function is called with this calling
11982 sequence with the function attribute @samp{stdcall}. You can also
11983 override the @option{-mrtd} option by using the function attribute
11984 @samp{cdecl}. @xref{Function Attributes}.
11986 @strong{Warning:} this calling convention is incompatible with the one
11987 normally used on Unix, so you cannot use it if you need to call
11988 libraries compiled with the Unix compiler.
11990 Also, you must provide function prototypes for all functions that
11991 take variable numbers of arguments (including @code{printf});
11992 otherwise incorrect code will be generated for calls to those
11995 In addition, seriously incorrect code will result if you call a
11996 function with too many arguments. (Normally, extra arguments are
11997 harmlessly ignored.)
11999 @item -mregparm=@var{num}
12001 Control how many registers are used to pass integer arguments. By
12002 default, no registers are used to pass arguments, and at most 3
12003 registers can be used. You can control this behavior for a specific
12004 function by using the function attribute @samp{regparm}.
12005 @xref{Function Attributes}.
12007 @strong{Warning:} if you use this switch, and
12008 @var{num} is nonzero, then you must build all modules with the same
12009 value, including any libraries. This includes the system libraries and
12013 @opindex msseregparm
12014 Use SSE register passing conventions for float and double arguments
12015 and return values. You can control this behavior for a specific
12016 function by using the function attribute @samp{sseregparm}.
12017 @xref{Function Attributes}.
12019 @strong{Warning:} if you use this switch then you must build all
12020 modules with the same value, including any libraries. This includes
12021 the system libraries and startup modules.
12030 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12031 is specified, the significands of results of floating-point operations are
12032 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12033 significands of results of floating-point operations to 53 bits (double
12034 precision) and @option{-mpc80} rounds the significands of results of
12035 floating-point operations to 64 bits (extended double precision), which is
12036 the default. When this option is used, floating-point operations in higher
12037 precisions are not available to the programmer without setting the FPU
12038 control word explicitly.
12040 Setting the rounding of floating-point operations to less than the default
12041 80 bits can speed some programs by 2% or more. Note that some mathematical
12042 libraries assume that extended precision (80 bit) floating-point operations
12043 are enabled by default; routines in such libraries could suffer significant
12044 loss of accuracy, typically through so-called "catastrophic cancellation",
12045 when this option is used to set the precision to less than extended precision.
12047 @item -mstackrealign
12048 @opindex mstackrealign
12049 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12050 option will generate an alternate prologue and epilogue that realigns the
12051 runtime stack if necessary. This supports mixing legacy codes that keep
12052 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12053 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12054 applicable to individual functions.
12056 @item -mpreferred-stack-boundary=@var{num}
12057 @opindex mpreferred-stack-boundary
12058 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12059 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12060 the default is 4 (16 bytes or 128 bits).
12062 @item -mincoming-stack-boundary=@var{num}
12063 @opindex mincoming-stack-boundary
12064 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12065 boundary. If @option{-mincoming-stack-boundary} is not specified,
12066 the one specified by @option{-mpreferred-stack-boundary} will be used.
12068 On Pentium and PentiumPro, @code{double} and @code{long double} values
12069 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12070 suffer significant run time performance penalties. On Pentium III, the
12071 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12072 properly if it is not 16 byte aligned.
12074 To ensure proper alignment of this values on the stack, the stack boundary
12075 must be as aligned as that required by any value stored on the stack.
12076 Further, every function must be generated such that it keeps the stack
12077 aligned. Thus calling a function compiled with a higher preferred
12078 stack boundary from a function compiled with a lower preferred stack
12079 boundary will most likely misalign the stack. It is recommended that
12080 libraries that use callbacks always use the default setting.
12082 This extra alignment does consume extra stack space, and generally
12083 increases code size. Code that is sensitive to stack space usage, such
12084 as embedded systems and operating system kernels, may want to reduce the
12085 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12129 These switches enable or disable the use of instructions in the MMX,
12130 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, FMA4, XOP,
12131 LWP, ABM or 3DNow!@: extended instruction sets.
12132 These extensions are also available as built-in functions: see
12133 @ref{X86 Built-in Functions}, for details of the functions enabled and
12134 disabled by these switches.
12136 To have SSE/SSE2 instructions generated automatically from floating-point
12137 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12139 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12140 generates new AVX instructions or AVX equivalence for all SSEx instructions
12143 These options will enable GCC to use these extended instructions in
12144 generated code, even without @option{-mfpmath=sse}. Applications which
12145 perform runtime CPU detection must compile separate files for each
12146 supported architecture, using the appropriate flags. In particular,
12147 the file containing the CPU detection code should be compiled without
12151 @itemx -mno-fused-madd
12152 @opindex mfused-madd
12153 @opindex mno-fused-madd
12154 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12155 instructions. The default is to use these instructions.
12159 This option instructs GCC to emit a @code{cld} instruction in the prologue
12160 of functions that use string instructions. String instructions depend on
12161 the DF flag to select between autoincrement or autodecrement mode. While the
12162 ABI specifies the DF flag to be cleared on function entry, some operating
12163 systems violate this specification by not clearing the DF flag in their
12164 exception dispatchers. The exception handler can be invoked with the DF flag
12165 set which leads to wrong direction mode, when string instructions are used.
12166 This option can be enabled by default on 32-bit x86 targets by configuring
12167 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12168 instructions can be suppressed with the @option{-mno-cld} compiler option
12173 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12174 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12175 data types. This is useful for high resolution counters that could be updated
12176 by multiple processors (or cores). This instruction is generated as part of
12177 atomic built-in functions: see @ref{Atomic Builtins} for details.
12181 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12182 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12183 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12184 SAHF are load and store instructions, respectively, for certain status flags.
12185 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12186 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12190 This option will enable GCC to use movbe instruction to implement
12191 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12195 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12196 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12197 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12201 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12202 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12203 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12204 variants) for single precision floating point arguments. These instructions
12205 are generated only when @option{-funsafe-math-optimizations} is enabled
12206 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12207 Note that while the throughput of the sequence is higher than the throughput
12208 of the non-reciprocal instruction, the precision of the sequence can be
12209 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12211 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12212 already with @option{-ffast-math} (or the above option combination), and
12213 doesn't need @option{-mrecip}.
12215 @item -mveclibabi=@var{type}
12216 @opindex mveclibabi
12217 Specifies the ABI type to use for vectorizing intrinsics using an
12218 external library. Supported types are @code{svml} for the Intel short
12219 vector math library and @code{acml} for the AMD math core library style
12220 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12221 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12222 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12223 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12224 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12225 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12226 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12227 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12228 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12229 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12230 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12231 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12232 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12233 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12234 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12235 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12236 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12237 compatible library will have to be specified at link time.
12239 @item -mabi=@var{name}
12241 Generate code for the specified calling convention. Permissible values
12242 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12243 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12244 ABI when targeting Windows. On all other systems, the default is the
12245 SYSV ABI. You can control this behavior for a specific function by
12246 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12247 @xref{Function Attributes}.
12250 @itemx -mno-push-args
12251 @opindex mpush-args
12252 @opindex mno-push-args
12253 Use PUSH operations to store outgoing parameters. This method is shorter
12254 and usually equally fast as method using SUB/MOV operations and is enabled
12255 by default. In some cases disabling it may improve performance because of
12256 improved scheduling and reduced dependencies.
12258 @item -maccumulate-outgoing-args
12259 @opindex maccumulate-outgoing-args
12260 If enabled, the maximum amount of space required for outgoing arguments will be
12261 computed in the function prologue. This is faster on most modern CPUs
12262 because of reduced dependencies, improved scheduling and reduced stack usage
12263 when preferred stack boundary is not equal to 2. The drawback is a notable
12264 increase in code size. This switch implies @option{-mno-push-args}.
12268 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12269 on thread-safe exception handling must compile and link all code with the
12270 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12271 @option{-D_MT}; when linking, it links in a special thread helper library
12272 @option{-lmingwthrd} which cleans up per thread exception handling data.
12274 @item -mno-align-stringops
12275 @opindex mno-align-stringops
12276 Do not align destination of inlined string operations. This switch reduces
12277 code size and improves performance in case the destination is already aligned,
12278 but GCC doesn't know about it.
12280 @item -minline-all-stringops
12281 @opindex minline-all-stringops
12282 By default GCC inlines string operations only when destination is known to be
12283 aligned at least to 4 byte boundary. This enables more inlining, increase code
12284 size, but may improve performance of code that depends on fast memcpy, strlen
12285 and memset for short lengths.
12287 @item -minline-stringops-dynamically
12288 @opindex minline-stringops-dynamically
12289 For string operation of unknown size, inline runtime checks so for small
12290 blocks inline code is used, while for large blocks library call is used.
12292 @item -mstringop-strategy=@var{alg}
12293 @opindex mstringop-strategy=@var{alg}
12294 Overwrite internal decision heuristic about particular algorithm to inline
12295 string operation with. The allowed values are @code{rep_byte},
12296 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12297 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12298 expanding inline loop, @code{libcall} for always expanding library call.
12300 @item -momit-leaf-frame-pointer
12301 @opindex momit-leaf-frame-pointer
12302 Don't keep the frame pointer in a register for leaf functions. This
12303 avoids the instructions to save, set up and restore frame pointers and
12304 makes an extra register available in leaf functions. The option
12305 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12306 which might make debugging harder.
12308 @item -mtls-direct-seg-refs
12309 @itemx -mno-tls-direct-seg-refs
12310 @opindex mtls-direct-seg-refs
12311 Controls whether TLS variables may be accessed with offsets from the
12312 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12313 or whether the thread base pointer must be added. Whether or not this
12314 is legal depends on the operating system, and whether it maps the
12315 segment to cover the entire TLS area.
12317 For systems that use GNU libc, the default is on.
12320 @itemx -mno-sse2avx
12322 Specify that the assembler should encode SSE instructions with VEX
12323 prefix. The option @option{-mavx} turns this on by default.
12326 These @samp{-m} switches are supported in addition to the above
12327 on AMD x86-64 processors in 64-bit environments.
12334 Generate code for a 32-bit or 64-bit environment.
12335 The 32-bit environment sets int, long and pointer to 32 bits and
12336 generates code that runs on any i386 system.
12337 The 64-bit environment sets int to 32 bits and long and pointer
12338 to 64 bits and generates code for AMD's x86-64 architecture. For
12339 darwin only the -m64 option turns off the @option{-fno-pic} and
12340 @option{-mdynamic-no-pic} options.
12342 @item -mno-red-zone
12343 @opindex mno-red-zone
12344 Do not use a so called red zone for x86-64 code. The red zone is mandated
12345 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12346 stack pointer that will not be modified by signal or interrupt handlers
12347 and therefore can be used for temporary data without adjusting the stack
12348 pointer. The flag @option{-mno-red-zone} disables this red zone.
12350 @item -mcmodel=small
12351 @opindex mcmodel=small
12352 Generate code for the small code model: the program and its symbols must
12353 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12354 Programs can be statically or dynamically linked. This is the default
12357 @item -mcmodel=kernel
12358 @opindex mcmodel=kernel
12359 Generate code for the kernel code model. The kernel runs in the
12360 negative 2 GB of the address space.
12361 This model has to be used for Linux kernel code.
12363 @item -mcmodel=medium
12364 @opindex mcmodel=medium
12365 Generate code for the medium model: The program is linked in the lower 2
12366 GB of the address space. Small symbols are also placed there. Symbols
12367 with sizes larger than @option{-mlarge-data-threshold} are put into
12368 large data or bss sections and can be located above 2GB. Programs can
12369 be statically or dynamically linked.
12371 @item -mcmodel=large
12372 @opindex mcmodel=large
12373 Generate code for the large model: This model makes no assumptions
12374 about addresses and sizes of sections.
12377 @node IA-64 Options
12378 @subsection IA-64 Options
12379 @cindex IA-64 Options
12381 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12385 @opindex mbig-endian
12386 Generate code for a big endian target. This is the default for HP-UX@.
12388 @item -mlittle-endian
12389 @opindex mlittle-endian
12390 Generate code for a little endian target. This is the default for AIX5
12396 @opindex mno-gnu-as
12397 Generate (or don't) code for the GNU assembler. This is the default.
12398 @c Also, this is the default if the configure option @option{--with-gnu-as}
12404 @opindex mno-gnu-ld
12405 Generate (or don't) code for the GNU linker. This is the default.
12406 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12411 Generate code that does not use a global pointer register. The result
12412 is not position independent code, and violates the IA-64 ABI@.
12414 @item -mvolatile-asm-stop
12415 @itemx -mno-volatile-asm-stop
12416 @opindex mvolatile-asm-stop
12417 @opindex mno-volatile-asm-stop
12418 Generate (or don't) a stop bit immediately before and after volatile asm
12421 @item -mregister-names
12422 @itemx -mno-register-names
12423 @opindex mregister-names
12424 @opindex mno-register-names
12425 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12426 the stacked registers. This may make assembler output more readable.
12432 Disable (or enable) optimizations that use the small data section. This may
12433 be useful for working around optimizer bugs.
12435 @item -mconstant-gp
12436 @opindex mconstant-gp
12437 Generate code that uses a single constant global pointer value. This is
12438 useful when compiling kernel code.
12442 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12443 This is useful when compiling firmware code.
12445 @item -minline-float-divide-min-latency
12446 @opindex minline-float-divide-min-latency
12447 Generate code for inline divides of floating point values
12448 using the minimum latency algorithm.
12450 @item -minline-float-divide-max-throughput
12451 @opindex minline-float-divide-max-throughput
12452 Generate code for inline divides of floating point values
12453 using the maximum throughput algorithm.
12455 @item -mno-inline-float-divide
12456 @opindex mno-inline-float-divide
12457 Do not generate inline code for divides of floating point values.
12459 @item -minline-int-divide-min-latency
12460 @opindex minline-int-divide-min-latency
12461 Generate code for inline divides of integer values
12462 using the minimum latency algorithm.
12464 @item -minline-int-divide-max-throughput
12465 @opindex minline-int-divide-max-throughput
12466 Generate code for inline divides of integer values
12467 using the maximum throughput algorithm.
12469 @item -mno-inline-int-divide
12470 @opindex mno-inline-int-divide
12471 Do not generate inline code for divides of integer values.
12473 @item -minline-sqrt-min-latency
12474 @opindex minline-sqrt-min-latency
12475 Generate code for inline square roots
12476 using the minimum latency algorithm.
12478 @item -minline-sqrt-max-throughput
12479 @opindex minline-sqrt-max-throughput
12480 Generate code for inline square roots
12481 using the maximum throughput algorithm.
12483 @item -mno-inline-sqrt
12484 @opindex mno-inline-sqrt
12485 Do not generate inline code for sqrt.
12488 @itemx -mno-fused-madd
12489 @opindex mfused-madd
12490 @opindex mno-fused-madd
12491 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12492 instructions. The default is to use these instructions.
12494 @item -mno-dwarf2-asm
12495 @itemx -mdwarf2-asm
12496 @opindex mno-dwarf2-asm
12497 @opindex mdwarf2-asm
12498 Don't (or do) generate assembler code for the DWARF2 line number debugging
12499 info. This may be useful when not using the GNU assembler.
12501 @item -mearly-stop-bits
12502 @itemx -mno-early-stop-bits
12503 @opindex mearly-stop-bits
12504 @opindex mno-early-stop-bits
12505 Allow stop bits to be placed earlier than immediately preceding the
12506 instruction that triggered the stop bit. This can improve instruction
12507 scheduling, but does not always do so.
12509 @item -mfixed-range=@var{register-range}
12510 @opindex mfixed-range
12511 Generate code treating the given register range as fixed registers.
12512 A fixed register is one that the register allocator can not use. This is
12513 useful when compiling kernel code. A register range is specified as
12514 two registers separated by a dash. Multiple register ranges can be
12515 specified separated by a comma.
12517 @item -mtls-size=@var{tls-size}
12519 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12522 @item -mtune=@var{cpu-type}
12524 Tune the instruction scheduling for a particular CPU, Valid values are
12525 itanium, itanium1, merced, itanium2, and mckinley.
12531 Generate code for a 32-bit or 64-bit environment.
12532 The 32-bit environment sets int, long and pointer to 32 bits.
12533 The 64-bit environment sets int to 32 bits and long and pointer
12534 to 64 bits. These are HP-UX specific flags.
12536 @item -mno-sched-br-data-spec
12537 @itemx -msched-br-data-spec
12538 @opindex mno-sched-br-data-spec
12539 @opindex msched-br-data-spec
12540 (Dis/En)able data speculative scheduling before reload.
12541 This will result in generation of the ld.a instructions and
12542 the corresponding check instructions (ld.c / chk.a).
12543 The default is 'disable'.
12545 @item -msched-ar-data-spec
12546 @itemx -mno-sched-ar-data-spec
12547 @opindex msched-ar-data-spec
12548 @opindex mno-sched-ar-data-spec
12549 (En/Dis)able data speculative scheduling after reload.
12550 This will result in generation of the ld.a instructions and
12551 the corresponding check instructions (ld.c / chk.a).
12552 The default is 'enable'.
12554 @item -mno-sched-control-spec
12555 @itemx -msched-control-spec
12556 @opindex mno-sched-control-spec
12557 @opindex msched-control-spec
12558 (Dis/En)able control speculative scheduling. This feature is
12559 available only during region scheduling (i.e.@: before reload).
12560 This will result in generation of the ld.s instructions and
12561 the corresponding check instructions chk.s .
12562 The default is 'disable'.
12564 @item -msched-br-in-data-spec
12565 @itemx -mno-sched-br-in-data-spec
12566 @opindex msched-br-in-data-spec
12567 @opindex mno-sched-br-in-data-spec
12568 (En/Dis)able speculative scheduling of the instructions that
12569 are dependent on the data speculative loads before reload.
12570 This is effective only with @option{-msched-br-data-spec} enabled.
12571 The default is 'enable'.
12573 @item -msched-ar-in-data-spec
12574 @itemx -mno-sched-ar-in-data-spec
12575 @opindex msched-ar-in-data-spec
12576 @opindex mno-sched-ar-in-data-spec
12577 (En/Dis)able speculative scheduling of the instructions that
12578 are dependent on the data speculative loads after reload.
12579 This is effective only with @option{-msched-ar-data-spec} enabled.
12580 The default is 'enable'.
12582 @item -msched-in-control-spec
12583 @itemx -mno-sched-in-control-spec
12584 @opindex msched-in-control-spec
12585 @opindex mno-sched-in-control-spec
12586 (En/Dis)able speculative scheduling of the instructions that
12587 are dependent on the control speculative loads.
12588 This is effective only with @option{-msched-control-spec} enabled.
12589 The default is 'enable'.
12591 @item -mno-sched-prefer-non-data-spec-insns
12592 @itemx -msched-prefer-non-data-spec-insns
12593 @opindex mno-sched-prefer-non-data-spec-insns
12594 @opindex msched-prefer-non-data-spec-insns
12595 If enabled, data speculative instructions will be chosen for schedule
12596 only if there are no other choices at the moment. This will make
12597 the use of the data speculation much more conservative.
12598 The default is 'disable'.
12600 @item -mno-sched-prefer-non-control-spec-insns
12601 @itemx -msched-prefer-non-control-spec-insns
12602 @opindex mno-sched-prefer-non-control-spec-insns
12603 @opindex msched-prefer-non-control-spec-insns
12604 If enabled, control speculative instructions will be chosen for schedule
12605 only if there are no other choices at the moment. This will make
12606 the use of the control speculation much more conservative.
12607 The default is 'disable'.
12609 @item -mno-sched-count-spec-in-critical-path
12610 @itemx -msched-count-spec-in-critical-path
12611 @opindex mno-sched-count-spec-in-critical-path
12612 @opindex msched-count-spec-in-critical-path
12613 If enabled, speculative dependencies will be considered during
12614 computation of the instructions priorities. This will make the use of the
12615 speculation a bit more conservative.
12616 The default is 'disable'.
12618 @item -msched-spec-ldc
12619 @opindex msched-spec-ldc
12620 Use a simple data speculation check. This option is on by default.
12622 @item -msched-control-spec-ldc
12623 @opindex msched-spec-ldc
12624 Use a simple check for control speculation. This option is on by default.
12626 @item -msched-stop-bits-after-every-cycle
12627 @opindex msched-stop-bits-after-every-cycle
12628 Place a stop bit after every cycle when scheduling. This option is on
12631 @item -msched-fp-mem-deps-zero-cost
12632 @opindex msched-fp-mem-deps-zero-cost
12633 Assume that floating-point stores and loads are not likely to cause a conflict
12634 when placed into the same instruction group. This option is disabled by
12637 @item -msel-sched-dont-check-control-spec
12638 @opindex msel-sched-dont-check-control-spec
12639 Generate checks for control speculation in selective scheduling.
12640 This flag is disabled by default.
12642 @item -msched-max-memory-insns=@var{max-insns}
12643 @opindex msched-max-memory-insns
12644 Limit on the number of memory insns per instruction group, giving lower
12645 priority to subsequent memory insns attempting to schedule in the same
12646 instruction group. Frequently useful to prevent cache bank conflicts.
12647 The default value is 1.
12649 @item -msched-max-memory-insns-hard-limit
12650 @opindex msched-max-memory-insns-hard-limit
12651 Disallow more than `msched-max-memory-insns' in instruction group.
12652 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12653 when limit is reached but may still schedule memory operations.
12657 @node IA-64/VMS Options
12658 @subsection IA-64/VMS Options
12660 These @samp{-m} options are defined for the IA-64/VMS implementations:
12663 @item -mvms-return-codes
12664 @opindex mvms-return-codes
12665 Return VMS condition codes from main. The default is to return POSIX
12666 style condition (e.g.@ error) codes.
12668 @item -mdebug-main=@var{prefix}
12669 @opindex mdebug-main=@var{prefix}
12670 Flag the first routine whose name starts with @var{prefix} as the main
12671 routine for the debugger.
