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, 2011
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, 2011
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.3 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
168 -fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
174 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
175 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
176 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
177 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
178 -fsigned-bitfields -fsigned-char @gol
179 -funsigned-bitfields -funsigned-char}
181 @item C++ Language Options
182 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
183 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
184 -fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-abi-version=@var{n} @gol
212 -fobjc-call-cxx-cdtors @gol
213 -fobjc-direct-dispatch @gol
214 -fobjc-exceptions @gol
217 -fobjc-std=objc1 @gol
218 -freplace-objc-classes @gol
221 -Wassign-intercept @gol
222 -Wno-protocol -Wselector @gol
223 -Wstrict-selector-match @gol
224 -Wundeclared-selector}
226 @item Language Independent Options
227 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
228 @gccoptlist{-fmessage-length=@var{n} @gol
229 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
230 -fno-diagnostics-show-option}
232 @item Warning Options
233 @xref{Warning Options,,Options to Request or Suppress Warnings}.
234 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
235 -pedantic-errors @gol
236 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
237 -Wno-attributes -Wno-builtin-macro-redefined @gol
238 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
239 -Wchar-subscripts -Wclobbered -Wcomment @gol
240 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol
241 -Wno-deprecated-declarations -Wdisabled-optimization @gol
242 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
243 -Wno-endif-labels -Werror -Werror=* @gol
244 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
245 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
246 -Wformat-security -Wformat-y2k @gol
247 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
248 -Wignored-qualifiers @gol
249 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
250 -Winit-self -Winline -Wmaybe-uninitialized @gol
251 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
252 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
253 -Wlogical-op -Wlong-long @gol
254 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
255 -Wmissing-format-attribute -Wmissing-include-dirs @gol
257 -Wno-multichar -Wnonnull -Wno-overflow @gol
258 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
259 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
260 -Wpointer-arith -Wno-pointer-to-int-cast @gol
261 -Wredundant-decls @gol
262 -Wreturn-type -Wsequence-point -Wshadow @gol
263 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
264 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
265 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
266 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
267 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
268 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
269 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
270 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
271 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
272 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
273 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
274 -Wvariadic-macros -Wvector-operation-performance -Wvla
275 -Wvolatile-register-var -Wwrite-strings}
277 @item C and Objective-C-only Warning Options
278 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
279 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
280 -Wold-style-declaration -Wold-style-definition @gol
281 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
282 -Wdeclaration-after-statement -Wpointer-sign}
284 @item Debugging Options
285 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
286 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
287 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
288 -fdisable-ipa-@var{pass_name} @gol
289 -fdisable-rtl-@var{pass_name} @gol
290 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
291 -fdisable-tree-@var{pass_name} @gol
292 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
293 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
294 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
295 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
296 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
298 -fdump-statistics @gol
300 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
304 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-nrv -fdump-tree-vect @gol
314 -fdump-tree-sink @gol
315 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
317 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
318 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
319 -ftree-vectorizer-verbose=@var{n} @gol
320 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
321 -fdump-final-insns=@var{file} @gol
322 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
323 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
324 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
325 -fenable-@var{kind}-@var{pass} @gol
326 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
327 -fdebug-types-section @gol
328 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
329 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
330 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
331 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
332 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
333 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
334 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
335 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
336 -gvms -gxcoff -gxcoff+ @gol
337 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
338 -fdebug-prefix-map=@var{old}=@var{new} @gol
339 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
340 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
341 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
342 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
343 -print-prog-name=@var{program} -print-search-dirs -Q @gol
344 -print-sysroot -print-sysroot-headers-suffix @gol
345 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
347 @item Optimization Options
348 @xref{Optimize Options,,Options that Control Optimization}.
349 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
350 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
351 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
352 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
353 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
354 -fcompare-elim -fcprop-registers -fcrossjumping @gol
355 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
356 -fcx-limited-range @gol
357 -fdata-sections -fdce -fdce -fdelayed-branch @gol
358 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
359 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
360 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
361 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
362 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
363 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
364 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
365 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
366 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
367 -fira-algorithm=@var{algorithm} @gol
368 -fira-region=@var{region} @gol
369 -fira-loop-pressure -fno-ira-share-save-slots @gol
370 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
371 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
372 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
373 -floop-parallelize-all -flto -flto-compression-level @gol
374 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
375 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
376 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
377 -fno-default-inline @gol
378 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
379 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
380 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
381 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
382 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
383 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
384 -fprefetch-loop-arrays @gol
385 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
386 -fprofile-generate=@var{path} @gol
387 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
388 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
389 -freorder-blocks-and-partition -freorder-functions @gol
390 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
391 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
392 -fsched-spec-load -fsched-spec-load-dangerous @gol
393 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
394 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
395 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
396 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
397 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
398 -fselective-scheduling -fselective-scheduling2 @gol
399 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
400 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
401 -fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol
402 -fstack-protector-all -fstrict-aliasing -fstrict-overflow @gol
403 -fthread-jumps -ftracer -ftree-bit-ccp @gol
404 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
405 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
406 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
407 -ftree-loop-if-convert-stores -ftree-loop-im @gol
408 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
409 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
410 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
411 -ftree-sink -ftree-sra -ftree-switch-conversion -ftree-tail-merge @gol
412 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
413 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
414 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
415 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
416 -fwhole-program -fwpa -fuse-linker-plugin @gol
417 --param @var{name}=@var{value}
418 -O -O0 -O1 -O2 -O3 -Os -Ofast}
420 @item Preprocessor Options
421 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
422 @gccoptlist{-A@var{question}=@var{answer} @gol
423 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
424 -C -dD -dI -dM -dN @gol
425 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
426 -idirafter @var{dir} @gol
427 -include @var{file} -imacros @var{file} @gol
428 -iprefix @var{file} -iwithprefix @var{dir} @gol
429 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
430 -imultilib @var{dir} -isysroot @var{dir} @gol
431 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
432 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
433 -remap -trigraphs -undef -U@var{macro} @gol
434 -Wp,@var{option} -Xpreprocessor @var{option}}
436 @item Assembler Option
437 @xref{Assembler Options,,Passing Options to the Assembler}.
438 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
441 @xref{Link Options,,Options for Linking}.
442 @gccoptlist{@var{object-file-name} -l@var{library} @gol
443 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
444 -s -static -static-libgcc -static-libstdc++ -shared @gol
445 -shared-libgcc -symbolic @gol
446 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
449 @item Directory Options
450 @xref{Directory Options,,Options for Directory Search}.
451 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
452 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
455 @item Machine Dependent Options
456 @xref{Submodel Options,,Hardware Models and Configurations}.
457 @c This list is ordered alphanumerically by subsection name.
458 @c Try and put the significant identifier (CPU or system) first,
459 @c so users have a clue at guessing where the ones they want will be.
462 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
463 -mabi=@var{name} @gol
464 -mapcs-stack-check -mno-apcs-stack-check @gol
465 -mapcs-float -mno-apcs-float @gol
466 -mapcs-reentrant -mno-apcs-reentrant @gol
467 -msched-prolog -mno-sched-prolog @gol
468 -mlittle-endian -mbig-endian -mwords-little-endian @gol
469 -mfloat-abi=@var{name} -mfpe @gol
470 -mfp16-format=@var{name}
471 -mthumb-interwork -mno-thumb-interwork @gol
472 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
473 -mstructure-size-boundary=@var{n} @gol
474 -mabort-on-noreturn @gol
475 -mlong-calls -mno-long-calls @gol
476 -msingle-pic-base -mno-single-pic-base @gol
477 -mpic-register=@var{reg} @gol
478 -mnop-fun-dllimport @gol
479 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
480 -mpoke-function-name @gol
482 -mtpcs-frame -mtpcs-leaf-frame @gol
483 -mcaller-super-interworking -mcallee-super-interworking @gol
484 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
485 -mword-relocations @gol
486 -mfix-cortex-m3-ldrd}
489 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
490 -mcall-prologues -mtiny-stack -mint8 -mstrict-X}
492 @emph{Blackfin Options}
493 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
494 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
495 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
496 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
497 -mno-id-shared-library -mshared-library-id=@var{n} @gol
498 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
499 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
500 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
504 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
505 -msim -msdata=@var{sdata-type}}
508 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
509 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
510 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
511 -mstack-align -mdata-align -mconst-align @gol
512 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
513 -melf -maout -melinux -mlinux -sim -sim2 @gol
514 -mmul-bug-workaround -mno-mul-bug-workaround}
516 @emph{Darwin Options}
517 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
518 -arch_only -bind_at_load -bundle -bundle_loader @gol
519 -client_name -compatibility_version -current_version @gol
521 -dependency-file -dylib_file -dylinker_install_name @gol
522 -dynamic -dynamiclib -exported_symbols_list @gol
523 -filelist -flat_namespace -force_cpusubtype_ALL @gol
524 -force_flat_namespace -headerpad_max_install_names @gol
526 -image_base -init -install_name -keep_private_externs @gol
527 -multi_module -multiply_defined -multiply_defined_unused @gol
528 -noall_load -no_dead_strip_inits_and_terms @gol
529 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
530 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
531 -private_bundle -read_only_relocs -sectalign @gol
532 -sectobjectsymbols -whyload -seg1addr @gol
533 -sectcreate -sectobjectsymbols -sectorder @gol
534 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
535 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
536 -segprot -segs_read_only_addr -segs_read_write_addr @gol
537 -single_module -static -sub_library -sub_umbrella @gol
538 -twolevel_namespace -umbrella -undefined @gol
539 -unexported_symbols_list -weak_reference_mismatches @gol
540 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
541 -mkernel -mone-byte-bool}
543 @emph{DEC Alpha Options}
544 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
545 -mieee -mieee-with-inexact -mieee-conformant @gol
546 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
547 -mtrap-precision=@var{mode} -mbuild-constants @gol
548 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
549 -mbwx -mmax -mfix -mcix @gol
550 -mfloat-vax -mfloat-ieee @gol
551 -mexplicit-relocs -msmall-data -mlarge-data @gol
552 -msmall-text -mlarge-text @gol
553 -mmemory-latency=@var{time}}
555 @emph{DEC Alpha/VMS Options}
556 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
559 @gccoptlist{-msmall-model -mno-lsim}
562 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
563 -mhard-float -msoft-float @gol
564 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
565 -mdouble -mno-double @gol
566 -mmedia -mno-media -mmuladd -mno-muladd @gol
567 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
568 -mlinked-fp -mlong-calls -malign-labels @gol
569 -mlibrary-pic -macc-4 -macc-8 @gol
570 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
571 -moptimize-membar -mno-optimize-membar @gol
572 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
573 -mvliw-branch -mno-vliw-branch @gol
574 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
575 -mno-nested-cond-exec -mtomcat-stats @gol
579 @emph{GNU/Linux Options}
580 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
581 -tno-android-cc -tno-android-ld}
583 @emph{H8/300 Options}
584 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
587 @gccoptlist{-march=@var{architecture-type} @gol
588 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
589 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
590 -mfixed-range=@var{register-range} @gol
591 -mjump-in-delay -mlinker-opt -mlong-calls @gol
592 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
593 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
594 -mno-jump-in-delay -mno-long-load-store @gol
595 -mno-portable-runtime -mno-soft-float @gol
596 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
597 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
598 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
599 -munix=@var{unix-std} -nolibdld -static -threads}
601 @emph{i386 and x86-64 Options}
602 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
603 -mfpmath=@var{unit} @gol
604 -masm=@var{dialect} -mno-fancy-math-387 @gol
605 -mno-fp-ret-in-387 -msoft-float @gol
606 -mno-wide-multiply -mrtd -malign-double @gol
607 -mpreferred-stack-boundary=@var{num} @gol
608 -mincoming-stack-boundary=@var{num} @gol
609 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
610 -mrecip -mrecip=@var{opt} @gol
612 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
613 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
614 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
615 -mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol
616 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
617 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
618 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
619 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
620 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
621 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
622 -mcmodel=@var{code-model} -mabi=@var{name} @gol
623 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
624 -msse2avx -mfentry -m8bit-idiv @gol
625 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
627 @emph{i386 and x86-64 Windows Options}
628 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
629 -mnop-fun-dllimport -mthread @gol
630 -municode -mwin32 -mwindows -fno-set-stack-executable}
633 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
634 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
635 -mconstant-gp -mauto-pic -mfused-madd @gol
636 -minline-float-divide-min-latency @gol
637 -minline-float-divide-max-throughput @gol
638 -mno-inline-float-divide @gol
639 -minline-int-divide-min-latency @gol
640 -minline-int-divide-max-throughput @gol
641 -mno-inline-int-divide @gol
642 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
643 -mno-inline-sqrt @gol
644 -mdwarf2-asm -mearly-stop-bits @gol
645 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
646 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
647 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
648 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
649 -msched-spec-ldc -msched-spec-control-ldc @gol
650 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
651 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
652 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
653 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
655 @emph{IA-64/VMS Options}
656 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
659 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
660 -msign-extend-enabled -muser-enabled}
662 @emph{M32R/D Options}
663 @gccoptlist{-m32r2 -m32rx -m32r @gol
665 -malign-loops -mno-align-loops @gol
666 -missue-rate=@var{number} @gol
667 -mbranch-cost=@var{number} @gol
668 -mmodel=@var{code-size-model-type} @gol
669 -msdata=@var{sdata-type} @gol
670 -mno-flush-func -mflush-func=@var{name} @gol
671 -mno-flush-trap -mflush-trap=@var{number} @gol
675 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
677 @emph{M680x0 Options}
678 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
679 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
680 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
681 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
682 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
683 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
684 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
685 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
689 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
690 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
691 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
692 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
693 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
696 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
697 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
698 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
699 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
702 @emph{MicroBlaze Options}
703 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
704 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
705 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
706 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
707 -mxl-mode-@var{app-model}}
710 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
711 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
712 -mips64 -mips64r2 @gol
713 -mips16 -mno-mips16 -mflip-mips16 @gol
714 -minterlink-mips16 -mno-interlink-mips16 @gol
715 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
716 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
717 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
718 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
719 -mfpu=@var{fpu-type} @gol
720 -msmartmips -mno-smartmips @gol
721 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
722 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
723 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
724 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
725 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
726 -membedded-data -mno-embedded-data @gol
727 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
728 -mcode-readable=@var{setting} @gol
729 -msplit-addresses -mno-split-addresses @gol
730 -mexplicit-relocs -mno-explicit-relocs @gol
731 -mcheck-zero-division -mno-check-zero-division @gol
732 -mdivide-traps -mdivide-breaks @gol
733 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
734 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
735 -mfix-24k -mno-fix-24k @gol
736 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
737 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
738 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
739 -mflush-func=@var{func} -mno-flush-func @gol
740 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
741 -mfp-exceptions -mno-fp-exceptions @gol
742 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
743 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
746 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
747 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
748 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
749 -mno-base-addresses -msingle-exit -mno-single-exit}
751 @emph{MN10300 Options}
752 @gccoptlist{-mmult-bug -mno-mult-bug @gol
753 -mno-am33 -mam33 -mam33-2 -mam34 @gol
754 -mtune=@var{cpu-type} @gol
755 -mreturn-pointer-on-d0 @gol
756 -mno-crt0 -mrelax -mliw -msetlb}
758 @emph{PDP-11 Options}
759 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
760 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
761 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
762 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
763 -mbranch-expensive -mbranch-cheap @gol
764 -munix-asm -mdec-asm}
766 @emph{picoChip Options}
767 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
768 -msymbol-as-address -mno-inefficient-warnings}
770 @emph{PowerPC Options}
771 See RS/6000 and PowerPC Options.
773 @emph{RS/6000 and PowerPC Options}
774 @gccoptlist{-mcpu=@var{cpu-type} @gol
775 -mtune=@var{cpu-type} @gol
776 -mcmodel=@var{code-model} @gol
777 -mpower -mno-power -mpower2 -mno-power2 @gol
778 -mpowerpc -mpowerpc64 -mno-powerpc @gol
779 -maltivec -mno-altivec @gol
780 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
781 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
782 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
783 -mfprnd -mno-fprnd @gol
784 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
785 -mnew-mnemonics -mold-mnemonics @gol
786 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
787 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
788 -malign-power -malign-natural @gol
789 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
790 -msingle-float -mdouble-float -msimple-fpu @gol
791 -mstring -mno-string -mupdate -mno-update @gol
792 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
793 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
794 -mstrict-align -mno-strict-align -mrelocatable @gol
795 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
796 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
797 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
798 -mprioritize-restricted-insns=@var{priority} @gol
799 -msched-costly-dep=@var{dependence_type} @gol
800 -minsert-sched-nops=@var{scheme} @gol
801 -mcall-sysv -mcall-netbsd @gol
802 -maix-struct-return -msvr4-struct-return @gol
803 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
804 -mblock-move-inline-limit=@var{num} @gol
805 -misel -mno-isel @gol
806 -misel=yes -misel=no @gol
808 -mspe=yes -mspe=no @gol
810 -mgen-cell-microcode -mwarn-cell-microcode @gol
811 -mvrsave -mno-vrsave @gol
812 -mmulhw -mno-mulhw @gol
813 -mdlmzb -mno-dlmzb @gol
814 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
815 -mprototype -mno-prototype @gol
816 -msim -mmvme -mads -myellowknife -memb -msdata @gol
817 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
818 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
819 -mno-recip-precision @gol
820 -mveclibabi=@var{type} -mfriz -mno-friz @gol
821 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
822 -msave-toc-indirect -mno-save-toc-indirect}
825 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
827 -mbig-endian-data -mlittle-endian-data @gol
830 -mas100-syntax -mno-as100-syntax@gol
832 -mmax-constant-size=@gol
835 -msave-acc-in-interrupts}
837 @emph{S/390 and zSeries Options}
838 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
839 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
840 -mlong-double-64 -mlong-double-128 @gol
841 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
842 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
843 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
844 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
845 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
848 @gccoptlist{-meb -mel @gol
852 -mscore5 -mscore5u -mscore7 -mscore7d}
855 @gccoptlist{-m1 -m2 -m2e @gol
856 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
858 -m4-nofpu -m4-single-only -m4-single -m4 @gol
859 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
860 -m5-64media -m5-64media-nofpu @gol
861 -m5-32media -m5-32media-nofpu @gol
862 -m5-compact -m5-compact-nofpu @gol
863 -mb -ml -mdalign -mrelax @gol
864 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
865 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
866 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
867 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
868 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
869 -maccumulate-outgoing-args -minvalid-symbols}
871 @emph{Solaris 2 Options}
872 @gccoptlist{-mimpure-text -mno-impure-text @gol
876 @gccoptlist{-mcpu=@var{cpu-type} @gol
877 -mtune=@var{cpu-type} @gol
878 -mcmodel=@var{code-model} @gol
879 -m32 -m64 -mapp-regs -mno-app-regs @gol
880 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
881 -mfpu -mno-fpu -mhard-float -msoft-float @gol
882 -mhard-quad-float -msoft-quad-float @gol
884 -mstack-bias -mno-stack-bias @gol
885 -munaligned-doubles -mno-unaligned-doubles @gol
886 -mv8plus -mno-v8plus -mvis -mno-vis @gol
887 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
888 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
892 @gccoptlist{-mwarn-reloc -merror-reloc @gol
893 -msafe-dma -munsafe-dma @gol
895 -msmall-mem -mlarge-mem -mstdmain @gol
896 -mfixed-range=@var{register-range} @gol
898 -maddress-space-conversion -mno-address-space-conversion @gol
899 -mcache-size=@var{cache-size} @gol
900 -matomic-updates -mno-atomic-updates}
902 @emph{System V Options}
903 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
906 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
907 -mprolog-function -mno-prolog-function -mspace @gol
908 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
909 -mapp-regs -mno-app-regs @gol
910 -mdisable-callt -mno-disable-callt @gol
913 -mv850e1 -mv850es @gol
918 @gccoptlist{-mg -mgnu -munix}
920 @emph{VxWorks Options}
921 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
922 -Xbind-lazy -Xbind-now}
924 @emph{x86-64 Options}
925 See i386 and x86-64 Options.
927 @emph{Xstormy16 Options}
930 @emph{Xtensa Options}
931 @gccoptlist{-mconst16 -mno-const16 @gol
932 -mfused-madd -mno-fused-madd @gol
934 -mserialize-volatile -mno-serialize-volatile @gol
935 -mtext-section-literals -mno-text-section-literals @gol
936 -mtarget-align -mno-target-align @gol
937 -mlongcalls -mno-longcalls}
939 @emph{zSeries Options}
940 See S/390 and zSeries Options.
942 @item Code Generation Options
943 @xref{Code Gen Options,,Options for Code Generation Conventions}.
944 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
945 -ffixed-@var{reg} -fexceptions @gol
946 -fnon-call-exceptions -funwind-tables @gol
947 -fasynchronous-unwind-tables @gol
948 -finhibit-size-directive -finstrument-functions @gol
949 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
950 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
951 -fno-common -fno-ident @gol
952 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
953 -fno-jump-tables @gol
954 -frecord-gcc-switches @gol
955 -freg-struct-return -fshort-enums @gol
956 -fshort-double -fshort-wchar @gol
957 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
958 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
959 -fno-stack-limit -fsplit-stack @gol
960 -fleading-underscore -ftls-model=@var{model} @gol
961 -ftrapv -fwrapv -fbounds-check @gol
962 -fvisibility -fstrict-volatile-bitfields}
966 * Overall Options:: Controlling the kind of output:
967 an executable, object files, assembler files,
968 or preprocessed source.
969 * C Dialect Options:: Controlling the variant of C language compiled.
970 * C++ Dialect Options:: Variations on C++.
971 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
973 * Language Independent Options:: Controlling how diagnostics should be
975 * Warning Options:: How picky should the compiler be?
976 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
977 * Optimize Options:: How much optimization?
978 * Preprocessor Options:: Controlling header files and macro definitions.
979 Also, getting dependency information for Make.
980 * Assembler Options:: Passing options to the assembler.
981 * Link Options:: Specifying libraries and so on.
982 * Directory Options:: Where to find header files and libraries.
983 Where to find the compiler executable files.
984 * Spec Files:: How to pass switches to sub-processes.
985 * Target Options:: Running a cross-compiler, or an old version of GCC.
988 @node Overall Options
989 @section Options Controlling the Kind of Output
991 Compilation can involve up to four stages: preprocessing, compilation
992 proper, assembly and linking, always in that order. GCC is capable of
993 preprocessing and compiling several files either into several
994 assembler input files, or into one assembler input file; then each
995 assembler input file produces an object file, and linking combines all
996 the object files (those newly compiled, and those specified as input)
997 into an executable file.
999 @cindex file name suffix
1000 For any given input file, the file name suffix determines what kind of
1001 compilation is done:
1005 C source code which must be preprocessed.
1008 C source code which should not be preprocessed.
1011 C++ source code which should not be preprocessed.
1014 Objective-C source code. Note that you must link with the @file{libobjc}
1015 library to make an Objective-C program work.
1018 Objective-C source code which should not be preprocessed.
1022 Objective-C++ source code. Note that you must link with the @file{libobjc}
1023 library to make an Objective-C++ program work. Note that @samp{.M} refers
1024 to a literal capital M@.
1026 @item @var{file}.mii
1027 Objective-C++ source code which should not be preprocessed.
1030 C, C++, Objective-C or Objective-C++ header file to be turned into a
1031 precompiled header (default), or C, C++ header file to be turned into an
1032 Ada spec (via the @option{-fdump-ada-spec} switch).
1035 @itemx @var{file}.cp
1036 @itemx @var{file}.cxx
1037 @itemx @var{file}.cpp
1038 @itemx @var{file}.CPP
1039 @itemx @var{file}.c++
1041 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1042 the last two letters must both be literally @samp{x}. Likewise,
1043 @samp{.C} refers to a literal capital C@.
1047 Objective-C++ source code which must be preprocessed.
1049 @item @var{file}.mii
1050 Objective-C++ source code which should not be preprocessed.
1054 @itemx @var{file}.hp
1055 @itemx @var{file}.hxx
1056 @itemx @var{file}.hpp
1057 @itemx @var{file}.HPP
1058 @itemx @var{file}.h++
1059 @itemx @var{file}.tcc
1060 C++ header file to be turned into a precompiled header or Ada spec.
1063 @itemx @var{file}.for
1064 @itemx @var{file}.ftn
1065 Fixed form Fortran source code which should not be preprocessed.
1068 @itemx @var{file}.FOR
1069 @itemx @var{file}.fpp
1070 @itemx @var{file}.FPP
1071 @itemx @var{file}.FTN
1072 Fixed form Fortran source code which must be preprocessed (with the traditional
1075 @item @var{file}.f90
1076 @itemx @var{file}.f95
1077 @itemx @var{file}.f03
1078 @itemx @var{file}.f08
1079 Free form Fortran source code which should not be preprocessed.
1081 @item @var{file}.F90
1082 @itemx @var{file}.F95
1083 @itemx @var{file}.F03
1084 @itemx @var{file}.F08
1085 Free form Fortran source code which must be preprocessed (with the
1086 traditional preprocessor).
1091 @c FIXME: Descriptions of Java file types.
1097 @item @var{file}.ads
1098 Ada source code file which contains a library unit declaration (a
1099 declaration of a package, subprogram, or generic, or a generic
1100 instantiation), or a library unit renaming declaration (a package,
1101 generic, or subprogram renaming declaration). Such files are also
1104 @item @var{file}.adb
1105 Ada source code file containing a library unit body (a subprogram or
1106 package body). Such files are also called @dfn{bodies}.
1108 @c GCC also knows about some suffixes for languages not yet included:
1119 @itemx @var{file}.sx
1120 Assembler code which must be preprocessed.
1123 An object file to be fed straight into linking.
1124 Any file name with no recognized suffix is treated this way.
1128 You can specify the input language explicitly with the @option{-x} option:
1131 @item -x @var{language}
1132 Specify explicitly the @var{language} for the following input files
1133 (rather than letting the compiler choose a default based on the file
1134 name suffix). This option applies to all following input files until
1135 the next @option{-x} option. Possible values for @var{language} are:
1137 c c-header cpp-output
1138 c++ c++-header c++-cpp-output
1139 objective-c objective-c-header objective-c-cpp-output
1140 objective-c++ objective-c++-header objective-c++-cpp-output
1141 assembler assembler-with-cpp
1143 f77 f77-cpp-input f95 f95-cpp-input
1149 Turn off any specification of a language, so that subsequent files are
1150 handled according to their file name suffixes (as they are if @option{-x}
1151 has not been used at all).
1153 @item -pass-exit-codes
1154 @opindex pass-exit-codes
1155 Normally the @command{gcc} program will exit with the code of 1 if any
1156 phase of the compiler returns a non-success return code. If you specify
1157 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1158 numerically highest error produced by any phase that returned an error
1159 indication. The C, C++, and Fortran frontends return 4, if an internal
1160 compiler error is encountered.
1163 If you only want some of the stages of compilation, you can use
1164 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1165 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1166 @command{gcc} is to stop. Note that some combinations (for example,
1167 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1172 Compile or assemble the source files, but do not link. The linking
1173 stage simply is not done. The ultimate output is in the form of an
1174 object file for each source file.
1176 By default, the object file name for a source file is made by replacing
1177 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1179 Unrecognized input files, not requiring compilation or assembly, are
1184 Stop after the stage of compilation proper; do not assemble. The output
1185 is in the form of an assembler code file for each non-assembler input
1188 By default, the assembler file name for a source file is made by
1189 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1191 Input files that don't require compilation are ignored.
1195 Stop after the preprocessing stage; do not run the compiler proper. The
1196 output is in the form of preprocessed source code, which is sent to the
1199 Input files which don't require preprocessing are ignored.
1201 @cindex output file option
1204 Place output in file @var{file}. This applies regardless to whatever
1205 sort of output is being produced, whether it be an executable file,
1206 an object file, an assembler file or preprocessed C code.
1208 If @option{-o} is not specified, the default is to put an executable
1209 file in @file{a.out}, the object file for
1210 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1211 assembler file in @file{@var{source}.s}, a precompiled header file in
1212 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1217 Print (on standard error output) the commands executed to run the stages
1218 of compilation. Also print the version number of the compiler driver
1219 program and of the preprocessor and the compiler proper.
1223 Like @option{-v} except the commands are not executed and arguments
1224 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1225 This is useful for shell scripts to capture the driver-generated command lines.
1229 Use pipes rather than temporary files for communication between the
1230 various stages of compilation. This fails to work on some systems where
1231 the assembler is unable to read from a pipe; but the GNU assembler has
1236 Print (on the standard output) a description of the command line options
1237 understood by @command{gcc}. If the @option{-v} option is also specified
1238 then @option{--help} will also be passed on to the various processes
1239 invoked by @command{gcc}, so that they can display the command line options
1240 they accept. If the @option{-Wextra} option has also been specified
1241 (prior to the @option{--help} option), then command line options which
1242 have no documentation associated with them will also be displayed.
1245 @opindex target-help
1246 Print (on the standard output) a description of target-specific command
1247 line options for each tool. For some targets extra target-specific
1248 information may also be printed.
1250 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1251 Print (on the standard output) a description of the command line
1252 options understood by the compiler that fit into all specified classes
1253 and qualifiers. These are the supported classes:
1256 @item @samp{optimizers}
1257 This will display all of the optimization options supported by the
1260 @item @samp{warnings}
1261 This will display all of the options controlling warning messages
1262 produced by the compiler.
1265 This will display target-specific options. Unlike the
1266 @option{--target-help} option however, target-specific options of the
1267 linker and assembler will not be displayed. This is because those
1268 tools do not currently support the extended @option{--help=} syntax.
1271 This will display the values recognized by the @option{--param}
1274 @item @var{language}
1275 This will display the options supported for @var{language}, where
1276 @var{language} is the name of one of the languages supported in this
1280 This will display the options that are common to all languages.
1283 These are the supported qualifiers:
1286 @item @samp{undocumented}
1287 Display only those options which are undocumented.
1290 Display options which take an argument that appears after an equal
1291 sign in the same continuous piece of text, such as:
1292 @samp{--help=target}.
1294 @item @samp{separate}
1295 Display options which take an argument that appears as a separate word
1296 following the original option, such as: @samp{-o output-file}.
1299 Thus for example to display all the undocumented target-specific
1300 switches supported by the compiler the following can be used:
1303 --help=target,undocumented
1306 The sense of a qualifier can be inverted by prefixing it with the
1307 @samp{^} character, so for example to display all binary warning
1308 options (i.e., ones that are either on or off and that do not take an
1309 argument), which have a description the following can be used:
1312 --help=warnings,^joined,^undocumented
1315 The argument to @option{--help=} should not consist solely of inverted
1318 Combining several classes is possible, although this usually
1319 restricts the output by so much that there is nothing to display. One
1320 case where it does work however is when one of the classes is
1321 @var{target}. So for example to display all the target-specific
1322 optimization options the following can be used:
1325 --help=target,optimizers
1328 The @option{--help=} option can be repeated on the command line. Each
1329 successive use will display its requested class of options, skipping
1330 those that have already been displayed.
1332 If the @option{-Q} option appears on the command line before the
1333 @option{--help=} option, then the descriptive text displayed by
1334 @option{--help=} is changed. Instead of describing the displayed
1335 options, an indication is given as to whether the option is enabled,
1336 disabled or set to a specific value (assuming that the compiler
1337 knows this at the point where the @option{--help=} option is used).
1339 Here is a truncated example from the ARM port of @command{gcc}:
1342 % gcc -Q -mabi=2 --help=target -c
1343 The following options are target specific:
1345 -mabort-on-noreturn [disabled]
1349 The output is sensitive to the effects of previous command line
1350 options, so for example it is possible to find out which optimizations
1351 are enabled at @option{-O2} by using:
1354 -Q -O2 --help=optimizers
1357 Alternatively you can discover which binary optimizations are enabled
1358 by @option{-O3} by using:
1361 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1362 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1363 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1366 @item -no-canonical-prefixes
1367 @opindex no-canonical-prefixes
1368 Do not expand any symbolic links, resolve references to @samp{/../}
1369 or @samp{/./}, or make the path absolute when generating a relative
1374 Display the version number and copyrights of the invoked GCC@.
1378 Invoke all subcommands under a wrapper program. The name of the
1379 wrapper program and its parameters are passed as a comma separated
1383 gcc -c t.c -wrapper gdb,--args
1386 This will invoke all subprograms of @command{gcc} under
1387 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1388 @samp{gdb --args cc1 @dots{}}.
1390 @item -fplugin=@var{name}.so
1391 Load the plugin code in file @var{name}.so, assumed to be a
1392 shared object to be dlopen'd by the compiler. The base name of
1393 the shared object file is used to identify the plugin for the
1394 purposes of argument parsing (See
1395 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1396 Each plugin should define the callback functions specified in the
1399 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1400 Define an argument called @var{key} with a value of @var{value}
1401 for the plugin called @var{name}.
1403 @item -fdump-ada-spec@r{[}-slim@r{]}
1404 For C and C++ source and include files, generate corresponding Ada
1405 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1406 GNAT User's Guide}, which provides detailed documentation on this feature.
1408 @item -fdump-go-spec=@var{file}
1409 For input files in any language, generate corresponding Go
1410 declarations in @var{file}. This generates Go @code{const},
1411 @code{type}, @code{var}, and @code{func} declarations which may be a
1412 useful way to start writing a Go interface to code written in some
1415 @include @value{srcdir}/../libiberty/at-file.texi
1419 @section Compiling C++ Programs
1421 @cindex suffixes for C++ source
1422 @cindex C++ source file suffixes
1423 C++ source files conventionally use one of the suffixes @samp{.C},
1424 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1425 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1426 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1427 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1428 files with these names and compiles them as C++ programs even if you
1429 call the compiler the same way as for compiling C programs (usually
1430 with the name @command{gcc}).
1434 However, the use of @command{gcc} does not add the C++ library.
1435 @command{g++} is a program that calls GCC and treats @samp{.c},
1436 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1437 files unless @option{-x} is used, and automatically specifies linking
1438 against the C++ library. This program is also useful when
1439 precompiling a C header file with a @samp{.h} extension for use in C++
1440 compilations. On many systems, @command{g++} is also installed with
1441 the name @command{c++}.
1443 @cindex invoking @command{g++}
1444 When you compile C++ programs, you may specify many of the same
1445 command-line options that you use for compiling programs in any
1446 language; or command-line options meaningful for C and related
1447 languages; or options that are meaningful only for C++ programs.
1448 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1449 explanations of options for languages related to C@.
1450 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1451 explanations of options that are meaningful only for C++ programs.
1453 @node C Dialect Options
1454 @section Options Controlling C Dialect
1455 @cindex dialect options
1456 @cindex language dialect options
1457 @cindex options, dialect
1459 The following options control the dialect of C (or languages derived
1460 from C, such as C++, Objective-C and Objective-C++) that the compiler
1464 @cindex ANSI support
1468 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1469 equivalent to @samp{-std=c++98}.
1471 This turns off certain features of GCC that are incompatible with ISO
1472 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1473 such as the @code{asm} and @code{typeof} keywords, and
1474 predefined macros such as @code{unix} and @code{vax} that identify the
1475 type of system you are using. It also enables the undesirable and
1476 rarely used ISO trigraph feature. For the C compiler,
1477 it disables recognition of C++ style @samp{//} comments as well as
1478 the @code{inline} keyword.
1480 The alternate keywords @code{__asm__}, @code{__extension__},
1481 @code{__inline__} and @code{__typeof__} continue to work despite
1482 @option{-ansi}. You would not want to use them in an ISO C program, of
1483 course, but it is useful to put them in header files that might be included
1484 in compilations done with @option{-ansi}. Alternate predefined macros
1485 such as @code{__unix__} and @code{__vax__} are also available, with or
1486 without @option{-ansi}.
1488 The @option{-ansi} option does not cause non-ISO programs to be
1489 rejected gratuitously. For that, @option{-pedantic} is required in
1490 addition to @option{-ansi}. @xref{Warning Options}.
1492 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1493 option is used. Some header files may notice this macro and refrain
1494 from declaring certain functions or defining certain macros that the
1495 ISO standard doesn't call for; this is to avoid interfering with any
1496 programs that might use these names for other things.
1498 Functions that would normally be built in but do not have semantics
1499 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1500 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1501 built-in functions provided by GCC}, for details of the functions
1506 Determine the language standard. @xref{Standards,,Language Standards
1507 Supported by GCC}, for details of these standard versions. This option
1508 is currently only supported when compiling C or C++.
1510 The compiler can accept several base standards, such as @samp{c90} or
1511 @samp{c++98}, and GNU dialects of those standards, such as
1512 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1513 compiler will accept all programs following that standard and those
1514 using GNU extensions that do not contradict it. For example,
1515 @samp{-std=c90} turns off certain features of GCC that are
1516 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1517 keywords, but not other GNU extensions that do not have a meaning in
1518 ISO C90, such as omitting the middle term of a @code{?:}
1519 expression. On the other hand, by specifying a GNU dialect of a
1520 standard, all features the compiler support are enabled, even when
1521 those features change the meaning of the base standard and some
1522 strict-conforming programs may be rejected. The particular standard
1523 is used by @option{-pedantic} to identify which features are GNU
1524 extensions given that version of the standard. For example
1525 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1526 comments, while @samp{-std=gnu99 -pedantic} would not.
1528 A value for this option must be provided; possible values are
1534 Support all ISO C90 programs (certain GNU extensions that conflict
1535 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1537 @item iso9899:199409
1538 ISO C90 as modified in amendment 1.
1544 ISO C99. Note that this standard is not yet fully supported; see
1545 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1546 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1549 ISO C1X, the draft of the next revision of the ISO C standard.
1550 Support is limited and experimental and features enabled by this
1551 option may be changed or removed if changed in or removed from the
1556 GNU dialect of ISO C90 (including some C99 features). This
1557 is the default for C code.
1561 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1562 this will become the default. The name @samp{gnu9x} is deprecated.
1565 GNU dialect of ISO C1X. Support is limited and experimental and
1566 features enabled by this option may be changed or removed if changed
1567 in or removed from the standard draft.
1570 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1574 GNU dialect of @option{-std=c++98}. This is the default for
1578 The working draft of the upcoming ISO C++0x standard. This option
1579 enables experimental features that are likely to be included in
1580 C++0x. The working draft is constantly changing, and any feature that is
1581 enabled by this flag may be removed from future versions of GCC if it is
1582 not part of the C++0x standard.
1585 GNU dialect of @option{-std=c++0x}. This option enables
1586 experimental features that may be removed in future versions of GCC.
1589 @item -fgnu89-inline
1590 @opindex fgnu89-inline
1591 The option @option{-fgnu89-inline} tells GCC to use the traditional
1592 GNU semantics for @code{inline} functions when in C99 mode.
1593 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1594 is accepted and ignored by GCC versions 4.1.3 up to but not including
1595 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1596 C99 mode. Using this option is roughly equivalent to adding the
1597 @code{gnu_inline} function attribute to all inline functions
1598 (@pxref{Function Attributes}).
1600 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1601 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1602 specifies the default behavior). This option was first supported in
1603 GCC 4.3. This option is not supported in @option{-std=c90} or
1604 @option{-std=gnu90} mode.
1606 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1607 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1608 in effect for @code{inline} functions. @xref{Common Predefined
1609 Macros,,,cpp,The C Preprocessor}.
1611 @item -aux-info @var{filename}
1613 Output to the given filename prototyped declarations for all functions
1614 declared and/or defined in a translation unit, including those in header
1615 files. This option is silently ignored in any language other than C@.
1617 Besides declarations, the file indicates, in comments, the origin of
1618 each declaration (source file and line), whether the declaration was
1619 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1620 @samp{O} for old, respectively, in the first character after the line
1621 number and the colon), and whether it came from a declaration or a
1622 definition (@samp{C} or @samp{F}, respectively, in the following
1623 character). In the case of function definitions, a K&R-style list of
1624 arguments followed by their declarations is also provided, inside
1625 comments, after the declaration.
1627 @item -fallow-parameterless-variadic-functions
1628 Accept variadic functions without named parameters.
1630 Although it is possible to define such a function, this is not very
1631 useful as it is not possible to read the arguments. This is only
1632 supported for C as this construct is allowed by C++.
1636 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1637 keyword, so that code can use these words as identifiers. You can use
1638 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1639 instead. @option{-ansi} implies @option{-fno-asm}.
1641 In C++, this switch only affects the @code{typeof} keyword, since
1642 @code{asm} and @code{inline} are standard keywords. You may want to
1643 use the @option{-fno-gnu-keywords} flag instead, which has the same
1644 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1645 switch only affects the @code{asm} and @code{typeof} keywords, since
1646 @code{inline} is a standard keyword in ISO C99.
1649 @itemx -fno-builtin-@var{function}
1650 @opindex fno-builtin
1651 @cindex built-in functions
1652 Don't recognize built-in functions that do not begin with
1653 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1654 functions provided by GCC}, for details of the functions affected,
1655 including those which are not built-in functions when @option{-ansi} or
1656 @option{-std} options for strict ISO C conformance are used because they
1657 do not have an ISO standard meaning.
1659 GCC normally generates special code to handle certain built-in functions
1660 more efficiently; for instance, calls to @code{alloca} may become single
1661 instructions that adjust the stack directly, and calls to @code{memcpy}
1662 may become inline copy loops. The resulting code is often both smaller
1663 and faster, but since the function calls no longer appear as such, you
1664 cannot set a breakpoint on those calls, nor can you change the behavior
1665 of the functions by linking with a different library. In addition,
1666 when a function is recognized as a built-in function, GCC may use
1667 information about that function to warn about problems with calls to
1668 that function, or to generate more efficient code, even if the
1669 resulting code still contains calls to that function. For example,
1670 warnings are given with @option{-Wformat} for bad calls to
1671 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1672 known not to modify global memory.
1674 With the @option{-fno-builtin-@var{function}} option
1675 only the built-in function @var{function} is
1676 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1677 function is named that is not built-in in this version of GCC, this
1678 option is ignored. There is no corresponding
1679 @option{-fbuiltin-@var{function}} option; if you wish to enable
1680 built-in functions selectively when using @option{-fno-builtin} or
1681 @option{-ffreestanding}, you may define macros such as:
1684 #define abs(n) __builtin_abs ((n))
1685 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1690 @cindex hosted environment
1692 Assert that compilation takes place in a hosted environment. This implies
1693 @option{-fbuiltin}. A hosted environment is one in which the
1694 entire standard library is available, and in which @code{main} has a return
1695 type of @code{int}. Examples are nearly everything except a kernel.
1696 This is equivalent to @option{-fno-freestanding}.
1698 @item -ffreestanding
1699 @opindex ffreestanding
1700 @cindex hosted environment
1702 Assert that compilation takes place in a freestanding environment. This
1703 implies @option{-fno-builtin}. A freestanding environment
1704 is one in which the standard library may not exist, and program startup may
1705 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1706 This is equivalent to @option{-fno-hosted}.
1708 @xref{Standards,,Language Standards Supported by GCC}, for details of
1709 freestanding and hosted environments.
1713 @cindex OpenMP parallel
1714 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1715 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1716 compiler generates parallel code according to the OpenMP Application
1717 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1718 implies @option{-pthread}, and thus is only supported on targets that
1719 have support for @option{-pthread}.
1721 @item -fms-extensions
1722 @opindex fms-extensions
1723 Accept some non-standard constructs used in Microsoft header files.
1725 In C++ code, this allows member names in structures to be similar
1726 to previous types declarations.
1735 Some cases of unnamed fields in structures and unions are only
1736 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1737 fields within structs/unions}, for details.
1739 @item -fplan9-extensions
1740 Accept some non-standard constructs used in Plan 9 code.
1742 This enables @option{-fms-extensions}, permits passing pointers to
1743 structures with anonymous fields to functions which expect pointers to
1744 elements of the type of the field, and permits referring to anonymous
1745 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1746 struct/union fields within structs/unions}, for details. This is only
1747 supported for C, not C++.
1751 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1752 options for strict ISO C conformance) implies @option{-trigraphs}.
1754 @item -no-integrated-cpp
1755 @opindex no-integrated-cpp
1756 Performs a compilation in two passes: preprocessing and compiling. This
1757 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1758 @option{-B} option. The user supplied compilation step can then add in
1759 an additional preprocessing step after normal preprocessing but before
1760 compiling. The default is to use the integrated cpp (internal cpp)
1762 The semantics of this option will change if "cc1", "cc1plus", and
1763 "cc1obj" are merged.
1765 @cindex traditional C language
1766 @cindex C language, traditional
1768 @itemx -traditional-cpp
1769 @opindex traditional-cpp
1770 @opindex traditional
1771 Formerly, these options caused GCC to attempt to emulate a pre-standard
1772 C compiler. They are now only supported with the @option{-E} switch.
1773 The preprocessor continues to support a pre-standard mode. See the GNU
1774 CPP manual for details.
1776 @item -fcond-mismatch
1777 @opindex fcond-mismatch
1778 Allow conditional expressions with mismatched types in the second and
1779 third arguments. The value of such an expression is void. This option
1780 is not supported for C++.
1782 @item -flax-vector-conversions
1783 @opindex flax-vector-conversions
1784 Allow implicit conversions between vectors with differing numbers of
1785 elements and/or incompatible element types. This option should not be
1788 @item -funsigned-char
1789 @opindex funsigned-char
1790 Let the type @code{char} be unsigned, like @code{unsigned char}.
1792 Each kind of machine has a default for what @code{char} should
1793 be. It is either like @code{unsigned char} by default or like
1794 @code{signed char} by default.
1796 Ideally, a portable program should always use @code{signed char} or
1797 @code{unsigned char} when it depends on the signedness of an object.
1798 But many programs have been written to use plain @code{char} and
1799 expect it to be signed, or expect it to be unsigned, depending on the
1800 machines they were written for. This option, and its inverse, let you
1801 make such a program work with the opposite default.
1803 The type @code{char} is always a distinct type from each of
1804 @code{signed char} or @code{unsigned char}, even though its behavior
1805 is always just like one of those two.
1808 @opindex fsigned-char
1809 Let the type @code{char} be signed, like @code{signed char}.
1811 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1812 the negative form of @option{-funsigned-char}. Likewise, the option
1813 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1815 @item -fsigned-bitfields
1816 @itemx -funsigned-bitfields
1817 @itemx -fno-signed-bitfields
1818 @itemx -fno-unsigned-bitfields
1819 @opindex fsigned-bitfields
1820 @opindex funsigned-bitfields
1821 @opindex fno-signed-bitfields
1822 @opindex fno-unsigned-bitfields
1823 These options control whether a bit-field is signed or unsigned, when the
1824 declaration does not use either @code{signed} or @code{unsigned}. By
1825 default, such a bit-field is signed, because this is consistent: the
1826 basic integer types such as @code{int} are signed types.
1829 @node C++ Dialect Options
1830 @section Options Controlling C++ Dialect
1832 @cindex compiler options, C++
1833 @cindex C++ options, command line
1834 @cindex options, C++
1835 This section describes the command-line options that are only meaningful
1836 for C++ programs; but you can also use most of the GNU compiler options
1837 regardless of what language your program is in. For example, you
1838 might compile a file @code{firstClass.C} like this:
1841 g++ -g -frepo -O -c firstClass.C
1845 In this example, only @option{-frepo} is an option meant
1846 only for C++ programs; you can use the other options with any
1847 language supported by GCC@.
1849 Here is a list of options that are @emph{only} for compiling C++ programs:
1853 @item -fabi-version=@var{n}
1854 @opindex fabi-version
1855 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1856 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1857 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1858 the version that conforms most closely to the C++ ABI specification.
1859 Therefore, the ABI obtained using version 0 will change as ABI bugs
1862 The default is version 2.
1864 Version 3 corrects an error in mangling a constant address as a
1867 Version 4 implements a standard mangling for vector types.
1869 Version 5 corrects the mangling of attribute const/volatile on
1870 function pointer types, decltype of a plain decl, and use of a
1871 function parameter in the declaration of another parameter.
1873 Version 6 corrects the promotion behavior of C++0x scoped enums.
1875 See also @option{-Wabi}.
1877 @item -fno-access-control
1878 @opindex fno-access-control
1879 Turn off all access checking. This switch is mainly useful for working
1880 around bugs in the access control code.
1884 Check that the pointer returned by @code{operator new} is non-null
1885 before attempting to modify the storage allocated. This check is
1886 normally unnecessary because the C++ standard specifies that
1887 @code{operator new} will only return @code{0} if it is declared
1888 @samp{throw()}, in which case the compiler will always check the
1889 return value even without this option. In all other cases, when
1890 @code{operator new} has a non-empty exception specification, memory
1891 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1892 @samp{new (nothrow)}.
1894 @item -fconserve-space
1895 @opindex fconserve-space
1896 Put uninitialized or runtime-initialized global variables into the
1897 common segment, as C does. This saves space in the executable at the
1898 cost of not diagnosing duplicate definitions. If you compile with this
1899 flag and your program mysteriously crashes after @code{main()} has
1900 completed, you may have an object that is being destroyed twice because
1901 two definitions were merged.
1903 This option is no longer useful on most targets, now that support has
1904 been added for putting variables into BSS without making them common.
1906 @item -fconstexpr-depth=@var{n}
1907 @opindex fconstexpr-depth
1908 Set the maximum nested evaluation depth for C++0x constexpr functions
1909 to @var{n}. A limit is needed to detect endless recursion during
1910 constant expression evaluation. The minimum specified by the standard
1913 @item -fno-deduce-init-list
1914 @opindex fno-deduce-init-list
1915 Disable deduction of a template type parameter as
1916 std::initializer_list from a brace-enclosed initializer list, i.e.
1919 template <class T> auto forward(T t) -> decltype (realfn (t))
1926 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1930 This option is present because this deduction is an extension to the
1931 current specification in the C++0x working draft, and there was
1932 some concern about potential overload resolution problems.
1934 @item -ffriend-injection
1935 @opindex ffriend-injection
1936 Inject friend functions into the enclosing namespace, so that they are
1937 visible outside the scope of the class in which they are declared.
1938 Friend functions were documented to work this way in the old Annotated
1939 C++ Reference Manual, and versions of G++ before 4.1 always worked
1940 that way. However, in ISO C++ a friend function which is not declared
1941 in an enclosing scope can only be found using argument dependent
1942 lookup. This option causes friends to be injected as they were in
1945 This option is for compatibility, and may be removed in a future
1948 @item -fno-elide-constructors
1949 @opindex fno-elide-constructors
1950 The C++ standard allows an implementation to omit creating a temporary
1951 which is only used to initialize another object of the same type.
1952 Specifying this option disables that optimization, and forces G++ to
1953 call the copy constructor in all cases.
1955 @item -fno-enforce-eh-specs
1956 @opindex fno-enforce-eh-specs
1957 Don't generate code to check for violation of exception specifications
1958 at runtime. This option violates the C++ standard, but may be useful
1959 for reducing code size in production builds, much like defining
1960 @samp{NDEBUG}. This does not give user code permission to throw
1961 exceptions in violation of the exception specifications; the compiler
1962 will still optimize based on the specifications, so throwing an
1963 unexpected exception will result in undefined behavior.
1966 @itemx -fno-for-scope
1968 @opindex fno-for-scope
1969 If @option{-ffor-scope} is specified, the scope of variables declared in
1970 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1971 as specified by the C++ standard.
1972 If @option{-fno-for-scope} is specified, the scope of variables declared in
1973 a @i{for-init-statement} extends to the end of the enclosing scope,
1974 as was the case in old versions of G++, and other (traditional)
1975 implementations of C++.
1977 The default if neither flag is given to follow the standard,
1978 but to allow and give a warning for old-style code that would
1979 otherwise be invalid, or have different behavior.
1981 @item -fno-gnu-keywords
1982 @opindex fno-gnu-keywords
1983 Do not recognize @code{typeof} as a keyword, so that code can use this
1984 word as an identifier. You can use the keyword @code{__typeof__} instead.
1985 @option{-ansi} implies @option{-fno-gnu-keywords}.
1987 @item -fno-implicit-templates
1988 @opindex fno-implicit-templates
1989 Never emit code for non-inline templates which are instantiated
1990 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1991 @xref{Template Instantiation}, for more information.
1993 @item -fno-implicit-inline-templates
1994 @opindex fno-implicit-inline-templates
1995 Don't emit code for implicit instantiations of inline templates, either.
1996 The default is to handle inlines differently so that compiles with and
1997 without optimization will need the same set of explicit instantiations.
1999 @item -fno-implement-inlines
2000 @opindex fno-implement-inlines
2001 To save space, do not emit out-of-line copies of inline functions
2002 controlled by @samp{#pragma implementation}. This will cause linker
2003 errors if these functions are not inlined everywhere they are called.
2005 @item -fms-extensions
2006 @opindex fms-extensions
2007 Disable pedantic warnings about constructs used in MFC, such as implicit
2008 int and getting a pointer to member function via non-standard syntax.
2010 @item -fno-nonansi-builtins
2011 @opindex fno-nonansi-builtins
2012 Disable built-in declarations of functions that are not mandated by
2013 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2014 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2017 @opindex fnothrow-opt
2018 Treat a @code{throw()} exception specification as though it were a
2019 @code{noexcept} specification to reduce or eliminate the text size
2020 overhead relative to a function with no exception specification. If
2021 the function has local variables of types with non-trivial
2022 destructors, the exception specification will actually make the
2023 function smaller because the EH cleanups for those variables can be
2024 optimized away. The semantic effect is that an exception thrown out of
2025 a function with such an exception specification will result in a call
2026 to @code{terminate} rather than @code{unexpected}.
2028 @item -fno-operator-names
2029 @opindex fno-operator-names
2030 Do not treat the operator name keywords @code{and}, @code{bitand},
2031 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2032 synonyms as keywords.
2034 @item -fno-optional-diags
2035 @opindex fno-optional-diags
2036 Disable diagnostics that the standard says a compiler does not need to
2037 issue. Currently, the only such diagnostic issued by G++ is the one for
2038 a name having multiple meanings within a class.
2041 @opindex fpermissive
2042 Downgrade some diagnostics about nonconformant code from errors to
2043 warnings. Thus, using @option{-fpermissive} will allow some
2044 nonconforming code to compile.
2046 @item -fno-pretty-templates
2047 @opindex fno-pretty-templates
2048 When an error message refers to a specialization of a function
2049 template, the compiler will normally print the signature of the
2050 template followed by the template arguments and any typedefs or
2051 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2052 rather than @code{void f(int)}) so that it's clear which template is
2053 involved. When an error message refers to a specialization of a class
2054 template, the compiler will omit any template arguments which match
2055 the default template arguments for that template. If either of these
2056 behaviors make it harder to understand the error message rather than
2057 easier, using @option{-fno-pretty-templates} will disable them.
2061 Enable automatic template instantiation at link time. This option also
2062 implies @option{-fno-implicit-templates}. @xref{Template
2063 Instantiation}, for more information.
2067 Disable generation of information about every class with virtual
2068 functions for use by the C++ runtime type identification features
2069 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2070 of the language, you can save some space by using this flag. Note that
2071 exception handling uses the same information, but it will generate it as
2072 needed. The @samp{dynamic_cast} operator can still be used for casts that
2073 do not require runtime type information, i.e.@: casts to @code{void *} or to
2074 unambiguous base classes.
2078 Emit statistics about front-end processing at the end of the compilation.
2079 This information is generally only useful to the G++ development team.
2081 @item -fstrict-enums
2082 @opindex fstrict-enums
2083 Allow the compiler to optimize using the assumption that a value of
2084 enumeration type can only be one of the values of the enumeration (as
2085 defined in the C++ standard; basically, a value which can be
2086 represented in the minimum number of bits needed to represent all the
2087 enumerators). This assumption may not be valid if the program uses a
2088 cast to convert an arbitrary integer value to the enumeration type.
2090 @item -ftemplate-depth=@var{n}
2091 @opindex ftemplate-depth
2092 Set the maximum instantiation depth for template classes to @var{n}.
2093 A limit on the template instantiation depth is needed to detect
2094 endless recursions during template class instantiation. ANSI/ISO C++
2095 conforming programs must not rely on a maximum depth greater than 17
2096 (changed to 1024 in C++0x). The default value is 900, as the compiler
2097 can run out of stack space before hitting 1024 in some situations.
2099 @item -fno-threadsafe-statics
2100 @opindex fno-threadsafe-statics
2101 Do not emit the extra code to use the routines specified in the C++
2102 ABI for thread-safe initialization of local statics. You can use this
2103 option to reduce code size slightly in code that doesn't need to be
2106 @item -fuse-cxa-atexit
2107 @opindex fuse-cxa-atexit
2108 Register destructors for objects with static storage duration with the
2109 @code{__cxa_atexit} function rather than the @code{atexit} function.
2110 This option is required for fully standards-compliant handling of static
2111 destructors, but will only work if your C library supports
2112 @code{__cxa_atexit}.
2114 @item -fno-use-cxa-get-exception-ptr
2115 @opindex fno-use-cxa-get-exception-ptr
2116 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2117 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2118 if the runtime routine is not available.
2120 @item -fvisibility-inlines-hidden
2121 @opindex fvisibility-inlines-hidden
2122 This switch declares that the user does not attempt to compare
2123 pointers to inline methods where the addresses of the two functions
2124 were taken in different shared objects.
2126 The effect of this is that GCC may, effectively, mark inline methods with
2127 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2128 appear in the export table of a DSO and do not require a PLT indirection
2129 when used within the DSO@. Enabling this option can have a dramatic effect
2130 on load and link times of a DSO as it massively reduces the size of the
2131 dynamic export table when the library makes heavy use of templates.
2133 The behavior of this switch is not quite the same as marking the
2134 methods as hidden directly, because it does not affect static variables
2135 local to the function or cause the compiler to deduce that
2136 the function is defined in only one shared object.
2138 You may mark a method as having a visibility explicitly to negate the
2139 effect of the switch for that method. For example, if you do want to
2140 compare pointers to a particular inline method, you might mark it as
2141 having default visibility. Marking the enclosing class with explicit
2142 visibility will have no effect.
2144 Explicitly instantiated inline methods are unaffected by this option
2145 as their linkage might otherwise cross a shared library boundary.
2146 @xref{Template Instantiation}.
2148 @item -fvisibility-ms-compat
2149 @opindex fvisibility-ms-compat
2150 This flag attempts to use visibility settings to make GCC's C++
2151 linkage model compatible with that of Microsoft Visual Studio.
2153 The flag makes these changes to GCC's linkage model:
2157 It sets the default visibility to @code{hidden}, like
2158 @option{-fvisibility=hidden}.
2161 Types, but not their members, are not hidden by default.
2164 The One Definition Rule is relaxed for types without explicit
2165 visibility specifications which are defined in more than one different
2166 shared object: those declarations are permitted if they would have
2167 been permitted when this option was not used.
2170 In new code it is better to use @option{-fvisibility=hidden} and
2171 export those classes which are intended to be externally visible.
2172 Unfortunately it is possible for code to rely, perhaps accidentally,
2173 on the Visual Studio behavior.
2175 Among the consequences of these changes are that static data members
2176 of the same type with the same name but defined in different shared
2177 objects will be different, so changing one will not change the other;
2178 and that pointers to function members defined in different shared
2179 objects may not compare equal. When this flag is given, it is a
2180 violation of the ODR to define types with the same name differently.
2184 Do not use weak symbol support, even if it is provided by the linker.
2185 By default, G++ will use weak symbols if they are available. This
2186 option exists only for testing, and should not be used by end-users;
2187 it will result in inferior code and has no benefits. This option may
2188 be removed in a future release of G++.
2192 Do not search for header files in the standard directories specific to
2193 C++, but do still search the other standard directories. (This option
2194 is used when building the C++ library.)
2197 In addition, these optimization, warning, and code generation options
2198 have meanings only for C++ programs:
2201 @item -fno-default-inline
2202 @opindex fno-default-inline
2203 Do not assume @samp{inline} for functions defined inside a class scope.
2204 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2205 functions will have linkage like inline functions; they just won't be
2208 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2211 Warn when G++ generates code that is probably not compatible with the
2212 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2213 all such cases, there are probably some cases that are not warned about,
2214 even though G++ is generating incompatible code. There may also be
2215 cases where warnings are emitted even though the code that is generated
2218 You should rewrite your code to avoid these warnings if you are
2219 concerned about the fact that code generated by G++ may not be binary
2220 compatible with code generated by other compilers.
2222 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2227 A template with a non-type template parameter of reference type is
2228 mangled incorrectly:
2231 template <int &> struct S @{@};
2235 This is fixed in @option{-fabi-version=3}.
2238 SIMD vector types declared using @code{__attribute ((vector_size))} are
2239 mangled in a non-standard way that does not allow for overloading of
2240 functions taking vectors of different sizes.
2242 The mangling is changed in @option{-fabi-version=4}.
2245 The known incompatibilities in @option{-fabi-version=1} include:
2250 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2251 pack data into the same byte as a base class. For example:
2254 struct A @{ virtual void f(); int f1 : 1; @};
2255 struct B : public A @{ int f2 : 1; @};
2259 In this case, G++ will place @code{B::f2} into the same byte
2260 as@code{A::f1}; other compilers will not. You can avoid this problem
2261 by explicitly padding @code{A} so that its size is a multiple of the
2262 byte size on your platform; that will cause G++ and other compilers to
2263 layout @code{B} identically.
2266 Incorrect handling of tail-padding for virtual bases. G++ does not use
2267 tail padding when laying out virtual bases. For example:
2270 struct A @{ virtual void f(); char c1; @};
2271 struct B @{ B(); char c2; @};
2272 struct C : public A, public virtual B @{@};
2276 In this case, G++ will not place @code{B} into the tail-padding for
2277 @code{A}; other compilers will. You can avoid this problem by
2278 explicitly padding @code{A} so that its size is a multiple of its
2279 alignment (ignoring virtual base classes); that will cause G++ and other
2280 compilers to layout @code{C} identically.
2283 Incorrect handling of bit-fields with declared widths greater than that
2284 of their underlying types, when the bit-fields appear in a union. For
2288 union U @{ int i : 4096; @};
2292 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2293 union too small by the number of bits in an @code{int}.
2296 Empty classes can be placed at incorrect offsets. For example:
2306 struct C : public B, public A @{@};
2310 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2311 it should be placed at offset zero. G++ mistakenly believes that the
2312 @code{A} data member of @code{B} is already at offset zero.
2315 Names of template functions whose types involve @code{typename} or
2316 template template parameters can be mangled incorrectly.
2319 template <typename Q>
2320 void f(typename Q::X) @{@}
2322 template <template <typename> class Q>
2323 void f(typename Q<int>::X) @{@}
2327 Instantiations of these templates may be mangled incorrectly.
2331 It also warns psABI related changes. The known psABI changes at this
2337 For SYSV/x86-64, when passing union with long double, it is changed to
2338 pass in memory as specified in psABI. For example:
2348 @code{union U} will always be passed in memory.
2352 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2353 @opindex Wctor-dtor-privacy
2354 @opindex Wno-ctor-dtor-privacy
2355 Warn when a class seems unusable because all the constructors or
2356 destructors in that class are private, and it has neither friends nor
2357 public static member functions.
2359 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2360 @opindex Wdelete-non-virtual-dtor
2361 @opindex Wno-delete-non-virtual-dtor
2362 Warn when @samp{delete} is used to destroy an instance of a class which
2363 has virtual functions and non-virtual destructor. It is unsafe to delete
2364 an instance of a derived class through a pointer to a base class if the
2365 base class does not have a virtual destructor. This warning is enabled
2368 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2370 @opindex Wno-narrowing
2371 With -std=c++0x, suppress the diagnostic required by the standard for
2372 narrowing conversions within @samp{@{ @}}, e.g.
2375 int i = @{ 2.2 @}; // error: narrowing from double to int
2378 This flag can be useful for compiling valid C++98 code in C++0x mode
2380 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2382 @opindex Wno-noexcept
2383 Warn when a noexcept-expression evaluates to false because of a call
2384 to a function that does not have a non-throwing exception
2385 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2386 the compiler to never throw an exception.
2388 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2389 @opindex Wnon-virtual-dtor
2390 @opindex Wno-non-virtual-dtor
2391 Warn when a class has virtual functions and accessible non-virtual
2392 destructor, in which case it would be possible but unsafe to delete
2393 an instance of a derived class through a pointer to the base class.
2394 This warning is also enabled if -Weffc++ is specified.
2396 @item -Wreorder @r{(C++ and Objective-C++ only)}
2398 @opindex Wno-reorder
2399 @cindex reordering, warning
2400 @cindex warning for reordering of member initializers
2401 Warn when the order of member initializers given in the code does not
2402 match the order in which they must be executed. For instance:
2408 A(): j (0), i (1) @{ @}
2412 The compiler will rearrange the member initializers for @samp{i}
2413 and @samp{j} to match the declaration order of the members, emitting
2414 a warning to that effect. This warning is enabled by @option{-Wall}.
2417 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2420 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2423 Warn about violations of the following style guidelines from Scott Meyers'
2424 @cite{Effective C++} book:
2428 Item 11: Define a copy constructor and an assignment operator for classes
2429 with dynamically allocated memory.
2432 Item 12: Prefer initialization to assignment in constructors.
2435 Item 14: Make destructors virtual in base classes.
2438 Item 15: Have @code{operator=} return a reference to @code{*this}.
2441 Item 23: Don't try to return a reference when you must return an object.
2445 Also warn about violations of the following style guidelines from
2446 Scott Meyers' @cite{More Effective C++} book:
2450 Item 6: Distinguish between prefix and postfix forms of increment and
2451 decrement operators.
2454 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2458 When selecting this option, be aware that the standard library
2459 headers do not obey all of these guidelines; use @samp{grep -v}
2460 to filter out those warnings.
2462 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2463 @opindex Wstrict-null-sentinel
2464 @opindex Wno-strict-null-sentinel
2465 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2466 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2467 to @code{__null}. Although it is a null pointer constant not a null pointer,
2468 it is guaranteed to be of the same size as a pointer. But this use is
2469 not portable across different compilers.
2471 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2472 @opindex Wno-non-template-friend
2473 @opindex Wnon-template-friend
2474 Disable warnings when non-templatized friend functions are declared
2475 within a template. Since the advent of explicit template specification
2476 support in G++, if the name of the friend is an unqualified-id (i.e.,
2477 @samp{friend foo(int)}), the C++ language specification demands that the
2478 friend declare or define an ordinary, nontemplate function. (Section
2479 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2480 could be interpreted as a particular specialization of a templatized
2481 function. Because this non-conforming behavior is no longer the default
2482 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2483 check existing code for potential trouble spots and is on by default.
2484 This new compiler behavior can be turned off with
2485 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2486 but disables the helpful warning.
2488 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2489 @opindex Wold-style-cast
2490 @opindex Wno-old-style-cast
2491 Warn if an old-style (C-style) cast to a non-void type is used within
2492 a C++ program. The new-style casts (@samp{dynamic_cast},
2493 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2494 less vulnerable to unintended effects and much easier to search for.
2496 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2497 @opindex Woverloaded-virtual
2498 @opindex Wno-overloaded-virtual
2499 @cindex overloaded virtual function, warning
2500 @cindex warning for overloaded virtual function
2501 Warn when a function declaration hides virtual functions from a
2502 base class. For example, in:
2509 struct B: public A @{
2514 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2522 will fail to compile.
2524 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2525 @opindex Wno-pmf-conversions
2526 @opindex Wpmf-conversions
2527 Disable the diagnostic for converting a bound pointer to member function
2530 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2531 @opindex Wsign-promo
2532 @opindex Wno-sign-promo
2533 Warn when overload resolution chooses a promotion from unsigned or
2534 enumerated type to a signed type, over a conversion to an unsigned type of
2535 the same size. Previous versions of G++ would try to preserve
2536 unsignedness, but the standard mandates the current behavior.
2541 A& operator = (int);
2551 In this example, G++ will synthesize a default @samp{A& operator =
2552 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2555 @node Objective-C and Objective-C++ Dialect Options
2556 @section Options Controlling Objective-C and Objective-C++ Dialects
2558 @cindex compiler options, Objective-C and Objective-C++
2559 @cindex Objective-C and Objective-C++ options, command line
2560 @cindex options, Objective-C and Objective-C++
2561 (NOTE: This manual does not describe the Objective-C and Objective-C++
2562 languages themselves. @xref{Standards,,Language Standards
2563 Supported by GCC}, for references.)
2565 This section describes the command-line options that are only meaningful
2566 for Objective-C and Objective-C++ programs, but you can also use most of
2567 the language-independent GNU compiler options.
2568 For example, you might compile a file @code{some_class.m} like this:
2571 gcc -g -fgnu-runtime -O -c some_class.m
2575 In this example, @option{-fgnu-runtime} is an option meant only for
2576 Objective-C and Objective-C++ programs; you can use the other options with
2577 any language supported by GCC@.
2579 Note that since Objective-C is an extension of the C language, Objective-C
2580 compilations may also use options specific to the C front-end (e.g.,
2581 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2582 C++-specific options (e.g., @option{-Wabi}).
2584 Here is a list of options that are @emph{only} for compiling Objective-C
2585 and Objective-C++ programs:
2588 @item -fconstant-string-class=@var{class-name}
2589 @opindex fconstant-string-class
2590 Use @var{class-name} as the name of the class to instantiate for each
2591 literal string specified with the syntax @code{@@"@dots{}"}. The default
2592 class name is @code{NXConstantString} if the GNU runtime is being used, and
2593 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2594 @option{-fconstant-cfstrings} option, if also present, will override the
2595 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2596 to be laid out as constant CoreFoundation strings.
2599 @opindex fgnu-runtime
2600 Generate object code compatible with the standard GNU Objective-C
2601 runtime. This is the default for most types of systems.
2603 @item -fnext-runtime
2604 @opindex fnext-runtime
2605 Generate output compatible with the NeXT runtime. This is the default
2606 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2607 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2610 @item -fno-nil-receivers
2611 @opindex fno-nil-receivers
2612 Assume that all Objective-C message dispatches (@code{[receiver
2613 message:arg]}) in this translation unit ensure that the receiver is
2614 not @code{nil}. This allows for more efficient entry points in the
2615 runtime to be used. This option is only available in conjunction with
2616 the NeXT runtime and ABI version 0 or 1.
2618 @item -fobjc-abi-version=@var{n}
2619 @opindex fobjc-abi-version
2620 Use version @var{n} of the Objective-C ABI for the selected runtime.
2621 This option is currently supported only for the NeXT runtime. In that
2622 case, Version 0 is the traditional (32-bit) ABI without support for
2623 properties and other Objective-C 2.0 additions. Version 1 is the
2624 traditional (32-bit) ABI with support for properties and other
2625 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2626 nothing is specified, the default is Version 0 on 32-bit target
2627 machines, and Version 2 on 64-bit target machines.