12675 Default to 64bit memory allocation routines.
12679 @subsection LM32 Options
12680 @cindex LM32 options
12682 These @option{-m} options are defined for the Lattice Mico32 architecture:
12685 @item -mbarrel-shift-enabled
12686 @opindex mbarrel-shift-enabled
12687 Enable barrel-shift instructions.
12689 @item -mdivide-enabled
12690 @opindex mdivide-enabled
12691 Enable divide and modulus instructions.
12693 @item -mmultiply-enabled
12694 @opindex multiply-enabled
12695 Enable multiply instructions.
12697 @item -msign-extend-enabled
12698 @opindex msign-extend-enabled
12699 Enable sign extend instructions.
12701 @item -muser-enabled
12702 @opindex muser-enabled
12703 Enable user-defined instructions.
12708 @subsection M32C Options
12709 @cindex M32C options
12712 @item -mcpu=@var{name}
12714 Select the CPU for which code is generated. @var{name} may be one of
12715 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12716 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12717 the M32C/80 series.
12721 Specifies that the program will be run on the simulator. This causes
12722 an alternate runtime library to be linked in which supports, for
12723 example, file I/O@. You must not use this option when generating
12724 programs that will run on real hardware; you must provide your own
12725 runtime library for whatever I/O functions are needed.
12727 @item -memregs=@var{number}
12729 Specifies the number of memory-based pseudo-registers GCC will use
12730 during code generation. These pseudo-registers will be used like real
12731 registers, so there is a tradeoff between GCC's ability to fit the
12732 code into available registers, and the performance penalty of using
12733 memory instead of registers. Note that all modules in a program must
12734 be compiled with the same value for this option. Because of that, you
12735 must not use this option with the default runtime libraries gcc
12740 @node M32R/D Options
12741 @subsection M32R/D Options
12742 @cindex M32R/D options
12744 These @option{-m} options are defined for Renesas M32R/D architectures:
12749 Generate code for the M32R/2@.
12753 Generate code for the M32R/X@.
12757 Generate code for the M32R@. This is the default.
12759 @item -mmodel=small
12760 @opindex mmodel=small
12761 Assume all objects live in the lower 16MB of memory (so that their addresses
12762 can be loaded with the @code{ld24} instruction), and assume all subroutines
12763 are reachable with the @code{bl} instruction.
12764 This is the default.
12766 The addressability of a particular object can be set with the
12767 @code{model} attribute.
12769 @item -mmodel=medium
12770 @opindex mmodel=medium
12771 Assume objects may be anywhere in the 32-bit address space (the compiler
12772 will generate @code{seth/add3} instructions to load their addresses), and
12773 assume all subroutines are reachable with the @code{bl} instruction.
12775 @item -mmodel=large
12776 @opindex mmodel=large
12777 Assume objects may be anywhere in the 32-bit address space (the compiler
12778 will generate @code{seth/add3} instructions to load their addresses), and
12779 assume subroutines may not be reachable with the @code{bl} instruction
12780 (the compiler will generate the much slower @code{seth/add3/jl}
12781 instruction sequence).
12784 @opindex msdata=none
12785 Disable use of the small data area. Variables will be put into
12786 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12787 @code{section} attribute has been specified).
12788 This is the default.
12790 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12791 Objects may be explicitly put in the small data area with the
12792 @code{section} attribute using one of these sections.
12794 @item -msdata=sdata
12795 @opindex msdata=sdata
12796 Put small global and static data in the small data area, but do not
12797 generate special code to reference them.
12800 @opindex msdata=use
12801 Put small global and static data in the small data area, and generate
12802 special instructions to reference them.
12806 @cindex smaller data references
12807 Put global and static objects less than or equal to @var{num} bytes
12808 into the small data or bss sections instead of the normal data or bss
12809 sections. The default value of @var{num} is 8.
12810 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12811 for this option to have any effect.
12813 All modules should be compiled with the same @option{-G @var{num}} value.
12814 Compiling with different values of @var{num} may or may not work; if it
12815 doesn't the linker will give an error message---incorrect code will not be
12820 Makes the M32R specific code in the compiler display some statistics
12821 that might help in debugging programs.
12823 @item -malign-loops
12824 @opindex malign-loops
12825 Align all loops to a 32-byte boundary.
12827 @item -mno-align-loops
12828 @opindex mno-align-loops
12829 Do not enforce a 32-byte alignment for loops. This is the default.
12831 @item -missue-rate=@var{number}
12832 @opindex missue-rate=@var{number}
12833 Issue @var{number} instructions per cycle. @var{number} can only be 1
12836 @item -mbranch-cost=@var{number}
12837 @opindex mbranch-cost=@var{number}
12838 @var{number} can only be 1 or 2. If it is 1 then branches will be
12839 preferred over conditional code, if it is 2, then the opposite will
12842 @item -mflush-trap=@var{number}
12843 @opindex mflush-trap=@var{number}
12844 Specifies the trap number to use to flush the cache. The default is
12845 12. Valid numbers are between 0 and 15 inclusive.
12847 @item -mno-flush-trap
12848 @opindex mno-flush-trap
12849 Specifies that the cache cannot be flushed by using a trap.
12851 @item -mflush-func=@var{name}
12852 @opindex mflush-func=@var{name}
12853 Specifies the name of the operating system function to call to flush
12854 the cache. The default is @emph{_flush_cache}, but a function call
12855 will only be used if a trap is not available.
12857 @item -mno-flush-func
12858 @opindex mno-flush-func
12859 Indicates that there is no OS function for flushing the cache.
12863 @node M680x0 Options
12864 @subsection M680x0 Options
12865 @cindex M680x0 options
12867 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12868 The default settings depend on which architecture was selected when
12869 the compiler was configured; the defaults for the most common choices
12873 @item -march=@var{arch}
12875 Generate code for a specific M680x0 or ColdFire instruction set
12876 architecture. Permissible values of @var{arch} for M680x0
12877 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12878 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12879 architectures are selected according to Freescale's ISA classification
12880 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12881 @samp{isab} and @samp{isac}.
12883 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12884 code for a ColdFire target. The @var{arch} in this macro is one of the
12885 @option{-march} arguments given above.
12887 When used together, @option{-march} and @option{-mtune} select code
12888 that runs on a family of similar processors but that is optimized
12889 for a particular microarchitecture.
12891 @item -mcpu=@var{cpu}
12893 Generate code for a specific M680x0 or ColdFire processor.
12894 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12895 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12896 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12897 below, which also classifies the CPUs into families:
12899 @multitable @columnfractions 0.20 0.80
12900 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12901 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12902 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12903 @item @samp{5206e} @tab @samp{5206e}
12904 @item @samp{5208} @tab @samp{5207} @samp{5208}
12905 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12906 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12907 @item @samp{5216} @tab @samp{5214} @samp{5216}
12908 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12909 @item @samp{5225} @tab @samp{5224} @samp{5225}
12910 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12911 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12912 @item @samp{5249} @tab @samp{5249}
12913 @item @samp{5250} @tab @samp{5250}
12914 @item @samp{5271} @tab @samp{5270} @samp{5271}
12915 @item @samp{5272} @tab @samp{5272}
12916 @item @samp{5275} @tab @samp{5274} @samp{5275}
12917 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12918 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12919 @item @samp{5307} @tab @samp{5307}
12920 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12921 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12922 @item @samp{5407} @tab @samp{5407}
12923 @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}
12926 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12927 @var{arch} is compatible with @var{cpu}. Other combinations of
12928 @option{-mcpu} and @option{-march} are rejected.
12930 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12931 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12932 where the value of @var{family} is given by the table above.
12934 @item -mtune=@var{tune}
12936 Tune the code for a particular microarchitecture, within the
12937 constraints set by @option{-march} and @option{-mcpu}.
12938 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12939 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12940 and @samp{cpu32}. The ColdFire microarchitectures
12941 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12943 You can also use @option{-mtune=68020-40} for code that needs
12944 to run relatively well on 68020, 68030 and 68040 targets.
12945 @option{-mtune=68020-60} is similar but includes 68060 targets
12946 as well. These two options select the same tuning decisions as
12947 @option{-m68020-40} and @option{-m68020-60} respectively.
12949 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12950 when tuning for 680x0 architecture @var{arch}. It also defines
12951 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12952 option is used. If gcc is tuning for a range of architectures,
12953 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12954 it defines the macros for every architecture in the range.
12956 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12957 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12958 of the arguments given above.
12964 Generate output for a 68000. This is the default
12965 when the compiler is configured for 68000-based systems.
12966 It is equivalent to @option{-march=68000}.
12968 Use this option for microcontrollers with a 68000 or EC000 core,
12969 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12973 Generate output for a 68010. This is the default
12974 when the compiler is configured for 68010-based systems.
12975 It is equivalent to @option{-march=68010}.
12981 Generate output for a 68020. This is the default
12982 when the compiler is configured for 68020-based systems.
12983 It is equivalent to @option{-march=68020}.
12987 Generate output for a 68030. This is the default when the compiler is
12988 configured for 68030-based systems. It is equivalent to
12989 @option{-march=68030}.
12993 Generate output for a 68040. This is the default when the compiler is
12994 configured for 68040-based systems. It is equivalent to
12995 @option{-march=68040}.
12997 This option inhibits the use of 68881/68882 instructions that have to be
12998 emulated by software on the 68040. Use this option if your 68040 does not
12999 have code to emulate those instructions.
13003 Generate output for a 68060. This is the default when the compiler is
13004 configured for 68060-based systems. It is equivalent to
13005 @option{-march=68060}.
13007 This option inhibits the use of 68020 and 68881/68882 instructions that
13008 have to be emulated by software on the 68060. Use this option if your 68060
13009 does not have code to emulate those instructions.
13013 Generate output for a CPU32. This is the default
13014 when the compiler is configured for CPU32-based systems.
13015 It is equivalent to @option{-march=cpu32}.
13017 Use this option for microcontrollers with a
13018 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13019 68336, 68340, 68341, 68349 and 68360.
13023 Generate output for a 520X ColdFire CPU@. This is the default
13024 when the compiler is configured for 520X-based systems.
13025 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13026 in favor of that option.
13028 Use this option for microcontroller with a 5200 core, including
13029 the MCF5202, MCF5203, MCF5204 and MCF5206.
13033 Generate output for a 5206e ColdFire CPU@. The option is now
13034 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13038 Generate output for a member of the ColdFire 528X family.
13039 The option is now deprecated in favor of the equivalent
13040 @option{-mcpu=528x}.
13044 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13045 in favor of the equivalent @option{-mcpu=5307}.
13049 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13050 in favor of the equivalent @option{-mcpu=5407}.
13054 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13055 This includes use of hardware floating point instructions.
13056 The option is equivalent to @option{-mcpu=547x}, and is now
13057 deprecated in favor of that option.
13061 Generate output for a 68040, without using any of the new instructions.
13062 This results in code which can run relatively efficiently on either a
13063 68020/68881 or a 68030 or a 68040. The generated code does use the
13064 68881 instructions that are emulated on the 68040.
13066 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13070 Generate output for a 68060, without using any of the new instructions.
13071 This results in code which can run relatively efficiently on either a
13072 68020/68881 or a 68030 or a 68040. The generated code does use the
13073 68881 instructions that are emulated on the 68060.
13075 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13079 @opindex mhard-float
13081 Generate floating-point instructions. This is the default for 68020
13082 and above, and for ColdFire devices that have an FPU@. It defines the
13083 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13084 on ColdFire targets.
13087 @opindex msoft-float
13088 Do not generate floating-point instructions; use library calls instead.
13089 This is the default for 68000, 68010, and 68832 targets. It is also
13090 the default for ColdFire devices that have no FPU.
13096 Generate (do not generate) ColdFire hardware divide and remainder
13097 instructions. If @option{-march} is used without @option{-mcpu},
13098 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13099 architectures. Otherwise, the default is taken from the target CPU
13100 (either the default CPU, or the one specified by @option{-mcpu}). For
13101 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13102 @option{-mcpu=5206e}.
13104 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13108 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13109 Additionally, parameters passed on the stack are also aligned to a
13110 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13114 Do not consider type @code{int} to be 16 bits wide. This is the default.
13117 @itemx -mno-bitfield
13118 @opindex mnobitfield
13119 @opindex mno-bitfield
13120 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13121 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13125 Do use the bit-field instructions. The @option{-m68020} option implies
13126 @option{-mbitfield}. This is the default if you use a configuration
13127 designed for a 68020.
13131 Use a different function-calling convention, in which functions
13132 that take a fixed number of arguments return with the @code{rtd}
13133 instruction, which pops their arguments while returning. This
13134 saves one instruction in the caller since there is no need to pop
13135 the arguments there.
13137 This calling convention is incompatible with the one normally
13138 used on Unix, so you cannot use it if you need to call libraries
13139 compiled with the Unix compiler.
13141 Also, you must provide function prototypes for all functions that
13142 take variable numbers of arguments (including @code{printf});
13143 otherwise incorrect code will be generated for calls to those
13146 In addition, seriously incorrect code will result if you call a
13147 function with too many arguments. (Normally, extra arguments are
13148 harmlessly ignored.)
13150 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13151 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13155 Do not use the calling conventions selected by @option{-mrtd}.
13156 This is the default.
13159 @itemx -mno-align-int
13160 @opindex malign-int
13161 @opindex mno-align-int
13162 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13163 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13164 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13165 Aligning variables on 32-bit boundaries produces code that runs somewhat
13166 faster on processors with 32-bit busses at the expense of more memory.
13168 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13169 align structures containing the above types differently than
13170 most published application binary interface specifications for the m68k.
13174 Use the pc-relative addressing mode of the 68000 directly, instead of
13175 using a global offset table. At present, this option implies @option{-fpic},
13176 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13177 not presently supported with @option{-mpcrel}, though this could be supported for
13178 68020 and higher processors.
13180 @item -mno-strict-align
13181 @itemx -mstrict-align
13182 @opindex mno-strict-align
13183 @opindex mstrict-align
13184 Do not (do) assume that unaligned memory references will be handled by
13188 Generate code that allows the data segment to be located in a different
13189 area of memory from the text segment. This allows for execute in place in
13190 an environment without virtual memory management. This option implies
13193 @item -mno-sep-data
13194 Generate code that assumes that the data segment follows the text segment.
13195 This is the default.
13197 @item -mid-shared-library
13198 Generate code that supports shared libraries via the library ID method.
13199 This allows for execute in place and shared libraries in an environment
13200 without virtual memory management. This option implies @option{-fPIC}.
13202 @item -mno-id-shared-library
13203 Generate code that doesn't assume ID based shared libraries are being used.
13204 This is the default.
13206 @item -mshared-library-id=n
13207 Specified the identification number of the ID based shared library being
13208 compiled. Specifying a value of 0 will generate more compact code, specifying
13209 other values will force the allocation of that number to the current
13210 library but is no more space or time efficient than omitting this option.
13216 When generating position-independent code for ColdFire, generate code
13217 that works if the GOT has more than 8192 entries. This code is
13218 larger and slower than code generated without this option. On M680x0
13219 processors, this option is not needed; @option{-fPIC} suffices.
13221 GCC normally uses a single instruction to load values from the GOT@.
13222 While this is relatively efficient, it only works if the GOT
13223 is smaller than about 64k. Anything larger causes the linker
13224 to report an error such as:
13226 @cindex relocation truncated to fit (ColdFire)
13228 relocation truncated to fit: R_68K_GOT16O foobar
13231 If this happens, you should recompile your code with @option{-mxgot}.
13232 It should then work with very large GOTs. However, code generated with
13233 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13234 the value of a global symbol.
13236 Note that some linkers, including newer versions of the GNU linker,
13237 can create multiple GOTs and sort GOT entries. If you have such a linker,
13238 you should only need to use @option{-mxgot} when compiling a single
13239 object file that accesses more than 8192 GOT entries. Very few do.
13241 These options have no effect unless GCC is generating
13242 position-independent code.
13246 @node M68hc1x Options
13247 @subsection M68hc1x Options
13248 @cindex M68hc1x options
13250 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13251 microcontrollers. The default values for these options depends on
13252 which style of microcontroller was selected when the compiler was configured;
13253 the defaults for the most common choices are given below.
13260 Generate output for a 68HC11. This is the default
13261 when the compiler is configured for 68HC11-based systems.
13267 Generate output for a 68HC12. This is the default
13268 when the compiler is configured for 68HC12-based systems.
13274 Generate output for a 68HCS12.
13276 @item -mauto-incdec
13277 @opindex mauto-incdec
13278 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13285 Enable the use of 68HC12 min and max instructions.
13288 @itemx -mno-long-calls
13289 @opindex mlong-calls
13290 @opindex mno-long-calls
13291 Treat all calls as being far away (near). If calls are assumed to be
13292 far away, the compiler will use the @code{call} instruction to
13293 call a function and the @code{rtc} instruction for returning.
13297 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13299 @item -msoft-reg-count=@var{count}
13300 @opindex msoft-reg-count
13301 Specify the number of pseudo-soft registers which are used for the
13302 code generation. The maximum number is 32. Using more pseudo-soft
13303 register may or may not result in better code depending on the program.
13304 The default is 4 for 68HC11 and 2 for 68HC12.
13308 @node MCore Options
13309 @subsection MCore Options
13310 @cindex MCore options
13312 These are the @samp{-m} options defined for the Motorola M*Core
13318 @itemx -mno-hardlit
13320 @opindex mno-hardlit
13321 Inline constants into the code stream if it can be done in two
13322 instructions or less.
13328 Use the divide instruction. (Enabled by default).
13330 @item -mrelax-immediate
13331 @itemx -mno-relax-immediate
13332 @opindex mrelax-immediate
13333 @opindex mno-relax-immediate
13334 Allow arbitrary sized immediates in bit operations.
13336 @item -mwide-bitfields
13337 @itemx -mno-wide-bitfields
13338 @opindex mwide-bitfields
13339 @opindex mno-wide-bitfields
13340 Always treat bit-fields as int-sized.
13342 @item -m4byte-functions
13343 @itemx -mno-4byte-functions
13344 @opindex m4byte-functions
13345 @opindex mno-4byte-functions
13346 Force all functions to be aligned to a four byte boundary.
13348 @item -mcallgraph-data
13349 @itemx -mno-callgraph-data
13350 @opindex mcallgraph-data
13351 @opindex mno-callgraph-data
13352 Emit callgraph information.
13355 @itemx -mno-slow-bytes
13356 @opindex mslow-bytes
13357 @opindex mno-slow-bytes
13358 Prefer word access when reading byte quantities.
13360 @item -mlittle-endian
13361 @itemx -mbig-endian
13362 @opindex mlittle-endian
13363 @opindex mbig-endian
13364 Generate code for a little endian target.
13370 Generate code for the 210 processor.
13374 Assume that run-time support has been provided and so omit the
13375 simulator library (@file{libsim.a)} from the linker command line.
13377 @item -mstack-increment=@var{size}
13378 @opindex mstack-increment
13379 Set the maximum amount for a single stack increment operation. Large
13380 values can increase the speed of programs which contain functions
13381 that need a large amount of stack space, but they can also trigger a
13382 segmentation fault if the stack is extended too much. The default
13388 @subsection MeP Options
13389 @cindex MeP options
13395 Enables the @code{abs} instruction, which is the absolute difference
13396 between two registers.
13400 Enables all the optional instructions - average, multiply, divide, bit
13401 operations, leading zero, absolute difference, min/max, clip, and
13407 Enables the @code{ave} instruction, which computes the average of two
13410 @item -mbased=@var{n}
13412 Variables of size @var{n} bytes or smaller will be placed in the
13413 @code{.based} section by default. Based variables use the @code{$tp}
13414 register as a base register, and there is a 128 byte limit to the
13415 @code{.based} section.
13419 Enables the bit operation instructions - bit test (@code{btstm}), set
13420 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13421 test-and-set (@code{tas}).
13423 @item -mc=@var{name}
13425 Selects which section constant data will be placed in. @var{name} may
13426 be @code{tiny}, @code{near}, or @code{far}.
13430 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13431 useful unless you also provide @code{-mminmax}.
13433 @item -mconfig=@var{name}
13435 Selects one of the build-in core configurations. Each MeP chip has
13436 one or more modules in it; each module has a core CPU and a variety of
13437 coprocessors, optional instructions, and peripherals. The
13438 @code{MeP-Integrator} tool, not part of GCC, provides these
13439 configurations through this option; using this option is the same as
13440 using all the corresponding command line options. The default
13441 configuration is @code{default}.
13445 Enables the coprocessor instructions. By default, this is a 32-bit
13446 coprocessor. Note that the coprocessor is normally enabled via the
13447 @code{-mconfig=} option.
13451 Enables the 32-bit coprocessor's instructions.
13455 Enables the 64-bit coprocessor's instructions.
13459 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13463 Causes constant variables to be placed in the @code{.near} section.
13467 Enables the @code{div} and @code{divu} instructions.
13471 Generate big-endian code.
13475 Generate little-endian code.
13477 @item -mio-volatile
13478 @opindex mio-volatile
13479 Tells the compiler that any variable marked with the @code{io}
13480 attribute is to be considered volatile.
13484 Causes variables to be assigned to the @code{.far} section by default.
13488 Enables the @code{leadz} (leading zero) instruction.
13492 Causes variables to be assigned to the @code{.near} section by default.
13496 Enables the @code{min} and @code{max} instructions.
13500 Enables the multiplication and multiply-accumulate instructions.
13504 Disables all the optional instructions enabled by @code{-mall-opts}.
13508 Enables the @code{repeat} and @code{erepeat} instructions, used for
13509 low-overhead looping.
13513 Causes all variables to default to the @code{.tiny} section. Note
13514 that there is a 65536 byte limit to this section. Accesses to these
13515 variables use the @code{%gp} base register.
13519 Enables the saturation instructions. Note that the compiler does not
13520 currently generate these itself, but this option is included for
13521 compatibility with other tools, like @code{as}.
13525 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13529 Link the simulator runtime libraries.
13533 Link the simulator runtime libraries, excluding built-in support
13534 for reset and exception vectors and tables.
13538 Causes all functions to default to the @code{.far} section. Without
13539 this option, functions default to the @code{.near} section.
13541 @item -mtiny=@var{n}
13543 Variables that are @var{n} bytes or smaller will be allocated to the
13544 @code{.tiny} section. These variables use the @code{$gp} base
13545 register. The default for this option is 4, but note that there's a
13546 65536 byte limit to the @code{.tiny} section.