2629 @item -fobjc-call-cxx-cdtors
2630 @opindex fobjc-call-cxx-cdtors
2631 For each Objective-C class, check if any of its instance variables is a
2632 C++ object with a non-trivial default constructor. If so, synthesize a
2633 special @code{- (id) .cxx_construct} instance method that will run
2634 non-trivial default constructors on any such instance variables, in order,
2635 and then return @code{self}. Similarly, check if any instance variable
2636 is a C++ object with a non-trivial destructor, and if so, synthesize a
2637 special @code{- (void) .cxx_destruct} method that will run
2638 all such default destructors, in reverse order.
2640 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2641 methods thusly generated will only operate on instance variables
2642 declared in the current Objective-C class, and not those inherited
2643 from superclasses. It is the responsibility of the Objective-C
2644 runtime to invoke all such methods in an object's inheritance
2645 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2646 by the runtime immediately after a new object instance is allocated;
2647 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2648 before the runtime deallocates an object instance.
2650 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2651 support for invoking the @code{- (id) .cxx_construct} and
2652 @code{- (void) .cxx_destruct} methods.
2654 @item -fobjc-direct-dispatch
2655 @opindex fobjc-direct-dispatch
2656 Allow fast jumps to the message dispatcher. On Darwin this is
2657 accomplished via the comm page.
2659 @item -fobjc-exceptions
2660 @opindex fobjc-exceptions
2661 Enable syntactic support for structured exception handling in
2662 Objective-C, similar to what is offered by C++ and Java. This option
2663 is required to use the Objective-C keywords @code{@@try},
2664 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2665 @code{@@synchronized}. This option is available with both the GNU
2666 runtime and the NeXT runtime (but not available in conjunction with
2667 the NeXT runtime on Mac OS X 10.2 and earlier).
2671 Enable garbage collection (GC) in Objective-C and Objective-C++
2672 programs. This option is only available with the NeXT runtime; the
2673 GNU runtime has a different garbage collection implementation that
2674 does not require special compiler flags.
2676 @item -fobjc-nilcheck
2677 @opindex fobjc-nilcheck
2678 For the NeXT runtime with version 2 of the ABI, check for a nil
2679 receiver in method invocations before doing the actual method call.
2680 This is the default and can be disabled using
2681 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2682 checked for nil in this way no matter what this flag is set to.
2683 Currently this flag does nothing when the GNU runtime, or an older
2684 version of the NeXT runtime ABI, is used.
2686 @item -fobjc-std=objc1
2688 Conform to the language syntax of Objective-C 1.0, the language
2689 recognized by GCC 4.0. This only affects the Objective-C additions to
2690 the C/C++ language; it does not affect conformance to C/C++ standards,
2691 which is controlled by the separate C/C++ dialect option flags. When
2692 this option is used with the Objective-C or Objective-C++ compiler,
2693 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2694 This is useful if you need to make sure that your Objective-C code can
2695 be compiled with older versions of GCC.
2697 @item -freplace-objc-classes
2698 @opindex freplace-objc-classes
2699 Emit a special marker instructing @command{ld(1)} not to statically link in
2700 the resulting object file, and allow @command{dyld(1)} to load it in at
2701 run time instead. This is used in conjunction with the Fix-and-Continue
2702 debugging mode, where the object file in question may be recompiled and
2703 dynamically reloaded in the course of program execution, without the need
2704 to restart the program itself. Currently, Fix-and-Continue functionality
2705 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2710 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2711 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2712 compile time) with static class references that get initialized at load time,
2713 which improves run-time performance. Specifying the @option{-fzero-link} flag
2714 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2715 to be retained. This is useful in Zero-Link debugging mode, since it allows
2716 for individual class implementations to be modified during program execution.
2717 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2718 regardless of command line options.
2722 Dump interface declarations for all classes seen in the source file to a
2723 file named @file{@var{sourcename}.decl}.
2725 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2726 @opindex Wassign-intercept
2727 @opindex Wno-assign-intercept
2728 Warn whenever an Objective-C assignment is being intercepted by the
2731 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2732 @opindex Wno-protocol
2734 If a class is declared to implement a protocol, a warning is issued for
2735 every method in the protocol that is not implemented by the class. The
2736 default behavior is to issue a warning for every method not explicitly
2737 implemented in the class, even if a method implementation is inherited
2738 from the superclass. If you use the @option{-Wno-protocol} option, then
2739 methods inherited from the superclass are considered to be implemented,
2740 and no warning is issued for them.
2742 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2744 @opindex Wno-selector
2745 Warn if multiple methods of different types for the same selector are
2746 found during compilation. The check is performed on the list of methods
2747 in the final stage of compilation. Additionally, a check is performed
2748 for each selector appearing in a @code{@@selector(@dots{})}
2749 expression, and a corresponding method for that selector has been found
2750 during compilation. Because these checks scan the method table only at
2751 the end of compilation, these warnings are not produced if the final
2752 stage of compilation is not reached, for example because an error is
2753 found during compilation, or because the @option{-fsyntax-only} option is
2756 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2757 @opindex Wstrict-selector-match
2758 @opindex Wno-strict-selector-match
2759 Warn if multiple methods with differing argument and/or return types are
2760 found for a given selector when attempting to send a message using this
2761 selector to a receiver of type @code{id} or @code{Class}. When this flag
2762 is off (which is the default behavior), the compiler will omit such warnings
2763 if any differences found are confined to types which share the same size
2766 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2767 @opindex Wundeclared-selector
2768 @opindex Wno-undeclared-selector
2769 Warn if a @code{@@selector(@dots{})} expression referring to an
2770 undeclared selector is found. A selector is considered undeclared if no
2771 method with that name has been declared before the
2772 @code{@@selector(@dots{})} expression, either explicitly in an
2773 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2774 an @code{@@implementation} section. This option always performs its
2775 checks as soon as a @code{@@selector(@dots{})} expression is found,
2776 while @option{-Wselector} only performs its checks in the final stage of
2777 compilation. This also enforces the coding style convention
2778 that methods and selectors must be declared before being used.
2780 @item -print-objc-runtime-info
2781 @opindex print-objc-runtime-info
2782 Generate C header describing the largest structure that is passed by
2787 @node Language Independent Options
2788 @section Options to Control Diagnostic Messages Formatting
2789 @cindex options to control diagnostics formatting
2790 @cindex diagnostic messages
2791 @cindex message formatting
2793 Traditionally, diagnostic messages have been formatted irrespective of
2794 the output device's aspect (e.g.@: its width, @dots{}). The options described
2795 below can be used to control the diagnostic messages formatting
2796 algorithm, e.g.@: how many characters per line, how often source location
2797 information should be reported. Right now, only the C++ front end can
2798 honor these options. However it is expected, in the near future, that
2799 the remaining front ends would be able to digest them correctly.
2802 @item -fmessage-length=@var{n}
2803 @opindex fmessage-length
2804 Try to format error messages so that they fit on lines of about @var{n}
2805 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2806 the front ends supported by GCC@. If @var{n} is zero, then no
2807 line-wrapping will be done; each error message will appear on a single
2810 @opindex fdiagnostics-show-location
2811 @item -fdiagnostics-show-location=once
2812 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2813 reporter to emit @emph{once} source location information; that is, in
2814 case the message is too long to fit on a single physical line and has to
2815 be wrapped, the source location won't be emitted (as prefix) again,
2816 over and over, in subsequent continuation lines. This is the default
2819 @item -fdiagnostics-show-location=every-line
2820 Only meaningful in line-wrapping mode. Instructs the diagnostic
2821 messages reporter to emit the same source location information (as
2822 prefix) for physical lines that result from the process of breaking
2823 a message which is too long to fit on a single line.
2825 @item -fno-diagnostics-show-option
2826 @opindex fno-diagnostics-show-option
2827 @opindex fdiagnostics-show-option
2828 By default, each diagnostic emitted includes text which indicates the
2829 command line option that directly controls the diagnostic (if such an
2830 option is known to the diagnostic machinery). Specifying the
2831 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2833 @item -Wcoverage-mismatch
2834 @opindex Wcoverage-mismatch
2835 Warn if feedback profiles do not match when using the
2836 @option{-fprofile-use} option.
2837 If a source file was changed between @option{-fprofile-gen} and
2838 @option{-fprofile-use}, the files with the profile feedback can fail
2839 to match the source file and GCC can not use the profile feedback
2840 information. By default, this warning is enabled and is treated as an
2841 error. @option{-Wno-coverage-mismatch} can be used to disable the
2842 warning or @option{-Wno-error=coverage-mismatch} can be used to
2843 disable the error. Disable the error for this warning can result in
2844 poorly optimized code, so disabling the error is useful only in the
2845 case of very minor changes such as bug fixes to an existing code-base.
2846 Completely disabling the warning is not recommended.
2850 @node Warning Options
2851 @section Options to Request or Suppress Warnings
2852 @cindex options to control warnings
2853 @cindex warning messages
2854 @cindex messages, warning
2855 @cindex suppressing warnings
2857 Warnings are diagnostic messages that report constructions which
2858 are not inherently erroneous but which are risky or suggest there
2859 may have been an error.
2861 The following language-independent options do not enable specific
2862 warnings but control the kinds of diagnostics produced by GCC.
2865 @cindex syntax checking
2867 @opindex fsyntax-only
2868 Check the code for syntax errors, but don't do anything beyond that.
2870 @item -fmax-errors=@var{n}
2871 @opindex fmax-errors
2872 Limits the maximum number of error messages to @var{n}, at which point
2873 GCC bails out rather than attempting to continue processing the source
2874 code. If @var{n} is 0 (the default), there is no limit on the number
2875 of error messages produced. If @option{-Wfatal-errors} is also
2876 specified, then @option{-Wfatal-errors} takes precedence over this
2881 Inhibit all warning messages.
2886 Make all warnings into errors.
2891 Make the specified warning into an error. The specifier for a warning
2892 is appended, for example @option{-Werror=switch} turns the warnings
2893 controlled by @option{-Wswitch} into errors. This switch takes a
2894 negative form, to be used to negate @option{-Werror} for specific
2895 warnings, for example @option{-Wno-error=switch} makes
2896 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2899 The warning message for each controllable warning includes the
2900 option which controls the warning. That option can then be used with
2901 @option{-Werror=} and @option{-Wno-error=} as described above.
2902 (Printing of the option in the warning message can be disabled using the
2903 @option{-fno-diagnostics-show-option} flag.)
2905 Note that specifying @option{-Werror=}@var{foo} automatically implies
2906 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2909 @item -Wfatal-errors
2910 @opindex Wfatal-errors
2911 @opindex Wno-fatal-errors
2912 This option causes the compiler to abort compilation on the first error
2913 occurred rather than trying to keep going and printing further error
2918 You can request many specific warnings with options beginning
2919 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2920 implicit declarations. Each of these specific warning options also
2921 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2922 example, @option{-Wno-implicit}. This manual lists only one of the
2923 two forms, whichever is not the default. For further,
2924 language-specific options also refer to @ref{C++ Dialect Options} and
2925 @ref{Objective-C and Objective-C++ Dialect Options}.
2927 When an unrecognized warning option is requested (e.g.,
2928 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2929 that the option is not recognized. However, if the @option{-Wno-} form
2930 is used, the behavior is slightly different: No diagnostic will be
2931 produced for @option{-Wno-unknown-warning} unless other diagnostics
2932 are being produced. This allows the use of new @option{-Wno-} options
2933 with old compilers, but if something goes wrong, the compiler will
2934 warn that an unrecognized option was used.
2939 Issue all the warnings demanded by strict ISO C and ISO C++;
2940 reject all programs that use forbidden extensions, and some other
2941 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2942 version of the ISO C standard specified by any @option{-std} option used.
2944 Valid ISO C and ISO C++ programs should compile properly with or without
2945 this option (though a rare few will require @option{-ansi} or a
2946 @option{-std} option specifying the required version of ISO C)@. However,
2947 without this option, certain GNU extensions and traditional C and C++
2948 features are supported as well. With this option, they are rejected.
2950 @option{-pedantic} does not cause warning messages for use of the
2951 alternate keywords whose names begin and end with @samp{__}. Pedantic
2952 warnings are also disabled in the expression that follows
2953 @code{__extension__}. However, only system header files should use
2954 these escape routes; application programs should avoid them.
2955 @xref{Alternate Keywords}.
2957 Some users try to use @option{-pedantic} to check programs for strict ISO
2958 C conformance. They soon find that it does not do quite what they want:
2959 it finds some non-ISO practices, but not all---only those for which
2960 ISO C @emph{requires} a diagnostic, and some others for which
2961 diagnostics have been added.
2963 A feature to report any failure to conform to ISO C might be useful in
2964 some instances, but would require considerable additional work and would
2965 be quite different from @option{-pedantic}. We don't have plans to
2966 support such a feature in the near future.
2968 Where the standard specified with @option{-std} represents a GNU
2969 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2970 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2971 extended dialect is based. Warnings from @option{-pedantic} are given
2972 where they are required by the base standard. (It would not make sense
2973 for such warnings to be given only for features not in the specified GNU
2974 C dialect, since by definition the GNU dialects of C include all
2975 features the compiler supports with the given option, and there would be
2976 nothing to warn about.)
2978 @item -pedantic-errors
2979 @opindex pedantic-errors
2980 Like @option{-pedantic}, except that errors are produced rather than
2986 This enables all the warnings about constructions that some users
2987 consider questionable, and that are easy to avoid (or modify to
2988 prevent the warning), even in conjunction with macros. This also
2989 enables some language-specific warnings described in @ref{C++ Dialect
2990 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2992 @option{-Wall} turns on the following warning flags:
2994 @gccoptlist{-Waddress @gol
2995 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2997 -Wchar-subscripts @gol
2998 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2999 -Wimplicit-int @r{(C and Objective-C only)} @gol
3000 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3003 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3004 -Wmaybe-uninitialized @gol
3005 -Wmissing-braces @gol
3011 -Wsequence-point @gol
3012 -Wsign-compare @r{(only in C++)} @gol
3013 -Wstrict-aliasing @gol
3014 -Wstrict-overflow=1 @gol
3017 -Wuninitialized @gol
3018 -Wunknown-pragmas @gol
3019 -Wunused-function @gol
3022 -Wunused-variable @gol
3023 -Wvolatile-register-var @gol
3026 Note that some warning flags are not implied by @option{-Wall}. Some of
3027 them warn about constructions that users generally do not consider
3028 questionable, but which occasionally you might wish to check for;
3029 others warn about constructions that are necessary or hard to avoid in
3030 some cases, and there is no simple way to modify the code to suppress
3031 the warning. Some of them are enabled by @option{-Wextra} but many of
3032 them must be enabled individually.
3038 This enables some extra warning flags that are not enabled by
3039 @option{-Wall}. (This option used to be called @option{-W}. The older
3040 name is still supported, but the newer name is more descriptive.)
3042 @gccoptlist{-Wclobbered @gol
3044 -Wignored-qualifiers @gol
3045 -Wmissing-field-initializers @gol
3046 -Wmissing-parameter-type @r{(C only)} @gol
3047 -Wold-style-declaration @r{(C only)} @gol
3048 -Woverride-init @gol
3051 -Wuninitialized @gol
3052 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3053 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3056 The option @option{-Wextra} also prints warning messages for the
3062 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3063 @samp{>}, or @samp{>=}.
3066 (C++ only) An enumerator and a non-enumerator both appear in a
3067 conditional expression.
3070 (C++ only) Ambiguous virtual bases.
3073 (C++ only) Subscripting an array which has been declared @samp{register}.
3076 (C++ only) Taking the address of a variable which has been declared
3080 (C++ only) A base class is not initialized in a derived class' copy
3085 @item -Wchar-subscripts
3086 @opindex Wchar-subscripts
3087 @opindex Wno-char-subscripts
3088 Warn if an array subscript has type @code{char}. This is a common cause
3089 of error, as programmers often forget that this type is signed on some
3091 This warning is enabled by @option{-Wall}.
3095 @opindex Wno-comment
3096 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3097 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3098 This warning is enabled by @option{-Wall}.
3101 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3103 Suppress warning messages emitted by @code{#warning} directives.
3105 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3106 @opindex Wdouble-promotion
3107 @opindex Wno-double-promotion
3108 Give a warning when a value of type @code{float} is implicitly
3109 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3110 floating-point unit implement @code{float} in hardware, but emulate
3111 @code{double} in software. On such a machine, doing computations
3112 using @code{double} values is much more expensive because of the
3113 overhead required for software emulation.
3115 It is easy to accidentally do computations with @code{double} because
3116 floating-point literals are implicitly of type @code{double}. For
3120 float area(float radius)
3122 return 3.14159 * radius * radius;
3126 the compiler will perform the entire computation with @code{double}
3127 because the floating-point literal is a @code{double}.
3132 @opindex ffreestanding
3133 @opindex fno-builtin
3134 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3135 the arguments supplied have types appropriate to the format string
3136 specified, and that the conversions specified in the format string make
3137 sense. This includes standard functions, and others specified by format
3138 attributes (@pxref{Function Attributes}), in the @code{printf},
3139 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3140 not in the C standard) families (or other target-specific families).
3141 Which functions are checked without format attributes having been
3142 specified depends on the standard version selected, and such checks of
3143 functions without the attribute specified are disabled by
3144 @option{-ffreestanding} or @option{-fno-builtin}.
3146 The formats are checked against the format features supported by GNU
3147 libc version 2.2. These include all ISO C90 and C99 features, as well
3148 as features from the Single Unix Specification and some BSD and GNU
3149 extensions. Other library implementations may not support all these
3150 features; GCC does not support warning about features that go beyond a
3151 particular library's limitations. However, if @option{-pedantic} is used
3152 with @option{-Wformat}, warnings will be given about format features not
3153 in the selected standard version (but not for @code{strfmon} formats,
3154 since those are not in any version of the C standard). @xref{C Dialect
3155 Options,,Options Controlling C Dialect}.
3157 Since @option{-Wformat} also checks for null format arguments for
3158 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3160 @option{-Wformat} is included in @option{-Wall}. For more control over some
3161 aspects of format checking, the options @option{-Wformat-y2k},
3162 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3163 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3164 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3167 @opindex Wformat-y2k
3168 @opindex Wno-format-y2k
3169 If @option{-Wformat} is specified, also warn about @code{strftime}
3170 formats which may yield only a two-digit year.
3172 @item -Wno-format-contains-nul
3173 @opindex Wno-format-contains-nul
3174 @opindex Wformat-contains-nul
3175 If @option{-Wformat} is specified, do not warn about format strings that
3178 @item -Wno-format-extra-args
3179 @opindex Wno-format-extra-args
3180 @opindex Wformat-extra-args
3181 If @option{-Wformat} is specified, do not warn about excess arguments to a
3182 @code{printf} or @code{scanf} format function. The C standard specifies
3183 that such arguments are ignored.
3185 Where the unused arguments lie between used arguments that are
3186 specified with @samp{$} operand number specifications, normally
3187 warnings are still given, since the implementation could not know what
3188 type to pass to @code{va_arg} to skip the unused arguments. However,
3189 in the case of @code{scanf} formats, this option will suppress the
3190 warning if the unused arguments are all pointers, since the Single
3191 Unix Specification says that such unused arguments are allowed.
3193 @item -Wno-format-zero-length
3194 @opindex Wno-format-zero-length
3195 @opindex Wformat-zero-length
3196 If @option{-Wformat} is specified, do not warn about zero-length formats.
3197 The C standard specifies that zero-length formats are allowed.
3199 @item -Wformat-nonliteral
3200 @opindex Wformat-nonliteral
3201 @opindex Wno-format-nonliteral
3202 If @option{-Wformat} is specified, also warn if the format string is not a
3203 string literal and so cannot be checked, unless the format function
3204 takes its format arguments as a @code{va_list}.
3206 @item -Wformat-security
3207 @opindex Wformat-security
3208 @opindex Wno-format-security
3209 If @option{-Wformat} is specified, also warn about uses of format
3210 functions that represent possible security problems. At present, this
3211 warns about calls to @code{printf} and @code{scanf} functions where the
3212 format string is not a string literal and there are no format arguments,
3213 as in @code{printf (foo);}. This may be a security hole if the format
3214 string came from untrusted input and contains @samp{%n}. (This is
3215 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3216 in future warnings may be added to @option{-Wformat-security} that are not
3217 included in @option{-Wformat-nonliteral}.)
3221 @opindex Wno-format=2
3222 Enable @option{-Wformat} plus format checks not included in
3223 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3224 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3228 @opindex Wno-nonnull
3229 Warn about passing a null pointer for arguments marked as
3230 requiring a non-null value by the @code{nonnull} function attribute.
3232 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3233 can be disabled with the @option{-Wno-nonnull} option.
3235 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3237 @opindex Wno-init-self
3238 Warn about uninitialized variables which are initialized with themselves.
3239 Note this option can only be used with the @option{-Wuninitialized} option.
3241 For example, GCC will warn about @code{i} being uninitialized in the
3242 following snippet only when @option{-Winit-self} has been specified:
3253 @item -Wimplicit-int @r{(C and Objective-C only)}
3254 @opindex Wimplicit-int
3255 @opindex Wno-implicit-int
3256 Warn when a declaration does not specify a type.
3257 This warning is enabled by @option{-Wall}.
3259 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3260 @opindex Wimplicit-function-declaration
3261 @opindex Wno-implicit-function-declaration
3262 Give a warning whenever a function is used before being declared. In
3263 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3264 enabled by default and it is made into an error by
3265 @option{-pedantic-errors}. This warning is also enabled by
3268 @item -Wimplicit @r{(C and Objective-C only)}
3270 @opindex Wno-implicit
3271 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3272 This warning is enabled by @option{-Wall}.
3274 @item -Wignored-qualifiers @r{(C and C++ only)}
3275 @opindex Wignored-qualifiers
3276 @opindex Wno-ignored-qualifiers
3277 Warn if the return type of a function has a type qualifier
3278 such as @code{const}. For ISO C such a type qualifier has no effect,
3279 since the value returned by a function is not an lvalue.
3280 For C++, the warning is only emitted for scalar types or @code{void}.
3281 ISO C prohibits qualified @code{void} return types on function
3282 definitions, so such return types always receive a warning
3283 even without this option.
3285 This warning is also enabled by @option{-Wextra}.
3290 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3291 a function with external linkage, returning int, taking either zero
3292 arguments, two, or three arguments of appropriate types. This warning
3293 is enabled by default in C++ and is enabled by either @option{-Wall}
3294 or @option{-pedantic}.
3296 @item -Wmissing-braces
3297 @opindex Wmissing-braces
3298 @opindex Wno-missing-braces
3299 Warn if an aggregate or union initializer is not fully bracketed. In
3300 the following example, the initializer for @samp{a} is not fully
3301 bracketed, but that for @samp{b} is fully bracketed.
3304 int a[2][2] = @{ 0, 1, 2, 3 @};
3305 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3308 This warning is enabled by @option{-Wall}.
3310 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3311 @opindex Wmissing-include-dirs
3312 @opindex Wno-missing-include-dirs
3313 Warn if a user-supplied include directory does not exist.
3316 @opindex Wparentheses
3317 @opindex Wno-parentheses
3318 Warn if parentheses are omitted in certain contexts, such
3319 as when there is an assignment in a context where a truth value
3320 is expected, or when operators are nested whose precedence people
3321 often get confused about.
3323 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3324 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3325 interpretation from that of ordinary mathematical notation.
3327 Also warn about constructions where there may be confusion to which
3328 @code{if} statement an @code{else} branch belongs. Here is an example of
3343 In C/C++, every @code{else} branch belongs to the innermost possible
3344 @code{if} statement, which in this example is @code{if (b)}. This is
3345 often not what the programmer expected, as illustrated in the above
3346 example by indentation the programmer chose. When there is the
3347 potential for this confusion, GCC will issue a warning when this flag
3348 is specified. To eliminate the warning, add explicit braces around
3349 the innermost @code{if} statement so there is no way the @code{else}
3350 could belong to the enclosing @code{if}. The resulting code would
3367 Also warn for dangerous uses of the
3368 ?: with omitted middle operand GNU extension. When the condition
3369 in the ?: operator is a boolean expression the omitted value will
3370 be always 1. Often the user expects it to be a value computed
3371 inside the conditional expression instead.
3373 This warning is enabled by @option{-Wall}.
3375 @item -Wsequence-point
3376 @opindex Wsequence-point
3377 @opindex Wno-sequence-point
3378 Warn about code that may have undefined semantics because of violations
3379 of sequence point rules in the C and C++ standards.
3381 The C and C++ standards defines the order in which expressions in a C/C++
3382 program are evaluated in terms of @dfn{sequence points}, which represent
3383 a partial ordering between the execution of parts of the program: those
3384 executed before the sequence point, and those executed after it. These
3385 occur after the evaluation of a full expression (one which is not part
3386 of a larger expression), after the evaluation of the first operand of a
3387 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3388 function is called (but after the evaluation of its arguments and the
3389 expression denoting the called function), and in certain other places.
3390 Other than as expressed by the sequence point rules, the order of
3391 evaluation of subexpressions of an expression is not specified. All
3392 these rules describe only a partial order rather than a total order,
3393 since, for example, if two functions are called within one expression
3394 with no sequence point between them, the order in which the functions
3395 are called is not specified. However, the standards committee have
3396 ruled that function calls do not overlap.
3398 It is not specified when between sequence points modifications to the
3399 values of objects take effect. Programs whose behavior depends on this
3400 have undefined behavior; the C and C++ standards specify that ``Between
3401 the previous and next sequence point an object shall have its stored
3402 value modified at most once by the evaluation of an expression.
3403 Furthermore, the prior value shall be read only to determine the value
3404 to be stored.''. If a program breaks these rules, the results on any
3405 particular implementation are entirely unpredictable.
3407 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3408 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3409 diagnosed by this option, and it may give an occasional false positive
3410 result, but in general it has been found fairly effective at detecting
3411 this sort of problem in programs.
3413 The standard is worded confusingly, therefore there is some debate
3414 over the precise meaning of the sequence point rules in subtle cases.
3415 Links to discussions of the problem, including proposed formal
3416 definitions, may be found on the GCC readings page, at
3417 @uref{http://gcc.gnu.org/@/readings.html}.
3419 This warning is enabled by @option{-Wall} for C and C++.
3422 @opindex Wreturn-type
3423 @opindex Wno-return-type
3424 Warn whenever a function is defined with a return-type that defaults
3425 to @code{int}. Also warn about any @code{return} statement with no
3426 return-value in a function whose return-type is not @code{void}
3427 (falling off the end of the function body is considered returning
3428 without a value), and about a @code{return} statement with an
3429 expression in a function whose return-type is @code{void}.
3431 For C++, a function without return type always produces a diagnostic
3432 message, even when @option{-Wno-return-type} is specified. The only
3433 exceptions are @samp{main} and functions defined in system headers.
3435 This warning is enabled by @option{-Wall}.
3440 Warn whenever a @code{switch} statement has an index of enumerated type
3441 and lacks a @code{case} for one or more of the named codes of that
3442 enumeration. (The presence of a @code{default} label prevents this
3443 warning.) @code{case} labels outside the enumeration range also
3444 provoke warnings when this option is used (even if there is a
3445 @code{default} label).
3446 This warning is enabled by @option{-Wall}.
3448 @item -Wswitch-default
3449 @opindex Wswitch-default
3450 @opindex Wno-switch-default
3451 Warn whenever a @code{switch} statement does not have a @code{default}
3455 @opindex Wswitch-enum
3456 @opindex Wno-switch-enum
3457 Warn whenever a @code{switch} statement has an index of enumerated type
3458 and lacks a @code{case} for one or more of the named codes of that
3459 enumeration. @code{case} labels outside the enumeration range also
3460 provoke warnings when this option is used. The only difference
3461 between @option{-Wswitch} and this option is that this option gives a
3462 warning about an omitted enumeration code even if there is a
3463 @code{default} label.
3465 @item -Wsync-nand @r{(C and C++ only)}
3467 @opindex Wno-sync-nand
3468 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3469 built-in functions are used. These functions changed semantics in GCC 4.4.
3473 @opindex Wno-trigraphs
3474 Warn if any trigraphs are encountered that might change the meaning of
3475 the program (trigraphs within comments are not warned about).
3476 This warning is enabled by @option{-Wall}.
3478 @item -Wunused-but-set-parameter
3479 @opindex Wunused-but-set-parameter
3480 @opindex Wno-unused-but-set-parameter
3481 Warn whenever a function parameter is assigned to, but otherwise unused
3482 (aside from its declaration).
3484 To suppress this warning use the @samp{unused} attribute
3485 (@pxref{Variable Attributes}).
3487 This warning is also enabled by @option{-Wunused} together with
3490 @item -Wunused-but-set-variable
3491 @opindex Wunused-but-set-variable
3492 @opindex Wno-unused-but-set-variable
3493 Warn whenever a local variable is assigned to, but otherwise unused
3494 (aside from its declaration).
3495 This warning is enabled by @option{-Wall}.
3497 To suppress this warning use the @samp{unused} attribute
3498 (@pxref{Variable Attributes}).
3500 This warning is also enabled by @option{-Wunused}, which is enabled
3503 @item -Wunused-function
3504 @opindex Wunused-function
3505 @opindex Wno-unused-function
3506 Warn whenever a static function is declared but not defined or a
3507 non-inline static function is unused.
3508 This warning is enabled by @option{-Wall}.
3510 @item -Wunused-label
3511 @opindex Wunused-label
3512 @opindex Wno-unused-label
3513 Warn whenever a label is declared but not used.
3514 This warning is enabled by @option{-Wall}.
3516 To suppress this warning use the @samp{unused} attribute
3517 (@pxref{Variable Attributes}).
3519 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3520 @opindex Wunused-local-typedefs
3521 Warn when a typedef locally defined in a function is not used.
3523 @item -Wunused-parameter
3524 @opindex Wunused-parameter
3525 @opindex Wno-unused-parameter
3526 Warn whenever a function parameter is unused aside from its declaration.
3528 To suppress this warning use the @samp{unused} attribute
3529 (@pxref{Variable Attributes}).
3531 @item -Wno-unused-result
3532 @opindex Wunused-result
3533 @opindex Wno-unused-result
3534 Do not warn if a caller of a function marked with attribute
3535 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3536 its return value. The default is @option{-Wunused-result}.
3538 @item -Wunused-variable
3539 @opindex Wunused-variable
3540 @opindex Wno-unused-variable
3541 Warn whenever a local variable or non-constant static variable is unused
3542 aside from its declaration.
3543 This warning is enabled by @option{-Wall}.
3545 To suppress this warning use the @samp{unused} attribute
3546 (@pxref{Variable Attributes}).
3548 @item -Wunused-value
3549 @opindex Wunused-value
3550 @opindex Wno-unused-value
3551 Warn whenever a statement computes a result that is explicitly not
3552 used. To suppress this warning cast the unused expression to
3553 @samp{void}. This includes an expression-statement or the left-hand
3554 side of a comma expression that contains no side effects. For example,
3555 an expression such as @samp{x[i,j]} will cause a warning, while
3556 @samp{x[(void)i,j]} will not.
3558 This warning is enabled by @option{-Wall}.
3563 All the above @option{-Wunused} options combined.
3565 In order to get a warning about an unused function parameter, you must
3566 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3567 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3569 @item -Wuninitialized
3570 @opindex Wuninitialized
3571 @opindex Wno-uninitialized
3572 Warn if an automatic variable is used without first being initialized
3573 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3574 warn if a non-static reference or non-static @samp{const} member
3575 appears in a class without constructors.
3577 If you want to warn about code which uses the uninitialized value of the
3578 variable in its own initializer, use the @option{-Winit-self} option.
3580 These warnings occur for individual uninitialized or clobbered
3581 elements of structure, union or array variables as well as for
3582 variables which are uninitialized or clobbered as a whole. They do
3583 not occur for variables or elements declared @code{volatile}. Because
3584 these warnings depend on optimization, the exact variables or elements
3585 for which there are warnings will depend on the precise optimization
3586 options and version of GCC used.
3588 Note that there may be no warning about a variable that is used only
3589 to compute a value that itself is never used, because such
3590 computations may be deleted by data flow analysis before the warnings
3593 @item -Wmaybe-uninitialized
3594 @opindex Wmaybe-uninitialized
3595 @opindex Wno-maybe-uninitialized
3596 For an automatic variable, if there exists a path from the function
3597 entry to a use of the variable that is initialized, but there exist
3598 some other paths the variable is not initialized, the compiler will
3599 emit a warning if it can not prove the uninitialized paths do not
3600 happen at runtime. These warnings are made optional because GCC is
3601 not smart enough to see all the reasons why the code might be correct
3602 despite appearing to have an error. Here is one example of how
3623 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3624 always initialized, but GCC doesn't know this. To suppress the
3625 warning, the user needs to provide a default case with assert(0) or
3628 @cindex @code{longjmp} warnings
3629 This option also warns when a non-volatile automatic variable might be
3630 changed by a call to @code{longjmp}. These warnings as well are possible
3631 only in optimizing compilation.
3633 The compiler sees only the calls to @code{setjmp}. It cannot know
3634 where @code{longjmp} will be called; in fact, a signal handler could
3635 call it at any point in the code. As a result, you may get a warning
3636 even when there is in fact no problem because @code{longjmp} cannot
3637 in fact be called at the place which would cause a problem.
3639 Some spurious warnings can be avoided if you declare all the functions
3640 you use that never return as @code{noreturn}. @xref{Function
3643 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3645 @item -Wunknown-pragmas
3646 @opindex Wunknown-pragmas
3647 @opindex Wno-unknown-pragmas
3648 @cindex warning for unknown pragmas
3649 @cindex unknown pragmas, warning
3650 @cindex pragmas, warning of unknown
3651 Warn when a #pragma directive is encountered which is not understood by
3652 GCC@. If this command line option is used, warnings will even be issued
3653 for unknown pragmas in system header files. This is not the case if
3654 the warnings were only enabled by the @option{-Wall} command line option.
3657 @opindex Wno-pragmas
3659 Do not warn about misuses of pragmas, such as incorrect parameters,
3660 invalid syntax, or conflicts between pragmas. See also
3661 @samp{-Wunknown-pragmas}.
3663 @item -Wstrict-aliasing
3664 @opindex Wstrict-aliasing
3665 @opindex Wno-strict-aliasing
3666 This option is only active when @option{-fstrict-aliasing} is active.
3667 It warns about code which might break the strict aliasing rules that the
3668 compiler is using for optimization. The warning does not catch all
3669 cases, but does attempt to catch the more common pitfalls. It is
3670 included in @option{-Wall}.
3671 It is equivalent to @option{-Wstrict-aliasing=3}
3673 @item -Wstrict-aliasing=n
3674 @opindex Wstrict-aliasing=n
3675 @opindex Wno-strict-aliasing=n
3676 This option is only active when @option{-fstrict-aliasing} is active.
3677 It warns about code which might break the strict aliasing rules that the
3678 compiler is using for optimization.
3679 Higher levels correspond to higher accuracy (fewer false positives).
3680 Higher levels also correspond to more effort, similar to the way -O works.
3681 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3684 Level 1: Most aggressive, quick, least accurate.
3685 Possibly useful when higher levels
3686 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3687 false negatives. However, it has many false positives.
3688 Warns for all pointer conversions between possibly incompatible types,
3689 even if never dereferenced. Runs in the frontend only.
3691 Level 2: Aggressive, quick, not too precise.
3692 May still have many false positives (not as many as level 1 though),
3693 and few false negatives (but possibly more than level 1).
3694 Unlike level 1, it only warns when an address is taken. Warns about
3695 incomplete types. Runs in the frontend only.
3697 Level 3 (default for @option{-Wstrict-aliasing}):
3698 Should have very few false positives and few false
3699 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3700 Takes care of the common pun+dereference pattern in the frontend:
3701 @code{*(int*)&some_float}.
3702 If optimization is enabled, it also runs in the backend, where it deals
3703 with multiple statement cases using flow-sensitive points-to information.
3704 Only warns when the converted pointer is dereferenced.
3705 Does not warn about incomplete types.
3707 @item -Wstrict-overflow
3708 @itemx -Wstrict-overflow=@var{n}
3709 @opindex Wstrict-overflow
3710 @opindex Wno-strict-overflow
3711 This option is only active when @option{-fstrict-overflow} is active.
3712 It warns about cases where the compiler optimizes based on the
3713 assumption that signed overflow does not occur. Note that it does not
3714 warn about all cases where the code might overflow: it only warns
3715 about cases where the compiler implements some optimization. Thus
3716 this warning depends on the optimization level.
3718 An optimization which assumes that signed overflow does not occur is
3719 perfectly safe if the values of the variables involved are such that
3720 overflow never does, in fact, occur. Therefore this warning can
3721 easily give a false positive: a warning about code which is not
3722 actually a problem. To help focus on important issues, several
3723 warning levels are defined. No warnings are issued for the use of
3724 undefined signed overflow when estimating how many iterations a loop
3725 will require, in particular when determining whether a loop will be
3729 @item -Wstrict-overflow=1
3730 Warn about cases which are both questionable and easy to avoid. For
3731 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3732 compiler will simplify this to @code{1}. This level of
3733 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3734 are not, and must be explicitly requested.
3736 @item -Wstrict-overflow=2
3737 Also warn about other cases where a comparison is simplified to a
3738 constant. For example: @code{abs (x) >= 0}. This can only be
3739 simplified when @option{-fstrict-overflow} is in effect, because
3740 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3741 zero. @option{-Wstrict-overflow} (with no level) is the same as
3742 @option{-Wstrict-overflow=2}.
3744 @item -Wstrict-overflow=3
3745 Also warn about other cases where a comparison is simplified. For
3746 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3748 @item -Wstrict-overflow=4
3749 Also warn about other simplifications not covered by the above cases.
3750 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3752 @item -Wstrict-overflow=5
3753 Also warn about cases where the compiler reduces the magnitude of a
3754 constant involved in a comparison. For example: @code{x + 2 > y} will
3755 be simplified to @code{x + 1 >= y}. This is reported only at the
3756 highest warning level because this simplification applies to many
3757 comparisons, so this warning level will give a very large number of
3761 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3762 @opindex Wsuggest-attribute=
3763 @opindex Wno-suggest-attribute=
3764 Warn for cases where adding an attribute may be beneficial. The
3765 attributes currently supported are listed below.
3768 @item -Wsuggest-attribute=pure
3769 @itemx -Wsuggest-attribute=const
3770 @itemx -Wsuggest-attribute=noreturn
3771 @opindex Wsuggest-attribute=pure
3772 @opindex Wno-suggest-attribute=pure
3773 @opindex Wsuggest-attribute=const
3774 @opindex Wno-suggest-attribute=const
3775 @opindex Wsuggest-attribute=noreturn
3776 @opindex Wno-suggest-attribute=noreturn
3778 Warn about functions which might be candidates for attributes
3779 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3780 functions visible in other compilation units or (in the case of @code{pure} and
3781 @code{const}) if it cannot prove that the function returns normally. A function
3782 returns normally if it doesn't contain an infinite loop nor returns abnormally
3783 by throwing, calling @code{abort()} or trapping. This analysis requires option
3784 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3785 higher. Higher optimization levels improve the accuracy of the analysis.
3788 @item -Warray-bounds
3789 @opindex Wno-array-bounds
3790 @opindex Warray-bounds
3791 This option is only active when @option{-ftree-vrp} is active
3792 (default for @option{-O2} and above). It warns about subscripts to arrays
3793 that are always out of bounds. This warning is enabled by @option{-Wall}.
3795 @item -Wno-div-by-zero
3796 @opindex Wno-div-by-zero
3797 @opindex Wdiv-by-zero
3798 Do not warn about compile-time integer division by zero. Floating point
3799 division by zero is not warned about, as it can be a legitimate way of
3800 obtaining infinities and NaNs.
3802 @item -Wsystem-headers
3803 @opindex Wsystem-headers
3804 @opindex Wno-system-headers
3805 @cindex warnings from system headers
3806 @cindex system headers, warnings from
3807 Print warning messages for constructs found in system header files.
3808 Warnings from system headers are normally suppressed, on the assumption
3809 that they usually do not indicate real problems and would only make the
3810 compiler output harder to read. Using this command line option tells
3811 GCC to emit warnings from system headers as if they occurred in user
3812 code. However, note that using @option{-Wall} in conjunction with this
3813 option will @emph{not} warn about unknown pragmas in system
3814 headers---for that, @option{-Wunknown-pragmas} must also be used.
3817 @opindex Wtrampolines
3818 @opindex Wno-trampolines
3819 Warn about trampolines generated for pointers to nested functions.
3821 A trampoline is a small piece of data or code that is created at run
3822 time on the stack when the address of a nested function is taken, and
3823 is used to call the nested function indirectly. For some targets, it
3824 is made up of data only and thus requires no special treatment. But,
3825 for most targets, it is made up of code and thus requires the stack
3826 to be made executable in order for the program to work properly.
3829 @opindex Wfloat-equal
3830 @opindex Wno-float-equal
3831 Warn if floating point values are used in equality comparisons.
3833 The idea behind this is that sometimes it is convenient (for the
3834 programmer) to consider floating-point values as approximations to
3835 infinitely precise real numbers. If you are doing this, then you need
3836 to compute (by analyzing the code, or in some other way) the maximum or
3837 likely maximum error that the computation introduces, and allow for it
3838 when performing comparisons (and when producing output, but that's a
3839 different problem). In particular, instead of testing for equality, you
3840 would check to see whether the two values have ranges that overlap; and
3841 this is done with the relational operators, so equality comparisons are
3844 @item -Wtraditional @r{(C and Objective-C only)}
3845 @opindex Wtraditional
3846 @opindex Wno-traditional
3847 Warn about certain constructs that behave differently in traditional and
3848 ISO C@. Also warn about ISO C constructs that have no traditional C
3849 equivalent, and/or problematic constructs which should be avoided.
3853 Macro parameters that appear within string literals in the macro body.
3854 In traditional C macro replacement takes place within string literals,
3855 but does not in ISO C@.
3858 In traditional C, some preprocessor directives did not exist.
3859 Traditional preprocessors would only consider a line to be a directive
3860 if the @samp{#} appeared in column 1 on the line. Therefore
3861 @option{-Wtraditional} warns about directives that traditional C
3862 understands but would ignore because the @samp{#} does not appear as the
3863 first character on the line. It also suggests you hide directives like
3864 @samp{#pragma} not understood by traditional C by indenting them. Some
3865 traditional implementations would not recognize @samp{#elif}, so it
3866 suggests avoiding it altogether.
3869 A function-like macro that appears without arguments.
3872 The unary plus operator.
3875 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3876 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3877 constants.) Note, these suffixes appear in macros defined in the system
3878 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3879 Use of these macros in user code might normally lead to spurious
3880 warnings, however GCC's integrated preprocessor has enough context to
3881 avoid warning in these cases.
3884 A function declared external in one block and then used after the end of
3888 A @code{switch} statement has an operand of type @code{long}.
3891 A non-@code{static} function declaration follows a @code{static} one.
3892 This construct is not accepted by some traditional C compilers.
3895 The ISO type of an integer constant has a different width or
3896 signedness from its traditional type. This warning is only issued if
3897 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3898 typically represent bit patterns, are not warned about.
3901 Usage of ISO string concatenation is detected.
3904 Initialization of automatic aggregates.
3907 Identifier conflicts with labels. Traditional C lacks a separate
3908 namespace for labels.
3911 Initialization of unions. If the initializer is zero, the warning is
3912 omitted. This is done under the assumption that the zero initializer in
3913 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3914 initializer warnings and relies on default initialization to zero in the
3918 Conversions by prototypes between fixed/floating point values and vice
3919 versa. The absence of these prototypes when compiling with traditional
3920 C would cause serious problems. This is a subset of the possible
3921 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3924 Use of ISO C style function definitions. This warning intentionally is
3925 @emph{not} issued for prototype declarations or variadic functions
3926 because these ISO C features will appear in your code when using
3927 libiberty's traditional C compatibility macros, @code{PARAMS} and
3928 @code{VPARAMS}. This warning is also bypassed for nested functions
3929 because that feature is already a GCC extension and thus not relevant to
3930 traditional C compatibility.
3933 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3934 @opindex Wtraditional-conversion
3935 @opindex Wno-traditional-conversion
3936 Warn if a prototype causes a type conversion that is different from what
3937 would happen to the same argument in the absence of a prototype. This
3938 includes conversions of fixed point to floating and vice versa, and
3939 conversions changing the width or signedness of a fixed point argument
3940 except when the same as the default promotion.
3942 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3943 @opindex Wdeclaration-after-statement
3944 @opindex Wno-declaration-after-statement
3945 Warn when a declaration is found after a statement in a block. This
3946 construct, known from C++, was introduced with ISO C99 and is by default
3947 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3948 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3953 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3955 @item -Wno-endif-labels
3956 @opindex Wno-endif-labels
3957 @opindex Wendif-labels
3958 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3963 Warn whenever a local variable or type declaration shadows another variable,
3964 parameter, type, or class member (in C++), or whenever a built-in function
3965 is shadowed. Note that in C++, the compiler will not warn if a local variable
3966 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3968 @item -Wlarger-than=@var{len}
3969 @opindex Wlarger-than=@var{len}
3970 @opindex Wlarger-than-@var{len}
3971 Warn whenever an object of larger than @var{len} bytes is defined.
3973 @item -Wframe-larger-than=@var{len}
3974 @opindex Wframe-larger-than
3975 Warn if the size of a function frame is larger than @var{len} bytes.
3976 The computation done to determine the stack frame size is approximate
3977 and not conservative.
3978 The actual requirements may be somewhat greater than @var{len}
3979 even if you do not get a warning. In addition, any space allocated
3980 via @code{alloca}, variable-length arrays, or related constructs
3981 is not included by the compiler when determining
3982 whether or not to issue a warning.
3984 @item -Wno-free-nonheap-object
3985 @opindex Wno-free-nonheap-object
3986 @opindex Wfree-nonheap-object
3987 Do not warn when attempting to free an object which was not allocated
3990 @item -Wstack-usage=@var{len}
3991 @opindex Wstack-usage
3992 Warn if the stack usage of a function might be larger than @var{len} bytes.
3993 The computation done to determine the stack usage is conservative.
3994 Any space allocated via @code{alloca}, variable-length arrays, or related
3995 constructs is included by the compiler when determining whether or not to
3998 The message is in keeping with the output of @option{-fstack-usage}.
4002 If the stack usage is fully static but exceeds the specified amount, it's:
4005 warning: stack usage is 1120 bytes
4008 If the stack usage is (partly) dynamic but bounded, it's:
4011 warning: stack usage might be 1648 bytes
4014 If the stack usage is (partly) dynamic and not bounded, it's:
4017 warning: stack usage might be unbounded
4021 @item -Wunsafe-loop-optimizations
4022 @opindex Wunsafe-loop-optimizations
4023 @opindex Wno-unsafe-loop-optimizations
4024 Warn if the loop cannot be optimized because the compiler could not
4025 assume anything on the bounds of the loop indices. With
4026 @option{-funsafe-loop-optimizations} warn if the compiler made
4029 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4030 @opindex Wno-pedantic-ms-format
4031 @opindex Wpedantic-ms-format
4032 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4033 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4034 depending on the MS runtime, when you are using the options @option{-Wformat}
4035 and @option{-pedantic} without gnu-extensions.
4037 @item -Wpointer-arith
4038 @opindex Wpointer-arith
4039 @opindex Wno-pointer-arith
4040 Warn about anything that depends on the ``size of'' a function type or
4041 of @code{void}. GNU C assigns these types a size of 1, for
4042 convenience in calculations with @code{void *} pointers and pointers
4043 to functions. In C++, warn also when an arithmetic operation involves
4044 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4047 @opindex Wtype-limits
4048 @opindex Wno-type-limits
4049 Warn if a comparison is always true or always false due to the limited
4050 range of the data type, but do not warn for constant expressions. For
4051 example, warn if an unsigned variable is compared against zero with
4052 @samp{<} or @samp{>=}. This warning is also enabled by
4055 @item -Wbad-function-cast @r{(C and Objective-C only)}
4056 @opindex Wbad-function-cast
4057 @opindex Wno-bad-function-cast
4058 Warn whenever a function call is cast to a non-matching type.
4059 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4061 @item -Wc++-compat @r{(C and Objective-C only)}
4062 Warn about ISO C constructs that are outside of the common subset of
4063 ISO C and ISO C++, e.g.@: request for implicit conversion from
4064 @code{void *} to a pointer to non-@code{void} type.
4066 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
4067 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
4068 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
4069 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
4073 @opindex Wno-cast-qual
4074 Warn whenever a pointer is cast so as to remove a type qualifier from
4075 the target type. For example, warn if a @code{const char *} is cast
4076 to an ordinary @code{char *}.
4078 Also warn when making a cast which introduces a type qualifier in an
4079 unsafe way. For example, casting @code{char **} to @code{const char **}
4080 is unsafe, as in this example:
4083 /* p is char ** value. */
4084 const char **q = (const char **) p;
4085 /* Assignment of readonly string to const char * is OK. */
4087 /* Now char** pointer points to read-only memory. */
4092 @opindex Wcast-align
4093 @opindex Wno-cast-align
4094 Warn whenever a pointer is cast such that the required alignment of the
4095 target is increased. For example, warn if a @code{char *} is cast to
4096 an @code{int *} on machines where integers can only be accessed at
4097 two- or four-byte boundaries.
4099 @item -Wwrite-strings
4100 @opindex Wwrite-strings
4101 @opindex Wno-write-strings
4102 When compiling C, give string constants the type @code{const
4103 char[@var{length}]} so that copying the address of one into a
4104 non-@code{const} @code{char *} pointer will get a warning. These
4105 warnings will help you find at compile time code that can try to write
4106 into a string constant, but only if you have been very careful about
4107 using @code{const} in declarations and prototypes. Otherwise, it will
4108 just be a nuisance. This is why we did not make @option{-Wall} request
4111 When compiling C++, warn about the deprecated conversion from string
4112 literals to @code{char *}. This warning is enabled by default for C++
4117 @opindex Wno-clobbered
4118 Warn for variables that might be changed by @samp{longjmp} or
4119 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4122 @opindex Wconversion
4123 @opindex Wno-conversion
4124 Warn for implicit conversions that may alter a value. This includes
4125 conversions between real and integer, like @code{abs (x)} when
4126 @code{x} is @code{double}; conversions between signed and unsigned,
4127 like @code{unsigned ui = -1}; and conversions to smaller types, like
4128 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4129 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4130 changed by the conversion like in @code{abs (2.0)}. Warnings about
4131 conversions between signed and unsigned integers can be disabled by
4132 using @option{-Wno-sign-conversion}.
4134 For C++, also warn for confusing overload resolution for user-defined
4135 conversions; and conversions that will never use a type conversion
4136 operator: conversions to @code{void}, the same type, a base class or a
4137 reference to them. Warnings about conversions between signed and
4138 unsigned integers are disabled by default in C++ unless
4139 @option{-Wsign-conversion} is explicitly enabled.
4141 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4142 @opindex Wconversion-null
4143 @opindex Wno-conversion-null
4144 Do not warn for conversions between @code{NULL} and non-pointer
4145 types. @option{-Wconversion-null} is enabled by default.
4148 @opindex Wempty-body
4149 @opindex Wno-empty-body
4150 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4151 while} statement. This warning is also enabled by @option{-Wextra}.
4153 @item -Wenum-compare
4154 @opindex Wenum-compare
4155 @opindex Wno-enum-compare
4156 Warn about a comparison between values of different enum types. In C++
4157 this warning is enabled by default. In C this warning is enabled by
4160 @item -Wjump-misses-init @r{(C, Objective-C only)}
4161 @opindex Wjump-misses-init
4162 @opindex Wno-jump-misses-init
4163 Warn if a @code{goto} statement or a @code{switch} statement jumps
4164 forward across the initialization of a variable, or jumps backward to a
4165 label after the variable has been initialized. This only warns about
4166 variables which are initialized when they are declared. This warning is
4167 only supported for C and Objective C; in C++ this sort of branch is an
4170 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4171 can be disabled with the @option{-Wno-jump-misses-init} option.
4173 @item -Wsign-compare
4174 @opindex Wsign-compare
4175 @opindex Wno-sign-compare
4176 @cindex warning for comparison of signed and unsigned values
4177 @cindex comparison of signed and unsigned values, warning
4178 @cindex signed and unsigned values, comparison warning
4179 Warn when a comparison between signed and unsigned values could produce
4180 an incorrect result when the signed value is converted to unsigned.
4181 This warning is also enabled by @option{-Wextra}; to get the other warnings
4182 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4184 @item -Wsign-conversion
4185 @opindex Wsign-conversion
4186 @opindex Wno-sign-conversion
4187 Warn for implicit conversions that may change the sign of an integer
4188 value, like assigning a signed integer expression to an unsigned
4189 integer variable. An explicit cast silences the warning. In C, this
4190 option is enabled also by @option{-Wconversion}.
4194 @opindex Wno-address
4195 Warn about suspicious uses of memory addresses. These include using
4196 the address of a function in a conditional expression, such as
4197 @code{void func(void); if (func)}, and comparisons against the memory
4198 address of a string literal, such as @code{if (x == "abc")}. Such
4199 uses typically indicate a programmer error: the address of a function
4200 always evaluates to true, so their use in a conditional usually
4201 indicate that the programmer forgot the parentheses in a function
4202 call; and comparisons against string literals result in unspecified
4203 behavior and are not portable in C, so they usually indicate that the
4204 programmer intended to use @code{strcmp}. This warning is enabled by
4208 @opindex Wlogical-op
4209 @opindex Wno-logical-op
4210 Warn about suspicious uses of logical operators in expressions.
4211 This includes using logical operators in contexts where a
4212 bit-wise operator is likely to be expected.
4214 @item -Waggregate-return
4215 @opindex Waggregate-return
4216 @opindex Wno-aggregate-return
4217 Warn if any functions that return structures or unions are defined or
4218 called. (In languages where you can return an array, this also elicits
4221 @item -Wno-attributes
4222 @opindex Wno-attributes
4223 @opindex Wattributes
4224 Do not warn if an unexpected @code{__attribute__} is used, such as
4225 unrecognized attributes, function attributes applied to variables,
4226 etc. This will not stop errors for incorrect use of supported
4229 @item -Wno-builtin-macro-redefined
4230 @opindex Wno-builtin-macro-redefined
4231 @opindex Wbuiltin-macro-redefined
4232 Do not warn if certain built-in macros are redefined. This suppresses
4233 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4234 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4236 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4237 @opindex Wstrict-prototypes
4238 @opindex Wno-strict-prototypes
4239 Warn if a function is declared or defined without specifying the
4240 argument types. (An old-style function definition is permitted without
4241 a warning if preceded by a declaration which specifies the argument
4244 @item -Wold-style-declaration @r{(C and Objective-C only)}
4245 @opindex Wold-style-declaration
4246 @opindex Wno-old-style-declaration
4247 Warn for obsolescent usages, according to the C Standard, in a
4248 declaration. For example, warn if storage-class specifiers like
4249 @code{static} are not the first things in a declaration. This warning
4250 is also enabled by @option{-Wextra}.
4252 @item -Wold-style-definition @r{(C and Objective-C only)}
4253 @opindex Wold-style-definition
4254 @opindex Wno-old-style-definition
4255 Warn if an old-style function definition is used. A warning is given
4256 even if there is a previous prototype.
4258 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4259 @opindex Wmissing-parameter-type
4260 @opindex Wno-missing-parameter-type
4261 A function parameter is declared without a type specifier in K&R-style
4268 This warning is also enabled by @option{-Wextra}.
4270 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4271 @opindex Wmissing-prototypes
4272 @opindex Wno-missing-prototypes
4273 Warn if a global function is defined without a previous prototype
4274 declaration. This warning is issued even if the definition itself
4275 provides a prototype. The aim is to detect global functions that fail
4276 to be declared in header files.
4278 @item -Wmissing-declarations
4279 @opindex Wmissing-declarations
4280 @opindex Wno-missing-declarations
4281 Warn if a global function is defined without a previous declaration.
4282 Do so even if the definition itself provides a prototype.
4283 Use this option to detect global functions that are not declared in
4284 header files. In C++, no warnings are issued for function templates,
4285 or for inline functions, or for functions in anonymous namespaces.
4287 @item -Wmissing-field-initializers
4288 @opindex Wmissing-field-initializers
4289 @opindex Wno-missing-field-initializers
4293 Warn if a structure's initializer has some fields missing. For
4294 example, the following code would cause such a warning, because
4295 @code{x.h} is implicitly zero:
4298 struct s @{ int f, g, h; @};
4299 struct s x = @{ 3, 4 @};
4302 This option does not warn about designated initializers, so the following
4303 modification would not trigger a warning:
4306 struct s @{ int f, g, h; @};
4307 struct s x = @{ .f = 3, .g = 4 @};
4310 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4311 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4313 @item -Wmissing-format-attribute
4314 @opindex Wmissing-format-attribute
4315 @opindex Wno-missing-format-attribute
4318 Warn about function pointers which might be candidates for @code{format}
4319 attributes. Note these are only possible candidates, not absolute ones.
4320 GCC will guess that function pointers with @code{format} attributes that
4321 are used in assignment, initialization, parameter passing or return
4322 statements should have a corresponding @code{format} attribute in the
4323 resulting type. I.e.@: the left-hand side of the assignment or
4324 initialization, the type of the parameter variable, or the return type
4325 of the containing function respectively should also have a @code{format}
4326 attribute to avoid the warning.
4328 GCC will also warn about function definitions which might be
4329 candidates for @code{format} attributes. Again, these are only
4330 possible candidates. GCC will guess that @code{format} attributes
4331 might be appropriate for any function that calls a function like
4332 @code{vprintf} or @code{vscanf}, but this might not always be the
4333 case, and some functions for which @code{format} attributes are
4334 appropriate may not be detected.
4336 @item -Wno-multichar
4337 @opindex Wno-multichar
4339 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4340 Usually they indicate a typo in the user's code, as they have
4341 implementation-defined values, and should not be used in portable code.
4343 @item -Wnormalized=<none|id|nfc|nfkc>
4344 @opindex Wnormalized=
4347 @cindex character set, input normalization
4348 In ISO C and ISO C++, two identifiers are different if they are
4349 different sequences of characters. However, sometimes when characters
4350 outside the basic ASCII character set are used, you can have two
4351 different character sequences that look the same. To avoid confusion,
4352 the ISO 10646 standard sets out some @dfn{normalization rules} which
4353 when applied ensure that two sequences that look the same are turned into
4354 the same sequence. GCC can warn you if you are using identifiers which
4355 have not been normalized; this option controls that warning.
4357 There are four levels of warning that GCC supports. The default is
4358 @option{-Wnormalized=nfc}, which warns about any identifier which is
4359 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4360 recommended form for most uses.
4362 Unfortunately, there are some characters which ISO C and ISO C++ allow
4363 in identifiers that when turned into NFC aren't allowable as
4364 identifiers. That is, there's no way to use these symbols in portable
4365 ISO C or C++ and have all your identifiers in NFC@.
4366 @option{-Wnormalized=id} suppresses the warning for these characters.
4367 It is hoped that future versions of the standards involved will correct
4368 this, which is why this option is not the default.
4370 You can switch the warning off for all characters by writing
4371 @option{-Wnormalized=none}. You would only want to do this if you
4372 were using some other normalization scheme (like ``D''), because
4373 otherwise you can easily create bugs that are literally impossible to see.
4375 Some characters in ISO 10646 have distinct meanings but look identical
4376 in some fonts or display methodologies, especially once formatting has
4377 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4378 LETTER N'', will display just like a regular @code{n} which has been
4379 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4380 normalization scheme to convert all these into a standard form as
4381 well, and GCC will warn if your code is not in NFKC if you use
4382 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4383 about every identifier that contains the letter O because it might be
4384 confused with the digit 0, and so is not the default, but may be
4385 useful as a local coding convention if the programming environment is
4386 unable to be fixed to display these characters distinctly.
4388 @item -Wno-deprecated
4389 @opindex Wno-deprecated
4390 @opindex Wdeprecated
4391 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4393 @item -Wno-deprecated-declarations
4394 @opindex Wno-deprecated-declarations
4395 @opindex Wdeprecated-declarations
4396 Do not warn about uses of functions (@pxref{Function Attributes}),
4397 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4398 Attributes}) marked as deprecated by using the @code{deprecated}
4402 @opindex Wno-overflow
4404 Do not warn about compile-time overflow in constant expressions.
4406 @item -Woverride-init @r{(C and Objective-C only)}
4407 @opindex Woverride-init
4408 @opindex Wno-override-init
4412 Warn if an initialized field without side effects is overridden when
4413 using designated initializers (@pxref{Designated Inits, , Designated
4416 This warning is included in @option{-Wextra}. To get other
4417 @option{-Wextra} warnings without this one, use @samp{-Wextra
4418 -Wno-override-init}.
4423 Warn if a structure is given the packed attribute, but the packed
4424 attribute has no effect on the layout or size of the structure.
4425 Such structures may be mis-aligned for little benefit. For
4426 instance, in this code, the variable @code{f.x} in @code{struct bar}
4427 will be misaligned even though @code{struct bar} does not itself
4428 have the packed attribute:
4435 @} __attribute__((packed));
4443 @item -Wpacked-bitfield-compat
4444 @opindex Wpacked-bitfield-compat
4445 @opindex Wno-packed-bitfield-compat
4446 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4447 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4448 the change can lead to differences in the structure layout. GCC
4449 informs you when the offset of such a field has changed in GCC 4.4.
4450 For example there is no longer a 4-bit padding between field @code{a}
4451 and @code{b} in this structure:
4458 @} __attribute__ ((packed));
4461 This warning is enabled by default. Use
4462 @option{-Wno-packed-bitfield-compat} to disable this warning.
4467 Warn if padding is included in a structure, either to align an element
4468 of the structure or to align the whole structure. Sometimes when this
4469 happens it is possible to rearrange the fields of the structure to
4470 reduce the padding and so make the structure smaller.
4472 @item -Wredundant-decls
4473 @opindex Wredundant-decls
4474 @opindex Wno-redundant-decls
4475 Warn if anything is declared more than once in the same scope, even in
4476 cases where multiple declaration is valid and changes nothing.
4478 @item -Wnested-externs @r{(C and Objective-C only)}
4479 @opindex Wnested-externs
4480 @opindex Wno-nested-externs
4481 Warn if an @code{extern} declaration is encountered within a function.
4486 Warn if a function can not be inlined and it was declared as inline.
4487 Even with this option, the compiler will not warn about failures to
4488 inline functions declared in system headers.
4490 The compiler uses a variety of heuristics to determine whether or not
4491 to inline a function. For example, the compiler takes into account
4492 the size of the function being inlined and the amount of inlining
4493 that has already been done in the current function. Therefore,
4494 seemingly insignificant changes in the source program can cause the
4495 warnings produced by @option{-Winline} to appear or disappear.
4497 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4498 @opindex Wno-invalid-offsetof
4499 @opindex Winvalid-offsetof
4500 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4501 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4502 to a non-POD type is undefined. In existing C++ implementations,
4503 however, @samp{offsetof} typically gives meaningful results even when
4504 applied to certain kinds of non-POD types. (Such as a simple
4505 @samp{struct} that fails to be a POD type only by virtue of having a
4506 constructor.) This flag is for users who are aware that they are
4507 writing nonportable code and who have deliberately chosen to ignore the
4510 The restrictions on @samp{offsetof} may be relaxed in a future version
4511 of the C++ standard.
4513 @item -Wno-int-to-pointer-cast
4514 @opindex Wno-int-to-pointer-cast
4515 @opindex Wint-to-pointer-cast
4516 Suppress warnings from casts to pointer type of an integer of a
4517 different size. In C++, casting to a pointer type of smaller size is
4518 an error. @option{Wint-to-pointer-cast} is enabled by default.
4521 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4522 @opindex Wno-pointer-to-int-cast
4523 @opindex Wpointer-to-int-cast
4524 Suppress warnings from casts from a pointer to an integer type of a
4528 @opindex Winvalid-pch
4529 @opindex Wno-invalid-pch
4530 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4531 the search path but can't be used.
4535 @opindex Wno-long-long
4536 Warn if @samp{long long} type is used. This is enabled by either
4537 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4538 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4540 @item -Wvariadic-macros
4541 @opindex Wvariadic-macros
4542 @opindex Wno-variadic-macros
4543 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4544 alternate syntax when in pedantic ISO C99 mode. This is default.
4545 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4547 @item -Wvector-operation-performance
4548 @opindex Wvector-operation-performance
4549 @opindex Wno-vector-operation-performance
4550 Warn if vector operation is not implemented via SIMD capabilities of the
4551 architecture. Mainly useful for the performance tuning.
4552 Vector operation can be implemented @code{piecewise} which means that the
4553 scalar operation is performed on every vector element;
4554 @code{in parallel} which means that the vector operation is implemented
4555 using scalars of wider type, which normally is more performance efficient;
4556 and @code{as a single scalar} which means that vector fits into a
4562 Warn if variable length array is used in the code.
4563 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4564 the variable length array.
4566 @item -Wvolatile-register-var
4567 @opindex Wvolatile-register-var
4568 @opindex Wno-volatile-register-var
4569 Warn if a register variable is declared volatile. The volatile
4570 modifier does not inhibit all optimizations that may eliminate reads
4571 and/or writes to register variables. This warning is enabled by
4574 @item -Wdisabled-optimization
4575 @opindex Wdisabled-optimization
4576 @opindex Wno-disabled-optimization
4577 Warn if a requested optimization pass is disabled. This warning does
4578 not generally indicate that there is anything wrong with your code; it
4579 merely indicates that GCC's optimizers were unable to handle the code
4580 effectively. Often, the problem is that your code is too big or too
4581 complex; GCC will refuse to optimize programs when the optimization
4582 itself is likely to take inordinate amounts of time.
4584 @item -Wpointer-sign @r{(C and Objective-C only)}
4585 @opindex Wpointer-sign
4586 @opindex Wno-pointer-sign
4587 Warn for pointer argument passing or assignment with different signedness.
4588 This option is only supported for C and Objective-C@. It is implied by
4589 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4590 @option{-Wno-pointer-sign}.
4592 @item -Wstack-protector
4593 @opindex Wstack-protector
4594 @opindex Wno-stack-protector
4595 This option is only active when @option{-fstack-protector} is active. It
4596 warns about functions that will not be protected against stack smashing.
4599 @opindex Wno-mudflap
4600 Suppress warnings about constructs that cannot be instrumented by
4603 @item -Woverlength-strings
4604 @opindex Woverlength-strings
4605 @opindex Wno-overlength-strings
4606 Warn about string constants which are longer than the ``minimum
4607 maximum'' length specified in the C standard. Modern compilers
4608 generally allow string constants which are much longer than the
4609 standard's minimum limit, but very portable programs should avoid
4610 using longer strings.
4612 The limit applies @emph{after} string constant concatenation, and does
4613 not count the trailing NUL@. In C90, the limit was 509 characters; in
4614 C99, it was raised to 4095. C++98 does not specify a normative
4615 minimum maximum, so we do not diagnose overlength strings in C++@.
4617 This option is implied by @option{-pedantic}, and can be disabled with
4618 @option{-Wno-overlength-strings}.
4620 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4621 @opindex Wunsuffixed-float-constants
4623 GCC will issue a warning for any floating constant that does not have
4624 a suffix. When used together with @option{-Wsystem-headers} it will
4625 warn about such constants in system header files. This can be useful
4626 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4627 from the decimal floating-point extension to C99.
4630 @node Debugging Options
4631 @section Options for Debugging Your Program or GCC
4632 @cindex options, debugging
4633 @cindex debugging information options
4635 GCC has various special options that are used for debugging
4636 either your program or GCC:
4641 Produce debugging information in the operating system's native format
4642 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4645 On most systems that use stabs format, @option{-g} enables use of extra
4646 debugging information that only GDB can use; this extra information
4647 makes debugging work better in GDB but will probably make other debuggers
4649 refuse to read the program. If you want to control for certain whether
4650 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4651 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4653 GCC allows you to use @option{-g} with
4654 @option{-O}. The shortcuts taken by optimized code may occasionally
4655 produce surprising results: some variables you declared may not exist
4656 at all; flow of control may briefly move where you did not expect it;
4657 some statements may not be executed because they compute constant
4658 results or their values were already at hand; some statements may
4659 execute in different places because they were moved out of loops.
4661 Nevertheless it proves possible to debug optimized output. This makes
4662 it reasonable to use the optimizer for programs that might have bugs.
4664 The following options are useful when GCC is generated with the
4665 capability for more than one debugging format.
4669 Produce debugging information for use by GDB@. This means to use the
4670 most expressive format available (DWARF 2, stabs, or the native format
4671 if neither of those are supported), including GDB extensions if at all
4676 Produce debugging information in stabs format (if that is supported),
4677 without GDB extensions. This is the format used by DBX on most BSD
4678 systems. On MIPS, Alpha and System V Release 4 systems this option
4679 produces stabs debugging output which is not understood by DBX or SDB@.
4680 On System V Release 4 systems this option requires the GNU assembler.
4682 @item -feliminate-unused-debug-symbols
4683 @opindex feliminate-unused-debug-symbols
4684 Produce debugging information in stabs format (if that is supported),
4685 for only symbols that are actually used.
4687 @item -femit-class-debug-always
4688 Instead of emitting debugging information for a C++ class in only one
4689 object file, emit it in all object files using the class. This option
4690 should be used only with debuggers that are unable to handle the way GCC
4691 normally emits debugging information for classes because using this
4692 option will increase the size of debugging information by as much as a
4695 @item -fno-debug-types-section
4696 @opindex fno-debug-types-section
4697 @opindex fdebug-types-section
4698 By default when using DWARF v4 or higher type DIEs will be put into
4699 their own .debug_types section instead of making them part of the
4700 .debug_info section. It is more efficient to put them in a separate
4701 comdat sections since the linker will then be able to remove duplicates.
4702 But not all DWARF consumers support .debug_types sections yet.
4706 Produce debugging information in stabs format (if that is supported),
4707 using GNU extensions understood only by the GNU debugger (GDB)@. The
4708 use of these extensions is likely to make other debuggers crash or
4709 refuse to read the program.
4713 Produce debugging information in COFF format (if that is supported).
4714 This is the format used by SDB on most System V systems prior to
4719 Produce debugging information in XCOFF format (if that is supported).
4720 This is the format used by the DBX debugger on IBM RS/6000 systems.
4724 Produce debugging information in XCOFF format (if that is supported),
4725 using GNU extensions understood only by the GNU debugger (GDB)@. The
4726 use of these extensions is likely to make other debuggers crash or
4727 refuse to read the program, and may cause assemblers other than the GNU
4728 assembler (GAS) to fail with an error.
4730 @item -gdwarf-@var{version}
4731 @opindex gdwarf-@var{version}
4732 Produce debugging information in DWARF format (if that is
4733 supported). This is the format used by DBX on IRIX 6. The value
4734 of @var{version} may be either 2, 3 or 4; the default version is 2.
4736 Note that with DWARF version 2 some ports require, and will always
4737 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4739 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4740 for maximum benefit.
4742 @item -grecord-gcc-switches
4743 @opindex grecord-gcc-switches
4744 This switch causes the command line options, that were used to invoke the
4745 compiler and may affect code generation, to be appended to the
4746 DW_AT_producer attribute in DWARF debugging information. The options
4747 are concatenated with spaces separating them from each other and from
4748 the compiler version. See also @option{-frecord-gcc-switches} for another
4749 way of storing compiler options into the object file.
4751 @item -gno-record-gcc-switches
4752 @opindex gno-record-gcc-switches
4753 Disallow appending command line options to the DW_AT_producer attribute
4754 in DWARF debugging information. This is the default.
4756 @item -gstrict-dwarf
4757 @opindex gstrict-dwarf
4758 Disallow using extensions of later DWARF standard version than selected
4759 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4760 DWARF extensions from later standard versions is allowed.
4762 @item -gno-strict-dwarf
4763 @opindex gno-strict-dwarf
4764 Allow using extensions of later DWARF standard version than selected with
4765 @option{-gdwarf-@var{version}}.
4769 Produce debugging information in VMS debug format (if that is
4770 supported). This is the format used by DEBUG on VMS systems.
4773 @itemx -ggdb@var{level}
4774 @itemx -gstabs@var{level}
4775 @itemx -gcoff@var{level}
4776 @itemx -gxcoff@var{level}
4777 @itemx -gvms@var{level}
4778 Request debugging information and also use @var{level} to specify how
4779 much information. The default level is 2.
4781 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4784 Level 1 produces minimal information, enough for making backtraces in
4785 parts of the program that you don't plan to debug. This includes
4786 descriptions of functions and external variables, but no information
4787 about local variables and no line numbers.
4789 Level 3 includes extra information, such as all the macro definitions
4790 present in the program. Some debuggers support macro expansion when
4791 you use @option{-g3}.
4793 @option{-gdwarf-2} does not accept a concatenated debug level, because
4794 GCC used to support an option @option{-gdwarf} that meant to generate
4795 debug information in version 1 of the DWARF format (which is very
4796 different from version 2), and it would have been too confusing. That
4797 debug format is long obsolete, but the option cannot be changed now.
4798 Instead use an additional @option{-g@var{level}} option to change the
4799 debug level for DWARF.
4803 Turn off generation of debug info, if leaving out this option would have
4804 generated it, or turn it on at level 2 otherwise. The position of this
4805 argument in the command line does not matter, it takes effect after all
4806 other options are processed, and it does so only once, no matter how
4807 many times it is given. This is mainly intended to be used with
4808 @option{-fcompare-debug}.
4810 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4811 @opindex fdump-final-insns
4812 Dump the final internal representation (RTL) to @var{file}. If the
4813 optional argument is omitted (or if @var{file} is @code{.}), the name
4814 of the dump file will be determined by appending @code{.gkd} to the
4815 compilation output file name.
4817 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4818 @opindex fcompare-debug
4819 @opindex fno-compare-debug
4820 If no error occurs during compilation, run the compiler a second time,
4821 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4822 passed to the second compilation. Dump the final internal
4823 representation in both compilations, and print an error if they differ.
4825 If the equal sign is omitted, the default @option{-gtoggle} is used.
4827 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4828 and nonzero, implicitly enables @option{-fcompare-debug}. If
4829 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4830 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4833 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4834 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4835 of the final representation and the second compilation, preventing even
4836 @env{GCC_COMPARE_DEBUG} from taking effect.
4838 To verify full coverage during @option{-fcompare-debug} testing, set
4839 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4840 which GCC will reject as an invalid option in any actual compilation
4841 (rather than preprocessing, assembly or linking). To get just a
4842 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4843 not overridden} will do.
4845 @item -fcompare-debug-second
4846 @opindex fcompare-debug-second
4847 This option is implicitly passed to the compiler for the second
4848 compilation requested by @option{-fcompare-debug}, along with options to
4849 silence warnings, and omitting other options that would cause
4850 side-effect compiler outputs to files or to the standard output. Dump
4851 files and preserved temporary files are renamed so as to contain the
4852 @code{.gk} additional extension during the second compilation, to avoid
4853 overwriting those generated by the first.
4855 When this option is passed to the compiler driver, it causes the
4856 @emph{first} compilation to be skipped, which makes it useful for little
4857 other than debugging the compiler proper.
4859 @item -feliminate-dwarf2-dups
4860 @opindex feliminate-dwarf2-dups
4861 Compress DWARF2 debugging information by eliminating duplicated
4862 information about each symbol. This option only makes sense when
4863 generating DWARF2 debugging information with @option{-gdwarf-2}.
4865 @item -femit-struct-debug-baseonly
4866 Emit debug information for struct-like types
4867 only when the base name of the compilation source file
4868 matches the base name of file in which the struct was defined.
4870 This option substantially reduces the size of debugging information,
4871 but at significant potential loss in type information to the debugger.
4872 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4873 See @option{-femit-struct-debug-detailed} for more detailed control.
4875 This option works only with DWARF 2.
4877 @item -femit-struct-debug-reduced
4878 Emit debug information for struct-like types
4879 only when the base name of the compilation source file
4880 matches the base name of file in which the type was defined,
4881 unless the struct is a template or defined in a system header.
4883 This option significantly reduces the size of debugging information,
4884 with some potential loss in type information to the debugger.
4885 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4886 See @option{-femit-struct-debug-detailed} for more detailed control.
4888 This option works only with DWARF 2.
4890 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4891 Specify the struct-like types
4892 for which the compiler will generate debug information.
4893 The intent is to reduce duplicate struct debug information
4894 between different object files within the same program.
4896 This option is a detailed version of
4897 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4898 which will serve for most needs.
4900 A specification has the syntax@*
4901 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4903 The optional first word limits the specification to
4904 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4905 A struct type is used directly when it is the type of a variable, member.
4906 Indirect uses arise through pointers to structs.
4907 That is, when use of an incomplete struct would be legal, the use is indirect.
4909 @samp{struct one direct; struct two * indirect;}.
4911 The optional second word limits the specification to
4912 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4913 Generic structs are a bit complicated to explain.
4914 For C++, these are non-explicit specializations of template classes,
4915 or non-template classes within the above.
4916 Other programming languages have generics,
4917 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4919 The third word specifies the source files for those
4920 structs for which the compiler will emit debug information.
4921 The values @samp{none} and @samp{any} have the normal meaning.
4922 The value @samp{base} means that
4923 the base of name of the file in which the type declaration appears
4924 must match the base of the name of the main compilation file.
4925 In practice, this means that
4926 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4927 but types declared in other header will not.
4928 The value @samp{sys} means those types satisfying @samp{base}
4929 or declared in system or compiler headers.
4931 You may need to experiment to determine the best settings for your application.
4933 The default is @samp{-femit-struct-debug-detailed=all}.
4935 This option works only with DWARF 2.
4937 @item -fno-merge-debug-strings
4938 @opindex fmerge-debug-strings
4939 @opindex fno-merge-debug-strings
4940 Direct the linker to not merge together strings in the debugging
4941 information which are identical in different object files. Merging is
4942 not supported by all assemblers or linkers. Merging decreases the size
4943 of the debug information in the output file at the cost of increasing
4944 link processing time. Merging is enabled by default.
4946 @item -fdebug-prefix-map=@var{old}=@var{new}
4947 @opindex fdebug-prefix-map
4948 When compiling files in directory @file{@var{old}}, record debugging
4949 information describing them as in @file{@var{new}} instead.
4951 @item -fno-dwarf2-cfi-asm
4952 @opindex fdwarf2-cfi-asm
4953 @opindex fno-dwarf2-cfi-asm
4954 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4955 instead of using GAS @code{.cfi_*} directives.
4957 @cindex @command{prof}
4960 Generate extra code to write profile information suitable for the
4961 analysis program @command{prof}. You must use this option when compiling
4962 the source files you want data about, and you must also use it when
4965 @cindex @command{gprof}
4968 Generate extra code to write profile information suitable for the
4969 analysis program @command{gprof}. You must use this option when compiling
4970 the source files you want data about, and you must also use it when
4975 Makes the compiler print out each function name as it is compiled, and
4976 print some statistics about each pass when it finishes.
4979 @opindex ftime-report
4980 Makes the compiler print some statistics about the time consumed by each
4981 pass when it finishes.
4984 @opindex fmem-report
4985 Makes the compiler print some statistics about permanent memory
4986 allocation when it finishes.
4988 @item -fpre-ipa-mem-report
4989 @opindex fpre-ipa-mem-report
4990 @item -fpost-ipa-mem-report
4991 @opindex fpost-ipa-mem-report
4992 Makes the compiler print some statistics about permanent memory
4993 allocation before or after interprocedural optimization.
4996 @opindex fstack-usage
4997 Makes the compiler output stack usage information for the program, on a
4998 per-function basis. The filename for the dump is made by appending
4999 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5000 the output file, if explicitly specified and it is not an executable,
5001 otherwise it is the basename of the source file. An entry is made up
5006 The name of the function.
5010 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5013 The qualifier @code{static} means that the function manipulates the stack
5014 statically: a fixed number of bytes are allocated for the frame on function
5015 entry and released on function exit; no stack adjustments are otherwise made
5016 in the function. The second field is this fixed number of bytes.
5018 The qualifier @code{dynamic} means that the function manipulates the stack
5019 dynamically: in addition to the static allocation described above, stack
5020 adjustments are made in the body of the function, for example to push/pop
5021 arguments around function calls. If the qualifier @code{bounded} is also
5022 present, the amount of these adjustments is bounded at compile-time and
5023 the second field is an upper bound of the total amount of stack used by
5024 the function. If it is not present, the amount of these adjustments is
5025 not bounded at compile-time and the second field only represents the
5028 @item -fprofile-arcs
5029 @opindex fprofile-arcs
5030 Add code so that program flow @dfn{arcs} are instrumented. During
5031 execution the program records how many times each branch and call is
5032 executed and how many times it is taken or returns. When the compiled
5033 program exits it saves this data to a file called
5034 @file{@var{auxname}.gcda} for each source file. The data may be used for
5035 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5036 test coverage analysis (@option{-ftest-coverage}). Each object file's
5037 @var{auxname} is generated from the name of the output file, if
5038 explicitly specified and it is not the final executable, otherwise it is
5039 the basename of the source file. In both cases any suffix is removed
5040 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5041 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5042 @xref{Cross-profiling}.
5044 @cindex @command{gcov}
5048 This option is used to compile and link code instrumented for coverage
5049 analysis. The option is a synonym for @option{-fprofile-arcs}
5050 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5051 linking). See the documentation for those options for more details.
5056 Compile the source files with @option{-fprofile-arcs} plus optimization
5057 and code generation options. For test coverage analysis, use the
5058 additional @option{-ftest-coverage} option. You do not need to profile
5059 every source file in a program.
5062 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5063 (the latter implies the former).
5066 Run the program on a representative workload to generate the arc profile
5067 information. This may be repeated any number of times. You can run
5068 concurrent instances of your program, and provided that the file system
5069 supports locking, the data files will be correctly updated. Also
5070 @code{fork} calls are detected and correctly handled (double counting
5074 For profile-directed optimizations, compile the source files again with
5075 the same optimization and code generation options plus
5076 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5077 Control Optimization}).
5080 For test coverage analysis, use @command{gcov} to produce human readable
5081 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5082 @command{gcov} documentation for further information.
5086 With @option{-fprofile-arcs}, for each function of your program GCC
5087 creates a program flow graph, then finds a spanning tree for the graph.
5088 Only arcs that are not on the spanning tree have to be instrumented: the
5089 compiler adds code to count the number of times that these arcs are
5090 executed. When an arc is the only exit or only entrance to a block, the
5091 instrumentation code can be added to the block; otherwise, a new basic
5092 block must be created to hold the instrumentation code.
5095 @item -ftest-coverage
5096 @opindex ftest-coverage
5097 Produce a notes file that the @command{gcov} code-coverage utility
5098 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5099 show program coverage. Each source file's note file is called
5100 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5101 above for a description of @var{auxname} and instructions on how to
5102 generate test coverage data. Coverage data will match the source files
5103 more closely, if you do not optimize.
5105 @item -fdbg-cnt-list
5106 @opindex fdbg-cnt-list
5107 Print the name and the counter upper bound for all debug counters.
5110 @item -fdbg-cnt=@var{counter-value-list}
5112 Set the internal debug counter upper bound. @var{counter-value-list}
5113 is a comma-separated list of @var{name}:@var{value} pairs
5114 which sets the upper bound of each debug counter @var{name} to @var{value}.
5115 All debug counters have the initial upper bound of @var{UINT_MAX},
5116 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5117 e.g. With -fdbg-cnt=dce:10,tail_call:0
5118 dbg_cnt(dce) will return true only for first 10 invocations
5120 @itemx -fenable-@var{kind}-@var{pass}
5121 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5125 This is a set of debugging options that are used to explicitly disable/enable
5126 optimization passes. For compiler users, regular options for enabling/disabling
5127 passes should be used instead.
5131 @item -fdisable-ipa-@var{pass}
5132 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5133 statically invoked in the compiler multiple times, the pass name should be
5134 appended with a sequential number starting from 1.
5136 @item -fdisable-rtl-@var{pass}
5137 @item -fdisable-rtl-@var{pass}=@var{range-list}
5138 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5139 statically invoked in the compiler multiple times, the pass name should be
5140 appended with a sequential number starting from 1. @var{range-list} is a comma
5141 seperated list of function ranges or assembler names. Each range is a number
5142 pair seperated by a colon. The range is inclusive in both ends. If the range
5143 is trivial, the number pair can be simplified as a single number. If the
5144 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5145 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5146 function header of a dump file, and the pass names can be dumped by using
5147 option @option{-fdump-passes}.
5149 @item -fdisable-tree-@var{pass}
5150 @item -fdisable-tree-@var{pass}=@var{range-list}
5151 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5154 @item -fenable-ipa-@var{pass}
5155 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5156 statically invoked in the compiler multiple times, the pass name should be
5157 appended with a sequential number starting from 1.
5159 @item -fenable-rtl-@var{pass}
5160 @item -fenable-rtl-@var{pass}=@var{range-list}
5161 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5162 description and examples.
5164 @item -fenable-tree-@var{pass}
5165 @item -fenable-tree-@var{pass}=@var{range-list}
5166 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5167 of option arguments.
5171 # disable ccp1 for all functions
5173 # disable complete unroll for function whose cgraph node uid is 1
5174 -fenable-tree-cunroll=1
5175 # disable gcse2 for functions at the following ranges [1,1],
5176 # [300,400], and [400,1000]
5177 # disable gcse2 for functions foo and foo2
5178 -fdisable-rtl-gcse2=foo,foo2
5179 # disable early inlining
5180 -fdisable-tree-einline
5181 # disable ipa inlining
5182 -fdisable-ipa-inline
5183 # enable tree full unroll
5184 -fenable-tree-unroll
5190 @item -d@var{letters}
5191 @itemx -fdump-rtl-@var{pass}
5193 Says to make debugging dumps during compilation at times specified by
5194 @var{letters}. This is used for debugging the RTL-based passes of the
5195 compiler. The file names for most of the dumps are made by appending
5196 a pass number and a word to the @var{dumpname}, and the files are
5197 created in the directory of the output file. Note that the pass
5198 number is computed statically as passes get registered into the pass
5199 manager. Thus the numbering is not related to the dynamic order of
5200 execution of passes. In particular, a pass installed by a plugin
5201 could have a number over 200 even if it executed quite early.
5202 @var{dumpname} is generated from the name of the output file, if
5203 explicitly specified and it is not an executable, otherwise it is the
5204 basename of the source file. These switches may have different effects
5205 when @option{-E} is used for preprocessing.
5207 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5208 @option{-d} option @var{letters}. Here are the possible
5209 letters for use in @var{pass} and @var{letters}, and their meanings:
5213 @item -fdump-rtl-alignments
5214 @opindex fdump-rtl-alignments
5215 Dump after branch alignments have been computed.
5217 @item -fdump-rtl-asmcons
5218 @opindex fdump-rtl-asmcons
5219 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5221 @item -fdump-rtl-auto_inc_dec
5222 @opindex fdump-rtl-auto_inc_dec
5223 Dump after auto-inc-dec discovery. This pass is only run on
5224 architectures that have auto inc or auto dec instructions.
5226 @item -fdump-rtl-barriers
5227 @opindex fdump-rtl-barriers
5228 Dump after cleaning up the barrier instructions.
5230 @item -fdump-rtl-bbpart
5231 @opindex fdump-rtl-bbpart
5232 Dump after partitioning hot and cold basic blocks.
5234 @item -fdump-rtl-bbro
5235 @opindex fdump-rtl-bbro
5236 Dump after block reordering.
5238 @item -fdump-rtl-btl1
5239 @itemx -fdump-rtl-btl2
5240 @opindex fdump-rtl-btl2
5241 @opindex fdump-rtl-btl2
5242 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5243 after the two branch
5244 target load optimization passes.
5246 @item -fdump-rtl-bypass
5247 @opindex fdump-rtl-bypass
5248 Dump after jump bypassing and control flow optimizations.
5250 @item -fdump-rtl-combine
5251 @opindex fdump-rtl-combine
5252 Dump after the RTL instruction combination pass.
5254 @item -fdump-rtl-compgotos
5255 @opindex fdump-rtl-compgotos
5256 Dump after duplicating the computed gotos.
5258 @item -fdump-rtl-ce1
5259 @itemx -fdump-rtl-ce2
5260 @itemx -fdump-rtl-ce3
5261 @opindex fdump-rtl-ce1
5262 @opindex fdump-rtl-ce2
5263 @opindex fdump-rtl-ce3
5264 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5265 @option{-fdump-rtl-ce3} enable dumping after the three
5266 if conversion passes.
5268 @itemx -fdump-rtl-cprop_hardreg
5269 @opindex fdump-rtl-cprop_hardreg
5270 Dump after hard register copy propagation.
5272 @itemx -fdump-rtl-csa
5273 @opindex fdump-rtl-csa
5274 Dump after combining stack adjustments.
5276 @item -fdump-rtl-cse1
5277 @itemx -fdump-rtl-cse2
5278 @opindex fdump-rtl-cse1
5279 @opindex fdump-rtl-cse2
5280 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5281 the two common sub-expression elimination passes.
5283 @itemx -fdump-rtl-dce
5284 @opindex fdump-rtl-dce
5285 Dump after the standalone dead code elimination passes.
5287 @itemx -fdump-rtl-dbr
5288 @opindex fdump-rtl-dbr
5289 Dump after delayed branch scheduling.
5291 @item -fdump-rtl-dce1
5292 @itemx -fdump-rtl-dce2
5293 @opindex fdump-rtl-dce1
5294 @opindex fdump-rtl-dce2
5295 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5296 the two dead store elimination passes.
5299 @opindex fdump-rtl-eh
5300 Dump after finalization of EH handling code.
5302 @item -fdump-rtl-eh_ranges
5303 @opindex fdump-rtl-eh_ranges
5304 Dump after conversion of EH handling range regions.
5306 @item -fdump-rtl-expand
5307 @opindex fdump-rtl-expand
5308 Dump after RTL generation.
5310 @item -fdump-rtl-fwprop1
5311 @itemx -fdump-rtl-fwprop2
5312 @opindex fdump-rtl-fwprop1
5313 @opindex fdump-rtl-fwprop2
5314 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5315 dumping after the two forward propagation passes.
5317 @item -fdump-rtl-gcse1
5318 @itemx -fdump-rtl-gcse2
5319 @opindex fdump-rtl-gcse1
5320 @opindex fdump-rtl-gcse2
5321 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5322 after global common subexpression elimination.
5324 @item -fdump-rtl-init-regs
5325 @opindex fdump-rtl-init-regs
5326 Dump after the initialization of the registers.
5328 @item -fdump-rtl-initvals
5329 @opindex fdump-rtl-initvals
5330 Dump after the computation of the initial value sets.
5332 @itemx -fdump-rtl-into_cfglayout
5333 @opindex fdump-rtl-into_cfglayout
5334 Dump after converting to cfglayout mode.
5336 @item -fdump-rtl-ira
5337 @opindex fdump-rtl-ira
5338 Dump after iterated register allocation.
5340 @item -fdump-rtl-jump
5341 @opindex fdump-rtl-jump
5342 Dump after the second jump optimization.
5344 @item -fdump-rtl-loop2
5345 @opindex fdump-rtl-loop2
5346 @option{-fdump-rtl-loop2} enables dumping after the rtl
5347 loop optimization passes.
5349 @item -fdump-rtl-mach
5350 @opindex fdump-rtl-mach
5351 Dump after performing the machine dependent reorganization pass, if that
5354 @item -fdump-rtl-mode_sw
5355 @opindex fdump-rtl-mode_sw
5356 Dump after removing redundant mode switches.
5358 @item -fdump-rtl-rnreg
5359 @opindex fdump-rtl-rnreg
5360 Dump after register renumbering.
5362 @itemx -fdump-rtl-outof_cfglayout
5363 @opindex fdump-rtl-outof_cfglayout
5364 Dump after converting from cfglayout mode.
5366 @item -fdump-rtl-peephole2
5367 @opindex fdump-rtl-peephole2
5368 Dump after the peephole pass.
5370 @item -fdump-rtl-postreload
5371 @opindex fdump-rtl-postreload
5372 Dump after post-reload optimizations.
5374 @itemx -fdump-rtl-pro_and_epilogue
5375 @opindex fdump-rtl-pro_and_epilogue
5376 Dump after generating the function pro and epilogues.
5378 @item -fdump-rtl-regmove
5379 @opindex fdump-rtl-regmove
5380 Dump after the register move pass.
5382 @item -fdump-rtl-sched1
5383 @itemx -fdump-rtl-sched2
5384 @opindex fdump-rtl-sched1
5385 @opindex fdump-rtl-sched2
5386 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5387 after the basic block scheduling passes.
5389 @item -fdump-rtl-see
5390 @opindex fdump-rtl-see
5391 Dump after sign extension elimination.
5393 @item -fdump-rtl-seqabstr
5394 @opindex fdump-rtl-seqabstr
5395 Dump after common sequence discovery.
5397 @item -fdump-rtl-shorten
5398 @opindex fdump-rtl-shorten
5399 Dump after shortening branches.
5401 @item -fdump-rtl-sibling
5402 @opindex fdump-rtl-sibling
5403 Dump after sibling call optimizations.
5405 @item -fdump-rtl-split1
5406 @itemx -fdump-rtl-split2
5407 @itemx -fdump-rtl-split3
5408 @itemx -fdump-rtl-split4
5409 @itemx -fdump-rtl-split5
5410 @opindex fdump-rtl-split1
5411 @opindex fdump-rtl-split2
5412 @opindex fdump-rtl-split3
5413 @opindex fdump-rtl-split4
5414 @opindex fdump-rtl-split5
5415 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5416 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5417 @option{-fdump-rtl-split5} enable dumping after five rounds of
5418 instruction splitting.
5420 @item -fdump-rtl-sms
5421 @opindex fdump-rtl-sms
5422 Dump after modulo scheduling. This pass is only run on some
5425 @item -fdump-rtl-stack
5426 @opindex fdump-rtl-stack
5427 Dump after conversion from GCC's "flat register file" registers to the
5428 x87's stack-like registers. This pass is only run on x86 variants.
5430 @item -fdump-rtl-subreg1
5431 @itemx -fdump-rtl-subreg2
5432 @opindex fdump-rtl-subreg1
5433 @opindex fdump-rtl-subreg2
5434 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5435 the two subreg expansion passes.
5437 @item -fdump-rtl-unshare
5438 @opindex fdump-rtl-unshare
5439 Dump after all rtl has been unshared.
5441 @item -fdump-rtl-vartrack
5442 @opindex fdump-rtl-vartrack
5443 Dump after variable tracking.
5445 @item -fdump-rtl-vregs
5446 @opindex fdump-rtl-vregs
5447 Dump after converting virtual registers to hard registers.
5449 @item -fdump-rtl-web
5450 @opindex fdump-rtl-web
5451 Dump after live range splitting.
5453 @item -fdump-rtl-regclass
5454 @itemx -fdump-rtl-subregs_of_mode_init
5455 @itemx -fdump-rtl-subregs_of_mode_finish
5456 @itemx -fdump-rtl-dfinit
5457 @itemx -fdump-rtl-dfinish
5458 @opindex fdump-rtl-regclass
5459 @opindex fdump-rtl-subregs_of_mode_init
5460 @opindex fdump-rtl-subregs_of_mode_finish
5461 @opindex fdump-rtl-dfinit
5462 @opindex fdump-rtl-dfinish
5463 These dumps are defined but always produce empty files.
5465 @item -fdump-rtl-all
5466 @opindex fdump-rtl-all
5467 Produce all the dumps listed above.
5471 Annotate the assembler output with miscellaneous debugging information.
5475 Dump all macro definitions, at the end of preprocessing, in addition to
5480 Produce a core dump whenever an error occurs.
5484 Print statistics on memory usage, at the end of the run, to
5489 Annotate the assembler output with a comment indicating which
5490 pattern and alternative was used. The length of each instruction is
5495 Dump the RTL in the assembler output as a comment before each instruction.
5496 Also turns on @option{-dp} annotation.
5500 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5501 dump a representation of the control flow graph suitable for viewing with VCG
5502 to @file{@var{file}.@var{pass}.vcg}.
5506 Just generate RTL for a function instead of compiling it. Usually used
5507 with @option{-fdump-rtl-expand}.
5511 @opindex fdump-noaddr
5512 When doing debugging dumps, suppress address output. This makes it more
5513 feasible to use diff on debugging dumps for compiler invocations with
5514 different compiler binaries and/or different
5515 text / bss / data / heap / stack / dso start locations.
5517 @item -fdump-unnumbered
5518 @opindex fdump-unnumbered
5519 When doing debugging dumps, suppress instruction numbers and address output.
5520 This makes it more feasible to use diff on debugging dumps for compiler
5521 invocations with different options, in particular with and without
5524 @item -fdump-unnumbered-links
5525 @opindex fdump-unnumbered-links
5526 When doing debugging dumps (see @option{-d} option above), suppress
5527 instruction numbers for the links to the previous and next instructions
5530 @item -fdump-translation-unit @r{(C++ only)}
5531 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5532 @opindex fdump-translation-unit
5533 Dump a representation of the tree structure for the entire translation
5534 unit to a file. The file name is made by appending @file{.tu} to the
5535 source file name, and the file is created in the same directory as the
5536 output file. If the @samp{-@var{options}} form is used, @var{options}
5537 controls the details of the dump as described for the
5538 @option{-fdump-tree} options.
5540 @item -fdump-class-hierarchy @r{(C++ only)}
5541 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5542 @opindex fdump-class-hierarchy
5543 Dump a representation of each class's hierarchy and virtual function
5544 table layout to a file. The file name is made by appending
5545 @file{.class} to the source file name, and the file is created in the
5546 same directory as the output file. If the @samp{-@var{options}} form
5547 is used, @var{options} controls the details of the dump as described
5548 for the @option{-fdump-tree} options.
5550 @item -fdump-ipa-@var{switch}
5552 Control the dumping at various stages of inter-procedural analysis
5553 language tree to a file. The file name is generated by appending a
5554 switch specific suffix to the source file name, and the file is created
5555 in the same directory as the output file. The following dumps are
5560 Enables all inter-procedural analysis dumps.
5563 Dumps information about call-graph optimization, unused function removal,
5564 and inlining decisions.
5567 Dump after function inlining.
5572 @opindex fdump-passes
5573 Dump the list of optimization passes that are turned on and off by
5574 the current command line options.
5576 @item -fdump-statistics-@var{option}
5577 @opindex fdump-statistics
5578 Enable and control dumping of pass statistics in a separate file. The
5579 file name is generated by appending a suffix ending in
5580 @samp{.statistics} to the source file name, and the file is created in
5581 the same directory as the output file. If the @samp{-@var{option}}
5582 form is used, @samp{-stats} will cause counters to be summed over the
5583 whole compilation unit while @samp{-details} will dump every event as
5584 the passes generate them. The default with no option is to sum
5585 counters for each function compiled.
5587 @item -fdump-tree-@var{switch}
5588 @itemx -fdump-tree-@var{switch}-@var{options}
5590 Control the dumping at various stages of processing the intermediate
5591 language tree to a file. The file name is generated by appending a
5592 switch specific suffix to the source file name, and the file is
5593 created in the same directory as the output file. If the
5594 @samp{-@var{options}} form is used, @var{options} is a list of
5595 @samp{-} separated options that control the details of the dump. Not
5596 all options are applicable to all dumps, those which are not
5597 meaningful will be ignored. The following options are available
5601 Print the address of each node. Usually this is not meaningful as it
5602 changes according to the environment and source file. Its primary use
5603 is for tying up a dump file with a debug environment.
5605 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5606 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5607 use working backward from mangled names in the assembly file.
5609 Inhibit dumping of members of a scope or body of a function merely
5610 because that scope has been reached. Only dump such items when they
5611 are directly reachable by some other path. When dumping pretty-printed
5612 trees, this option inhibits dumping the bodies of control structures.
5614 Print a raw representation of the tree. By default, trees are
5615 pretty-printed into a C-like representation.
5617 Enable more detailed dumps (not honored by every dump option).
5619 Enable dumping various statistics about the pass (not honored by every dump
5622 Enable showing basic block boundaries (disabled in raw dumps).
5624 Enable showing virtual operands for every statement.
5626 Enable showing line numbers for statements.
5628 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5630 Enable showing the tree dump for each statement.
5632 Enable showing the EH region number holding each statement.
5634 Enable showing scalar evolution analysis details.
5636 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5637 and @option{lineno}.
5640 The following tree dumps are possible:
5644 @opindex fdump-tree-original
5645 Dump before any tree based optimization, to @file{@var{file}.original}.
5648 @opindex fdump-tree-optimized
5649 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5652 @opindex fdump-tree-gimple
5653 Dump each function before and after the gimplification pass to a file. The
5654 file name is made by appending @file{.gimple} to the source file name.
5657 @opindex fdump-tree-cfg
5658 Dump the control flow graph of each function to a file. The file name is
5659 made by appending @file{.cfg} to the source file name.
5662 @opindex fdump-tree-vcg
5663 Dump the control flow graph of each function to a file in VCG format. The
5664 file name is made by appending @file{.vcg} to the source file name. Note
5665 that if the file contains more than one function, the generated file cannot
5666 be used directly by VCG@. You will need to cut and paste each function's
5667 graph into its own separate file first.
5670 @opindex fdump-tree-ch
5671 Dump each function after copying loop headers. The file name is made by
5672 appending @file{.ch} to the source file name.
5675 @opindex fdump-tree-ssa
5676 Dump SSA related information to a file. The file name is made by appending
5677 @file{.ssa} to the source file name.
5680 @opindex fdump-tree-alias
5681 Dump aliasing information for each function. The file name is made by
5682 appending @file{.alias} to the source file name.
5685 @opindex fdump-tree-ccp
5686 Dump each function after CCP@. The file name is made by appending
5687 @file{.ccp} to the source file name.
5690 @opindex fdump-tree-storeccp
5691 Dump each function after STORE-CCP@. The file name is made by appending
5692 @file{.storeccp} to the source file name.
5695 @opindex fdump-tree-pre
5696 Dump trees after partial redundancy elimination. The file name is made
5697 by appending @file{.pre} to the source file name.
5700 @opindex fdump-tree-fre
5701 Dump trees after full redundancy elimination. The file name is made
5702 by appending @file{.fre} to the source file name.
5705 @opindex fdump-tree-copyprop
5706 Dump trees after copy propagation. The file name is made
5707 by appending @file{.copyprop} to the source file name.
5709 @item store_copyprop
5710 @opindex fdump-tree-store_copyprop
5711 Dump trees after store copy-propagation. The file name is made
5712 by appending @file{.store_copyprop} to the source file name.
5715 @opindex fdump-tree-dce
5716 Dump each function after dead code elimination. The file name is made by
5717 appending @file{.dce} to the source file name.
5720 @opindex fdump-tree-mudflap
5721 Dump each function after adding mudflap instrumentation. The file name is
5722 made by appending @file{.mudflap} to the source file name.
5725 @opindex fdump-tree-sra
5726 Dump each function after performing scalar replacement of aggregates. The
5727 file name is made by appending @file{.sra} to the source file name.
5730 @opindex fdump-tree-sink
5731 Dump each function after performing code sinking. The file name is made
5732 by appending @file{.sink} to the source file name.
5735 @opindex fdump-tree-dom
5736 Dump each function after applying dominator tree optimizations. The file
5737 name is made by appending @file{.dom} to the source file name.
5740 @opindex fdump-tree-dse
5741 Dump each function after applying dead store elimination. The file
5742 name is made by appending @file{.dse} to the source file name.
5745 @opindex fdump-tree-phiopt
5746 Dump each function after optimizing PHI nodes into straightline code. The file
5747 name is made by appending @file{.phiopt} to the source file name.
5750 @opindex fdump-tree-forwprop
5751 Dump each function after forward propagating single use variables. The file
5752 name is made by appending @file{.forwprop} to the source file name.
5755 @opindex fdump-tree-copyrename
5756 Dump each function after applying the copy rename optimization. The file
5757 name is made by appending @file{.copyrename} to the source file name.
5760 @opindex fdump-tree-nrv
5761 Dump each function after applying the named return value optimization on
5762 generic trees. The file name is made by appending @file{.nrv} to the source
5766 @opindex fdump-tree-vect
5767 Dump each function after applying vectorization of loops. The file name is
5768 made by appending @file{.vect} to the source file name.
5771 @opindex fdump-tree-slp
5772 Dump each function after applying vectorization of basic blocks. The file name
5773 is made by appending @file{.slp} to the source file name.
5776 @opindex fdump-tree-vrp
5777 Dump each function after Value Range Propagation (VRP). The file name
5778 is made by appending @file{.vrp} to the source file name.
5781 @opindex fdump-tree-all
5782 Enable all the available tree dumps with the flags provided in this option.
5785 @item -ftree-vectorizer-verbose=@var{n}
5786 @opindex ftree-vectorizer-verbose
5787 This option controls the amount of debugging output the vectorizer prints.
5788 This information is written to standard error, unless
5789 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5790 in which case it is output to the usual dump listing file, @file{.vect}.
5791 For @var{n}=0 no diagnostic information is reported.
5792 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5793 and the total number of loops that got vectorized.
5794 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5795 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5796 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5797 level that @option{-fdump-tree-vect-stats} uses.
5798 Higher verbosity levels mean either more information dumped for each
5799 reported loop, or same amount of information reported for more loops:
5800 if @var{n}=3, vectorizer cost model information is reported.
5801 If @var{n}=4, alignment related information is added to the reports.
5802 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5803 memory access-patterns) is added to the reports.
5804 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5805 that did not pass the first analysis phase (i.e., may not be countable, or
5806 may have complicated control-flow).
5807 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5808 If @var{n}=8, SLP related information is added to the reports.
5809 For @var{n}=9, all the information the vectorizer generates during its
5810 analysis and transformation is reported. This is the same verbosity level
5811 that @option{-fdump-tree-vect-details} uses.
5813 @item -frandom-seed=@var{string}
5814 @opindex frandom-seed
5815 This option provides a seed that GCC uses when it would otherwise use
5816 random numbers. It is used to generate certain symbol names
5817 that have to be different in every compiled file. It is also used to
5818 place unique stamps in coverage data files and the object files that
5819 produce them. You can use the @option{-frandom-seed} option to produce
5820 reproducibly identical object files.
5822 The @var{string} should be different for every file you compile.
5824 @item -fsched-verbose=@var{n}
5825 @opindex fsched-verbose
5826 On targets that use instruction scheduling, this option controls the
5827 amount of debugging output the scheduler prints. This information is
5828 written to standard error, unless @option{-fdump-rtl-sched1} or
5829 @option{-fdump-rtl-sched2} is specified, in which case it is output
5830 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5831 respectively. However for @var{n} greater than nine, the output is
5832 always printed to standard error.
5834 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5835 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5836 For @var{n} greater than one, it also output basic block probabilities,
5837 detailed ready list information and unit/insn info. For @var{n} greater
5838 than two, it includes RTL at abort point, control-flow and regions info.
5839 And for @var{n} over four, @option{-fsched-verbose} also includes
5843 @itemx -save-temps=cwd
5845 Store the usual ``temporary'' intermediate files permanently; place them
5846 in the current directory and name them based on the source file. Thus,
5847 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5848 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5849 preprocessed @file{foo.i} output file even though the compiler now
5850 normally uses an integrated preprocessor.
5852 When used in combination with the @option{-x} command line option,
5853 @option{-save-temps} is sensible enough to avoid over writing an
5854 input source file with the same extension as an intermediate file.
5855 The corresponding intermediate file may be obtained by renaming the
5856 source file before using @option{-save-temps}.
5858 If you invoke GCC in parallel, compiling several different source
5859 files that share a common base name in different subdirectories or the
5860 same source file compiled for multiple output destinations, it is
5861 likely that the different parallel compilers will interfere with each
5862 other, and overwrite the temporary files. For instance:
5865 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5866 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5869 may result in @file{foo.i} and @file{foo.o} being written to
5870 simultaneously by both compilers.
5872 @item -save-temps=obj
5873 @opindex save-temps=obj
5874 Store the usual ``temporary'' intermediate files permanently. If the
5875 @option{-o} option is used, the temporary files are based on the
5876 object file. If the @option{-o} option is not used, the
5877 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5882 gcc -save-temps=obj -c foo.c
5883 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5884 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5887 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5888 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5889 @file{dir2/yfoobar.o}.
5891 @item -time@r{[}=@var{file}@r{]}
5893 Report the CPU time taken by each subprocess in the compilation
5894 sequence. For C source files, this is the compiler proper and assembler
5895 (plus the linker if linking is done).
5897 Without the specification of an output file, the output looks like this:
5904 The first number on each line is the ``user time'', that is time spent
5905 executing the program itself. The second number is ``system time'',
5906 time spent executing operating system routines on behalf of the program.
5907 Both numbers are in seconds.
5909 With the specification of an output file, the output is appended to the
5910 named file, and it looks like this:
5913 0.12 0.01 cc1 @var{options}
5914 0.00 0.01 as @var{options}
5917 The ``user time'' and the ``system time'' are moved before the program
5918 name, and the options passed to the program are displayed, so that one
5919 can later tell what file was being compiled, and with which options.
5921 @item -fvar-tracking
5922 @opindex fvar-tracking
5923 Run variable tracking pass. It computes where variables are stored at each
5924 position in code. Better debugging information is then generated
5925 (if the debugging information format supports this information).
5927 It is enabled by default when compiling with optimization (@option{-Os},
5928 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5929 the debug info format supports it.
5931 @item -fvar-tracking-assignments
5932 @opindex fvar-tracking-assignments
5933 @opindex fno-var-tracking-assignments
5934 Annotate assignments to user variables early in the compilation and
5935 attempt to carry the annotations over throughout the compilation all the
5936 way to the end, in an attempt to improve debug information while
5937 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5939 It can be enabled even if var-tracking is disabled, in which case
5940 annotations will be created and maintained, but discarded at the end.
5942 @item -fvar-tracking-assignments-toggle
5943 @opindex fvar-tracking-assignments-toggle
5944 @opindex fno-var-tracking-assignments-toggle
5945 Toggle @option{-fvar-tracking-assignments}, in the same way that
5946 @option{-gtoggle} toggles @option{-g}.
5948 @item -print-file-name=@var{library}
5949 @opindex print-file-name
5950 Print the full absolute name of the library file @var{library} that
5951 would be used when linking---and don't do anything else. With this
5952 option, GCC does not compile or link anything; it just prints the
5955 @item -print-multi-directory
5956 @opindex print-multi-directory
5957 Print the directory name corresponding to the multilib selected by any
5958 other switches present in the command line. This directory is supposed
5959 to exist in @env{GCC_EXEC_PREFIX}.
5961 @item -print-multi-lib
5962 @opindex print-multi-lib
5963 Print the mapping from multilib directory names to compiler switches
5964 that enable them. The directory name is separated from the switches by
5965 @samp{;}, and each switch starts with an @samp{@@} instead of the
5966 @samp{-}, without spaces between multiple switches. This is supposed to
5967 ease shell-processing.
5969 @item -print-multi-os-directory
5970 @opindex print-multi-os-directory
5971 Print the path to OS libraries for the selected
5972 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5973 present in the @file{lib} subdirectory and no multilibs are used, this is
5974 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5975 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5976 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5977 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5979 @item -print-prog-name=@var{program}
5980 @opindex print-prog-name
5981 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5983 @item -print-libgcc-file-name
5984 @opindex print-libgcc-file-name
5985 Same as @option{-print-file-name=libgcc.a}.
5987 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5988 but you do want to link with @file{libgcc.a}. You can do
5991 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5994 @item -print-search-dirs
5995 @opindex print-search-dirs
5996 Print the name of the configured installation directory and a list of
5997 program and library directories @command{gcc} will search---and don't do anything else.
5999 This is useful when @command{gcc} prints the error message
6000 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6001 To resolve this you either need to put @file{cpp0} and the other compiler
6002 components where @command{gcc} expects to find them, or you can set the environment
6003 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6004 Don't forget the trailing @samp{/}.
6005 @xref{Environment Variables}.
6007 @item -print-sysroot
6008 @opindex print-sysroot
6009 Print the target sysroot directory that will be used during
6010 compilation. This is the target sysroot specified either at configure
6011 time or using the @option{--sysroot} option, possibly with an extra
6012 suffix that depends on compilation options. If no target sysroot is
6013 specified, the option prints nothing.
6015 @item -print-sysroot-headers-suffix
6016 @opindex print-sysroot-headers-suffix
6017 Print the suffix added to the target sysroot when searching for
6018 headers, or give an error if the compiler is not configured with such
6019 a suffix---and don't do anything else.
6022 @opindex dumpmachine
6023 Print the compiler's target machine (for example,
6024 @samp{i686-pc-linux-gnu})---and don't do anything else.
6027 @opindex dumpversion
6028 Print the compiler version (for example, @samp{3.0})---and don't do
6033 Print the compiler's built-in specs---and don't do anything else. (This
6034 is used when GCC itself is being built.) @xref{Spec Files}.
6036 @item -feliminate-unused-debug-types
6037 @opindex feliminate-unused-debug-types
6038 Normally, when producing DWARF2 output, GCC will emit debugging
6039 information for all types declared in a compilation
6040 unit, regardless of whether or not they are actually used
6041 in that compilation unit. Sometimes this is useful, such as
6042 if, in the debugger, you want to cast a value to a type that is
6043 not actually used in your program (but is declared). More often,
6044 however, this results in a significant amount of wasted space.
6045 With this option, GCC will avoid producing debug symbol output
6046 for types that are nowhere used in the source file being compiled.
6049 @node Optimize Options
6050 @section Options That Control Optimization
6051 @cindex optimize options
6052 @cindex options, optimization
6054 These options control various sorts of optimizations.
6056 Without any optimization option, the compiler's goal is to reduce the
6057 cost of compilation and to make debugging produce the expected
6058 results. Statements are independent: if you stop the program with a
6059 breakpoint between statements, you can then assign a new value to any
6060 variable or change the program counter to any other statement in the
6061 function and get exactly the results you would expect from the source
6064 Turning on optimization flags makes the compiler attempt to improve
6065 the performance and/or code size at the expense of compilation time
6066 and possibly the ability to debug the program.
6068 The compiler performs optimization based on the knowledge it has of the
6069 program. Compiling multiple files at once to a single output file mode allows
6070 the compiler to use information gained from all of the files when compiling
6073 Not all optimizations are controlled directly by a flag. Only
6074 optimizations that have a flag are listed in this section.
6076 Most optimizations are only enabled if an @option{-O} level is set on
6077 the command line. Otherwise they are disabled, even if individual
6078 optimization flags are specified.
6080 Depending on the target and how GCC was configured, a slightly different
6081 set of optimizations may be enabled at each @option{-O} level than
6082 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6083 to find out the exact set of optimizations that are enabled at each level.
6084 @xref{Overall Options}, for examples.
6091 Optimize. Optimizing compilation takes somewhat more time, and a lot
6092 more memory for a large function.
6094 With @option{-O}, the compiler tries to reduce code size and execution
6095 time, without performing any optimizations that take a great deal of
6098 @option{-O} turns on the following optimization flags:
6102 -fcprop-registers @gol
6105 -fdelayed-branch @gol
6107 -fguess-branch-probability @gol
6108 -fif-conversion2 @gol
6109 -fif-conversion @gol
6110 -fipa-pure-const @gol
6112 -fipa-reference @gol
6114 -fsplit-wide-types @gol
6116 -ftree-builtin-call-dce @gol
6119 -ftree-copyrename @gol
6121 -ftree-dominator-opts @gol
6123 -ftree-forwprop @gol
6131 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6132 where doing so does not interfere with debugging.
6136 Optimize even more. GCC performs nearly all supported optimizations
6137 that do not involve a space-speed tradeoff.
6138 As compared to @option{-O}, this option increases both compilation time
6139 and the performance of the generated code.
6141 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6142 also turns on the following optimization flags:
6143 @gccoptlist{-fthread-jumps @gol
6144 -falign-functions -falign-jumps @gol
6145 -falign-loops -falign-labels @gol
6148 -fcse-follow-jumps -fcse-skip-blocks @gol
6149 -fdelete-null-pointer-checks @gol
6151 -fexpensive-optimizations @gol
6152 -fgcse -fgcse-lm @gol
6153 -finline-small-functions @gol
6154 -findirect-inlining @gol
6156 -foptimize-sibling-calls @gol
6157 -fpartial-inlining @gol
6160 -freorder-blocks -freorder-functions @gol
6161 -frerun-cse-after-loop @gol
6162 -fsched-interblock -fsched-spec @gol
6163 -fschedule-insns -fschedule-insns2 @gol
6164 -fstrict-aliasing -fstrict-overflow @gol
6165 -ftree-switch-conversion -ftree-tail-merge @gol
6169 Please note the warning under @option{-fgcse} about
6170 invoking @option{-O2} on programs that use computed gotos.
6174 Optimize yet more. @option{-O3} turns on all optimizations specified
6175 by @option{-O2} and also turns on the @option{-finline-functions},
6176 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6177 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6178 @option{-fipa-cp-clone} options.
6182 Reduce compilation time and make debugging produce the expected
6183 results. This is the default.
6187 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6188 do not typically increase code size. It also performs further
6189 optimizations designed to reduce code size.
6191 @option{-Os} disables the following optimization flags:
6192 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6193 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6194 -fprefetch-loop-arrays -ftree-vect-loop-version}
6198 Disregard strict standards compliance. @option{-Ofast} enables all
6199 @option{-O3} optimizations. It also enables optimizations that are not
6200 valid for all standard compliant programs.
6201 It turns on @option{-ffast-math} and the Fortran-specific
6202 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6204 If you use multiple @option{-O} options, with or without level numbers,
6205 the last such option is the one that is effective.
6208 Options of the form @option{-f@var{flag}} specify machine-independent
6209 flags. Most flags have both positive and negative forms; the negative
6210 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6211 below, only one of the forms is listed---the one you typically will
6212 use. You can figure out the other form by either removing @samp{no-}
6215 The following options control specific optimizations. They are either
6216 activated by @option{-O} options or are related to ones that are. You
6217 can use the following flags in the rare cases when ``fine-tuning'' of
6218 optimizations to be performed is desired.
6221 @item -fno-default-inline
6222 @opindex fno-default-inline
6223 Do not make member functions inline by default merely because they are
6224 defined inside the class scope (C++ only). Otherwise, when you specify
6225 @w{@option{-O}}, member functions defined inside class scope are compiled
6226 inline by default; i.e., you don't need to add @samp{inline} in front of
6227 the member function name.
6229 @item -fno-defer-pop
6230 @opindex fno-defer-pop
6231 Always pop the arguments to each function call as soon as that function
6232 returns. For machines which must pop arguments after a function call,
6233 the compiler normally lets arguments accumulate on the stack for several
6234 function calls and pops them all at once.
6236 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6238 @item -fforward-propagate
6239 @opindex fforward-propagate
6240 Perform a forward propagation pass on RTL@. The pass tries to combine two
6241 instructions and checks if the result can be simplified. If loop unrolling
6242 is active, two passes are performed and the second is scheduled after
6245 This option is enabled by default at optimization levels @option{-O},
6246 @option{-O2}, @option{-O3}, @option{-Os}.
6248 @item -ffp-contract=@var{style}
6249 @opindex ffp-contract
6250 @option{-ffp-contract=off} disables floating-point expression contraction.
6251 @option{-ffp-contract=fast} enables floating-point expression contraction
6252 such as forming of fused multiply-add operations if the target has
6253 native support for them.
6254 @option{-ffp-contract=on} enables floating-point expression contraction
6255 if allowed by the language standard. This is currently not implemented
6256 and treated equal to @option{-ffp-contract=off}.
6258 The default is @option{-ffp-contract=fast}.
6260 @item -fomit-frame-pointer
6261 @opindex fomit-frame-pointer
6262 Don't keep the frame pointer in a register for functions that
6263 don't need one. This avoids the instructions to save, set up and
6264 restore frame pointers; it also makes an extra register available
6265 in many functions. @strong{It also makes debugging impossible on
6268 On some machines, such as the VAX, this flag has no effect, because
6269 the standard calling sequence automatically handles the frame pointer
6270 and nothing is saved by pretending it doesn't exist. The
6271 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6272 whether a target machine supports this flag. @xref{Registers,,Register
6273 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6275 Starting with GCC version 4.6, the default setting (when not optimizing for
6276 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6277 @option{-fomit-frame-pointer}. The default can be reverted to
6278 @option{-fno-omit-frame-pointer} by configuring GCC with the
6279 @option{--enable-frame-pointer} configure option.
6281 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6283 @item -foptimize-sibling-calls
6284 @opindex foptimize-sibling-calls
6285 Optimize sibling and tail recursive calls.
6287 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6291 Don't pay attention to the @code{inline} keyword. Normally this option
6292 is used to keep the compiler from expanding any functions inline.
6293 Note that if you are not optimizing, no functions can be expanded inline.
6295 @item -finline-small-functions
6296 @opindex finline-small-functions
6297 Integrate functions into their callers when their body is smaller than expected
6298 function call code (so overall size of program gets smaller). The compiler
6299 heuristically decides which functions are simple enough to be worth integrating
6302 Enabled at level @option{-O2}.
6304 @item -findirect-inlining
6305 @opindex findirect-inlining
6306 Inline also indirect calls that are discovered to be known at compile
6307 time thanks to previous inlining. This option has any effect only
6308 when inlining itself is turned on by the @option{-finline-functions}
6309 or @option{-finline-small-functions} options.
6311 Enabled at level @option{-O2}.
6313 @item -finline-functions
6314 @opindex finline-functions
6315 Integrate all simple functions into their callers. The compiler
6316 heuristically decides which functions are simple enough to be worth
6317 integrating in this way.
6319 If all calls to a given function are integrated, and the function is
6320 declared @code{static}, then the function is normally not output as
6321 assembler code in its own right.
6323 Enabled at level @option{-O3}.
6325 @item -finline-functions-called-once
6326 @opindex finline-functions-called-once
6327 Consider all @code{static} functions called once for inlining into their
6328 caller even if they are not marked @code{inline}. If a call to a given
6329 function is integrated, then the function is not output as assembler code
6332 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6334 @item -fearly-inlining
6335 @opindex fearly-inlining
6336 Inline functions marked by @code{always_inline} and functions whose body seems
6337 smaller than the function call overhead early before doing
6338 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6339 makes profiling significantly cheaper and usually inlining faster on programs
6340 having large chains of nested wrapper functions.
6346 Perform interprocedural scalar replacement of aggregates, removal of
6347 unused parameters and replacement of parameters passed by reference
6348 by parameters passed by value.
6350 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6352 @item -finline-limit=@var{n}
6353 @opindex finline-limit
6354 By default, GCC limits the size of functions that can be inlined. This flag
6355 allows coarse control of this limit. @var{n} is the size of functions that
6356 can be inlined in number of pseudo instructions.
6358 Inlining is actually controlled by a number of parameters, which may be
6359 specified individually by using @option{--param @var{name}=@var{value}}.
6360 The @option{-finline-limit=@var{n}} option sets some of these parameters
6364 @item max-inline-insns-single
6365 is set to @var{n}/2.
6366 @item max-inline-insns-auto
6367 is set to @var{n}/2.
6370 See below for a documentation of the individual
6371 parameters controlling inlining and for the defaults of these parameters.
6373 @emph{Note:} there may be no value to @option{-finline-limit} that results
6374 in default behavior.
6376 @emph{Note:} pseudo instruction represents, in this particular context, an
6377 abstract measurement of function's size. In no way does it represent a count
6378 of assembly instructions and as such its exact meaning might change from one
6379 release to an another.
6381 @item -fno-keep-inline-dllexport
6382 @opindex -fno-keep-inline-dllexport
6383 This is a more fine-grained version of @option{-fkeep-inline-functions},
6384 which applies only to functions that are declared using the @code{dllexport}
6385 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6388 @item -fkeep-inline-functions
6389 @opindex fkeep-inline-functions
6390 In C, emit @code{static} functions that are declared @code{inline}
6391 into the object file, even if the function has been inlined into all
6392 of its callers. This switch does not affect functions using the
6393 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6394 inline functions into the object file.
6396 @item -fkeep-static-consts
6397 @opindex fkeep-static-consts
6398 Emit variables declared @code{static const} when optimization isn't turned
6399 on, even if the variables aren't referenced.
6401 GCC enables this option by default. If you want to force the compiler to
6402 check if the variable was referenced, regardless of whether or not
6403 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6405 @item -fmerge-constants
6406 @opindex fmerge-constants
6407 Attempt to merge identical constants (string constants and floating point
6408 constants) across compilation units.
6410 This option is the default for optimized compilation if the assembler and
6411 linker support it. Use @option{-fno-merge-constants} to inhibit this
6414 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6416 @item -fmerge-all-constants
6417 @opindex fmerge-all-constants
6418 Attempt to merge identical constants and identical variables.
6420 This option implies @option{-fmerge-constants}. In addition to
6421 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6422 arrays or initialized constant variables with integral or floating point
6423 types. Languages like C or C++ require each variable, including multiple
6424 instances of the same variable in recursive calls, to have distinct locations,
6425 so using this option will result in non-conforming
6428 @item -fmodulo-sched
6429 @opindex fmodulo-sched
6430 Perform swing modulo scheduling immediately before the first scheduling
6431 pass. This pass looks at innermost loops and reorders their
6432 instructions by overlapping different iterations.
6434 @item -fmodulo-sched-allow-regmoves
6435 @opindex fmodulo-sched-allow-regmoves
6436 Perform more aggressive SMS based modulo scheduling with register moves
6437 allowed. By setting this flag certain anti-dependences edges will be
6438 deleted which will trigger the generation of reg-moves based on the
6439 life-range analysis. This option is effective only with
6440 @option{-fmodulo-sched} enabled.
6442 @item -fno-branch-count-reg
6443 @opindex fno-branch-count-reg
6444 Do not use ``decrement and branch'' instructions on a count register,
6445 but instead generate a sequence of instructions that decrement a
6446 register, compare it against zero, then branch based upon the result.
6447 This option is only meaningful on architectures that support such
6448 instructions, which include x86, PowerPC, IA-64 and S/390.
6450 The default is @option{-fbranch-count-reg}.
6452 @item -fno-function-cse
6453 @opindex fno-function-cse
6454 Do not put function addresses in registers; make each instruction that
6455 calls a constant function contain the function's address explicitly.
6457 This option results in less efficient code, but some strange hacks
6458 that alter the assembler output may be confused by the optimizations
6459 performed when this option is not used.
6461 The default is @option{-ffunction-cse}
6463 @item -fno-zero-initialized-in-bss
6464 @opindex fno-zero-initialized-in-bss
6465 If the target supports a BSS section, GCC by default puts variables that
6466 are initialized to zero into BSS@. This can save space in the resulting
6469 This option turns off this behavior because some programs explicitly
6470 rely on variables going to the data section. E.g., so that the
6471 resulting executable can find the beginning of that section and/or make
6472 assumptions based on that.
6474 The default is @option{-fzero-initialized-in-bss}.
6476 @item -fmudflap -fmudflapth -fmudflapir
6480 @cindex bounds checking
6482 For front-ends that support it (C and C++), instrument all risky
6483 pointer/array dereferencing operations, some standard library
6484 string/heap functions, and some other associated constructs with
6485 range/validity tests. Modules so instrumented should be immune to
6486 buffer overflows, invalid heap use, and some other classes of C/C++
6487 programming errors. The instrumentation relies on a separate runtime
6488 library (@file{libmudflap}), which will be linked into a program if
6489 @option{-fmudflap} is given at link time. Run-time behavior of the
6490 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6491 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6494 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6495 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6496 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6497 instrumentation should ignore pointer reads. This produces less
6498 instrumentation (and therefore faster execution) and still provides
6499 some protection against outright memory corrupting writes, but allows
6500 erroneously read data to propagate within a program.
6502 @item -fthread-jumps
6503 @opindex fthread-jumps
6504 Perform optimizations where we check to see if a jump branches to a
6505 location where another comparison subsumed by the first is found. If
6506 so, the first branch is redirected to either the destination of the
6507 second branch or a point immediately following it, depending on whether
6508 the condition is known to be true or false.
6510 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6512 @item -fsplit-wide-types
6513 @opindex fsplit-wide-types
6514 When using a type that occupies multiple registers, such as @code{long
6515 long} on a 32-bit system, split the registers apart and allocate them
6516 independently. This normally generates better code for those types,
6517 but may make debugging more difficult.
6519 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6522 @item -fcse-follow-jumps
6523 @opindex fcse-follow-jumps
6524 In common subexpression elimination (CSE), scan through jump instructions
6525 when the target of the jump is not reached by any other path. For
6526 example, when CSE encounters an @code{if} statement with an
6527 @code{else} clause, CSE will follow the jump when the condition
6530 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6532 @item -fcse-skip-blocks
6533 @opindex fcse-skip-blocks
6534 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6535 follow jumps which conditionally skip over blocks. When CSE
6536 encounters a simple @code{if} statement with no else clause,
6537 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6538 body of the @code{if}.
6540 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6542 @item -frerun-cse-after-loop
6543 @opindex frerun-cse-after-loop
6544 Re-run common subexpression elimination after loop optimizations has been
6547 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6551 Perform a global common subexpression elimination pass.
6552 This pass also performs global constant and copy propagation.
6554 @emph{Note:} When compiling a program using computed gotos, a GCC
6555 extension, you may get better runtime performance if you disable
6556 the global common subexpression elimination pass by adding
6557 @option{-fno-gcse} to the command line.
6559 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6563 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6564 attempt to move loads which are only killed by stores into themselves. This
6565 allows a loop containing a load/store sequence to be changed to a load outside
6566 the loop, and a copy/store within the loop.
6568 Enabled by default when gcse is enabled.
6572 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6573 global common subexpression elimination. This pass will attempt to move
6574 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6575 loops containing a load/store sequence can be changed to a load before
6576 the loop and a store after the loop.
6578 Not enabled at any optimization level.
6582 When @option{-fgcse-las} is enabled, the global common subexpression
6583 elimination pass eliminates redundant loads that come after stores to the
6584 same memory location (both partial and full redundancies).
6586 Not enabled at any optimization level.
6588 @item -fgcse-after-reload
6589 @opindex fgcse-after-reload
6590 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6591 pass is performed after reload. The purpose of this pass is to cleanup
6594 @item -funsafe-loop-optimizations
6595 @opindex funsafe-loop-optimizations
6596 If given, the loop optimizer will assume that loop indices do not
6597 overflow, and that the loops with nontrivial exit condition are not
6598 infinite. This enables a wider range of loop optimizations even if
6599 the loop optimizer itself cannot prove that these assumptions are valid.
6600 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6601 if it finds this kind of loop.
6603 @item -fcrossjumping
6604 @opindex fcrossjumping
6605 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6606 resulting code may or may not perform better than without cross-jumping.
6608 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6610 @item -fauto-inc-dec
6611 @opindex fauto-inc-dec
6612 Combine increments or decrements of addresses with memory accesses.
6613 This pass is always skipped on architectures that do not have
6614 instructions to support this. Enabled by default at @option{-O} and
6615 higher on architectures that support this.
6619 Perform dead code elimination (DCE) on RTL@.
6620 Enabled by default at @option{-O} and higher.
6624 Perform dead store elimination (DSE) on RTL@.
6625 Enabled by default at @option{-O} and higher.
6627 @item -fif-conversion
6628 @opindex fif-conversion
6629 Attempt to transform conditional jumps into branch-less equivalents. This
6630 include use of conditional moves, min, max, set flags and abs instructions, and
6631 some tricks doable by standard arithmetics. The use of conditional execution
6632 on chips where it is available is controlled by @code{if-conversion2}.
6634 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6636 @item -fif-conversion2
6637 @opindex fif-conversion2
6638 Use conditional execution (where available) to transform conditional jumps into
6639 branch-less equivalents.
6641 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6643 @item -fdelete-null-pointer-checks
6644 @opindex fdelete-null-pointer-checks
6645 Assume that programs cannot safely dereference null pointers, and that
6646 no code or data element resides there. This enables simple constant
6647 folding optimizations at all optimization levels. In addition, other
6648 optimization passes in GCC use this flag to control global dataflow
6649 analyses that eliminate useless checks for null pointers; these assume
6650 that if a pointer is checked after it has already been dereferenced,
6653 Note however that in some environments this assumption is not true.
6654 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6655 for programs which depend on that behavior.
6657 Some targets, especially embedded ones, disable this option at all levels.
6658 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6659 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6660 are enabled independently at different optimization levels.
6662 @item -fdevirtualize
6663 @opindex fdevirtualize
6664 Attempt to convert calls to virtual functions to direct calls. This
6665 is done both within a procedure and interprocedurally as part of
6666 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6667 propagation (@option{-fipa-cp}).
6668 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6670 @item -fexpensive-optimizations
6671 @opindex fexpensive-optimizations
6672 Perform a number of minor optimizations that are relatively expensive.
6674 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6676 @item -foptimize-register-move
6678 @opindex foptimize-register-move
6680 Attempt to reassign register numbers in move instructions and as
6681 operands of other simple instructions in order to maximize the amount of
6682 register tying. This is especially helpful on machines with two-operand
6685 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6688 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6690 @item -fira-algorithm=@var{algorithm}
6691 Use specified coloring algorithm for the integrated register
6692 allocator. The @var{algorithm} argument should be @code{priority} or
6693 @code{CB}. The first algorithm specifies Chow's priority coloring,
6694 the second one specifies Chaitin-Briggs coloring. The second
6695 algorithm can be unimplemented for some architectures. If it is
6696 implemented, it is the default because Chaitin-Briggs coloring as a
6697 rule generates a better code.
6699 @item -fira-region=@var{region}
6700 Use specified regions for the integrated register allocator. The
6701 @var{region} argument should be one of @code{all}, @code{mixed}, or
6702 @code{one}. The first value means using all loops as register
6703 allocation regions, the second value which is the default means using
6704 all loops except for loops with small register pressure as the
6705 regions, and third one means using all function as a single region.
6706 The first value can give best result for machines with small size and
6707 irregular register set, the third one results in faster and generates
6708 decent code and the smallest size code, and the default value usually
6709 give the best results in most cases and for most architectures.
6711 @item -fira-loop-pressure
6712 @opindex fira-loop-pressure
6713 Use IRA to evaluate register pressure in loops for decision to move
6714 loop invariants. Usage of this option usually results in generation
6715 of faster and smaller code on machines with big register files (>= 32
6716 registers) but it can slow compiler down.
6718 This option is enabled at level @option{-O3} for some targets.
6720 @item -fno-ira-share-save-slots
6721 @opindex fno-ira-share-save-slots
6722 Switch off sharing stack slots used for saving call used hard
6723 registers living through a call. Each hard register will get a
6724 separate stack slot and as a result function stack frame will be
6727 @item -fno-ira-share-spill-slots
6728 @opindex fno-ira-share-spill-slots
6729 Switch off sharing stack slots allocated for pseudo-registers. Each
6730 pseudo-register which did not get a hard register will get a separate
6731 stack slot and as a result function stack frame will be bigger.
6733 @item -fira-verbose=@var{n}
6734 @opindex fira-verbose
6735 Set up how verbose dump file for the integrated register allocator
6736 will be. Default value is 5. If the value is greater or equal to 10,
6737 the dump file will be stderr as if the value were @var{n} minus 10.
6739 @item -fdelayed-branch
6740 @opindex fdelayed-branch
6741 If supported for the target machine, attempt to reorder instructions
6742 to exploit instruction slots available after delayed branch
6745 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6747 @item -fschedule-insns
6748 @opindex fschedule-insns
6749 If supported for the target machine, attempt to reorder instructions to
6750 eliminate execution stalls due to required data being unavailable. This
6751 helps machines that have slow floating point or memory load instructions
6752 by allowing other instructions to be issued until the result of the load
6753 or floating point instruction is required.
6755 Enabled at levels @option{-O2}, @option{-O3}.
6757 @item -fschedule-insns2
6758 @opindex fschedule-insns2
6759 Similar to @option{-fschedule-insns}, but requests an additional pass of
6760 instruction scheduling after register allocation has been done. This is
6761 especially useful on machines with a relatively small number of
6762 registers and where memory load instructions take more than one cycle.
6764 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6766 @item -fno-sched-interblock
6767 @opindex fno-sched-interblock
6768 Don't schedule instructions across basic blocks. This is normally
6769 enabled by default when scheduling before register allocation, i.e.@:
6770 with @option{-fschedule-insns} or at @option{-O2} or higher.
6772 @item -fno-sched-spec
6773 @opindex fno-sched-spec
6774 Don't allow speculative motion of non-load instructions. This is normally
6775 enabled by default when scheduling before register allocation, i.e.@:
6776 with @option{-fschedule-insns} or at @option{-O2} or higher.
6778 @item -fsched-pressure
6779 @opindex fsched-pressure
6780 Enable register pressure sensitive insn scheduling before the register
6781 allocation. This only makes sense when scheduling before register
6782 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6783 @option{-O2} or higher. Usage of this option can improve the
6784 generated code and decrease its size by preventing register pressure
6785 increase above the number of available hard registers and as a
6786 consequence register spills in the register allocation.
6788 @item -fsched-spec-load
6789 @opindex fsched-spec-load
6790 Allow speculative motion of some load instructions. This only makes
6791 sense when scheduling before register allocation, i.e.@: with
6792 @option{-fschedule-insns} or at @option{-O2} or higher.
6794 @item -fsched-spec-load-dangerous
6795 @opindex fsched-spec-load-dangerous
6796 Allow speculative motion of more load instructions. This only makes
6797 sense when scheduling before register allocation, i.e.@: with
6798 @option{-fschedule-insns} or at @option{-O2} or higher.
6800 @item -fsched-stalled-insns
6801 @itemx -fsched-stalled-insns=@var{n}
6802 @opindex fsched-stalled-insns
6803 Define how many insns (if any) can be moved prematurely from the queue
6804 of stalled insns into the ready list, during the second scheduling pass.
6805 @option{-fno-sched-stalled-insns} means that no insns will be moved
6806 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6807 on how many queued insns can be moved prematurely.
6808 @option{-fsched-stalled-insns} without a value is equivalent to
6809 @option{-fsched-stalled-insns=1}.
6811 @item -fsched-stalled-insns-dep
6812 @itemx -fsched-stalled-insns-dep=@var{n}
6813 @opindex fsched-stalled-insns-dep
6814 Define how many insn groups (cycles) will be examined for a dependency
6815 on a stalled insn that is candidate for premature removal from the queue
6816 of stalled insns. This has an effect only during the second scheduling pass,
6817 and only if @option{-fsched-stalled-insns} is used.
6818 @option{-fno-sched-stalled-insns-dep} is equivalent to
6819 @option{-fsched-stalled-insns-dep=0}.
6820 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6821 @option{-fsched-stalled-insns-dep=1}.
6823 @item -fsched2-use-superblocks
6824 @opindex fsched2-use-superblocks
6825 When scheduling after register allocation, do use superblock scheduling
6826 algorithm. Superblock scheduling allows motion across basic block boundaries
6827 resulting on faster schedules. This option is experimental, as not all machine
6828 descriptions used by GCC model the CPU closely enough to avoid unreliable
6829 results from the algorithm.
6831 This only makes sense when scheduling after register allocation, i.e.@: with
6832 @option{-fschedule-insns2} or at @option{-O2} or higher.
6834 @item -fsched-group-heuristic
6835 @opindex fsched-group-heuristic
6836 Enable the group heuristic in the scheduler. This heuristic favors
6837 the instruction that belongs to a schedule group. This is enabled
6838 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6839 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6841 @item -fsched-critical-path-heuristic
6842 @opindex fsched-critical-path-heuristic
6843 Enable the critical-path heuristic in the scheduler. This heuristic favors
6844 instructions on the critical path. This is enabled by default when
6845 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6846 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6848 @item -fsched-spec-insn-heuristic
6849 @opindex fsched-spec-insn-heuristic
6850 Enable the speculative instruction heuristic in the scheduler. This
6851 heuristic favors speculative instructions with greater dependency weakness.
6852 This is enabled by default when scheduling is enabled, i.e.@:
6853 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6854 or at @option{-O2} or higher.
6856 @item -fsched-rank-heuristic
6857 @opindex fsched-rank-heuristic
6858 Enable the rank heuristic in the scheduler. This heuristic favors
6859 the instruction belonging to a basic block with greater size or frequency.
6860 This is enabled by default when scheduling is enabled, i.e.@:
6861 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6862 at @option{-O2} or higher.
6864 @item -fsched-last-insn-heuristic
6865 @opindex fsched-last-insn-heuristic
6866 Enable the last-instruction heuristic in the scheduler. This heuristic
6867 favors the instruction that is less dependent on the last instruction
6868 scheduled. This is enabled by default when scheduling is enabled,
6869 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6870 at @option{-O2} or higher.
6872 @item -fsched-dep-count-heuristic
6873 @opindex fsched-dep-count-heuristic
6874 Enable the dependent-count heuristic in the scheduler. This heuristic
6875 favors the instruction that has more instructions depending on it.