13551 @subsection MIPS Options
13552 @cindex MIPS options
13558 Generate big-endian code.
13562 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13565 @item -march=@var{arch}
13567 Generate code that will run on @var{arch}, which can be the name of a
13568 generic MIPS ISA, or the name of a particular processor.
13570 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13571 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13572 The processor names are:
13573 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13574 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13575 @samp{5kc}, @samp{5kf},
13577 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13578 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13579 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13580 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13581 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13582 @samp{loongson2e}, @samp{loongson2f},
13586 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13587 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13588 @samp{rm7000}, @samp{rm9000},
13589 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13592 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13593 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13595 The special value @samp{from-abi} selects the
13596 most compatible architecture for the selected ABI (that is,
13597 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13599 Native Linux/GNU toolchains also support the value @samp{native},
13600 which selects the best architecture option for the host processor.
13601 @option{-march=native} has no effect if GCC does not recognize
13604 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13605 (for example, @samp{-march=r2k}). Prefixes are optional, and
13606 @samp{vr} may be written @samp{r}.
13608 Names of the form @samp{@var{n}f2_1} refer to processors with
13609 FPUs clocked at half the rate of the core, names of the form
13610 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13611 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13612 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13613 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13614 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13615 accepted as synonyms for @samp{@var{n}f1_1}.
13617 GCC defines two macros based on the value of this option. The first
13618 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13619 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13620 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13621 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13622 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13624 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13625 above. In other words, it will have the full prefix and will not
13626 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13627 the macro names the resolved architecture (either @samp{"mips1"} or
13628 @samp{"mips3"}). It names the default architecture when no
13629 @option{-march} option is given.
13631 @item -mtune=@var{arch}
13633 Optimize for @var{arch}. Among other things, this option controls
13634 the way instructions are scheduled, and the perceived cost of arithmetic
13635 operations. The list of @var{arch} values is the same as for
13638 When this option is not used, GCC will optimize for the processor
13639 specified by @option{-march}. By using @option{-march} and
13640 @option{-mtune} together, it is possible to generate code that will
13641 run on a family of processors, but optimize the code for one
13642 particular member of that family.
13644 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13645 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13646 @samp{-march} ones described above.
13650 Equivalent to @samp{-march=mips1}.
13654 Equivalent to @samp{-march=mips2}.
13658 Equivalent to @samp{-march=mips3}.
13662 Equivalent to @samp{-march=mips4}.
13666 Equivalent to @samp{-march=mips32}.
13670 Equivalent to @samp{-march=mips32r2}.
13674 Equivalent to @samp{-march=mips64}.
13678 Equivalent to @samp{-march=mips64r2}.
13683 @opindex mno-mips16
13684 Generate (do not generate) MIPS16 code. If GCC is targetting a
13685 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13687 MIPS16 code generation can also be controlled on a per-function basis
13688 by means of @code{mips16} and @code{nomips16} attributes.
13689 @xref{Function Attributes}, for more information.
13691 @item -mflip-mips16
13692 @opindex mflip-mips16
13693 Generate MIPS16 code on alternating functions. This option is provided
13694 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13695 not intended for ordinary use in compiling user code.
13697 @item -minterlink-mips16
13698 @itemx -mno-interlink-mips16
13699 @opindex minterlink-mips16
13700 @opindex mno-interlink-mips16
13701 Require (do not require) that non-MIPS16 code be link-compatible with
13704 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13705 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13706 therefore disables direct jumps unless GCC knows that the target of the
13707 jump is not MIPS16.
13719 Generate code for the given ABI@.
13721 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13722 generates 64-bit code when you select a 64-bit architecture, but you
13723 can use @option{-mgp32} to get 32-bit code instead.
13725 For information about the O64 ABI, see
13726 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13728 GCC supports a variant of the o32 ABI in which floating-point registers
13729 are 64 rather than 32 bits wide. You can select this combination with
13730 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13731 and @samp{mfhc1} instructions and is therefore only supported for
13732 MIPS32R2 processors.
13734 The register assignments for arguments and return values remain the
13735 same, but each scalar value is passed in a single 64-bit register
13736 rather than a pair of 32-bit registers. For example, scalar
13737 floating-point values are returned in @samp{$f0} only, not a
13738 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13739 remains the same, but all 64 bits are saved.
13742 @itemx -mno-abicalls
13744 @opindex mno-abicalls
13745 Generate (do not generate) code that is suitable for SVR4-style
13746 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13751 Generate (do not generate) code that is fully position-independent,
13752 and that can therefore be linked into shared libraries. This option
13753 only affects @option{-mabicalls}.
13755 All @option{-mabicalls} code has traditionally been position-independent,
13756 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13757 as an extension, the GNU toolchain allows executables to use absolute
13758 accesses for locally-binding symbols. It can also use shorter GP
13759 initialization sequences and generate direct calls to locally-defined
13760 functions. This mode is selected by @option{-mno-shared}.
13762 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13763 objects that can only be linked by the GNU linker. However, the option
13764 does not affect the ABI of the final executable; it only affects the ABI
13765 of relocatable objects. Using @option{-mno-shared} will generally make
13766 executables both smaller and quicker.
13768 @option{-mshared} is the default.
13774 Assume (do not assume) that the static and dynamic linkers
13775 support PLTs and copy relocations. This option only affects
13776 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13777 has no effect without @samp{-msym32}.
13779 You can make @option{-mplt} the default by configuring
13780 GCC with @option{--with-mips-plt}. The default is
13781 @option{-mno-plt} otherwise.
13787 Lift (do not lift) the usual restrictions on the size of the global
13790 GCC normally uses a single instruction to load values from the GOT@.
13791 While this is relatively efficient, it will only work if the GOT
13792 is smaller than about 64k. Anything larger will cause the linker
13793 to report an error such as:
13795 @cindex relocation truncated to fit (MIPS)
13797 relocation truncated to fit: R_MIPS_GOT16 foobar
13800 If this happens, you should recompile your code with @option{-mxgot}.
13801 It should then work with very large GOTs, although it will also be
13802 less efficient, since it will take three instructions to fetch the
13803 value of a global symbol.
13805 Note that some linkers can create multiple GOTs. If you have such a
13806 linker, you should only need to use @option{-mxgot} when a single object
13807 file accesses more than 64k's worth of GOT entries. Very few do.
13809 These options have no effect unless GCC is generating position
13814 Assume that general-purpose registers are 32 bits wide.
13818 Assume that general-purpose registers are 64 bits wide.
13822 Assume that floating-point registers are 32 bits wide.
13826 Assume that floating-point registers are 64 bits wide.
13829 @opindex mhard-float
13830 Use floating-point coprocessor instructions.
13833 @opindex msoft-float
13834 Do not use floating-point coprocessor instructions. Implement
13835 floating-point calculations using library calls instead.
13837 @item -msingle-float
13838 @opindex msingle-float
13839 Assume that the floating-point coprocessor only supports single-precision
13842 @item -mdouble-float
13843 @opindex mdouble-float
13844 Assume that the floating-point coprocessor supports double-precision
13845 operations. This is the default.
13851 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13852 implement atomic memory built-in functions. When neither option is
13853 specified, GCC will use the instructions if the target architecture
13856 @option{-mllsc} is useful if the runtime environment can emulate the
13857 instructions and @option{-mno-llsc} can be useful when compiling for
13858 nonstandard ISAs. You can make either option the default by
13859 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13860 respectively. @option{--with-llsc} is the default for some
13861 configurations; see the installation documentation for details.
13867 Use (do not use) revision 1 of the MIPS DSP ASE@.
13868 @xref{MIPS DSP Built-in Functions}. This option defines the
13869 preprocessor macro @samp{__mips_dsp}. It also defines
13870 @samp{__mips_dsp_rev} to 1.
13876 Use (do not use) revision 2 of the MIPS DSP ASE@.
13877 @xref{MIPS DSP Built-in Functions}. This option defines the
13878 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13879 It also defines @samp{__mips_dsp_rev} to 2.
13882 @itemx -mno-smartmips
13883 @opindex msmartmips
13884 @opindex mno-smartmips
13885 Use (do not use) the MIPS SmartMIPS ASE.
13887 @item -mpaired-single
13888 @itemx -mno-paired-single
13889 @opindex mpaired-single
13890 @opindex mno-paired-single
13891 Use (do not use) paired-single floating-point instructions.
13892 @xref{MIPS Paired-Single Support}. This option requires
13893 hardware floating-point support to be enabled.
13899 Use (do not use) MIPS Digital Media Extension instructions.
13900 This option can only be used when generating 64-bit code and requires
13901 hardware floating-point support to be enabled.
13906 @opindex mno-mips3d
13907 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13908 The option @option{-mips3d} implies @option{-mpaired-single}.
13914 Use (do not use) MT Multithreading instructions.
13918 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13919 an explanation of the default and the way that the pointer size is
13924 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13926 The default size of @code{int}s, @code{long}s and pointers depends on
13927 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13928 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13929 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13930 or the same size as integer registers, whichever is smaller.
13936 Assume (do not assume) that all symbols have 32-bit values, regardless
13937 of the selected ABI@. This option is useful in combination with
13938 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13939 to generate shorter and faster references to symbolic addresses.
13943 Put definitions of externally-visible data in a small data section
13944 if that data is no bigger than @var{num} bytes. GCC can then access
13945 the data more efficiently; see @option{-mgpopt} for details.
13947 The default @option{-G} option depends on the configuration.
13949 @item -mlocal-sdata
13950 @itemx -mno-local-sdata
13951 @opindex mlocal-sdata
13952 @opindex mno-local-sdata
13953 Extend (do not extend) the @option{-G} behavior to local data too,
13954 such as to static variables in C@. @option{-mlocal-sdata} is the
13955 default for all configurations.
13957 If the linker complains that an application is using too much small data,
13958 you might want to try rebuilding the less performance-critical parts with
13959 @option{-mno-local-sdata}. You might also want to build large
13960 libraries with @option{-mno-local-sdata}, so that the libraries leave
13961 more room for the main program.
13963 @item -mextern-sdata
13964 @itemx -mno-extern-sdata
13965 @opindex mextern-sdata
13966 @opindex mno-extern-sdata
13967 Assume (do not assume) that externally-defined data will be in
13968 a small data section if that data is within the @option{-G} limit.
13969 @option{-mextern-sdata} is the default for all configurations.
13971 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13972 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13973 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13974 is placed in a small data section. If @var{Var} is defined by another
13975 module, you must either compile that module with a high-enough
13976 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13977 definition. If @var{Var} is common, you must link the application
13978 with a high-enough @option{-G} setting.
13980 The easiest way of satisfying these restrictions is to compile
13981 and link every module with the same @option{-G} option. However,
13982 you may wish to build a library that supports several different
13983 small data limits. You can do this by compiling the library with
13984 the highest supported @option{-G} setting and additionally using
13985 @option{-mno-extern-sdata} to stop the library from making assumptions
13986 about externally-defined data.
13992 Use (do not use) GP-relative accesses for symbols that are known to be
13993 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13994 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13997 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13998 might not hold the value of @code{_gp}. For example, if the code is
13999 part of a library that might be used in a boot monitor, programs that
14000 call boot monitor routines will pass an unknown value in @code{$gp}.
14001 (In such situations, the boot monitor itself would usually be compiled
14002 with @option{-G0}.)
14004 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14005 @option{-mno-extern-sdata}.
14007 @item -membedded-data
14008 @itemx -mno-embedded-data
14009 @opindex membedded-data
14010 @opindex mno-embedded-data
14011 Allocate variables to the read-only data section first if possible, then
14012 next in the small data section if possible, otherwise in data. This gives
14013 slightly slower code than the default, but reduces the amount of RAM required
14014 when executing, and thus may be preferred for some embedded systems.
14016 @item -muninit-const-in-rodata
14017 @itemx -mno-uninit-const-in-rodata
14018 @opindex muninit-const-in-rodata
14019 @opindex mno-uninit-const-in-rodata
14020 Put uninitialized @code{const} variables in the read-only data section.
14021 This option is only meaningful in conjunction with @option{-membedded-data}.
14023 @item -mcode-readable=@var{setting}
14024 @opindex mcode-readable
14025 Specify whether GCC may generate code that reads from executable sections.
14026 There are three possible settings:
14029 @item -mcode-readable=yes
14030 Instructions may freely access executable sections. This is the
14033 @item -mcode-readable=pcrel
14034 MIPS16 PC-relative load instructions can access executable sections,
14035 but other instructions must not do so. This option is useful on 4KSc
14036 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14037 It is also useful on processors that can be configured to have a dual
14038 instruction/data SRAM interface and that, like the M4K, automatically
14039 redirect PC-relative loads to the instruction RAM.
14041 @item -mcode-readable=no
14042 Instructions must not access executable sections. This option can be
14043 useful on targets that are configured to have a dual instruction/data
14044 SRAM interface but that (unlike the M4K) do not automatically redirect
14045 PC-relative loads to the instruction RAM.
14048 @item -msplit-addresses
14049 @itemx -mno-split-addresses
14050 @opindex msplit-addresses
14051 @opindex mno-split-addresses
14052 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14053 relocation operators. This option has been superseded by
14054 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14056 @item -mexplicit-relocs
14057 @itemx -mno-explicit-relocs
14058 @opindex mexplicit-relocs
14059 @opindex mno-explicit-relocs
14060 Use (do not use) assembler relocation operators when dealing with symbolic
14061 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14062 is to use assembler macros instead.
14064 @option{-mexplicit-relocs} is the default if GCC was configured
14065 to use an assembler that supports relocation operators.
14067 @item -mcheck-zero-division
14068 @itemx -mno-check-zero-division
14069 @opindex mcheck-zero-division
14070 @opindex mno-check-zero-division
14071 Trap (do not trap) on integer division by zero.
14073 The default is @option{-mcheck-zero-division}.
14075 @item -mdivide-traps
14076 @itemx -mdivide-breaks
14077 @opindex mdivide-traps
14078 @opindex mdivide-breaks
14079 MIPS systems check for division by zero by generating either a
14080 conditional trap or a break instruction. Using traps results in
14081 smaller code, but is only supported on MIPS II and later. Also, some
14082 versions of the Linux kernel have a bug that prevents trap from
14083 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14084 allow conditional traps on architectures that support them and
14085 @option{-mdivide-breaks} to force the use of breaks.
14087 The default is usually @option{-mdivide-traps}, but this can be
14088 overridden at configure time using @option{--with-divide=breaks}.
14089 Divide-by-zero checks can be completely disabled using
14090 @option{-mno-check-zero-division}.
14095 @opindex mno-memcpy
14096 Force (do not force) the use of @code{memcpy()} for non-trivial block
14097 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14098 most constant-sized copies.
14101 @itemx -mno-long-calls
14102 @opindex mlong-calls
14103 @opindex mno-long-calls
14104 Disable (do not disable) use of the @code{jal} instruction. Calling
14105 functions using @code{jal} is more efficient but requires the caller
14106 and callee to be in the same 256 megabyte segment.
14108 This option has no effect on abicalls code. The default is
14109 @option{-mno-long-calls}.
14115 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14116 instructions, as provided by the R4650 ISA@.
14119 @itemx -mno-fused-madd
14120 @opindex mfused-madd
14121 @opindex mno-fused-madd
14122 Enable (disable) use of the floating point multiply-accumulate
14123 instructions, when they are available. The default is
14124 @option{-mfused-madd}.
14126 When multiply-accumulate instructions are used, the intermediate
14127 product is calculated to infinite precision and is not subject to
14128 the FCSR Flush to Zero bit. This may be undesirable in some
14133 Tell the MIPS assembler to not run its preprocessor over user
14134 assembler files (with a @samp{.s} suffix) when assembling them.
14137 @itemx -mno-fix-r4000
14138 @opindex mfix-r4000
14139 @opindex mno-fix-r4000
14140 Work around certain R4000 CPU errata:
14143 A double-word or a variable shift may give an incorrect result if executed
14144 immediately after starting an integer division.
14146 A double-word or a variable shift may give an incorrect result if executed
14147 while an integer multiplication is in progress.
14149 An integer division may give an incorrect result if started in a delay slot
14150 of a taken branch or a jump.
14154 @itemx -mno-fix-r4400
14155 @opindex mfix-r4400
14156 @opindex mno-fix-r4400
14157 Work around certain R4400 CPU errata:
14160 A double-word or a variable shift may give an incorrect result if executed
14161 immediately after starting an integer division.
14165 @itemx -mno-fix-r10000
14166 @opindex mfix-r10000
14167 @opindex mno-fix-r10000
14168 Work around certain R10000 errata:
14171 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14172 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14175 This option can only be used if the target architecture supports
14176 branch-likely instructions. @option{-mfix-r10000} is the default when
14177 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14181 @itemx -mno-fix-vr4120
14182 @opindex mfix-vr4120
14183 Work around certain VR4120 errata:
14186 @code{dmultu} does not always produce the correct result.
14188 @code{div} and @code{ddiv} do not always produce the correct result if one
14189 of the operands is negative.
14191 The workarounds for the division errata rely on special functions in
14192 @file{libgcc.a}. At present, these functions are only provided by
14193 the @code{mips64vr*-elf} configurations.
14195 Other VR4120 errata require a nop to be inserted between certain pairs of
14196 instructions. These errata are handled by the assembler, not by GCC itself.
14199 @opindex mfix-vr4130
14200 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14201 workarounds are implemented by the assembler rather than by GCC,
14202 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14203 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14204 instructions are available instead.
14207 @itemx -mno-fix-sb1
14209 Work around certain SB-1 CPU core errata.
14210 (This flag currently works around the SB-1 revision 2
14211 ``F1'' and ``F2'' floating point errata.)
14213 @item -mr10k-cache-barrier=@var{setting}
14214 @opindex mr10k-cache-barrier
14215 Specify whether GCC should insert cache barriers to avoid the
14216 side-effects of speculation on R10K processors.
14218 In common with many processors, the R10K tries to predict the outcome
14219 of a conditional branch and speculatively executes instructions from
14220 the ``taken'' branch. It later aborts these instructions if the
14221 predicted outcome was wrong. However, on the R10K, even aborted
14222 instructions can have side effects.
14224 This problem only affects kernel stores and, depending on the system,
14225 kernel loads. As an example, a speculatively-executed store may load
14226 the target memory into cache and mark the cache line as dirty, even if
14227 the store itself is later aborted. If a DMA operation writes to the
14228 same area of memory before the ``dirty'' line is flushed, the cached
14229 data will overwrite the DMA-ed data. See the R10K processor manual
14230 for a full description, including other potential problems.
14232 One workaround is to insert cache barrier instructions before every memory
14233 access that might be speculatively executed and that might have side
14234 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14235 controls GCC's implementation of this workaround. It assumes that
14236 aborted accesses to any byte in the following regions will not have
14241 the memory occupied by the current function's stack frame;
14244 the memory occupied by an incoming stack argument;
14247 the memory occupied by an object with a link-time-constant address.
14250 It is the kernel's responsibility to ensure that speculative
14251 accesses to these regions are indeed safe.
14253 If the input program contains a function declaration such as:
14259 then the implementation of @code{foo} must allow @code{j foo} and
14260 @code{jal foo} to be executed speculatively. GCC honors this
14261 restriction for functions it compiles itself. It expects non-GCC
14262 functions (such as hand-written assembly code) to do the same.
14264 The option has three forms:
14267 @item -mr10k-cache-barrier=load-store
14268 Insert a cache barrier before a load or store that might be
14269 speculatively executed and that might have side effects even
14272 @item -mr10k-cache-barrier=store
14273 Insert a cache barrier before a store that might be speculatively
14274 executed and that might have side effects even if aborted.
14276 @item -mr10k-cache-barrier=none
14277 Disable the insertion of cache barriers. This is the default setting.
14280 @item -mflush-func=@var{func}
14281 @itemx -mno-flush-func
14282 @opindex mflush-func
14283 Specifies the function to call to flush the I and D caches, or to not
14284 call any such function. If called, the function must take the same
14285 arguments as the common @code{_flush_func()}, that is, the address of the
14286 memory range for which the cache is being flushed, the size of the
14287 memory range, and the number 3 (to flush both caches). The default
14288 depends on the target GCC was configured for, but commonly is either
14289 @samp{_flush_func} or @samp{__cpu_flush}.
14291 @item mbranch-cost=@var{num}
14292 @opindex mbranch-cost
14293 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14294 This cost is only a heuristic and is not guaranteed to produce
14295 consistent results across releases. A zero cost redundantly selects
14296 the default, which is based on the @option{-mtune} setting.
14298 @item -mbranch-likely
14299 @itemx -mno-branch-likely
14300 @opindex mbranch-likely
14301 @opindex mno-branch-likely
14302 Enable or disable use of Branch Likely instructions, regardless of the
14303 default for the selected architecture. By default, Branch Likely
14304 instructions may be generated if they are supported by the selected
14305 architecture. An exception is for the MIPS32 and MIPS64 architectures
14306 and processors which implement those architectures; for those, Branch
14307 Likely instructions will not be generated by default because the MIPS32
14308 and MIPS64 architectures specifically deprecate their use.
14310 @item -mfp-exceptions
14311 @itemx -mno-fp-exceptions
14312 @opindex mfp-exceptions
14313 Specifies whether FP exceptions are enabled. This affects how we schedule
14314 FP instructions for some processors. The default is that FP exceptions are
14317 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14318 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14321 @item -mvr4130-align
14322 @itemx -mno-vr4130-align
14323 @opindex mvr4130-align
14324 The VR4130 pipeline is two-way superscalar, but can only issue two
14325 instructions together if the first one is 8-byte aligned. When this
14326 option is enabled, GCC will align pairs of instructions that it
14327 thinks should execute in parallel.
14329 This option only has an effect when optimizing for the VR4130.
14330 It normally makes code faster, but at the expense of making it bigger.
14331 It is enabled by default at optimization level @option{-O3}.
14336 Enable (disable) generation of @code{synci} instructions on
14337 architectures that support it. The @code{synci} instructions (if
14338 enabled) will be generated when @code{__builtin___clear_cache()} is
14341 This option defaults to @code{-mno-synci}, but the default can be
14342 overridden by configuring with @code{--with-synci}.
14344 When compiling code for single processor systems, it is generally safe
14345 to use @code{synci}. However, on many multi-core (SMP) systems, it
14346 will not invalidate the instruction caches on all cores and may lead
14347 to undefined behavior.
14349 @item -mrelax-pic-calls
14350 @itemx -mno-relax-pic-calls
14351 @opindex mrelax-pic-calls
14352 Try to turn PIC calls that are normally dispatched via register
14353 @code{$25} into direct calls. This is only possible if the linker can
14354 resolve the destination at link-time and if the destination is within
14355 range for a direct call.
14357 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14358 an assembler and a linker that supports the @code{.reloc} assembly
14359 directive and @code{-mexplicit-relocs} is in effect. With
14360 @code{-mno-explicit-relocs}, this optimization can be performed by the
14361 assembler and the linker alone without help from the compiler.
14363 @item -mmcount-ra-address
14364 @itemx -mno-mcount-ra-address
14365 @opindex mmcount-ra-address
14366 @opindex mno-mcount-ra-address
14367 Emit (do not emit) code that allows @code{_mcount} to modify the
14368 calling function's return address. When enabled, this option extends
14369 the usual @code{_mcount} interface with a new @var{ra-address}
14370 parameter, which has type @code{intptr_t *} and is passed in register
14371 @code{$12}. @code{_mcount} can then modify the return address by
14372 doing both of the following:
14375 Returning the new address in register @code{$31}.
14377 Storing the new address in @code{*@var{ra-address}},
14378 if @var{ra-address} is nonnull.
14381 The default is @option{-mno-mcount-ra-address}.
14386 @subsection MMIX Options
14387 @cindex MMIX Options
14389 These options are defined for the MMIX:
14393 @itemx -mno-libfuncs
14395 @opindex mno-libfuncs
14396 Specify that intrinsic library functions are being compiled, passing all
14397 values in registers, no matter the size.
14400 @itemx -mno-epsilon
14402 @opindex mno-epsilon
14403 Generate floating-point comparison instructions that compare with respect
14404 to the @code{rE} epsilon register.
14406 @item -mabi=mmixware
14408 @opindex mabi=mmixware
14410 Generate code that passes function parameters and return values that (in
14411 the called function) are seen as registers @code{$0} and up, as opposed to
14412 the GNU ABI which uses global registers @code{$231} and up.
14414 @item -mzero-extend
14415 @itemx -mno-zero-extend
14416 @opindex mzero-extend
14417 @opindex mno-zero-extend
14418 When reading data from memory in sizes shorter than 64 bits, use (do not
14419 use) zero-extending load instructions by default, rather than
14420 sign-extending ones.
14423 @itemx -mno-knuthdiv
14425 @opindex mno-knuthdiv
14426 Make the result of a division yielding a remainder have the same sign as
14427 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14428 remainder follows the sign of the dividend. Both methods are
14429 arithmetically valid, the latter being almost exclusively used.
14431 @item -mtoplevel-symbols
14432 @itemx -mno-toplevel-symbols
14433 @opindex mtoplevel-symbols
14434 @opindex mno-toplevel-symbols
14435 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14436 code can be used with the @code{PREFIX} assembly directive.
14440 Generate an executable in the ELF format, rather than the default
14441 @samp{mmo} format used by the @command{mmix} simulator.
14443 @item -mbranch-predict
14444 @itemx -mno-branch-predict
14445 @opindex mbranch-predict
14446 @opindex mno-branch-predict
14447 Use (do not use) the probable-branch instructions, when static branch
14448 prediction indicates a probable branch.
14450 @item -mbase-addresses
14451 @itemx -mno-base-addresses
14452 @opindex mbase-addresses
14453 @opindex mno-base-addresses
14454 Generate (do not generate) code that uses @emph{base addresses}. Using a
14455 base address automatically generates a request (handled by the assembler
14456 and the linker) for a constant to be set up in a global register. The
14457 register is used for one or more base address requests within the range 0
14458 to 255 from the value held in the register. The generally leads to short
14459 and fast code, but the number of different data items that can be
14460 addressed is limited. This means that a program that uses lots of static
14461 data may require @option{-mno-base-addresses}.
14463 @item -msingle-exit
14464 @itemx -mno-single-exit
14465 @opindex msingle-exit
14466 @opindex mno-single-exit
14467 Force (do not force) generated code to have a single exit point in each
14471 @node MN10300 Options
14472 @subsection MN10300 Options
14473 @cindex MN10300 options
14475 These @option{-m} options are defined for Matsushita MN10300 architectures:
14480 Generate code to avoid bugs in the multiply instructions for the MN10300
14481 processors. This is the default.
14483 @item -mno-mult-bug
14484 @opindex mno-mult-bug
14485 Do not generate code to avoid bugs in the multiply instructions for the
14486 MN10300 processors.
14490 Generate code which uses features specific to the AM33 processor.
14494 Do not generate code which uses features specific to the AM33 processor. This
14497 @item -mreturn-pointer-on-d0
14498 @opindex mreturn-pointer-on-d0
14499 When generating a function which returns a pointer, return the pointer
14500 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14501 only in a0, and attempts to call such functions without a prototype
14502 would result in errors. Note that this option is on by default; use
14503 @option{-mno-return-pointer-on-d0} to disable it.
14507 Do not link in the C run-time initialization object file.
14511 Indicate to the linker that it should perform a relaxation optimization pass
14512 to shorten branches, calls and absolute memory addresses. This option only
14513 has an effect when used on the command line for the final link step.
14515 This option makes symbolic debugging impossible.
14518 @node PDP-11 Options
14519 @subsection PDP-11 Options
14520 @cindex PDP-11 Options
14522 These options are defined for the PDP-11:
14527 Use hardware FPP floating point. This is the default. (FIS floating
14528 point on the PDP-11/40 is not supported.)
14531 @opindex msoft-float
14532 Do not use hardware floating point.
14536 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14540 Return floating-point results in memory. This is the default.
14544 Generate code for a PDP-11/40.
14548 Generate code for a PDP-11/45. This is the default.
14552 Generate code for a PDP-11/10.
14554 @item -mbcopy-builtin
14555 @opindex mbcopy-builtin
14556 Use inline @code{movmemhi} patterns for copying memory. This is the
14561 Do not use inline @code{movmemhi} patterns for copying memory.
14567 Use 16-bit @code{int}. This is the default.
14573 Use 32-bit @code{int}.
14576 @itemx -mno-float32
14578 @opindex mno-float32
14579 Use 64-bit @code{float}. This is the default.
14582 @itemx -mno-float64
14584 @opindex mno-float64
14585 Use 32-bit @code{float}.
14589 Use @code{abshi2} pattern. This is the default.
14593 Do not use @code{abshi2} pattern.
14595 @item -mbranch-expensive
14596 @opindex mbranch-expensive
14597 Pretend that branches are expensive. This is for experimenting with
14598 code generation only.
14600 @item -mbranch-cheap
14601 @opindex mbranch-cheap
14602 Do not pretend that branches are expensive. This is the default.
14606 Generate code for a system with split I&D@.
14610 Generate code for a system without split I&D@. This is the default.
14614 Use Unix assembler syntax. This is the default when configured for
14615 @samp{pdp11-*-bsd}.
14619 Use DEC assembler syntax. This is the default when configured for any
14620 PDP-11 target other than @samp{pdp11-*-bsd}.
14623 @node picoChip Options
14624 @subsection picoChip Options
14625 @cindex picoChip options
14627 These @samp{-m} options are defined for picoChip implementations:
14631 @item -mae=@var{ae_type}
14633 Set the instruction set, register set, and instruction scheduling
14634 parameters for array element type @var{ae_type}. Supported values
14635 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14637 @option{-mae=ANY} selects a completely generic AE type. Code
14638 generated with this option will run on any of the other AE types. The
14639 code will not be as efficient as it would be if compiled for a specific
14640 AE type, and some types of operation (e.g., multiplication) will not
14641 work properly on all types of AE.
14643 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14644 for compiled code, and is the default.
14646 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14647 option may suffer from poor performance of byte (char) manipulation,
14648 since the DSP AE does not provide hardware support for byte load/stores.
14650 @item -msymbol-as-address
14651 Enable the compiler to directly use a symbol name as an address in a
14652 load/store instruction, without first loading it into a
14653 register. Typically, the use of this option will generate larger
14654 programs, which run faster than when the option isn't used. However, the
14655 results vary from program to program, so it is left as a user option,
14656 rather than being permanently enabled.
14658 @item -mno-inefficient-warnings
14659 Disables warnings about the generation of inefficient code. These
14660 warnings can be generated, for example, when compiling code which
14661 performs byte-level memory operations on the MAC AE type. The MAC AE has
14662 no hardware support for byte-level memory operations, so all byte
14663 load/stores must be synthesized from word load/store operations. This is
14664 inefficient and a warning will be generated indicating to the programmer
14665 that they should rewrite the code to avoid byte operations, or to target
14666 an AE type which has the necessary hardware support. This option enables
14667 the warning to be turned off.
14671 @node PowerPC Options
14672 @subsection PowerPC Options
14673 @cindex PowerPC options
14675 These are listed under @xref{RS/6000 and PowerPC Options}.
14677 @node RS/6000 and PowerPC Options
14678 @subsection IBM RS/6000 and PowerPC Options
14679 @cindex RS/6000 and PowerPC Options
14680 @cindex IBM RS/6000 and PowerPC Options
14682 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14689 @itemx -mno-powerpc
14690 @itemx -mpowerpc-gpopt
14691 @itemx -mno-powerpc-gpopt
14692 @itemx -mpowerpc-gfxopt
14693 @itemx -mno-powerpc-gfxopt
14695 @itemx -mno-powerpc64
14699 @itemx -mno-popcntb
14701 @itemx -mno-popcntd
14709 @itemx -mno-hard-dfp
14713 @opindex mno-power2
14715 @opindex mno-powerpc
14716 @opindex mpowerpc-gpopt
14717 @opindex mno-powerpc-gpopt
14718 @opindex mpowerpc-gfxopt
14719 @opindex mno-powerpc-gfxopt
14720 @opindex mpowerpc64
14721 @opindex mno-powerpc64
14725 @opindex mno-popcntb
14727 @opindex mno-popcntd
14733 @opindex mno-mfpgpr
14735 @opindex mno-hard-dfp
14736 GCC supports two related instruction set architectures for the
14737 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14738 instructions supported by the @samp{rios} chip set used in the original
14739 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14740 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14741 the IBM 4xx, 6xx, and follow-on microprocessors.
14743 Neither architecture is a subset of the other. However there is a
14744 large common subset of instructions supported by both. An MQ
14745 register is included in processors supporting the POWER architecture.
14747 You use these options to specify which instructions are available on the
14748 processor you are using. The default value of these options is
14749 determined when configuring GCC@. Specifying the
14750 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14751 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14752 rather than the options listed above.
14754 The @option{-mpower} option allows GCC to generate instructions that
14755 are found only in the POWER architecture and to use the MQ register.
14756 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14757 to generate instructions that are present in the POWER2 architecture but
14758 not the original POWER architecture.
14760 The @option{-mpowerpc} option allows GCC to generate instructions that
14761 are found only in the 32-bit subset of the PowerPC architecture.
14762 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14763 GCC to use the optional PowerPC architecture instructions in the
14764 General Purpose group, including floating-point square root. Specifying
14765 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14766 use the optional PowerPC architecture instructions in the Graphics
14767 group, including floating-point select.
14769 The @option{-mmfcrf} option allows GCC to generate the move from
14770 condition register field instruction implemented on the POWER4
14771 processor and other processors that support the PowerPC V2.01
14773 The @option{-mpopcntb} option allows GCC to generate the popcount and
14774 double precision FP reciprocal estimate instruction implemented on the
14775 POWER5 processor and other processors that support the PowerPC V2.02
14777 The @option{-mpopcntd} option allows GCC to generate the popcount
14778 instruction implemented on the POWER7 processor and other processors
14779 that support the PowerPC V2.06 architecture.
14780 The @option{-mfprnd} option allows GCC to generate the FP round to
14781 integer instructions implemented on the POWER5+ processor and other
14782 processors that support the PowerPC V2.03 architecture.
14783 The @option{-mcmpb} option allows GCC to generate the compare bytes
14784 instruction implemented on the POWER6 processor and other processors
14785 that support the PowerPC V2.05 architecture.
14786 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14787 general purpose register instructions implemented on the POWER6X
14788 processor and other processors that support the extended PowerPC V2.05
14790 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14791 point instructions implemented on some POWER processors.
14793 The @option{-mpowerpc64} option allows GCC to generate the additional
14794 64-bit instructions that are found in the full PowerPC64 architecture
14795 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14796 @option{-mno-powerpc64}.
14798 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14799 will use only the instructions in the common subset of both
14800 architectures plus some special AIX common-mode calls, and will not use
14801 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14802 permits GCC to use any instruction from either architecture and to
14803 allow use of the MQ register; specify this for the Motorola MPC601.
14805 @item -mnew-mnemonics
14806 @itemx -mold-mnemonics
14807 @opindex mnew-mnemonics
14808 @opindex mold-mnemonics
14809 Select which mnemonics to use in the generated assembler code. With
14810 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14811 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14812 assembler mnemonics defined for the POWER architecture. Instructions
14813 defined in only one architecture have only one mnemonic; GCC uses that
14814 mnemonic irrespective of which of these options is specified.
14816 GCC defaults to the mnemonics appropriate for the architecture in
14817 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14818 value of these option. Unless you are building a cross-compiler, you
14819 should normally not specify either @option{-mnew-mnemonics} or
14820 @option{-mold-mnemonics}, but should instead accept the default.
14822 @item -mcpu=@var{cpu_type}
14824 Set architecture type, register usage, choice of mnemonics, and
14825 instruction scheduling parameters for machine type @var{cpu_type}.
14826 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14827 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14828 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
14829 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
14830 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14831 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14832 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
14833 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
14834 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
14835 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14836 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14838 @option{-mcpu=common} selects a completely generic processor. Code
14839 generated under this option will run on any POWER or PowerPC processor.
14840 GCC will use only the instructions in the common subset of both
14841 architectures, and will not use the MQ register. GCC assumes a generic
14842 processor model for scheduling purposes.
14844 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14845 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14846 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14847 types, with an appropriate, generic processor model assumed for
14848 scheduling purposes.
14850 The other options specify a specific processor. Code generated under
14851 those options will run best on that processor, and may not run at all on
14854 The @option{-mcpu} options automatically enable or disable the
14857 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14858 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14859 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14860 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14862 The particular options set for any particular CPU will vary between
14863 compiler versions, depending on what setting seems to produce optimal
14864 code for that CPU; it doesn't necessarily reflect the actual hardware's
14865 capabilities. If you wish to set an individual option to a particular
14866 value, you may specify it after the @option{-mcpu} option, like
14867 @samp{-mcpu=970 -mno-altivec}.
14869 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14870 not enabled or disabled by the @option{-mcpu} option at present because
14871 AIX does not have full support for these options. You may still
14872 enable or disable them individually if you're sure it'll work in your
14875 @item -mtune=@var{cpu_type}
14877 Set the instruction scheduling parameters for machine type
14878 @var{cpu_type}, but do not set the architecture type, register usage, or
14879 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14880 values for @var{cpu_type} are used for @option{-mtune} as for
14881 @option{-mcpu}. If both are specified, the code generated will use the
14882 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14883 scheduling parameters set by @option{-mtune}.
14889 Generate code to compute division as reciprocal estimate and iterative
14890 refinement, creating opportunities for increased throughput. This
14891 feature requires: optional PowerPC Graphics instruction set for single
14892 precision and FRE instruction for double precision, assuming divides
14893 cannot generate user-visible traps, and the domain values not include
14894 Infinities, denormals or zero denominator.
14897 @itemx -mno-altivec
14899 @opindex mno-altivec
14900 Generate code that uses (does not use) AltiVec instructions, and also
14901 enable the use of built-in functions that allow more direct access to
14902 the AltiVec instruction set. You may also need to set
14903 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14909 @opindex mno-vrsave
14910 Generate VRSAVE instructions when generating AltiVec code.
14912 @item -mgen-cell-microcode
14913 @opindex mgen-cell-microcode
14914 Generate Cell microcode instructions
14916 @item -mwarn-cell-microcode
14917 @opindex mwarn-cell-microcode
14918 Warning when a Cell microcode instruction is going to emitted. An example
14919 of a Cell microcode instruction is a variable shift.
14922 @opindex msecure-plt
14923 Generate code that allows ld and ld.so to build executables and shared
14924 libraries with non-exec .plt and .got sections. This is a PowerPC
14925 32-bit SYSV ABI option.
14929 Generate code that uses a BSS .plt section that ld.so fills in, and
14930 requires .plt and .got sections that are both writable and executable.
14931 This is a PowerPC 32-bit SYSV ABI option.
14937 This switch enables or disables the generation of ISEL instructions.
14939 @item -misel=@var{yes/no}
14940 This switch has been deprecated. Use @option{-misel} and
14941 @option{-mno-isel} instead.
14947 This switch enables or disables the generation of SPE simd
14953 @opindex mno-paired
14954 This switch enables or disables the generation of PAIRED simd
14957 @item -mspe=@var{yes/no}
14958 This option has been deprecated. Use @option{-mspe} and
14959 @option{-mno-spe} instead.
14965 Generate code that uses (does not use) vector/scalar (VSX)
14966 instructions, and also enable the use of built-in functions that allow
14967 more direct access to the VSX instruction set.
14969 @item -mfloat-gprs=@var{yes/single/double/no}
14970 @itemx -mfloat-gprs
14971 @opindex mfloat-gprs
14972 This switch enables or disables the generation of floating point
14973 operations on the general purpose registers for architectures that
14976 The argument @var{yes} or @var{single} enables the use of
14977 single-precision floating point operations.
14979 The argument @var{double} enables the use of single and
14980 double-precision floating point operations.
14982 The argument @var{no} disables floating point operations on the
14983 general purpose registers.
14985 This option is currently only available on the MPC854x.
14991 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14992 targets (including GNU/Linux). The 32-bit environment sets int, long
14993 and pointer to 32 bits and generates code that runs on any PowerPC
14994 variant. The 64-bit environment sets int to 32 bits and long and
14995 pointer to 64 bits, and generates code for PowerPC64, as for
14996 @option{-mpowerpc64}.
14999 @itemx -mno-fp-in-toc
15000 @itemx -mno-sum-in-toc
15001 @itemx -mminimal-toc
15003 @opindex mno-fp-in-toc
15004 @opindex mno-sum-in-toc
15005 @opindex mminimal-toc
15006 Modify generation of the TOC (Table Of Contents), which is created for
15007 every executable file. The @option{-mfull-toc} option is selected by
15008 default. In that case, GCC will allocate at least one TOC entry for
15009 each unique non-automatic variable reference in your program. GCC
15010 will also place floating-point constants in the TOC@. However, only
15011 16,384 entries are available in the TOC@.
15013 If you receive a linker error message that saying you have overflowed
15014 the available TOC space, you can reduce the amount of TOC space used
15015 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15016 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15017 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15018 generate code to calculate the sum of an address and a constant at
15019 run-time instead of putting that sum into the TOC@. You may specify one
15020 or both of these options. Each causes GCC to produce very slightly
15021 slower and larger code at the expense of conserving TOC space.
15023 If you still run out of space in the TOC even when you specify both of
15024 these options, specify @option{-mminimal-toc} instead. This option causes
15025 GCC to make only one TOC entry for every file. When you specify this
15026 option, GCC will produce code that is slower and larger but which
15027 uses extremely little TOC space. You may wish to use this option
15028 only on files that contain less frequently executed code.
15034 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15035 @code{long} type, and the infrastructure needed to support them.
15036 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15037 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15038 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15041 @itemx -mno-xl-compat
15042 @opindex mxl-compat
15043 @opindex mno-xl-compat
15044 Produce code that conforms more closely to IBM XL compiler semantics
15045 when using AIX-compatible ABI@. Pass floating-point arguments to
15046 prototyped functions beyond the register save area (RSA) on the stack
15047 in addition to argument FPRs. Do not assume that most significant
15048 double in 128-bit long double value is properly rounded when comparing
15049 values and converting to double. Use XL symbol names for long double
15052 The AIX calling convention was extended but not initially documented to
15053 handle an obscure K&R C case of calling a function that takes the
15054 address of its arguments with fewer arguments than declared. IBM XL
15055 compilers access floating point arguments which do not fit in the
15056 RSA from the stack when a subroutine is compiled without
15057 optimization. Because always storing floating-point arguments on the
15058 stack is inefficient and rarely needed, this option is not enabled by
15059 default and only is necessary when calling subroutines compiled by IBM
15060 XL compilers without optimization.
15064 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15065 application written to use message passing with special startup code to
15066 enable the application to run. The system must have PE installed in the
15067 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15068 must be overridden with the @option{-specs=} option to specify the
15069 appropriate directory location. The Parallel Environment does not
15070 support threads, so the @option{-mpe} option and the @option{-pthread}
15071 option are incompatible.
15073 @item -malign-natural
15074 @itemx -malign-power
15075 @opindex malign-natural
15076 @opindex malign-power
15077 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15078 @option{-malign-natural} overrides the ABI-defined alignment of larger
15079 types, such as floating-point doubles, on their natural size-based boundary.
15080 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15081 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15083 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15087 @itemx -mhard-float
15088 @opindex msoft-float
15089 @opindex mhard-float
15090 Generate code that does not use (uses) the floating-point register set.
15091 Software floating point emulation is provided if you use the
15092 @option{-msoft-float} option, and pass the option to GCC when linking.
15094 @item -msingle-float
15095 @itemx -mdouble-float
15096 @opindex msingle-float
15097 @opindex mdouble-float
15098 Generate code for single or double-precision floating point operations.
15099 @option{-mdouble-float} implies @option{-msingle-float}.
15102 @opindex msimple-fpu
15103 Do not generate sqrt and div instructions for hardware floating point unit.
15107 Specify type of floating point unit. Valid values are @var{sp_lite}
15108 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15109 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15110 and @var{dp_full} (equivalent to -mdouble-float).