6876 This is enabled by default when scheduling is enabled, i.e.@:
6877 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6878 at @option{-O2} or higher.
6880 @item -freschedule-modulo-scheduled-loops
6881 @opindex freschedule-modulo-scheduled-loops
6882 The modulo scheduling comes before the traditional scheduling, if a loop
6883 was modulo scheduled we may want to prevent the later scheduling passes
6884 from changing its schedule, we use this option to control that.
6886 @item -fselective-scheduling
6887 @opindex fselective-scheduling
6888 Schedule instructions using selective scheduling algorithm. Selective
6889 scheduling runs instead of the first scheduler pass.
6891 @item -fselective-scheduling2
6892 @opindex fselective-scheduling2
6893 Schedule instructions using selective scheduling algorithm. Selective
6894 scheduling runs instead of the second scheduler pass.
6896 @item -fsel-sched-pipelining
6897 @opindex fsel-sched-pipelining
6898 Enable software pipelining of innermost loops during selective scheduling.
6899 This option has no effect until one of @option{-fselective-scheduling} or
6900 @option{-fselective-scheduling2} is turned on.
6902 @item -fsel-sched-pipelining-outer-loops
6903 @opindex fsel-sched-pipelining-outer-loops
6904 When pipelining loops during selective scheduling, also pipeline outer loops.
6905 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6908 @opindex fshrink-wrap
6909 Emit function prologues only before parts of the function that need it,
6910 rather than at the top of the function. This flag is enabled by default at
6911 @option{-O} and higher.
6913 @item -fcaller-saves
6914 @opindex fcaller-saves
6915 Enable values to be allocated in registers that will be clobbered by
6916 function calls, by emitting extra instructions to save and restore the
6917 registers around such calls. Such allocation is done only when it
6918 seems to result in better code than would otherwise be produced.
6920 This option is always enabled by default on certain machines, usually
6921 those which have no call-preserved registers to use instead.
6923 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6925 @item -fcombine-stack-adjustments
6926 @opindex fcombine-stack-adjustments
6927 Tracks stack adjustments (pushes and pops) and stack memory references
6928 and then tries to find ways to combine them.
6930 Enabled by default at @option{-O1} and higher.
6932 @item -fconserve-stack
6933 @opindex fconserve-stack
6934 Attempt to minimize stack usage. The compiler will attempt to use less
6935 stack space, even if that makes the program slower. This option
6936 implies setting the @option{large-stack-frame} parameter to 100
6937 and the @option{large-stack-frame-growth} parameter to 400.
6939 @item -ftree-reassoc
6940 @opindex ftree-reassoc
6941 Perform reassociation on trees. This flag is enabled by default
6942 at @option{-O} and higher.
6946 Perform partial redundancy elimination (PRE) on trees. This flag is
6947 enabled by default at @option{-O2} and @option{-O3}.
6949 @item -ftree-forwprop
6950 @opindex ftree-forwprop
6951 Perform forward propagation on trees. This flag is enabled by default
6952 at @option{-O} and higher.
6956 Perform full redundancy elimination (FRE) on trees. The difference
6957 between FRE and PRE is that FRE only considers expressions
6958 that are computed on all paths leading to the redundant computation.
6959 This analysis is faster than PRE, though it exposes fewer redundancies.
6960 This flag is enabled by default at @option{-O} and higher.
6962 @item -ftree-phiprop
6963 @opindex ftree-phiprop
6964 Perform hoisting of loads from conditional pointers on trees. This
6965 pass is enabled by default at @option{-O} and higher.
6967 @item -ftree-copy-prop
6968 @opindex ftree-copy-prop
6969 Perform copy propagation on trees. This pass eliminates unnecessary
6970 copy operations. This flag is enabled by default at @option{-O} and
6973 @item -fipa-pure-const
6974 @opindex fipa-pure-const
6975 Discover which functions are pure or constant.
6976 Enabled by default at @option{-O} and higher.
6978 @item -fipa-reference
6979 @opindex fipa-reference
6980 Discover which static variables do not escape cannot escape the
6982 Enabled by default at @option{-O} and higher.
6986 Perform interprocedural pointer analysis and interprocedural modification
6987 and reference analysis. This option can cause excessive memory and
6988 compile-time usage on large compilation units. It is not enabled by
6989 default at any optimization level.
6992 @opindex fipa-profile
6993 Perform interprocedural profile propagation. The functions called only from
6994 cold functions are marked as cold. Also functions executed once (such as
6995 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6996 functions and loop less parts of functions executed once are then optimized for
6998 Enabled by default at @option{-O} and higher.
7002 Perform interprocedural constant propagation.
7003 This optimization analyzes the program to determine when values passed
7004 to functions are constants and then optimizes accordingly.
7005 This optimization can substantially increase performance
7006 if the application has constants passed to functions.
7007 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7009 @item -fipa-cp-clone
7010 @opindex fipa-cp-clone
7011 Perform function cloning to make interprocedural constant propagation stronger.
7012 When enabled, interprocedural constant propagation will perform function cloning
7013 when externally visible function can be called with constant arguments.
7014 Because this optimization can create multiple copies of functions,
7015 it may significantly increase code size
7016 (see @option{--param ipcp-unit-growth=@var{value}}).
7017 This flag is enabled by default at @option{-O3}.
7019 @item -fipa-matrix-reorg
7020 @opindex fipa-matrix-reorg
7021 Perform matrix flattening and transposing.
7022 Matrix flattening tries to replace an @math{m}-dimensional matrix
7023 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
7024 This reduces the level of indirection needed for accessing the elements
7025 of the matrix. The second optimization is matrix transposing that
7026 attempts to change the order of the matrix's dimensions in order to
7027 improve cache locality.
7028 Both optimizations need the @option{-fwhole-program} flag.
7029 Transposing is enabled only if profiling information is available.
7033 Perform forward store motion on trees. This flag is
7034 enabled by default at @option{-O} and higher.
7036 @item -ftree-bit-ccp
7037 @opindex ftree-bit-ccp
7038 Perform sparse conditional bit constant propagation on trees and propagate
7039 pointer alignment information.
7040 This pass only operates on local scalar variables and is enabled by default
7041 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7045 Perform sparse conditional constant propagation (CCP) on trees. This
7046 pass only operates on local scalar variables and is enabled by default
7047 at @option{-O} and higher.
7049 @item -ftree-switch-conversion
7050 Perform conversion of simple initializations in a switch to
7051 initializations from a scalar array. This flag is enabled by default
7052 at @option{-O2} and higher.
7054 @item -ftree-tail-merge
7055 Look for identical code sequences. When found, replace one with a jump to the
7056 other. This optimization is known as tail merging or cross jumping. This flag
7057 is enabled by default at @option{-O2} and higher. The run time of this pass can
7058 be limited using @option{max-tail-merge-comparisons} parameter and
7059 @option{max-tail-merge-iterations} parameter.
7063 Perform dead code elimination (DCE) on trees. This flag is enabled by
7064 default at @option{-O} and higher.
7066 @item -ftree-builtin-call-dce
7067 @opindex ftree-builtin-call-dce
7068 Perform conditional dead code elimination (DCE) for calls to builtin functions
7069 that may set @code{errno} but are otherwise side-effect free. This flag is
7070 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7073 @item -ftree-dominator-opts
7074 @opindex ftree-dominator-opts
7075 Perform a variety of simple scalar cleanups (constant/copy
7076 propagation, redundancy elimination, range propagation and expression
7077 simplification) based on a dominator tree traversal. This also
7078 performs jump threading (to reduce jumps to jumps). This flag is
7079 enabled by default at @option{-O} and higher.
7083 Perform dead store elimination (DSE) on trees. A dead store is a store into
7084 a memory location which will later be overwritten by another store without
7085 any intervening loads. In this case the earlier store can be deleted. This
7086 flag is enabled by default at @option{-O} and higher.
7090 Perform loop header copying on trees. This is beneficial since it increases
7091 effectiveness of code motion optimizations. It also saves one jump. This flag
7092 is enabled by default at @option{-O} and higher. It is not enabled
7093 for @option{-Os}, since it usually increases code size.
7095 @item -ftree-loop-optimize
7096 @opindex ftree-loop-optimize
7097 Perform loop optimizations on trees. This flag is enabled by default
7098 at @option{-O} and higher.
7100 @item -ftree-loop-linear
7101 @opindex ftree-loop-linear
7102 Perform loop interchange transformations on tree. Same as
7103 @option{-floop-interchange}. To use this code transformation, GCC has
7104 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7105 enable the Graphite loop transformation infrastructure.
7107 @item -floop-interchange
7108 @opindex floop-interchange
7109 Perform loop interchange transformations on loops. Interchanging two
7110 nested loops switches the inner and outer loops. For example, given a
7115 A(J, I) = A(J, I) * C
7119 loop interchange will transform the loop as if the user had written:
7123 A(J, I) = A(J, I) * C
7127 which can be beneficial when @code{N} is larger than the caches,
7128 because in Fortran, the elements of an array are stored in memory
7129 contiguously by column, and the original loop iterates over rows,
7130 potentially creating at each access a cache miss. This optimization
7131 applies to all the languages supported by GCC and is not limited to
7132 Fortran. To use this code transformation, GCC has to be configured
7133 with @option{--with-ppl} and @option{--with-cloog} to enable the
7134 Graphite loop transformation infrastructure.
7136 @item -floop-strip-mine
7137 @opindex floop-strip-mine
7138 Perform loop strip mining transformations on loops. Strip mining
7139 splits a loop into two nested loops. The outer loop has strides
7140 equal to the strip size and the inner loop has strides of the
7141 original loop within a strip. The strip length can be changed
7142 using the @option{loop-block-tile-size} parameter. For example,
7149 loop strip mining will transform the loop as if the user had written:
7152 DO I = II, min (II + 50, N)
7157 This optimization applies to all the languages supported by GCC and is
7158 not limited to Fortran. To use this code transformation, GCC has to
7159 be configured with @option{--with-ppl} and @option{--with-cloog} to
7160 enable the Graphite loop transformation infrastructure.
7163 @opindex floop-block
7164 Perform loop blocking transformations on loops. Blocking strip mines
7165 each loop in the loop nest such that the memory accesses of the
7166 element loops fit inside caches. The strip length can be changed
7167 using the @option{loop-block-tile-size} parameter. For example, given
7172 A(J, I) = B(I) + C(J)
7176 loop blocking will transform the loop as if the user had written:
7180 DO I = II, min (II + 50, N)
7181 DO J = JJ, min (JJ + 50, M)
7182 A(J, I) = B(I) + C(J)
7188 which can be beneficial when @code{M} is larger than the caches,
7189 because the innermost loop will iterate over a smaller amount of data
7190 that can be kept in the caches. This optimization applies to all the
7191 languages supported by GCC and is not limited to Fortran. To use this
7192 code transformation, GCC has to be configured with @option{--with-ppl}
7193 and @option{--with-cloog} to enable the Graphite loop transformation
7196 @item -fgraphite-identity
7197 @opindex fgraphite-identity
7198 Enable the identity transformation for graphite. For every SCoP we generate
7199 the polyhedral representation and transform it back to gimple. Using
7200 @option{-fgraphite-identity} we can check the costs or benefits of the
7201 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7202 are also performed by the code generator CLooG, like index splitting and
7203 dead code elimination in loops.
7205 @item -floop-flatten
7206 @opindex floop-flatten
7207 Removes the loop nesting structure: transforms the loop nest into a
7208 single loop. This transformation can be useful to vectorize all the
7209 levels of the loop nest.
7211 @item -floop-parallelize-all
7212 @opindex floop-parallelize-all
7213 Use the Graphite data dependence analysis to identify loops that can
7214 be parallelized. Parallelize all the loops that can be analyzed to
7215 not contain loop carried dependences without checking that it is
7216 profitable to parallelize the loops.
7218 @item -fcheck-data-deps
7219 @opindex fcheck-data-deps
7220 Compare the results of several data dependence analyzers. This option
7221 is used for debugging the data dependence analyzers.
7223 @item -ftree-loop-if-convert
7224 Attempt to transform conditional jumps in the innermost loops to
7225 branch-less equivalents. The intent is to remove control-flow from
7226 the innermost loops in order to improve the ability of the
7227 vectorization pass to handle these loops. This is enabled by default
7228 if vectorization is enabled.
7230 @item -ftree-loop-if-convert-stores
7231 Attempt to also if-convert conditional jumps containing memory writes.
7232 This transformation can be unsafe for multi-threaded programs as it
7233 transforms conditional memory writes into unconditional memory writes.
7236 for (i = 0; i < N; i++)
7240 would be transformed to
7242 for (i = 0; i < N; i++)
7243 A[i] = cond ? expr : A[i];
7245 potentially producing data races.
7247 @item -ftree-loop-distribution
7248 Perform loop distribution. This flag can improve cache performance on
7249 big loop bodies and allow further loop optimizations, like
7250 parallelization or vectorization, to take place. For example, the loop
7267 @item -ftree-loop-distribute-patterns
7268 Perform loop distribution of patterns that can be code generated with
7269 calls to a library. This flag is enabled by default at @option{-O3}.
7271 This pass distributes the initialization loops and generates a call to
7272 memset zero. For example, the loop
7288 and the initialization loop is transformed into a call to memset zero.
7290 @item -ftree-loop-im
7291 @opindex ftree-loop-im
7292 Perform loop invariant motion on trees. This pass moves only invariants that
7293 would be hard to handle at RTL level (function calls, operations that expand to
7294 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7295 operands of conditions that are invariant out of the loop, so that we can use
7296 just trivial invariantness analysis in loop unswitching. The pass also includes
7299 @item -ftree-loop-ivcanon
7300 @opindex ftree-loop-ivcanon
7301 Create a canonical counter for number of iterations in the loop for that
7302 determining number of iterations requires complicated analysis. Later
7303 optimizations then may determine the number easily. Useful especially
7304 in connection with unrolling.
7308 Perform induction variable optimizations (strength reduction, induction
7309 variable merging and induction variable elimination) on trees.
7311 @item -ftree-parallelize-loops=n
7312 @opindex ftree-parallelize-loops
7313 Parallelize loops, i.e., split their iteration space to run in n threads.
7314 This is only possible for loops whose iterations are independent
7315 and can be arbitrarily reordered. The optimization is only
7316 profitable on multiprocessor machines, for loops that are CPU-intensive,
7317 rather than constrained e.g.@: by memory bandwidth. This option
7318 implies @option{-pthread}, and thus is only supported on targets
7319 that have support for @option{-pthread}.
7323 Perform function-local points-to analysis on trees. This flag is
7324 enabled by default at @option{-O} and higher.
7328 Perform scalar replacement of aggregates. This pass replaces structure
7329 references with scalars to prevent committing structures to memory too
7330 early. This flag is enabled by default at @option{-O} and higher.
7332 @item -ftree-copyrename
7333 @opindex ftree-copyrename
7334 Perform copy renaming on trees. This pass attempts to rename compiler
7335 temporaries to other variables at copy locations, usually resulting in
7336 variable names which more closely resemble the original variables. This flag
7337 is enabled by default at @option{-O} and higher.
7341 Perform temporary expression replacement during the SSA->normal phase. Single
7342 use/single def temporaries are replaced at their use location with their
7343 defining expression. This results in non-GIMPLE code, but gives the expanders
7344 much more complex trees to work on resulting in better RTL generation. This is
7345 enabled by default at @option{-O} and higher.
7347 @item -ftree-vectorize
7348 @opindex ftree-vectorize
7349 Perform loop vectorization on trees. This flag is enabled by default at
7352 @item -ftree-slp-vectorize
7353 @opindex ftree-slp-vectorize
7354 Perform basic block vectorization on trees. This flag is enabled by default at
7355 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7357 @item -ftree-vect-loop-version
7358 @opindex ftree-vect-loop-version
7359 Perform loop versioning when doing loop vectorization on trees. When a loop
7360 appears to be vectorizable except that data alignment or data dependence cannot
7361 be determined at compile time then vectorized and non-vectorized versions of
7362 the loop are generated along with runtime checks for alignment or dependence
7363 to control which version is executed. This option is enabled by default
7364 except at level @option{-Os} where it is disabled.
7366 @item -fvect-cost-model
7367 @opindex fvect-cost-model
7368 Enable cost model for vectorization.
7372 Perform Value Range Propagation on trees. This is similar to the
7373 constant propagation pass, but instead of values, ranges of values are
7374 propagated. This allows the optimizers to remove unnecessary range
7375 checks like array bound checks and null pointer checks. This is
7376 enabled by default at @option{-O2} and higher. Null pointer check
7377 elimination is only done if @option{-fdelete-null-pointer-checks} is
7382 Perform tail duplication to enlarge superblock size. This transformation
7383 simplifies the control flow of the function allowing other optimizations to do
7386 @item -funroll-loops
7387 @opindex funroll-loops
7388 Unroll loops whose number of iterations can be determined at compile
7389 time or upon entry to the loop. @option{-funroll-loops} implies
7390 @option{-frerun-cse-after-loop}. This option makes code larger,
7391 and may or may not make it run faster.
7393 @item -funroll-all-loops
7394 @opindex funroll-all-loops
7395 Unroll all loops, even if their number of iterations is uncertain when
7396 the loop is entered. This usually makes programs run more slowly.
7397 @option{-funroll-all-loops} implies the same options as
7398 @option{-funroll-loops},
7400 @item -fsplit-ivs-in-unroller
7401 @opindex fsplit-ivs-in-unroller
7402 Enables expressing of values of induction variables in later iterations
7403 of the unrolled loop using the value in the first iteration. This breaks
7404 long dependency chains, thus improving efficiency of the scheduling passes.
7406 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7407 same effect. However in cases the loop body is more complicated than
7408 a single basic block, this is not reliable. It also does not work at all
7409 on some of the architectures due to restrictions in the CSE pass.
7411 This optimization is enabled by default.
7413 @item -fvariable-expansion-in-unroller
7414 @opindex fvariable-expansion-in-unroller
7415 With this option, the compiler will create multiple copies of some
7416 local variables when unrolling a loop which can result in superior code.
7418 @item -fpartial-inlining
7419 @opindex fpartial-inlining
7420 Inline parts of functions. This option has any effect only
7421 when inlining itself is turned on by the @option{-finline-functions}
7422 or @option{-finline-small-functions} options.
7424 Enabled at level @option{-O2}.
7426 @item -fpredictive-commoning
7427 @opindex fpredictive-commoning
7428 Perform predictive commoning optimization, i.e., reusing computations
7429 (especially memory loads and stores) performed in previous
7430 iterations of loops.
7432 This option is enabled at level @option{-O3}.
7434 @item -fprefetch-loop-arrays
7435 @opindex fprefetch-loop-arrays
7436 If supported by the target machine, generate instructions to prefetch
7437 memory to improve the performance of loops that access large arrays.
7439 This option may generate better or worse code; results are highly
7440 dependent on the structure of loops within the source code.
7442 Disabled at level @option{-Os}.
7445 @itemx -fno-peephole2
7446 @opindex fno-peephole
7447 @opindex fno-peephole2
7448 Disable any machine-specific peephole optimizations. The difference
7449 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7450 are implemented in the compiler; some targets use one, some use the
7451 other, a few use both.
7453 @option{-fpeephole} is enabled by default.
7454 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7456 @item -fno-guess-branch-probability
7457 @opindex fno-guess-branch-probability
7458 Do not guess branch probabilities using heuristics.
7460 GCC will use heuristics to guess branch probabilities if they are
7461 not provided by profiling feedback (@option{-fprofile-arcs}). These
7462 heuristics are based on the control flow graph. If some branch probabilities
7463 are specified by @samp{__builtin_expect}, then the heuristics will be
7464 used to guess branch probabilities for the rest of the control flow graph,
7465 taking the @samp{__builtin_expect} info into account. The interactions
7466 between the heuristics and @samp{__builtin_expect} can be complex, and in
7467 some cases, it may be useful to disable the heuristics so that the effects
7468 of @samp{__builtin_expect} are easier to understand.
7470 The default is @option{-fguess-branch-probability} at levels
7471 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7473 @item -freorder-blocks
7474 @opindex freorder-blocks
7475 Reorder basic blocks in the compiled function in order to reduce number of
7476 taken branches and improve code locality.
7478 Enabled at levels @option{-O2}, @option{-O3}.
7480 @item -freorder-blocks-and-partition
7481 @opindex freorder-blocks-and-partition
7482 In addition to reordering basic blocks in the compiled function, in order
7483 to reduce number of taken branches, partitions hot and cold basic blocks
7484 into separate sections of the assembly and .o files, to improve
7485 paging and cache locality performance.
7487 This optimization is automatically turned off in the presence of
7488 exception handling, for linkonce sections, for functions with a user-defined
7489 section attribute and on any architecture that does not support named
7492 @item -freorder-functions
7493 @opindex freorder-functions
7494 Reorder functions in the object file in order to
7495 improve code locality. This is implemented by using special
7496 subsections @code{.text.hot} for most frequently executed functions and
7497 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7498 the linker so object file format must support named sections and linker must
7499 place them in a reasonable way.
7501 Also profile feedback must be available in to make this option effective. See
7502 @option{-fprofile-arcs} for details.
7504 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7506 @item -fstrict-aliasing
7507 @opindex fstrict-aliasing
7508 Allow the compiler to assume the strictest aliasing rules applicable to
7509 the language being compiled. For C (and C++), this activates
7510 optimizations based on the type of expressions. In particular, an
7511 object of one type is assumed never to reside at the same address as an
7512 object of a different type, unless the types are almost the same. For
7513 example, an @code{unsigned int} can alias an @code{int}, but not a
7514 @code{void*} or a @code{double}. A character type may alias any other
7517 @anchor{Type-punning}Pay special attention to code like this:
7530 The practice of reading from a different union member than the one most
7531 recently written to (called ``type-punning'') is common. Even with
7532 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7533 is accessed through the union type. So, the code above will work as
7534 expected. @xref{Structures unions enumerations and bit-fields
7535 implementation}. However, this code might not:
7546 Similarly, access by taking the address, casting the resulting pointer
7547 and dereferencing the result has undefined behavior, even if the cast
7548 uses a union type, e.g.:
7552 return ((union a_union *) &d)->i;
7556 The @option{-fstrict-aliasing} option is enabled at levels
7557 @option{-O2}, @option{-O3}, @option{-Os}.
7559 @item -fstrict-overflow
7560 @opindex fstrict-overflow
7561 Allow the compiler to assume strict signed overflow rules, depending
7562 on the language being compiled. For C (and C++) this means that
7563 overflow when doing arithmetic with signed numbers is undefined, which
7564 means that the compiler may assume that it will not happen. This
7565 permits various optimizations. For example, the compiler will assume
7566 that an expression like @code{i + 10 > i} will always be true for
7567 signed @code{i}. This assumption is only valid if signed overflow is
7568 undefined, as the expression is false if @code{i + 10} overflows when
7569 using twos complement arithmetic. When this option is in effect any
7570 attempt to determine whether an operation on signed numbers will
7571 overflow must be written carefully to not actually involve overflow.
7573 This option also allows the compiler to assume strict pointer
7574 semantics: given a pointer to an object, if adding an offset to that
7575 pointer does not produce a pointer to the same object, the addition is
7576 undefined. This permits the compiler to conclude that @code{p + u >
7577 p} is always true for a pointer @code{p} and unsigned integer
7578 @code{u}. This assumption is only valid because pointer wraparound is
7579 undefined, as the expression is false if @code{p + u} overflows using
7580 twos complement arithmetic.
7582 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7583 that integer signed overflow is fully defined: it wraps. When
7584 @option{-fwrapv} is used, there is no difference between
7585 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7586 integers. With @option{-fwrapv} certain types of overflow are
7587 permitted. For example, if the compiler gets an overflow when doing
7588 arithmetic on constants, the overflowed value can still be used with
7589 @option{-fwrapv}, but not otherwise.
7591 The @option{-fstrict-overflow} option is enabled at levels
7592 @option{-O2}, @option{-O3}, @option{-Os}.
7594 @item -falign-functions
7595 @itemx -falign-functions=@var{n}
7596 @opindex falign-functions
7597 Align the start of functions to the next power-of-two greater than
7598 @var{n}, skipping up to @var{n} bytes. For instance,
7599 @option{-falign-functions=32} aligns functions to the next 32-byte
7600 boundary, but @option{-falign-functions=24} would align to the next
7601 32-byte boundary only if this can be done by skipping 23 bytes or less.
7603 @option{-fno-align-functions} and @option{-falign-functions=1} are
7604 equivalent and mean that functions will not be aligned.
7606 Some assemblers only support this flag when @var{n} is a power of two;
7607 in that case, it is rounded up.
7609 If @var{n} is not specified or is zero, use a machine-dependent default.
7611 Enabled at levels @option{-O2}, @option{-O3}.
7613 @item -falign-labels
7614 @itemx -falign-labels=@var{n}
7615 @opindex falign-labels
7616 Align all branch targets to a power-of-two boundary, skipping up to
7617 @var{n} bytes like @option{-falign-functions}. This option can easily
7618 make code slower, because it must insert dummy operations for when the
7619 branch target is reached in the usual flow of the code.
7621 @option{-fno-align-labels} and @option{-falign-labels=1} are
7622 equivalent and mean that labels will not be aligned.
7624 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7625 are greater than this value, then their values are used instead.
7627 If @var{n} is not specified or is zero, use a machine-dependent default
7628 which is very likely to be @samp{1}, meaning no alignment.
7630 Enabled at levels @option{-O2}, @option{-O3}.
7633 @itemx -falign-loops=@var{n}
7634 @opindex falign-loops
7635 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7636 like @option{-falign-functions}. The hope is that the loop will be
7637 executed many times, which will make up for any execution of the dummy
7640 @option{-fno-align-loops} and @option{-falign-loops=1} are
7641 equivalent and mean that loops will not be aligned.
7643 If @var{n} is not specified or is zero, use a machine-dependent default.
7645 Enabled at levels @option{-O2}, @option{-O3}.
7648 @itemx -falign-jumps=@var{n}
7649 @opindex falign-jumps
7650 Align branch targets to a power-of-two boundary, for branch targets
7651 where the targets can only be reached by jumping, skipping up to @var{n}
7652 bytes like @option{-falign-functions}. In this case, no dummy operations
7655 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7656 equivalent and mean that loops will not be aligned.
7658 If @var{n} is not specified or is zero, use a machine-dependent default.
7660 Enabled at levels @option{-O2}, @option{-O3}.
7662 @item -funit-at-a-time
7663 @opindex funit-at-a-time
7664 This option is left for compatibility reasons. @option{-funit-at-a-time}
7665 has no effect, while @option{-fno-unit-at-a-time} implies
7666 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7670 @item -fno-toplevel-reorder
7671 @opindex fno-toplevel-reorder
7672 Do not reorder top-level functions, variables, and @code{asm}
7673 statements. Output them in the same order that they appear in the
7674 input file. When this option is used, unreferenced static variables
7675 will not be removed. This option is intended to support existing code
7676 which relies on a particular ordering. For new code, it is better to
7679 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7680 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7685 Constructs webs as commonly used for register allocation purposes and assign
7686 each web individual pseudo register. This allows the register allocation pass
7687 to operate on pseudos directly, but also strengthens several other optimization
7688 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7689 however, make debugging impossible, since variables will no longer stay in a
7692 Enabled by default with @option{-funroll-loops}.
7694 @item -fwhole-program
7695 @opindex fwhole-program
7696 Assume that the current compilation unit represents the whole program being
7697 compiled. All public functions and variables with the exception of @code{main}
7698 and those merged by attribute @code{externally_visible} become static functions
7699 and in effect are optimized more aggressively by interprocedural optimizers. If @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary.
7700 While this option is equivalent to proper use of the @code{static} keyword for
7701 programs consisting of a single file, in combination with option
7702 @option{-flto} this flag can be used to
7703 compile many smaller scale programs since the functions and variables become
7704 local for the whole combined compilation unit, not for the single source file
7707 This option implies @option{-fwhole-file} for Fortran programs.
7709 @item -flto[=@var{n}]
7711 This option runs the standard link-time optimizer. When invoked
7712 with source code, it generates GIMPLE (one of GCC's internal
7713 representations) and writes it to special ELF sections in the object
7714 file. When the object files are linked together, all the function
7715 bodies are read from these ELF sections and instantiated as if they
7716 had been part of the same translation unit.
7718 To use the link-timer optimizer, @option{-flto} needs to be specified at
7719 compile time and during the final link. For example,
7722 gcc -c -O2 -flto foo.c
7723 gcc -c -O2 -flto bar.c
7724 gcc -o myprog -flto -O2 foo.o bar.o
7727 The first two invocations to GCC will save a bytecode representation
7728 of GIMPLE into special ELF sections inside @file{foo.o} and
7729 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7730 @file{foo.o} and @file{bar.o}, merge the two files into a single
7731 internal image, and compile the result as usual. Since both
7732 @file{foo.o} and @file{bar.o} are merged into a single image, this
7733 causes all the inter-procedural analyses and optimizations in GCC to
7734 work across the two files as if they were a single one. This means,
7735 for example, that the inliner will be able to inline functions in
7736 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7738 Another (simpler) way to enable link-time optimization is,
7741 gcc -o myprog -flto -O2 foo.c bar.c
7744 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7745 merge them together into a single GIMPLE representation and optimize
7746 them as usual to produce @file{myprog}.
7748 The only important thing to keep in mind is that to enable link-time
7749 optimizations the @option{-flto} flag needs to be passed to both the
7750 compile and the link commands.
7752 To make whole program optimization effective, it is necessary to make
7753 certain whole program assumptions. The compiler needs to know
7754 what functions and variables can be accessed by libraries and runtime
7755 outside of the link time optimized unit. When supported by the linker,
7756 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7757 compiler information about used and externally visible symbols. When
7758 the linker plugin is not available, @option{-fwhole-program} should be
7759 used to allow the compiler to make these assumptions, which will lead
7760 to more aggressive optimization decisions.
7762 Note that when a file is compiled with @option{-flto}, the generated
7763 object file will be larger than a regular object file because it will
7764 contain GIMPLE bytecodes and the usual final code. This means that
7765 object files with LTO information can be linked as a normal object
7766 file. So, in the previous example, if the final link is done with
7769 gcc -o myprog foo.o bar.o
7772 The only difference will be that no inter-procedural optimizations
7773 will be applied to produce @file{myprog}. The two object files
7774 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7777 Additionally, the optimization flags used to compile individual files
7778 are not necessarily related to those used at link-time. For instance,
7781 gcc -c -O0 -flto foo.c
7782 gcc -c -O0 -flto bar.c
7783 gcc -o myprog -flto -O3 foo.o bar.o
7786 This will produce individual object files with unoptimized assembler
7787 code, but the resulting binary @file{myprog} will be optimized at
7788 @option{-O3}. Now, if the final binary is generated without
7789 @option{-flto}, then @file{myprog} will not be optimized.
7791 When producing the final binary with @option{-flto}, GCC will only
7792 apply link-time optimizations to those files that contain bytecode.
7793 Therefore, you can mix and match object files and libraries with
7794 GIMPLE bytecodes and final object code. GCC will automatically select
7795 which files to optimize in LTO mode and which files to link without
7798 There are some code generation flags that GCC will preserve when
7799 generating bytecodes, as they need to be used during the final link
7800 stage. Currently, the following options are saved into the GIMPLE
7801 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7802 @option{-m} target flags.
7804 At link time, these options are read-in and reapplied. Note that the
7805 current implementation makes no attempt at recognizing conflicting
7806 values for these options. If two or more files have a conflicting
7807 value (e.g., one file is compiled with @option{-fPIC} and another
7808 isn't), the compiler will simply use the last value read from the
7809 bytecode files. It is recommended, then, that all the files
7810 participating in the same link be compiled with the same options.
7812 Another feature of LTO is that it is possible to apply interprocedural
7813 optimizations on files written in different languages. This requires
7814 some support in the language front end. Currently, the C, C++ and
7815 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7816 something like this should work
7821 gfortran -c -flto baz.f90
7822 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7825 Notice that the final link is done with @command{g++} to get the C++
7826 runtime libraries and @option{-lgfortran} is added to get the Fortran
7827 runtime libraries. In general, when mixing languages in LTO mode, you
7828 should use the same link command used when mixing languages in a
7829 regular (non-LTO) compilation. This means that if your build process
7830 was mixing languages before, all you need to add is @option{-flto} to
7831 all the compile and link commands.
7833 If LTO encounters objects with C linkage declared with incompatible
7834 types in separate translation units to be linked together (undefined
7835 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7836 issued. The behavior is still undefined at runtime.
7838 If object files containing GIMPLE bytecode are stored in a library archive, say
7839 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7840 are using a linker with linker plugin support. To enable this feature, use
7841 the flag @option{-fuse-linker-plugin} at link-time:
7844 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7847 With the linker plugin enabled, the linker will extract the needed
7848 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7849 to make them part of the aggregated GIMPLE image to be optimized.
7851 If you are not using a linker with linker plugin support and/or do not
7852 enable linker plugin then the objects inside @file{libfoo.a}
7853 will be extracted and linked as usual, but they will not participate
7854 in the LTO optimization process.
7856 Link time optimizations do not require the presence of the whole program to
7857 operate. If the program does not require any symbols to be exported, it is
7858 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7859 the interprocedural optimizers to use more aggressive assumptions which may
7860 lead to improved optimization opportunities.
7861 Use of @option{-fwhole-program} is not needed when linker plugin is
7862 active (see @option{-fuse-linker-plugin}).
7864 Regarding portability: the current implementation of LTO makes no
7865 attempt at generating bytecode that can be ported between different
7866 types of hosts. The bytecode files are versioned and there is a
7867 strict version check, so bytecode files generated in one version of
7868 GCC will not work with an older/newer version of GCC.
7870 Link time optimization does not play well with generating debugging
7871 information. Combining @option{-flto} with
7872 @option{-g} is currently experimental and expected to produce wrong
7875 If you specify the optional @var{n}, the optimization and code
7876 generation done at link time is executed in parallel using @var{n}
7877 parallel jobs by utilizing an installed @command{make} program. The
7878 environment variable @env{MAKE} may be used to override the program
7879 used. The default value for @var{n} is 1.
7881 You can also specify @option{-flto=jobserver} to use GNU make's
7882 job server mode to determine the number of parallel jobs. This
7883 is useful when the Makefile calling GCC is already executing in parallel.
7884 The parent Makefile will need a @samp{+} prepended to the command recipe
7885 for this to work. This will likely only work if @env{MAKE} is
7888 This option is disabled by default.
7890 @item -flto-partition=@var{alg}
7891 @opindex flto-partition
7892 Specify the partitioning algorithm used by the link time optimizer.
7893 The value is either @code{1to1} to specify a partitioning mirroring
7894 the original source files or @code{balanced} to specify partitioning
7895 into equally sized chunks (whenever possible). Specifying @code{none}
7896 as an algorithm disables partitioning and streaming completely. The
7897 default value is @code{balanced}.
7899 @item -flto-compression-level=@var{n}
7900 This option specifies the level of compression used for intermediate
7901 language written to LTO object files, and is only meaningful in
7902 conjunction with LTO mode (@option{-flto}). Valid
7903 values are 0 (no compression) to 9 (maximum compression). Values
7904 outside this range are clamped to either 0 or 9. If the option is not
7905 given, a default balanced compression setting is used.
7908 Prints a report with internal details on the workings of the link-time
7909 optimizer. The contents of this report vary from version to version,
7910 it is meant to be useful to GCC developers when processing object
7911 files in LTO mode (via @option{-flto}).
7913 Disabled by default.
7915 @item -fuse-linker-plugin
7916 Enables the use of a linker plugin during link time optimization. This
7917 option relies on plugin support in the linker, which is available in gold
7918 or in GNU ld 2.21 or newer.
7920 This option enables the extraction of object files with GIMPLE bytecode out
7921 of library archives. This improves the quality of optimization by exposing
7922 more code to the link time optimizer. This information specifies what
7923 symbols can be accessed externally (by non-LTO object or during dynamic
7924 linking). Resulting code quality improvements on binaries (and shared
7925 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7926 See @option{-flto} for a description of the effect of this flag and how to
7929 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7930 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7932 @item -ffat-lto-objects
7933 @opindex ffat-lto-objects
7934 Fat LTO objects are object files that contain both the intermediate language
7935 and the object code. This makes them useable for both LTO linking and normal
7936 linking. This option makes effect only with @option{-flto} and is ignored
7939 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
7940 requires the complete toolchain to be aware of LTO. It requires a linker with
7941 linker plugin support for basic functionality. Additionally, nm, ar and ranlib
7942 need to support linker plugins to allow a full-featured build environment
7943 (capable of building static libraries etc).
7945 The default is @option{-ffat-lto-objects} but this default is intended to
7946 change in future releases when linker plugin enabled environments become more
7948 @item -fcompare-elim
7949 @opindex fcompare-elim
7950 After register allocation and post-register allocation instruction splitting,
7951 identify arithmetic instructions that compute processor flags similar to a
7952 comparison operation based on that arithmetic. If possible, eliminate the
7953 explicit comparison operation.
7955 This pass only applies to certain targets that cannot explicitly represent
7956 the comparison operation before register allocation is complete.
7958 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7960 @item -fcprop-registers
7961 @opindex fcprop-registers
7962 After register allocation and post-register allocation instruction splitting,
7963 we perform a copy-propagation pass to try to reduce scheduling dependencies
7964 and occasionally eliminate the copy.
7966 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7968 @item -fprofile-correction
7969 @opindex fprofile-correction
7970 Profiles collected using an instrumented binary for multi-threaded programs may
7971 be inconsistent due to missed counter updates. When this option is specified,
7972 GCC will use heuristics to correct or smooth out such inconsistencies. By
7973 default, GCC will emit an error message when an inconsistent profile is detected.
7975 @item -fprofile-dir=@var{path}
7976 @opindex fprofile-dir
7978 Set the directory to search for the profile data files in to @var{path}.
7979 This option affects only the profile data generated by
7980 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7981 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7982 and its related options. Both absolute and relative paths can be used.
7983 By default, GCC will use the current directory as @var{path}, thus the
7984 profile data file will appear in the same directory as the object file.
7986 @item -fprofile-generate
7987 @itemx -fprofile-generate=@var{path}
7988 @opindex fprofile-generate
7990 Enable options usually used for instrumenting application to produce
7991 profile useful for later recompilation with profile feedback based
7992 optimization. You must use @option{-fprofile-generate} both when
7993 compiling and when linking your program.
7995 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7997 If @var{path} is specified, GCC will look at the @var{path} to find
7998 the profile feedback data files. See @option{-fprofile-dir}.
8001 @itemx -fprofile-use=@var{path}
8002 @opindex fprofile-use
8003 Enable profile feedback directed optimizations, and optimizations
8004 generally profitable only with profile feedback available.
8006 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
8007 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
8009 By default, GCC emits an error message if the feedback profiles do not
8010 match the source code. This error can be turned into a warning by using
8011 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8014 If @var{path} is specified, GCC will look at the @var{path} to find
8015 the profile feedback data files. See @option{-fprofile-dir}.
8018 The following options control compiler behavior regarding floating
8019 point arithmetic. These options trade off between speed and
8020 correctness. All must be specifically enabled.
8024 @opindex ffloat-store
8025 Do not store floating point variables in registers, and inhibit other
8026 options that might change whether a floating point value is taken from a
8029 @cindex floating point precision
8030 This option prevents undesirable excess precision on machines such as
8031 the 68000 where the floating registers (of the 68881) keep more
8032 precision than a @code{double} is supposed to have. Similarly for the
8033 x86 architecture. For most programs, the excess precision does only
8034 good, but a few programs rely on the precise definition of IEEE floating
8035 point. Use @option{-ffloat-store} for such programs, after modifying
8036 them to store all pertinent intermediate computations into variables.
8038 @item -fexcess-precision=@var{style}
8039 @opindex fexcess-precision
8040 This option allows further control over excess precision on machines
8041 where floating-point registers have more precision than the IEEE
8042 @code{float} and @code{double} types and the processor does not
8043 support operations rounding to those types. By default,
8044 @option{-fexcess-precision=fast} is in effect; this means that
8045 operations are carried out in the precision of the registers and that
8046 it is unpredictable when rounding to the types specified in the source
8047 code takes place. When compiling C, if
8048 @option{-fexcess-precision=standard} is specified then excess
8049 precision will follow the rules specified in ISO C99; in particular,
8050 both casts and assignments cause values to be rounded to their
8051 semantic types (whereas @option{-ffloat-store} only affects
8052 assignments). This option is enabled by default for C if a strict
8053 conformance option such as @option{-std=c99} is used.
8056 @option{-fexcess-precision=standard} is not implemented for languages
8057 other than C, and has no effect if
8058 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8059 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8060 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8061 semantics apply without excess precision, and in the latter, rounding
8066 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8067 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8068 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8070 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8072 This option is not turned on by any @option{-O} option besides
8073 @option{-Ofast} since it can result in incorrect output for programs
8074 which depend on an exact implementation of IEEE or ISO rules/specifications
8075 for math functions. It may, however, yield faster code for programs
8076 that do not require the guarantees of these specifications.
8078 @item -fno-math-errno
8079 @opindex fno-math-errno
8080 Do not set ERRNO after calling math functions that are executed
8081 with a single instruction, e.g., sqrt. A program that relies on
8082 IEEE exceptions for math error handling may want to use this flag
8083 for speed while maintaining IEEE arithmetic compatibility.
8085 This option is not turned on by any @option{-O} option since
8086 it can result in incorrect output for programs which depend on
8087 an exact implementation of IEEE or ISO rules/specifications for
8088 math functions. It may, however, yield faster code for programs
8089 that do not require the guarantees of these specifications.
8091 The default is @option{-fmath-errno}.
8093 On Darwin systems, the math library never sets @code{errno}. There is
8094 therefore no reason for the compiler to consider the possibility that
8095 it might, and @option{-fno-math-errno} is the default.
8097 @item -funsafe-math-optimizations
8098 @opindex funsafe-math-optimizations
8100 Allow optimizations for floating-point arithmetic that (a) assume
8101 that arguments and results are valid and (b) may violate IEEE or
8102 ANSI standards. When used at link-time, it may include libraries
8103 or startup files that change the default FPU control word or other
8104 similar optimizations.
8106 This option is not turned on by any @option{-O} option since
8107 it can result in incorrect output for programs which depend on
8108 an exact implementation of IEEE or ISO rules/specifications for
8109 math functions. It may, however, yield faster code for programs
8110 that do not require the guarantees of these specifications.
8111 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8112 @option{-fassociative-math} and @option{-freciprocal-math}.
8114 The default is @option{-fno-unsafe-math-optimizations}.
8116 @item -fassociative-math
8117 @opindex fassociative-math
8119 Allow re-association of operands in series of floating-point operations.
8120 This violates the ISO C and C++ language standard by possibly changing
8121 computation result. NOTE: re-ordering may change the sign of zero as
8122 well as ignore NaNs and inhibit or create underflow or overflow (and
8123 thus cannot be used on a code which relies on rounding behavior like
8124 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
8125 and thus may not be used when ordered comparisons are required.
8126 This option requires that both @option{-fno-signed-zeros} and
8127 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8128 much sense with @option{-frounding-math}. For Fortran the option
8129 is automatically enabled when both @option{-fno-signed-zeros} and
8130 @option{-fno-trapping-math} are in effect.
8132 The default is @option{-fno-associative-math}.
8134 @item -freciprocal-math
8135 @opindex freciprocal-math
8137 Allow the reciprocal of a value to be used instead of dividing by
8138 the value if this enables optimizations. For example @code{x / y}
8139 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8140 is subject to common subexpression elimination. Note that this loses
8141 precision and increases the number of flops operating on the value.
8143 The default is @option{-fno-reciprocal-math}.
8145 @item -ffinite-math-only
8146 @opindex ffinite-math-only
8147 Allow optimizations for floating-point arithmetic that assume
8148 that arguments and results are not NaNs or +-Infs.
8150 This option is not turned on by any @option{-O} option since
8151 it can result in incorrect output for programs which depend on
8152 an exact implementation of IEEE or ISO rules/specifications for
8153 math functions. It may, however, yield faster code for programs
8154 that do not require the guarantees of these specifications.
8156 The default is @option{-fno-finite-math-only}.
8158 @item -fno-signed-zeros
8159 @opindex fno-signed-zeros
8160 Allow optimizations for floating point arithmetic that ignore the
8161 signedness of zero. IEEE arithmetic specifies the behavior of
8162 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8163 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8164 This option implies that the sign of a zero result isn't significant.
8166 The default is @option{-fsigned-zeros}.
8168 @item -fno-trapping-math
8169 @opindex fno-trapping-math
8170 Compile code assuming that floating-point operations cannot generate
8171 user-visible traps. These traps include division by zero, overflow,
8172 underflow, inexact result and invalid operation. This option requires
8173 that @option{-fno-signaling-nans} be in effect. Setting this option may
8174 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8176 This option should never be turned on by any @option{-O} option since
8177 it can result in incorrect output for programs which depend on
8178 an exact implementation of IEEE or ISO rules/specifications for
8181 The default is @option{-ftrapping-math}.
8183 @item -frounding-math
8184 @opindex frounding-math
8185 Disable transformations and optimizations that assume default floating
8186 point rounding behavior. This is round-to-zero for all floating point
8187 to integer conversions, and round-to-nearest for all other arithmetic
8188 truncations. This option should be specified for programs that change
8189 the FP rounding mode dynamically, or that may be executed with a
8190 non-default rounding mode. This option disables constant folding of
8191 floating point expressions at compile-time (which may be affected by
8192 rounding mode) and arithmetic transformations that are unsafe in the
8193 presence of sign-dependent rounding modes.
8195 The default is @option{-fno-rounding-math}.
8197 This option is experimental and does not currently guarantee to
8198 disable all GCC optimizations that are affected by rounding mode.
8199 Future versions of GCC may provide finer control of this setting
8200 using C99's @code{FENV_ACCESS} pragma. This command line option
8201 will be used to specify the default state for @code{FENV_ACCESS}.
8203 @item -fsignaling-nans
8204 @opindex fsignaling-nans
8205 Compile code assuming that IEEE signaling NaNs may generate user-visible
8206 traps during floating-point operations. Setting this option disables
8207 optimizations that may change the number of exceptions visible with
8208 signaling NaNs. This option implies @option{-ftrapping-math}.
8210 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8213 The default is @option{-fno-signaling-nans}.
8215 This option is experimental and does not currently guarantee to
8216 disable all GCC optimizations that affect signaling NaN behavior.
8218 @item -fsingle-precision-constant
8219 @opindex fsingle-precision-constant
8220 Treat floating point constant as single precision constant instead of
8221 implicitly converting it to double precision constant.
8223 @item -fcx-limited-range
8224 @opindex fcx-limited-range
8225 When enabled, this option states that a range reduction step is not
8226 needed when performing complex division. Also, there is no checking
8227 whether the result of a complex multiplication or division is @code{NaN
8228 + I*NaN}, with an attempt to rescue the situation in that case. The
8229 default is @option{-fno-cx-limited-range}, but is enabled by
8230 @option{-ffast-math}.
8232 This option controls the default setting of the ISO C99
8233 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8236 @item -fcx-fortran-rules
8237 @opindex fcx-fortran-rules
8238 Complex multiplication and division follow Fortran rules. Range
8239 reduction is done as part of complex division, but there is no checking
8240 whether the result of a complex multiplication or division is @code{NaN
8241 + I*NaN}, with an attempt to rescue the situation in that case.
8243 The default is @option{-fno-cx-fortran-rules}.
8247 The following options control optimizations that may improve
8248 performance, but are not enabled by any @option{-O} options. This
8249 section includes experimental options that may produce broken code.
8252 @item -fbranch-probabilities
8253 @opindex fbranch-probabilities
8254 After running a program compiled with @option{-fprofile-arcs}
8255 (@pxref{Debugging Options,, Options for Debugging Your Program or
8256 @command{gcc}}), you can compile it a second time using
8257 @option{-fbranch-probabilities}, to improve optimizations based on
8258 the number of times each branch was taken. When the program
8259 compiled with @option{-fprofile-arcs} exits it saves arc execution
8260 counts to a file called @file{@var{sourcename}.gcda} for each source
8261 file. The information in this data file is very dependent on the
8262 structure of the generated code, so you must use the same source code
8263 and the same optimization options for both compilations.
8265 With @option{-fbranch-probabilities}, GCC puts a
8266 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8267 These can be used to improve optimization. Currently, they are only
8268 used in one place: in @file{reorg.c}, instead of guessing which path a
8269 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8270 exactly determine which path is taken more often.
8272 @item -fprofile-values
8273 @opindex fprofile-values
8274 If combined with @option{-fprofile-arcs}, it adds code so that some
8275 data about values of expressions in the program is gathered.
8277 With @option{-fbranch-probabilities}, it reads back the data gathered
8278 from profiling values of expressions for usage in optimizations.
8280 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8284 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8285 a code to gather information about values of expressions.
8287 With @option{-fbranch-probabilities}, it reads back the data gathered
8288 and actually performs the optimizations based on them.
8289 Currently the optimizations include specialization of division operation
8290 using the knowledge about the value of the denominator.
8292 @item -frename-registers
8293 @opindex frename-registers
8294 Attempt to avoid false dependencies in scheduled code by making use
8295 of registers left over after register allocation. This optimization
8296 will most benefit processors with lots of registers. Depending on the
8297 debug information format adopted by the target, however, it can
8298 make debugging impossible, since variables will no longer stay in
8299 a ``home register''.
8301 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8305 Perform tail duplication to enlarge superblock size. This transformation
8306 simplifies the control flow of the function allowing other optimizations to do
8309 Enabled with @option{-fprofile-use}.
8311 @item -funroll-loops
8312 @opindex funroll-loops
8313 Unroll loops whose number of iterations can be determined at compile time or
8314 upon entry to the loop. @option{-funroll-loops} implies
8315 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8316 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8317 small constant number of iterations). This option makes code larger, and may
8318 or may not make it run faster.
8320 Enabled with @option{-fprofile-use}.
8322 @item -funroll-all-loops
8323 @opindex funroll-all-loops
8324 Unroll all loops, even if their number of iterations is uncertain when
8325 the loop is entered. This usually makes programs run more slowly.
8326 @option{-funroll-all-loops} implies the same options as
8327 @option{-funroll-loops}.
8330 @opindex fpeel-loops
8331 Peels the loops for that there is enough information that they do not
8332 roll much (from profile feedback). It also turns on complete loop peeling
8333 (i.e.@: complete removal of loops with small constant number of iterations).
8335 Enabled with @option{-fprofile-use}.
8337 @item -fmove-loop-invariants
8338 @opindex fmove-loop-invariants
8339 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8340 at level @option{-O1}
8342 @item -funswitch-loops
8343 @opindex funswitch-loops
8344 Move branches with loop invariant conditions out of the loop, with duplicates
8345 of the loop on both branches (modified according to result of the condition).
8347 @item -ffunction-sections
8348 @itemx -fdata-sections
8349 @opindex ffunction-sections
8350 @opindex fdata-sections
8351 Place each function or data item into its own section in the output
8352 file if the target supports arbitrary sections. The name of the
8353 function or the name of the data item determines the section's name
8356 Use these options on systems where the linker can perform optimizations
8357 to improve locality of reference in the instruction space. Most systems
8358 using the ELF object format and SPARC processors running Solaris 2 have
8359 linkers with such optimizations. AIX may have these optimizations in
8362 Only use these options when there are significant benefits from doing
8363 so. When you specify these options, the assembler and linker will
8364 create larger object and executable files and will also be slower.
8365 You will not be able to use @code{gprof} on all systems if you
8366 specify this option and you may have problems with debugging if
8367 you specify both this option and @option{-g}.
8369 @item -fbranch-target-load-optimize
8370 @opindex fbranch-target-load-optimize
8371 Perform branch target register load optimization before prologue / epilogue
8373 The use of target registers can typically be exposed only during reload,
8374 thus hoisting loads out of loops and doing inter-block scheduling needs
8375 a separate optimization pass.
8377 @item -fbranch-target-load-optimize2
8378 @opindex fbranch-target-load-optimize2
8379 Perform branch target register load optimization after prologue / epilogue
8382 @item -fbtr-bb-exclusive
8383 @opindex fbtr-bb-exclusive
8384 When performing branch target register load optimization, don't reuse
8385 branch target registers in within any basic block.
8387 @item -fstack-protector
8388 @opindex fstack-protector
8389 Emit extra code to check for buffer overflows, such as stack smashing
8390 attacks. This is done by adding a guard variable to functions with
8391 vulnerable objects. This includes functions that call alloca, and
8392 functions with buffers larger than 8 bytes. The guards are initialized
8393 when a function is entered and then checked when the function exits.
8394 If a guard check fails, an error message is printed and the program exits.
8396 @item -fstack-protector-all
8397 @opindex fstack-protector-all
8398 Like @option{-fstack-protector} except that all functions are protected.
8400 @item -fsection-anchors
8401 @opindex fsection-anchors
8402 Try to reduce the number of symbolic address calculations by using
8403 shared ``anchor'' symbols to address nearby objects. This transformation
8404 can help to reduce the number of GOT entries and GOT accesses on some
8407 For example, the implementation of the following function @code{foo}:
8411 int foo (void) @{ return a + b + c; @}
8414 would usually calculate the addresses of all three variables, but if you
8415 compile it with @option{-fsection-anchors}, it will access the variables
8416 from a common anchor point instead. The effect is similar to the
8417 following pseudocode (which isn't valid C):
8422 register int *xr = &x;
8423 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8427 Not all targets support this option.
8429 @item --param @var{name}=@var{value}
8431 In some places, GCC uses various constants to control the amount of
8432 optimization that is done. For example, GCC will not inline functions
8433 that contain more that a certain number of instructions. You can
8434 control some of these constants on the command-line using the
8435 @option{--param} option.
8437 The names of specific parameters, and the meaning of the values, are
8438 tied to the internals of the compiler, and are subject to change
8439 without notice in future releases.
8441 In each case, the @var{value} is an integer. The allowable choices for
8442 @var{name} are given in the following table:
8445 @item predictable-branch-outcome
8446 When branch is predicted to be taken with probability lower than this threshold
8447 (in percent), then it is considered well predictable. The default is 10.
8449 @item max-crossjump-edges
8450 The maximum number of incoming edges to consider for crossjumping.
8451 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8452 the number of edges incoming to each block. Increasing values mean
8453 more aggressive optimization, making the compile time increase with
8454 probably small improvement in executable size.
8456 @item min-crossjump-insns
8457 The minimum number of instructions which must be matched at the end
8458 of two blocks before crossjumping will be performed on them. This
8459 value is ignored in the case where all instructions in the block being
8460 crossjumped from are matched. The default value is 5.
8462 @item max-grow-copy-bb-insns
8463 The maximum code size expansion factor when copying basic blocks
8464 instead of jumping. The expansion is relative to a jump instruction.
8465 The default value is 8.
8467 @item max-goto-duplication-insns
8468 The maximum number of instructions to duplicate to a block that jumps
8469 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8470 passes, GCC factors computed gotos early in the compilation process,
8471 and unfactors them as late as possible. Only computed jumps at the
8472 end of a basic blocks with no more than max-goto-duplication-insns are
8473 unfactored. The default value is 8.
8475 @item max-delay-slot-insn-search
8476 The maximum number of instructions to consider when looking for an
8477 instruction to fill a delay slot. If more than this arbitrary number of
8478 instructions is searched, the time savings from filling the delay slot
8479 will be minimal so stop searching. Increasing values mean more
8480 aggressive optimization, making the compile time increase with probably
8481 small improvement in executable run time.
8483 @item max-delay-slot-live-search
8484 When trying to fill delay slots, the maximum number of instructions to
8485 consider when searching for a block with valid live register
8486 information. Increasing this arbitrarily chosen value means more
8487 aggressive optimization, increasing the compile time. This parameter
8488 should be removed when the delay slot code is rewritten to maintain the
8491 @item max-gcse-memory
8492 The approximate maximum amount of memory that will be allocated in
8493 order to perform the global common subexpression elimination
8494 optimization. If more memory than specified is required, the
8495 optimization will not be done.
8497 @item max-gcse-insertion-ratio
8498 If the ratio of expression insertions to deletions is larger than this value
8499 for any expression, then RTL PRE will insert or remove the expression and thus
8500 leave partially redundant computations in the instruction stream. The default value is 20.
8502 @item max-pending-list-length
8503 The maximum number of pending dependencies scheduling will allow
8504 before flushing the current state and starting over. Large functions
8505 with few branches or calls can create excessively large lists which
8506 needlessly consume memory and resources.
8508 @item max-modulo-backtrack-attempts
8509 The maximum number of backtrack attempts the scheduler should make
8510 when modulo scheduling a loop. Larger values can exponentially increase
8513 @item max-inline-insns-single
8514 Several parameters control the tree inliner used in gcc.
8515 This number sets the maximum number of instructions (counted in GCC's
8516 internal representation) in a single function that the tree inliner
8517 will consider for inlining. This only affects functions declared
8518 inline and methods implemented in a class declaration (C++).
8519 The default value is 400.
8521 @item max-inline-insns-auto
8522 When you use @option{-finline-functions} (included in @option{-O3}),
8523 a lot of functions that would otherwise not be considered for inlining
8524 by the compiler will be investigated. To those functions, a different
8525 (more restrictive) limit compared to functions declared inline can
8527 The default value is 40.
8529 @item large-function-insns
8530 The limit specifying really large functions. For functions larger than this
8531 limit after inlining, inlining is constrained by
8532 @option{--param large-function-growth}. This parameter is useful primarily
8533 to avoid extreme compilation time caused by non-linear algorithms used by the
8535 The default value is 2700.
8537 @item large-function-growth
8538 Specifies maximal growth of large function caused by inlining in percents.
8539 The default value is 100 which limits large function growth to 2.0 times
8542 @item large-unit-insns
8543 The limit specifying large translation unit. Growth caused by inlining of
8544 units larger than this limit is limited by @option{--param inline-unit-growth}.
8545 For small units this might be too tight (consider unit consisting of function A
8546 that is inline and B that just calls A three time. If B is small relative to
8547 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8548 large units consisting of small inlineable functions however the overall unit
8549 growth limit is needed to avoid exponential explosion of code size. Thus for
8550 smaller units, the size is increased to @option{--param large-unit-insns}
8551 before applying @option{--param inline-unit-growth}. The default is 10000
8553 @item inline-unit-growth
8554 Specifies maximal overall growth of the compilation unit caused by inlining.
8555 The default value is 30 which limits unit growth to 1.3 times the original
8558 @item ipcp-unit-growth
8559 Specifies maximal overall growth of the compilation unit caused by
8560 interprocedural constant propagation. The default value is 10 which limits
8561 unit growth to 1.1 times the original size.
8563 @item large-stack-frame
8564 The limit specifying large stack frames. While inlining the algorithm is trying
8565 to not grow past this limit too much. Default value is 256 bytes.
8567 @item large-stack-frame-growth
8568 Specifies maximal growth of large stack frames caused by inlining in percents.
8569 The default value is 1000 which limits large stack frame growth to 11 times
8572 @item max-inline-insns-recursive
8573 @itemx max-inline-insns-recursive-auto
8574 Specifies maximum number of instructions out-of-line copy of self recursive inline
8575 function can grow into by performing recursive inlining.
8577 For functions declared inline @option{--param max-inline-insns-recursive} is
8578 taken into account. For function not declared inline, recursive inlining
8579 happens only when @option{-finline-functions} (included in @option{-O3}) is
8580 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8581 default value is 450.
8583 @item max-inline-recursive-depth
8584 @itemx max-inline-recursive-depth-auto
8585 Specifies maximum recursion depth used by the recursive inlining.
8587 For functions declared inline @option{--param max-inline-recursive-depth} is
8588 taken into account. For function not declared inline, recursive inlining
8589 happens only when @option{-finline-functions} (included in @option{-O3}) is
8590 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8593 @item min-inline-recursive-probability
8594 Recursive inlining is profitable only for function having deep recursion
8595 in average and can hurt for function having little recursion depth by
8596 increasing the prologue size or complexity of function body to other
8599 When profile feedback is available (see @option{-fprofile-generate}) the actual
8600 recursion depth can be guessed from probability that function will recurse via
8601 given call expression. This parameter limits inlining only to call expression
8602 whose probability exceeds given threshold (in percents). The default value is
8605 @item early-inlining-insns
8606 Specify growth that early inliner can make. In effect it increases amount of
8607 inlining for code having large abstraction penalty. The default value is 10.
8609 @item max-early-inliner-iterations
8610 @itemx max-early-inliner-iterations
8611 Limit of iterations of early inliner. This basically bounds number of nested
8612 indirect calls early inliner can resolve. Deeper chains are still handled by
8615 @item comdat-sharing-probability
8616 @itemx comdat-sharing-probability
8617 Probability (in percent) that C++ inline function with comdat visibility
8618 will be shared across multiple compilation units. The default value is 20.
8620 @item min-vect-loop-bound
8621 The minimum number of iterations under which a loop will not get vectorized
8622 when @option{-ftree-vectorize} is used. The number of iterations after
8623 vectorization needs to be greater than the value specified by this option
8624 to allow vectorization. The default value is 0.
8626 @item gcse-cost-distance-ratio
8627 Scaling factor in calculation of maximum distance an expression
8628 can be moved by GCSE optimizations. This is currently supported only in the
8629 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8630 will be with simple expressions, i.e., the expressions which have cost
8631 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8632 hoisting of simple expressions. The default value is 10.
8634 @item gcse-unrestricted-cost
8635 Cost, roughly measured as the cost of a single typical machine
8636 instruction, at which GCSE optimizations will not constrain
8637 the distance an expression can travel. This is currently
8638 supported only in the code hoisting pass. The lesser the cost,
8639 the more aggressive code hoisting will be. Specifying 0 will
8640 allow all expressions to travel unrestricted distances.
8641 The default value is 3.
8643 @item max-hoist-depth
8644 The depth of search in the dominator tree for expressions to hoist.
8645 This is used to avoid quadratic behavior in hoisting algorithm.
8646 The value of 0 will avoid limiting the search, but may slow down compilation
8647 of huge functions. The default value is 30.
8649 @item max-tail-merge-comparisons
8650 The maximum amount of similar bbs to compare a bb with. This is used to
8651 avoid quadratic behaviour in tree tail merging. The default value is 10.
8653 @item max-tail-merge-iterations
8654 The maximum amount of iterations of the pass over the function. This is used to
8655 limit run time in tree tail merging. The default value is 2.
8657 @item max-unrolled-insns
8658 The maximum number of instructions that a loop should have if that loop
8659 is unrolled, and if the loop is unrolled, it determines how many times
8660 the loop code is unrolled.
8662 @item max-average-unrolled-insns
8663 The maximum number of instructions biased by probabilities of their execution
8664 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8665 it determines how many times the loop code is unrolled.
8667 @item max-unroll-times
8668 The maximum number of unrollings of a single loop.
8670 @item max-peeled-insns
8671 The maximum number of instructions that a loop should have if that loop
8672 is peeled, and if the loop is peeled, it determines how many times
8673 the loop code is peeled.
8675 @item max-peel-times
8676 The maximum number of peelings of a single loop.
8678 @item max-completely-peeled-insns
8679 The maximum number of insns of a completely peeled loop.
8681 @item max-completely-peel-times
8682 The maximum number of iterations of a loop to be suitable for complete peeling.
8684 @item max-completely-peel-loop-nest-depth
8685 The maximum depth of a loop nest suitable for complete peeling.
8687 @item max-unswitch-insns
8688 The maximum number of insns of an unswitched loop.
8690 @item max-unswitch-level
8691 The maximum number of branches unswitched in a single loop.
8694 The minimum cost of an expensive expression in the loop invariant motion.
8696 @item iv-consider-all-candidates-bound
8697 Bound on number of candidates for induction variables below that
8698 all candidates are considered for each use in induction variable
8699 optimizations. Only the most relevant candidates are considered
8700 if there are more candidates, to avoid quadratic time complexity.
8702 @item iv-max-considered-uses
8703 The induction variable optimizations give up on loops that contain more
8704 induction variable uses.
8706 @item iv-always-prune-cand-set-bound
8707 If number of candidates in the set is smaller than this value,
8708 we always try to remove unnecessary ivs from the set during its
8709 optimization when a new iv is added to the set.
8711 @item scev-max-expr-size
8712 Bound on size of expressions used in the scalar evolutions analyzer.
8713 Large expressions slow the analyzer.
8715 @item scev-max-expr-complexity
8716 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8717 Complex expressions slow the analyzer.
8719 @item omega-max-vars
8720 The maximum number of variables in an Omega constraint system.
8721 The default value is 128.
8723 @item omega-max-geqs
8724 The maximum number of inequalities in an Omega constraint system.
8725 The default value is 256.
8728 The maximum number of equalities in an Omega constraint system.
8729 The default value is 128.
8731 @item omega-max-wild-cards
8732 The maximum number of wildcard variables that the Omega solver will
8733 be able to insert. The default value is 18.
8735 @item omega-hash-table-size
8736 The size of the hash table in the Omega solver. The default value is
8739 @item omega-max-keys
8740 The maximal number of keys used by the Omega solver. The default
8743 @item omega-eliminate-redundant-constraints
8744 When set to 1, use expensive methods to eliminate all redundant
8745 constraints. The default value is 0.
8747 @item vect-max-version-for-alignment-checks
8748 The maximum number of runtime checks that can be performed when
8749 doing loop versioning for alignment in the vectorizer. See option
8750 ftree-vect-loop-version for more information.
8752 @item vect-max-version-for-alias-checks
8753 The maximum number of runtime checks that can be performed when
8754 doing loop versioning for alias in the vectorizer. See option
8755 ftree-vect-loop-version for more information.
8757 @item max-iterations-to-track
8759 The maximum number of iterations of a loop the brute force algorithm
8760 for analysis of # of iterations of the loop tries to evaluate.
8762 @item hot-bb-count-fraction
8763 Select fraction of the maximal count of repetitions of basic block in program
8764 given basic block needs to have to be considered hot.
8766 @item hot-bb-frequency-fraction
8767 Select fraction of the entry block frequency of executions of basic block in
8768 function given basic block needs to have to be considered hot.
8770 @item max-predicted-iterations
8771 The maximum number of loop iterations we predict statically. This is useful
8772 in cases where function contain single loop with known bound and other loop
8773 with unknown. We predict the known number of iterations correctly, while
8774 the unknown number of iterations average to roughly 10. This means that the
8775 loop without bounds would appear artificially cold relative to the other one.
8777 @item align-threshold
8779 Select fraction of the maximal frequency of executions of basic block in
8780 function given basic block will get aligned.
8782 @item align-loop-iterations
8784 A loop expected to iterate at lest the selected number of iterations will get
8787 @item tracer-dynamic-coverage
8788 @itemx tracer-dynamic-coverage-feedback
8790 This value is used to limit superblock formation once the given percentage of
8791 executed instructions is covered. This limits unnecessary code size
8794 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8795 feedback is available. The real profiles (as opposed to statically estimated
8796 ones) are much less balanced allowing the threshold to be larger value.
8798 @item tracer-max-code-growth
8799 Stop tail duplication once code growth has reached given percentage. This is
8800 rather hokey argument, as most of the duplicates will be eliminated later in
8801 cross jumping, so it may be set to much higher values than is the desired code
8804 @item tracer-min-branch-ratio
8806 Stop reverse growth when the reverse probability of best edge is less than this
8807 threshold (in percent).
8809 @item tracer-min-branch-ratio
8810 @itemx tracer-min-branch-ratio-feedback
8812 Stop forward growth if the best edge do have probability lower than this
8815 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8816 compilation for profile feedback and one for compilation without. The value
8817 for compilation with profile feedback needs to be more conservative (higher) in
8818 order to make tracer effective.
8820 @item max-cse-path-length
8822 Maximum number of basic blocks on path that cse considers. The default is 10.
8825 The maximum instructions CSE process before flushing. The default is 1000.
8827 @item ggc-min-expand
8829 GCC uses a garbage collector to manage its own memory allocation. This
8830 parameter specifies the minimum percentage by which the garbage
8831 collector's heap should be allowed to expand between collections.
8832 Tuning this may improve compilation speed; it has no effect on code
8835 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8836 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8837 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8838 GCC is not able to calculate RAM on a particular platform, the lower
8839 bound of 30% is used. Setting this parameter and
8840 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8841 every opportunity. This is extremely slow, but can be useful for
8844 @item ggc-min-heapsize
8846 Minimum size of the garbage collector's heap before it begins bothering
8847 to collect garbage. The first collection occurs after the heap expands
8848 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8849 tuning this may improve compilation speed, and has no effect on code
8852 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8853 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8854 with a lower bound of 4096 (four megabytes) and an upper bound of
8855 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8856 particular platform, the lower bound is used. Setting this parameter
8857 very large effectively disables garbage collection. Setting this
8858 parameter and @option{ggc-min-expand} to zero causes a full collection
8859 to occur at every opportunity.
8861 @item max-reload-search-insns
8862 The maximum number of instruction reload should look backward for equivalent
8863 register. Increasing values mean more aggressive optimization, making the
8864 compile time increase with probably slightly better performance. The default
8867 @item max-cselib-memory-locations
8868 The maximum number of memory locations cselib should take into account.
8869 Increasing values mean more aggressive optimization, making the compile time
8870 increase with probably slightly better performance. The default value is 500.
8872 @item reorder-blocks-duplicate
8873 @itemx reorder-blocks-duplicate-feedback
8875 Used by basic block reordering pass to decide whether to use unconditional
8876 branch or duplicate the code on its destination. Code is duplicated when its
8877 estimated size is smaller than this value multiplied by the estimated size of
8878 unconditional jump in the hot spots of the program.
8880 The @option{reorder-block-duplicate-feedback} is used only when profile
8881 feedback is available and may be set to higher values than
8882 @option{reorder-block-duplicate} since information about the hot spots is more
8885 @item max-sched-ready-insns
8886 The maximum number of instructions ready to be issued the scheduler should
8887 consider at any given time during the first scheduling pass. Increasing
8888 values mean more thorough searches, making the compilation time increase
8889 with probably little benefit. The default value is 100.
8891 @item max-sched-region-blocks
8892 The maximum number of blocks in a region to be considered for
8893 interblock scheduling. The default value is 10.
8895 @item max-pipeline-region-blocks
8896 The maximum number of blocks in a region to be considered for
8897 pipelining in the selective scheduler. The default value is 15.
8899 @item max-sched-region-insns
8900 The maximum number of insns in a region to be considered for
8901 interblock scheduling. The default value is 100.
8903 @item max-pipeline-region-insns
8904 The maximum number of insns in a region to be considered for
8905 pipelining in the selective scheduler. The default value is 200.