15113 @opindex mxilinx-fpu
15114 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15117 @itemx -mno-multiple
15119 @opindex mno-multiple
15120 Generate code that uses (does not use) the load multiple word
15121 instructions and the store multiple word instructions. These
15122 instructions are generated by default on POWER systems, and not
15123 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15124 endian PowerPC systems, since those instructions do not work when the
15125 processor is in little endian mode. The exceptions are PPC740 and
15126 PPC750 which permit the instructions usage in little endian mode.
15131 @opindex mno-string
15132 Generate code that uses (does not use) the load string instructions
15133 and the store string word instructions to save multiple registers and
15134 do small block moves. These instructions are generated by default on
15135 POWER systems, and not generated on PowerPC systems. Do not use
15136 @option{-mstring} on little endian PowerPC systems, since those
15137 instructions do not work when the processor is in little endian mode.
15138 The exceptions are PPC740 and PPC750 which permit the instructions
15139 usage in little endian mode.
15144 @opindex mno-update
15145 Generate code that uses (does not use) the load or store instructions
15146 that update the base register to the address of the calculated memory
15147 location. These instructions are generated by default. If you use
15148 @option{-mno-update}, there is a small window between the time that the
15149 stack pointer is updated and the address of the previous frame is
15150 stored, which means code that walks the stack frame across interrupts or
15151 signals may get corrupted data.
15153 @item -mavoid-indexed-addresses
15154 @itemx -mno-avoid-indexed-addresses
15155 @opindex mavoid-indexed-addresses
15156 @opindex mno-avoid-indexed-addresses
15157 Generate code that tries to avoid (not avoid) the use of indexed load
15158 or store instructions. These instructions can incur a performance
15159 penalty on Power6 processors in certain situations, such as when
15160 stepping through large arrays that cross a 16M boundary. This option
15161 is enabled by default when targetting Power6 and disabled otherwise.
15164 @itemx -mno-fused-madd
15165 @opindex mfused-madd
15166 @opindex mno-fused-madd
15167 Generate code that uses (does not use) the floating point multiply and
15168 accumulate instructions. These instructions are generated by default if
15169 hardware floating is used.
15175 Generate code that uses (does not use) the half-word multiply and
15176 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15177 These instructions are generated by default when targetting those
15184 Generate code that uses (does not use) the string-search @samp{dlmzb}
15185 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15186 generated by default when targetting those processors.
15188 @item -mno-bit-align
15190 @opindex mno-bit-align
15191 @opindex mbit-align
15192 On System V.4 and embedded PowerPC systems do not (do) force structures
15193 and unions that contain bit-fields to be aligned to the base type of the
15196 For example, by default a structure containing nothing but 8
15197 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15198 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15199 the structure would be aligned to a 1 byte boundary and be one byte in
15202 @item -mno-strict-align
15203 @itemx -mstrict-align
15204 @opindex mno-strict-align
15205 @opindex mstrict-align
15206 On System V.4 and embedded PowerPC systems do not (do) assume that
15207 unaligned memory references will be handled by the system.
15209 @item -mrelocatable
15210 @itemx -mno-relocatable
15211 @opindex mrelocatable
15212 @opindex mno-relocatable
15213 On embedded PowerPC systems generate code that allows (does not allow)
15214 the program to be relocated to a different address at runtime. If you
15215 use @option{-mrelocatable} on any module, all objects linked together must
15216 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
15218 @item -mrelocatable-lib
15219 @itemx -mno-relocatable-lib
15220 @opindex mrelocatable-lib
15221 @opindex mno-relocatable-lib
15222 On embedded PowerPC systems generate code that allows (does not allow)
15223 the program to be relocated to a different address at runtime. Modules
15224 compiled with @option{-mrelocatable-lib} can be linked with either modules
15225 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15226 with modules compiled with the @option{-mrelocatable} options.
15232 On System V.4 and embedded PowerPC systems do not (do) assume that
15233 register 2 contains a pointer to a global area pointing to the addresses
15234 used in the program.
15237 @itemx -mlittle-endian
15239 @opindex mlittle-endian
15240 On System V.4 and embedded PowerPC systems compile code for the
15241 processor in little endian mode. The @option{-mlittle-endian} option is
15242 the same as @option{-mlittle}.
15245 @itemx -mbig-endian
15247 @opindex mbig-endian
15248 On System V.4 and embedded PowerPC systems compile code for the
15249 processor in big endian mode. The @option{-mbig-endian} option is
15250 the same as @option{-mbig}.
15252 @item -mdynamic-no-pic
15253 @opindex mdynamic-no-pic
15254 On Darwin and Mac OS X systems, compile code so that it is not
15255 relocatable, but that its external references are relocatable. The
15256 resulting code is suitable for applications, but not shared
15259 @item -mprioritize-restricted-insns=@var{priority}
15260 @opindex mprioritize-restricted-insns
15261 This option controls the priority that is assigned to
15262 dispatch-slot restricted instructions during the second scheduling
15263 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15264 @var{no/highest/second-highest} priority to dispatch slot restricted
15267 @item -msched-costly-dep=@var{dependence_type}
15268 @opindex msched-costly-dep
15269 This option controls which dependences are considered costly
15270 by the target during instruction scheduling. The argument
15271 @var{dependence_type} takes one of the following values:
15272 @var{no}: no dependence is costly,
15273 @var{all}: all dependences are costly,
15274 @var{true_store_to_load}: a true dependence from store to load is costly,
15275 @var{store_to_load}: any dependence from store to load is costly,
15276 @var{number}: any dependence which latency >= @var{number} is costly.
15278 @item -minsert-sched-nops=@var{scheme}
15279 @opindex minsert-sched-nops
15280 This option controls which nop insertion scheme will be used during
15281 the second scheduling pass. The argument @var{scheme} takes one of the
15283 @var{no}: Don't insert nops.
15284 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15285 according to the scheduler's grouping.
15286 @var{regroup_exact}: Insert nops to force costly dependent insns into
15287 separate groups. Insert exactly as many nops as needed to force an insn
15288 to a new group, according to the estimated processor grouping.
15289 @var{number}: Insert nops to force costly dependent insns into
15290 separate groups. Insert @var{number} nops to force an insn to a new group.
15293 @opindex mcall-sysv
15294 On System V.4 and embedded PowerPC systems compile code using calling
15295 conventions that adheres to the March 1995 draft of the System V
15296 Application Binary Interface, PowerPC processor supplement. This is the
15297 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15299 @item -mcall-sysv-eabi
15301 @opindex mcall-sysv-eabi
15302 @opindex mcall-eabi
15303 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15305 @item -mcall-sysv-noeabi
15306 @opindex mcall-sysv-noeabi
15307 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15309 @item -mcall-aixdesc
15311 On System V.4 and embedded PowerPC systems compile code for the AIX
15315 @opindex mcall-linux
15316 On System V.4 and embedded PowerPC systems compile code for the
15317 Linux-based GNU system.
15321 On System V.4 and embedded PowerPC systems compile code for the
15322 Hurd-based GNU system.
15324 @item -mcall-freebsd
15325 @opindex mcall-freebsd
15326 On System V.4 and embedded PowerPC systems compile code for the
15327 FreeBSD operating system.
15329 @item -mcall-netbsd
15330 @opindex mcall-netbsd
15331 On System V.4 and embedded PowerPC systems compile code for the
15332 NetBSD operating system.
15334 @item -mcall-openbsd
15335 @opindex mcall-netbsd
15336 On System V.4 and embedded PowerPC systems compile code for the
15337 OpenBSD operating system.
15339 @item -maix-struct-return
15340 @opindex maix-struct-return
15341 Return all structures in memory (as specified by the AIX ABI)@.
15343 @item -msvr4-struct-return
15344 @opindex msvr4-struct-return
15345 Return structures smaller than 8 bytes in registers (as specified by the
15348 @item -mabi=@var{abi-type}
15350 Extend the current ABI with a particular extension, or remove such extension.
15351 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15352 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15356 Extend the current ABI with SPE ABI extensions. This does not change
15357 the default ABI, instead it adds the SPE ABI extensions to the current
15361 @opindex mabi=no-spe
15362 Disable Booke SPE ABI extensions for the current ABI@.
15364 @item -mabi=ibmlongdouble
15365 @opindex mabi=ibmlongdouble
15366 Change the current ABI to use IBM extended precision long double.
15367 This is a PowerPC 32-bit SYSV ABI option.
15369 @item -mabi=ieeelongdouble
15370 @opindex mabi=ieeelongdouble
15371 Change the current ABI to use IEEE extended precision long double.
15372 This is a PowerPC 32-bit Linux ABI option.
15375 @itemx -mno-prototype
15376 @opindex mprototype
15377 @opindex mno-prototype
15378 On System V.4 and embedded PowerPC systems assume that all calls to
15379 variable argument functions are properly prototyped. Otherwise, the
15380 compiler must insert an instruction before every non prototyped call to
15381 set or clear bit 6 of the condition code register (@var{CR}) to
15382 indicate whether floating point values were passed in the floating point
15383 registers in case the function takes a variable arguments. With
15384 @option{-mprototype}, only calls to prototyped variable argument functions
15385 will set or clear the bit.
15389 On embedded PowerPC systems, assume that the startup module is called
15390 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15391 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15396 On embedded PowerPC systems, assume that the startup module is called
15397 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15402 On embedded PowerPC systems, assume that the startup module is called
15403 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15406 @item -myellowknife
15407 @opindex myellowknife
15408 On embedded PowerPC systems, assume that the startup module is called
15409 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15414 On System V.4 and embedded PowerPC systems, specify that you are
15415 compiling for a VxWorks system.
15419 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15420 header to indicate that @samp{eabi} extended relocations are used.
15426 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15427 Embedded Applications Binary Interface (eabi) which is a set of
15428 modifications to the System V.4 specifications. Selecting @option{-meabi}
15429 means that the stack is aligned to an 8 byte boundary, a function
15430 @code{__eabi} is called to from @code{main} to set up the eabi
15431 environment, and the @option{-msdata} option can use both @code{r2} and
15432 @code{r13} to point to two separate small data areas. Selecting
15433 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15434 do not call an initialization function from @code{main}, and the
15435 @option{-msdata} option will only use @code{r13} to point to a single
15436 small data area. The @option{-meabi} option is on by default if you
15437 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15440 @opindex msdata=eabi
15441 On System V.4 and embedded PowerPC systems, put small initialized
15442 @code{const} global and static data in the @samp{.sdata2} section, which
15443 is pointed to by register @code{r2}. Put small initialized
15444 non-@code{const} global and static data in the @samp{.sdata} section,
15445 which is pointed to by register @code{r13}. Put small uninitialized
15446 global and static data in the @samp{.sbss} section, which is adjacent to
15447 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15448 incompatible with the @option{-mrelocatable} option. The
15449 @option{-msdata=eabi} option also sets the @option{-memb} option.
15452 @opindex msdata=sysv
15453 On System V.4 and embedded PowerPC systems, put small global and static
15454 data in the @samp{.sdata} section, which is pointed to by register
15455 @code{r13}. Put small uninitialized global and static data in the
15456 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15457 The @option{-msdata=sysv} option is incompatible with the
15458 @option{-mrelocatable} option.
15460 @item -msdata=default
15462 @opindex msdata=default
15464 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15465 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15466 same as @option{-msdata=sysv}.
15469 @opindex msdata=data
15470 On System V.4 and embedded PowerPC systems, put small global
15471 data in the @samp{.sdata} section. Put small uninitialized global
15472 data in the @samp{.sbss} section. Do not use register @code{r13}
15473 to address small data however. This is the default behavior unless
15474 other @option{-msdata} options are used.
15478 @opindex msdata=none
15480 On embedded PowerPC systems, put all initialized global and static data
15481 in the @samp{.data} section, and all uninitialized data in the
15482 @samp{.bss} section.
15486 @cindex smaller data references (PowerPC)
15487 @cindex .sdata/.sdata2 references (PowerPC)
15488 On embedded PowerPC systems, put global and static items less than or
15489 equal to @var{num} bytes into the small data or bss sections instead of
15490 the normal data or bss section. By default, @var{num} is 8. The
15491 @option{-G @var{num}} switch is also passed to the linker.
15492 All modules should be compiled with the same @option{-G @var{num}} value.
15495 @itemx -mno-regnames
15497 @opindex mno-regnames
15498 On System V.4 and embedded PowerPC systems do (do not) emit register
15499 names in the assembly language output using symbolic forms.
15502 @itemx -mno-longcall
15504 @opindex mno-longcall
15505 By default assume that all calls are far away so that a longer more
15506 expensive calling sequence is required. This is required for calls
15507 further than 32 megabytes (33,554,432 bytes) from the current location.
15508 A short call will be generated if the compiler knows
15509 the call cannot be that far away. This setting can be overridden by
15510 the @code{shortcall} function attribute, or by @code{#pragma
15513 Some linkers are capable of detecting out-of-range calls and generating
15514 glue code on the fly. On these systems, long calls are unnecessary and
15515 generate slower code. As of this writing, the AIX linker can do this,
15516 as can the GNU linker for PowerPC/64. It is planned to add this feature
15517 to the GNU linker for 32-bit PowerPC systems as well.
15519 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15520 callee, L42'', plus a ``branch island'' (glue code). The two target
15521 addresses represent the callee and the ``branch island''. The
15522 Darwin/PPC linker will prefer the first address and generate a ``bl
15523 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15524 otherwise, the linker will generate ``bl L42'' to call the ``branch
15525 island''. The ``branch island'' is appended to the body of the
15526 calling function; it computes the full 32-bit address of the callee
15529 On Mach-O (Darwin) systems, this option directs the compiler emit to
15530 the glue for every direct call, and the Darwin linker decides whether
15531 to use or discard it.
15533 In the future, we may cause GCC to ignore all longcall specifications
15534 when the linker is known to generate glue.
15536 @item -mtls-markers
15537 @itemx -mno-tls-markers
15538 @opindex mtls-markers
15539 @opindex mno-tls-markers
15540 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15541 specifying the function argument. The relocation allows ld to
15542 reliably associate function call with argument setup instructions for
15543 TLS optimization, which in turn allows gcc to better schedule the
15548 Adds support for multithreading with the @dfn{pthreads} library.
15549 This option sets flags for both the preprocessor and linker.
15554 @subsection RX Options
15557 These command line options are defined for RX targets:
15560 @item -m64bit-doubles
15561 @itemx -m32bit-doubles
15562 @opindex m64bit-doubles
15563 @opindex m32bit-doubles
15564 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
15565 or 32-bits (@option{-m32bit-doubles}) in size. The default is
15566 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
15567 works on 32-bit values, which is why the default is
15568 @option{-m32bit-doubles}.
15574 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
15575 floating point hardware. The default is enabled for the @var{RX600}
15576 series and disabled for the @var{RX200} series.
15578 Floating point instructions will only be generated for 32-bit floating
15579 point values however, so if the @option{-m64bit-doubles} option is in
15580 use then the FPU hardware will not be used for doubles.
15582 @emph{Note} If the @option{-fpu} option is enabled then
15583 @option{-funsafe-math-optimizations} is also enabled automatically.
15584 This is because the RX FPU instructions are themselves unsafe.
15586 @item -mcpu=@var{name}
15587 @itemx -patch=@var{name}
15590 Selects the type of RX CPU to be targeted. Currently three types are
15591 supported, the generic @var{RX600} and @var{RX200} series hardware and
15592 the specific @var{RX610} cpu. The default is @var{RX600}.
15594 The only difference between @var{RX600} and @var{RX610} is that the
15595 @var{RX610} does not support the @code{MVTIPL} instruction.
15597 The @var{RX200} series does not have a hardware floating point unit
15598 and so @option{-nofpu} is enabled by default when this type is
15601 @item -mbig-endian-data
15602 @itemx -mlittle-endian-data
15603 @opindex mbig-endian-data
15604 @opindex mlittle-endian-data
15605 Store data (but not code) in the big-endian format. The default is
15606 @option{-mlittle-endian-data}, ie to store data in the little endian
15609 @item -msmall-data-limit=@var{N}
15610 @opindex msmall-data-limit
15611 Specifies the maximum size in bytes of global and static variables
15612 which can be placed into the small data area. Using the small data
15613 area can lead to smaller and faster code, but the size of area is
15614 limited and it is up to the programmer to ensure that the area does
15615 not overflow. Also when the small data area is used one of the RX's
15616 registers (@code{r13}) is reserved for use pointing to this area, so
15617 it is no longer available for use by the compiler. This could result
15618 in slower and/or larger code if variables which once could have been
15619 held in @code{r13} are now pushed onto the stack.
15621 Note, common variables (variables which have not been initialised) and
15622 constants are not placed into the small data area as they are assigned
15623 to other sections in the output executable.
15625 The default value is zero, which disables this feature. Note, this
15626 feature is not enabled by default with higher optimization levels
15627 (@option{-O2} etc) because of the potentially detrimental effects of
15628 reserving register @code{r13}. It is up to the programmer to
15629 experiment and discover whether this feature is of benefit to their
15636 Use the simulator runtime. The default is to use the libgloss board
15639 @item -mas100-syntax
15640 @itemx -mno-as100-syntax
15641 @opindex mas100-syntax
15642 @opindex mno-as100-syntax
15643 When generating assembler output use a syntax that is compatible with
15644 Renesas's AS100 assembler. This syntax can also be handled by the GAS
15645 assembler but it has some restrictions so generating it is not the
15648 @item -mmax-constant-size=@var{N}
15649 @opindex mmax-constant-size
15650 Specifies the maximum size, in bytes, of a constant that can be used as
15651 an operand in a RX instruction. Although the RX instruction set does
15652 allow constants of up to 4 bytes in length to be used in instructions,
15653 a longer value equates to a longer instruction. Thus in some
15654 circumstances it can be beneficial to restrict the size of constants
15655 that are used in instructions. Constants that are too big are instead
15656 placed into a constant pool and referenced via register indirection.
15658 The value @var{N} can be between 0 and 4. A value of 0 (the default)
15659 or 4 means that constants of any size are allowed.
15663 Enable linker relaxation. Linker relaxation is a process whereby the
15664 linker will attempt to reduce the size of a program by finding shorter
15665 versions of various instructions. Disabled by default.
15667 @item -mint-register=@var{N}
15668 @opindex mint-register
15669 Specify the number of registers to reserve for fast interrupt handler
15670 functions. The value @var{N} can be between 0 and 4. A value of 1
15671 means that register @code{r13} will be reserved for the exclusive use
15672 of fast interrupt handlers. A value of 2 reserves @code{r13} and
15673 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
15674 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
15675 A value of 0, the default, does not reserve any registers.
15677 @item -msave-acc-in-interrupts
15678 @opindex msave-acc-in-interrupts
15679 Specifies that interrupt handler functions should preserve the
15680 accumulator register. This is only necessary if normal code might use
15681 the accumulator register, for example because it performs 64-bit
15682 multiplications. The default is to ignore the accumulator as this
15683 makes the interrupt handlers faster.
15687 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
15688 has special significance to the RX port when used with the
15689 @code{interrupt} function attribute. This attribute indicates a
15690 function intended to process fast interrupts. GCC will will ensure
15691 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
15692 and/or @code{r13} and only provided that the normal use of the
15693 corresponding registers have been restricted via the
15694 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
15697 @node S/390 and zSeries Options
15698 @subsection S/390 and zSeries Options
15699 @cindex S/390 and zSeries Options
15701 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15705 @itemx -msoft-float
15706 @opindex mhard-float
15707 @opindex msoft-float
15708 Use (do not use) the hardware floating-point instructions and registers
15709 for floating-point operations. When @option{-msoft-float} is specified,
15710 functions in @file{libgcc.a} will be used to perform floating-point
15711 operations. When @option{-mhard-float} is specified, the compiler
15712 generates IEEE floating-point instructions. This is the default.
15715 @itemx -mno-hard-dfp
15717 @opindex mno-hard-dfp
15718 Use (do not use) the hardware decimal-floating-point instructions for
15719 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15720 specified, functions in @file{libgcc.a} will be used to perform
15721 decimal-floating-point operations. When @option{-mhard-dfp} is
15722 specified, the compiler generates decimal-floating-point hardware
15723 instructions. This is the default for @option{-march=z9-ec} or higher.
15725 @item -mlong-double-64
15726 @itemx -mlong-double-128
15727 @opindex mlong-double-64
15728 @opindex mlong-double-128
15729 These switches control the size of @code{long double} type. A size
15730 of 64bit makes the @code{long double} type equivalent to the @code{double}
15731 type. This is the default.
15734 @itemx -mno-backchain
15735 @opindex mbackchain
15736 @opindex mno-backchain
15737 Store (do not store) the address of the caller's frame as backchain pointer
15738 into the callee's stack frame.
15739 A backchain may be needed to allow debugging using tools that do not understand
15740 DWARF-2 call frame information.
15741 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15742 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15743 the backchain is placed into the topmost word of the 96/160 byte register
15746 In general, code compiled with @option{-mbackchain} is call-compatible with
15747 code compiled with @option{-mmo-backchain}; however, use of the backchain
15748 for debugging purposes usually requires that the whole binary is built with
15749 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15750 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15751 to build a linux kernel use @option{-msoft-float}.
15753 The default is to not maintain the backchain.
15755 @item -mpacked-stack
15756 @itemx -mno-packed-stack
15757 @opindex mpacked-stack
15758 @opindex mno-packed-stack
15759 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15760 specified, the compiler uses the all fields of the 96/160 byte register save
15761 area only for their default purpose; unused fields still take up stack space.
15762 When @option{-mpacked-stack} is specified, register save slots are densely
15763 packed at the top of the register save area; unused space is reused for other
15764 purposes, allowing for more efficient use of the available stack space.
15765 However, when @option{-mbackchain} is also in effect, the topmost word of
15766 the save area is always used to store the backchain, and the return address
15767 register is always saved two words below the backchain.
15769 As long as the stack frame backchain is not used, code generated with
15770 @option{-mpacked-stack} is call-compatible with code generated with
15771 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15772 S/390 or zSeries generated code that uses the stack frame backchain at run
15773 time, not just for debugging purposes. Such code is not call-compatible
15774 with code compiled with @option{-mpacked-stack}. Also, note that the
15775 combination of @option{-mbackchain},
15776 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15777 to build a linux kernel use @option{-msoft-float}.
15779 The default is to not use the packed stack layout.
15782 @itemx -mno-small-exec
15783 @opindex msmall-exec
15784 @opindex mno-small-exec
15785 Generate (or do not generate) code using the @code{bras} instruction
15786 to do subroutine calls.
15787 This only works reliably if the total executable size does not
15788 exceed 64k. The default is to use the @code{basr} instruction instead,
15789 which does not have this limitation.
15795 When @option{-m31} is specified, generate code compliant to the
15796 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15797 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15798 particular to generate 64-bit instructions. For the @samp{s390}
15799 targets, the default is @option{-m31}, while the @samp{s390x}
15800 targets default to @option{-m64}.
15806 When @option{-mzarch} is specified, generate code using the
15807 instructions available on z/Architecture.
15808 When @option{-mesa} is specified, generate code using the
15809 instructions available on ESA/390. Note that @option{-mesa} is
15810 not possible with @option{-m64}.
15811 When generating code compliant to the GNU/Linux for S/390 ABI,
15812 the default is @option{-mesa}. When generating code compliant
15813 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15819 Generate (or do not generate) code using the @code{mvcle} instruction
15820 to perform block moves. When @option{-mno-mvcle} is specified,
15821 use a @code{mvc} loop instead. This is the default unless optimizing for
15828 Print (or do not print) additional debug information when compiling.
15829 The default is to not print debug information.
15831 @item -march=@var{cpu-type}
15833 Generate code that will run on @var{cpu-type}, which is the name of a system
15834 representing a certain processor type. Possible values for
15835 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15836 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15837 When generating code using the instructions available on z/Architecture,
15838 the default is @option{-march=z900}. Otherwise, the default is
15839 @option{-march=g5}.
15841 @item -mtune=@var{cpu-type}
15843 Tune to @var{cpu-type} everything applicable about the generated code,
15844 except for the ABI and the set of available instructions.
15845 The list of @var{cpu-type} values is the same as for @option{-march}.
15846 The default is the value used for @option{-march}.
15849 @itemx -mno-tpf-trace
15850 @opindex mtpf-trace
15851 @opindex mno-tpf-trace
15852 Generate code that adds (does not add) in TPF OS specific branches to trace
15853 routines in the operating system. This option is off by default, even
15854 when compiling for the TPF OS@.
15857 @itemx -mno-fused-madd
15858 @opindex mfused-madd
15859 @opindex mno-fused-madd
15860 Generate code that uses (does not use) the floating point multiply and
15861 accumulate instructions. These instructions are generated by default if
15862 hardware floating point is used.
15864 @item -mwarn-framesize=@var{framesize}
15865 @opindex mwarn-framesize
15866 Emit a warning if the current function exceeds the given frame size. Because
15867 this is a compile time check it doesn't need to be a real problem when the program
15868 runs. It is intended to identify functions which most probably cause
15869 a stack overflow. It is useful to be used in an environment with limited stack
15870 size e.g.@: the linux kernel.
15872 @item -mwarn-dynamicstack
15873 @opindex mwarn-dynamicstack
15874 Emit a warning if the function calls alloca or uses dynamically
15875 sized arrays. This is generally a bad idea with a limited stack size.
15877 @item -mstack-guard=@var{stack-guard}
15878 @itemx -mstack-size=@var{stack-size}
15879 @opindex mstack-guard
15880 @opindex mstack-size
15881 If these options are provided the s390 back end emits additional instructions in
15882 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15883 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15884 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15885 the frame size of the compiled function is chosen.
15886 These options are intended to be used to help debugging stack overflow problems.
15887 The additionally emitted code causes only little overhead and hence can also be
15888 used in production like systems without greater performance degradation. The given
15889 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15890 @var{stack-guard} without exceeding 64k.
15891 In order to be efficient the extra code makes the assumption that the stack starts
15892 at an address aligned to the value given by @var{stack-size}.
15893 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15896 @node Score Options
15897 @subsection Score Options
15898 @cindex Score Options
15900 These options are defined for Score implementations:
15905 Compile code for big endian mode. This is the default.
15909 Compile code for little endian mode.
15913 Disable generate bcnz instruction.
15917 Enable generate unaligned load and store instruction.
15921 Enable the use of multiply-accumulate instructions. Disabled by default.
15925 Specify the SCORE5 as the target architecture.
15929 Specify the SCORE5U of the target architecture.
15933 Specify the SCORE7 as the target architecture. This is the default.
15937 Specify the SCORE7D as the target architecture.
15941 @subsection SH Options
15943 These @samp{-m} options are defined for the SH implementations:
15948 Generate code for the SH1.
15952 Generate code for the SH2.
15955 Generate code for the SH2e.
15959 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15960 that the floating-point unit is not used.
15962 @item -m2a-single-only
15963 @opindex m2a-single-only
15964 Generate code for the SH2a-FPU, in such a way that no double-precision
15965 floating point operations are used.
15968 @opindex m2a-single
15969 Generate code for the SH2a-FPU assuming the floating-point unit is in
15970 single-precision mode by default.
15974 Generate code for the SH2a-FPU assuming the floating-point unit is in
15975 double-precision mode by default.
15979 Generate code for the SH3.
15983 Generate code for the SH3e.
15987 Generate code for the SH4 without a floating-point unit.
15989 @item -m4-single-only
15990 @opindex m4-single-only
15991 Generate code for the SH4 with a floating-point unit that only
15992 supports single-precision arithmetic.
15996 Generate code for the SH4 assuming the floating-point unit is in
15997 single-precision mode by default.
16001 Generate code for the SH4.
16005 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16006 floating-point unit is not used.
16008 @item -m4a-single-only
16009 @opindex m4a-single-only
16010 Generate code for the SH4a, in such a way that no double-precision
16011 floating point operations are used.
16014 @opindex m4a-single
16015 Generate code for the SH4a assuming the floating-point unit is in
16016 single-precision mode by default.
16020 Generate code for the SH4a.
16024 Same as @option{-m4a-nofpu}, except that it implicitly passes
16025 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16026 instructions at the moment.
16030 Compile code for the processor in big endian mode.
16034 Compile code for the processor in little endian mode.
16038 Align doubles at 64-bit boundaries. Note that this changes the calling
16039 conventions, and thus some functions from the standard C library will
16040 not work unless you recompile it first with @option{-mdalign}.
16044 Shorten some address references at link time, when possible; uses the
16045 linker option @option{-relax}.
16049 Use 32-bit offsets in @code{switch} tables. The default is to use
16054 Enable the use of bit manipulation instructions on SH2A.
16058 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16059 alignment constraints.
16063 Comply with the calling conventions defined by Renesas.
16067 Comply with the calling conventions defined by Renesas.
16071 Comply with the calling conventions defined for GCC before the Renesas
16072 conventions were available. This option is the default for all
16073 targets of the SH toolchain except for @samp{sh-symbianelf}.
16076 @opindex mnomacsave
16077 Mark the @code{MAC} register as call-clobbered, even if
16078 @option{-mhitachi} is given.
16082 Increase IEEE-compliance of floating-point code.
16083 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16084 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16085 comparisons of NANs / infinities incurs extra overhead in every
16086 floating point comparison, therefore the default is set to
16087 @option{-ffinite-math-only}.
16089 @item -minline-ic_invalidate
16090 @opindex minline-ic_invalidate
16091 Inline code to invalidate instruction cache entries after setting up
16092 nested function trampolines.
16093 This option has no effect if -musermode is in effect and the selected
16094 code generation option (e.g. -m4) does not allow the use of the icbi
16096 If the selected code generation option does not allow the use of the icbi
16097 instruction, and -musermode is not in effect, the inlined code will
16098 manipulate the instruction cache address array directly with an associative
16099 write. This not only requires privileged mode, but it will also
16100 fail if the cache line had been mapped via the TLB and has become unmapped.
16104 Dump instruction size and location in the assembly code.
16107 @opindex mpadstruct
16108 This option is deprecated. It pads structures to multiple of 4 bytes,
16109 which is incompatible with the SH ABI@.
16113 Optimize for space instead of speed. Implied by @option{-Os}.
16116 @opindex mprefergot
16117 When generating position-independent code, emit function calls using
16118 the Global Offset Table instead of the Procedure Linkage Table.
16122 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16123 if the inlined code would not work in user mode.
16124 This is the default when the target is @code{sh-*-linux*}.
16126 @item -multcost=@var{number}
16127 @opindex multcost=@var{number}
16128 Set the cost to assume for a multiply insn.
16130 @item -mdiv=@var{strategy}
16131 @opindex mdiv=@var{strategy}
16132 Set the division strategy to use for SHmedia code. @var{strategy} must be
16133 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16134 inv:call2, inv:fp .
16135 "fp" performs the operation in floating point. This has a very high latency,
16136 but needs only a few instructions, so it might be a good choice if
16137 your code has enough easily exploitable ILP to allow the compiler to
16138 schedule the floating point instructions together with other instructions.
16139 Division by zero causes a floating point exception.
16140 "inv" uses integer operations to calculate the inverse of the divisor,
16141 and then multiplies the dividend with the inverse. This strategy allows
16142 cse and hoisting of the inverse calculation. Division by zero calculates
16143 an unspecified result, but does not trap.
16144 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16145 have been found, or if the entire operation has been hoisted to the same
16146 place, the last stages of the inverse calculation are intertwined with the
16147 final multiply to reduce the overall latency, at the expense of using a few
16148 more instructions, and thus offering fewer scheduling opportunities with
16150 "call" calls a library function that usually implements the inv:minlat
16152 This gives high code density for m5-*media-nofpu compilations.
16153 "call2" uses a different entry point of the same library function, where it
16154 assumes that a pointer to a lookup table has already been set up, which
16155 exposes the pointer load to cse / code hoisting optimizations.
16156 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16157 code generation, but if the code stays unoptimized, revert to the "call",
16158 "call2", or "fp" strategies, respectively. Note that the
16159 potentially-trapping side effect of division by zero is carried by a
16160 separate instruction, so it is possible that all the integer instructions
16161 are hoisted out, but the marker for the side effect stays where it is.
16162 A recombination to fp operations or a call is not possible in that case.
16163 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16164 that the inverse calculation was nor separated from the multiply, they speed
16165 up division where the dividend fits into 20 bits (plus sign where applicable),
16166 by inserting a test to skip a number of operations in this case; this test
16167 slows down the case of larger dividends. inv20u assumes the case of a such
16168 a small dividend to be unlikely, and inv20l assumes it to be likely.
16170 @item -maccumulate-outgoing-args
16171 @opindex maccumulate-outgoing-args
16172 Reserve space once for outgoing arguments in the function prologue rather
16173 than around each call. Generally beneficial for performance and size. Also
16174 needed for unwinding to avoid changing the stack frame around conditional code.
16176 @item -mdivsi3_libfunc=@var{name}
16177 @opindex mdivsi3_libfunc=@var{name}
16178 Set the name of the library function used for 32 bit signed division to
16179 @var{name}. This only affect the name used in the call and inv:call
16180 division strategies, and the compiler will still expect the same
16181 sets of input/output/clobbered registers as if this option was not present.
16183 @item -mfixed-range=@var{register-range}
16184 @opindex mfixed-range
16185 Generate code treating the given register range as fixed registers.
16186 A fixed register is one that the register allocator can not use. This is
16187 useful when compiling kernel code. A register range is specified as
16188 two registers separated by a dash. Multiple register ranges can be
16189 specified separated by a comma.
16191 @item -madjust-unroll
16192 @opindex madjust-unroll
16193 Throttle unrolling to avoid thrashing target registers.
16194 This option only has an effect if the gcc code base supports the
16195 TARGET_ADJUST_UNROLL_MAX target hook.
16197 @item -mindexed-addressing
16198 @opindex mindexed-addressing
16199 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16200 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16201 semantics for the indexed addressing mode. The architecture allows the
16202 implementation of processors with 64 bit MMU, which the OS could use to
16203 get 32 bit addressing, but since no current hardware implementation supports
16204 this or any other way to make the indexed addressing mode safe to use in
16205 the 32 bit ABI, the default is -mno-indexed-addressing.
16207 @item -mgettrcost=@var{number}
16208 @opindex mgettrcost=@var{number}
16209 Set the cost assumed for the gettr instruction to @var{number}.
16210 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16214 Assume pt* instructions won't trap. This will generally generate better
16215 scheduled code, but is unsafe on current hardware. The current architecture
16216 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16217 This has the unintentional effect of making it unsafe to schedule ptabs /
16218 ptrel before a branch, or hoist it out of a loop. For example,
16219 __do_global_ctors, a part of libgcc that runs constructors at program
16220 startup, calls functions in a list which is delimited by @minus{}1. With the
16221 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16222 That means that all the constructors will be run a bit quicker, but when
16223 the loop comes to the end of the list, the program crashes because ptabs
16224 loads @minus{}1 into a target register. Since this option is unsafe for any
16225 hardware implementing the current architecture specification, the default
16226 is -mno-pt-fixed. Unless the user specifies a specific cost with
16227 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16228 this deters register allocation using target registers for storing
16231 @item -minvalid-symbols
16232 @opindex minvalid-symbols
16233 Assume symbols might be invalid. Ordinary function symbols generated by
16234 the compiler will always be valid to load with movi/shori/ptabs or
16235 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16236 to generate symbols that will cause ptabs / ptrel to trap.
16237 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16238 It will then prevent cross-basic-block cse, hoisting and most scheduling
16239 of symbol loads. The default is @option{-mno-invalid-symbols}.
16242 @node SPARC Options
16243 @subsection SPARC Options
16244 @cindex SPARC options
16246 These @samp{-m} options are supported on the SPARC:
16249 @item -mno-app-regs
16251 @opindex mno-app-regs
16253 Specify @option{-mapp-regs} to generate output using the global registers
16254 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16257 To be fully SVR4 ABI compliant at the cost of some performance loss,
16258 specify @option{-mno-app-regs}. You should compile libraries and system
16259 software with this option.
16262 @itemx -mhard-float
16264 @opindex mhard-float
16265 Generate output containing floating point instructions. This is the
16269 @itemx -msoft-float
16271 @opindex msoft-float
16272 Generate output containing library calls for floating point.
16273 @strong{Warning:} the requisite libraries are not available for all SPARC
16274 targets. Normally the facilities of the machine's usual C compiler are
16275 used, but this cannot be done directly in cross-compilation. You must make
16276 your own arrangements to provide suitable library functions for
16277 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16278 @samp{sparclite-*-*} do provide software floating point support.
16280 @option{-msoft-float} changes the calling convention in the output file;
16281 therefore, it is only useful if you compile @emph{all} of a program with
16282 this option. In particular, you need to compile @file{libgcc.a}, the
16283 library that comes with GCC, with @option{-msoft-float} in order for
16286 @item -mhard-quad-float
16287 @opindex mhard-quad-float
16288 Generate output containing quad-word (long double) floating point
16291 @item -msoft-quad-float
16292 @opindex msoft-quad-float
16293 Generate output containing library calls for quad-word (long double)
16294 floating point instructions. The functions called are those specified
16295 in the SPARC ABI@. This is the default.
16297 As of this writing, there are no SPARC implementations that have hardware
16298 support for the quad-word floating point instructions. They all invoke
16299 a trap handler for one of these instructions, and then the trap handler
16300 emulates the effect of the instruction. Because of the trap handler overhead,
16301 this is much slower than calling the ABI library routines. Thus the
16302 @option{-msoft-quad-float} option is the default.
16304 @item -mno-unaligned-doubles
16305 @itemx -munaligned-doubles
16306 @opindex mno-unaligned-doubles
16307 @opindex munaligned-doubles
16308 Assume that doubles have 8 byte alignment. This is the default.
16310 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16311 alignment only if they are contained in another type, or if they have an
16312 absolute address. Otherwise, it assumes they have 4 byte alignment.
16313 Specifying this option avoids some rare compatibility problems with code
16314 generated by other compilers. It is not the default because it results
16315 in a performance loss, especially for floating point code.
16317 @item -mno-faster-structs
16318 @itemx -mfaster-structs
16319 @opindex mno-faster-structs
16320 @opindex mfaster-structs
16321 With @option{-mfaster-structs}, the compiler assumes that structures
16322 should have 8 byte alignment. This enables the use of pairs of
16323 @code{ldd} and @code{std} instructions for copies in structure
16324 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16325 However, the use of this changed alignment directly violates the SPARC
16326 ABI@. Thus, it's intended only for use on targets where the developer
16327 acknowledges that their resulting code will not be directly in line with
16328 the rules of the ABI@.
16330 @item -mimpure-text
16331 @opindex mimpure-text
16332 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16333 the compiler to not pass @option{-z text} to the linker when linking a
16334 shared object. Using this option, you can link position-dependent
16335 code into a shared object.
16337 @option{-mimpure-text} suppresses the ``relocations remain against
16338 allocatable but non-writable sections'' linker error message.
16339 However, the necessary relocations will trigger copy-on-write, and the
16340 shared object is not actually shared across processes. Instead of
16341 using @option{-mimpure-text}, you should compile all source code with
16342 @option{-fpic} or @option{-fPIC}.
16344 This option is only available on SunOS and Solaris.
16346 @item -mcpu=@var{cpu_type}
16348 Set the instruction set, register set, and instruction scheduling parameters
16349 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16350 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16351 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16352 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16353 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16355 Default instruction scheduling parameters are used for values that select
16356 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16357 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16359 Here is a list of each supported architecture and their supported
16364 v8: supersparc, hypersparc
16365 sparclite: f930, f934, sparclite86x
16367 v9: ultrasparc, ultrasparc3, niagara, niagara2
16370 By default (unless configured otherwise), GCC generates code for the V7
16371 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16372 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16373 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16374 SPARCStation 1, 2, IPX etc.
16376 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16377 architecture. The only difference from V7 code is that the compiler emits
16378 the integer multiply and integer divide instructions which exist in SPARC-V8
16379 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16380 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16383 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16384 the SPARC architecture. This adds the integer multiply, integer divide step
16385 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16386 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16387 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16388 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16389 MB86934 chip, which is the more recent SPARClite with FPU@.
16391 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16392 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16393 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16394 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16395 optimizes it for the TEMIC SPARClet chip.
16397 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16398 architecture. This adds 64-bit integer and floating-point move instructions,
16399 3 additional floating-point condition code registers and conditional move
16400 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16401 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16402 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16403 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16404 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16405 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16406 additionally optimizes it for Sun UltraSPARC T2 chips.
16408 @item -mtune=@var{cpu_type}
16410 Set the instruction scheduling parameters for machine type
16411 @var{cpu_type}, but do not set the instruction set or register set that the
16412 option @option{-mcpu=@var{cpu_type}} would.
16414 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16415 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16416 that select a particular cpu implementation. Those are @samp{cypress},
16417 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16418 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16419 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16424 @opindex mno-v8plus
16425 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16426 difference from the V8 ABI is that the global and out registers are
16427 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16428 mode for all SPARC-V9 processors.
16434 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16435 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16438 These @samp{-m} options are supported in addition to the above
16439 on SPARC-V9 processors in 64-bit environments:
16442 @item -mlittle-endian
16443 @opindex mlittle-endian
16444 Generate code for a processor running in little-endian mode. It is only
16445 available for a few configurations and most notably not on Solaris and Linux.
16451 Generate code for a 32-bit or 64-bit environment.
16452 The 32-bit environment sets int, long and pointer to 32 bits.
16453 The 64-bit environment sets int to 32 bits and long and pointer
16456 @item -mcmodel=medlow
16457 @opindex mcmodel=medlow
16458 Generate code for the Medium/Low code model: 64-bit addresses, programs
16459 must be linked in the low 32 bits of memory. Programs can be statically
16460 or dynamically linked.
16462 @item -mcmodel=medmid
16463 @opindex mcmodel=medmid
16464 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16465 must be linked in the low 44 bits of memory, the text and data segments must
16466 be less than 2GB in size and the data segment must be located within 2GB of
16469 @item -mcmodel=medany
16470 @opindex mcmodel=medany
16471 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16472 may be linked anywhere in memory, the text and data segments must be less
16473 than 2GB in size and the data segment must be located within 2GB of the
16476 @item -mcmodel=embmedany
16477 @opindex mcmodel=embmedany
16478 Generate code for the Medium/Anywhere code model for embedded systems:
16479 64-bit addresses, the text and data segments must be less than 2GB in
16480 size, both starting anywhere in memory (determined at link time). The
16481 global register %g4 points to the base of the data segment. Programs
16482 are statically linked and PIC is not supported.
16485 @itemx -mno-stack-bias
16486 @opindex mstack-bias
16487 @opindex mno-stack-bias
16488 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16489 frame pointer if present, are offset by @minus{}2047 which must be added back
16490 when making stack frame references. This is the default in 64-bit mode.
16491 Otherwise, assume no such offset is present.
16494 These switches are supported in addition to the above on Solaris:
16499 Add support for multithreading using the Solaris threads library. This
16500 option sets flags for both the preprocessor and linker. This option does
16501 not affect the thread safety of object code produced by the compiler or
16502 that of libraries supplied with it.
16506 Add support for multithreading using the POSIX threads library. This
16507 option sets flags for both the preprocessor and linker. This option does
16508 not affect the thread safety of object code produced by the compiler or
16509 that of libraries supplied with it.
16513 This is a synonym for @option{-pthreads}.
16517 @subsection SPU Options
16518 @cindex SPU options
16520 These @samp{-m} options are supported on the SPU:
16524 @itemx -merror-reloc
16525 @opindex mwarn-reloc
16526 @opindex merror-reloc
16528 The loader for SPU does not handle dynamic relocations. By default, GCC
16529 will give an error when it generates code that requires a dynamic
16530 relocation. @option{-mno-error-reloc} disables the error,
16531 @option{-mwarn-reloc} will generate a warning instead.