8908 The minimum probability (in percents) of reaching a source block
8909 for interblock speculative scheduling. The default value is 40.
8911 @item max-sched-extend-regions-iters
8912 The maximum number of iterations through CFG to extend regions.
8913 0 - disable region extension,
8914 N - do at most N iterations.
8915 The default value is 0.
8917 @item max-sched-insn-conflict-delay
8918 The maximum conflict delay for an insn to be considered for speculative motion.
8919 The default value is 3.
8921 @item sched-spec-prob-cutoff
8922 The minimal probability of speculation success (in percents), so that
8923 speculative insn will be scheduled.
8924 The default value is 40.
8926 @item sched-mem-true-dep-cost
8927 Minimal distance (in CPU cycles) between store and load targeting same
8928 memory locations. The default value is 1.
8930 @item selsched-max-lookahead
8931 The maximum size of the lookahead window of selective scheduling. It is a
8932 depth of search for available instructions.
8933 The default value is 50.
8935 @item selsched-max-sched-times
8936 The maximum number of times that an instruction will be scheduled during
8937 selective scheduling. This is the limit on the number of iterations
8938 through which the instruction may be pipelined. The default value is 2.
8940 @item selsched-max-insns-to-rename
8941 The maximum number of best instructions in the ready list that are considered
8942 for renaming in the selective scheduler. The default value is 2.
8945 The minimum value of stage count that swing modulo scheduler will
8946 generate. The default value is 2.
8948 @item max-last-value-rtl
8949 The maximum size measured as number of RTLs that can be recorded in an expression
8950 in combiner for a pseudo register as last known value of that register. The default
8953 @item integer-share-limit
8954 Small integer constants can use a shared data structure, reducing the
8955 compiler's memory usage and increasing its speed. This sets the maximum
8956 value of a shared integer constant. The default value is 256.
8958 @item min-virtual-mappings
8959 Specifies the minimum number of virtual mappings in the incremental
8960 SSA updater that should be registered to trigger the virtual mappings
8961 heuristic defined by virtual-mappings-ratio. The default value is
8964 @item virtual-mappings-ratio
8965 If the number of virtual mappings is virtual-mappings-ratio bigger
8966 than the number of virtual symbols to be updated, then the incremental
8967 SSA updater switches to a full update for those symbols. The default
8970 @item ssp-buffer-size
8971 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8972 protection when @option{-fstack-protection} is used.
8974 @item max-jump-thread-duplication-stmts
8975 Maximum number of statements allowed in a block that needs to be
8976 duplicated when threading jumps.
8978 @item max-fields-for-field-sensitive
8979 Maximum number of fields in a structure we will treat in
8980 a field sensitive manner during pointer analysis. The default is zero
8981 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8983 @item prefetch-latency
8984 Estimate on average number of instructions that are executed before
8985 prefetch finishes. The distance we prefetch ahead is proportional
8986 to this constant. Increasing this number may also lead to less
8987 streams being prefetched (see @option{simultaneous-prefetches}).
8989 @item simultaneous-prefetches
8990 Maximum number of prefetches that can run at the same time.
8992 @item l1-cache-line-size
8993 The size of cache line in L1 cache, in bytes.
8996 The size of L1 cache, in kilobytes.
8999 The size of L2 cache, in kilobytes.
9001 @item min-insn-to-prefetch-ratio
9002 The minimum ratio between the number of instructions and the
9003 number of prefetches to enable prefetching in a loop.
9005 @item prefetch-min-insn-to-mem-ratio
9006 The minimum ratio between the number of instructions and the
9007 number of memory references to enable prefetching in a loop.
9009 @item use-canonical-types
9010 Whether the compiler should use the ``canonical'' type system. By
9011 default, this should always be 1, which uses a more efficient internal
9012 mechanism for comparing types in C++ and Objective-C++. However, if
9013 bugs in the canonical type system are causing compilation failures,
9014 set this value to 0 to disable canonical types.
9016 @item switch-conversion-max-branch-ratio
9017 Switch initialization conversion will refuse to create arrays that are
9018 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9019 branches in the switch.
9021 @item max-partial-antic-length
9022 Maximum length of the partial antic set computed during the tree
9023 partial redundancy elimination optimization (@option{-ftree-pre}) when
9024 optimizing at @option{-O3} and above. For some sorts of source code
9025 the enhanced partial redundancy elimination optimization can run away,
9026 consuming all of the memory available on the host machine. This
9027 parameter sets a limit on the length of the sets that are computed,
9028 which prevents the runaway behavior. Setting a value of 0 for
9029 this parameter will allow an unlimited set length.
9031 @item sccvn-max-scc-size
9032 Maximum size of a strongly connected component (SCC) during SCCVN
9033 processing. If this limit is hit, SCCVN processing for the whole
9034 function will not be done and optimizations depending on it will
9035 be disabled. The default maximum SCC size is 10000.
9037 @item ira-max-loops-num
9038 IRA uses a regional register allocation by default. If a function
9039 contains loops more than number given by the parameter, only at most
9040 given number of the most frequently executed loops will form regions
9041 for the regional register allocation. The default value of the
9044 @item ira-max-conflict-table-size
9045 Although IRA uses a sophisticated algorithm of compression conflict
9046 table, the table can be still big for huge functions. If the conflict
9047 table for a function could be more than size in MB given by the
9048 parameter, the conflict table is not built and faster, simpler, and
9049 lower quality register allocation algorithm will be used. The
9050 algorithm do not use pseudo-register conflicts. The default value of
9051 the parameter is 2000.
9053 @item ira-loop-reserved-regs
9054 IRA can be used to evaluate more accurate register pressure in loops
9055 for decision to move loop invariants (see @option{-O3}). The number
9056 of available registers reserved for some other purposes is described
9057 by this parameter. The default value of the parameter is 2 which is
9058 minimal number of registers needed for execution of typical
9059 instruction. This value is the best found from numerous experiments.
9061 @item loop-invariant-max-bbs-in-loop
9062 Loop invariant motion can be very expensive, both in compile time and
9063 in amount of needed compile time memory, with very large loops. Loops
9064 with more basic blocks than this parameter won't have loop invariant
9065 motion optimization performed on them. The default value of the
9066 parameter is 1000 for -O1 and 10000 for -O2 and above.
9068 @item max-vartrack-size
9069 Sets a maximum number of hash table slots to use during variable
9070 tracking dataflow analysis of any function. If this limit is exceeded
9071 with variable tracking at assignments enabled, analysis for that
9072 function is retried without it, after removing all debug insns from
9073 the function. If the limit is exceeded even without debug insns, var
9074 tracking analysis is completely disabled for the function. Setting
9075 the parameter to zero makes it unlimited.
9077 @item max-vartrack-expr-depth
9078 Sets a maximum number of recursion levels when attempting to map
9079 variable names or debug temporaries to value expressions. This trades
9080 compile time for more complete debug information. If this is set too
9081 low, value expressions that are available and could be represented in
9082 debug information may end up not being used; setting this higher may
9083 enable the compiler to find more complex debug expressions, but compile
9084 time and memory use may grow. The default is 12.
9086 @item min-nondebug-insn-uid
9087 Use uids starting at this parameter for nondebug insns. The range below
9088 the parameter is reserved exclusively for debug insns created by
9089 @option{-fvar-tracking-assignments}, but debug insns may get
9090 (non-overlapping) uids above it if the reserved range is exhausted.
9092 @item ipa-sra-ptr-growth-factor
9093 IPA-SRA will replace a pointer to an aggregate with one or more new
9094 parameters only when their cumulative size is less or equal to
9095 @option{ipa-sra-ptr-growth-factor} times the size of the original
9098 @item graphite-max-nb-scop-params
9099 To avoid exponential effects in the Graphite loop transforms, the
9100 number of parameters in a Static Control Part (SCoP) is bounded. The
9101 default value is 10 parameters. A variable whose value is unknown at
9102 compile time and defined outside a SCoP is a parameter of the SCoP.
9104 @item graphite-max-bbs-per-function
9105 To avoid exponential effects in the detection of SCoPs, the size of
9106 the functions analyzed by Graphite is bounded. The default value is
9109 @item loop-block-tile-size
9110 Loop blocking or strip mining transforms, enabled with
9111 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9112 loop in the loop nest by a given number of iterations. The strip
9113 length can be changed using the @option{loop-block-tile-size}
9114 parameter. The default value is 51 iterations.
9116 @item ipa-cp-value-list-size
9117 IPA-CP attempts to track all possible values and types passed to a function's
9118 parameter in order to propagate them and perform devirtualization.
9119 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9120 stores per one formal parameter of a function.
9122 @item lto-partitions
9123 Specify desired number of partitions produced during WHOPR compilation.
9124 The number of partitions should exceed the number of CPUs used for compilation.
9125 The default value is 32.
9127 @item lto-minpartition
9128 Size of minimal partition for WHOPR (in estimated instructions).
9129 This prevents expenses of splitting very small programs into too many
9132 @item cxx-max-namespaces-for-diagnostic-help
9133 The maximum number of namespaces to consult for suggestions when C++
9134 name lookup fails for an identifier. The default is 1000.
9136 @item sink-frequency-threshold
9137 The maximum relative execution frequency (in percents) of the target block
9138 relative to a statement's original block to allow statement sinking of a
9139 statement. Larger numbers result in more aggressive statement sinking.
9140 The default value is 75. A small positive adjustment is applied for
9141 statements with memory operands as those are even more profitable so sink.
9143 @item max-stores-to-sink
9144 The maximum number of conditional stores paires that can be sunk. Set to 0
9145 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9146 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9148 @item allow-store-data-races
9149 Allow optimizers to introduce new data races on stores.
9150 Set to 1 to allow, otherwise to 0. This option is enabled by default
9151 unless implicitly set by the @option{-fmemory-model=} option.
9153 @item case-values-threshold
9154 The smallest number of different values for which it is best to use a
9155 jump-table instead of a tree of conditional branches. If the value is
9156 0, use the default for the machine. The default is 0.
9158 @item tree-reassoc-width
9159 Set the maximum number of instructions executed in parallel in
9160 reassociated tree. This parameter overrides target dependent
9161 heuristics used by default if has non zero value.
9166 @node Preprocessor Options
9167 @section Options Controlling the Preprocessor
9168 @cindex preprocessor options
9169 @cindex options, preprocessor
9171 These options control the C preprocessor, which is run on each C source
9172 file before actual compilation.
9174 If you use the @option{-E} option, nothing is done except preprocessing.
9175 Some of these options make sense only together with @option{-E} because
9176 they cause the preprocessor output to be unsuitable for actual
9180 @item -Wp,@var{option}
9182 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9183 and pass @var{option} directly through to the preprocessor. If
9184 @var{option} contains commas, it is split into multiple options at the
9185 commas. However, many options are modified, translated or interpreted
9186 by the compiler driver before being passed to the preprocessor, and
9187 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9188 interface is undocumented and subject to change, so whenever possible
9189 you should avoid using @option{-Wp} and let the driver handle the
9192 @item -Xpreprocessor @var{option}
9193 @opindex Xpreprocessor
9194 Pass @var{option} as an option to the preprocessor. You can use this to
9195 supply system-specific preprocessor options which GCC does not know how to
9198 If you want to pass an option that takes an argument, you must use
9199 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9202 @include cppopts.texi
9204 @node Assembler Options
9205 @section Passing Options to the Assembler
9207 @c prevent bad page break with this line
9208 You can pass options to the assembler.
9211 @item -Wa,@var{option}
9213 Pass @var{option} as an option to the assembler. If @var{option}
9214 contains commas, it is split into multiple options at the commas.
9216 @item -Xassembler @var{option}
9218 Pass @var{option} as an option to the assembler. You can use this to
9219 supply system-specific assembler options which GCC does not know how to
9222 If you want to pass an option that takes an argument, you must use
9223 @option{-Xassembler} twice, once for the option and once for the argument.
9228 @section Options for Linking
9229 @cindex link options
9230 @cindex options, linking
9232 These options come into play when the compiler links object files into
9233 an executable output file. They are meaningless if the compiler is
9234 not doing a link step.
9238 @item @var{object-file-name}
9239 A file name that does not end in a special recognized suffix is
9240 considered to name an object file or library. (Object files are
9241 distinguished from libraries by the linker according to the file
9242 contents.) If linking is done, these object files are used as input
9251 If any of these options is used, then the linker is not run, and
9252 object file names should not be used as arguments. @xref{Overall
9256 @item -l@var{library}
9257 @itemx -l @var{library}
9259 Search the library named @var{library} when linking. (The second
9260 alternative with the library as a separate argument is only for
9261 POSIX compliance and is not recommended.)
9263 It makes a difference where in the command you write this option; the
9264 linker searches and processes libraries and object files in the order they
9265 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9266 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9267 to functions in @samp{z}, those functions may not be loaded.
9269 The linker searches a standard list of directories for the library,
9270 which is actually a file named @file{lib@var{library}.a}. The linker
9271 then uses this file as if it had been specified precisely by name.
9273 The directories searched include several standard system directories
9274 plus any that you specify with @option{-L}.
9276 Normally the files found this way are library files---archive files
9277 whose members are object files. The linker handles an archive file by
9278 scanning through it for members which define symbols that have so far
9279 been referenced but not defined. But if the file that is found is an
9280 ordinary object file, it is linked in the usual fashion. The only
9281 difference between using an @option{-l} option and specifying a file name
9282 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9283 and searches several directories.
9287 You need this special case of the @option{-l} option in order to
9288 link an Objective-C or Objective-C++ program.
9291 @opindex nostartfiles
9292 Do not use the standard system startup files when linking.
9293 The standard system libraries are used normally, unless @option{-nostdlib}
9294 or @option{-nodefaultlibs} is used.
9296 @item -nodefaultlibs
9297 @opindex nodefaultlibs
9298 Do not use the standard system libraries when linking.
9299 Only the libraries you specify will be passed to the linker, options
9300 specifying linkage of the system libraries, such as @code{-static-libgcc}
9301 or @code{-shared-libgcc}, will be ignored.
9302 The standard startup files are used normally, unless @option{-nostartfiles}
9303 is used. The compiler may generate calls to @code{memcmp},
9304 @code{memset}, @code{memcpy} and @code{memmove}.
9305 These entries are usually resolved by entries in
9306 libc. These entry points should be supplied through some other
9307 mechanism when this option is specified.
9311 Do not use the standard system startup files or libraries when linking.
9312 No startup files and only the libraries you specify will be passed to
9313 the linker, options specifying linkage of the system libraries, such as
9314 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9315 The compiler may generate calls to @code{memcmp}, @code{memset},
9316 @code{memcpy} and @code{memmove}.
9317 These entries are usually resolved by entries in
9318 libc. These entry points should be supplied through some other
9319 mechanism when this option is specified.
9321 @cindex @option{-lgcc}, use with @option{-nostdlib}
9322 @cindex @option{-nostdlib} and unresolved references
9323 @cindex unresolved references and @option{-nostdlib}
9324 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9325 @cindex @option{-nodefaultlibs} and unresolved references
9326 @cindex unresolved references and @option{-nodefaultlibs}
9327 One of the standard libraries bypassed by @option{-nostdlib} and
9328 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9329 that GCC uses to overcome shortcomings of particular machines, or special
9330 needs for some languages.
9331 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9332 Collection (GCC) Internals},
9333 for more discussion of @file{libgcc.a}.)
9334 In most cases, you need @file{libgcc.a} even when you want to avoid
9335 other standard libraries. In other words, when you specify @option{-nostdlib}
9336 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9337 This ensures that you have no unresolved references to internal GCC
9338 library subroutines. (For example, @samp{__main}, used to ensure C++
9339 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9340 GNU Compiler Collection (GCC) Internals}.)
9344 Produce a position independent executable on targets which support it.
9345 For predictable results, you must also specify the same set of options
9346 that were used to generate code (@option{-fpie}, @option{-fPIE},
9347 or model suboptions) when you specify this option.
9351 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9352 that support it. This instructs the linker to add all symbols, not
9353 only used ones, to the dynamic symbol table. This option is needed
9354 for some uses of @code{dlopen} or to allow obtaining backtraces
9355 from within a program.
9359 Remove all symbol table and relocation information from the executable.
9363 On systems that support dynamic linking, this prevents linking with the shared
9364 libraries. On other systems, this option has no effect.
9368 Produce a shared object which can then be linked with other objects to
9369 form an executable. Not all systems support this option. For predictable
9370 results, you must also specify the same set of options that were used to
9371 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9372 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9373 needs to build supplementary stub code for constructors to work. On
9374 multi-libbed systems, @samp{gcc -shared} must select the correct support
9375 libraries to link against. Failing to supply the correct flags may lead
9376 to subtle defects. Supplying them in cases where they are not necessary
9379 @item -shared-libgcc
9380 @itemx -static-libgcc
9381 @opindex shared-libgcc
9382 @opindex static-libgcc
9383 On systems that provide @file{libgcc} as a shared library, these options
9384 force the use of either the shared or static version respectively.
9385 If no shared version of @file{libgcc} was built when the compiler was
9386 configured, these options have no effect.
9388 There are several situations in which an application should use the
9389 shared @file{libgcc} instead of the static version. The most common
9390 of these is when the application wishes to throw and catch exceptions
9391 across different shared libraries. In that case, each of the libraries
9392 as well as the application itself should use the shared @file{libgcc}.
9394 Therefore, the G++ and GCJ drivers automatically add
9395 @option{-shared-libgcc} whenever you build a shared library or a main
9396 executable, because C++ and Java programs typically use exceptions, so
9397 this is the right thing to do.
9399 If, instead, you use the GCC driver to create shared libraries, you may
9400 find that they will not always be linked with the shared @file{libgcc}.
9401 If GCC finds, at its configuration time, that you have a non-GNU linker
9402 or a GNU linker that does not support option @option{--eh-frame-hdr},
9403 it will link the shared version of @file{libgcc} into shared libraries
9404 by default. Otherwise, it will take advantage of the linker and optimize
9405 away the linking with the shared version of @file{libgcc}, linking with
9406 the static version of libgcc by default. This allows exceptions to
9407 propagate through such shared libraries, without incurring relocation
9408 costs at library load time.
9410 However, if a library or main executable is supposed to throw or catch
9411 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9412 for the languages used in the program, or using the option
9413 @option{-shared-libgcc}, such that it is linked with the shared
9416 @item -static-libstdc++
9417 When the @command{g++} program is used to link a C++ program, it will
9418 normally automatically link against @option{libstdc++}. If
9419 @file{libstdc++} is available as a shared library, and the
9420 @option{-static} option is not used, then this will link against the
9421 shared version of @file{libstdc++}. That is normally fine. However, it
9422 is sometimes useful to freeze the version of @file{libstdc++} used by
9423 the program without going all the way to a fully static link. The
9424 @option{-static-libstdc++} option directs the @command{g++} driver to
9425 link @file{libstdc++} statically, without necessarily linking other
9426 libraries statically.
9430 Bind references to global symbols when building a shared object. Warn
9431 about any unresolved references (unless overridden by the link editor
9432 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9435 @item -T @var{script}
9437 @cindex linker script
9438 Use @var{script} as the linker script. This option is supported by most
9439 systems using the GNU linker. On some targets, such as bare-board
9440 targets without an operating system, the @option{-T} option may be required
9441 when linking to avoid references to undefined symbols.
9443 @item -Xlinker @var{option}
9445 Pass @var{option} as an option to the linker. You can use this to
9446 supply system-specific linker options which GCC does not know how to
9449 If you want to pass an option that takes a separate argument, you must use
9450 @option{-Xlinker} twice, once for the option and once for the argument.
9451 For example, to pass @option{-assert definitions}, you must write
9452 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9453 @option{-Xlinker "-assert definitions"}, because this passes the entire
9454 string as a single argument, which is not what the linker expects.
9456 When using the GNU linker, it is usually more convenient to pass
9457 arguments to linker options using the @option{@var{option}=@var{value}}
9458 syntax than as separate arguments. For example, you can specify
9459 @samp{-Xlinker -Map=output.map} rather than
9460 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9461 this syntax for command-line options.
9463 @item -Wl,@var{option}
9465 Pass @var{option} as an option to the linker. If @var{option} contains
9466 commas, it is split into multiple options at the commas. You can use this
9467 syntax to pass an argument to the option.
9468 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9469 linker. When using the GNU linker, you can also get the same effect with
9470 @samp{-Wl,-Map=output.map}.
9472 @item -u @var{symbol}
9474 Pretend the symbol @var{symbol} is undefined, to force linking of
9475 library modules to define it. You can use @option{-u} multiple times with
9476 different symbols to force loading of additional library modules.
9479 @node Directory Options
9480 @section Options for Directory Search
9481 @cindex directory options
9482 @cindex options, directory search
9485 These options specify directories to search for header files, for
9486 libraries and for parts of the compiler:
9491 Add the directory @var{dir} to the head of the list of directories to be
9492 searched for header files. This can be used to override a system header
9493 file, substituting your own version, since these directories are
9494 searched before the system header file directories. However, you should
9495 not use this option to add directories that contain vendor-supplied
9496 system header files (use @option{-isystem} for that). If you use more than
9497 one @option{-I} option, the directories are scanned in left-to-right
9498 order; the standard system directories come after.
9500 If a standard system include directory, or a directory specified with
9501 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9502 option will be ignored. The directory will still be searched but as a
9503 system directory at its normal position in the system include chain.
9504 This is to ensure that GCC's procedure to fix buggy system headers and
9505 the ordering for the include_next directive are not inadvertently changed.
9506 If you really need to change the search order for system directories,
9507 use the @option{-nostdinc} and/or @option{-isystem} options.
9509 @item -iplugindir=@var{dir}
9510 Set the directory to search for plugins which are passed
9511 by @option{-fplugin=@var{name}} instead of
9512 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9513 to be used by the user, but only passed by the driver.
9515 @item -iquote@var{dir}
9517 Add the directory @var{dir} to the head of the list of directories to
9518 be searched for header files only for the case of @samp{#include
9519 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9520 otherwise just like @option{-I}.
9524 Add directory @var{dir} to the list of directories to be searched
9527 @item -B@var{prefix}
9529 This option specifies where to find the executables, libraries,
9530 include files, and data files of the compiler itself.
9532 The compiler driver program runs one or more of the subprograms
9533 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9534 @var{prefix} as a prefix for each program it tries to run, both with and
9535 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9537 For each subprogram to be run, the compiler driver first tries the
9538 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9539 was not specified, the driver tries two standard prefixes, which are
9540 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9541 those results in a file name that is found, the unmodified program
9542 name is searched for using the directories specified in your
9543 @env{PATH} environment variable.
9545 The compiler will check to see if the path provided by the @option{-B}
9546 refers to a directory, and if necessary it will add a directory
9547 separator character at the end of the path.
9549 @option{-B} prefixes that effectively specify directory names also apply
9550 to libraries in the linker, because the compiler translates these
9551 options into @option{-L} options for the linker. They also apply to
9552 includes files in the preprocessor, because the compiler translates these
9553 options into @option{-isystem} options for the preprocessor. In this case,
9554 the compiler appends @samp{include} to the prefix.
9556 The run-time support file @file{libgcc.a} can also be searched for using
9557 the @option{-B} prefix, if needed. If it is not found there, the two
9558 standard prefixes above are tried, and that is all. The file is left
9559 out of the link if it is not found by those means.
9561 Another way to specify a prefix much like the @option{-B} prefix is to use
9562 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9565 As a special kludge, if the path provided by @option{-B} is
9566 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9567 9, then it will be replaced by @file{[dir/]include}. This is to help
9568 with boot-strapping the compiler.
9570 @item -specs=@var{file}
9572 Process @var{file} after the compiler reads in the standard @file{specs}
9573 file, in order to override the defaults that the @file{gcc} driver
9574 program uses when determining what switches to pass to @file{cc1},
9575 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9576 @option{-specs=@var{file}} can be specified on the command line, and they
9577 are processed in order, from left to right.
9579 @item --sysroot=@var{dir}
9581 Use @var{dir} as the logical root directory for headers and libraries.
9582 For example, if the compiler would normally search for headers in
9583 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9584 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9586 If you use both this option and the @option{-isysroot} option, then
9587 the @option{--sysroot} option will apply to libraries, but the
9588 @option{-isysroot} option will apply to header files.
9590 The GNU linker (beginning with version 2.16) has the necessary support
9591 for this option. If your linker does not support this option, the
9592 header file aspect of @option{--sysroot} will still work, but the
9593 library aspect will not.
9597 This option has been deprecated. Please use @option{-iquote} instead for
9598 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9599 Any directories you specify with @option{-I} options before the @option{-I-}
9600 option are searched only for the case of @samp{#include "@var{file}"};
9601 they are not searched for @samp{#include <@var{file}>}.
9603 If additional directories are specified with @option{-I} options after
9604 the @option{-I-}, these directories are searched for all @samp{#include}
9605 directives. (Ordinarily @emph{all} @option{-I} directories are used
9608 In addition, the @option{-I-} option inhibits the use of the current
9609 directory (where the current input file came from) as the first search
9610 directory for @samp{#include "@var{file}"}. There is no way to
9611 override this effect of @option{-I-}. With @option{-I.} you can specify
9612 searching the directory which was current when the compiler was
9613 invoked. That is not exactly the same as what the preprocessor does
9614 by default, but it is often satisfactory.
9616 @option{-I-} does not inhibit the use of the standard system directories
9617 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9624 @section Specifying subprocesses and the switches to pass to them
9627 @command{gcc} is a driver program. It performs its job by invoking a
9628 sequence of other programs to do the work of compiling, assembling and
9629 linking. GCC interprets its command-line parameters and uses these to
9630 deduce which programs it should invoke, and which command-line options
9631 it ought to place on their command lines. This behavior is controlled
9632 by @dfn{spec strings}. In most cases there is one spec string for each
9633 program that GCC can invoke, but a few programs have multiple spec
9634 strings to control their behavior. The spec strings built into GCC can
9635 be overridden by using the @option{-specs=} command-line switch to specify
9638 @dfn{Spec files} are plaintext files that are used to construct spec
9639 strings. They consist of a sequence of directives separated by blank
9640 lines. The type of directive is determined by the first non-whitespace
9641 character on the line and it can be one of the following:
9644 @item %@var{command}
9645 Issues a @var{command} to the spec file processor. The commands that can
9649 @item %include <@var{file}>
9650 @cindex @code{%include}
9651 Search for @var{file} and insert its text at the current point in the
9654 @item %include_noerr <@var{file}>
9655 @cindex @code{%include_noerr}
9656 Just like @samp{%include}, but do not generate an error message if the include
9657 file cannot be found.
9659 @item %rename @var{old_name} @var{new_name}
9660 @cindex @code{%rename}
9661 Rename the spec string @var{old_name} to @var{new_name}.
9665 @item *[@var{spec_name}]:
9666 This tells the compiler to create, override or delete the named spec
9667 string. All lines after this directive up to the next directive or
9668 blank line are considered to be the text for the spec string. If this
9669 results in an empty string then the spec will be deleted. (Or, if the
9670 spec did not exist, then nothing will happen.) Otherwise, if the spec
9671 does not currently exist a new spec will be created. If the spec does
9672 exist then its contents will be overridden by the text of this
9673 directive, unless the first character of that text is the @samp{+}
9674 character, in which case the text will be appended to the spec.
9676 @item [@var{suffix}]:
9677 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9678 and up to the next directive or blank line are considered to make up the
9679 spec string for the indicated suffix. When the compiler encounters an
9680 input file with the named suffix, it will processes the spec string in
9681 order to work out how to compile that file. For example:
9688 This says that any input file whose name ends in @samp{.ZZ} should be
9689 passed to the program @samp{z-compile}, which should be invoked with the
9690 command-line switch @option{-input} and with the result of performing the
9691 @samp{%i} substitution. (See below.)
9693 As an alternative to providing a spec string, the text that follows a
9694 suffix directive can be one of the following:
9697 @item @@@var{language}
9698 This says that the suffix is an alias for a known @var{language}. This is
9699 similar to using the @option{-x} command-line switch to GCC to specify a
9700 language explicitly. For example:
9707 Says that .ZZ files are, in fact, C++ source files.
9710 This causes an error messages saying:
9713 @var{name} compiler not installed on this system.
9717 GCC already has an extensive list of suffixes built into it.
9718 This directive will add an entry to the end of the list of suffixes, but
9719 since the list is searched from the end backwards, it is effectively
9720 possible to override earlier entries using this technique.
9724 GCC has the following spec strings built into it. Spec files can
9725 override these strings or create their own. Note that individual
9726 targets can also add their own spec strings to this list.
9729 asm Options to pass to the assembler
9730 asm_final Options to pass to the assembler post-processor
9731 cpp Options to pass to the C preprocessor
9732 cc1 Options to pass to the C compiler
9733 cc1plus Options to pass to the C++ compiler
9734 endfile Object files to include at the end of the link
9735 link Options to pass to the linker
9736 lib Libraries to include on the command line to the linker
9737 libgcc Decides which GCC support library to pass to the linker
9738 linker Sets the name of the linker
9739 predefines Defines to be passed to the C preprocessor
9740 signed_char Defines to pass to CPP to say whether @code{char} is signed
9742 startfile Object files to include at the start of the link
9745 Here is a small example of a spec file:
9751 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9754 This example renames the spec called @samp{lib} to @samp{old_lib} and
9755 then overrides the previous definition of @samp{lib} with a new one.
9756 The new definition adds in some extra command-line options before
9757 including the text of the old definition.
9759 @dfn{Spec strings} are a list of command-line options to be passed to their
9760 corresponding program. In addition, the spec strings can contain
9761 @samp{%}-prefixed sequences to substitute variable text or to
9762 conditionally insert text into the command line. Using these constructs
9763 it is possible to generate quite complex command lines.
9765 Here is a table of all defined @samp{%}-sequences for spec
9766 strings. Note that spaces are not generated automatically around the
9767 results of expanding these sequences. Therefore you can concatenate them
9768 together or combine them with constant text in a single argument.
9772 Substitute one @samp{%} into the program name or argument.
9775 Substitute the name of the input file being processed.
9778 Substitute the basename of the input file being processed.
9779 This is the substring up to (and not including) the last period
9780 and not including the directory.
9783 This is the same as @samp{%b}, but include the file suffix (text after
9787 Marks the argument containing or following the @samp{%d} as a
9788 temporary file name, so that that file will be deleted if GCC exits
9789 successfully. Unlike @samp{%g}, this contributes no text to the
9792 @item %g@var{suffix}
9793 Substitute a file name that has suffix @var{suffix} and is chosen
9794 once per compilation, and mark the argument in the same way as
9795 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9796 name is now chosen in a way that is hard to predict even when previously
9797 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9798 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9799 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9800 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9801 was simply substituted with a file name chosen once per compilation,
9802 without regard to any appended suffix (which was therefore treated
9803 just like ordinary text), making such attacks more likely to succeed.
9805 @item %u@var{suffix}
9806 Like @samp{%g}, but generates a new temporary file name even if
9807 @samp{%u@var{suffix}} was already seen.
9809 @item %U@var{suffix}
9810 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9811 new one if there is no such last file name. In the absence of any
9812 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9813 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9814 would involve the generation of two distinct file names, one
9815 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9816 simply substituted with a file name chosen for the previous @samp{%u},
9817 without regard to any appended suffix.
9819 @item %j@var{suffix}
9820 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9821 writable, and if save-temps is off; otherwise, substitute the name
9822 of a temporary file, just like @samp{%u}. This temporary file is not
9823 meant for communication between processes, but rather as a junk
9826 @item %|@var{suffix}
9827 @itemx %m@var{suffix}
9828 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9829 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9830 all. These are the two most common ways to instruct a program that it
9831 should read from standard input or write to standard output. If you
9832 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9833 construct: see for example @file{f/lang-specs.h}.
9835 @item %.@var{SUFFIX}
9836 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9837 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9838 terminated by the next space or %.
9841 Marks the argument containing or following the @samp{%w} as the
9842 designated output file of this compilation. This puts the argument
9843 into the sequence of arguments that @samp{%o} will substitute later.
9846 Substitutes the names of all the output files, with spaces
9847 automatically placed around them. You should write spaces
9848 around the @samp{%o} as well or the results are undefined.
9849 @samp{%o} is for use in the specs for running the linker.
9850 Input files whose names have no recognized suffix are not compiled
9851 at all, but they are included among the output files, so they will
9855 Substitutes the suffix for object files. Note that this is
9856 handled specially when it immediately follows @samp{%g, %u, or %U},
9857 because of the need for those to form complete file names. The
9858 handling is such that @samp{%O} is treated exactly as if it had already
9859 been substituted, except that @samp{%g, %u, and %U} do not currently
9860 support additional @var{suffix} characters following @samp{%O} as they would
9861 following, for example, @samp{.o}.
9864 Substitutes the standard macro predefinitions for the
9865 current target machine. Use this when running @code{cpp}.
9868 Like @samp{%p}, but puts @samp{__} before and after the name of each
9869 predefined macro, except for macros that start with @samp{__} or with
9870 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9874 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9875 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9876 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9877 and @option{-imultilib} as necessary.
9880 Current argument is the name of a library or startup file of some sort.
9881 Search for that file in a standard list of directories and substitute
9882 the full name found. The current working directory is included in the
9883 list of directories scanned.
9886 Current argument is the name of a linker script. Search for that file
9887 in the current list of directories to scan for libraries. If the file
9888 is located insert a @option{--script} option into the command line
9889 followed by the full path name found. If the file is not found then
9890 generate an error message. Note: the current working directory is not
9894 Print @var{str} as an error message. @var{str} is terminated by a newline.
9895 Use this when inconsistent options are detected.
9898 Substitute the contents of spec string @var{name} at this point.
9900 @item %x@{@var{option}@}
9901 Accumulate an option for @samp{%X}.
9904 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9908 Output the accumulated assembler options specified by @option{-Wa}.
9911 Output the accumulated preprocessor options specified by @option{-Wp}.
9914 Process the @code{asm} spec. This is used to compute the
9915 switches to be passed to the assembler.
9918 Process the @code{asm_final} spec. This is a spec string for
9919 passing switches to an assembler post-processor, if such a program is
9923 Process the @code{link} spec. This is the spec for computing the
9924 command line passed to the linker. Typically it will make use of the
9925 @samp{%L %G %S %D and %E} sequences.
9928 Dump out a @option{-L} option for each directory that GCC believes might
9929 contain startup files. If the target supports multilibs then the
9930 current multilib directory will be prepended to each of these paths.
9933 Process the @code{lib} spec. This is a spec string for deciding which
9934 libraries should be included on the command line to the linker.
9937 Process the @code{libgcc} spec. This is a spec string for deciding
9938 which GCC support library should be included on the command line to the linker.
9941 Process the @code{startfile} spec. This is a spec for deciding which
9942 object files should be the first ones passed to the linker. Typically
9943 this might be a file named @file{crt0.o}.
9946 Process the @code{endfile} spec. This is a spec string that specifies
9947 the last object files that will be passed to the linker.
9950 Process the @code{cpp} spec. This is used to construct the arguments
9951 to be passed to the C preprocessor.
9954 Process the @code{cc1} spec. This is used to construct the options to be
9955 passed to the actual C compiler (@samp{cc1}).
9958 Process the @code{cc1plus} spec. This is used to construct the options to be
9959 passed to the actual C++ compiler (@samp{cc1plus}).
9962 Substitute the variable part of a matched option. See below.
9963 Note that each comma in the substituted string is replaced by
9967 Remove all occurrences of @code{-S} from the command line. Note---this
9968 command is position dependent. @samp{%} commands in the spec string
9969 before this one will see @code{-S}, @samp{%} commands in the spec string
9970 after this one will not.
9972 @item %:@var{function}(@var{args})
9973 Call the named function @var{function}, passing it @var{args}.
9974 @var{args} is first processed as a nested spec string, then split
9975 into an argument vector in the usual fashion. The function returns
9976 a string which is processed as if it had appeared literally as part
9977 of the current spec.
9979 The following built-in spec functions are provided:
9983 The @code{getenv} spec function takes two arguments: an environment
9984 variable name and a string. If the environment variable is not
9985 defined, a fatal error is issued. Otherwise, the return value is the
9986 value of the environment variable concatenated with the string. For
9987 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9990 %:getenv(TOPDIR /include)
9993 expands to @file{/path/to/top/include}.
9995 @item @code{if-exists}
9996 The @code{if-exists} spec function takes one argument, an absolute
9997 pathname to a file. If the file exists, @code{if-exists} returns the
9998 pathname. Here is a small example of its usage:
10002 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
10005 @item @code{if-exists-else}
10006 The @code{if-exists-else} spec function is similar to the @code{if-exists}
10007 spec function, except that it takes two arguments. The first argument is
10008 an absolute pathname to a file. If the file exists, @code{if-exists-else}
10009 returns the pathname. If it does not exist, it returns the second argument.
10010 This way, @code{if-exists-else} can be used to select one file or another,
10011 based on the existence of the first. Here is a small example of its usage:
10015 crt0%O%s %:if-exists(crti%O%s) \
10016 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
10019 @item @code{replace-outfile}
10020 The @code{replace-outfile} spec function takes two arguments. It looks for the
10021 first argument in the outfiles array and replaces it with the second argument. Here
10022 is a small example of its usage:
10025 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
10028 @item @code{remove-outfile}
10029 The @code{remove-outfile} spec function takes one argument. It looks for the
10030 first argument in the outfiles array and removes it. Here is a small example
10034 %:remove-outfile(-lm)
10037 @item @code{pass-through-libs}
10038 The @code{pass-through-libs} spec function takes any number of arguments. It
10039 finds any @option{-l} options and any non-options ending in ".a" (which it
10040 assumes are the names of linker input library archive files) and returns a
10041 result containing all the found arguments each prepended by
10042 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
10043 intended to be passed to the LTO linker plugin.
10046 %:pass-through-libs(%G %L %G)
10049 @item @code{print-asm-header}
10050 The @code{print-asm-header} function takes no arguments and simply
10051 prints a banner like:
10057 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
10060 It is used to separate compiler options from assembler options
10061 in the @option{--target-help} output.
10064 @item %@{@code{S}@}
10065 Substitutes the @code{-S} switch, if that switch was given to GCC@.
10066 If that switch was not specified, this substitutes nothing. Note that
10067 the leading dash is omitted when specifying this option, and it is
10068 automatically inserted if the substitution is performed. Thus the spec
10069 string @samp{%@{foo@}} would match the command-line option @option{-foo}
10070 and would output the command line option @option{-foo}.
10072 @item %W@{@code{S}@}
10073 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10074 deleted on failure.
10076 @item %@{@code{S}*@}
10077 Substitutes all the switches specified to GCC whose names start
10078 with @code{-S}, but which also take an argument. This is used for
10079 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10080 GCC considers @option{-o foo} as being
10081 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
10082 text, including the space. Thus two arguments would be generated.
10084 @item %@{@code{S}*&@code{T}*@}
10085 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10086 (the order of @code{S} and @code{T} in the spec is not significant).
10087 There can be any number of ampersand-separated variables; for each the
10088 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10090 @item %@{@code{S}:@code{X}@}
10091 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10093 @item %@{!@code{S}:@code{X}@}
10094 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10096 @item %@{@code{S}*:@code{X}@}
10097 Substitutes @code{X} if one or more switches whose names start with
10098 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10099 once, no matter how many such switches appeared. However, if @code{%*}
10100 appears somewhere in @code{X}, then @code{X} will be substituted once
10101 for each matching switch, with the @code{%*} replaced by the part of
10102 that switch that matched the @code{*}.
10104 @item %@{.@code{S}:@code{X}@}
10105 Substitutes @code{X}, if processing a file with suffix @code{S}.
10107 @item %@{!.@code{S}:@code{X}@}
10108 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10110 @item %@{,@code{S}:@code{X}@}
10111 Substitutes @code{X}, if processing a file for language @code{S}.
10113 @item %@{!,@code{S}:@code{X}@}
10114 Substitutes @code{X}, if not processing a file for language @code{S}.
10116 @item %@{@code{S}|@code{P}:@code{X}@}
10117 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10118 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10119 @code{*} sequences as well, although they have a stronger binding than
10120 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10121 alternatives must be starred, and only the first matching alternative
10124 For example, a spec string like this:
10127 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10130 will output the following command-line options from the following input
10131 command-line options:
10136 -d fred.c -foo -baz -boggle
10137 -d jim.d -bar -baz -boggle
10140 @item %@{S:X; T:Y; :D@}
10142 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10143 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10144 be as many clauses as you need. This may be combined with @code{.},
10145 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10150 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10151 construct may contain other nested @samp{%} constructs or spaces, or
10152 even newlines. They are processed as usual, as described above.
10153 Trailing white space in @code{X} is ignored. White space may also
10154 appear anywhere on the left side of the colon in these constructs,
10155 except between @code{.} or @code{*} and the corresponding word.
10157 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10158 handled specifically in these constructs. If another value of
10159 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10160 @option{-W} switch is found later in the command line, the earlier
10161 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10162 just one letter, which passes all matching options.
10164 The character @samp{|} at the beginning of the predicate text is used to
10165 indicate that a command should be piped to the following command, but
10166 only if @option{-pipe} is specified.
10168 It is built into GCC which switches take arguments and which do not.
10169 (You might think it would be useful to generalize this to allow each
10170 compiler's spec to say which switches take arguments. But this cannot
10171 be done in a consistent fashion. GCC cannot even decide which input
10172 files have been specified without knowing which switches take arguments,
10173 and it must know which input files to compile in order to tell which
10176 GCC also knows implicitly that arguments starting in @option{-l} are to be
10177 treated as compiler output files, and passed to the linker in their
10178 proper position among the other output files.
10180 @c man begin OPTIONS
10182 @node Target Options
10183 @section Specifying Target Machine and Compiler Version
10184 @cindex target options
10185 @cindex cross compiling
10186 @cindex specifying machine version
10187 @cindex specifying compiler version and target machine
10188 @cindex compiler version, specifying
10189 @cindex target machine, specifying
10191 The usual way to run GCC is to run the executable called @command{gcc}, or
10192 @command{@var{machine}-gcc} when cross-compiling, or
10193 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10194 one that was installed last.
10196 @node Submodel Options
10197 @section Hardware Models and Configurations
10198 @cindex submodel options
10199 @cindex specifying hardware config
10200 @cindex hardware models and configurations, specifying
10201 @cindex machine dependent options
10203 Each target machine types can have its own
10204 special options, starting with @samp{-m}, to choose among various
10205 hardware models or configurations---for example, 68010 vs 68020,
10206 floating coprocessor or none. A single installed version of the
10207 compiler can compile for any model or configuration, according to the
10210 Some configurations of the compiler also support additional special
10211 options, usually for compatibility with other compilers on the same
10214 @c This list is ordered alphanumerically by subsection name.
10215 @c It should be the same order and spelling as these options are listed
10216 @c in Machine Dependent Options
10221 * Blackfin Options::
10225 * DEC Alpha Options::
10226 * DEC Alpha/VMS Options::
10229 * GNU/Linux Options::
10232 * i386 and x86-64 Options::
10233 * i386 and x86-64 Windows Options::
10235 * IA-64/VMS Options::
10242 * MicroBlaze Options::
10245 * MN10300 Options::
10247 * picoChip Options::
10248 * PowerPC Options::
10249 * RS/6000 and PowerPC Options::
10251 * S/390 and zSeries Options::
10254 * Solaris 2 Options::
10257 * System V Options::
10260 * VxWorks Options::
10262 * Xstormy16 Options::
10264 * zSeries Options::
10268 @subsection ARM Options
10269 @cindex ARM options
10271 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10275 @item -mabi=@var{name}
10277 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10278 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10281 @opindex mapcs-frame
10282 Generate a stack frame that is compliant with the ARM Procedure Call
10283 Standard for all functions, even if this is not strictly necessary for
10284 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10285 with this option will cause the stack frames not to be generated for
10286 leaf functions. The default is @option{-mno-apcs-frame}.
10290 This is a synonym for @option{-mapcs-frame}.
10293 @c not currently implemented
10294 @item -mapcs-stack-check
10295 @opindex mapcs-stack-check
10296 Generate code to check the amount of stack space available upon entry to
10297 every function (that actually uses some stack space). If there is
10298 insufficient space available then either the function
10299 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10300 called, depending upon the amount of stack space required. The run time
10301 system is required to provide these functions. The default is
10302 @option{-mno-apcs-stack-check}, since this produces smaller code.
10304 @c not currently implemented
10306 @opindex mapcs-float
10307 Pass floating point arguments using the float point registers. This is
10308 one of the variants of the APCS@. This option is recommended if the
10309 target hardware has a floating point unit or if a lot of floating point
10310 arithmetic is going to be performed by the code. The default is
10311 @option{-mno-apcs-float}, since integer only code is slightly increased in
10312 size if @option{-mapcs-float} is used.
10314 @c not currently implemented
10315 @item -mapcs-reentrant
10316 @opindex mapcs-reentrant
10317 Generate reentrant, position independent code. The default is
10318 @option{-mno-apcs-reentrant}.
10321 @item -mthumb-interwork
10322 @opindex mthumb-interwork
10323 Generate code which supports calling between the ARM and Thumb
10324 instruction sets. Without this option, on pre-v5 architectures, the
10325 two instruction sets cannot be reliably used inside one program. The
10326 default is @option{-mno-thumb-interwork}, since slightly larger code
10327 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10328 configurations this option is meaningless.
10330 @item -mno-sched-prolog
10331 @opindex mno-sched-prolog
10332 Prevent the reordering of instructions in the function prolog, or the
10333 merging of those instruction with the instructions in the function's
10334 body. This means that all functions will start with a recognizable set
10335 of instructions (or in fact one of a choice from a small set of
10336 different function prologues), and this information can be used to
10337 locate the start if functions inside an executable piece of code. The
10338 default is @option{-msched-prolog}.
10340 @item -mfloat-abi=@var{name}
10341 @opindex mfloat-abi
10342 Specifies which floating-point ABI to use. Permissible values
10343 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10345 Specifying @samp{soft} causes GCC to generate output containing
10346 library calls for floating-point operations.
10347 @samp{softfp} allows the generation of code using hardware floating-point
10348 instructions, but still uses the soft-float calling conventions.
10349 @samp{hard} allows generation of floating-point instructions
10350 and uses FPU-specific calling conventions.
10352 The default depends on the specific target configuration. Note that
10353 the hard-float and soft-float ABIs are not link-compatible; you must
10354 compile your entire program with the same ABI, and link with a
10355 compatible set of libraries.
10357 @item -mlittle-endian
10358 @opindex mlittle-endian
10359 Generate code for a processor running in little-endian mode. This is
10360 the default for all standard configurations.
10363 @opindex mbig-endian
10364 Generate code for a processor running in big-endian mode; the default is
10365 to compile code for a little-endian processor.
10367 @item -mwords-little-endian
10368 @opindex mwords-little-endian
10369 This option only applies when generating code for big-endian processors.
10370 Generate code for a little-endian word order but a big-endian byte
10371 order. That is, a byte order of the form @samp{32107654}. Note: this
10372 option should only be used if you require compatibility with code for
10373 big-endian ARM processors generated by versions of the compiler prior to
10374 2.8. This option is now deprecated.
10376 @item -mcpu=@var{name}
10378 This specifies the name of the target ARM processor. GCC uses this name
10379 to determine what kind of instructions it can emit when generating
10380 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10381 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10382 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10383 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10384 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10386 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10387 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10388 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10389 @samp{strongarm1110},
10390 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10391 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10392 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10393 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10394 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10395 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10396 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10397 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10398 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10399 @samp{cortex-m4}, @samp{cortex-m3},
10402 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10403 @samp{fa526}, @samp{fa626},
10404 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10407 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10408 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10409 See @option{-mtune} for more information.
10411 @option{-mcpu=native} causes the compiler to auto-detect the CPU
10412 of the build computer. At present, this feature is only supported on
10413 Linux, and not all architectures are recognised. If the auto-detect is
10414 unsuccessful the option has no effect.
10416 @item -mtune=@var{name}
10418 This option is very similar to the @option{-mcpu=} option, except that
10419 instead of specifying the actual target processor type, and hence
10420 restricting which instructions can be used, it specifies that GCC should
10421 tune the performance of the code as if the target were of the type
10422 specified in this option, but still choosing the instructions that it
10423 will generate based on the CPU specified by a @option{-mcpu=} option.
10424 For some ARM implementations better performance can be obtained by using
10427 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10428 performance for a blend of processors within architecture @var{arch}.
10429 The aim is to generate code that run well on the current most popular
10430 processors, balancing between optimizations that benefit some CPUs in the
10431 range, and avoiding performance pitfalls of other CPUs. The effects of
10432 this option may change in future GCC versions as CPU models come and go.
10434 @option{-mtune=native} causes the compiler to auto-detect the CPU
10435 of the build computer. At present, this feature is only supported on
10436 Linux, and not all architectures are recognised. If the auto-detect is
10437 unsuccessful the option has no effect.
10439 @item -march=@var{name}
10441 This specifies the name of the target ARM architecture. GCC uses this
10442 name to determine what kind of instructions it can emit when generating
10443 assembly code. This option can be used in conjunction with or instead
10444 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10445 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10446 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10447 @samp{armv6}, @samp{armv6j},
10448 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10449 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10450 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10452 @option{-march=native} causes the compiler to auto-detect the architecture
10453 of the build computer. At present, this feature is only supported on
10454 Linux, and not all architectures are recognised. If the auto-detect is
10455 unsuccessful the option has no effect.
10457 @item -mfpu=@var{name}
10458 @itemx -mfpe=@var{number}
10459 @itemx -mfp=@var{number}
10463 This specifies what floating point hardware (or hardware emulation) is
10464 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10465 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10466 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10467 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10468 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10469 @option{-mfp} and @option{-mfpe} are synonyms for
10470 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10473 If @option{-msoft-float} is specified this specifies the format of
10474 floating point values.
10476 If the selected floating-point hardware includes the NEON extension
10477 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10478 operations will not be used by GCC's auto-vectorization pass unless
10479 @option{-funsafe-math-optimizations} is also specified. This is
10480 because NEON hardware does not fully implement the IEEE 754 standard for
10481 floating-point arithmetic (in particular denormal values are treated as
10482 zero), so the use of NEON instructions may lead to a loss of precision.
10484 @item -mfp16-format=@var{name}
10485 @opindex mfp16-format
10486 Specify the format of the @code{__fp16} half-precision floating-point type.
10487 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10488 the default is @samp{none}, in which case the @code{__fp16} type is not
10489 defined. @xref{Half-Precision}, for more information.
10491 @item -mstructure-size-boundary=@var{n}
10492 @opindex mstructure-size-boundary
10493 The size of all structures and unions will be rounded up to a multiple
10494 of the number of bits set by this option. Permissible values are 8, 32
10495 and 64. The default value varies for different toolchains. For the COFF
10496 targeted toolchain the default value is 8. A value of 64 is only allowed
10497 if the underlying ABI supports it.
10499 Specifying the larger number can produce faster, more efficient code, but
10500 can also increase the size of the program. Different values are potentially
10501 incompatible. Code compiled with one value cannot necessarily expect to
10502 work with code or libraries compiled with another value, if they exchange
10503 information using structures or unions.
10505 @item -mabort-on-noreturn
10506 @opindex mabort-on-noreturn
10507 Generate a call to the function @code{abort} at the end of a
10508 @code{noreturn} function. It will be executed if the function tries to
10512 @itemx -mno-long-calls
10513 @opindex mlong-calls
10514 @opindex mno-long-calls
10515 Tells the compiler to perform function calls by first loading the
10516 address of the function into a register and then performing a subroutine
10517 call on this register. This switch is needed if the target function
10518 will lie outside of the 64 megabyte addressing range of the offset based
10519 version of subroutine call instruction.
10521 Even if this switch is enabled, not all function calls will be turned
10522 into long calls. The heuristic is that static functions, functions
10523 which have the @samp{short-call} attribute, functions that are inside
10524 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10525 definitions have already been compiled within the current compilation
10526 unit, will not be turned into long calls. The exception to this rule is
10527 that weak function definitions, functions with the @samp{long-call}
10528 attribute or the @samp{section} attribute, and functions that are within
10529 the scope of a @samp{#pragma long_calls} directive, will always be
10530 turned into long calls.
10532 This feature is not enabled by default. Specifying
10533 @option{-mno-long-calls} will restore the default behavior, as will
10534 placing the function calls within the scope of a @samp{#pragma
10535 long_calls_off} directive. Note these switches have no effect on how
10536 the compiler generates code to handle function calls via function
10539 @item -msingle-pic-base
10540 @opindex msingle-pic-base
10541 Treat the register used for PIC addressing as read-only, rather than
10542 loading it in the prologue for each function. The run-time system is
10543 responsible for initializing this register with an appropriate value
10544 before execution begins.
10546 @item -mpic-register=@var{reg}
10547 @opindex mpic-register
10548 Specify the register to be used for PIC addressing. The default is R10
10549 unless stack-checking is enabled, when R9 is used.
10551 @item -mcirrus-fix-invalid-insns
10552 @opindex mcirrus-fix-invalid-insns
10553 @opindex mno-cirrus-fix-invalid-insns
10554 Insert NOPs into the instruction stream to in order to work around
10555 problems with invalid Maverick instruction combinations. This option
10556 is only valid if the @option{-mcpu=ep9312} option has been used to
10557 enable generation of instructions for the Cirrus Maverick floating
10558 point co-processor. This option is not enabled by default, since the
10559 problem is only present in older Maverick implementations. The default
10560 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10563 @item -mpoke-function-name
10564 @opindex mpoke-function-name
10565 Write the name of each function into the text section, directly
10566 preceding the function prologue. The generated code is similar to this:
10570 .ascii "arm_poke_function_name", 0
10573 .word 0xff000000 + (t1 - t0)
10574 arm_poke_function_name
10576 stmfd sp!, @{fp, ip, lr, pc@}
10580 When performing a stack backtrace, code can inspect the value of
10581 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10582 location @code{pc - 12} and the top 8 bits are set, then we know that
10583 there is a function name embedded immediately preceding this location
10584 and has length @code{((pc[-3]) & 0xff000000)}.
10591 Select between generating code that executes in ARM and Thumb
10592 states. The default for most configurations is to generate code
10593 that executes in ARM state, but the default can be changed by
10594 configuring GCC with the @option{--with-mode=}@var{state}
10598 @opindex mtpcs-frame
10599 Generate a stack frame that is compliant with the Thumb Procedure Call
10600 Standard for all non-leaf functions. (A leaf function is one that does
10601 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10603 @item -mtpcs-leaf-frame
10604 @opindex mtpcs-leaf-frame
10605 Generate a stack frame that is compliant with the Thumb Procedure Call
10606 Standard for all leaf functions. (A leaf function is one that does
10607 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10609 @item -mcallee-super-interworking
10610 @opindex mcallee-super-interworking
10611 Gives all externally visible functions in the file being compiled an ARM
10612 instruction set header which switches to Thumb mode before executing the
10613 rest of the function. This allows these functions to be called from
10614 non-interworking code. This option is not valid in AAPCS configurations
10615 because interworking is enabled by default.
10617 @item -mcaller-super-interworking
10618 @opindex mcaller-super-interworking
10619 Allows calls via function pointers (including virtual functions) to
10620 execute correctly regardless of whether the target code has been
10621 compiled for interworking or not. There is a small overhead in the cost
10622 of executing a function pointer if this option is enabled. This option
10623 is not valid in AAPCS configurations because interworking is enabled
10626 @item -mtp=@var{name}
10628 Specify the access model for the thread local storage pointer. The valid
10629 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10630 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10631 (supported in the arm6k architecture), and @option{auto}, which uses the
10632 best available method for the selected processor. The default setting is
10635 @item -mtls-dialect=@var{dialect}
10636 @opindex mtls-dialect
10637 Specify the dialect to use for accessing thread local storage. Two
10638 dialects are supported --- @option{gnu} and @option{gnu2}. The
10639 @option{gnu} dialect selects the original GNU scheme for supporting
10640 local and global dynamic TLS models. The @option{gnu2} dialect
10641 selects the GNU descriptor scheme, which provides better performance
10642 for shared libraries. The GNU descriptor scheme is compatible with
10643 the original scheme, but does require new assembler, linker and
10644 library support. Initial and local exec TLS models are unaffected by
10645 this option and always use the original scheme.
10647 @item -mword-relocations
10648 @opindex mword-relocations
10649 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10650 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10651 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10654 @item -mfix-cortex-m3-ldrd
10655 @opindex mfix-cortex-m3-ldrd
10656 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10657 with overlapping destination and base registers are used. This option avoids
10658 generating these instructions. This option is enabled by default when
10659 @option{-mcpu=cortex-m3} is specified.
10664 @subsection AVR Options
10665 @cindex AVR Options
10667 These options are defined for AVR implementations:
10670 @item -mmcu=@var{mcu}
10672 Specify ATMEL AVR instruction set or MCU type.
10674 Instruction set avr1 is for the minimal AVR core, not supported by the C
10675 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10676 attiny11, attiny12, attiny15, attiny28).
10678 Instruction set avr2 (default) is for the classic AVR core with up to
10679 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10680 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10681 at90c8534, at90s8535).
10683 Instruction set avr3 is for the classic AVR core with up to 128K program
10684 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10686 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10687 memory space (MCU types: atmega8, atmega83, atmega85).
10689 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10690 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10691 atmega64, atmega128, at43usb355, at94k).
10693 @item -mno-interrupts
10694 @opindex mno-interrupts
10695 Generated code is not compatible with hardware interrupts.
10696 Code size will be smaller.
10698 @item -mcall-prologues
10699 @opindex mcall-prologues
10700 Functions prologues/epilogues expanded as call to appropriate
10701 subroutines. Code size will be smaller.
10704 @opindex mtiny-stack
10705 Change only the low 8 bits of the stack pointer.
10709 Assume int to be 8 bit integer. This affects the sizes of all types: A
10710 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10711 and long long will be 4 bytes. Please note that this option does not
10712 comply to the C standards, but it will provide you with smaller code
10717 Use register @code{X} in a way proposed by the hardware. This means
10718 that @code{X} will only be used in indirect, post-increment or
10719 pre-decrement addressing.
10721 Without this option, the @code{X} register may be used in the same way
10722 as @code{Y} or @code{Z} which then is emulated by additional
10724 For example, loading a value with @code{X+const} addressing with a
10725 small @code{const <= 63} to a register @var{Rn} will be printed as
10733 @subsubsection @code{EIND} and Devices with more than 128k Bytes of Flash
10735 Pointers in the implementation are 16 bits wide.
10736 The address of a function or label is represented as word address so
10737 that indirect jumps and calls can address any code address in the
10738 range of 64k words.
10740 In order to faciliate indirect jump on devices with more than 128k
10741 bytes of program memory space, there is a special function register called
10742 @code{EIND} that serves as most significant part of the target address
10743 when @code{EICALL} or @code{EIJMP} instructions are used.
10745 Indirect jumps and calls on these devices are handled as follows and
10746 are subject to some limitations:
10751 The compiler never sets @code{EIND}.
10754 The startup code from libgcc never sets @code{EIND}.
10755 Notice that startup code is a blend of code from libgcc and avr-libc.
10756 For the impact of avr-libc on @code{EIND}, see the
10757 @w{@uref{http://nongnu.org/avr-libc/user-manual,avr-libc user manual}}.
10760 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
10761 instructions or might read @code{EIND} directly.
10764 The compiler assumes that @code{EIND} never changes during the startup
10765 code or run of the application. In particular, @code{EIND} is not
10766 saved/restored in function or interrupt service routine
10770 It is legitimate for user-specific startup code to set up @code{EIND}
10771 early, for example by means of initialization code located in
10772 section @code{.init3}, and thus prior to general startup code that
10773 initializes RAM and calls constructors.
10776 For indirect calls to functions and computed goto, the linker will
10777 generate @emph{stubs}. Stubs are jump pads sometimes also called
10778 @emph{trampolines}. Thus, the indirect call/jump will jump to such a stub.
10779 The stub contains a direct jump to the desired address.
10782 Stubs will be generated automatically by the linker if
10783 the following two conditions are met:
10786 @item The address of a label is taken by means of the @code{gs} modifier
10787 (short for @emph{generate stubs}) like so:
10789 LDI r24, lo8(gs(@var{func}))
10790 LDI r25, hi8(gs(@var{func}))
10792 @item The final location of that label is in a code segment
10793 @emph{outside} the segment where the stubs are located.
10797 The compiler will emit such @code{gs} modifiers for code labels in the
10798 following situations:
10800 @item Taking address of a function or code label.
10801 @item Computed goto.
10802 @item If prologue-save function is used, see @option{-mcall-prologues}
10803 command line option.
10804 @item Switch/case dispatch tables. If you do not want such dispatch
10805 tables you can specify the @option{-fno-jump-tables} command line option.
10806 @item C and C++ constructors/destructors called during startup/shutdown.
10807 @item If the tools hit a @code{gs()} modifier explained above.
10811 The default linker script is arranged for code with @code{EIND = 0}.
10812 If code is supposed to work for a setup with @code{EIND != 0}, a custom
10813 linker script has to be used in order to place the sections whose
10814 name start with @code{.trampolines} into the segment where @code{EIND}
10818 Jumping to non-symbolic addresses like so is @emph{not} supported:
10823 /* Call function at word address 0x2 */
10824 return ((int(*)(void)) 0x2)();
10828 Instead, a stub has to be set up:
10833 extern int func_4 (void);
10835 /* Call function at byte address 0x4 */
10840 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
10841 Alternatively, @code{func_4} can be defined in the linker script.
10844 @node Blackfin Options
10845 @subsection Blackfin Options
10846 @cindex Blackfin Options
10849 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10851 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10852 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10853 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10854 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10855 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10856 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10857 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10858 @samp{bf561}, @samp{bf592}.
10859 The optional @var{sirevision} specifies the silicon revision of the target
10860 Blackfin processor. Any workarounds available for the targeted silicon revision
10861 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10862 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10863 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10864 hexadecimal digits representing the major and minor numbers in the silicon
10865 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10866 is not defined. If @var{sirevision} is @samp{any}, the
10867 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10868 If this optional @var{sirevision} is not used, GCC assumes the latest known
10869 silicon revision of the targeted Blackfin processor.
10871 Support for @samp{bf561} is incomplete. For @samp{bf561},
10872 Only the processor macro is defined.
10873 Without this option, @samp{bf532} is used as the processor by default.
10874 The corresponding predefined processor macros for @var{cpu} is to
10875 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10876 provided by libgloss to be linked in if @option{-msim} is not given.
10880 Specifies that the program will be run on the simulator. This causes
10881 the simulator BSP provided by libgloss to be linked in. This option
10882 has effect only for @samp{bfin-elf} toolchain.
10883 Certain other options, such as @option{-mid-shared-library} and
10884 @option{-mfdpic}, imply @option{-msim}.
10886 @item -momit-leaf-frame-pointer
10887 @opindex momit-leaf-frame-pointer
10888 Don't keep the frame pointer in a register for leaf functions. This
10889 avoids the instructions to save, set up and restore frame pointers and
10890 makes an extra register available in leaf functions. The option
10891 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10892 which might make debugging harder.
10894 @item -mspecld-anomaly
10895 @opindex mspecld-anomaly
10896 When enabled, the compiler will ensure that the generated code does not
10897 contain speculative loads after jump instructions. If this option is used,
10898 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10900 @item -mno-specld-anomaly
10901 @opindex mno-specld-anomaly
10902 Don't generate extra code to prevent speculative loads from occurring.
10904 @item -mcsync-anomaly
10905 @opindex mcsync-anomaly
10906 When enabled, the compiler will ensure that the generated code does not
10907 contain CSYNC or SSYNC instructions too soon after conditional branches.
10908 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10910 @item -mno-csync-anomaly
10911 @opindex mno-csync-anomaly
10912 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10913 occurring too soon after a conditional branch.
10917 When enabled, the compiler is free to take advantage of the knowledge that
10918 the entire program fits into the low 64k of memory.
10921 @opindex mno-low-64k
10922 Assume that the program is arbitrarily large. This is the default.
10924 @item -mstack-check-l1
10925 @opindex mstack-check-l1
10926 Do stack checking using information placed into L1 scratchpad memory by the
10929 @item -mid-shared-library
10930 @opindex mid-shared-library
10931 Generate code that supports shared libraries via the library ID method.
10932 This allows for execute in place and shared libraries in an environment
10933 without virtual memory management. This option implies @option{-fPIC}.
10934 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10936 @item -mno-id-shared-library
10937 @opindex mno-id-shared-library
10938 Generate code that doesn't assume ID based shared libraries are being used.
10939 This is the default.
10941 @item -mleaf-id-shared-library
10942 @opindex mleaf-id-shared-library
10943 Generate code that supports shared libraries via the library ID method,
10944 but assumes that this library or executable won't link against any other
10945 ID shared libraries. That allows the compiler to use faster code for jumps
10948 @item -mno-leaf-id-shared-library
10949 @opindex mno-leaf-id-shared-library
10950 Do not assume that the code being compiled won't link against any ID shared
10951 libraries. Slower code will be generated for jump and call insns.
10953 @item -mshared-library-id=n
10954 @opindex mshared-library-id
10955 Specified the identification number of the ID based shared library being
10956 compiled. Specifying a value of 0 will generate more compact code, specifying
10957 other values will force the allocation of that number to the current
10958 library but is no more space or time efficient than omitting this option.
10962 Generate code that allows the data segment to be located in a different
10963 area of memory from the text segment. This allows for execute in place in
10964 an environment without virtual memory management by eliminating relocations
10965 against the text section.
10967 @item -mno-sep-data
10968 @opindex mno-sep-data
10969 Generate code that assumes that the data segment follows the text segment.
10970 This is the default.
10973 @itemx -mno-long-calls
10974 @opindex mlong-calls
10975 @opindex mno-long-calls
10976 Tells the compiler to perform function calls by first loading the
10977 address of the function into a register and then performing a subroutine
10978 call on this register. This switch is needed if the target function
10979 will lie outside of the 24 bit addressing range of the offset based
10980 version of subroutine call instruction.
10982 This feature is not enabled by default. Specifying
10983 @option{-mno-long-calls} will restore the default behavior. Note these
10984 switches have no effect on how the compiler generates code to handle
10985 function calls via function pointers.
10989 Link with the fast floating-point library. This library relaxes some of
10990 the IEEE floating-point standard's rules for checking inputs against
10991 Not-a-Number (NAN), in the interest of performance.
10994 @opindex minline-plt
10995 Enable inlining of PLT entries in function calls to functions that are
10996 not known to bind locally. It has no effect without @option{-mfdpic}.
10999 @opindex mmulticore
11000 Build standalone application for multicore Blackfin processor. Proper
11001 start files and link scripts will be used to support multicore.
11002 This option defines @code{__BFIN_MULTICORE}. It can only be used with
11003 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
11004 @option{-mcorea} or @option{-mcoreb}. If it's used without
11005 @option{-mcorea} or @option{-mcoreb}, single application/dual core
11006 programming model is used. In this model, the main function of Core B
11007 should be named as coreb_main. If it's used with @option{-mcorea} or
11008 @option{-mcoreb}, one application per core programming model is used.
11009 If this option is not used, single core application programming
11014 Build standalone application for Core A of BF561 when using
11015 one application per core programming model. Proper start files
11016 and link scripts will be used to support Core A. This option
11017 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
11021 Build standalone application for Core B of BF561 when using
11022 one application per core programming model. Proper start files
11023 and link scripts will be used to support Core B. This option
11024 defines @code{__BFIN_COREB}. When this option is used, coreb_main
11025 should be used instead of main. It must be used with
11026 @option{-mmulticore}.
11030 Build standalone application for SDRAM. Proper start files and
11031 link scripts will be used to put the application into SDRAM.
11032 Loader should initialize SDRAM before loading the application
11033 into SDRAM. This option defines @code{__BFIN_SDRAM}.
11037 Assume that ICPLBs are enabled at runtime. This has an effect on certain
11038 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
11039 are enabled; for standalone applications the default is off.
11043 @subsection C6X Options
11044 @cindex C6X Options
11047 @item -march=@var{name}
11049 This specifies the name of the target architecture. GCC uses this
11050 name to determine what kind of instructions it can emit when generating
11051 assembly code. Permissible names are: @samp{c62x},
11052 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
11055 @opindex mbig-endian
11056 Generate code for a big endian target.
11058 @item -mlittle-endian
11059 @opindex mlittle-endian
11060 Generate code for a little endian target. This is the default.
11064 Choose startup files and linker script suitable for the simulator.
11066 @item -msdata=default
11067 @opindex msdata=default
11068 Put small global and static data in the @samp{.neardata} section,
11069 which is pointed to by register @code{B14}. Put small uninitialized
11070 global and static data in the @samp{.bss} section, which is adjacent
11071 to the @samp{.neardata} section. Put small read-only data into the
11072 @samp{.rodata} section. The corresponding sections used for large
11073 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
11076 @opindex msdata=all
11077 Put all data, not just small objets, into the sections reserved for
11078 small data, and use addressing relative to the @code{B14} register to
11082 @opindex msdata=none
11083 Make no use of the sections reserved for small data, and use absolute
11084 addresses to access all data. Put all initialized global and static
11085 data in the @samp{.fardata} section, and all uninitialized data in the
11086 @samp{.far} section. Put all constant data into the @samp{.const}
11091 @subsection CRIS Options
11092 @cindex CRIS Options
11094 These options are defined specifically for the CRIS ports.
11097 @item -march=@var{architecture-type}
11098 @itemx -mcpu=@var{architecture-type}
11101 Generate code for the specified architecture. The choices for
11102 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
11103 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
11104 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
11107 @item -mtune=@var{architecture-type}
11109 Tune to @var{architecture-type} everything applicable about the generated
11110 code, except for the ABI and the set of available instructions. The
11111 choices for @var{architecture-type} are the same as for
11112 @option{-march=@var{architecture-type}}.
11114 @item -mmax-stack-frame=@var{n}
11115 @opindex mmax-stack-frame
11116 Warn when the stack frame of a function exceeds @var{n} bytes.
11122 The options @option{-metrax4} and @option{-metrax100} are synonyms for
11123 @option{-march=v3} and @option{-march=v8} respectively.
11125 @item -mmul-bug-workaround
11126 @itemx -mno-mul-bug-workaround
11127 @opindex mmul-bug-workaround
11128 @opindex mno-mul-bug-workaround
11129 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
11130 models where it applies. This option is active by default.
11134 Enable CRIS-specific verbose debug-related information in the assembly
11135 code. This option also has the effect to turn off the @samp{#NO_APP}
11136 formatted-code indicator to the assembler at the beginning of the
11141 Do not use condition-code results from previous instruction; always emit
11142 compare and test instructions before use of condition codes.
11144 @item -mno-side-effects
11145 @opindex mno-side-effects
11146 Do not emit instructions with side-effects in addressing modes other than
11149 @item -mstack-align
11150 @itemx -mno-stack-align
11151 @itemx -mdata-align
11152 @itemx -mno-data-align
11153 @itemx -mconst-align
11154 @itemx -mno-const-align
11155 @opindex mstack-align
11156 @opindex mno-stack-align
11157 @opindex mdata-align
11158 @opindex mno-data-align
11159 @opindex mconst-align
11160 @opindex mno-const-align
11161 These options (no-options) arranges (eliminate arrangements) for the
11162 stack-frame, individual data and constants to be aligned for the maximum
11163 single data access size for the chosen CPU model. The default is to
11164 arrange for 32-bit alignment. ABI details such as structure layout are
11165 not affected by these options.
11173 Similar to the stack- data- and const-align options above, these options
11174 arrange for stack-frame, writable data and constants to all be 32-bit,
11175 16-bit or 8-bit aligned. The default is 32-bit alignment.
11177 @item -mno-prologue-epilogue
11178 @itemx -mprologue-epilogue
11179 @opindex mno-prologue-epilogue
11180 @opindex mprologue-epilogue
11181 With @option{-mno-prologue-epilogue}, the normal function prologue and
11182 epilogue that sets up the stack-frame are omitted and no return
11183 instructions or return sequences are generated in the code. Use this
11184 option only together with visual inspection of the compiled code: no
11185 warnings or errors are generated when call-saved registers must be saved,
11186 or storage for local variable needs to be allocated.
11190 @opindex mno-gotplt
11192 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
11193 instruction sequences that load addresses for functions from the PLT part
11194 of the GOT rather than (traditional on other architectures) calls to the
11195 PLT@. The default is @option{-mgotplt}.
11199 Legacy no-op option only recognized with the cris-axis-elf and
11200 cris-axis-linux-gnu targets.
11204 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
11208 This option, recognized for the cris-axis-elf arranges
11209 to link with input-output functions from a simulator library. Code,
11210 initialized data and zero-initialized data are allocated consecutively.
11214 Like @option{-sim}, but pass linker options to locate initialized data at
11215 0x40000000 and zero-initialized data at 0x80000000.
11218 @node Darwin Options
11219 @subsection Darwin Options
11220 @cindex Darwin options
11222 These options are defined for all architectures running the Darwin operating
11225 FSF GCC on Darwin does not create ``fat'' object files; it will create
11226 an object file for the single architecture that it was built to
11227 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11228 @option{-arch} options are used; it does so by running the compiler or
11229 linker multiple times and joining the results together with
11232 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11233 @samp{i686}) is determined by the flags that specify the ISA
11234 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11235 @option{-force_cpusubtype_ALL} option can be used to override this.
11237 The Darwin tools vary in their behavior when presented with an ISA
11238 mismatch. The assembler, @file{as}, will only permit instructions to
11239 be used that are valid for the subtype of the file it is generating,
11240 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11241 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11242 and print an error if asked to create a shared library with a less
11243 restrictive subtype than its input files (for instance, trying to put
11244 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11245 for executables, @file{ld}, will quietly give the executable the most
11246 restrictive subtype of any of its input files.
11251 Add the framework directory @var{dir} to the head of the list of
11252 directories to be searched for header files. These directories are
11253 interleaved with those specified by @option{-I} options and are
11254 scanned in a left-to-right order.
11256 A framework directory is a directory with frameworks in it. A
11257 framework is a directory with a @samp{"Headers"} and/or
11258 @samp{"PrivateHeaders"} directory contained directly in it that ends
11259 in @samp{".framework"}. The name of a framework is the name of this
11260 directory excluding the @samp{".framework"}. Headers associated with
11261 the framework are found in one of those two directories, with
11262 @samp{"Headers"} being searched first. A subframework is a framework
11263 directory that is in a framework's @samp{"Frameworks"} directory.
11264 Includes of subframework headers can only appear in a header of a
11265 framework that contains the subframework, or in a sibling subframework
11266 header. Two subframeworks are siblings if they occur in the same
11267 framework. A subframework should not have the same name as a
11268 framework, a warning will be issued if this is violated. Currently a
11269 subframework cannot have subframeworks, in the future, the mechanism
11270 may be extended to support this. The standard frameworks can be found
11271 in @samp{"/System/Library/Frameworks"} and
11272 @samp{"/Library/Frameworks"}. An example include looks like
11273 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11274 the name of the framework and header.h is found in the
11275 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11277 @item -iframework@var{dir}
11278 @opindex iframework
11279 Like @option{-F} except the directory is a treated as a system
11280 directory. The main difference between this @option{-iframework} and
11281 @option{-F} is that with @option{-iframework} the compiler does not
11282 warn about constructs contained within header files found via
11283 @var{dir}. This option is valid only for the C family of languages.
11287 Emit debugging information for symbols that are used. For STABS
11288 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11289 This is by default ON@.
11293 Emit debugging information for all symbols and types.
11295 @item -mmacosx-version-min=@var{version}
11296 The earliest version of MacOS X that this executable will run on
11297 is @var{version}. Typical values of @var{version} include @code{10.1},
11298 @code{10.2}, and @code{10.3.9}.
11300 If the compiler was built to use the system's headers by default,
11301 then the default for this option is the system version on which the
11302 compiler is running, otherwise the default is to make choices which
11303 are compatible with as many systems and code bases as possible.
11307 Enable kernel development mode. The @option{-mkernel} option sets
11308 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11309 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11310 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11311 applicable. This mode also sets @option{-mno-altivec},
11312 @option{-msoft-float}, @option{-fno-builtin} and
11313 @option{-mlong-branch} for PowerPC targets.
11315 @item -mone-byte-bool
11316 @opindex mone-byte-bool
11317 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11318 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11319 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11320 option has no effect on x86.
11322 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11323 to generate code that is not binary compatible with code generated
11324 without that switch. Using this switch may require recompiling all
11325 other modules in a program, including system libraries. Use this
11326 switch to conform to a non-default data model.
11328 @item -mfix-and-continue
11329 @itemx -ffix-and-continue
11330 @itemx -findirect-data
11331 @opindex mfix-and-continue
11332 @opindex ffix-and-continue
11333 @opindex findirect-data
11334 Generate code suitable for fast turn around development. Needed to
11335 enable gdb to dynamically load @code{.o} files into already running
11336 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11337 are provided for backwards compatibility.
11341 Loads all members of static archive libraries.
11342 See man ld(1) for more information.
11344 @item -arch_errors_fatal
11345 @opindex arch_errors_fatal
11346 Cause the errors having to do with files that have the wrong architecture
11349 @item -bind_at_load
11350 @opindex bind_at_load
11351 Causes the output file to be marked such that the dynamic linker will
11352 bind all undefined references when the file is loaded or launched.
11356 Produce a Mach-o bundle format file.
11357 See man ld(1) for more information.
11359 @item -bundle_loader @var{executable}
11360 @opindex bundle_loader
11361 This option specifies the @var{executable} that will be loading the build
11362 output file being linked. See man ld(1) for more information.
11365 @opindex dynamiclib
11366 When passed this option, GCC will produce a dynamic library instead of
11367 an executable when linking, using the Darwin @file{libtool} command.
11369 @item -force_cpusubtype_ALL
11370 @opindex force_cpusubtype_ALL
11371 This causes GCC's output file to have the @var{ALL} subtype, instead of
11372 one controlled by the @option{-mcpu} or @option{-march} option.
11374 @item -allowable_client @var{client_name}
11375 @itemx -client_name
11376 @itemx -compatibility_version
11377 @itemx -current_version
11379 @itemx -dependency-file
11381 @itemx -dylinker_install_name
11383 @itemx -exported_symbols_list
11386 @itemx -flat_namespace
11387 @itemx -force_flat_namespace
11388 @itemx -headerpad_max_install_names
11391 @itemx -install_name
11392 @itemx -keep_private_externs
11393 @itemx -multi_module
11394 @itemx -multiply_defined
11395 @itemx -multiply_defined_unused
11398 @itemx -no_dead_strip_inits_and_terms
11399 @itemx -nofixprebinding
11400 @itemx -nomultidefs
11402 @itemx -noseglinkedit
11403 @itemx -pagezero_size
11405 @itemx -prebind_all_twolevel_modules
11406 @itemx -private_bundle
11408 @itemx -read_only_relocs
11410 @itemx -sectobjectsymbols
11414 @itemx -sectobjectsymbols
11417 @itemx -segs_read_only_addr
11419 @itemx -segs_read_write_addr
11420 @itemx -seg_addr_table
11421 @itemx -seg_addr_table_filename
11422 @itemx -seglinkedit
11424 @itemx -segs_read_only_addr
11425 @itemx -segs_read_write_addr
11426 @itemx -single_module
11428 @itemx -sub_library
11430 @itemx -sub_umbrella
11431 @itemx -twolevel_namespace
11434 @itemx -unexported_symbols_list
11435 @itemx -weak_reference_mismatches
11436 @itemx -whatsloaded
11437 @opindex allowable_client
11438 @opindex client_name
11439 @opindex compatibility_version
11440 @opindex current_version
11441 @opindex dead_strip
11442 @opindex dependency-file
11443 @opindex dylib_file
11444 @opindex dylinker_install_name
11446 @opindex exported_symbols_list
11448 @opindex flat_namespace
11449 @opindex force_flat_namespace
11450 @opindex headerpad_max_install_names
11451 @opindex image_base
11453 @opindex install_name
11454 @opindex keep_private_externs
11455 @opindex multi_module
11456 @opindex multiply_defined
11457 @opindex multiply_defined_unused
11458 @opindex noall_load
11459 @opindex no_dead_strip_inits_and_terms
11460 @opindex nofixprebinding
11461 @opindex nomultidefs
11463 @opindex noseglinkedit
11464 @opindex pagezero_size
11466 @opindex prebind_all_twolevel_modules
11467 @opindex private_bundle
11468 @opindex read_only_relocs
11470 @opindex sectobjectsymbols
11473 @opindex sectcreate
11474 @opindex sectobjectsymbols
11477 @opindex segs_read_only_addr
11478 @opindex segs_read_write_addr
11479 @opindex seg_addr_table
11480 @opindex seg_addr_table_filename
11481 @opindex seglinkedit
11483 @opindex segs_read_only_addr
11484 @opindex segs_read_write_addr
11485 @opindex single_module
11487 @opindex sub_library
11488 @opindex sub_umbrella
11489 @opindex twolevel_namespace
11492 @opindex unexported_symbols_list
11493 @opindex weak_reference_mismatches
11494 @opindex whatsloaded
11495 These options are passed to the Darwin linker. The Darwin linker man page
11496 describes them in detail.
11499 @node DEC Alpha Options
11500 @subsection DEC Alpha Options
11502 These @samp{-m} options are defined for the DEC Alpha implementations:
11505 @item -mno-soft-float
11506 @itemx -msoft-float
11507 @opindex mno-soft-float
11508 @opindex msoft-float
11509 Use (do not use) the hardware floating-point instructions for
11510 floating-point operations. When @option{-msoft-float} is specified,
11511 functions in @file{libgcc.a} will be used to perform floating-point
11512 operations. Unless they are replaced by routines that emulate the
11513 floating-point operations, or compiled in such a way as to call such
11514 emulations routines, these routines will issue floating-point
11515 operations. If you are compiling for an Alpha without floating-point
11516 operations, you must ensure that the library is built so as not to call
11519 Note that Alpha implementations without floating-point operations are
11520 required to have floating-point registers.
11523 @itemx -mno-fp-regs
11525 @opindex mno-fp-regs
11526 Generate code that uses (does not use) the floating-point register set.
11527 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11528 register set is not used, floating point operands are passed in integer
11529 registers as if they were integers and floating-point results are passed
11530 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11531 so any function with a floating-point argument or return value called by code
11532 compiled with @option{-mno-fp-regs} must also be compiled with that
11535 A typical use of this option is building a kernel that does not use,
11536 and hence need not save and restore, any floating-point registers.
11540 The Alpha architecture implements floating-point hardware optimized for
11541 maximum performance. It is mostly compliant with the IEEE floating
11542 point standard. However, for full compliance, software assistance is
11543 required. This option generates code fully IEEE compliant code
11544 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11545 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11546 defined during compilation. The resulting code is less efficient but is
11547 able to correctly support denormalized numbers and exceptional IEEE
11548 values such as not-a-number and plus/minus infinity. Other Alpha
11549 compilers call this option @option{-ieee_with_no_inexact}.
11551 @item -mieee-with-inexact
11552 @opindex mieee-with-inexact
11553 This is like @option{-mieee} except the generated code also maintains
11554 the IEEE @var{inexact-flag}. Turning on this option causes the
11555 generated code to implement fully-compliant IEEE math. In addition to
11556 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11557 macro. On some Alpha implementations the resulting code may execute
11558 significantly slower than the code generated by default. Since there is
11559 very little code that depends on the @var{inexact-flag}, you should
11560 normally not specify this option. Other Alpha compilers call this
11561 option @option{-ieee_with_inexact}.
11563 @item -mfp-trap-mode=@var{trap-mode}
11564 @opindex mfp-trap-mode
11565 This option controls what floating-point related traps are enabled.
11566 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11567 The trap mode can be set to one of four values:
11571 This is the default (normal) setting. The only traps that are enabled
11572 are the ones that cannot be disabled in software (e.g., division by zero
11576 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11580 Like @samp{u}, but the instructions are marked to be safe for software
11581 completion (see Alpha architecture manual for details).
11584 Like @samp{su}, but inexact traps are enabled as well.
11587 @item -mfp-rounding-mode=@var{rounding-mode}
11588 @opindex mfp-rounding-mode
11589 Selects the IEEE rounding mode. Other Alpha compilers call this option
11590 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11595 Normal IEEE rounding mode. Floating point numbers are rounded towards
11596 the nearest machine number or towards the even machine number in case
11600 Round towards minus infinity.
11603 Chopped rounding mode. Floating point numbers are rounded towards zero.
11606 Dynamic rounding mode. A field in the floating point control register
11607 (@var{fpcr}, see Alpha architecture reference manual) controls the
11608 rounding mode in effect. The C library initializes this register for
11609 rounding towards plus infinity. Thus, unless your program modifies the
11610 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11613 @item -mtrap-precision=@var{trap-precision}
11614 @opindex mtrap-precision
11615 In the Alpha architecture, floating point traps are imprecise. This
11616 means without software assistance it is impossible to recover from a
11617 floating trap and program execution normally needs to be terminated.
11618 GCC can generate code that can assist operating system trap handlers
11619 in determining the exact location that caused a floating point trap.
11620 Depending on the requirements of an application, different levels of
11621 precisions can be selected:
11625 Program precision. This option is the default and means a trap handler
11626 can only identify which program caused a floating point exception.
11629 Function precision. The trap handler can determine the function that
11630 caused a floating point exception.
11633 Instruction precision. The trap handler can determine the exact
11634 instruction that caused a floating point exception.
11637 Other Alpha compilers provide the equivalent options called
11638 @option{-scope_safe} and @option{-resumption_safe}.
11640 @item -mieee-conformant
11641 @opindex mieee-conformant
11642 This option marks the generated code as IEEE conformant. You must not
11643 use this option unless you also specify @option{-mtrap-precision=i} and either
11644 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11645 is to emit the line @samp{.eflag 48} in the function prologue of the
11646 generated assembly file. Under DEC Unix, this has the effect that
11647 IEEE-conformant math library routines will be linked in.
11649 @item -mbuild-constants
11650 @opindex mbuild-constants
11651 Normally GCC examines a 32- or 64-bit integer constant to
11652 see if it can construct it from smaller constants in two or three
11653 instructions. If it cannot, it will output the constant as a literal and
11654 generate code to load it from the data segment at runtime.
11656 Use this option to require GCC to construct @emph{all} integer constants
11657 using code, even if it takes more instructions (the maximum is six).
11659 You would typically use this option to build a shared library dynamic
11660 loader. Itself a shared library, it must relocate itself in memory
11661 before it can find the variables and constants in its own data segment.
11667 Select whether to generate code to be assembled by the vendor-supplied
11668 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11686 Indicate whether GCC should generate code to use the optional BWX,
11687 CIX, FIX and MAX instruction sets. The default is to use the instruction
11688 sets supported by the CPU type specified via @option{-mcpu=} option or that
11689 of the CPU on which GCC was built if none was specified.
11692 @itemx -mfloat-ieee
11693 @opindex mfloat-vax
11694 @opindex mfloat-ieee
11695 Generate code that uses (does not use) VAX F and G floating point
11696 arithmetic instead of IEEE single and double precision.
11698 @item -mexplicit-relocs
11699 @itemx -mno-explicit-relocs
11700 @opindex mexplicit-relocs
11701 @opindex mno-explicit-relocs
11702 Older Alpha assemblers provided no way to generate symbol relocations
11703 except via assembler macros. Use of these macros does not allow
11704 optimal instruction scheduling. GNU binutils as of version 2.12
11705 supports a new syntax that allows the compiler to explicitly mark
11706 which relocations should apply to which instructions. This option
11707 is mostly useful for debugging, as GCC detects the capabilities of
11708 the assembler when it is built and sets the default accordingly.
11711 @itemx -mlarge-data
11712 @opindex msmall-data
11713 @opindex mlarge-data
11714 When @option{-mexplicit-relocs} is in effect, static data is
11715 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11716 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11717 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11718 16-bit relocations off of the @code{$gp} register. This limits the
11719 size of the small data area to 64KB, but allows the variables to be
11720 directly accessed via a single instruction.
11722 The default is @option{-mlarge-data}. With this option the data area
11723 is limited to just below 2GB@. Programs that require more than 2GB of
11724 data must use @code{malloc} or @code{mmap} to allocate the data in the
11725 heap instead of in the program's data segment.
11727 When generating code for shared libraries, @option{-fpic} implies
11728 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11731 @itemx -mlarge-text
11732 @opindex msmall-text
11733 @opindex mlarge-text
11734 When @option{-msmall-text} is used, the compiler assumes that the
11735 code of the entire program (or shared library) fits in 4MB, and is
11736 thus reachable with a branch instruction. When @option{-msmall-data}
11737 is used, the compiler can assume that all local symbols share the
11738 same @code{$gp} value, and thus reduce the number of instructions
11739 required for a function call from 4 to 1.
11741 The default is @option{-mlarge-text}.
11743 @item -mcpu=@var{cpu_type}
11745 Set the instruction set and instruction scheduling parameters for
11746 machine type @var{cpu_type}. You can specify either the @samp{EV}
11747 style name or the corresponding chip number. GCC supports scheduling
11748 parameters for the EV4, EV5 and EV6 family of processors and will
11749 choose the default values for the instruction set from the processor
11750 you specify. If you do not specify a processor type, GCC will default
11751 to the processor on which the compiler was built.
11753 Supported values for @var{cpu_type} are
11759 Schedules as an EV4 and has no instruction set extensions.
11763 Schedules as an EV5 and has no instruction set extensions.
11767 Schedules as an EV5 and supports the BWX extension.
11772 Schedules as an EV5 and supports the BWX and MAX extensions.
11776 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11780 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11783 Native toolchains also support the value @samp{native},
11784 which selects the best architecture option for the host processor.
11785 @option{-mcpu=native} has no effect if GCC does not recognize
11788 @item -mtune=@var{cpu_type}
11790 Set only the instruction scheduling parameters for machine type
11791 @var{cpu_type}. The instruction set is not changed.
11793 Native toolchains also support the value @samp{native},
11794 which selects the best architecture option for the host processor.
11795 @option{-mtune=native} has no effect if GCC does not recognize
11798 @item -mmemory-latency=@var{time}
11799 @opindex mmemory-latency
11800 Sets the latency the scheduler should assume for typical memory
11801 references as seen by the application. This number is highly
11802 dependent on the memory access patterns used by the application
11803 and the size of the external cache on the machine.
11805 Valid options for @var{time} are
11809 A decimal number representing clock cycles.
11815 The compiler contains estimates of the number of clock cycles for
11816 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11817 (also called Dcache, Scache, and Bcache), as well as to main memory.
11818 Note that L3 is only valid for EV5.
11823 @node DEC Alpha/VMS Options
11824 @subsection DEC Alpha/VMS Options
11826 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11829 @item -mvms-return-codes
11830 @opindex mvms-return-codes
11831 Return VMS condition codes from main. The default is to return POSIX
11832 style condition (e.g.@: error) codes.
11834 @item -mdebug-main=@var{prefix}
11835 @opindex mdebug-main=@var{prefix}
11836 Flag the first routine whose name starts with @var{prefix} as the main
11837 routine for the debugger.
11841 Default to 64bit memory allocation routines.
11845 @subsection FR30 Options
11846 @cindex FR30 Options
11848 These options are defined specifically for the FR30 port.
11852 @item -msmall-model
11853 @opindex msmall-model
11854 Use the small address space model. This can produce smaller code, but
11855 it does assume that all symbolic values and addresses will fit into a
11860 Assume that run-time support has been provided and so there is no need
11861 to include the simulator library (@file{libsim.a}) on the linker
11867 @subsection FRV Options
11868 @cindex FRV Options
11874 Only use the first 32 general purpose registers.
11879 Use all 64 general purpose registers.
11884 Use only the first 32 floating point registers.
11889 Use all 64 floating point registers
11892 @opindex mhard-float
11894 Use hardware instructions for floating point operations.
11897 @opindex msoft-float
11899 Use library routines for floating point operations.
11904 Dynamically allocate condition code registers.
11909 Do not try to dynamically allocate condition code registers, only
11910 use @code{icc0} and @code{fcc0}.
11915 Change ABI to use double word insns.
11920 Do not use double word instructions.
11925 Use floating point double instructions.
11928 @opindex mno-double
11930 Do not use floating point double instructions.
11935 Use media instructions.
11940 Do not use media instructions.
11945 Use multiply and add/subtract instructions.
11948 @opindex mno-muladd
11950 Do not use multiply and add/subtract instructions.
11955 Select the FDPIC ABI, that uses function descriptors to represent
11956 pointers to functions. Without any PIC/PIE-related options, it
11957 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11958 assumes GOT entries and small data are within a 12-bit range from the
11959 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11960 are computed with 32 bits.
11961 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11964 @opindex minline-plt
11966 Enable inlining of PLT entries in function calls to functions that are
11967 not known to bind locally. It has no effect without @option{-mfdpic}.
11968 It's enabled by default if optimizing for speed and compiling for
11969 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11970 optimization option such as @option{-O3} or above is present in the
11976 Assume a large TLS segment when generating thread-local code.
11981 Do not assume a large TLS segment when generating thread-local code.
11986 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11987 that is known to be in read-only sections. It's enabled by default,
11988 except for @option{-fpic} or @option{-fpie}: even though it may help
11989 make the global offset table smaller, it trades 1 instruction for 4.
11990 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11991 one of which may be shared by multiple symbols, and it avoids the need
11992 for a GOT entry for the referenced symbol, so it's more likely to be a
11993 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11995 @item -multilib-library-pic
11996 @opindex multilib-library-pic
11998 Link with the (library, not FD) pic libraries. It's implied by
11999 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
12000 @option{-fpic} without @option{-mfdpic}. You should never have to use
12004 @opindex mlinked-fp
12006 Follow the EABI requirement of always creating a frame pointer whenever
12007 a stack frame is allocated. This option is enabled by default and can
12008 be disabled with @option{-mno-linked-fp}.
12011 @opindex mlong-calls
12013 Use indirect addressing to call functions outside the current
12014 compilation unit. This allows the functions to be placed anywhere
12015 within the 32-bit address space.
12017 @item -malign-labels
12018 @opindex malign-labels
12020 Try to align labels to an 8-byte boundary by inserting nops into the
12021 previous packet. This option only has an effect when VLIW packing
12022 is enabled. It doesn't create new packets; it merely adds nops to
12025 @item -mlibrary-pic
12026 @opindex mlibrary-pic
12028 Generate position-independent EABI code.
12033 Use only the first four media accumulator registers.
12038 Use all eight media accumulator registers.
12043 Pack VLIW instructions.
12048 Do not pack VLIW instructions.
12051 @opindex mno-eflags
12053 Do not mark ABI switches in e_flags.
12056 @opindex mcond-move
12058 Enable the use of conditional-move instructions (default).
12060 This switch is mainly for debugging the compiler and will likely be removed
12061 in a future version.
12063 @item -mno-cond-move
12064 @opindex mno-cond-move
12066 Disable the use of conditional-move instructions.
12068 This switch is mainly for debugging the compiler and will likely be removed
12069 in a future version.
12074 Enable the use of conditional set instructions (default).
12076 This switch is mainly for debugging the compiler and will likely be removed
12077 in a future version.
12082 Disable the use of conditional set instructions.
12084 This switch is mainly for debugging the compiler and will likely be removed
12085 in a future version.
12088 @opindex mcond-exec
12090 Enable the use of conditional execution (default).
12092 This switch is mainly for debugging the compiler and will likely be removed
12093 in a future version.
12095 @item -mno-cond-exec
12096 @opindex mno-cond-exec
12098 Disable the use of conditional execution.
12100 This switch is mainly for debugging the compiler and will likely be removed
12101 in a future version.
12103 @item -mvliw-branch
12104 @opindex mvliw-branch
12106 Run a pass to pack branches into VLIW instructions (default).
12108 This switch is mainly for debugging the compiler and will likely be removed
12109 in a future version.
12111 @item -mno-vliw-branch
12112 @opindex mno-vliw-branch
12114 Do not run a pass to pack branches into VLIW instructions.
12116 This switch is mainly for debugging the compiler and will likely be removed
12117 in a future version.
12119 @item -mmulti-cond-exec
12120 @opindex mmulti-cond-exec
12122 Enable optimization of @code{&&} and @code{||} in conditional execution
12125 This switch is mainly for debugging the compiler and will likely be removed
12126 in a future version.
12128 @item -mno-multi-cond-exec
12129 @opindex mno-multi-cond-exec
12131 Disable optimization of @code{&&} and @code{||} in conditional execution.
12133 This switch is mainly for debugging the compiler and will likely be removed
12134 in a future version.
12136 @item -mnested-cond-exec
12137 @opindex mnested-cond-exec
12139 Enable nested conditional execution optimizations (default).
12141 This switch is mainly for debugging the compiler and will likely be removed
12142 in a future version.
12144 @item -mno-nested-cond-exec
12145 @opindex mno-nested-cond-exec
12147 Disable nested conditional execution optimizations.
12149 This switch is mainly for debugging the compiler and will likely be removed
12150 in a future version.
12152 @item -moptimize-membar
12153 @opindex moptimize-membar
12155 This switch removes redundant @code{membar} instructions from the
12156 compiler generated code. It is enabled by default.
12158 @item -mno-optimize-membar
12159 @opindex mno-optimize-membar
12161 This switch disables the automatic removal of redundant @code{membar}
12162 instructions from the generated code.
12164 @item -mtomcat-stats
12165 @opindex mtomcat-stats
12167 Cause gas to print out tomcat statistics.
12169 @item -mcpu=@var{cpu}
12172 Select the processor type for which to generate code. Possible values are
12173 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
12174 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
12178 @node GNU/Linux Options
12179 @subsection GNU/Linux Options
12181 These @samp{-m} options are defined for GNU/Linux targets:
12186 Use the GNU C library. This is the default except
12187 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
12191 Use uClibc C library. This is the default on
12192 @samp{*-*-linux-*uclibc*} targets.
12196 Use Bionic C library. This is the default on
12197 @samp{*-*-linux-*android*} targets.
12201 Compile code compatible with Android platform. This is the default on
12202 @samp{*-*-linux-*android*} targets.
12204 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
12205 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
12206 this option makes the GCC driver pass Android-specific options to the linker.
12207 Finally, this option causes the preprocessor macro @code{__ANDROID__}
12210 @item -tno-android-cc
12211 @opindex tno-android-cc
12212 Disable compilation effects of @option{-mandroid}, i.e., do not enable
12213 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
12214 @option{-fno-rtti} by default.
12216 @item -tno-android-ld
12217 @opindex tno-android-ld
12218 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12219 linking options to the linker.
12223 @node H8/300 Options
12224 @subsection H8/300 Options
12226 These @samp{-m} options are defined for the H8/300 implementations:
12231 Shorten some address references at link time, when possible; uses the
12232 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12233 ld, Using ld}, for a fuller description.
12237 Generate code for the H8/300H@.
12241 Generate code for the H8S@.
12245 Generate code for the H8S and H8/300H in the normal mode. This switch
12246 must be used either with @option{-mh} or @option{-ms}.
12250 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12254 Make @code{int} data 32 bits by default.
12257 @opindex malign-300
12258 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12259 The default for the H8/300H and H8S is to align longs and floats on 4
12261 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
12262 This option has no effect on the H8/300.
12266 @subsection HPPA Options
12267 @cindex HPPA Options
12269 These @samp{-m} options are defined for the HPPA family of computers:
12272 @item -march=@var{architecture-type}
12274 Generate code for the specified architecture. The choices for
12275 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12276 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12277 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12278 architecture option for your machine. Code compiled for lower numbered
12279 architectures will run on higher numbered architectures, but not the
12282 @item -mpa-risc-1-0
12283 @itemx -mpa-risc-1-1
12284 @itemx -mpa-risc-2-0
12285 @opindex mpa-risc-1-0
12286 @opindex mpa-risc-1-1
12287 @opindex mpa-risc-2-0
12288 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12291 @opindex mbig-switch
12292 Generate code suitable for big switch tables. Use this option only if
12293 the assembler/linker complain about out of range branches within a switch
12296 @item -mjump-in-delay
12297 @opindex mjump-in-delay
12298 Fill delay slots of function calls with unconditional jump instructions
12299 by modifying the return pointer for the function call to be the target
12300 of the conditional jump.
12302 @item -mdisable-fpregs
12303 @opindex mdisable-fpregs
12304 Prevent floating point registers from being used in any manner. This is
12305 necessary for compiling kernels which perform lazy context switching of
12306 floating point registers. If you use this option and attempt to perform
12307 floating point operations, the compiler will abort.
12309 @item -mdisable-indexing
12310 @opindex mdisable-indexing
12311 Prevent the compiler from using indexing address modes. This avoids some
12312 rather obscure problems when compiling MIG generated code under MACH@.
12314 @item -mno-space-regs
12315 @opindex mno-space-regs
12316 Generate code that assumes the target has no space registers. This allows
12317 GCC to generate faster indirect calls and use unscaled index address modes.
12319 Such code is suitable for level 0 PA systems and kernels.
12321 @item -mfast-indirect-calls
12322 @opindex mfast-indirect-calls
12323 Generate code that assumes calls never cross space boundaries. This
12324 allows GCC to emit code which performs faster indirect calls.
12326 This option will not work in the presence of shared libraries or nested
12329 @item -mfixed-range=@var{register-range}
12330 @opindex mfixed-range
12331 Generate code treating the given register range as fixed registers.
12332 A fixed register is one that the register allocator can not use. This is
12333 useful when compiling kernel code. A register range is specified as
12334 two registers separated by a dash. Multiple register ranges can be
12335 specified separated by a comma.
12337 @item -mlong-load-store
12338 @opindex mlong-load-store
12339 Generate 3-instruction load and store sequences as sometimes required by
12340 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12343 @item -mportable-runtime
12344 @opindex mportable-runtime
12345 Use the portable calling conventions proposed by HP for ELF systems.
12349 Enable the use of assembler directives only GAS understands.
12351 @item -mschedule=@var{cpu-type}
12353 Schedule code according to the constraints for the machine type
12354 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12355 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12356 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12357 proper scheduling option for your machine. The default scheduling is
12361 @opindex mlinker-opt
12362 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12363 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12364 linkers in which they give bogus error messages when linking some programs.
12367 @opindex msoft-float
12368 Generate output containing library calls for floating point.
12369 @strong{Warning:} the requisite libraries are not available for all HPPA
12370 targets. Normally the facilities of the machine's usual C compiler are
12371 used, but this cannot be done directly in cross-compilation. You must make
12372 your own arrangements to provide suitable library functions for
12375 @option{-msoft-float} changes the calling convention in the output file;
12376 therefore, it is only useful if you compile @emph{all} of a program with
12377 this option. In particular, you need to compile @file{libgcc.a}, the
12378 library that comes with GCC, with @option{-msoft-float} in order for
12383 Generate the predefine, @code{_SIO}, for server IO@. The default is
12384 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12385 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12386 options are available under HP-UX and HI-UX@.
12390 Use GNU ld specific options. This passes @option{-shared} to ld when
12391 building a shared library. It is the default when GCC is configured,
12392 explicitly or implicitly, with the GNU linker. This option does not
12393 have any affect on which ld is called, it only changes what parameters
12394 are passed to that ld. The ld that is called is determined by the
12395 @option{--with-ld} configure option, GCC's program search path, and
12396 finally by the user's @env{PATH}. The linker used by GCC can be printed
12397 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12398 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12402 Use HP ld specific options. This passes @option{-b} to ld when building
12403 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12404 links. It is the default when GCC is configured, explicitly or
12405 implicitly, with the HP linker. This option does not have any affect on
12406 which ld is called, it only changes what parameters are passed to that
12407 ld. The ld that is called is determined by the @option{--with-ld}
12408 configure option, GCC's program search path, and finally by the user's
12409 @env{PATH}. The linker used by GCC can be printed using @samp{which
12410 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12411 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12414 @opindex mno-long-calls
12415 Generate code that uses long call sequences. This ensures that a call
12416 is always able to reach linker generated stubs. The default is to generate
12417 long calls only when the distance from the call site to the beginning
12418 of the function or translation unit, as the case may be, exceeds a
12419 predefined limit set by the branch type being used. The limits for
12420 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12421 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12424 Distances are measured from the beginning of functions when using the
12425 @option{-ffunction-sections} option, or when using the @option{-mgas}
12426 and @option{-mno-portable-runtime} options together under HP-UX with
12429 It is normally not desirable to use this option as it will degrade
12430 performance. However, it may be useful in large applications,
12431 particularly when partial linking is used to build the application.
12433 The types of long calls used depends on the capabilities of the
12434 assembler and linker, and the type of code being generated. The
12435 impact on systems that support long absolute calls, and long pic
12436 symbol-difference or pc-relative calls should be relatively small.
12437 However, an indirect call is used on 32-bit ELF systems in pic code
12438 and it is quite long.
12440 @item -munix=@var{unix-std}
12442 Generate compiler predefines and select a startfile for the specified
12443 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12444 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12445 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12446 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12447 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12450 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12451 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12452 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12453 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12454 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12455 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12457 It is @emph{important} to note that this option changes the interfaces
12458 for various library routines. It also affects the operational behavior
12459 of the C library. Thus, @emph{extreme} care is needed in using this
12462 Library code that is intended to operate with more than one UNIX
12463 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12464 as appropriate. Most GNU software doesn't provide this capability.
12468 Suppress the generation of link options to search libdld.sl when the
12469 @option{-static} option is specified on HP-UX 10 and later.
12473 The HP-UX implementation of setlocale in libc has a dependency on
12474 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12475 when the @option{-static} option is specified, special link options
12476 are needed to resolve this dependency.
12478 On HP-UX 10 and later, the GCC driver adds the necessary options to
12479 link with libdld.sl when the @option{-static} option is specified.
12480 This causes the resulting binary to be dynamic. On the 64-bit port,
12481 the linkers generate dynamic binaries by default in any case. The
12482 @option{-nolibdld} option can be used to prevent the GCC driver from
12483 adding these link options.
12487 Add support for multithreading with the @dfn{dce thread} library
12488 under HP-UX@. This option sets flags for both the preprocessor and
12492 @node i386 and x86-64 Options
12493 @subsection Intel 386 and AMD x86-64 Options
12494 @cindex i386 Options
12495 @cindex x86-64 Options
12496 @cindex Intel 386 Options
12497 @cindex AMD x86-64 Options
12499 These @samp{-m} options are defined for the i386 and x86-64 family of
12503 @item -mtune=@var{cpu-type}
12505 Tune to @var{cpu-type} everything applicable about the generated code, except
12506 for the ABI and the set of available instructions. The choices for
12507 @var{cpu-type} are:
12510 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12511 If you know the CPU on which your code will run, then you should use
12512 the corresponding @option{-mtune} option instead of
12513 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12514 of your application will have, then you should use this option.
12516 As new processors are deployed in the marketplace, the behavior of this
12517 option will change. Therefore, if you upgrade to a newer version of
12518 GCC, the code generated option will change to reflect the processors
12519 that were most common when that version of GCC was released.
12521 There is no @option{-march=generic} option because @option{-march}
12522 indicates the instruction set the compiler can use, and there is no
12523 generic instruction set applicable to all processors. In contrast,
12524 @option{-mtune} indicates the processor (or, in this case, collection of
12525 processors) for which the code is optimized.
12527 This selects the CPU to tune for at compilation time by determining
12528 the processor type of the compiling machine. Using @option{-mtune=native}
12529 will produce code optimized for the local machine under the constraints
12530 of the selected instruction set. Using @option{-march=native} will
12531 enable all instruction subsets supported by the local machine (hence
12532 the result might not run on different machines).
12534 Original Intel's i386 CPU@.
12536 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12537 @item i586, pentium
12538 Intel Pentium CPU with no MMX support.
12540 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12542 Intel PentiumPro CPU@.
12544 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12545 instruction set will be used, so the code will run on all i686 family chips.
12547 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12548 @item pentium3, pentium3m
12549 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12552 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12553 support. Used by Centrino notebooks.
12554 @item pentium4, pentium4m
12555 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12557 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12560 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12561 SSE2 and SSE3 instruction set support.
12563 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12564 instruction set support.
12566 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12567 and SSE4.2 instruction set support.
12569 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12570 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12572 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12573 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
12576 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12577 instruction set support.
12579 AMD K6 CPU with MMX instruction set support.
12581 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12582 @item athlon, athlon-tbird
12583 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12585 @item athlon-4, athlon-xp, athlon-mp
12586 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12587 instruction set support.
12588 @item k8, opteron, athlon64, athlon-fx
12589 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12590 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12591 @item k8-sse3, opteron-sse3, athlon64-sse3
12592 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12593 @item amdfam10, barcelona
12594 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12595 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12596 instruction set extensions.)
12598 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12601 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12602 instruction set support.
12604 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12605 implemented for this chip.)
12607 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12608 implemented for this chip.)
12610 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12613 While picking a specific @var{cpu-type} will schedule things appropriately
12614 for that particular chip, the compiler will not generate any code that
12615 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12618 @item -march=@var{cpu-type}
12620 Generate instructions for the machine type @var{cpu-type}. The choices
12621 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12622 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12624 @item -mcpu=@var{cpu-type}
12626 A deprecated synonym for @option{-mtune}.
12628 @item -mfpmath=@var{unit}
12630 Generate floating point arithmetics for selected unit @var{unit}. The choices
12631 for @var{unit} are:
12635 Use the standard 387 floating point coprocessor present majority of chips and
12636 emulated otherwise. Code compiled with this option will run almost everywhere.
12637 The temporary results are computed in 80bit precision instead of precision
12638 specified by the type resulting in slightly different results compared to most
12639 of other chips. See @option{-ffloat-store} for more detailed description.
12641 This is the default choice for i386 compiler.
12644 Use scalar floating point instructions present in the SSE instruction set.
12645 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12646 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12647 instruction set supports only single precision arithmetics, thus the double and
12648 extended precision arithmetics is still done using 387. Later version, present
12649 only in Pentium4 and the future AMD x86-64 chips supports double precision
12652 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12653 or @option{-msse2} switches to enable SSE extensions and make this option
12654 effective. For the x86-64 compiler, these extensions are enabled by default.
12656 The resulting code should be considerably faster in the majority of cases and avoid
12657 the numerical instability problems of 387 code, but may break some existing
12658 code that expects temporaries to be 80bit.
12660 This is the default choice for the x86-64 compiler.
12665 Attempt to utilize both instruction sets at once. This effectively double the
12666 amount of available registers and on chips with separate execution units for
12667 387 and SSE the execution resources too. Use this option with care, as it is
12668 still experimental, because the GCC register allocator does not model separate
12669 functional units well resulting in instable performance.
12672 @item -masm=@var{dialect}
12673 @opindex masm=@var{dialect}
12674 Output asm instructions using selected @var{dialect}. Supported
12675 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12676 not support @samp{intel}.
12679 @itemx -mno-ieee-fp
12681 @opindex mno-ieee-fp
12682 Control whether or not the compiler uses IEEE floating point
12683 comparisons. These handle correctly the case where the result of a
12684 comparison is unordered.
12687 @opindex msoft-float
12688 Generate output containing library calls for floating point.
12689 @strong{Warning:} the requisite libraries are not part of GCC@.
12690 Normally the facilities of the machine's usual C compiler are used, but
12691 this can't be done directly in cross-compilation. You must make your
12692 own arrangements to provide suitable library functions for
12695 On machines where a function returns floating point results in the 80387
12696 register stack, some floating point opcodes may be emitted even if
12697 @option{-msoft-float} is used.
12699 @item -mno-fp-ret-in-387
12700 @opindex mno-fp-ret-in-387
12701 Do not use the FPU registers for return values of functions.
12703 The usual calling convention has functions return values of types
12704 @code{float} and @code{double} in an FPU register, even if there
12705 is no FPU@. The idea is that the operating system should emulate
12708 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12709 in ordinary CPU registers instead.
12711 @item -mno-fancy-math-387
12712 @opindex mno-fancy-math-387
12713 Some 387 emulators do not support the @code{sin}, @code{cos} and
12714 @code{sqrt} instructions for the 387. Specify this option to avoid
12715 generating those instructions. This option is the default on FreeBSD,
12716 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12717 indicates that the target CPU will always have an FPU and so the
12718 instruction will not need emulation. As of revision 2.6.1, these
12719 instructions are not generated unless you also use the
12720 @option{-funsafe-math-optimizations} switch.
12722 @item -malign-double
12723 @itemx -mno-align-double
12724 @opindex malign-double
12725 @opindex mno-align-double
12726 Control whether GCC aligns @code{double}, @code{long double}, and
12727 @code{long long} variables on a two word boundary or a one word
12728 boundary. Aligning @code{double} variables on a two word boundary will
12729 produce code that runs somewhat faster on a @samp{Pentium} at the
12730 expense of more memory.
12732 On x86-64, @option{-malign-double} is enabled by default.
12734 @strong{Warning:} if you use the @option{-malign-double} switch,
12735 structures containing the above types will be aligned differently than
12736 the published application binary interface specifications for the 386
12737 and will not be binary compatible with structures in code compiled
12738 without that switch.
12740 @item -m96bit-long-double
12741 @itemx -m128bit-long-double
12742 @opindex m96bit-long-double
12743 @opindex m128bit-long-double
12744 These switches control the size of @code{long double} type. The i386
12745 application binary interface specifies the size to be 96 bits,
12746 so @option{-m96bit-long-double} is the default in 32 bit mode.
12748 Modern architectures (Pentium and newer) would prefer @code{long double}
12749 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12750 conforming to the ABI, this would not be possible. So specifying a
12751 @option{-m128bit-long-double} will align @code{long double}
12752 to a 16 byte boundary by padding the @code{long double} with an additional
12755 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12756 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12758 Notice that neither of these options enable any extra precision over the x87
12759 standard of 80 bits for a @code{long double}.
12761 @strong{Warning:} if you override the default value for your target ABI, the
12762 structures and arrays containing @code{long double} variables will change
12763 their size as well as function calling convention for function taking
12764 @code{long double} will be modified. Hence they will not be binary
12765 compatible with arrays or structures in code compiled without that switch.
12767 @item -mlarge-data-threshold=@var{number}
12768 @opindex mlarge-data-threshold=@var{number}
12769 When @option{-mcmodel=medium} is specified, the data greater than
12770 @var{threshold} are placed in large data section. This value must be the
12771 same across all object linked into the binary and defaults to 65535.
12775 Use a different function-calling convention, in which functions that
12776 take a fixed number of arguments return with the @code{ret} @var{num}
12777 instruction, which pops their arguments while returning. This saves one
12778 instruction in the caller since there is no need to pop the arguments
12781 You can specify that an individual function is called with this calling
12782 sequence with the function attribute @samp{stdcall}. You can also
12783 override the @option{-mrtd} option by using the function attribute
12784 @samp{cdecl}. @xref{Function Attributes}.
12786 @strong{Warning:} this calling convention is incompatible with the one
12787 normally used on Unix, so you cannot use it if you need to call
12788 libraries compiled with the Unix compiler.
12790 Also, you must provide function prototypes for all functions that
12791 take variable numbers of arguments (including @code{printf});
12792 otherwise incorrect code will be generated for calls to those
12795 In addition, seriously incorrect code will result if you call a
12796 function with too many arguments. (Normally, extra arguments are
12797 harmlessly ignored.)
12799 @item -mregparm=@var{num}
12801 Control how many registers are used to pass integer arguments. By
12802 default, no registers are used to pass arguments, and at most 3
12803 registers can be used. You can control this behavior for a specific
12804 function by using the function attribute @samp{regparm}.
12805 @xref{Function Attributes}.
12807 @strong{Warning:} if you use this switch, and
12808 @var{num} is nonzero, then you must build all modules with the same
12809 value, including any libraries. This includes the system libraries and
12813 @opindex msseregparm
12814 Use SSE register passing conventions for float and double arguments
12815 and return values. You can control this behavior for a specific
12816 function by using the function attribute @samp{sseregparm}.
12817 @xref{Function Attributes}.
12819 @strong{Warning:} if you use this switch then you must build all
12820 modules with the same value, including any libraries. This includes
12821 the system libraries and startup modules.
12823 @item -mvect8-ret-in-mem
12824 @opindex mvect8-ret-in-mem
12825 Return 8-byte vectors in memory instead of MMX registers. This is the
12826 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12827 Studio compilers until version 12. Later compiler versions (starting
12828 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12829 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12830 you need to remain compatible with existing code produced by those
12831 previous compiler versions or older versions of GCC.
12840 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12841 is specified, the significands of results of floating-point operations are
12842 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12843 significands of results of floating-point operations to 53 bits (double
12844 precision) and @option{-mpc80} rounds the significands of results of
12845 floating-point operations to 64 bits (extended double precision), which is
12846 the default. When this option is used, floating-point operations in higher
12847 precisions are not available to the programmer without setting the FPU
12848 control word explicitly.
12850 Setting the rounding of floating-point operations to less than the default
12851 80 bits can speed some programs by 2% or more. Note that some mathematical
12852 libraries assume that extended precision (80 bit) floating-point operations
12853 are enabled by default; routines in such libraries could suffer significant
12854 loss of accuracy, typically through so-called "catastrophic cancellation",
12855 when this option is used to set the precision to less than extended precision.
12857 @item -mstackrealign
12858 @opindex mstackrealign
12859 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12860 option will generate an alternate prologue and epilogue that realigns the
12861 runtime stack if necessary. This supports mixing legacy codes that keep
12862 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12863 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12864 applicable to individual functions.
12866 @item -mpreferred-stack-boundary=@var{num}
12867 @opindex mpreferred-stack-boundary
12868 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12869 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12870 the default is 4 (16 bytes or 128 bits).
12872 @item -mincoming-stack-boundary=@var{num}
12873 @opindex mincoming-stack-boundary
12874 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12875 boundary. If @option{-mincoming-stack-boundary} is not specified,
12876 the one specified by @option{-mpreferred-stack-boundary} will be used.
12878 On Pentium and PentiumPro, @code{double} and @code{long double} values
12879 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12880 suffer significant run time performance penalties. On Pentium III, the
12881 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12882 properly if it is not 16 byte aligned.
12884 To ensure proper alignment of this values on the stack, the stack boundary
12885 must be as aligned as that required by any value stored on the stack.
12886 Further, every function must be generated such that it keeps the stack
12887 aligned. Thus calling a function compiled with a higher preferred
12888 stack boundary from a function compiled with a lower preferred stack
12889 boundary will most likely misalign the stack. It is recommended that
12890 libraries that use callbacks always use the default setting.
12892 This extra alignment does consume extra stack space, and generally
12893 increases code size. Code that is sensitive to stack space usage, such
12894 as embedded systems and operating system kernels, may want to reduce the
12895 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12924 @itemx -mno-fsgsbase
12960 These switches enable or disable the use of instructions in the MMX, SSE,
12961 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
12962 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
12963 @: extended instruction sets.
12964 These extensions are also available as built-in functions: see
12965 @ref{X86 Built-in Functions}, for details of the functions enabled and
12966 disabled by these switches.
12968 To have SSE/SSE2 instructions generated automatically from floating-point
12969 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12971 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12972 generates new AVX instructions or AVX equivalence for all SSEx instructions
12975 These options will enable GCC to use these extended instructions in
12976 generated code, even without @option{-mfpmath=sse}. Applications which
12977 perform runtime CPU detection must compile separate files for each
12978 supported architecture, using the appropriate flags. In particular,
12979 the file containing the CPU detection code should be compiled without
12984 This option instructs GCC to emit a @code{cld} instruction in the prologue
12985 of functions that use string instructions. String instructions depend on
12986 the DF flag to select between autoincrement or autodecrement mode. While the
12987 ABI specifies the DF flag to be cleared on function entry, some operating
12988 systems violate this specification by not clearing the DF flag in their
12989 exception dispatchers. The exception handler can be invoked with the DF flag
12990 set which leads to wrong direction mode, when string instructions are used.
12991 This option can be enabled by default on 32-bit x86 targets by configuring
12992 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12993 instructions can be suppressed with the @option{-mno-cld} compiler option
12997 @opindex mvzeroupper
12998 This option instructs GCC to emit a @code{vzeroupper} instruction
12999 before a transfer of control flow out of the function to minimize
13000 AVX to SSE transition penalty as well as remove unnecessary zeroupper
13005 This option will enable GCC to use CMPXCHG16B instruction in generated code.
13006 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
13007 data types. This is useful for high resolution counters that could be updated
13008 by multiple processors (or cores). This instruction is generated as part of
13009 atomic built-in functions: see @ref{Atomic Builtins} for details.
13013 This option will enable GCC to use SAHF instruction in generated 64-bit code.
13014 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
13015 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
13016 SAHF are load and store instructions, respectively, for certain status flags.
13017 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
13018 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
13022 This option will enable GCC to use movbe instruction to implement
13023 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
13027 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
13028 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
13029 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
13033 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
13034 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
13035 to increase precision instead of DIVSS and SQRTSS (and their vectorized
13036 variants) for single precision floating point arguments. These instructions
13037 are generated only when @option{-funsafe-math-optimizations} is enabled
13038 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
13039 Note that while the throughput of the sequence is higher than the throughput
13040 of the non-reciprocal instruction, the precision of the sequence can be
13041 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
13043 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of RSQRTSS
13044 (or RSQRTPS) already with @option{-ffast-math} (or the above option
13045 combination), and doesn't need @option{-mrecip}.
13047 Also note that GCC emits the above sequence with additional Newton-Raphson step
13048 for vectorized single float division and vectorized @code{sqrtf(@var{x})}
13049 already with @option{-ffast-math} (or the above option combination), and
13050 doesn't need @option{-mrecip}.
13052 @item -mrecip=@var{opt}
13053 @opindex mrecip=opt
13054 This option allows to control which reciprocal estimate instructions
13055 may be used. @var{opt} is a comma separated list of options, that may
13056 be preceded by a @code{!} to invert the option:
13057 @code{all}: enable all estimate instructions,
13058 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
13059 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip},
13060 @code{div}: enable the approximation for scalar division,
13061 @code{vec-div}: enable the approximation for vectorized division,
13062 @code{sqrt}: enable the approximation for scalar square root,
13063 @code{vec-sqrt}: enable the approximation for vectorized square root.
13065 So for example, @option{-mrecip=all,!sqrt} would enable
13066 all of the reciprocal approximations, except for square root.
13068 @item -mveclibabi=@var{type}
13069 @opindex mveclibabi
13070 Specifies the ABI type to use for vectorizing intrinsics using an
13071 external library. Supported types are @code{svml} for the Intel short
13072 vector math library and @code{acml} for the AMD math core library style
13073 of interfacing. GCC will currently emit calls to @code{vmldExp2},
13074 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
13075 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
13076 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
13077 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
13078 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
13079 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
13080 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
13081 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
13082 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
13083 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
13084 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
13085 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
13086 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
13087 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
13088 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
13089 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
13090 compatible library will have to be specified at link time.
13092 @item -mabi=@var{name}
13094 Generate code for the specified calling convention. Permissible values
13095 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
13096 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
13097 ABI when targeting Windows. On all other systems, the default is the
13098 SYSV ABI. You can control this behavior for a specific function by
13099 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
13100 @xref{Function Attributes}.
13102 @item -mtls-dialect=@var{type}
13103 @opindex mtls-dialect
13104 Generate code to access thread-local storage using the @samp{gnu} or
13105 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
13106 @samp{gnu2} is more efficient, but it may add compile- and run-time
13107 requirements that cannot be satisfied on all systems.
13110 @itemx -mno-push-args
13111 @opindex mpush-args
13112 @opindex mno-push-args
13113 Use PUSH operations to store outgoing parameters. This method is shorter
13114 and usually equally fast as method using SUB/MOV operations and is enabled
13115 by default. In some cases disabling it may improve performance because of
13116 improved scheduling and reduced dependencies.
13118 @item -maccumulate-outgoing-args
13119 @opindex maccumulate-outgoing-args
13120 If enabled, the maximum amount of space required for outgoing arguments will be
13121 computed in the function prologue. This is faster on most modern CPUs
13122 because of reduced dependencies, improved scheduling and reduced stack usage
13123 when preferred stack boundary is not equal to 2. The drawback is a notable
13124 increase in code size. This switch implies @option{-mno-push-args}.
13128 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
13129 on thread-safe exception handling must compile and link all code with the
13130 @option{-mthreads} option. When compiling, @option{-mthreads} defines
13131 @option{-D_MT}; when linking, it links in a special thread helper library
13132 @option{-lmingwthrd} which cleans up per thread exception handling data.
13134 @item -mno-align-stringops
13135 @opindex mno-align-stringops
13136 Do not align destination of inlined string operations. This switch reduces
13137 code size and improves performance in case the destination is already aligned,
13138 but GCC doesn't know about it.
13140 @item -minline-all-stringops
13141 @opindex minline-all-stringops
13142 By default GCC inlines string operations only when destination is known to be
13143 aligned at least to 4 byte boundary. This enables more inlining, increase code
13144 size, but may improve performance of code that depends on fast memcpy, strlen
13145 and memset for short lengths.
13147 @item -minline-stringops-dynamically
13148 @opindex minline-stringops-dynamically
13149 For string operation of unknown size, inline runtime checks so for small
13150 blocks inline code is used, while for large blocks library call is used.
13152 @item -mstringop-strategy=@var{alg}
13153 @opindex mstringop-strategy=@var{alg}
13154 Overwrite internal decision heuristic about particular algorithm to inline
13155 string operation with. The allowed values are @code{rep_byte},
13156 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
13157 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
13158 expanding inline loop, @code{libcall} for always expanding library call.
13160 @item -momit-leaf-frame-pointer
13161 @opindex momit-leaf-frame-pointer
13162 Don't keep the frame pointer in a register for leaf functions. This
13163 avoids the instructions to save, set up and restore frame pointers and
13164 makes an extra register available in leaf functions. The option
13165 @option{-fomit-frame-pointer} removes the frame pointer for all functions
13166 which might make debugging harder.
13168 @item -mtls-direct-seg-refs
13169 @itemx -mno-tls-direct-seg-refs
13170 @opindex mtls-direct-seg-refs
13171 Controls whether TLS variables may be accessed with offsets from the
13172 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
13173 or whether the thread base pointer must be added. Whether or not this
13174 is legal depends on the operating system, and whether it maps the
13175 segment to cover the entire TLS area.
13177 For systems that use GNU libc, the default is on.
13180 @itemx -mno-sse2avx
13182 Specify that the assembler should encode SSE instructions with VEX
13183 prefix. The option @option{-mavx} turns this on by default.
13188 If profiling is active @option{-pg} put the profiling
13189 counter call before prologue.
13190 Note: On x86 architectures the attribute @code{ms_hook_prologue}
13191 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
13194 @itemx -mno-8bit-idiv
13196 On some processors, like Intel Atom, 8bit unsigned integer divide is
13197 much faster than 32bit/64bit integer divide. This option will generate a
13198 runt-time check. If both dividend and divisor are within range of 0
13199 to 255, 8bit unsigned integer divide will be used instead of
13200 32bit/64bit integer divide.
13202 @item -mavx256-split-unaligned-load
13203 @item -mavx256-split-unaligned-store
13204 @opindex avx256-split-unaligned-load
13205 @opindex avx256-split-unaligned-store
13206 Split 32-byte AVX unaligned load and store.
13210 These @samp{-m} switches are supported in addition to the above
13211 on AMD x86-64 processors in 64-bit environments.
13220 Generate code for a 32-bit or 64-bit environment.
13221 The @option{-m32} option sets int, long and pointer to 32 bits and
13222 generates code that runs on any i386 system.
13223 The @option{-m64} option sets int to 32 bits and long and pointer
13224 to 64 bits and generates code for AMD's x86-64 architecture.
13225 The @option{-mx32} option sets int, long and pointer to 32 bits and
13226 generates code for AMD's x86-64 architecture.
13227 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
13228 and @option{-mdynamic-no-pic} options.
13230 @item -mno-red-zone
13231 @opindex mno-red-zone
13232 Do not use a so called red zone for x86-64 code. The red zone is mandated
13233 by the x86-64 ABI, it is a 128-byte area beyond the location of the
13234 stack pointer that will not be modified by signal or interrupt handlers
13235 and therefore can be used for temporary data without adjusting the stack
13236 pointer. The flag @option{-mno-red-zone} disables this red zone.
13238 @item -mcmodel=small
13239 @opindex mcmodel=small
13240 Generate code for the small code model: the program and its symbols must
13241 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13242 Programs can be statically or dynamically linked. This is the default
13245 @item -mcmodel=kernel
13246 @opindex mcmodel=kernel
13247 Generate code for the kernel code model. The kernel runs in the
13248 negative 2 GB of the address space.
13249 This model has to be used for Linux kernel code.
13251 @item -mcmodel=medium
13252 @opindex mcmodel=medium
13253 Generate code for the medium model: The program is linked in the lower 2
13254 GB of the address space. Small symbols are also placed there. Symbols
13255 with sizes larger than @option{-mlarge-data-threshold} are put into
13256 large data or bss sections and can be located above 2GB. Programs can
13257 be statically or dynamically linked.
13259 @item -mcmodel=large
13260 @opindex mcmodel=large
13261 Generate code for the large model: This model makes no assumptions
13262 about addresses and sizes of sections.
13265 @node i386 and x86-64 Windows Options
13266 @subsection i386 and x86-64 Windows Options
13267 @cindex i386 and x86-64 Windows Options
13269 These additional options are available for Windows targets:
13274 This option is available for Cygwin and MinGW targets. It
13275 specifies that a console application is to be generated, by
13276 instructing the linker to set the PE header subsystem type
13277 required for console applications.
13278 This is the default behavior for Cygwin and MinGW targets.
13282 This option is available for Cygwin and MinGW targets. It
13283 specifies that a DLL - a dynamic link library - is to be
13284 generated, enabling the selection of the required runtime
13285 startup object and entry point.
13287 @item -mnop-fun-dllimport
13288 @opindex mnop-fun-dllimport
13289 This option is available for Cygwin and MinGW targets. It
13290 specifies that the dllimport attribute should be ignored.
13294 This option is available for MinGW targets. It specifies
13295 that MinGW-specific thread support is to be used.
13299 This option is available for mingw-w64 targets. It specifies
13300 that the UNICODE macro is getting pre-defined and that the
13301 unicode capable runtime startup code is chosen.
13305 This option is available for Cygwin and MinGW targets. It
13306 specifies that the typical Windows pre-defined macros are to
13307 be set in the pre-processor, but does not influence the choice
13308 of runtime library/startup code.
13312 This option is available for Cygwin and MinGW targets. It
13313 specifies that a GUI application is to be generated by
13314 instructing the linker to set the PE header subsystem type
13317 @item -fno-set-stack-executable
13318 @opindex fno-set-stack-executable
13319 This option is available for MinGW targets. It specifies that
13320 the executable flag for stack used by nested functions isn't
13321 set. This is necessary for binaries running in kernel mode of
13322 Windows, as there the user32 API, which is used to set executable
13323 privileges, isn't available.
13325 @item -mpe-aligned-commons
13326 @opindex mpe-aligned-commons
13327 This option is available for Cygwin and MinGW targets. It
13328 specifies that the GNU extension to the PE file format that
13329 permits the correct alignment of COMMON variables should be
13330 used when generating code. It will be enabled by default if
13331 GCC detects that the target assembler found during configuration
13332 supports the feature.
13335 See also under @ref{i386 and x86-64 Options} for standard options.
13337 @node IA-64 Options
13338 @subsection IA-64 Options
13339 @cindex IA-64 Options
13341 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13345 @opindex mbig-endian
13346 Generate code for a big endian target. This is the default for HP-UX@.
13348 @item -mlittle-endian
13349 @opindex mlittle-endian
13350 Generate code for a little endian target. This is the default for AIX5
13356 @opindex mno-gnu-as
13357 Generate (or don't) code for the GNU assembler. This is the default.
13358 @c Also, this is the default if the configure option @option{--with-gnu-as}
13364 @opindex mno-gnu-ld
13365 Generate (or don't) code for the GNU linker. This is the default.
13366 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13371 Generate code that does not use a global pointer register. The result
13372 is not position independent code, and violates the IA-64 ABI@.
13374 @item -mvolatile-asm-stop
13375 @itemx -mno-volatile-asm-stop
13376 @opindex mvolatile-asm-stop
13377 @opindex mno-volatile-asm-stop
13378 Generate (or don't) a stop bit immediately before and after volatile asm
13381 @item -mregister-names
13382 @itemx -mno-register-names
13383 @opindex mregister-names
13384 @opindex mno-register-names
13385 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13386 the stacked registers. This may make assembler output more readable.
13392 Disable (or enable) optimizations that use the small data section. This may
13393 be useful for working around optimizer bugs.
13395 @item -mconstant-gp
13396 @opindex mconstant-gp
13397 Generate code that uses a single constant global pointer value. This is
13398 useful when compiling kernel code.
13402 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13403 This is useful when compiling firmware code.
13405 @item -minline-float-divide-min-latency
13406 @opindex minline-float-divide-min-latency
13407 Generate code for inline divides of floating point values
13408 using the minimum latency algorithm.
13410 @item -minline-float-divide-max-throughput
13411 @opindex minline-float-divide-max-throughput
13412 Generate code for inline divides of floating point values
13413 using the maximum throughput algorithm.
13415 @item -mno-inline-float-divide
13416 @opindex mno-inline-float-divide
13417 Do not generate inline code for divides of floating point values.
13419 @item -minline-int-divide-min-latency
13420 @opindex minline-int-divide-min-latency
13421 Generate code for inline divides of integer values
13422 using the minimum latency algorithm.
13424 @item -minline-int-divide-max-throughput
13425 @opindex minline-int-divide-max-throughput
13426 Generate code for inline divides of integer values
13427 using the maximum throughput algorithm.
13429 @item -mno-inline-int-divide
13430 @opindex mno-inline-int-divide
13431 Do not generate inline code for divides of integer values.
13433 @item -minline-sqrt-min-latency
13434 @opindex minline-sqrt-min-latency
13435 Generate code for inline square roots
13436 using the minimum latency algorithm.
13438 @item -minline-sqrt-max-throughput
13439 @opindex minline-sqrt-max-throughput
13440 Generate code for inline square roots
13441 using the maximum throughput algorithm.
13443 @item -mno-inline-sqrt
13444 @opindex mno-inline-sqrt
13445 Do not generate inline code for sqrt.
13448 @itemx -mno-fused-madd
13449 @opindex mfused-madd
13450 @opindex mno-fused-madd
13451 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13452 instructions. The default is to use these instructions.
13454 @item -mno-dwarf2-asm
13455 @itemx -mdwarf2-asm
13456 @opindex mno-dwarf2-asm
13457 @opindex mdwarf2-asm
13458 Don't (or do) generate assembler code for the DWARF2 line number debugging
13459 info. This may be useful when not using the GNU assembler.
13461 @item -mearly-stop-bits
13462 @itemx -mno-early-stop-bits
13463 @opindex mearly-stop-bits
13464 @opindex mno-early-stop-bits
13465 Allow stop bits to be placed earlier than immediately preceding the
13466 instruction that triggered the stop bit. This can improve instruction
13467 scheduling, but does not always do so.
13469 @item -mfixed-range=@var{register-range}
13470 @opindex mfixed-range
13471 Generate code treating the given register range as fixed registers.
13472 A fixed register is one that the register allocator can not use. This is
13473 useful when compiling kernel code. A register range is specified as
13474 two registers separated by a dash. Multiple register ranges can be
13475 specified separated by a comma.
13477 @item -mtls-size=@var{tls-size}
13479 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13482 @item -mtune=@var{cpu-type}
13484 Tune the instruction scheduling for a particular CPU, Valid values are
13485 itanium, itanium1, merced, itanium2, and mckinley.
13491 Generate code for a 32-bit or 64-bit environment.
13492 The 32-bit environment sets int, long and pointer to 32 bits.
13493 The 64-bit environment sets int to 32 bits and long and pointer
13494 to 64 bits. These are HP-UX specific flags.
13496 @item -mno-sched-br-data-spec
13497 @itemx -msched-br-data-spec
13498 @opindex mno-sched-br-data-spec
13499 @opindex msched-br-data-spec
13500 (Dis/En)able data speculative scheduling before reload.
13501 This will result in generation of the ld.a instructions and
13502 the corresponding check instructions (ld.c / chk.a).
13503 The default is 'disable'.
13505 @item -msched-ar-data-spec
13506 @itemx -mno-sched-ar-data-spec
13507 @opindex msched-ar-data-spec
13508 @opindex mno-sched-ar-data-spec
13509 (En/Dis)able data speculative scheduling after reload.
13510 This will result in generation of the ld.a instructions and
13511 the corresponding check instructions (ld.c / chk.a).
13512 The default is 'enable'.
13514 @item -mno-sched-control-spec
13515 @itemx -msched-control-spec
13516 @opindex mno-sched-control-spec
13517 @opindex msched-control-spec
13518 (Dis/En)able control speculative scheduling. This feature is
13519 available only during region scheduling (i.e.@: before reload).
13520 This will result in generation of the ld.s instructions and
13521 the corresponding check instructions chk.s .
13522 The default is 'disable'.
13524 @item -msched-br-in-data-spec
13525 @itemx -mno-sched-br-in-data-spec
13526 @opindex msched-br-in-data-spec
13527 @opindex mno-sched-br-in-data-spec
13528 (En/Dis)able speculative scheduling of the instructions that
13529 are dependent on the data speculative loads before reload.
13530 This is effective only with @option{-msched-br-data-spec} enabled.
13531 The default is 'enable'.
13533 @item -msched-ar-in-data-spec
13534 @itemx -mno-sched-ar-in-data-spec
13535 @opindex msched-ar-in-data-spec
13536 @opindex mno-sched-ar-in-data-spec
13537 (En/Dis)able speculative scheduling of the instructions that
13538 are dependent on the data speculative loads after reload.
13539 This is effective only with @option{-msched-ar-data-spec} enabled.
13540 The default is 'enable'.
13542 @item -msched-in-control-spec
13543 @itemx -mno-sched-in-control-spec
13544 @opindex msched-in-control-spec
13545 @opindex mno-sched-in-control-spec
13546 (En/Dis)able speculative scheduling of the instructions that
13547 are dependent on the control speculative loads.
13548 This is effective only with @option{-msched-control-spec} enabled.
13549 The default is 'enable'.
13551 @item -mno-sched-prefer-non-data-spec-insns
13552 @itemx -msched-prefer-non-data-spec-insns
13553 @opindex mno-sched-prefer-non-data-spec-insns
13554 @opindex msched-prefer-non-data-spec-insns
13555 If enabled, data speculative instructions will be chosen for schedule
13556 only if there are no other choices at the moment. This will make
13557 the use of the data speculation much more conservative.
13558 The default is 'disable'.
13560 @item -mno-sched-prefer-non-control-spec-insns
13561 @itemx -msched-prefer-non-control-spec-insns
13562 @opindex mno-sched-prefer-non-control-spec-insns
13563 @opindex msched-prefer-non-control-spec-insns
13564 If enabled, control speculative instructions will be chosen for schedule
13565 only if there are no other choices at the moment. This will make
13566 the use of the control speculation much more conservative.
13567 The default is 'disable'.
13569 @item -mno-sched-count-spec-in-critical-path
13570 @itemx -msched-count-spec-in-critical-path
13571 @opindex mno-sched-count-spec-in-critical-path
13572 @opindex msched-count-spec-in-critical-path
13573 If enabled, speculative dependencies will be considered during
13574 computation of the instructions priorities. This will make the use of the
13575 speculation a bit more conservative.
13576 The default is 'disable'.
13578 @item -msched-spec-ldc
13579 @opindex msched-spec-ldc
13580 Use a simple data speculation check. This option is on by default.
13582 @item -msched-control-spec-ldc
13583 @opindex msched-spec-ldc
13584 Use a simple check for control speculation. This option is on by default.
13586 @item -msched-stop-bits-after-every-cycle
13587 @opindex msched-stop-bits-after-every-cycle
13588 Place a stop bit after every cycle when scheduling. This option is on
13591 @item -msched-fp-mem-deps-zero-cost
13592 @opindex msched-fp-mem-deps-zero-cost
13593 Assume that floating-point stores and loads are not likely to cause a conflict
13594 when placed into the same instruction group. This option is disabled by
13597 @item -msel-sched-dont-check-control-spec
13598 @opindex msel-sched-dont-check-control-spec
13599 Generate checks for control speculation in selective scheduling.
13600 This flag is disabled by default.
13602 @item -msched-max-memory-insns=@var{max-insns}
13603 @opindex msched-max-memory-insns
13604 Limit on the number of memory insns per instruction group, giving lower
13605 priority to subsequent memory insns attempting to schedule in the same
13606 instruction group. Frequently useful to prevent cache bank conflicts.
13607 The default value is 1.
13609 @item -msched-max-memory-insns-hard-limit
13610 @opindex msched-max-memory-insns-hard-limit
13611 Disallow more than `msched-max-memory-insns' in instruction group.
13612 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13613 when limit is reached but may still schedule memory operations.