16534 @itemx -munsafe-dma
16536 @opindex munsafe-dma
16538 Instructions which initiate or test completion of DMA must not be
16539 reordered with respect to loads and stores of the memory which is being
16540 accessed. Users typically address this problem using the volatile
16541 keyword, but that can lead to inefficient code in places where the
16542 memory is known to not change. Rather than mark the memory as volatile
16543 we treat the DMA instructions as potentially effecting all memory. With
16544 @option{-munsafe-dma} users must use the volatile keyword to protect
16547 @item -mbranch-hints
16548 @opindex mbranch-hints
16550 By default, GCC will generate a branch hint instruction to avoid
16551 pipeline stalls for always taken or probably taken branches. A hint
16552 will not be generated closer than 8 instructions away from its branch.
16553 There is little reason to disable them, except for debugging purposes,
16554 or to make an object a little bit smaller.
16558 @opindex msmall-mem
16559 @opindex mlarge-mem
16561 By default, GCC generates code assuming that addresses are never larger
16562 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16563 a full 32 bit address.
16568 By default, GCC links against startup code that assumes the SPU-style
16569 main function interface (which has an unconventional parameter list).
16570 With @option{-mstdmain}, GCC will link your program against startup
16571 code that assumes a C99-style interface to @code{main}, including a
16572 local copy of @code{argv} strings.
16574 @item -mfixed-range=@var{register-range}
16575 @opindex mfixed-range
16576 Generate code treating the given register range as fixed registers.
16577 A fixed register is one that the register allocator can not use. This is
16578 useful when compiling kernel code. A register range is specified as
16579 two registers separated by a dash. Multiple register ranges can be
16580 specified separated by a comma.
16586 Compile code assuming that pointers to the PPU address space accessed
16587 via the @code{__ea} named address space qualifier are either 32 or 64
16588 bits wide. The default is 32 bits. As this is an ABI changing option,
16589 all object code in an executable must be compiled with the same setting.
16591 @item -maddress-space-conversion
16592 @itemx -mno-address-space-conversion
16593 @opindex maddress-space-conversion
16594 @opindex mno-address-space-conversion
16595 Allow/disallow treating the @code{__ea} address space as superset
16596 of the generic address space. This enables explicit type casts
16597 between @code{__ea} and generic pointer as well as implicit
16598 conversions of generic pointers to @code{__ea} pointers. The
16599 default is to allow address space pointer conversions.
16601 @item -mcache-size=@var{cache-size}
16602 @opindex mcache-size
16603 This option controls the version of libgcc that the compiler links to an
16604 executable and selects a software-managed cache for accessing variables
16605 in the @code{__ea} address space with a particular cache size. Possible
16606 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
16607 and @samp{128}. The default cache size is 64KB.
16609 @item -matomic-updates
16610 @itemx -mno-atomic-updates
16611 @opindex matomic-updates
16612 @opindex mno-atomic-updates
16613 This option controls the version of libgcc that the compiler links to an
16614 executable and selects whether atomic updates to the software-managed
16615 cache of PPU-side variables are used. If you use atomic updates, changes
16616 to a PPU variable from SPU code using the @code{__ea} named address space
16617 qualifier will not interfere with changes to other PPU variables residing
16618 in the same cache line from PPU code. If you do not use atomic updates,
16619 such interference may occur; however, writing back cache lines will be
16620 more efficient. The default behavior is to use atomic updates.
16623 @itemx -mdual-nops=@var{n}
16624 @opindex mdual-nops
16625 By default, GCC will insert nops to increase dual issue when it expects
16626 it to increase performance. @var{n} can be a value from 0 to 10. A
16627 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16628 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16630 @item -mhint-max-nops=@var{n}
16631 @opindex mhint-max-nops
16632 Maximum number of nops to insert for a branch hint. A branch hint must
16633 be at least 8 instructions away from the branch it is effecting. GCC
16634 will insert up to @var{n} nops to enforce this, otherwise it will not
16635 generate the branch hint.
16637 @item -mhint-max-distance=@var{n}
16638 @opindex mhint-max-distance
16639 The encoding of the branch hint instruction limits the hint to be within
16640 256 instructions of the branch it is effecting. By default, GCC makes
16641 sure it is within 125.
16644 @opindex msafe-hints
16645 Work around a hardware bug which causes the SPU to stall indefinitely.
16646 By default, GCC will insert the @code{hbrp} instruction to make sure
16647 this stall won't happen.
16651 @node System V Options
16652 @subsection Options for System V
16654 These additional options are available on System V Release 4 for
16655 compatibility with other compilers on those systems:
16660 Create a shared object.
16661 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16665 Identify the versions of each tool used by the compiler, in a
16666 @code{.ident} assembler directive in the output.
16670 Refrain from adding @code{.ident} directives to the output file (this is
16673 @item -YP,@var{dirs}
16675 Search the directories @var{dirs}, and no others, for libraries
16676 specified with @option{-l}.
16678 @item -Ym,@var{dir}
16680 Look in the directory @var{dir} to find the M4 preprocessor.
16681 The assembler uses this option.
16682 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16683 @c the generic assembler that comes with Solaris takes just -Ym.
16687 @subsection V850 Options
16688 @cindex V850 Options
16690 These @samp{-m} options are defined for V850 implementations:
16694 @itemx -mno-long-calls
16695 @opindex mlong-calls
16696 @opindex mno-long-calls
16697 Treat all calls as being far away (near). If calls are assumed to be
16698 far away, the compiler will always load the functions address up into a
16699 register, and call indirect through the pointer.
16705 Do not optimize (do optimize) basic blocks that use the same index
16706 pointer 4 or more times to copy pointer into the @code{ep} register, and
16707 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16708 option is on by default if you optimize.
16710 @item -mno-prolog-function
16711 @itemx -mprolog-function
16712 @opindex mno-prolog-function
16713 @opindex mprolog-function
16714 Do not use (do use) external functions to save and restore registers
16715 at the prologue and epilogue of a function. The external functions
16716 are slower, but use less code space if more than one function saves
16717 the same number of registers. The @option{-mprolog-function} option
16718 is on by default if you optimize.
16722 Try to make the code as small as possible. At present, this just turns
16723 on the @option{-mep} and @option{-mprolog-function} options.
16725 @item -mtda=@var{n}
16727 Put static or global variables whose size is @var{n} bytes or less into
16728 the tiny data area that register @code{ep} points to. The tiny data
16729 area can hold up to 256 bytes in total (128 bytes for byte references).
16731 @item -msda=@var{n}
16733 Put static or global variables whose size is @var{n} bytes or less into
16734 the small data area that register @code{gp} points to. The small data
16735 area can hold up to 64 kilobytes.
16737 @item -mzda=@var{n}
16739 Put static or global variables whose size is @var{n} bytes or less into
16740 the first 32 kilobytes of memory.
16744 Specify that the target processor is the V850.
16747 @opindex mbig-switch
16748 Generate code suitable for big switch tables. Use this option only if
16749 the assembler/linker complain about out of range branches within a switch
16754 This option will cause r2 and r5 to be used in the code generated by
16755 the compiler. This setting is the default.
16757 @item -mno-app-regs
16758 @opindex mno-app-regs
16759 This option will cause r2 and r5 to be treated as fixed registers.
16763 Specify that the target processor is the V850E1. The preprocessor
16764 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16765 this option is used.
16769 Specify that the target processor is the V850E@. The preprocessor
16770 constant @samp{__v850e__} will be defined if this option is used.
16772 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16773 are defined then a default target processor will be chosen and the
16774 relevant @samp{__v850*__} preprocessor constant will be defined.
16776 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16777 defined, regardless of which processor variant is the target.
16779 @item -mdisable-callt
16780 @opindex mdisable-callt
16781 This option will suppress generation of the CALLT instruction for the
16782 v850e and v850e1 flavors of the v850 architecture. The default is
16783 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16788 @subsection VAX Options
16789 @cindex VAX options
16791 These @samp{-m} options are defined for the VAX:
16796 Do not output certain jump instructions (@code{aobleq} and so on)
16797 that the Unix assembler for the VAX cannot handle across long
16802 Do output those jump instructions, on the assumption that you
16803 will assemble with the GNU assembler.
16807 Output code for g-format floating point numbers instead of d-format.
16810 @node VxWorks Options
16811 @subsection VxWorks Options
16812 @cindex VxWorks Options
16814 The options in this section are defined for all VxWorks targets.
16815 Options specific to the target hardware are listed with the other
16816 options for that target.
16821 GCC can generate code for both VxWorks kernels and real time processes
16822 (RTPs). This option switches from the former to the latter. It also
16823 defines the preprocessor macro @code{__RTP__}.
16826 @opindex non-static
16827 Link an RTP executable against shared libraries rather than static
16828 libraries. The options @option{-static} and @option{-shared} can
16829 also be used for RTPs (@pxref{Link Options}); @option{-static}
16836 These options are passed down to the linker. They are defined for
16837 compatibility with Diab.
16840 @opindex Xbind-lazy
16841 Enable lazy binding of function calls. This option is equivalent to
16842 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16846 Disable lazy binding of function calls. This option is the default and
16847 is defined for compatibility with Diab.
16850 @node x86-64 Options
16851 @subsection x86-64 Options
16852 @cindex x86-64 options
16854 These are listed under @xref{i386 and x86-64 Options}.
16856 @node i386 and x86-64 Windows Options
16857 @subsection i386 and x86-64 Windows Options
16858 @cindex i386 and x86-64 Windows Options
16860 These additional options are available for Windows targets:
16865 This option is available for Cygwin and MinGW targets. It
16866 specifies that a console application is to be generated, by
16867 instructing the linker to set the PE header subsystem type
16868 required for console applications.
16869 This is the default behavior for Cygwin and MinGW targets.
16873 This option is available for Cygwin targets. It specifies that
16874 the Cygwin internal interface is to be used for predefined
16875 preprocessor macros, C runtime libraries and related linker
16876 paths and options. For Cygwin targets this is the default behavior.
16877 This option is deprecated and will be removed in a future release.
16880 @opindex mno-cygwin
16881 This option is available for Cygwin targets. It specifies that
16882 the MinGW internal interface is to be used instead of Cygwin's, by
16883 setting MinGW-related predefined macros and linker paths and default
16885 This option is deprecated and will be removed in a future release.
16889 This option is available for Cygwin and MinGW targets. It
16890 specifies that a DLL - a dynamic link library - is to be
16891 generated, enabling the selection of the required runtime
16892 startup object and entry point.
16894 @item -mnop-fun-dllimport
16895 @opindex mnop-fun-dllimport
16896 This option is available for Cygwin and MinGW targets. It
16897 specifies that the dllimport attribute should be ignored.
16901 This option is available for MinGW targets. It specifies
16902 that MinGW-specific thread support is to be used.
16906 This option is available for mingw-w64 targets. It specifies
16907 that the UNICODE macro is getting pre-defined and that the
16908 unicode capable runtime startup code is chosen.
16912 This option is available for Cygwin and MinGW targets. It
16913 specifies that the typical Windows pre-defined macros are to
16914 be set in the pre-processor, but does not influence the choice
16915 of runtime library/startup code.
16919 This option is available for Cygwin and MinGW targets. It
16920 specifies that a GUI application is to be generated by
16921 instructing the linker to set the PE header subsystem type
16924 @item -fno-set-stack-executable
16925 @opindex fno-set-stack-executable
16926 This option is available for MinGW targets. It specifies that
16927 the executable flag for stack used by nested functions isn't
16928 set. This is necessary for binaries running in kernel mode of
16929 Windows, as there the user32 API, which is used to set executable
16930 privileges, isn't available.
16932 @item -mpe-aligned-commons
16933 @opindex mpe-aligned-commons
16934 This option is available for Cygwin and MinGW targets. It
16935 specifies that the GNU extension to the PE file format that
16936 permits the correct alignment of COMMON variables should be
16937 used when generating code. It will be enabled by default if
16938 GCC detects that the target assembler found during configuration
16939 supports the feature.
16942 See also under @ref{i386 and x86-64 Options} for standard options.
16944 @node Xstormy16 Options
16945 @subsection Xstormy16 Options
16946 @cindex Xstormy16 Options
16948 These options are defined for Xstormy16:
16953 Choose startup files and linker script suitable for the simulator.
16956 @node Xtensa Options
16957 @subsection Xtensa Options
16958 @cindex Xtensa Options
16960 These options are supported for Xtensa targets:
16964 @itemx -mno-const16
16966 @opindex mno-const16
16967 Enable or disable use of @code{CONST16} instructions for loading
16968 constant values. The @code{CONST16} instruction is currently not a
16969 standard option from Tensilica. When enabled, @code{CONST16}
16970 instructions are always used in place of the standard @code{L32R}
16971 instructions. The use of @code{CONST16} is enabled by default only if
16972 the @code{L32R} instruction is not available.
16975 @itemx -mno-fused-madd
16976 @opindex mfused-madd
16977 @opindex mno-fused-madd
16978 Enable or disable use of fused multiply/add and multiply/subtract
16979 instructions in the floating-point option. This has no effect if the
16980 floating-point option is not also enabled. Disabling fused multiply/add
16981 and multiply/subtract instructions forces the compiler to use separate
16982 instructions for the multiply and add/subtract operations. This may be
16983 desirable in some cases where strict IEEE 754-compliant results are
16984 required: the fused multiply add/subtract instructions do not round the
16985 intermediate result, thereby producing results with @emph{more} bits of
16986 precision than specified by the IEEE standard. Disabling fused multiply
16987 add/subtract instructions also ensures that the program output is not
16988 sensitive to the compiler's ability to combine multiply and add/subtract
16991 @item -mserialize-volatile
16992 @itemx -mno-serialize-volatile
16993 @opindex mserialize-volatile
16994 @opindex mno-serialize-volatile
16995 When this option is enabled, GCC inserts @code{MEMW} instructions before
16996 @code{volatile} memory references to guarantee sequential consistency.
16997 The default is @option{-mserialize-volatile}. Use
16998 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17000 @item -mtext-section-literals
17001 @itemx -mno-text-section-literals
17002 @opindex mtext-section-literals
17003 @opindex mno-text-section-literals
17004 Control the treatment of literal pools. The default is
17005 @option{-mno-text-section-literals}, which places literals in a separate
17006 section in the output file. This allows the literal pool to be placed
17007 in a data RAM/ROM, and it also allows the linker to combine literal
17008 pools from separate object files to remove redundant literals and
17009 improve code size. With @option{-mtext-section-literals}, the literals
17010 are interspersed in the text section in order to keep them as close as
17011 possible to their references. This may be necessary for large assembly
17014 @item -mtarget-align
17015 @itemx -mno-target-align
17016 @opindex mtarget-align
17017 @opindex mno-target-align
17018 When this option is enabled, GCC instructs the assembler to
17019 automatically align instructions to reduce branch penalties at the
17020 expense of some code density. The assembler attempts to widen density
17021 instructions to align branch targets and the instructions following call
17022 instructions. If there are not enough preceding safe density
17023 instructions to align a target, no widening will be performed. The
17024 default is @option{-mtarget-align}. These options do not affect the
17025 treatment of auto-aligned instructions like @code{LOOP}, which the
17026 assembler will always align, either by widening density instructions or
17027 by inserting no-op instructions.
17030 @itemx -mno-longcalls
17031 @opindex mlongcalls
17032 @opindex mno-longcalls
17033 When this option is enabled, GCC instructs the assembler to translate
17034 direct calls to indirect calls unless it can determine that the target
17035 of a direct call is in the range allowed by the call instruction. This
17036 translation typically occurs for calls to functions in other source
17037 files. Specifically, the assembler translates a direct @code{CALL}
17038 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17039 The default is @option{-mno-longcalls}. This option should be used in
17040 programs where the call target can potentially be out of range. This
17041 option is implemented in the assembler, not the compiler, so the
17042 assembly code generated by GCC will still show direct call
17043 instructions---look at the disassembled object code to see the actual
17044 instructions. Note that the assembler will use an indirect call for
17045 every cross-file call, not just those that really will be out of range.
17048 @node zSeries Options
17049 @subsection zSeries Options
17050 @cindex zSeries options
17052 These are listed under @xref{S/390 and zSeries Options}.
17054 @node Code Gen Options
17055 @section Options for Code Generation Conventions
17056 @cindex code generation conventions
17057 @cindex options, code generation
17058 @cindex run-time options
17060 These machine-independent options control the interface conventions
17061 used in code generation.
17063 Most of them have both positive and negative forms; the negative form
17064 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17065 one of the forms is listed---the one which is not the default. You
17066 can figure out the other form by either removing @samp{no-} or adding
17070 @item -fbounds-check
17071 @opindex fbounds-check
17072 For front-ends that support it, generate additional code to check that
17073 indices used to access arrays are within the declared range. This is
17074 currently only supported by the Java and Fortran front-ends, where
17075 this option defaults to true and false respectively.
17079 This option generates traps for signed overflow on addition, subtraction,
17080 multiplication operations.
17084 This option instructs the compiler to assume that signed arithmetic
17085 overflow of addition, subtraction and multiplication wraps around
17086 using twos-complement representation. This flag enables some optimizations
17087 and disables others. This option is enabled by default for the Java
17088 front-end, as required by the Java language specification.
17091 @opindex fexceptions
17092 Enable exception handling. Generates extra code needed to propagate
17093 exceptions. For some targets, this implies GCC will generate frame
17094 unwind information for all functions, which can produce significant data
17095 size overhead, although it does not affect execution. If you do not
17096 specify this option, GCC will enable it by default for languages like
17097 C++ which normally require exception handling, and disable it for
17098 languages like C that do not normally require it. However, you may need
17099 to enable this option when compiling C code that needs to interoperate
17100 properly with exception handlers written in C++. You may also wish to
17101 disable this option if you are compiling older C++ programs that don't
17102 use exception handling.
17104 @item -fnon-call-exceptions
17105 @opindex fnon-call-exceptions
17106 Generate code that allows trapping instructions to throw exceptions.
17107 Note that this requires platform-specific runtime support that does
17108 not exist everywhere. Moreover, it only allows @emph{trapping}
17109 instructions to throw exceptions, i.e.@: memory references or floating
17110 point instructions. It does not allow exceptions to be thrown from
17111 arbitrary signal handlers such as @code{SIGALRM}.
17113 @item -funwind-tables
17114 @opindex funwind-tables
17115 Similar to @option{-fexceptions}, except that it will just generate any needed
17116 static data, but will not affect the generated code in any other way.
17117 You will normally not enable this option; instead, a language processor
17118 that needs this handling would enable it on your behalf.
17120 @item -fasynchronous-unwind-tables
17121 @opindex fasynchronous-unwind-tables
17122 Generate unwind table in dwarf2 format, if supported by target machine. The
17123 table is exact at each instruction boundary, so it can be used for stack
17124 unwinding from asynchronous events (such as debugger or garbage collector).
17126 @item -fpcc-struct-return
17127 @opindex fpcc-struct-return
17128 Return ``short'' @code{struct} and @code{union} values in memory like
17129 longer ones, rather than in registers. This convention is less
17130 efficient, but it has the advantage of allowing intercallability between
17131 GCC-compiled files and files compiled with other compilers, particularly
17132 the Portable C Compiler (pcc).
17134 The precise convention for returning structures in memory depends
17135 on the target configuration macros.
17137 Short structures and unions are those whose size and alignment match
17138 that of some integer type.
17140 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17141 switch is not binary compatible with code compiled with the
17142 @option{-freg-struct-return} switch.
17143 Use it to conform to a non-default application binary interface.
17145 @item -freg-struct-return
17146 @opindex freg-struct-return
17147 Return @code{struct} and @code{union} values in registers when possible.
17148 This is more efficient for small structures than
17149 @option{-fpcc-struct-return}.
17151 If you specify neither @option{-fpcc-struct-return} nor
17152 @option{-freg-struct-return}, GCC defaults to whichever convention is
17153 standard for the target. If there is no standard convention, GCC
17154 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17155 the principal compiler. In those cases, we can choose the standard, and
17156 we chose the more efficient register return alternative.
17158 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17159 switch is not binary compatible with code compiled with the
17160 @option{-fpcc-struct-return} switch.
17161 Use it to conform to a non-default application binary interface.
17163 @item -fshort-enums
17164 @opindex fshort-enums
17165 Allocate to an @code{enum} type only as many bytes as it needs for the
17166 declared range of possible values. Specifically, the @code{enum} type
17167 will be equivalent to the smallest integer type which has enough room.
17169 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17170 code that is not binary compatible with code generated without that switch.
17171 Use it to conform to a non-default application binary interface.
17173 @item -fshort-double
17174 @opindex fshort-double
17175 Use the same size for @code{double} as for @code{float}.
17177 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17178 code that is not binary compatible with code generated without that switch.
17179 Use it to conform to a non-default application binary interface.
17181 @item -fshort-wchar
17182 @opindex fshort-wchar
17183 Override the underlying type for @samp{wchar_t} to be @samp{short
17184 unsigned int} instead of the default for the target. This option is
17185 useful for building programs to run under WINE@.
17187 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17188 code that is not binary compatible with code generated without that switch.
17189 Use it to conform to a non-default application binary interface.
17192 @opindex fno-common
17193 In C code, controls the placement of uninitialized global variables.
17194 Unix C compilers have traditionally permitted multiple definitions of
17195 such variables in different compilation units by placing the variables
17197 This is the behavior specified by @option{-fcommon}, and is the default
17198 for GCC on most targets.
17199 On the other hand, this behavior is not required by ISO C, and on some
17200 targets may carry a speed or code size penalty on variable references.
17201 The @option{-fno-common} option specifies that the compiler should place
17202 uninitialized global variables in the data section of the object file,
17203 rather than generating them as common blocks.
17204 This has the effect that if the same variable is declared
17205 (without @code{extern}) in two different compilations,
17206 you will get a multiple-definition error when you link them.
17207 In this case, you must compile with @option{-fcommon} instead.
17208 Compiling with @option{-fno-common} is useful on targets for which
17209 it provides better performance, or if you wish to verify that the
17210 program will work on other systems which always treat uninitialized
17211 variable declarations this way.
17215 Ignore the @samp{#ident} directive.
17217 @item -finhibit-size-directive
17218 @opindex finhibit-size-directive
17219 Don't output a @code{.size} assembler directive, or anything else that
17220 would cause trouble if the function is split in the middle, and the
17221 two halves are placed at locations far apart in memory. This option is
17222 used when compiling @file{crtstuff.c}; you should not need to use it
17225 @item -fverbose-asm
17226 @opindex fverbose-asm
17227 Put extra commentary information in the generated assembly code to
17228 make it more readable. This option is generally only of use to those
17229 who actually need to read the generated assembly code (perhaps while
17230 debugging the compiler itself).
17232 @option{-fno-verbose-asm}, the default, causes the
17233 extra information to be omitted and is useful when comparing two assembler
17236 @item -frecord-gcc-switches
17237 @opindex frecord-gcc-switches
17238 This switch causes the command line that was used to invoke the
17239 compiler to be recorded into the object file that is being created.
17240 This switch is only implemented on some targets and the exact format
17241 of the recording is target and binary file format dependent, but it
17242 usually takes the form of a section containing ASCII text. This
17243 switch is related to the @option{-fverbose-asm} switch, but that
17244 switch only records information in the assembler output file as
17245 comments, so it never reaches the object file.
17249 @cindex global offset table
17251 Generate position-independent code (PIC) suitable for use in a shared
17252 library, if supported for the target machine. Such code accesses all
17253 constant addresses through a global offset table (GOT)@. The dynamic
17254 loader resolves the GOT entries when the program starts (the dynamic
17255 loader is not part of GCC; it is part of the operating system). If
17256 the GOT size for the linked executable exceeds a machine-specific
17257 maximum size, you get an error message from the linker indicating that
17258 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17259 instead. (These maximums are 8k on the SPARC and 32k
17260 on the m68k and RS/6000. The 386 has no such limit.)
17262 Position-independent code requires special support, and therefore works
17263 only on certain machines. For the 386, GCC supports PIC for System V
17264 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17265 position-independent.
17267 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17272 If supported for the target machine, emit position-independent code,
17273 suitable for dynamic linking and avoiding any limit on the size of the
17274 global offset table. This option makes a difference on the m68k,
17275 PowerPC and SPARC@.
17277 Position-independent code requires special support, and therefore works
17278 only on certain machines.
17280 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17287 These options are similar to @option{-fpic} and @option{-fPIC}, but
17288 generated position independent code can be only linked into executables.
17289 Usually these options are used when @option{-pie} GCC option will be
17290 used during linking.
17292 @option{-fpie} and @option{-fPIE} both define the macros
17293 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17294 for @option{-fpie} and 2 for @option{-fPIE}.
17296 @item -fno-jump-tables
17297 @opindex fno-jump-tables
17298 Do not use jump tables for switch statements even where it would be
17299 more efficient than other code generation strategies. This option is
17300 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17301 building code which forms part of a dynamic linker and cannot
17302 reference the address of a jump table. On some targets, jump tables
17303 do not require a GOT and this option is not needed.
17305 @item -ffixed-@var{reg}
17307 Treat the register named @var{reg} as a fixed register; generated code
17308 should never refer to it (except perhaps as a stack pointer, frame
17309 pointer or in some other fixed role).
17311 @var{reg} must be the name of a register. The register names accepted
17312 are machine-specific and are defined in the @code{REGISTER_NAMES}
17313 macro in the machine description macro file.
17315 This flag does not have a negative form, because it specifies a
17318 @item -fcall-used-@var{reg}
17319 @opindex fcall-used
17320 Treat the register named @var{reg} as an allocable register that is
17321 clobbered by function calls. It may be allocated for temporaries or
17322 variables that do not live across a call. Functions compiled this way
17323 will not save and restore the register @var{reg}.
17325 It is an error to used this flag with the frame pointer or stack pointer.
17326 Use of this flag for other registers that have fixed pervasive roles in
17327 the machine's execution model will produce disastrous results.
17329 This flag does not have a negative form, because it specifies a
17332 @item -fcall-saved-@var{reg}
17333 @opindex fcall-saved
17334 Treat the register named @var{reg} as an allocable register saved by
17335 functions. It may be allocated even for temporaries or variables that
17336 live across a call. Functions compiled this way will save and restore
17337 the register @var{reg} if they use it.
17339 It is an error to used this flag with the frame pointer or stack pointer.
17340 Use of this flag for other registers that have fixed pervasive roles in
17341 the machine's execution model will produce disastrous results.
17343 A different sort of disaster will result from the use of this flag for
17344 a register in which function values may be returned.
17346 This flag does not have a negative form, because it specifies a
17349 @item -fpack-struct[=@var{n}]
17350 @opindex fpack-struct
17351 Without a value specified, pack all structure members together without
17352 holes. When a value is specified (which must be a small power of two), pack
17353 structure members according to this value, representing the maximum
17354 alignment (that is, objects with default alignment requirements larger than
17355 this will be output potentially unaligned at the next fitting location.
17357 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17358 code that is not binary compatible with code generated without that switch.
17359 Additionally, it makes the code suboptimal.
17360 Use it to conform to a non-default application binary interface.
17362 @item -finstrument-functions
17363 @opindex finstrument-functions
17364 Generate instrumentation calls for entry and exit to functions. Just
17365 after function entry and just before function exit, the following
17366 profiling functions will be called with the address of the current
17367 function and its call site. (On some platforms,
17368 @code{__builtin_return_address} does not work beyond the current
17369 function, so the call site information may not be available to the
17370 profiling functions otherwise.)
17373 void __cyg_profile_func_enter (void *this_fn,
17375 void __cyg_profile_func_exit (void *this_fn,
17379 The first argument is the address of the start of the current function,
17380 which may be looked up exactly in the symbol table.
17382 This instrumentation is also done for functions expanded inline in other
17383 functions. The profiling calls will indicate where, conceptually, the
17384 inline function is entered and exited. This means that addressable
17385 versions of such functions must be available. If all your uses of a
17386 function are expanded inline, this may mean an additional expansion of
17387 code size. If you use @samp{extern inline} in your C code, an
17388 addressable version of such functions must be provided. (This is
17389 normally the case anyways, but if you get lucky and the optimizer always
17390 expands the functions inline, you might have gotten away without
17391 providing static copies.)
17393 A function may be given the attribute @code{no_instrument_function}, in
17394 which case this instrumentation will not be done. This can be used, for
17395 example, for the profiling functions listed above, high-priority
17396 interrupt routines, and any functions from which the profiling functions
17397 cannot safely be called (perhaps signal handlers, if the profiling
17398 routines generate output or allocate memory).
17400 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17401 @opindex finstrument-functions-exclude-file-list
17403 Set the list of functions that are excluded from instrumentation (see
17404 the description of @code{-finstrument-functions}). If the file that
17405 contains a function definition matches with one of @var{file}, then
17406 that function is not instrumented. The match is done on substrings:
17407 if the @var{file} parameter is a substring of the file name, it is
17408 considered to be a match.
17411 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
17412 will exclude any inline function defined in files whose pathnames
17413 contain @code{/bits/stl} or @code{include/sys}.
17415 If, for some reason, you want to include letter @code{','} in one of
17416 @var{sym}, write @code{'\,'}. For example,
17417 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17418 (note the single quote surrounding the option).
17420 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17421 @opindex finstrument-functions-exclude-function-list
17423 This is similar to @code{-finstrument-functions-exclude-file-list},
17424 but this option sets the list of function names to be excluded from
17425 instrumentation. The function name to be matched is its user-visible
17426 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17427 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17428 match is done on substrings: if the @var{sym} parameter is a substring
17429 of the function name, it is considered to be a match. For C99 and C++
17430 extended identifiers, the function name must be given in UTF-8, not
17431 using universal character names.
17433 @item -fstack-check
17434 @opindex fstack-check
17435 Generate code to verify that you do not go beyond the boundary of the
17436 stack. You should specify this flag if you are running in an
17437 environment with multiple threads, but only rarely need to specify it in
17438 a single-threaded environment since stack overflow is automatically
17439 detected on nearly all systems if there is only one stack.
17441 Note that this switch does not actually cause checking to be done; the
17442 operating system or the language runtime must do that. The switch causes
17443 generation of code to ensure that they see the stack being extended.
17445 You can additionally specify a string parameter: @code{no} means no
17446 checking, @code{generic} means force the use of old-style checking,
17447 @code{specific} means use the best checking method and is equivalent
17448 to bare @option{-fstack-check}.
17450 Old-style checking is a generic mechanism that requires no specific
17451 target support in the compiler but comes with the following drawbacks:
17455 Modified allocation strategy for large objects: they will always be
17456 allocated dynamically if their size exceeds a fixed threshold.
17459 Fixed limit on the size of the static frame of functions: when it is
17460 topped by a particular function, stack checking is not reliable and
17461 a warning is issued by the compiler.
17464 Inefficiency: because of both the modified allocation strategy and the
17465 generic implementation, the performances of the code are hampered.
17468 Note that old-style stack checking is also the fallback method for
17469 @code{specific} if no target support has been added in the compiler.
17471 @item -fstack-limit-register=@var{reg}
17472 @itemx -fstack-limit-symbol=@var{sym}
17473 @itemx -fno-stack-limit
17474 @opindex fstack-limit-register
17475 @opindex fstack-limit-symbol
17476 @opindex fno-stack-limit
17477 Generate code to ensure that the stack does not grow beyond a certain value,
17478 either the value of a register or the address of a symbol. If the stack
17479 would grow beyond the value, a signal is raised. For most targets,
17480 the signal is raised before the stack overruns the boundary, so
17481 it is possible to catch the signal without taking special precautions.
17483 For instance, if the stack starts at absolute address @samp{0x80000000}
17484 and grows downwards, you can use the flags
17485 @option{-fstack-limit-symbol=__stack_limit} and
17486 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17487 of 128KB@. Note that this may only work with the GNU linker.
17489 @item -fleading-underscore
17490 @opindex fleading-underscore
17491 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17492 change the way C symbols are represented in the object file. One use
17493 is to help link with legacy assembly code.
17495 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17496 generate code that is not binary compatible with code generated without that
17497 switch. Use it to conform to a non-default application binary interface.
17498 Not all targets provide complete support for this switch.
17500 @item -ftls-model=@var{model}
17501 @opindex ftls-model
17502 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17503 The @var{model} argument should be one of @code{global-dynamic},
17504 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17506 The default without @option{-fpic} is @code{initial-exec}; with
17507 @option{-fpic} the default is @code{global-dynamic}.
17509 @item -fvisibility=@var{default|internal|hidden|protected}
17510 @opindex fvisibility
17511 Set the default ELF image symbol visibility to the specified option---all
17512 symbols will be marked with this unless overridden within the code.
17513 Using this feature can very substantially improve linking and
17514 load times of shared object libraries, produce more optimized
17515 code, provide near-perfect API export and prevent symbol clashes.
17516 It is @strong{strongly} recommended that you use this in any shared objects
17519 Despite the nomenclature, @code{default} always means public ie;
17520 available to be linked against from outside the shared object.
17521 @code{protected} and @code{internal} are pretty useless in real-world
17522 usage so the only other commonly used option will be @code{hidden}.
17523 The default if @option{-fvisibility} isn't specified is
17524 @code{default}, i.e., make every
17525 symbol public---this causes the same behavior as previous versions of
17528 A good explanation of the benefits offered by ensuring ELF
17529 symbols have the correct visibility is given by ``How To Write
17530 Shared Libraries'' by Ulrich Drepper (which can be found at
17531 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17532 solution made possible by this option to marking things hidden when
17533 the default is public is to make the default hidden and mark things
17534 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17535 and @code{__attribute__ ((visibility("default")))} instead of
17536 @code{__declspec(dllexport)} you get almost identical semantics with
17537 identical syntax. This is a great boon to those working with
17538 cross-platform projects.
17540 For those adding visibility support to existing code, you may find
17541 @samp{#pragma GCC visibility} of use. This works by you enclosing
17542 the declarations you wish to set visibility for with (for example)
17543 @samp{#pragma GCC visibility push(hidden)} and
17544 @samp{#pragma GCC visibility pop}.
17545 Bear in mind that symbol visibility should be viewed @strong{as
17546 part of the API interface contract} and thus all new code should
17547 always specify visibility when it is not the default ie; declarations
17548 only for use within the local DSO should @strong{always} be marked explicitly
17549 as hidden as so to avoid PLT indirection overheads---making this
17550 abundantly clear also aids readability and self-documentation of the code.
17551 Note that due to ISO C++ specification requirements, operator new and
17552 operator delete must always be of default visibility.
17554 Be aware that headers from outside your project, in particular system
17555 headers and headers from any other library you use, may not be
17556 expecting to be compiled with visibility other than the default. You
17557 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17558 before including any such headers.
17560 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17561 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17562 no modifications. However, this means that calls to @samp{extern}
17563 functions with no explicit visibility will use the PLT, so it is more
17564 effective to use @samp{__attribute ((visibility))} and/or
17565 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17566 declarations should be treated as hidden.
17568 Note that @samp{-fvisibility} does affect C++ vague linkage
17569 entities. This means that, for instance, an exception class that will
17570 be thrown between DSOs must be explicitly marked with default
17571 visibility so that the @samp{type_info} nodes will be unified between
17574 An overview of these techniques, their benefits and how to use them
17575 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17581 @node Environment Variables
17582 @section Environment Variables Affecting GCC
17583 @cindex environment variables
17585 @c man begin ENVIRONMENT
17586 This section describes several environment variables that affect how GCC
17587 operates. Some of them work by specifying directories or prefixes to use
17588 when searching for various kinds of files. Some are used to specify other
17589 aspects of the compilation environment.
17591 Note that you can also specify places to search using options such as
17592 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17593 take precedence over places specified using environment variables, which
17594 in turn take precedence over those specified by the configuration of GCC@.
17595 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17596 GNU Compiler Collection (GCC) Internals}.
17601 @c @itemx LC_COLLATE
17603 @c @itemx LC_MONETARY
17604 @c @itemx LC_NUMERIC
17609 @c @findex LC_COLLATE
17610 @findex LC_MESSAGES
17611 @c @findex LC_MONETARY
17612 @c @findex LC_NUMERIC
17616 These environment variables control the way that GCC uses
17617 localization information that allow GCC to work with different
17618 national conventions. GCC inspects the locale categories
17619 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17620 so. These locale categories can be set to any value supported by your
17621 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17622 Kingdom encoded in UTF-8.
17624 The @env{LC_CTYPE} environment variable specifies character
17625 classification. GCC uses it to determine the character boundaries in
17626 a string; this is needed for some multibyte encodings that contain quote
17627 and escape characters that would otherwise be interpreted as a string
17630 The @env{LC_MESSAGES} environment variable specifies the language to
17631 use in diagnostic messages.
17633 If the @env{LC_ALL} environment variable is set, it overrides the value
17634 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17635 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17636 environment variable. If none of these variables are set, GCC
17637 defaults to traditional C English behavior.
17641 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17642 files. GCC uses temporary files to hold the output of one stage of
17643 compilation which is to be used as input to the next stage: for example,
17644 the output of the preprocessor, which is the input to the compiler
17647 @item GCC_EXEC_PREFIX
17648 @findex GCC_EXEC_PREFIX
17649 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17650 names of the subprograms executed by the compiler. No slash is added
17651 when this prefix is combined with the name of a subprogram, but you can
17652 specify a prefix that ends with a slash if you wish.
17654 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17655 an appropriate prefix to use based on the pathname it was invoked with.
17657 If GCC cannot find the subprogram using the specified prefix, it
17658 tries looking in the usual places for the subprogram.
17660 The default value of @env{GCC_EXEC_PREFIX} is
17661 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17662 the installed compiler. In many cases @var{prefix} is the value
17663 of @code{prefix} when you ran the @file{configure} script.
17665 Other prefixes specified with @option{-B} take precedence over this prefix.
17667 This prefix is also used for finding files such as @file{crt0.o} that are
17670 In addition, the prefix is used in an unusual way in finding the
17671 directories to search for header files. For each of the standard
17672 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17673 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17674 replacing that beginning with the specified prefix to produce an
17675 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17676 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17677 These alternate directories are searched first; the standard directories
17678 come next. If a standard directory begins with the configured
17679 @var{prefix} then the value of @var{prefix} is replaced by
17680 @env{GCC_EXEC_PREFIX} when looking for header files.
17682 @item COMPILER_PATH
17683 @findex COMPILER_PATH
17684 The value of @env{COMPILER_PATH} is a colon-separated list of
17685 directories, much like @env{PATH}. GCC tries the directories thus
17686 specified when searching for subprograms, if it can't find the
17687 subprograms using @env{GCC_EXEC_PREFIX}.
17690 @findex LIBRARY_PATH
17691 The value of @env{LIBRARY_PATH} is a colon-separated list of
17692 directories, much like @env{PATH}. When configured as a native compiler,
17693 GCC tries the directories thus specified when searching for special
17694 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17695 using GCC also uses these directories when searching for ordinary
17696 libraries for the @option{-l} option (but directories specified with
17697 @option{-L} come first).
17701 @cindex locale definition
17702 This variable is used to pass locale information to the compiler. One way in
17703 which this information is used is to determine the character set to be used
17704 when character literals, string literals and comments are parsed in C and C++.
17705 When the compiler is configured to allow multibyte characters,
17706 the following values for @env{LANG} are recognized:
17710 Recognize JIS characters.
17712 Recognize SJIS characters.
17714 Recognize EUCJP characters.
17717 If @env{LANG} is not defined, or if it has some other value, then the
17718 compiler will use mblen and mbtowc as defined by the default locale to
17719 recognize and translate multibyte characters.
17723 Some additional environments variables affect the behavior of the
17726 @include cppenv.texi
17730 @node Precompiled Headers
17731 @section Using Precompiled Headers
17732 @cindex precompiled headers
17733 @cindex speed of compilation
17735 Often large projects have many header files that are included in every
17736 source file. The time the compiler takes to process these header files
17737 over and over again can account for nearly all of the time required to
17738 build the project. To make builds faster, GCC allows users to
17739 `precompile' a header file; then, if builds can use the precompiled
17740 header file they will be much faster.
17742 To create a precompiled header file, simply compile it as you would any
17743 other file, if necessary using the @option{-x} option to make the driver
17744 treat it as a C or C++ header file. You will probably want to use a
17745 tool like @command{make} to keep the precompiled header up-to-date when
17746 the headers it contains change.
17748 A precompiled header file will be searched for when @code{#include} is
17749 seen in the compilation. As it searches for the included file
17750 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17751 compiler looks for a precompiled header in each directory just before it
17752 looks for the include file in that directory. The name searched for is
17753 the name specified in the @code{#include} with @samp{.gch} appended. If
17754 the precompiled header file can't be used, it is ignored.
17756 For instance, if you have @code{#include "all.h"}, and you have
17757 @file{all.h.gch} in the same directory as @file{all.h}, then the
17758 precompiled header file will be used if possible, and the original
17759 header will be used otherwise.
17761 Alternatively, you might decide to put the precompiled header file in a
17762 directory and use @option{-I} to ensure that directory is searched
17763 before (or instead of) the directory containing the original header.
17764 Then, if you want to check that the precompiled header file is always
17765 used, you can put a file of the same name as the original header in this
17766 directory containing an @code{#error} command.
17768 This also works with @option{-include}. So yet another way to use
17769 precompiled headers, good for projects not designed with precompiled
17770 header files in mind, is to simply take most of the header files used by
17771 a project, include them from another header file, precompile that header
17772 file, and @option{-include} the precompiled header. If the header files
17773 have guards against multiple inclusion, they will be skipped because
17774 they've already been included (in the precompiled header).
17776 If you need to precompile the same header file for different
17777 languages, targets, or compiler options, you can instead make a
17778 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17779 header in the directory, perhaps using @option{-o}. It doesn't matter
17780 what you call the files in the directory, every precompiled header in
17781 the directory will be considered. The first precompiled header
17782 encountered in the directory that is valid for this compilation will
17783 be used; they're searched in no particular order.
17785 There are many other possibilities, limited only by your imagination,
17786 good sense, and the constraints of your build system.
17788 A precompiled header file can be used only when these conditions apply:
17792 Only one precompiled header can be used in a particular compilation.
17795 A precompiled header can't be used once the first C token is seen. You
17796 can have preprocessor directives before a precompiled header; you can
17797 even include a precompiled header from inside another header, so long as
17798 there are no C tokens before the @code{#include}.
17801 The precompiled header file must be produced for the same language as
17802 the current compilation. You can't use a C precompiled header for a C++
17806 The precompiled header file must have been produced by the same compiler
17807 binary as the current compilation is using.
17810 Any macros defined before the precompiled header is included must
17811 either be defined in the same way as when the precompiled header was
17812 generated, or must not affect the precompiled header, which usually
17813 means that they don't appear in the precompiled header at all.
17815 The @option{-D} option is one way to define a macro before a
17816 precompiled header is included; using a @code{#define} can also do it.
17817 There are also some options that define macros implicitly, like
17818 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17821 @item If debugging information is output when using the precompiled
17822 header, using @option{-g} or similar, the same kind of debugging information
17823 must have been output when building the precompiled header. However,
17824 a precompiled header built using @option{-g} can be used in a compilation
17825 when no debugging information is being output.
17827 @item The same @option{-m} options must generally be used when building
17828 and using the precompiled header. @xref{Submodel Options},
17829 for any cases where this rule is relaxed.
17831 @item Each of the following options must be the same when building and using
17832 the precompiled header:
17834 @gccoptlist{-fexceptions}
17837 Some other command-line options starting with @option{-f},
17838 @option{-p}, or @option{-O} must be defined in the same way as when
17839 the precompiled header was generated. At present, it's not clear
17840 which options are safe to change and which are not; the safest choice
17841 is to use exactly the same options when generating and using the
17842 precompiled header. The following are known to be safe:
17844 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17845 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17846 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17851 For all of these except the last, the compiler will automatically
17852 ignore the precompiled header if the conditions aren't met. If you
17853 find an option combination that doesn't work and doesn't cause the
17854 precompiled header to be ignored, please consider filing a bug report,
17857 If you do use differing options when generating and using the
17858 precompiled header, the actual behavior will be a mixture of the
17859 behavior for the options. For instance, if you use @option{-g} to
17860 generate the precompiled header but not when using it, you may or may
17861 not get debugging information for routines in the precompiled header.