13617 @node IA-64/VMS Options
13618 @subsection IA-64/VMS Options
13620 These @samp{-m} options are defined for the IA-64/VMS implementations:
13623 @item -mvms-return-codes
13624 @opindex mvms-return-codes
13625 Return VMS condition codes from main. The default is to return POSIX
13626 style condition (e.g.@ error) codes.
13628 @item -mdebug-main=@var{prefix}
13629 @opindex mdebug-main=@var{prefix}
13630 Flag the first routine whose name starts with @var{prefix} as the main
13631 routine for the debugger.
13635 Default to 64bit memory allocation routines.
13639 @subsection LM32 Options
13640 @cindex LM32 options
13642 These @option{-m} options are defined for the Lattice Mico32 architecture:
13645 @item -mbarrel-shift-enabled
13646 @opindex mbarrel-shift-enabled
13647 Enable barrel-shift instructions.
13649 @item -mdivide-enabled
13650 @opindex mdivide-enabled
13651 Enable divide and modulus instructions.
13653 @item -mmultiply-enabled
13654 @opindex multiply-enabled
13655 Enable multiply instructions.
13657 @item -msign-extend-enabled
13658 @opindex msign-extend-enabled
13659 Enable sign extend instructions.
13661 @item -muser-enabled
13662 @opindex muser-enabled
13663 Enable user-defined instructions.
13668 @subsection M32C Options
13669 @cindex M32C options
13672 @item -mcpu=@var{name}
13674 Select the CPU for which code is generated. @var{name} may be one of
13675 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13676 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13677 the M32C/80 series.
13681 Specifies that the program will be run on the simulator. This causes
13682 an alternate runtime library to be linked in which supports, for
13683 example, file I/O@. You must not use this option when generating
13684 programs that will run on real hardware; you must provide your own
13685 runtime library for whatever I/O functions are needed.
13687 @item -memregs=@var{number}
13689 Specifies the number of memory-based pseudo-registers GCC will use
13690 during code generation. These pseudo-registers will be used like real
13691 registers, so there is a tradeoff between GCC's ability to fit the
13692 code into available registers, and the performance penalty of using
13693 memory instead of registers. Note that all modules in a program must
13694 be compiled with the same value for this option. Because of that, you
13695 must not use this option with the default runtime libraries gcc
13700 @node M32R/D Options
13701 @subsection M32R/D Options
13702 @cindex M32R/D options
13704 These @option{-m} options are defined for Renesas M32R/D architectures:
13709 Generate code for the M32R/2@.
13713 Generate code for the M32R/X@.
13717 Generate code for the M32R@. This is the default.
13719 @item -mmodel=small
13720 @opindex mmodel=small
13721 Assume all objects live in the lower 16MB of memory (so that their addresses
13722 can be loaded with the @code{ld24} instruction), and assume all subroutines
13723 are reachable with the @code{bl} instruction.
13724 This is the default.
13726 The addressability of a particular object can be set with the
13727 @code{model} attribute.
13729 @item -mmodel=medium
13730 @opindex mmodel=medium
13731 Assume objects may be anywhere in the 32-bit address space (the compiler
13732 will generate @code{seth/add3} instructions to load their addresses), and
13733 assume all subroutines are reachable with the @code{bl} instruction.
13735 @item -mmodel=large
13736 @opindex mmodel=large
13737 Assume objects may be anywhere in the 32-bit address space (the compiler
13738 will generate @code{seth/add3} instructions to load their addresses), and
13739 assume subroutines may not be reachable with the @code{bl} instruction
13740 (the compiler will generate the much slower @code{seth/add3/jl}
13741 instruction sequence).
13744 @opindex msdata=none
13745 Disable use of the small data area. Variables will be put into
13746 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13747 @code{section} attribute has been specified).
13748 This is the default.
13750 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13751 Objects may be explicitly put in the small data area with the
13752 @code{section} attribute using one of these sections.
13754 @item -msdata=sdata
13755 @opindex msdata=sdata
13756 Put small global and static data in the small data area, but do not
13757 generate special code to reference them.
13760 @opindex msdata=use
13761 Put small global and static data in the small data area, and generate
13762 special instructions to reference them.
13766 @cindex smaller data references
13767 Put global and static objects less than or equal to @var{num} bytes
13768 into the small data or bss sections instead of the normal data or bss
13769 sections. The default value of @var{num} is 8.
13770 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13771 for this option to have any effect.
13773 All modules should be compiled with the same @option{-G @var{num}} value.
13774 Compiling with different values of @var{num} may or may not work; if it
13775 doesn't the linker will give an error message---incorrect code will not be
13780 Makes the M32R specific code in the compiler display some statistics
13781 that might help in debugging programs.
13783 @item -malign-loops
13784 @opindex malign-loops
13785 Align all loops to a 32-byte boundary.
13787 @item -mno-align-loops
13788 @opindex mno-align-loops
13789 Do not enforce a 32-byte alignment for loops. This is the default.
13791 @item -missue-rate=@var{number}
13792 @opindex missue-rate=@var{number}
13793 Issue @var{number} instructions per cycle. @var{number} can only be 1
13796 @item -mbranch-cost=@var{number}
13797 @opindex mbranch-cost=@var{number}
13798 @var{number} can only be 1 or 2. If it is 1 then branches will be
13799 preferred over conditional code, if it is 2, then the opposite will
13802 @item -mflush-trap=@var{number}
13803 @opindex mflush-trap=@var{number}
13804 Specifies the trap number to use to flush the cache. The default is
13805 12. Valid numbers are between 0 and 15 inclusive.
13807 @item -mno-flush-trap
13808 @opindex mno-flush-trap
13809 Specifies that the cache cannot be flushed by using a trap.
13811 @item -mflush-func=@var{name}
13812 @opindex mflush-func=@var{name}
13813 Specifies the name of the operating system function to call to flush
13814 the cache. The default is @emph{_flush_cache}, but a function call
13815 will only be used if a trap is not available.
13817 @item -mno-flush-func
13818 @opindex mno-flush-func
13819 Indicates that there is no OS function for flushing the cache.
13823 @node M680x0 Options
13824 @subsection M680x0 Options
13825 @cindex M680x0 options
13827 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13828 The default settings depend on which architecture was selected when
13829 the compiler was configured; the defaults for the most common choices
13833 @item -march=@var{arch}
13835 Generate code for a specific M680x0 or ColdFire instruction set
13836 architecture. Permissible values of @var{arch} for M680x0
13837 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13838 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13839 architectures are selected according to Freescale's ISA classification
13840 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13841 @samp{isab} and @samp{isac}.
13843 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13844 code for a ColdFire target. The @var{arch} in this macro is one of the
13845 @option{-march} arguments given above.
13847 When used together, @option{-march} and @option{-mtune} select code
13848 that runs on a family of similar processors but that is optimized
13849 for a particular microarchitecture.
13851 @item -mcpu=@var{cpu}
13853 Generate code for a specific M680x0 or ColdFire processor.
13854 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13855 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13856 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13857 below, which also classifies the CPUs into families:
13859 @multitable @columnfractions 0.20 0.80
13860 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13861 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13862 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13863 @item @samp{5206e} @tab @samp{5206e}
13864 @item @samp{5208} @tab @samp{5207} @samp{5208}
13865 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13866 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13867 @item @samp{5216} @tab @samp{5214} @samp{5216}
13868 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13869 @item @samp{5225} @tab @samp{5224} @samp{5225}
13870 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13871 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13872 @item @samp{5249} @tab @samp{5249}
13873 @item @samp{5250} @tab @samp{5250}
13874 @item @samp{5271} @tab @samp{5270} @samp{5271}
13875 @item @samp{5272} @tab @samp{5272}
13876 @item @samp{5275} @tab @samp{5274} @samp{5275}
13877 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13878 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13879 @item @samp{5307} @tab @samp{5307}
13880 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13881 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13882 @item @samp{5407} @tab @samp{5407}
13883 @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}
13886 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13887 @var{arch} is compatible with @var{cpu}. Other combinations of
13888 @option{-mcpu} and @option{-march} are rejected.
13890 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13891 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13892 where the value of @var{family} is given by the table above.
13894 @item -mtune=@var{tune}
13896 Tune the code for a particular microarchitecture, within the
13897 constraints set by @option{-march} and @option{-mcpu}.
13898 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13899 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13900 and @samp{cpu32}. The ColdFire microarchitectures
13901 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13903 You can also use @option{-mtune=68020-40} for code that needs
13904 to run relatively well on 68020, 68030 and 68040 targets.
13905 @option{-mtune=68020-60} is similar but includes 68060 targets
13906 as well. These two options select the same tuning decisions as
13907 @option{-m68020-40} and @option{-m68020-60} respectively.
13909 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13910 when tuning for 680x0 architecture @var{arch}. It also defines
13911 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13912 option is used. If gcc is tuning for a range of architectures,
13913 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13914 it defines the macros for every architecture in the range.
13916 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13917 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13918 of the arguments given above.
13924 Generate output for a 68000. This is the default
13925 when the compiler is configured for 68000-based systems.
13926 It is equivalent to @option{-march=68000}.
13928 Use this option for microcontrollers with a 68000 or EC000 core,
13929 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13933 Generate output for a 68010. This is the default
13934 when the compiler is configured for 68010-based systems.
13935 It is equivalent to @option{-march=68010}.
13941 Generate output for a 68020. This is the default
13942 when the compiler is configured for 68020-based systems.
13943 It is equivalent to @option{-march=68020}.
13947 Generate output for a 68030. This is the default when the compiler is
13948 configured for 68030-based systems. It is equivalent to
13949 @option{-march=68030}.
13953 Generate output for a 68040. This is the default when the compiler is
13954 configured for 68040-based systems. It is equivalent to
13955 @option{-march=68040}.
13957 This option inhibits the use of 68881/68882 instructions that have to be
13958 emulated by software on the 68040. Use this option if your 68040 does not
13959 have code to emulate those instructions.
13963 Generate output for a 68060. This is the default when the compiler is
13964 configured for 68060-based systems. It is equivalent to
13965 @option{-march=68060}.
13967 This option inhibits the use of 68020 and 68881/68882 instructions that
13968 have to be emulated by software on the 68060. Use this option if your 68060
13969 does not have code to emulate those instructions.
13973 Generate output for a CPU32. This is the default
13974 when the compiler is configured for CPU32-based systems.
13975 It is equivalent to @option{-march=cpu32}.
13977 Use this option for microcontrollers with a
13978 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13979 68336, 68340, 68341, 68349 and 68360.
13983 Generate output for a 520X ColdFire CPU@. This is the default
13984 when the compiler is configured for 520X-based systems.
13985 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13986 in favor of that option.
13988 Use this option for microcontroller with a 5200 core, including
13989 the MCF5202, MCF5203, MCF5204 and MCF5206.
13993 Generate output for a 5206e ColdFire CPU@. The option is now
13994 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13998 Generate output for a member of the ColdFire 528X family.
13999 The option is now deprecated in favor of the equivalent
14000 @option{-mcpu=528x}.
14004 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
14005 in favor of the equivalent @option{-mcpu=5307}.
14009 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
14010 in favor of the equivalent @option{-mcpu=5407}.
14014 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
14015 This includes use of hardware floating point instructions.
14016 The option is equivalent to @option{-mcpu=547x}, and is now
14017 deprecated in favor of that option.
14021 Generate output for a 68040, without using any of the new instructions.
14022 This results in code which can run relatively efficiently on either a
14023 68020/68881 or a 68030 or a 68040. The generated code does use the
14024 68881 instructions that are emulated on the 68040.
14026 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
14030 Generate output for a 68060, without using any of the new instructions.
14031 This results in code which can run relatively efficiently on either a
14032 68020/68881 or a 68030 or a 68040. The generated code does use the
14033 68881 instructions that are emulated on the 68060.
14035 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
14039 @opindex mhard-float
14041 Generate floating-point instructions. This is the default for 68020
14042 and above, and for ColdFire devices that have an FPU@. It defines the
14043 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
14044 on ColdFire targets.
14047 @opindex msoft-float
14048 Do not generate floating-point instructions; use library calls instead.
14049 This is the default for 68000, 68010, and 68832 targets. It is also
14050 the default for ColdFire devices that have no FPU.
14056 Generate (do not generate) ColdFire hardware divide and remainder
14057 instructions. If @option{-march} is used without @option{-mcpu},
14058 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
14059 architectures. Otherwise, the default is taken from the target CPU
14060 (either the default CPU, or the one specified by @option{-mcpu}). For
14061 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
14062 @option{-mcpu=5206e}.
14064 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
14068 Consider type @code{int} to be 16 bits wide, like @code{short int}.
14069 Additionally, parameters passed on the stack are also aligned to a
14070 16-bit boundary even on targets whose API mandates promotion to 32-bit.
14074 Do not consider type @code{int} to be 16 bits wide. This is the default.
14077 @itemx -mno-bitfield
14078 @opindex mnobitfield
14079 @opindex mno-bitfield
14080 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
14081 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
14085 Do use the bit-field instructions. The @option{-m68020} option implies
14086 @option{-mbitfield}. This is the default if you use a configuration
14087 designed for a 68020.
14091 Use a different function-calling convention, in which functions
14092 that take a fixed number of arguments return with the @code{rtd}
14093 instruction, which pops their arguments while returning. This
14094 saves one instruction in the caller since there is no need to pop
14095 the arguments there.
14097 This calling convention is incompatible with the one normally
14098 used on Unix, so you cannot use it if you need to call libraries
14099 compiled with the Unix compiler.
14101 Also, you must provide function prototypes for all functions that
14102 take variable numbers of arguments (including @code{printf});
14103 otherwise incorrect code will be generated for calls to those
14106 In addition, seriously incorrect code will result if you call a
14107 function with too many arguments. (Normally, extra arguments are
14108 harmlessly ignored.)
14110 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
14111 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
14115 Do not use the calling conventions selected by @option{-mrtd}.
14116 This is the default.
14119 @itemx -mno-align-int
14120 @opindex malign-int
14121 @opindex mno-align-int
14122 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
14123 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
14124 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
14125 Aligning variables on 32-bit boundaries produces code that runs somewhat
14126 faster on processors with 32-bit busses at the expense of more memory.
14128 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
14129 align structures containing the above types differently than
14130 most published application binary interface specifications for the m68k.
14134 Use the pc-relative addressing mode of the 68000 directly, instead of
14135 using a global offset table. At present, this option implies @option{-fpic},
14136 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
14137 not presently supported with @option{-mpcrel}, though this could be supported for
14138 68020 and higher processors.
14140 @item -mno-strict-align
14141 @itemx -mstrict-align
14142 @opindex mno-strict-align
14143 @opindex mstrict-align
14144 Do not (do) assume that unaligned memory references will be handled by
14148 Generate code that allows the data segment to be located in a different
14149 area of memory from the text segment. This allows for execute in place in
14150 an environment without virtual memory management. This option implies
14153 @item -mno-sep-data
14154 Generate code that assumes that the data segment follows the text segment.
14155 This is the default.
14157 @item -mid-shared-library
14158 Generate code that supports shared libraries via the library ID method.
14159 This allows for execute in place and shared libraries in an environment
14160 without virtual memory management. This option implies @option{-fPIC}.
14162 @item -mno-id-shared-library
14163 Generate code that doesn't assume ID based shared libraries are being used.
14164 This is the default.
14166 @item -mshared-library-id=n
14167 Specified the identification number of the ID based shared library being
14168 compiled. Specifying a value of 0 will generate more compact code, specifying
14169 other values will force the allocation of that number to the current
14170 library but is no more space or time efficient than omitting this option.
14176 When generating position-independent code for ColdFire, generate code
14177 that works if the GOT has more than 8192 entries. This code is
14178 larger and slower than code generated without this option. On M680x0
14179 processors, this option is not needed; @option{-fPIC} suffices.
14181 GCC normally uses a single instruction to load values from the GOT@.
14182 While this is relatively efficient, it only works if the GOT
14183 is smaller than about 64k. Anything larger causes the linker
14184 to report an error such as:
14186 @cindex relocation truncated to fit (ColdFire)
14188 relocation truncated to fit: R_68K_GOT16O foobar
14191 If this happens, you should recompile your code with @option{-mxgot}.
14192 It should then work with very large GOTs. However, code generated with
14193 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
14194 the value of a global symbol.
14196 Note that some linkers, including newer versions of the GNU linker,
14197 can create multiple GOTs and sort GOT entries. If you have such a linker,
14198 you should only need to use @option{-mxgot} when compiling a single
14199 object file that accesses more than 8192 GOT entries. Very few do.
14201 These options have no effect unless GCC is generating
14202 position-independent code.
14206 @node MCore Options
14207 @subsection MCore Options
14208 @cindex MCore options
14210 These are the @samp{-m} options defined for the Motorola M*Core
14216 @itemx -mno-hardlit
14218 @opindex mno-hardlit
14219 Inline constants into the code stream if it can be done in two
14220 instructions or less.
14226 Use the divide instruction. (Enabled by default).
14228 @item -mrelax-immediate
14229 @itemx -mno-relax-immediate
14230 @opindex mrelax-immediate
14231 @opindex mno-relax-immediate
14232 Allow arbitrary sized immediates in bit operations.
14234 @item -mwide-bitfields
14235 @itemx -mno-wide-bitfields
14236 @opindex mwide-bitfields
14237 @opindex mno-wide-bitfields
14238 Always treat bit-fields as int-sized.
14240 @item -m4byte-functions
14241 @itemx -mno-4byte-functions
14242 @opindex m4byte-functions
14243 @opindex mno-4byte-functions
14244 Force all functions to be aligned to a four byte boundary.
14246 @item -mcallgraph-data
14247 @itemx -mno-callgraph-data
14248 @opindex mcallgraph-data
14249 @opindex mno-callgraph-data
14250 Emit callgraph information.
14253 @itemx -mno-slow-bytes
14254 @opindex mslow-bytes
14255 @opindex mno-slow-bytes
14256 Prefer word access when reading byte quantities.
14258 @item -mlittle-endian
14259 @itemx -mbig-endian
14260 @opindex mlittle-endian
14261 @opindex mbig-endian
14262 Generate code for a little endian target.
14268 Generate code for the 210 processor.
14272 Assume that run-time support has been provided and so omit the
14273 simulator library (@file{libsim.a)} from the linker command line.
14275 @item -mstack-increment=@var{size}
14276 @opindex mstack-increment
14277 Set the maximum amount for a single stack increment operation. Large
14278 values can increase the speed of programs which contain functions
14279 that need a large amount of stack space, but they can also trigger a
14280 segmentation fault if the stack is extended too much. The default
14286 @subsection MeP Options
14287 @cindex MeP options
14293 Enables the @code{abs} instruction, which is the absolute difference
14294 between two registers.
14298 Enables all the optional instructions - average, multiply, divide, bit
14299 operations, leading zero, absolute difference, min/max, clip, and
14305 Enables the @code{ave} instruction, which computes the average of two
14308 @item -mbased=@var{n}
14310 Variables of size @var{n} bytes or smaller will be placed in the
14311 @code{.based} section by default. Based variables use the @code{$tp}
14312 register as a base register, and there is a 128 byte limit to the
14313 @code{.based} section.
14317 Enables the bit operation instructions - bit test (@code{btstm}), set
14318 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14319 test-and-set (@code{tas}).
14321 @item -mc=@var{name}
14323 Selects which section constant data will be placed in. @var{name} may
14324 be @code{tiny}, @code{near}, or @code{far}.
14328 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14329 useful unless you also provide @code{-mminmax}.
14331 @item -mconfig=@var{name}
14333 Selects one of the build-in core configurations. Each MeP chip has
14334 one or more modules in it; each module has a core CPU and a variety of
14335 coprocessors, optional instructions, and peripherals. The
14336 @code{MeP-Integrator} tool, not part of GCC, provides these
14337 configurations through this option; using this option is the same as
14338 using all the corresponding command line options. The default
14339 configuration is @code{default}.
14343 Enables the coprocessor instructions. By default, this is a 32-bit
14344 coprocessor. Note that the coprocessor is normally enabled via the
14345 @code{-mconfig=} option.
14349 Enables the 32-bit coprocessor's instructions.
14353 Enables the 64-bit coprocessor's instructions.
14357 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14361 Causes constant variables to be placed in the @code{.near} section.
14365 Enables the @code{div} and @code{divu} instructions.
14369 Generate big-endian code.
14373 Generate little-endian code.
14375 @item -mio-volatile
14376 @opindex mio-volatile
14377 Tells the compiler that any variable marked with the @code{io}
14378 attribute is to be considered volatile.
14382 Causes variables to be assigned to the @code{.far} section by default.
14386 Enables the @code{leadz} (leading zero) instruction.
14390 Causes variables to be assigned to the @code{.near} section by default.
14394 Enables the @code{min} and @code{max} instructions.
14398 Enables the multiplication and multiply-accumulate instructions.
14402 Disables all the optional instructions enabled by @code{-mall-opts}.
14406 Enables the @code{repeat} and @code{erepeat} instructions, used for
14407 low-overhead looping.
14411 Causes all variables to default to the @code{.tiny} section. Note
14412 that there is a 65536 byte limit to this section. Accesses to these
14413 variables use the @code{%gp} base register.
14417 Enables the saturation instructions. Note that the compiler does not
14418 currently generate these itself, but this option is included for
14419 compatibility with other tools, like @code{as}.
14423 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14427 Link the simulator runtime libraries.
14431 Link the simulator runtime libraries, excluding built-in support
14432 for reset and exception vectors and tables.
14436 Causes all functions to default to the @code{.far} section. Without
14437 this option, functions default to the @code{.near} section.
14439 @item -mtiny=@var{n}
14441 Variables that are @var{n} bytes or smaller will be allocated to the
14442 @code{.tiny} section. These variables use the @code{$gp} base
14443 register. The default for this option is 4, but note that there's a
14444 65536 byte limit to the @code{.tiny} section.
14448 @node MicroBlaze Options
14449 @subsection MicroBlaze Options
14450 @cindex MicroBlaze Options
14455 @opindex msoft-float
14456 Use software emulation for floating point (default).
14459 @opindex mhard-float
14460 Use hardware floating point instructions.
14464 Do not optimize block moves, use @code{memcpy}.
14466 @item -mno-clearbss
14467 @opindex mno-clearbss
14468 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14470 @item -mcpu=@var{cpu-type}
14472 Use features of and schedule code for given CPU.
14473 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14474 where @var{X} is a major version, @var{YY} is the minor version, and
14475 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14476 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14478 @item -mxl-soft-mul
14479 @opindex mxl-soft-mul
14480 Use software multiply emulation (default).
14482 @item -mxl-soft-div
14483 @opindex mxl-soft-div
14484 Use software emulation for divides (default).
14486 @item -mxl-barrel-shift
14487 @opindex mxl-barrel-shift
14488 Use the hardware barrel shifter.
14490 @item -mxl-pattern-compare
14491 @opindex mxl-pattern-compare
14492 Use pattern compare instructions.
14494 @item -msmall-divides
14495 @opindex msmall-divides
14496 Use table lookup optimization for small signed integer divisions.
14498 @item -mxl-stack-check
14499 @opindex mxl-stack-check
14500 This option is deprecated. Use -fstack-check instead.
14503 @opindex mxl-gp-opt
14504 Use GP relative sdata/sbss sections.
14506 @item -mxl-multiply-high
14507 @opindex mxl-multiply-high
14508 Use multiply high instructions for high part of 32x32 multiply.
14510 @item -mxl-float-convert
14511 @opindex mxl-float-convert
14512 Use hardware floating point conversion instructions.
14514 @item -mxl-float-sqrt
14515 @opindex mxl-float-sqrt
14516 Use hardware floating point square root instruction.
14518 @item -mxl-mode-@var{app-model}
14519 Select application model @var{app-model}. Valid models are
14522 normal executable (default), uses startup code @file{crt0.o}.
14525 for use with Xilinx Microprocessor Debugger (XMD) based
14526 software intrusive debug agent called xmdstub. This uses startup file
14527 @file{crt1.o} and sets the start address of the program to be 0x800.
14530 for applications that are loaded using a bootloader.
14531 This model uses startup file @file{crt2.o} which does not contain a processor
14532 reset vector handler. This is suitable for transferring control on a
14533 processor reset to the bootloader rather than the application.
14536 for applications that do not require any of the
14537 MicroBlaze vectors. This option may be useful for applications running
14538 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14541 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14542 @option{-mxl-mode-@var{app-model}}.
14547 @subsection MIPS Options
14548 @cindex MIPS options
14554 Generate big-endian code.
14558 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14561 @item -march=@var{arch}
14563 Generate code that will run on @var{arch}, which can be the name of a
14564 generic MIPS ISA, or the name of a particular processor.
14566 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14567 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14568 The processor names are:
14569 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14570 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14571 @samp{5kc}, @samp{5kf},
14573 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14574 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14575 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14576 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14577 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14578 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14582 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14583 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14584 @samp{rm7000}, @samp{rm9000},
14585 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14588 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14589 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14591 The special value @samp{from-abi} selects the
14592 most compatible architecture for the selected ABI (that is,
14593 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14595 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
14596 which selects the best architecture option for the host processor.
14597 @option{-march=native} has no effect if GCC does not recognize
14600 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14601 (for example, @samp{-march=r2k}). Prefixes are optional, and
14602 @samp{vr} may be written @samp{r}.
14604 Names of the form @samp{@var{n}f2_1} refer to processors with
14605 FPUs clocked at half the rate of the core, names of the form
14606 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14607 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14608 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14609 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14610 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14611 accepted as synonyms for @samp{@var{n}f1_1}.
14613 GCC defines two macros based on the value of this option. The first
14614 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14615 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14616 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14617 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14618 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14620 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14621 above. In other words, it will have the full prefix and will not
14622 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14623 the macro names the resolved architecture (either @samp{"mips1"} or
14624 @samp{"mips3"}). It names the default architecture when no
14625 @option{-march} option is given.
14627 @item -mtune=@var{arch}
14629 Optimize for @var{arch}. Among other things, this option controls
14630 the way instructions are scheduled, and the perceived cost of arithmetic
14631 operations. The list of @var{arch} values is the same as for
14634 When this option is not used, GCC will optimize for the processor
14635 specified by @option{-march}. By using @option{-march} and
14636 @option{-mtune} together, it is possible to generate code that will
14637 run on a family of processors, but optimize the code for one
14638 particular member of that family.
14640 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14641 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14642 @samp{-march} ones described above.
14646 Equivalent to @samp{-march=mips1}.
14650 Equivalent to @samp{-march=mips2}.
14654 Equivalent to @samp{-march=mips3}.
14658 Equivalent to @samp{-march=mips4}.
14662 Equivalent to @samp{-march=mips32}.
14666 Equivalent to @samp{-march=mips32r2}.
14670 Equivalent to @samp{-march=mips64}.
14674 Equivalent to @samp{-march=mips64r2}.
14679 @opindex mno-mips16
14680 Generate (do not generate) MIPS16 code. If GCC is targetting a
14681 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14683 MIPS16 code generation can also be controlled on a per-function basis
14684 by means of @code{mips16} and @code{nomips16} attributes.
14685 @xref{Function Attributes}, for more information.
14687 @item -mflip-mips16
14688 @opindex mflip-mips16
14689 Generate MIPS16 code on alternating functions. This option is provided
14690 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14691 not intended for ordinary use in compiling user code.
14693 @item -minterlink-mips16
14694 @itemx -mno-interlink-mips16
14695 @opindex minterlink-mips16
14696 @opindex mno-interlink-mips16
14697 Require (do not require) that non-MIPS16 code be link-compatible with
14700 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14701 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14702 therefore disables direct jumps unless GCC knows that the target of the
14703 jump is not MIPS16.
14715 Generate code for the given ABI@.
14717 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14718 generates 64-bit code when you select a 64-bit architecture, but you
14719 can use @option{-mgp32} to get 32-bit code instead.
14721 For information about the O64 ABI, see
14722 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14724 GCC supports a variant of the o32 ABI in which floating-point registers
14725 are 64 rather than 32 bits wide. You can select this combination with
14726 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14727 and @samp{mfhc1} instructions and is therefore only supported for
14728 MIPS32R2 processors.
14730 The register assignments for arguments and return values remain the
14731 same, but each scalar value is passed in a single 64-bit register
14732 rather than a pair of 32-bit registers. For example, scalar
14733 floating-point values are returned in @samp{$f0} only, not a
14734 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14735 remains the same, but all 64 bits are saved.
14738 @itemx -mno-abicalls
14740 @opindex mno-abicalls
14741 Generate (do not generate) code that is suitable for SVR4-style
14742 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14747 Generate (do not generate) code that is fully position-independent,
14748 and that can therefore be linked into shared libraries. This option
14749 only affects @option{-mabicalls}.
14751 All @option{-mabicalls} code has traditionally been position-independent,
14752 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14753 as an extension, the GNU toolchain allows executables to use absolute
14754 accesses for locally-binding symbols. It can also use shorter GP
14755 initialization sequences and generate direct calls to locally-defined
14756 functions. This mode is selected by @option{-mno-shared}.
14758 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14759 objects that can only be linked by the GNU linker. However, the option
14760 does not affect the ABI of the final executable; it only affects the ABI
14761 of relocatable objects. Using @option{-mno-shared} will generally make
14762 executables both smaller and quicker.
14764 @option{-mshared} is the default.
14770 Assume (do not assume) that the static and dynamic linkers
14771 support PLTs and copy relocations. This option only affects
14772 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14773 has no effect without @samp{-msym32}.
14775 You can make @option{-mplt} the default by configuring
14776 GCC with @option{--with-mips-plt}. The default is
14777 @option{-mno-plt} otherwise.
14783 Lift (do not lift) the usual restrictions on the size of the global
14786 GCC normally uses a single instruction to load values from the GOT@.
14787 While this is relatively efficient, it will only work if the GOT
14788 is smaller than about 64k. Anything larger will cause the linker
14789 to report an error such as:
14791 @cindex relocation truncated to fit (MIPS)
14793 relocation truncated to fit: R_MIPS_GOT16 foobar
14796 If this happens, you should recompile your code with @option{-mxgot}.
14797 It should then work with very large GOTs, although it will also be
14798 less efficient, since it will take three instructions to fetch the
14799 value of a global symbol.
14801 Note that some linkers can create multiple GOTs. If you have such a
14802 linker, you should only need to use @option{-mxgot} when a single object
14803 file accesses more than 64k's worth of GOT entries. Very few do.
14805 These options have no effect unless GCC is generating position
14810 Assume that general-purpose registers are 32 bits wide.
14814 Assume that general-purpose registers are 64 bits wide.
14818 Assume that floating-point registers are 32 bits wide.
14822 Assume that floating-point registers are 64 bits wide.
14825 @opindex mhard-float
14826 Use floating-point coprocessor instructions.
14829 @opindex msoft-float
14830 Do not use floating-point coprocessor instructions. Implement
14831 floating-point calculations using library calls instead.
14833 @item -msingle-float
14834 @opindex msingle-float
14835 Assume that the floating-point coprocessor only supports single-precision
14838 @item -mdouble-float
14839 @opindex mdouble-float
14840 Assume that the floating-point coprocessor supports double-precision
14841 operations. This is the default.
14847 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14848 implement atomic memory built-in functions. When neither option is
14849 specified, GCC will use the instructions if the target architecture
14852 @option{-mllsc} is useful if the runtime environment can emulate the
14853 instructions and @option{-mno-llsc} can be useful when compiling for
14854 nonstandard ISAs. You can make either option the default by
14855 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14856 respectively. @option{--with-llsc} is the default for some
14857 configurations; see the installation documentation for details.
14863 Use (do not use) revision 1 of the MIPS DSP ASE@.
14864 @xref{MIPS DSP Built-in Functions}. This option defines the
14865 preprocessor macro @samp{__mips_dsp}. It also defines
14866 @samp{__mips_dsp_rev} to 1.
14872 Use (do not use) revision 2 of the MIPS DSP ASE@.
14873 @xref{MIPS DSP Built-in Functions}. This option defines the
14874 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14875 It also defines @samp{__mips_dsp_rev} to 2.
14878 @itemx -mno-smartmips
14879 @opindex msmartmips
14880 @opindex mno-smartmips
14881 Use (do not use) the MIPS SmartMIPS ASE.
14883 @item -mpaired-single
14884 @itemx -mno-paired-single
14885 @opindex mpaired-single
14886 @opindex mno-paired-single
14887 Use (do not use) paired-single floating-point instructions.
14888 @xref{MIPS Paired-Single Support}. This option requires
14889 hardware floating-point support to be enabled.
14895 Use (do not use) MIPS Digital Media Extension instructions.
14896 This option can only be used when generating 64-bit code and requires
14897 hardware floating-point support to be enabled.
14902 @opindex mno-mips3d
14903 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14904 The option @option{-mips3d} implies @option{-mpaired-single}.
14910 Use (do not use) MT Multithreading instructions.
14914 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14915 an explanation of the default and the way that the pointer size is
14920 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14922 The default size of @code{int}s, @code{long}s and pointers depends on
14923 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14924 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14925 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14926 or the same size as integer registers, whichever is smaller.
14932 Assume (do not assume) that all symbols have 32-bit values, regardless
14933 of the selected ABI@. This option is useful in combination with
14934 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14935 to generate shorter and faster references to symbolic addresses.
14939 Put definitions of externally-visible data in a small data section
14940 if that data is no bigger than @var{num} bytes. GCC can then access
14941 the data more efficiently; see @option{-mgpopt} for details.
14943 The default @option{-G} option depends on the configuration.
14945 @item -mlocal-sdata
14946 @itemx -mno-local-sdata
14947 @opindex mlocal-sdata
14948 @opindex mno-local-sdata
14949 Extend (do not extend) the @option{-G} behavior to local data too,
14950 such as to static variables in C@. @option{-mlocal-sdata} is the
14951 default for all configurations.
14953 If the linker complains that an application is using too much small data,
14954 you might want to try rebuilding the less performance-critical parts with
14955 @option{-mno-local-sdata}. You might also want to build large
14956 libraries with @option{-mno-local-sdata}, so that the libraries leave
14957 more room for the main program.
14959 @item -mextern-sdata
14960 @itemx -mno-extern-sdata
14961 @opindex mextern-sdata
14962 @opindex mno-extern-sdata
14963 Assume (do not assume) that externally-defined data will be in
14964 a small data section if that data is within the @option{-G} limit.
14965 @option{-mextern-sdata} is the default for all configurations.
14967 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14968 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14969 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14970 is placed in a small data section. If @var{Var} is defined by another
14971 module, you must either compile that module with a high-enough
14972 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14973 definition. If @var{Var} is common, you must link the application
14974 with a high-enough @option{-G} setting.
14976 The easiest way of satisfying these restrictions is to compile
14977 and link every module with the same @option{-G} option. However,
14978 you may wish to build a library that supports several different
14979 small data limits. You can do this by compiling the library with
14980 the highest supported @option{-G} setting and additionally using
14981 @option{-mno-extern-sdata} to stop the library from making assumptions
14982 about externally-defined data.
14988 Use (do not use) GP-relative accesses for symbols that are known to be
14989 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14990 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14993 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14994 might not hold the value of @code{_gp}. For example, if the code is
14995 part of a library that might be used in a boot monitor, programs that
14996 call boot monitor routines will pass an unknown value in @code{$gp}.
14997 (In such situations, the boot monitor itself would usually be compiled
14998 with @option{-G0}.)
15000 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
15001 @option{-mno-extern-sdata}.
15003 @item -membedded-data
15004 @itemx -mno-embedded-data
15005 @opindex membedded-data
15006 @opindex mno-embedded-data
15007 Allocate variables to the read-only data section first if possible, then
15008 next in the small data section if possible, otherwise in data. This gives
15009 slightly slower code than the default, but reduces the amount of RAM required
15010 when executing, and thus may be preferred for some embedded systems.
15012 @item -muninit-const-in-rodata
15013 @itemx -mno-uninit-const-in-rodata
15014 @opindex muninit-const-in-rodata
15015 @opindex mno-uninit-const-in-rodata
15016 Put uninitialized @code{const} variables in the read-only data section.
15017 This option is only meaningful in conjunction with @option{-membedded-data}.
15019 @item -mcode-readable=@var{setting}
15020 @opindex mcode-readable
15021 Specify whether GCC may generate code that reads from executable sections.
15022 There are three possible settings:
15025 @item -mcode-readable=yes
15026 Instructions may freely access executable sections. This is the
15029 @item -mcode-readable=pcrel
15030 MIPS16 PC-relative load instructions can access executable sections,
15031 but other instructions must not do so. This option is useful on 4KSc
15032 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
15033 It is also useful on processors that can be configured to have a dual
15034 instruction/data SRAM interface and that, like the M4K, automatically
15035 redirect PC-relative loads to the instruction RAM.
15037 @item -mcode-readable=no
15038 Instructions must not access executable sections. This option can be
15039 useful on targets that are configured to have a dual instruction/data
15040 SRAM interface but that (unlike the M4K) do not automatically redirect
15041 PC-relative loads to the instruction RAM.
15044 @item -msplit-addresses
15045 @itemx -mno-split-addresses
15046 @opindex msplit-addresses
15047 @opindex mno-split-addresses
15048 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
15049 relocation operators. This option has been superseded by
15050 @option{-mexplicit-relocs} but is retained for backwards compatibility.
15052 @item -mexplicit-relocs
15053 @itemx -mno-explicit-relocs
15054 @opindex mexplicit-relocs
15055 @opindex mno-explicit-relocs
15056 Use (do not use) assembler relocation operators when dealing with symbolic
15057 addresses. The alternative, selected by @option{-mno-explicit-relocs},
15058 is to use assembler macros instead.
15060 @option{-mexplicit-relocs} is the default if GCC was configured
15061 to use an assembler that supports relocation operators.
15063 @item -mcheck-zero-division
15064 @itemx -mno-check-zero-division
15065 @opindex mcheck-zero-division
15066 @opindex mno-check-zero-division
15067 Trap (do not trap) on integer division by zero.
15069 The default is @option{-mcheck-zero-division}.
15071 @item -mdivide-traps
15072 @itemx -mdivide-breaks
15073 @opindex mdivide-traps
15074 @opindex mdivide-breaks
15075 MIPS systems check for division by zero by generating either a
15076 conditional trap or a break instruction. Using traps results in
15077 smaller code, but is only supported on MIPS II and later. Also, some
15078 versions of the Linux kernel have a bug that prevents trap from
15079 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
15080 allow conditional traps on architectures that support them and
15081 @option{-mdivide-breaks} to force the use of breaks.
15083 The default is usually @option{-mdivide-traps}, but this can be
15084 overridden at configure time using @option{--with-divide=breaks}.
15085 Divide-by-zero checks can be completely disabled using
15086 @option{-mno-check-zero-division}.
15091 @opindex mno-memcpy
15092 Force (do not force) the use of @code{memcpy()} for non-trivial block
15093 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
15094 most constant-sized copies.
15097 @itemx -mno-long-calls
15098 @opindex mlong-calls
15099 @opindex mno-long-calls
15100 Disable (do not disable) use of the @code{jal} instruction. Calling
15101 functions using @code{jal} is more efficient but requires the caller
15102 and callee to be in the same 256 megabyte segment.
15104 This option has no effect on abicalls code. The default is
15105 @option{-mno-long-calls}.
15111 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
15112 instructions, as provided by the R4650 ISA@.
15115 @itemx -mno-fused-madd
15116 @opindex mfused-madd
15117 @opindex mno-fused-madd
15118 Enable (disable) use of the floating point multiply-accumulate
15119 instructions, when they are available. The default is
15120 @option{-mfused-madd}.
15122 When multiply-accumulate instructions are used, the intermediate
15123 product is calculated to infinite precision and is not subject to
15124 the FCSR Flush to Zero bit. This may be undesirable in some
15129 Tell the MIPS assembler to not run its preprocessor over user
15130 assembler files (with a @samp{.s} suffix) when assembling them.
15135 @opindex mno-fix-24k
15136 Work around the 24K E48 (lost data on stores during refill) errata.
15137 The workarounds are implemented by the assembler rather than by GCC.
15140 @itemx -mno-fix-r4000
15141 @opindex mfix-r4000
15142 @opindex mno-fix-r4000
15143 Work around certain R4000 CPU errata:
15146 A double-word or a variable shift may give an incorrect result if executed
15147 immediately after starting an integer division.
15149 A double-word or a variable shift may give an incorrect result if executed
15150 while an integer multiplication is in progress.
15152 An integer division may give an incorrect result if started in a delay slot
15153 of a taken branch or a jump.
15157 @itemx -mno-fix-r4400
15158 @opindex mfix-r4400
15159 @opindex mno-fix-r4400
15160 Work around certain R4400 CPU errata:
15163 A double-word or a variable shift may give an incorrect result if executed
15164 immediately after starting an integer division.
15168 @itemx -mno-fix-r10000
15169 @opindex mfix-r10000
15170 @opindex mno-fix-r10000
15171 Work around certain R10000 errata:
15174 @code{ll}/@code{sc} sequences may not behave atomically on revisions
15175 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
15178 This option can only be used if the target architecture supports
15179 branch-likely instructions. @option{-mfix-r10000} is the default when
15180 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
15184 @itemx -mno-fix-vr4120
15185 @opindex mfix-vr4120
15186 Work around certain VR4120 errata:
15189 @code{dmultu} does not always produce the correct result.
15191 @code{div} and @code{ddiv} do not always produce the correct result if one
15192 of the operands is negative.
15194 The workarounds for the division errata rely on special functions in
15195 @file{libgcc.a}. At present, these functions are only provided by
15196 the @code{mips64vr*-elf} configurations.
15198 Other VR4120 errata require a nop to be inserted between certain pairs of
15199 instructions. These errata are handled by the assembler, not by GCC itself.
15202 @opindex mfix-vr4130
15203 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
15204 workarounds are implemented by the assembler rather than by GCC,
15205 although GCC will avoid using @code{mflo} and @code{mfhi} if the
15206 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
15207 instructions are available instead.
15210 @itemx -mno-fix-sb1
15212 Work around certain SB-1 CPU core errata.
15213 (This flag currently works around the SB-1 revision 2
15214 ``F1'' and ``F2'' floating point errata.)
15216 @item -mr10k-cache-barrier=@var{setting}
15217 @opindex mr10k-cache-barrier
15218 Specify whether GCC should insert cache barriers to avoid the
15219 side-effects of speculation on R10K processors.
15221 In common with many processors, the R10K tries to predict the outcome
15222 of a conditional branch and speculatively executes instructions from
15223 the ``taken'' branch. It later aborts these instructions if the
15224 predicted outcome was wrong. However, on the R10K, even aborted
15225 instructions can have side effects.
15227 This problem only affects kernel stores and, depending on the system,
15228 kernel loads. As an example, a speculatively-executed store may load
15229 the target memory into cache and mark the cache line as dirty, even if
15230 the store itself is later aborted. If a DMA operation writes to the
15231 same area of memory before the ``dirty'' line is flushed, the cached
15232 data will overwrite the DMA-ed data. See the R10K processor manual
15233 for a full description, including other potential problems.
15235 One workaround is to insert cache barrier instructions before every memory
15236 access that might be speculatively executed and that might have side
15237 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15238 controls GCC's implementation of this workaround. It assumes that
15239 aborted accesses to any byte in the following regions will not have
15244 the memory occupied by the current function's stack frame;
15247 the memory occupied by an incoming stack argument;
15250 the memory occupied by an object with a link-time-constant address.
15253 It is the kernel's responsibility to ensure that speculative
15254 accesses to these regions are indeed safe.
15256 If the input program contains a function declaration such as:
15262 then the implementation of @code{foo} must allow @code{j foo} and
15263 @code{jal foo} to be executed speculatively. GCC honors this
15264 restriction for functions it compiles itself. It expects non-GCC
15265 functions (such as hand-written assembly code) to do the same.
15267 The option has three forms:
15270 @item -mr10k-cache-barrier=load-store
15271 Insert a cache barrier before a load or store that might be
15272 speculatively executed and that might have side effects even
15275 @item -mr10k-cache-barrier=store
15276 Insert a cache barrier before a store that might be speculatively
15277 executed and that might have side effects even if aborted.
15279 @item -mr10k-cache-barrier=none
15280 Disable the insertion of cache barriers. This is the default setting.
15283 @item -mflush-func=@var{func}
15284 @itemx -mno-flush-func
15285 @opindex mflush-func
15286 Specifies the function to call to flush the I and D caches, or to not
15287 call any such function. If called, the function must take the same
15288 arguments as the common @code{_flush_func()}, that is, the address of the
15289 memory range for which the cache is being flushed, the size of the
15290 memory range, and the number 3 (to flush both caches). The default
15291 depends on the target GCC was configured for, but commonly is either
15292 @samp{_flush_func} or @samp{__cpu_flush}.
15294 @item mbranch-cost=@var{num}
15295 @opindex mbranch-cost
15296 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15297 This cost is only a heuristic and is not guaranteed to produce
15298 consistent results across releases. A zero cost redundantly selects
15299 the default, which is based on the @option{-mtune} setting.
15301 @item -mbranch-likely
15302 @itemx -mno-branch-likely
15303 @opindex mbranch-likely
15304 @opindex mno-branch-likely
15305 Enable or disable use of Branch Likely instructions, regardless of the
15306 default for the selected architecture. By default, Branch Likely
15307 instructions may be generated if they are supported by the selected
15308 architecture. An exception is for the MIPS32 and MIPS64 architectures
15309 and processors which implement those architectures; for those, Branch
15310 Likely instructions will not be generated by default because the MIPS32
15311 and MIPS64 architectures specifically deprecate their use.
15313 @item -mfp-exceptions
15314 @itemx -mno-fp-exceptions
15315 @opindex mfp-exceptions
15316 Specifies whether FP exceptions are enabled. This affects how we schedule
15317 FP instructions for some processors. The default is that FP exceptions are
15320 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15321 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15324 @item -mvr4130-align
15325 @itemx -mno-vr4130-align
15326 @opindex mvr4130-align
15327 The VR4130 pipeline is two-way superscalar, but can only issue two
15328 instructions together if the first one is 8-byte aligned. When this
15329 option is enabled, GCC will align pairs of instructions that it
15330 thinks should execute in parallel.
15332 This option only has an effect when optimizing for the VR4130.
15333 It normally makes code faster, but at the expense of making it bigger.
15334 It is enabled by default at optimization level @option{-O3}.
15339 Enable (disable) generation of @code{synci} instructions on
15340 architectures that support it. The @code{synci} instructions (if
15341 enabled) will be generated when @code{__builtin___clear_cache()} is
15344 This option defaults to @code{-mno-synci}, but the default can be
15345 overridden by configuring with @code{--with-synci}.
15347 When compiling code for single processor systems, it is generally safe
15348 to use @code{synci}. However, on many multi-core (SMP) systems, it
15349 will not invalidate the instruction caches on all cores and may lead
15350 to undefined behavior.
15352 @item -mrelax-pic-calls
15353 @itemx -mno-relax-pic-calls
15354 @opindex mrelax-pic-calls
15355 Try to turn PIC calls that are normally dispatched via register
15356 @code{$25} into direct calls. This is only possible if the linker can
15357 resolve the destination at link-time and if the destination is within
15358 range for a direct call.
15360 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15361 an assembler and a linker that supports the @code{.reloc} assembly
15362 directive and @code{-mexplicit-relocs} is in effect. With
15363 @code{-mno-explicit-relocs}, this optimization can be performed by the
15364 assembler and the linker alone without help from the compiler.
15366 @item -mmcount-ra-address
15367 @itemx -mno-mcount-ra-address
15368 @opindex mmcount-ra-address
15369 @opindex mno-mcount-ra-address
15370 Emit (do not emit) code that allows @code{_mcount} to modify the
15371 calling function's return address. When enabled, this option extends
15372 the usual @code{_mcount} interface with a new @var{ra-address}
15373 parameter, which has type @code{intptr_t *} and is passed in register
15374 @code{$12}. @code{_mcount} can then modify the return address by
15375 doing both of the following:
15378 Returning the new address in register @code{$31}.
15380 Storing the new address in @code{*@var{ra-address}},
15381 if @var{ra-address} is nonnull.
15384 The default is @option{-mno-mcount-ra-address}.
15389 @subsection MMIX Options
15390 @cindex MMIX Options
15392 These options are defined for the MMIX:
15396 @itemx -mno-libfuncs
15398 @opindex mno-libfuncs
15399 Specify that intrinsic library functions are being compiled, passing all
15400 values in registers, no matter the size.
15403 @itemx -mno-epsilon
15405 @opindex mno-epsilon
15406 Generate floating-point comparison instructions that compare with respect
15407 to the @code{rE} epsilon register.
15409 @item -mabi=mmixware
15411 @opindex mabi=mmixware
15413 Generate code that passes function parameters and return values that (in
15414 the called function) are seen as registers @code{$0} and up, as opposed to
15415 the GNU ABI which uses global registers @code{$231} and up.
15417 @item -mzero-extend
15418 @itemx -mno-zero-extend
15419 @opindex mzero-extend
15420 @opindex mno-zero-extend
15421 When reading data from memory in sizes shorter than 64 bits, use (do not
15422 use) zero-extending load instructions by default, rather than
15423 sign-extending ones.
15426 @itemx -mno-knuthdiv
15428 @opindex mno-knuthdiv
15429 Make the result of a division yielding a remainder have the same sign as
15430 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15431 remainder follows the sign of the dividend. Both methods are
15432 arithmetically valid, the latter being almost exclusively used.
15434 @item -mtoplevel-symbols
15435 @itemx -mno-toplevel-symbols
15436 @opindex mtoplevel-symbols
15437 @opindex mno-toplevel-symbols
15438 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15439 code can be used with the @code{PREFIX} assembly directive.
15443 Generate an executable in the ELF format, rather than the default
15444 @samp{mmo} format used by the @command{mmix} simulator.
15446 @item -mbranch-predict
15447 @itemx -mno-branch-predict
15448 @opindex mbranch-predict
15449 @opindex mno-branch-predict
15450 Use (do not use) the probable-branch instructions, when static branch
15451 prediction indicates a probable branch.
15453 @item -mbase-addresses
15454 @itemx -mno-base-addresses
15455 @opindex mbase-addresses
15456 @opindex mno-base-addresses
15457 Generate (do not generate) code that uses @emph{base addresses}. Using a
15458 base address automatically generates a request (handled by the assembler
15459 and the linker) for a constant to be set up in a global register. The
15460 register is used for one or more base address requests within the range 0
15461 to 255 from the value held in the register. The generally leads to short
15462 and fast code, but the number of different data items that can be
15463 addressed is limited. This means that a program that uses lots of static
15464 data may require @option{-mno-base-addresses}.
15466 @item -msingle-exit
15467 @itemx -mno-single-exit
15468 @opindex msingle-exit
15469 @opindex mno-single-exit
15470 Force (do not force) generated code to have a single exit point in each
15474 @node MN10300 Options
15475 @subsection MN10300 Options
15476 @cindex MN10300 options
15478 These @option{-m} options are defined for Matsushita MN10300 architectures:
15483 Generate code to avoid bugs in the multiply instructions for the MN10300
15484 processors. This is the default.
15486 @item -mno-mult-bug
15487 @opindex mno-mult-bug
15488 Do not generate code to avoid bugs in the multiply instructions for the
15489 MN10300 processors.
15493 Generate code which uses features specific to the AM33 processor.
15497 Do not generate code which uses features specific to the AM33 processor. This
15502 Generate code which uses features specific to the AM33/2.0 processor.
15506 Generate code which uses features specific to the AM34 processor.
15508 @item -mtune=@var{cpu-type}
15510 Use the timing characteristics of the indicated CPU type when
15511 scheduling instructions. This does not change the targeted processor
15512 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15513 @samp{am33-2} or @samp{am34}.
15515 @item -mreturn-pointer-on-d0
15516 @opindex mreturn-pointer-on-d0
15517 When generating a function which returns a pointer, return the pointer
15518 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15519 only in a0, and attempts to call such functions without a prototype
15520 would result in errors. Note that this option is on by default; use
15521 @option{-mno-return-pointer-on-d0} to disable it.
15525 Do not link in the C run-time initialization object file.
15529 Indicate to the linker that it should perform a relaxation optimization pass
15530 to shorten branches, calls and absolute memory addresses. This option only
15531 has an effect when used on the command line for the final link step.
15533 This option makes symbolic debugging impossible.
15537 Allow the compiler to generate @emph{Long Instruction Word}
15538 instructions if the target is the @samp{AM33} or later. This is the
15539 default. This option defines the preprocessor macro @samp{__LIW__}.
15543 Do not allow the compiler to generate @emph{Long Instruction Word}
15544 instructions. This option defines the preprocessor macro
15549 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15550 instructions if the target is the @samp{AM33} or later. This is the
15551 default. This option defines the preprocessor macro @samp{__SETLB__}.
15555 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15556 instructions. This option defines the preprocessor macro
15557 @samp{__NO_SETLB__}.
15561 @node PDP-11 Options
15562 @subsection PDP-11 Options
15563 @cindex PDP-11 Options
15565 These options are defined for the PDP-11:
15570 Use hardware FPP floating point. This is the default. (FIS floating
15571 point on the PDP-11/40 is not supported.)
15574 @opindex msoft-float
15575 Do not use hardware floating point.
15579 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15583 Return floating-point results in memory. This is the default.
15587 Generate code for a PDP-11/40.
15591 Generate code for a PDP-11/45. This is the default.
15595 Generate code for a PDP-11/10.
15597 @item -mbcopy-builtin
15598 @opindex mbcopy-builtin
15599 Use inline @code{movmemhi} patterns for copying memory. This is the
15604 Do not use inline @code{movmemhi} patterns for copying memory.
15610 Use 16-bit @code{int}. This is the default.
15616 Use 32-bit @code{int}.
15619 @itemx -mno-float32
15621 @opindex mno-float32
15622 Use 64-bit @code{float}. This is the default.
15625 @itemx -mno-float64
15627 @opindex mno-float64
15628 Use 32-bit @code{float}.
15632 Use @code{abshi2} pattern. This is the default.
15636 Do not use @code{abshi2} pattern.
15638 @item -mbranch-expensive
15639 @opindex mbranch-expensive
15640 Pretend that branches are expensive. This is for experimenting with
15641 code generation only.
15643 @item -mbranch-cheap
15644 @opindex mbranch-cheap
15645 Do not pretend that branches are expensive. This is the default.
15649 Use Unix assembler syntax. This is the default when configured for
15650 @samp{pdp11-*-bsd}.
15654 Use DEC assembler syntax. This is the default when configured for any
15655 PDP-11 target other than @samp{pdp11-*-bsd}.
15658 @node picoChip Options
15659 @subsection picoChip Options
15660 @cindex picoChip options
15662 These @samp{-m} options are defined for picoChip implementations:
15666 @item -mae=@var{ae_type}
15668 Set the instruction set, register set, and instruction scheduling
15669 parameters for array element type @var{ae_type}. Supported values
15670 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15672 @option{-mae=ANY} selects a completely generic AE type. Code
15673 generated with this option will run on any of the other AE types. The
15674 code will not be as efficient as it would be if compiled for a specific
15675 AE type, and some types of operation (e.g., multiplication) will not
15676 work properly on all types of AE.
15678 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15679 for compiled code, and is the default.
15681 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15682 option may suffer from poor performance of byte (char) manipulation,
15683 since the DSP AE does not provide hardware support for byte load/stores.
15685 @item -msymbol-as-address
15686 Enable the compiler to directly use a symbol name as an address in a
15687 load/store instruction, without first loading it into a
15688 register. Typically, the use of this option will generate larger
15689 programs, which run faster than when the option isn't used. However, the
15690 results vary from program to program, so it is left as a user option,
15691 rather than being permanently enabled.
15693 @item -mno-inefficient-warnings
15694 Disables warnings about the generation of inefficient code. These
15695 warnings can be generated, for example, when compiling code which
15696 performs byte-level memory operations on the MAC AE type. The MAC AE has
15697 no hardware support for byte-level memory operations, so all byte
15698 load/stores must be synthesized from word load/store operations. This is
15699 inefficient and a warning will be generated indicating to the programmer
15700 that they should rewrite the code to avoid byte operations, or to target
15701 an AE type which has the necessary hardware support. This option enables
15702 the warning to be turned off.
15706 @node PowerPC Options
15707 @subsection PowerPC Options
15708 @cindex PowerPC options
15710 These are listed under @xref{RS/6000 and PowerPC Options}.
15712 @node RS/6000 and PowerPC Options
15713 @subsection IBM RS/6000 and PowerPC Options
15714 @cindex RS/6000 and PowerPC Options
15715 @cindex IBM RS/6000 and PowerPC Options
15717 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15724 @itemx -mno-powerpc
15725 @itemx -mpowerpc-gpopt
15726 @itemx -mno-powerpc-gpopt
15727 @itemx -mpowerpc-gfxopt
15728 @itemx -mno-powerpc-gfxopt
15731 @itemx -mno-powerpc64
15735 @itemx -mno-popcntb
15737 @itemx -mno-popcntd
15746 @itemx -mno-hard-dfp
15750 @opindex mno-power2
15752 @opindex mno-powerpc
15753 @opindex mpowerpc-gpopt
15754 @opindex mno-powerpc-gpopt
15755 @opindex mpowerpc-gfxopt
15756 @opindex mno-powerpc-gfxopt
15757 @opindex mpowerpc64
15758 @opindex mno-powerpc64
15762 @opindex mno-popcntb
15764 @opindex mno-popcntd
15770 @opindex mno-mfpgpr
15772 @opindex mno-hard-dfp
15773 GCC supports two related instruction set architectures for the
15774 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15775 instructions supported by the @samp{rios} chip set used in the original
15776 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15777 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15778 the IBM 4xx, 6xx, and follow-on microprocessors.
15780 Neither architecture is a subset of the other. However there is a
15781 large common subset of instructions supported by both. An MQ
15782 register is included in processors supporting the POWER architecture.
15784 You use these options to specify which instructions are available on the
15785 processor you are using. The default value of these options is
15786 determined when configuring GCC@. Specifying the
15787 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15788 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15789 rather than the options listed above.
15791 The @option{-mpower} option allows GCC to generate instructions that
15792 are found only in the POWER architecture and to use the MQ register.
15793 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15794 to generate instructions that are present in the POWER2 architecture but
15795 not the original POWER architecture.
15797 The @option{-mpowerpc} option allows GCC to generate instructions that
15798 are found only in the 32-bit subset of the PowerPC architecture.
15799 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15800 GCC to use the optional PowerPC architecture instructions in the
15801 General Purpose group, including floating-point square root. Specifying
15802 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15803 use the optional PowerPC architecture instructions in the Graphics
15804 group, including floating-point select.
15806 The @option{-mmfcrf} option allows GCC to generate the move from
15807 condition register field instruction implemented on the POWER4
15808 processor and other processors that support the PowerPC V2.01
15810 The @option{-mpopcntb} option allows GCC to generate the popcount and
15811 double precision FP reciprocal estimate instruction implemented on the
15812 POWER5 processor and other processors that support the PowerPC V2.02
15814 The @option{-mpopcntd} option allows GCC to generate the popcount
15815 instruction implemented on the POWER7 processor and other processors
15816 that support the PowerPC V2.06 architecture.
15817 The @option{-mfprnd} option allows GCC to generate the FP round to
15818 integer instructions implemented on the POWER5+ processor and other
15819 processors that support the PowerPC V2.03 architecture.
15820 The @option{-mcmpb} option allows GCC to generate the compare bytes
15821 instruction implemented on the POWER6 processor and other processors
15822 that support the PowerPC V2.05 architecture.
15823 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15824 general purpose register instructions implemented on the POWER6X
15825 processor and other processors that support the extended PowerPC V2.05
15827 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15828 point instructions implemented on some POWER processors.
15830 The @option{-mpowerpc64} option allows GCC to generate the additional
15831 64-bit instructions that are found in the full PowerPC64 architecture
15832 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15833 @option{-mno-powerpc64}.
15835 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15836 will use only the instructions in the common subset of both
15837 architectures plus some special AIX common-mode calls, and will not use
15838 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15839 permits GCC to use any instruction from either architecture and to
15840 allow use of the MQ register; specify this for the Motorola MPC601.
15842 @item -mnew-mnemonics
15843 @itemx -mold-mnemonics
15844 @opindex mnew-mnemonics
15845 @opindex mold-mnemonics
15846 Select which mnemonics to use in the generated assembler code. With
15847 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15848 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15849 assembler mnemonics defined for the POWER architecture. Instructions
15850 defined in only one architecture have only one mnemonic; GCC uses that
15851 mnemonic irrespective of which of these options is specified.
15853 GCC defaults to the mnemonics appropriate for the architecture in
15854 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15855 value of these option. Unless you are building a cross-compiler, you
15856 should normally not specify either @option{-mnew-mnemonics} or
15857 @option{-mold-mnemonics}, but should instead accept the default.
15859 @item -mcpu=@var{cpu_type}
15861 Set architecture type, register usage, choice of mnemonics, and
15862 instruction scheduling parameters for machine type @var{cpu_type}.
15863 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15864 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15865 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15866 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15867 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15868 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15869 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15870 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15871 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15872 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15873 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15875 @option{-mcpu=common} selects a completely generic processor. Code
15876 generated under this option will run on any POWER or PowerPC processor.
15877 GCC will use only the instructions in the common subset of both
15878 architectures, and will not use the MQ register. GCC assumes a generic
15879 processor model for scheduling purposes.
15881 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15882 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15883 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15884 types, with an appropriate, generic processor model assumed for
15885 scheduling purposes.
15887 The other options specify a specific processor. Code generated under
15888 those options will run best on that processor, and may not run at all on
15891 The @option{-mcpu} options automatically enable or disable the
15894 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15895 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15896 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15897 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15899 The particular options set for any particular CPU will vary between
15900 compiler versions, depending on what setting seems to produce optimal
15901 code for that CPU; it doesn't necessarily reflect the actual hardware's
15902 capabilities. If you wish to set an individual option to a particular
15903 value, you may specify it after the @option{-mcpu} option, like
15904 @samp{-mcpu=970 -mno-altivec}.
15906 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15907 not enabled or disabled by the @option{-mcpu} option at present because
15908 AIX does not have full support for these options. You may still
15909 enable or disable them individually if you're sure it'll work in your
15912 @item -mtune=@var{cpu_type}
15914 Set the instruction scheduling parameters for machine type
15915 @var{cpu_type}, but do not set the architecture type, register usage, or
15916 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15917 values for @var{cpu_type} are used for @option{-mtune} as for
15918 @option{-mcpu}. If both are specified, the code generated will use the
15919 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15920 scheduling parameters set by @option{-mtune}.
15922 @item -mcmodel=small
15923 @opindex mcmodel=small
15924 Generate PowerPC64 code for the small model: The TOC is limited to
15927 @item -mcmodel=medium
15928 @opindex mcmodel=medium
15929 Generate PowerPC64 code for the medium model: The TOC and other static
15930 data may be up to a total of 4G in size.
15932 @item -mcmodel=large
15933 @opindex mcmodel=large
15934 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15935 in size. Other data and code is only limited by the 64-bit address
15939 @itemx -mno-altivec
15941 @opindex mno-altivec
15942 Generate code that uses (does not use) AltiVec instructions, and also
15943 enable the use of built-in functions that allow more direct access to
15944 the AltiVec instruction set. You may also need to set
15945 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15951 @opindex mno-vrsave
15952 Generate VRSAVE instructions when generating AltiVec code.
15954 @item -mgen-cell-microcode
15955 @opindex mgen-cell-microcode
15956 Generate Cell microcode instructions
15958 @item -mwarn-cell-microcode
15959 @opindex mwarn-cell-microcode
15960 Warning when a Cell microcode instruction is going to emitted. An example
15961 of a Cell microcode instruction is a variable shift.
15964 @opindex msecure-plt
15965 Generate code that allows ld and ld.so to build executables and shared
15966 libraries with non-exec .plt and .got sections. This is a PowerPC
15967 32-bit SYSV ABI option.
15971 Generate code that uses a BSS .plt section that ld.so fills in, and
15972 requires .plt and .got sections that are both writable and executable.
15973 This is a PowerPC 32-bit SYSV ABI option.
15979 This switch enables or disables the generation of ISEL instructions.
15981 @item -misel=@var{yes/no}
15982 This switch has been deprecated. Use @option{-misel} and
15983 @option{-mno-isel} instead.
15989 This switch enables or disables the generation of SPE simd
15995 @opindex mno-paired
15996 This switch enables or disables the generation of PAIRED simd
15999 @item -mspe=@var{yes/no}
16000 This option has been deprecated. Use @option{-mspe} and
16001 @option{-mno-spe} instead.
16007 Generate code that uses (does not use) vector/scalar (VSX)
16008 instructions, and also enable the use of built-in functions that allow
16009 more direct access to the VSX instruction set.
16011 @item -mfloat-gprs=@var{yes/single/double/no}
16012 @itemx -mfloat-gprs
16013 @opindex mfloat-gprs
16014 This switch enables or disables the generation of floating point
16015 operations on the general purpose registers for architectures that
16018 The argument @var{yes} or @var{single} enables the use of
16019 single-precision floating point operations.
16021 The argument @var{double} enables the use of single and
16022 double-precision floating point operations.
16024 The argument @var{no} disables floating point operations on the
16025 general purpose registers.
16027 This option is currently only available on the MPC854x.
16033 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
16034 targets (including GNU/Linux). The 32-bit environment sets int, long
16035 and pointer to 32 bits and generates code that runs on any PowerPC
16036 variant. The 64-bit environment sets int to 32 bits and long and
16037 pointer to 64 bits, and generates code for PowerPC64, as for
16038 @option{-mpowerpc64}.
16041 @itemx -mno-fp-in-toc
16042 @itemx -mno-sum-in-toc
16043 @itemx -mminimal-toc
16045 @opindex mno-fp-in-toc
16046 @opindex mno-sum-in-toc
16047 @opindex mminimal-toc
16048 Modify generation of the TOC (Table Of Contents), which is created for
16049 every executable file. The @option{-mfull-toc} option is selected by
16050 default. In that case, GCC will allocate at least one TOC entry for
16051 each unique non-automatic variable reference in your program. GCC
16052 will also place floating-point constants in the TOC@. However, only
16053 16,384 entries are available in the TOC@.
16055 If you receive a linker error message that saying you have overflowed
16056 the available TOC space, you can reduce the amount of TOC space used
16057 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
16058 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
16059 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
16060 generate code to calculate the sum of an address and a constant at
16061 run-time instead of putting that sum into the TOC@. You may specify one
16062 or both of these options. Each causes GCC to produce very slightly
16063 slower and larger code at the expense of conserving TOC space.
16065 If you still run out of space in the TOC even when you specify both of
16066 these options, specify @option{-mminimal-toc} instead. This option causes
16067 GCC to make only one TOC entry for every file. When you specify this
16068 option, GCC will produce code that is slower and larger but which
16069 uses extremely little TOC space. You may wish to use this option
16070 only on files that contain less frequently executed code.
16076 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
16077 @code{long} type, and the infrastructure needed to support them.
16078 Specifying @option{-maix64} implies @option{-mpowerpc64} and
16079 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
16080 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
16083 @itemx -mno-xl-compat
16084 @opindex mxl-compat
16085 @opindex mno-xl-compat
16086 Produce code that conforms more closely to IBM XL compiler semantics
16087 when using AIX-compatible ABI@. Pass floating-point arguments to
16088 prototyped functions beyond the register save area (RSA) on the stack
16089 in addition to argument FPRs. Do not assume that most significant
16090 double in 128-bit long double value is properly rounded when comparing
16091 values and converting to double. Use XL symbol names for long double
16094 The AIX calling convention was extended but not initially documented to
16095 handle an obscure K&R C case of calling a function that takes the
16096 address of its arguments with fewer arguments than declared. IBM XL
16097 compilers access floating point arguments which do not fit in the
16098 RSA from the stack when a subroutine is compiled without
16099 optimization. Because always storing floating-point arguments on the
16100 stack is inefficient and rarely needed, this option is not enabled by
16101 default and only is necessary when calling subroutines compiled by IBM
16102 XL compilers without optimization.
16106 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
16107 application written to use message passing with special startup code to
16108 enable the application to run. The system must have PE installed in the
16109 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
16110 must be overridden with the @option{-specs=} option to specify the
16111 appropriate directory location. The Parallel Environment does not
16112 support threads, so the @option{-mpe} option and the @option{-pthread}
16113 option are incompatible.
16115 @item -malign-natural
16116 @itemx -malign-power
16117 @opindex malign-natural
16118 @opindex malign-power
16119 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
16120 @option{-malign-natural} overrides the ABI-defined alignment of larger
16121 types, such as floating-point doubles, on their natural size-based boundary.
16122 The option @option{-malign-power} instructs GCC to follow the ABI-specified
16123 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
16125 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
16129 @itemx -mhard-float
16130 @opindex msoft-float
16131 @opindex mhard-float
16132 Generate code that does not use (uses) the floating-point register set.
16133 Software floating point emulation is provided if you use the
16134 @option{-msoft-float} option, and pass the option to GCC when linking.
16136 @item -msingle-float
16137 @itemx -mdouble-float
16138 @opindex msingle-float
16139 @opindex mdouble-float
16140 Generate code for single or double-precision floating point operations.
16141 @option{-mdouble-float} implies @option{-msingle-float}.
16144 @opindex msimple-fpu
16145 Do not generate sqrt and div instructions for hardware floating point unit.
16149 Specify type of floating point unit. Valid values are @var{sp_lite}
16150 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
16151 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
16152 and @var{dp_full} (equivalent to -mdouble-float).
16155 @opindex mxilinx-fpu
16156 Perform optimizations for floating point unit on Xilinx PPC 405/440.
16159 @itemx -mno-multiple
16161 @opindex mno-multiple
16162 Generate code that uses (does not use) the load multiple word
16163 instructions and the store multiple word instructions. These
16164 instructions are generated by default on POWER systems, and not
16165 generated on PowerPC systems. Do not use @option{-mmultiple} on little
16166 endian PowerPC systems, since those instructions do not work when the
16167 processor is in little endian mode. The exceptions are PPC740 and
16168 PPC750 which permit the instructions usage in little endian mode.
16173 @opindex mno-string
16174 Generate code that uses (does not use) the load string instructions
16175 and the store string word instructions to save multiple registers and
16176 do small block moves. These instructions are generated by default on
16177 POWER systems, and not generated on PowerPC systems. Do not use
16178 @option{-mstring} on little endian PowerPC systems, since those
16179 instructions do not work when the processor is in little endian mode.
16180 The exceptions are PPC740 and PPC750 which permit the instructions
16181 usage in little endian mode.
16186 @opindex mno-update
16187 Generate code that uses (does not use) the load or store instructions
16188 that update the base register to the address of the calculated memory
16189 location. These instructions are generated by default. If you use
16190 @option{-mno-update}, there is a small window between the time that the
16191 stack pointer is updated and the address of the previous frame is
16192 stored, which means code that walks the stack frame across interrupts or
16193 signals may get corrupted data.
16195 @item -mavoid-indexed-addresses
16196 @itemx -mno-avoid-indexed-addresses
16197 @opindex mavoid-indexed-addresses
16198 @opindex mno-avoid-indexed-addresses
16199 Generate code that tries to avoid (not avoid) the use of indexed load
16200 or store instructions. These instructions can incur a performance
16201 penalty on Power6 processors in certain situations, such as when
16202 stepping through large arrays that cross a 16M boundary. This option
16203 is enabled by default when targetting Power6 and disabled otherwise.
16206 @itemx -mno-fused-madd
16207 @opindex mfused-madd
16208 @opindex mno-fused-madd
16209 Generate code that uses (does not use) the floating point multiply and
16210 accumulate instructions. These instructions are generated by default
16211 if hardware floating point is used. The machine dependent
16212 @option{-mfused-madd} option is now mapped to the machine independent
16213 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
16214 mapped to @option{-ffp-contract=off}.
16220 Generate code that uses (does not use) the half-word multiply and
16221 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
16222 These instructions are generated by default when targetting those
16229 Generate code that uses (does not use) the string-search @samp{dlmzb}
16230 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
16231 generated by default when targetting those processors.
16233 @item -mno-bit-align
16235 @opindex mno-bit-align
16236 @opindex mbit-align
16237 On System V.4 and embedded PowerPC systems do not (do) force structures
16238 and unions that contain bit-fields to be aligned to the base type of the
16241 For example, by default a structure containing nothing but 8
16242 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
16243 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
16244 the structure would be aligned to a 1 byte boundary and be one byte in
16247 @item -mno-strict-align
16248 @itemx -mstrict-align
16249 @opindex mno-strict-align
16250 @opindex mstrict-align
16251 On System V.4 and embedded PowerPC systems do not (do) assume that
16252 unaligned memory references will be handled by the system.
16254 @item -mrelocatable
16255 @itemx -mno-relocatable
16256 @opindex mrelocatable
16257 @opindex mno-relocatable
16258 Generate code that allows (does not allow) a static executable to be
16259 relocated to a different address at runtime. A simple embedded
16260 PowerPC system loader should relocate the entire contents of
16261 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
16262 a table of 32-bit addresses generated by this option. For this to
16263 work, all objects linked together must be compiled with
16264 @option{-mrelocatable} or @option{-mrelocatable-lib}.
16265 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
16267 @item -mrelocatable-lib
16268 @itemx -mno-relocatable-lib
16269 @opindex mrelocatable-lib
16270 @opindex mno-relocatable-lib
16271 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
16272 @code{.fixup} section to allow static executables to be relocated at
16273 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
16274 alignment of @option{-mrelocatable}. Objects compiled with
16275 @option{-mrelocatable-lib} may be linked with objects compiled with
16276 any combination of the @option{-mrelocatable} options.
16282 On System V.4 and embedded PowerPC systems do not (do) assume that
16283 register 2 contains a pointer to a global area pointing to the addresses
16284 used in the program.
16287 @itemx -mlittle-endian
16289 @opindex mlittle-endian
16290 On System V.4 and embedded PowerPC systems compile code for the
16291 processor in little endian mode. The @option{-mlittle-endian} option is
16292 the same as @option{-mlittle}.
16295 @itemx -mbig-endian
16297 @opindex mbig-endian
16298 On System V.4 and embedded PowerPC systems compile code for the
16299 processor in big endian mode. The @option{-mbig-endian} option is
16300 the same as @option{-mbig}.
16302 @item -mdynamic-no-pic
16303 @opindex mdynamic-no-pic
16304 On Darwin and Mac OS X systems, compile code so that it is not
16305 relocatable, but that its external references are relocatable. The
16306 resulting code is suitable for applications, but not shared
16309 @item -msingle-pic-base
16310 @opindex msingle-pic-base
16311 Treat the register used for PIC addressing as read-only, rather than
16312 loading it in the prologue for each function. The run-time system is
16313 responsible for initializing this register with an appropriate value
16314 before execution begins.
16316 @item -mprioritize-restricted-insns=@var{priority}
16317 @opindex mprioritize-restricted-insns
16318 This option controls the priority that is assigned to
16319 dispatch-slot restricted instructions during the second scheduling
16320 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16321 @var{no/highest/second-highest} priority to dispatch slot restricted
16324 @item -msched-costly-dep=@var{dependence_type}
16325 @opindex msched-costly-dep
16326 This option controls which dependences are considered costly
16327 by the target during instruction scheduling. The argument
16328 @var{dependence_type} takes one of the following values:
16329 @var{no}: no dependence is costly,
16330 @var{all}: all dependences are costly,
16331 @var{true_store_to_load}: a true dependence from store to load is costly,
16332 @var{store_to_load}: any dependence from store to load is costly,
16333 @var{number}: any dependence which latency >= @var{number} is costly.
16335 @item -minsert-sched-nops=@var{scheme}
16336 @opindex minsert-sched-nops
16337 This option controls which nop insertion scheme will be used during
16338 the second scheduling pass. The argument @var{scheme} takes one of the
16340 @var{no}: Don't insert nops.
16341 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
16342 according to the scheduler's grouping.
16343 @var{regroup_exact}: Insert nops to force costly dependent insns into
16344 separate groups. Insert exactly as many nops as needed to force an insn
16345 to a new group, according to the estimated processor grouping.
16346 @var{number}: Insert nops to force costly dependent insns into
16347 separate groups. Insert @var{number} nops to force an insn to a new group.
16350 @opindex mcall-sysv
16351 On System V.4 and embedded PowerPC systems compile code using calling
16352 conventions that adheres to the March 1995 draft of the System V
16353 Application Binary Interface, PowerPC processor supplement. This is the
16354 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16356 @item -mcall-sysv-eabi
16358 @opindex mcall-sysv-eabi
16359 @opindex mcall-eabi
16360 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16362 @item -mcall-sysv-noeabi
16363 @opindex mcall-sysv-noeabi
16364 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16366 @item -mcall-aixdesc
16368 On System V.4 and embedded PowerPC systems compile code for the AIX
16372 @opindex mcall-linux
16373 On System V.4 and embedded PowerPC systems compile code for the
16374 Linux-based GNU system.
16376 @item -mcall-freebsd
16377 @opindex mcall-freebsd
16378 On System V.4 and embedded PowerPC systems compile code for the
16379 FreeBSD operating system.
16381 @item -mcall-netbsd
16382 @opindex mcall-netbsd
16383 On System V.4 and embedded PowerPC systems compile code for the
16384 NetBSD operating system.
16386 @item -mcall-openbsd
16387 @opindex mcall-netbsd
16388 On System V.4 and embedded PowerPC systems compile code for the
16389 OpenBSD operating system.
16391 @item -maix-struct-return
16392 @opindex maix-struct-return
16393 Return all structures in memory (as specified by the AIX ABI)@.
16395 @item -msvr4-struct-return
16396 @opindex msvr4-struct-return
16397 Return structures smaller than 8 bytes in registers (as specified by the
16400 @item -mabi=@var{abi-type}
16402 Extend the current ABI with a particular extension, or remove such extension.
16403 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16404 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16408 Extend the current ABI with SPE ABI extensions. This does not change
16409 the default ABI, instead it adds the SPE ABI extensions to the current
16413 @opindex mabi=no-spe
16414 Disable Booke SPE ABI extensions for the current ABI@.
16416 @item -mabi=ibmlongdouble
16417 @opindex mabi=ibmlongdouble
16418 Change the current ABI to use IBM extended precision long double.
16419 This is a PowerPC 32-bit SYSV ABI option.
16421 @item -mabi=ieeelongdouble
16422 @opindex mabi=ieeelongdouble
16423 Change the current ABI to use IEEE extended precision long double.
16424 This is a PowerPC 32-bit Linux ABI option.
16427 @itemx -mno-prototype
16428 @opindex mprototype
16429 @opindex mno-prototype
16430 On System V.4 and embedded PowerPC systems assume that all calls to
16431 variable argument functions are properly prototyped. Otherwise, the
16432 compiler must insert an instruction before every non prototyped call to
16433 set or clear bit 6 of the condition code register (@var{CR}) to
16434 indicate whether floating point values were passed in the floating point
16435 registers in case the function takes a variable arguments. With
16436 @option{-mprototype}, only calls to prototyped variable argument functions
16437 will set or clear the bit.
16441 On embedded PowerPC systems, assume that the startup module is called
16442 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16443 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16448 On embedded PowerPC systems, assume that the startup module is called
16449 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16454 On embedded PowerPC systems, assume that the startup module is called
16455 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16458 @item -myellowknife
16459 @opindex myellowknife
16460 On embedded PowerPC systems, assume that the startup module is called
16461 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16466 On System V.4 and embedded PowerPC systems, specify that you are
16467 compiling for a VxWorks system.
16471 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16472 header to indicate that @samp{eabi} extended relocations are used.
16478 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16479 Embedded Applications Binary Interface (eabi) which is a set of
16480 modifications to the System V.4 specifications. Selecting @option{-meabi}
16481 means that the stack is aligned to an 8 byte boundary, a function
16482 @code{__eabi} is called to from @code{main} to set up the eabi
16483 environment, and the @option{-msdata} option can use both @code{r2} and
16484 @code{r13} to point to two separate small data areas. Selecting
16485 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16486 do not call an initialization function from @code{main}, and the
16487 @option{-msdata} option will only use @code{r13} to point to a single
16488 small data area. The @option{-meabi} option is on by default if you
16489 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16492 @opindex msdata=eabi
16493 On System V.4 and embedded PowerPC systems, put small initialized
16494 @code{const} global and static data in the @samp{.sdata2} section, which
16495 is pointed to by register @code{r2}. Put small initialized
16496 non-@code{const} global and static data in the @samp{.sdata} section,
16497 which is pointed to by register @code{r13}. Put small uninitialized
16498 global and static data in the @samp{.sbss} section, which is adjacent to
16499 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16500 incompatible with the @option{-mrelocatable} option. The
16501 @option{-msdata=eabi} option also sets the @option{-memb} option.
16504 @opindex msdata=sysv
16505 On System V.4 and embedded PowerPC systems, put small global and static
16506 data in the @samp{.sdata} section, which is pointed to by register
16507 @code{r13}. Put small uninitialized global and static data in the
16508 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16509 The @option{-msdata=sysv} option is incompatible with the
16510 @option{-mrelocatable} option.
16512 @item -msdata=default
16514 @opindex msdata=default
16516 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16517 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16518 same as @option{-msdata=sysv}.
16521 @opindex msdata=data
16522 On System V.4 and embedded PowerPC systems, put small global
16523 data in the @samp{.sdata} section. Put small uninitialized global
16524 data in the @samp{.sbss} section. Do not use register @code{r13}
16525 to address small data however. This is the default behavior unless
16526 other @option{-msdata} options are used.
16530 @opindex msdata=none
16532 On embedded PowerPC systems, put all initialized global and static data
16533 in the @samp{.data} section, and all uninitialized data in the
16534 @samp{.bss} section.
16536 @item -mblock-move-inline-limit=@var{num}
16537 @opindex mblock-move-inline-limit
16538 Inline all block moves (such as calls to @code{memcpy} or structure
16539 copies) less than or equal to @var{num} bytes. The minimum value for
16540 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16541 targets. The default value is target-specific.
16545 @cindex smaller data references (PowerPC)
16546 @cindex .sdata/.sdata2 references (PowerPC)
16547 On embedded PowerPC systems, put global and static items less than or
16548 equal to @var{num} bytes into the small data or bss sections instead of
16549 the normal data or bss section. By default, @var{num} is 8. The
16550 @option{-G @var{num}} switch is also passed to the linker.
16551 All modules should be compiled with the same @option{-G @var{num}} value.
16554 @itemx -mno-regnames
16556 @opindex mno-regnames
16557 On System V.4 and embedded PowerPC systems do (do not) emit register
16558 names in the assembly language output using symbolic forms.
16561 @itemx -mno-longcall
16563 @opindex mno-longcall
16564 By default assume that all calls are far away so that a longer more
16565 expensive calling sequence is required. This is required for calls
16566 further than 32 megabytes (33,554,432 bytes) from the current location.
16567 A short call will be generated if the compiler knows
16568 the call cannot be that far away. This setting can be overridden by
16569 the @code{shortcall} function attribute, or by @code{#pragma
16572 Some linkers are capable of detecting out-of-range calls and generating
16573 glue code on the fly. On these systems, long calls are unnecessary and
16574 generate slower code. As of this writing, the AIX linker can do this,
16575 as can the GNU linker for PowerPC/64. It is planned to add this feature
16576 to the GNU linker for 32-bit PowerPC systems as well.
16578 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16579 callee, L42'', plus a ``branch island'' (glue code). The two target
16580 addresses represent the callee and the ``branch island''. The
16581 Darwin/PPC linker will prefer the first address and generate a ``bl
16582 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16583 otherwise, the linker will generate ``bl L42'' to call the ``branch
16584 island''. The ``branch island'' is appended to the body of the
16585 calling function; it computes the full 32-bit address of the callee
16588 On Mach-O (Darwin) systems, this option directs the compiler emit to
16589 the glue for every direct call, and the Darwin linker decides whether
16590 to use or discard it.
16592 In the future, we may cause GCC to ignore all longcall specifications
16593 when the linker is known to generate glue.
16595 @item -mtls-markers
16596 @itemx -mno-tls-markers
16597 @opindex mtls-markers
16598 @opindex mno-tls-markers
16599 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16600 specifying the function argument. The relocation allows ld to
16601 reliably associate function call with argument setup instructions for
16602 TLS optimization, which in turn allows gcc to better schedule the
16607 Adds support for multithreading with the @dfn{pthreads} library.
16608 This option sets flags for both the preprocessor and linker.
16613 This option will enable GCC to use the reciprocal estimate and
16614 reciprocal square root estimate instructions with additional
16615 Newton-Raphson steps to increase precision instead of doing a divide or
16616 square root and divide for floating point arguments. You should use
16617 the @option{-ffast-math} option when using @option{-mrecip} (or at
16618 least @option{-funsafe-math-optimizations},
16619 @option{-finite-math-only}, @option{-freciprocal-math} and
16620 @option{-fno-trapping-math}). Note that while the throughput of the
16621 sequence is generally higher than the throughput of the non-reciprocal
16622 instruction, the precision of the sequence can be decreased by up to 2
16623 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16626 @item -mrecip=@var{opt}
16627 @opindex mrecip=opt
16628 This option allows to control which reciprocal estimate instructions
16629 may be used. @var{opt} is a comma separated list of options, that may
16630 be preceded by a @code{!} to invert the option:
16631 @code{all}: enable all estimate instructions,
16632 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16633 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16634 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16635 @code{divf}: enable the single precision reciprocal approximation instructions;
16636 @code{divd}: enable the double precision reciprocal approximation instructions;
16637 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16638 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16639 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16641 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16642 all of the reciprocal estimate instructions, except for the
16643 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16644 which handle the double precision reciprocal square root calculations.
16646 @item -mrecip-precision
16647 @itemx -mno-recip-precision
16648 @opindex mrecip-precision
16649 Assume (do not assume) that the reciprocal estimate instructions
16650 provide higher precision estimates than is mandated by the powerpc
16651 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16652 automatically selects @option{-mrecip-precision}. The double
16653 precision square root estimate instructions are not generated by
16654 default on low precision machines, since they do not provide an
16655 estimate that converges after three steps.
16657 @item -mveclibabi=@var{type}
16658 @opindex mveclibabi
16659 Specifies the ABI type to use for vectorizing intrinsics using an
16660 external library. The only type supported at present is @code{mass},
16661 which specifies to use IBM's Mathematical Acceleration Subsystem
16662 (MASS) libraries for vectorizing intrinsics using external libraries.
16663 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16664 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16665 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16666 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16667 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16668 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16669 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16670 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16671 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16672 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16673 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16674 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16675 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16676 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16677 for power7. Both @option{-ftree-vectorize} and
16678 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16679 libraries will have to be specified at link time.
16684 Generate (do not generate) the @code{friz} instruction when the
16685 @option{-funsafe-math-optimizations} option is used to optimize
16686 rounding a floating point value to 64-bit integer and back to floating
16687 point. The @code{friz} instruction does not return the same value if
16688 the floating point number is too large to fit in an integer.
16690 @item -mpointers-to-nested-functions
16691 @itemx -mno-pointers-to-nested-functions
16692 @opindex mpointers-to-nested-functions
16693 Generate (do not generate) code to load up the static chain register
16694 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
16695 systems where a function pointer points to a 3 word descriptor giving
16696 the function address, TOC value to be loaded in register @var{r2}, and
16697 static chain value to be loaded in register @var{r11}. The
16698 @option{-mpointers-to-nested-functions} is on by default. You will
16699 not be able to call through pointers to nested functions or pointers
16700 to functions compiled in other languages that use the static chain if
16701 you use the @option{-mno-pointers-to-nested-functions}.
16703 @item -msave-toc-indirect
16704 @itemx -mno-save-toc-indirect
16705 @opindex msave-toc-indirect
16706 Generate (do not generate) code to save the TOC value in the reserved
16707 stack location in the function prologue if the function calls through
16708 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
16709 saved in the prologue, it is saved just before the call through the
16710 pointer. The @option{-mno-save-toc-indirect} option is the default.
16714 @subsection RX Options
16717 These command line options are defined for RX targets:
16720 @item -m64bit-doubles
16721 @itemx -m32bit-doubles
16722 @opindex m64bit-doubles
16723 @opindex m32bit-doubles
16724 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16725 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16726 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16727 works on 32-bit values, which is why the default is
16728 @option{-m32bit-doubles}.
16734 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16735 floating point hardware. The default is enabled for the @var{RX600}
16736 series and disabled for the @var{RX200} series.
16738 Floating point instructions will only be generated for 32-bit floating
16739 point values however, so if the @option{-m64bit-doubles} option is in
16740 use then the FPU hardware will not be used for doubles.
16742 @emph{Note} If the @option{-fpu} option is enabled then
16743 @option{-funsafe-math-optimizations} is also enabled automatically.
16744 This is because the RX FPU instructions are themselves unsafe.
16746 @item -mcpu=@var{name}
16748 Selects the type of RX CPU to be targeted. Currently three types are
16749 supported, the generic @var{RX600} and @var{RX200} series hardware and
16750 the specific @var{RX610} CPU. The default is @var{RX600}.
16752 The only difference between @var{RX600} and @var{RX610} is that the
16753 @var{RX610} does not support the @code{MVTIPL} instruction.
16755 The @var{RX200} series does not have a hardware floating point unit
16756 and so @option{-nofpu} is enabled by default when this type is
16759 @item -mbig-endian-data
16760 @itemx -mlittle-endian-data
16761 @opindex mbig-endian-data
16762 @opindex mlittle-endian-data
16763 Store data (but not code) in the big-endian format. The default is
16764 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16767 @item -msmall-data-limit=@var{N}
16768 @opindex msmall-data-limit
16769 Specifies the maximum size in bytes of global and static variables
16770 which can be placed into the small data area. Using the small data
16771 area can lead to smaller and faster code, but the size of area is
16772 limited and it is up to the programmer to ensure that the area does
16773 not overflow. Also when the small data area is used one of the RX's
16774 registers (usually @code{r13}) is reserved for use pointing to this
16775 area, so it is no longer available for use by the compiler. This
16776 could result in slower and/or larger code if variables which once
16777 could have been held in the reserved register are now pushed onto the
16780 Note, common variables (variables which have not been initialised) and
16781 constants are not placed into the small data area as they are assigned
16782 to other sections in the output executable.
16784 The default value is zero, which disables this feature. Note, this
16785 feature is not enabled by default with higher optimization levels
16786 (@option{-O2} etc) because of the potentially detrimental effects of
16787 reserving a register. It is up to the programmer to experiment and
16788 discover whether this feature is of benefit to their program. See the
16789 description of the @option{-mpid} option for a description of how the
16790 actual register to hold the small data area pointer is chosen.
16796 Use the simulator runtime. The default is to use the libgloss board
16799 @item -mas100-syntax
16800 @itemx -mno-as100-syntax
16801 @opindex mas100-syntax
16802 @opindex mno-as100-syntax
16803 When generating assembler output use a syntax that is compatible with
16804 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16805 assembler but it has some restrictions so generating it is not the
16808 @item -mmax-constant-size=@var{N}
16809 @opindex mmax-constant-size
16810 Specifies the maximum size, in bytes, of a constant that can be used as
16811 an operand in a RX instruction. Although the RX instruction set does
16812 allow constants of up to 4 bytes in length to be used in instructions,
16813 a longer value equates to a longer instruction. Thus in some
16814 circumstances it can be beneficial to restrict the size of constants
16815 that are used in instructions. Constants that are too big are instead
16816 placed into a constant pool and referenced via register indirection.
16818 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16819 or 4 means that constants of any size are allowed.
16823 Enable linker relaxation. Linker relaxation is a process whereby the
16824 linker will attempt to reduce the size of a program by finding shorter
16825 versions of various instructions. Disabled by default.
16827 @item -mint-register=@var{N}
16828 @opindex mint-register
16829 Specify the number of registers to reserve for fast interrupt handler
16830 functions. The value @var{N} can be between 0 and 4. A value of 1
16831 means that register @code{r13} will be reserved for the exclusive use
16832 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16833 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16834 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16835 A value of 0, the default, does not reserve any registers.
16837 @item -msave-acc-in-interrupts
16838 @opindex msave-acc-in-interrupts
16839 Specifies that interrupt handler functions should preserve the
16840 accumulator register. This is only necessary if normal code might use
16841 the accumulator register, for example because it performs 64-bit
16842 multiplications. The default is to ignore the accumulator as this
16843 makes the interrupt handlers faster.
16849 Enables the generation of position independent data. When enabled any
16850 access to constant data will done via an offset from a base address
16851 held in a register. This allows the location of constant data to be
16852 determined at run-time without requiring the executable to be
16853 relocated, which is a benefit to embedded applications with tight
16854 memory constraints. Data that can be modified is not affected by this
16857 Note, using this feature reserves a register, usually @code{r13}, for
16858 the constant data base address. This can result in slower and/or
16859 larger code, especially in complicated functions.
16861 The actual register chosen to hold the constant data base address
16862 depends upon whether the @option{-msmall-data-limit} and/or the
16863 @option{-mint-register} command line options are enabled. Starting
16864 with register @code{r13} and proceeding downwards, registers are
16865 allocated first to satisfy the requirements of @option{-mint-register},
16866 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
16867 is possible for the small data area register to be @code{r8} if both
16868 @option{-mint-register=4} and @option{-mpid} are specified on the
16871 By default this feature is not enabled. The default can be restored
16872 via the @option{-mno-pid} command line option.
16876 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16877 has special significance to the RX port when used with the
16878 @code{interrupt} function attribute. This attribute indicates a
16879 function intended to process fast interrupts. GCC will will ensure
16880 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16881 and/or @code{r13} and only provided that the normal use of the
16882 corresponding registers have been restricted via the
16883 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16886 @node S/390 and zSeries Options
16887 @subsection S/390 and zSeries Options
16888 @cindex S/390 and zSeries Options
16890 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16894 @itemx -msoft-float
16895 @opindex mhard-float
16896 @opindex msoft-float
16897 Use (do not use) the hardware floating-point instructions and registers
16898 for floating-point operations. When @option{-msoft-float} is specified,
16899 functions in @file{libgcc.a} will be used to perform floating-point
16900 operations. When @option{-mhard-float} is specified, the compiler
16901 generates IEEE floating-point instructions. This is the default.
16904 @itemx -mno-hard-dfp
16906 @opindex mno-hard-dfp
16907 Use (do not use) the hardware decimal-floating-point instructions for
16908 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16909 specified, functions in @file{libgcc.a} will be used to perform
16910 decimal-floating-point operations. When @option{-mhard-dfp} is
16911 specified, the compiler generates decimal-floating-point hardware
16912 instructions. This is the default for @option{-march=z9-ec} or higher.
16914 @item -mlong-double-64
16915 @itemx -mlong-double-128
16916 @opindex mlong-double-64
16917 @opindex mlong-double-128
16918 These switches control the size of @code{long double} type. A size
16919 of 64bit makes the @code{long double} type equivalent to the @code{double}
16920 type. This is the default.
16923 @itemx -mno-backchain
16924 @opindex mbackchain
16925 @opindex mno-backchain
16926 Store (do not store) the address of the caller's frame as backchain pointer
16927 into the callee's stack frame.
16928 A backchain may be needed to allow debugging using tools that do not understand
16929 DWARF-2 call frame information.
16930 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16931 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16932 the backchain is placed into the topmost word of the 96/160 byte register
16935 In general, code compiled with @option{-mbackchain} is call-compatible with
16936 code compiled with @option{-mmo-backchain}; however, use of the backchain
16937 for debugging purposes usually requires that the whole binary is built with
16938 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16939 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16940 to build a linux kernel use @option{-msoft-float}.
16942 The default is to not maintain the backchain.
16944 @item -mpacked-stack
16945 @itemx -mno-packed-stack
16946 @opindex mpacked-stack
16947 @opindex mno-packed-stack
16948 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16949 specified, the compiler uses the all fields of the 96/160 byte register save
16950 area only for their default purpose; unused fields still take up stack space.
16951 When @option{-mpacked-stack} is specified, register save slots are densely
16952 packed at the top of the register save area; unused space is reused for other
16953 purposes, allowing for more efficient use of the available stack space.
16954 However, when @option{-mbackchain} is also in effect, the topmost word of
16955 the save area is always used to store the backchain, and the return address
16956 register is always saved two words below the backchain.
16958 As long as the stack frame backchain is not used, code generated with
16959 @option{-mpacked-stack} is call-compatible with code generated with
16960 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16961 S/390 or zSeries generated code that uses the stack frame backchain at run
16962 time, not just for debugging purposes. Such code is not call-compatible
16963 with code compiled with @option{-mpacked-stack}. Also, note that the
16964 combination of @option{-mbackchain},
16965 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16966 to build a linux kernel use @option{-msoft-float}.
16968 The default is to not use the packed stack layout.
16971 @itemx -mno-small-exec
16972 @opindex msmall-exec
16973 @opindex mno-small-exec
16974 Generate (or do not generate) code using the @code{bras} instruction
16975 to do subroutine calls.
16976 This only works reliably if the total executable size does not
16977 exceed 64k. The default is to use the @code{basr} instruction instead,
16978 which does not have this limitation.
16984 When @option{-m31} is specified, generate code compliant to the
16985 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16986 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16987 particular to generate 64-bit instructions. For the @samp{s390}
16988 targets, the default is @option{-m31}, while the @samp{s390x}
16989 targets default to @option{-m64}.
16995 When @option{-mzarch} is specified, generate code using the
16996 instructions available on z/Architecture.
16997 When @option{-mesa} is specified, generate code using the
16998 instructions available on ESA/390. Note that @option{-mesa} is
16999 not possible with @option{-m64}.
17000 When generating code compliant to the GNU/Linux for S/390 ABI,
17001 the default is @option{-mesa}. When generating code compliant
17002 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
17008 Generate (or do not generate) code using the @code{mvcle} instruction
17009 to perform block moves. When @option{-mno-mvcle} is specified,
17010 use a @code{mvc} loop instead. This is the default unless optimizing for
17017 Print (or do not print) additional debug information when compiling.
17018 The default is to not print debug information.
17020 @item -march=@var{cpu-type}
17022 Generate code that will run on @var{cpu-type}, which is the name of a system
17023 representing a certain processor type. Possible values for
17024 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
17025 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
17026 When generating code using the instructions available on z/Architecture,
17027 the default is @option{-march=z900}. Otherwise, the default is
17028 @option{-march=g5}.
17030 @item -mtune=@var{cpu-type}
17032 Tune to @var{cpu-type} everything applicable about the generated code,
17033 except for the ABI and the set of available instructions.
17034 The list of @var{cpu-type} values is the same as for @option{-march}.
17035 The default is the value used for @option{-march}.
17038 @itemx -mno-tpf-trace
17039 @opindex mtpf-trace
17040 @opindex mno-tpf-trace
17041 Generate code that adds (does not add) in TPF OS specific branches to trace
17042 routines in the operating system. This option is off by default, even
17043 when compiling for the TPF OS@.
17046 @itemx -mno-fused-madd
17047 @opindex mfused-madd
17048 @opindex mno-fused-madd
17049 Generate code that uses (does not use) the floating point multiply and
17050 accumulate instructions. These instructions are generated by default if
17051 hardware floating point is used.
17053 @item -mwarn-framesize=@var{framesize}
17054 @opindex mwarn-framesize
17055 Emit a warning if the current function exceeds the given frame size. Because
17056 this is a compile time check it doesn't need to be a real problem when the program
17057 runs. It is intended to identify functions which most probably cause
17058 a stack overflow. It is useful to be used in an environment with limited stack
17059 size e.g.@: the linux kernel.
17061 @item -mwarn-dynamicstack
17062 @opindex mwarn-dynamicstack
17063 Emit a warning if the function calls alloca or uses dynamically
17064 sized arrays. This is generally a bad idea with a limited stack size.
17066 @item -mstack-guard=@var{stack-guard}
17067 @itemx -mstack-size=@var{stack-size}
17068 @opindex mstack-guard
17069 @opindex mstack-size
17070 If these options are provided the s390 back end emits additional instructions in
17071 the function prologue which trigger a trap if the stack size is @var{stack-guard}
17072 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
17073 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
17074 the frame size of the compiled function is chosen.
17075 These options are intended to be used to help debugging stack overflow problems.
17076 The additionally emitted code causes only little overhead and hence can also be
17077 used in production like systems without greater performance degradation. The given
17078 values have to be exact powers of 2 and @var{stack-size} has to be greater than
17079 @var{stack-guard} without exceeding 64k.
17080 In order to be efficient the extra code makes the assumption that the stack starts
17081 at an address aligned to the value given by @var{stack-size}.
17082 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
17085 @node Score Options
17086 @subsection Score Options
17087 @cindex Score Options
17089 These options are defined for Score implementations:
17094 Compile code for big endian mode. This is the default.
17098 Compile code for little endian mode.
17102 Disable generate bcnz instruction.
17106 Enable generate unaligned load and store instruction.
17110 Enable the use of multiply-accumulate instructions. Disabled by default.
17114 Specify the SCORE5 as the target architecture.
17118 Specify the SCORE5U of the target architecture.
17122 Specify the SCORE7 as the target architecture. This is the default.
17126 Specify the SCORE7D as the target architecture.
17130 @subsection SH Options
17132 These @samp{-m} options are defined for the SH implementations:
17137 Generate code for the SH1.
17141 Generate code for the SH2.
17144 Generate code for the SH2e.
17148 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
17149 that the floating-point unit is not used.
17151 @item -m2a-single-only
17152 @opindex m2a-single-only
17153 Generate code for the SH2a-FPU, in such a way that no double-precision
17154 floating point operations are used.
17157 @opindex m2a-single
17158 Generate code for the SH2a-FPU assuming the floating-point unit is in
17159 single-precision mode by default.
17163 Generate code for the SH2a-FPU assuming the floating-point unit is in
17164 double-precision mode by default.
17168 Generate code for the SH3.
17172 Generate code for the SH3e.
17176 Generate code for the SH4 without a floating-point unit.
17178 @item -m4-single-only
17179 @opindex m4-single-only
17180 Generate code for the SH4 with a floating-point unit that only
17181 supports single-precision arithmetic.
17185 Generate code for the SH4 assuming the floating-point unit is in
17186 single-precision mode by default.
17190 Generate code for the SH4.
17194 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
17195 floating-point unit is not used.
17197 @item -m4a-single-only
17198 @opindex m4a-single-only
17199 Generate code for the SH4a, in such a way that no double-precision
17200 floating point operations are used.
17203 @opindex m4a-single
17204 Generate code for the SH4a assuming the floating-point unit is in
17205 single-precision mode by default.
17209 Generate code for the SH4a.
17213 Same as @option{-m4a-nofpu}, except that it implicitly passes
17214 @option{-dsp} to the assembler. GCC doesn't generate any DSP
17215 instructions at the moment.
17219 Compile code for the processor in big endian mode.
17223 Compile code for the processor in little endian mode.
17227 Align doubles at 64-bit boundaries. Note that this changes the calling
17228 conventions, and thus some functions from the standard C library will
17229 not work unless you recompile it first with @option{-mdalign}.
17233 Shorten some address references at link time, when possible; uses the
17234 linker option @option{-relax}.
17238 Use 32-bit offsets in @code{switch} tables. The default is to use
17243 Enable the use of bit manipulation instructions on SH2A.
17247 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
17248 alignment constraints.
17252 Comply with the calling conventions defined by Renesas.
17256 Comply with the calling conventions defined by Renesas.
17260 Comply with the calling conventions defined for GCC before the Renesas
17261 conventions were available. This option is the default for all
17262 targets of the SH toolchain.
17265 @opindex mnomacsave
17266 Mark the @code{MAC} register as call-clobbered, even if
17267 @option{-mhitachi} is given.
17271 Increase IEEE-compliance of floating-point code.
17272 At the moment, this is equivalent to @option{-fno-finite-math-only}.
17273 When generating 16 bit SH opcodes, getting IEEE-conforming results for
17274 comparisons of NANs / infinities incurs extra overhead in every
17275 floating point comparison, therefore the default is set to
17276 @option{-ffinite-math-only}.
17278 @item -minline-ic_invalidate
17279 @opindex minline-ic_invalidate
17280 Inline code to invalidate instruction cache entries after setting up
17281 nested function trampolines.
17282 This option has no effect if -musermode is in effect and the selected
17283 code generation option (e.g. -m4) does not allow the use of the icbi
17285 If the selected code generation option does not allow the use of the icbi
17286 instruction, and -musermode is not in effect, the inlined code will
17287 manipulate the instruction cache address array directly with an associative
17288 write. This not only requires privileged mode, but it will also
17289 fail if the cache line had been mapped via the TLB and has become unmapped.
17293 Dump instruction size and location in the assembly code.
17296 @opindex mpadstruct
17297 This option is deprecated. It pads structures to multiple of 4 bytes,
17298 which is incompatible with the SH ABI@.
17302 Optimize for space instead of speed. Implied by @option{-Os}.
17305 @opindex mprefergot
17306 When generating position-independent code, emit function calls using
17307 the Global Offset Table instead of the Procedure Linkage Table.
17311 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
17312 if the inlined code would not work in user mode.
17313 This is the default when the target is @code{sh-*-linux*}.
17315 @item -multcost=@var{number}
17316 @opindex multcost=@var{number}
17317 Set the cost to assume for a multiply insn.
17319 @item -mdiv=@var{strategy}
17320 @opindex mdiv=@var{strategy}
17321 Set the division strategy to use for SHmedia code. @var{strategy} must be
17322 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
17323 inv:call2, inv:fp .
17324 "fp" performs the operation in floating point. This has a very high latency,
17325 but needs only a few instructions, so it might be a good choice if
17326 your code has enough easily exploitable ILP to allow the compiler to
17327 schedule the floating point instructions together with other instructions.
17328 Division by zero causes a floating point exception.
17329 "inv" uses integer operations to calculate the inverse of the divisor,
17330 and then multiplies the dividend with the inverse. This strategy allows
17331 cse and hoisting of the inverse calculation. Division by zero calculates
17332 an unspecified result, but does not trap.
17333 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17334 have been found, or if the entire operation has been hoisted to the same
17335 place, the last stages of the inverse calculation are intertwined with the
17336 final multiply to reduce the overall latency, at the expense of using a few
17337 more instructions, and thus offering fewer scheduling opportunities with
17339 "call" calls a library function that usually implements the inv:minlat
17341 This gives high code density for m5-*media-nofpu compilations.
17342 "call2" uses a different entry point of the same library function, where it
17343 assumes that a pointer to a lookup table has already been set up, which
17344 exposes the pointer load to cse / code hoisting optimizations.
17345 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17346 code generation, but if the code stays unoptimized, revert to the "call",
17347 "call2", or "fp" strategies, respectively. Note that the
17348 potentially-trapping side effect of division by zero is carried by a
17349 separate instruction, so it is possible that all the integer instructions
17350 are hoisted out, but the marker for the side effect stays where it is.
17351 A recombination to fp operations or a call is not possible in that case.
17352 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17353 that the inverse calculation was nor separated from the multiply, they speed
17354 up division where the dividend fits into 20 bits (plus sign where applicable),
17355 by inserting a test to skip a number of operations in this case; this test
17356 slows down the case of larger dividends. inv20u assumes the case of a such
17357 a small dividend to be unlikely, and inv20l assumes it to be likely.
17359 @item -maccumulate-outgoing-args
17360 @opindex maccumulate-outgoing-args
17361 Reserve space once for outgoing arguments in the function prologue rather
17362 than around each call. Generally beneficial for performance and size. Also
17363 needed for unwinding to avoid changing the stack frame around conditional code.
17365 @item -mdivsi3_libfunc=@var{name}
17366 @opindex mdivsi3_libfunc=@var{name}
17367 Set the name of the library function used for 32 bit signed division to
17368 @var{name}. This only affect the name used in the call and inv:call
17369 division strategies, and the compiler will still expect the same
17370 sets of input/output/clobbered registers as if this option was not present.
17372 @item -mfixed-range=@var{register-range}
17373 @opindex mfixed-range
17374 Generate code treating the given register range as fixed registers.
17375 A fixed register is one that the register allocator can not use. This is
17376 useful when compiling kernel code. A register range is specified as
17377 two registers separated by a dash. Multiple register ranges can be
17378 specified separated by a comma.
17380 @item -madjust-unroll
17381 @opindex madjust-unroll
17382 Throttle unrolling to avoid thrashing target registers.
17383 This option only has an effect if the gcc code base supports the
17384 TARGET_ADJUST_UNROLL_MAX target hook.
17386 @item -mindexed-addressing
17387 @opindex mindexed-addressing
17388 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17389 This is only safe if the hardware and/or OS implement 32 bit wrap-around
17390 semantics for the indexed addressing mode. The architecture allows the
17391 implementation of processors with 64 bit MMU, which the OS could use to
17392 get 32 bit addressing, but since no current hardware implementation supports
17393 this or any other way to make the indexed addressing mode safe to use in
17394 the 32 bit ABI, the default is -mno-indexed-addressing.
17396 @item -mgettrcost=@var{number}
17397 @opindex mgettrcost=@var{number}
17398 Set the cost assumed for the gettr instruction to @var{number}.
17399 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17403 Assume pt* instructions won't trap. This will generally generate better
17404 scheduled code, but is unsafe on current hardware. The current architecture
17405 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17406 This has the unintentional effect of making it unsafe to schedule ptabs /
17407 ptrel before a branch, or hoist it out of a loop. For example,
17408 __do_global_ctors, a part of libgcc that runs constructors at program
17409 startup, calls functions in a list which is delimited by @minus{}1. With the
17410 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17411 That means that all the constructors will be run a bit quicker, but when
17412 the loop comes to the end of the list, the program crashes because ptabs
17413 loads @minus{}1 into a target register. Since this option is unsafe for any
17414 hardware implementing the current architecture specification, the default
17415 is -mno-pt-fixed. Unless the user specifies a specific cost with
17416 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17417 this deters register allocation using target registers for storing
17420 @item -minvalid-symbols
17421 @opindex minvalid-symbols
17422 Assume symbols might be invalid. Ordinary function symbols generated by
17423 the compiler will always be valid to load with movi/shori/ptabs or
17424 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17425 to generate symbols that will cause ptabs / ptrel to trap.
17426 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17427 It will then prevent cross-basic-block cse, hoisting and most scheduling
17428 of symbol loads. The default is @option{-mno-invalid-symbols}.
17431 @node Solaris 2 Options
17432 @subsection Solaris 2 Options
17433 @cindex Solaris 2 options
17435 These @samp{-m} options are supported on Solaris 2:
17438 @item -mimpure-text
17439 @opindex mimpure-text
17440 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17441 the compiler to not pass @option{-z text} to the linker when linking a
17442 shared object. Using this option, you can link position-dependent
17443 code into a shared object.
17445 @option{-mimpure-text} suppresses the ``relocations remain against
17446 allocatable but non-writable sections'' linker error message.
17447 However, the necessary relocations will trigger copy-on-write, and the
17448 shared object is not actually shared across processes. Instead of
17449 using @option{-mimpure-text}, you should compile all source code with
17450 @option{-fpic} or @option{-fPIC}.
17454 These switches are supported in addition to the above on Solaris 2:
17459 Add support for multithreading using the POSIX threads library. This
17460 option sets flags for both the preprocessor and linker. This option does
17461 not affect the thread safety of object code produced by the compiler or
17462 that of libraries supplied with it.
17466 This is a synonym for @option{-pthreads}.
17469 @node SPARC Options
17470 @subsection SPARC Options
17471 @cindex SPARC options
17473 These @samp{-m} options are supported on the SPARC:
17476 @item -mno-app-regs
17478 @opindex mno-app-regs
17480 Specify @option{-mapp-regs} to generate output using the global registers
17481 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17484 To be fully SVR4 ABI compliant at the cost of some performance loss,
17485 specify @option{-mno-app-regs}. You should compile libraries and system
17486 software with this option.
17492 With @option{-mflat}, the compiler does not generate save/restore instructions
17493 and uses a ``flat'' or single register window model. This model is compatible
17494 with the regular register window model. The local registers and the input
17495 registers (0--5) are still treated as ``call-saved'' registers and will be
17496 saved on the stack as needed.
17498 With @option{-mno-flat} (the default), the compiler generates save/restore
17499 instructions (except for leaf functions). This is the normal operating mode.
17502 @itemx -mhard-float
17504 @opindex mhard-float
17505 Generate output containing floating point instructions. This is the
17509 @itemx -msoft-float
17511 @opindex msoft-float
17512 Generate output containing library calls for floating point.
17513 @strong{Warning:} the requisite libraries are not available for all SPARC
17514 targets. Normally the facilities of the machine's usual C compiler are
17515 used, but this cannot be done directly in cross-compilation. You must make
17516 your own arrangements to provide suitable library functions for
17517 cross-compilation. The embedded targets @samp{sparc-*-aout} and
17518 @samp{sparclite-*-*} do provide software floating point support.
17520 @option{-msoft-float} changes the calling convention in the output file;
17521 therefore, it is only useful if you compile @emph{all} of a program with
17522 this option. In particular, you need to compile @file{libgcc.a}, the
17523 library that comes with GCC, with @option{-msoft-float} in order for
17526 @item -mhard-quad-float
17527 @opindex mhard-quad-float
17528 Generate output containing quad-word (long double) floating point
17531 @item -msoft-quad-float
17532 @opindex msoft-quad-float
17533 Generate output containing library calls for quad-word (long double)
17534 floating point instructions. The functions called are those specified
17535 in the SPARC ABI@. This is the default.
17537 As of this writing, there are no SPARC implementations that have hardware
17538 support for the quad-word floating point instructions. They all invoke
17539 a trap handler for one of these instructions, and then the trap handler
17540 emulates the effect of the instruction. Because of the trap handler overhead,
17541 this is much slower than calling the ABI library routines. Thus the
17542 @option{-msoft-quad-float} option is the default.
17544 @item -mno-unaligned-doubles
17545 @itemx -munaligned-doubles
17546 @opindex mno-unaligned-doubles
17547 @opindex munaligned-doubles
17548 Assume that doubles have 8 byte alignment. This is the default.
17550 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17551 alignment only if they are contained in another type, or if they have an
17552 absolute address. Otherwise, it assumes they have 4 byte alignment.
17553 Specifying this option avoids some rare compatibility problems with code
17554 generated by other compilers. It is not the default because it results
17555 in a performance loss, especially for floating point code.
17557 @item -mno-faster-structs
17558 @itemx -mfaster-structs
17559 @opindex mno-faster-structs
17560 @opindex mfaster-structs
17561 With @option{-mfaster-structs}, the compiler assumes that structures
17562 should have 8 byte alignment. This enables the use of pairs of
17563 @code{ldd} and @code{std} instructions for copies in structure
17564 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17565 However, the use of this changed alignment directly violates the SPARC
17566 ABI@. Thus, it's intended only for use on targets where the developer
17567 acknowledges that their resulting code will not be directly in line with
17568 the rules of the ABI@.
17570 @item -mcpu=@var{cpu_type}
17572 Set the instruction set, register set, and instruction scheduling parameters
17573 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17574 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17575 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17576 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17577 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
17578 and @samp{niagara4}.
17580 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
17581 which selects the best architecture option for the host processor.
17582 @option{-mcpu=native} has no effect if GCC does not recognize
17585 Default instruction scheduling parameters are used for values that select
17586 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17587 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17589 Here is a list of each supported architecture and their supported
17594 v8: supersparc, hypersparc, leon
17595 sparclite: f930, f934, sparclite86x
17597 v9: ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
17600 By default (unless configured otherwise), GCC generates code for the V7
17601 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17602 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17603 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17604 SPARCStation 1, 2, IPX etc.
17606 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17607 architecture. The only difference from V7 code is that the compiler emits
17608 the integer multiply and integer divide instructions which exist in SPARC-V8
17609 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17610 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17613 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17614 the SPARC architecture. This adds the integer multiply, integer divide step
17615 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17616 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17617 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17618 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17619 MB86934 chip, which is the more recent SPARClite with FPU@.
17621 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17622 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17623 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17624 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17625 optimizes it for the TEMIC SPARClet chip.
17627 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17628 architecture. This adds 64-bit integer and floating-point move instructions,
17629 3 additional floating-point condition code registers and conditional move
17630 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17631 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17632 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17633 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17634 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17635 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17636 additionally optimizes it for Sun UltraSPARC T2 chips. With
17637 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
17638 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
17639 additionally optimizes it for Sun UltraSPARC T4 chips.
17641 @item -mtune=@var{cpu_type}
17643 Set the instruction scheduling parameters for machine type
17644 @var{cpu_type}, but do not set the instruction set or register set that the
17645 option @option{-mcpu=@var{cpu_type}} would.
17647 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17648 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17649 that select a particular CPU implementation. Those are @samp{cypress},
17650 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17651 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17652 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
17653 native Solaris and GNU/Linux toolchains, @samp{native} can also be used.
17658 @opindex mno-v8plus
17659 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17660 difference from the V8 ABI is that the global and out registers are
17661 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17662 mode for all SPARC-V9 processors.
17668 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17669 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17675 With @option{-mvis2}, GCC generates code that takes advantage of
17676 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
17677 default is @option{-mvis2} when targetting a cpu that supports such
17678 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
17679 also sets @option{-mvis}.
17685 With @option{-mvis3}, GCC generates code that takes advantage of
17686 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
17687 default is @option{-mvis3} when targetting a cpu that supports such
17688 instructions, such as niagara-3 and later. Setting @option{-mvis3}
17689 also sets @option{-mvis2} and @option{-mvis}.
17695 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
17696 population count instruction. The default is @option{-mpopc}
17697 when targetting a cpu that supports such instructions, such as Niagara-2 and
17704 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
17705 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
17706 when targetting a cpu that supports such instructions, such as Niagara-3 and
17710 @opindex mfix-at697f
17711 Enable the documented workaround for the single erratum of the Atmel AT697F
17712 processor (which corresponds to erratum #13 of the AT697E processor).
17715 These @samp{-m} options are supported in addition to the above
17716 on SPARC-V9 processors in 64-bit environments:
17719 @item -mlittle-endian
17720 @opindex mlittle-endian
17721 Generate code for a processor running in little-endian mode. It is only
17722 available for a few configurations and most notably not on Solaris and Linux.
17728 Generate code for a 32-bit or 64-bit environment.
17729 The 32-bit environment sets int, long and pointer to 32 bits.
17730 The 64-bit environment sets int to 32 bits and long and pointer
17733 @item -mcmodel=medlow
17734 @opindex mcmodel=medlow
17735 Generate code for the Medium/Low code model: 64-bit addresses, programs
17736 must be linked in the low 32 bits of memory. Programs can be statically
17737 or dynamically linked.
17739 @item -mcmodel=medmid
17740 @opindex mcmodel=medmid
17741 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17742 must be linked in the low 44 bits of memory, the text and data segments must
17743 be less than 2GB in size and the data segment must be located within 2GB of
17746 @item -mcmodel=medany
17747 @opindex mcmodel=medany
17748 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17749 may be linked anywhere in memory, the text and data segments must be less
17750 than 2GB in size and the data segment must be located within 2GB of the
17753 @item -mcmodel=embmedany
17754 @opindex mcmodel=embmedany
17755 Generate code for the Medium/Anywhere code model for embedded systems:
17756 64-bit addresses, the text and data segments must be less than 2GB in
17757 size, both starting anywhere in memory (determined at link time). The
17758 global register %g4 points to the base of the data segment. Programs
17759 are statically linked and PIC is not supported.
17762 @itemx -mno-stack-bias
17763 @opindex mstack-bias
17764 @opindex mno-stack-bias
17765 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17766 frame pointer if present, are offset by @minus{}2047 which must be added back
17767 when making stack frame references. This is the default in 64-bit mode.
17768 Otherwise, assume no such offset is present.
17772 @subsection SPU Options
17773 @cindex SPU options
17775 These @samp{-m} options are supported on the SPU:
17779 @itemx -merror-reloc
17780 @opindex mwarn-reloc
17781 @opindex merror-reloc
17783 The loader for SPU does not handle dynamic relocations. By default, GCC
17784 will give an error when it generates code that requires a dynamic
17785 relocation. @option{-mno-error-reloc} disables the error,
17786 @option{-mwarn-reloc} will generate a warning instead.
17789 @itemx -munsafe-dma
17791 @opindex munsafe-dma
17793 Instructions which initiate or test completion of DMA must not be
17794 reordered with respect to loads and stores of the memory which is being
17795 accessed. Users typically address this problem using the volatile
17796 keyword, but that can lead to inefficient code in places where the
17797 memory is known to not change. Rather than mark the memory as volatile
17798 we treat the DMA instructions as potentially effecting all memory. With
17799 @option{-munsafe-dma} users must use the volatile keyword to protect
17802 @item -mbranch-hints
17803 @opindex mbranch-hints
17805 By default, GCC will generate a branch hint instruction to avoid
17806 pipeline stalls for always taken or probably taken branches. A hint
17807 will not be generated closer than 8 instructions away from its branch.
17808 There is little reason to disable them, except for debugging purposes,
17809 or to make an object a little bit smaller.
17813 @opindex msmall-mem
17814 @opindex mlarge-mem
17816 By default, GCC generates code assuming that addresses are never larger
17817 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17818 a full 32 bit address.
17823 By default, GCC links against startup code that assumes the SPU-style
17824 main function interface (which has an unconventional parameter list).
17825 With @option{-mstdmain}, GCC will link your program against startup
17826 code that assumes a C99-style interface to @code{main}, including a
17827 local copy of @code{argv} strings.
17829 @item -mfixed-range=@var{register-range}
17830 @opindex mfixed-range
17831 Generate code treating the given register range as fixed registers.
17832 A fixed register is one that the register allocator can not use. This is
17833 useful when compiling kernel code. A register range is specified as
17834 two registers separated by a dash. Multiple register ranges can be
17835 specified separated by a comma.
17841 Compile code assuming that pointers to the PPU address space accessed
17842 via the @code{__ea} named address space qualifier are either 32 or 64
17843 bits wide. The default is 32 bits. As this is an ABI changing option,
17844 all object code in an executable must be compiled with the same setting.
17846 @item -maddress-space-conversion
17847 @itemx -mno-address-space-conversion
17848 @opindex maddress-space-conversion
17849 @opindex mno-address-space-conversion
17850 Allow/disallow treating the @code{__ea} address space as superset
17851 of the generic address space. This enables explicit type casts
17852 between @code{__ea} and generic pointer as well as implicit
17853 conversions of generic pointers to @code{__ea} pointers. The
17854 default is to allow address space pointer conversions.
17856 @item -mcache-size=@var{cache-size}
17857 @opindex mcache-size
17858 This option controls the version of libgcc that the compiler links to an
17859 executable and selects a software-managed cache for accessing variables
17860 in the @code{__ea} address space with a particular cache size. Possible
17861 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17862 and @samp{128}. The default cache size is 64KB.
17864 @item -matomic-updates
17865 @itemx -mno-atomic-updates
17866 @opindex matomic-updates
17867 @opindex mno-atomic-updates
17868 This option controls the version of libgcc that the compiler links to an
17869 executable and selects whether atomic updates to the software-managed
17870 cache of PPU-side variables are used. If you use atomic updates, changes
17871 to a PPU variable from SPU code using the @code{__ea} named address space
17872 qualifier will not interfere with changes to other PPU variables residing
17873 in the same cache line from PPU code. If you do not use atomic updates,
17874 such interference may occur; however, writing back cache lines will be
17875 more efficient. The default behavior is to use atomic updates.
17878 @itemx -mdual-nops=@var{n}
17879 @opindex mdual-nops
17880 By default, GCC will insert nops to increase dual issue when it expects
17881 it to increase performance. @var{n} can be a value from 0 to 10. A
17882 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17883 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17885 @item -mhint-max-nops=@var{n}
17886 @opindex mhint-max-nops
17887 Maximum number of nops to insert for a branch hint. A branch hint must
17888 be at least 8 instructions away from the branch it is effecting. GCC
17889 will insert up to @var{n} nops to enforce this, otherwise it will not
17890 generate the branch hint.
17892 @item -mhint-max-distance=@var{n}
17893 @opindex mhint-max-distance
17894 The encoding of the branch hint instruction limits the hint to be within
17895 256 instructions of the branch it is effecting. By default, GCC makes
17896 sure it is within 125.
17899 @opindex msafe-hints
17900 Work around a hardware bug which causes the SPU to stall indefinitely.
17901 By default, GCC will insert the @code{hbrp} instruction to make sure
17902 this stall won't happen.
17906 @node System V Options
17907 @subsection Options for System V
17909 These additional options are available on System V Release 4 for
17910 compatibility with other compilers on those systems:
17915 Create a shared object.
17916 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17920 Identify the versions of each tool used by the compiler, in a
17921 @code{.ident} assembler directive in the output.
17925 Refrain from adding @code{.ident} directives to the output file (this is
17928 @item -YP,@var{dirs}
17930 Search the directories @var{dirs}, and no others, for libraries
17931 specified with @option{-l}.
17933 @item -Ym,@var{dir}
17935 Look in the directory @var{dir} to find the M4 preprocessor.
17936 The assembler uses this option.
17937 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17938 @c the generic assembler that comes with Solaris takes just -Ym.
17942 @subsection V850 Options
17943 @cindex V850 Options
17945 These @samp{-m} options are defined for V850 implementations:
17949 @itemx -mno-long-calls
17950 @opindex mlong-calls
17951 @opindex mno-long-calls
17952 Treat all calls as being far away (near). If calls are assumed to be
17953 far away, the compiler will always load the functions address up into a
17954 register, and call indirect through the pointer.
17960 Do not optimize (do optimize) basic blocks that use the same index
17961 pointer 4 or more times to copy pointer into the @code{ep} register, and
17962 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17963 option is on by default if you optimize.
17965 @item -mno-prolog-function
17966 @itemx -mprolog-function
17967 @opindex mno-prolog-function
17968 @opindex mprolog-function
17969 Do not use (do use) external functions to save and restore registers
17970 at the prologue and epilogue of a function. The external functions
17971 are slower, but use less code space if more than one function saves
17972 the same number of registers. The @option{-mprolog-function} option
17973 is on by default if you optimize.
17977 Try to make the code as small as possible. At present, this just turns
17978 on the @option{-mep} and @option{-mprolog-function} options.
17980 @item -mtda=@var{n}
17982 Put static or global variables whose size is @var{n} bytes or less into
17983 the tiny data area that register @code{ep} points to. The tiny data
17984 area can hold up to 256 bytes in total (128 bytes for byte references).
17986 @item -msda=@var{n}
17988 Put static or global variables whose size is @var{n} bytes or less into
17989 the small data area that register @code{gp} points to. The small data
17990 area can hold up to 64 kilobytes.
17992 @item -mzda=@var{n}
17994 Put static or global variables whose size is @var{n} bytes or less into
17995 the first 32 kilobytes of memory.
17999 Specify that the target processor is the V850.
18002 @opindex mbig-switch
18003 Generate code suitable for big switch tables. Use this option only if
18004 the assembler/linker complain about out of range branches within a switch
18009 This option will cause r2 and r5 to be used in the code generated by
18010 the compiler. This setting is the default.
18012 @item -mno-app-regs
18013 @opindex mno-app-regs
18014 This option will cause r2 and r5 to be treated as fixed registers.
18018 Specify that the target processor is the V850E2V3. The preprocessor
18019 constants @samp{__v850e2v3__} will be defined if
18020 this option is used.
18024 Specify that the target processor is the V850E2. The preprocessor
18025 constants @samp{__v850e2__} will be defined if this option is used.
18029 Specify that the target processor is the V850E1. The preprocessor
18030 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
18031 this option is used.
18035 Specify that the target processor is the V850ES. This is an alias for
18036 the @option{-mv850e1} option.
18040 Specify that the target processor is the V850E@. The preprocessor
18041 constant @samp{__v850e__} will be defined if this option is used.
18043 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
18044 nor @option{-mv850e2} nor @option{-mv850e2v3}
18045 are defined then a default target processor will be chosen and the
18046 relevant @samp{__v850*__} preprocessor constant will be defined.
18048 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
18049 defined, regardless of which processor variant is the target.
18051 @item -mdisable-callt
18052 @opindex mdisable-callt
18053 This option will suppress generation of the CALLT instruction for the
18054 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
18055 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
18060 @subsection VAX Options
18061 @cindex VAX options
18063 These @samp{-m} options are defined for the VAX:
18068 Do not output certain jump instructions (@code{aobleq} and so on)
18069 that the Unix assembler for the VAX cannot handle across long
18074 Do output those jump instructions, on the assumption that you
18075 will assemble with the GNU assembler.
18079 Output code for g-format floating point numbers instead of d-format.
18082 @node VxWorks Options
18083 @subsection VxWorks Options
18084 @cindex VxWorks Options
18086 The options in this section are defined for all VxWorks targets.
18087 Options specific to the target hardware are listed with the other
18088 options for that target.
18093 GCC can generate code for both VxWorks kernels and real time processes
18094 (RTPs). This option switches from the former to the latter. It also
18095 defines the preprocessor macro @code{__RTP__}.
18098 @opindex non-static
18099 Link an RTP executable against shared libraries rather than static
18100 libraries. The options @option{-static} and @option{-shared} can
18101 also be used for RTPs (@pxref{Link Options}); @option{-static}
18108 These options are passed down to the linker. They are defined for
18109 compatibility with Diab.
18112 @opindex Xbind-lazy
18113 Enable lazy binding of function calls. This option is equivalent to
18114 @option{-Wl,-z,now} and is defined for compatibility with Diab.
18118 Disable lazy binding of function calls. This option is the default and
18119 is defined for compatibility with Diab.
18122 @node x86-64 Options
18123 @subsection x86-64 Options
18124 @cindex x86-64 options
18126 These are listed under @xref{i386 and x86-64 Options}.
18128 @node Xstormy16 Options
18129 @subsection Xstormy16 Options
18130 @cindex Xstormy16 Options
18132 These options are defined for Xstormy16:
18137 Choose startup files and linker script suitable for the simulator.
18140 @node Xtensa Options
18141 @subsection Xtensa Options
18142 @cindex Xtensa Options
18144 These options are supported for Xtensa targets:
18148 @itemx -mno-const16
18150 @opindex mno-const16
18151 Enable or disable use of @code{CONST16} instructions for loading
18152 constant values. The @code{CONST16} instruction is currently not a
18153 standard option from Tensilica. When enabled, @code{CONST16}
18154 instructions are always used in place of the standard @code{L32R}
18155 instructions. The use of @code{CONST16} is enabled by default only if
18156 the @code{L32R} instruction is not available.
18159 @itemx -mno-fused-madd
18160 @opindex mfused-madd
18161 @opindex mno-fused-madd
18162 Enable or disable use of fused multiply/add and multiply/subtract
18163 instructions in the floating-point option. This has no effect if the
18164 floating-point option is not also enabled. Disabling fused multiply/add
18165 and multiply/subtract instructions forces the compiler to use separate
18166 instructions for the multiply and add/subtract operations. This may be
18167 desirable in some cases where strict IEEE 754-compliant results are
18168 required: the fused multiply add/subtract instructions do not round the
18169 intermediate result, thereby producing results with @emph{more} bits of
18170 precision than specified by the IEEE standard. Disabling fused multiply
18171 add/subtract instructions also ensures that the program output is not
18172 sensitive to the compiler's ability to combine multiply and add/subtract
18175 @item -mserialize-volatile
18176 @itemx -mno-serialize-volatile
18177 @opindex mserialize-volatile
18178 @opindex mno-serialize-volatile
18179 When this option is enabled, GCC inserts @code{MEMW} instructions before
18180 @code{volatile} memory references to guarantee sequential consistency.
18181 The default is @option{-mserialize-volatile}. Use
18182 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
18184 @item -mforce-no-pic
18185 @opindex mforce-no-pic
18186 For targets, like GNU/Linux, where all user-mode Xtensa code must be
18187 position-independent code (PIC), this option disables PIC for compiling
18190 @item -mtext-section-literals
18191 @itemx -mno-text-section-literals
18192 @opindex mtext-section-literals
18193 @opindex mno-text-section-literals
18194 Control the treatment of literal pools. The default is
18195 @option{-mno-text-section-literals}, which places literals in a separate
18196 section in the output file. This allows the literal pool to be placed
18197 in a data RAM/ROM, and it also allows the linker to combine literal
18198 pools from separate object files to remove redundant literals and
18199 improve code size. With @option{-mtext-section-literals}, the literals
18200 are interspersed in the text section in order to keep them as close as
18201 possible to their references. This may be necessary for large assembly
18204 @item -mtarget-align
18205 @itemx -mno-target-align
18206 @opindex mtarget-align
18207 @opindex mno-target-align
18208 When this option is enabled, GCC instructs the assembler to
18209 automatically align instructions to reduce branch penalties at the
18210 expense of some code density. The assembler attempts to widen density
18211 instructions to align branch targets and the instructions following call
18212 instructions. If there are not enough preceding safe density
18213 instructions to align a target, no widening will be performed. The
18214 default is @option{-mtarget-align}. These options do not affect the
18215 treatment of auto-aligned instructions like @code{LOOP}, which the
18216 assembler will always align, either by widening density instructions or
18217 by inserting no-op instructions.
18220 @itemx -mno-longcalls
18221 @opindex mlongcalls
18222 @opindex mno-longcalls
18223 When this option is enabled, GCC instructs the assembler to translate
18224 direct calls to indirect calls unless it can determine that the target
18225 of a direct call is in the range allowed by the call instruction. This
18226 translation typically occurs for calls to functions in other source
18227 files. Specifically, the assembler translates a direct @code{CALL}
18228 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
18229 The default is @option{-mno-longcalls}. This option should be used in
18230 programs where the call target can potentially be out of range. This
18231 option is implemented in the assembler, not the compiler, so the
18232 assembly code generated by GCC will still show direct call
18233 instructions---look at the disassembled object code to see the actual
18234 instructions. Note that the assembler will use an indirect call for
18235 every cross-file call, not just those that really will be out of range.
18238 @node zSeries Options
18239 @subsection zSeries Options
18240 @cindex zSeries options
18242 These are listed under @xref{S/390 and zSeries Options}.
18244 @node Code Gen Options
18245 @section Options for Code Generation Conventions
18246 @cindex code generation conventions
18247 @cindex options, code generation
18248 @cindex run-time options
18250 These machine-independent options control the interface conventions
18251 used in code generation.
18253 Most of them have both positive and negative forms; the negative form
18254 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
18255 one of the forms is listed---the one which is not the default. You
18256 can figure out the other form by either removing @samp{no-} or adding
18260 @item -fbounds-check
18261 @opindex fbounds-check
18262 For front-ends that support it, generate additional code to check that
18263 indices used to access arrays are within the declared range. This is
18264 currently only supported by the Java and Fortran front-ends, where
18265 this option defaults to true and false respectively.
18269 This option generates traps for signed overflow on addition, subtraction,
18270 multiplication operations.
18274 This option instructs the compiler to assume that signed arithmetic
18275 overflow of addition, subtraction and multiplication wraps around
18276 using twos-complement representation. This flag enables some optimizations
18277 and disables others. This option is enabled by default for the Java
18278 front-end, as required by the Java language specification.
18281 @opindex fexceptions
18282 Enable exception handling. Generates extra code needed to propagate
18283 exceptions. For some targets, this implies GCC will generate frame
18284 unwind information for all functions, which can produce significant data
18285 size overhead, although it does not affect execution. If you do not
18286 specify this option, GCC will enable it by default for languages like
18287 C++ which normally require exception handling, and disable it for
18288 languages like C that do not normally require it. However, you may need
18289 to enable this option when compiling C code that needs to interoperate
18290 properly with exception handlers written in C++. You may also wish to
18291 disable this option if you are compiling older C++ programs that don't
18292 use exception handling.
18294 @item -fnon-call-exceptions
18295 @opindex fnon-call-exceptions
18296 Generate code that allows trapping instructions to throw exceptions.
18297 Note that this requires platform-specific runtime support that does
18298 not exist everywhere. Moreover, it only allows @emph{trapping}
18299 instructions to throw exceptions, i.e.@: memory references or floating
18300 point instructions. It does not allow exceptions to be thrown from
18301 arbitrary signal handlers such as @code{SIGALRM}.
18303 @item -funwind-tables
18304 @opindex funwind-tables
18305 Similar to @option{-fexceptions}, except that it will just generate any needed
18306 static data, but will not affect the generated code in any other way.
18307 You will normally not enable this option; instead, a language processor
18308 that needs this handling would enable it on your behalf.
18310 @item -fasynchronous-unwind-tables
18311 @opindex fasynchronous-unwind-tables
18312 Generate unwind table in dwarf2 format, if supported by target machine. The
18313 table is exact at each instruction boundary, so it can be used for stack
18314 unwinding from asynchronous events (such as debugger or garbage collector).
18316 @item -fpcc-struct-return
18317 @opindex fpcc-struct-return
18318 Return ``short'' @code{struct} and @code{union} values in memory like
18319 longer ones, rather than in registers. This convention is less
18320 efficient, but it has the advantage of allowing intercallability between
18321 GCC-compiled files and files compiled with other compilers, particularly
18322 the Portable C Compiler (pcc).
18324 The precise convention for returning structures in memory depends
18325 on the target configuration macros.
18327 Short structures and unions are those whose size and alignment match
18328 that of some integer type.
18330 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
18331 switch is not binary compatible with code compiled with the
18332 @option{-freg-struct-return} switch.
18333 Use it to conform to a non-default application binary interface.
18335 @item -freg-struct-return
18336 @opindex freg-struct-return
18337 Return @code{struct} and @code{union} values in registers when possible.
18338 This is more efficient for small structures than
18339 @option{-fpcc-struct-return}.
18341 If you specify neither @option{-fpcc-struct-return} nor
18342 @option{-freg-struct-return}, GCC defaults to whichever convention is
18343 standard for the target. If there is no standard convention, GCC
18344 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
18345 the principal compiler. In those cases, we can choose the standard, and
18346 we chose the more efficient register return alternative.
18348 @strong{Warning:} code compiled with the @option{-freg-struct-return}
18349 switch is not binary compatible with code compiled with the
18350 @option{-fpcc-struct-return} switch.
18351 Use it to conform to a non-default application binary interface.
18353 @item -fshort-enums
18354 @opindex fshort-enums
18355 Allocate to an @code{enum} type only as many bytes as it needs for the
18356 declared range of possible values. Specifically, the @code{enum} type
18357 will be equivalent to the smallest integer type which has enough room.
18359 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
18360 code that is not binary compatible with code generated without that switch.
18361 Use it to conform to a non-default application binary interface.
18363 @item -fshort-double
18364 @opindex fshort-double
18365 Use the same size for @code{double} as for @code{float}.
18367 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
18368 code that is not binary compatible with code generated without that switch.
18369 Use it to conform to a non-default application binary interface.
18371 @item -fshort-wchar
18372 @opindex fshort-wchar
18373 Override the underlying type for @samp{wchar_t} to be @samp{short
18374 unsigned int} instead of the default for the target. This option is
18375 useful for building programs to run under WINE@.
18377 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18378 code that is not binary compatible with code generated without that switch.
18379 Use it to conform to a non-default application binary interface.
18382 @opindex fno-common
18383 In C code, controls the placement of uninitialized global variables.
18384 Unix C compilers have traditionally permitted multiple definitions of
18385 such variables in different compilation units by placing the variables
18387 This is the behavior specified by @option{-fcommon}, and is the default
18388 for GCC on most targets.
18389 On the other hand, this behavior is not required by ISO C, and on some
18390 targets may carry a speed or code size penalty on variable references.
18391 The @option{-fno-common} option specifies that the compiler should place
18392 uninitialized global variables in the data section of the object file,
18393 rather than generating them as common blocks.
18394 This has the effect that if the same variable is declared
18395 (without @code{extern}) in two different compilations,
18396 you will get a multiple-definition error when you link them.
18397 In this case, you must compile with @option{-fcommon} instead.
18398 Compiling with @option{-fno-common} is useful on targets for which
18399 it provides better performance, or if you wish to verify that the
18400 program will work on other systems which always treat uninitialized
18401 variable declarations this way.
18405 Ignore the @samp{#ident} directive.
18407 @item -finhibit-size-directive
18408 @opindex finhibit-size-directive
18409 Don't output a @code{.size} assembler directive, or anything else that
18410 would cause trouble if the function is split in the middle, and the
18411 two halves are placed at locations far apart in memory. This option is
18412 used when compiling @file{crtstuff.c}; you should not need to use it
18415 @item -fverbose-asm
18416 @opindex fverbose-asm
18417 Put extra commentary information in the generated assembly code to
18418 make it more readable. This option is generally only of use to those
18419 who actually need to read the generated assembly code (perhaps while
18420 debugging the compiler itself).
18422 @option{-fno-verbose-asm}, the default, causes the
18423 extra information to be omitted and is useful when comparing two assembler
18426 @item -frecord-gcc-switches
18427 @opindex frecord-gcc-switches
18428 This switch causes the command line that was used to invoke the
18429 compiler to be recorded into the object file that is being created.
18430 This switch is only implemented on some targets and the exact format
18431 of the recording is target and binary file format dependent, but it
18432 usually takes the form of a section containing ASCII text. This
18433 switch is related to the @option{-fverbose-asm} switch, but that
18434 switch only records information in the assembler output file as
18435 comments, so it never reaches the object file.
18436 See also @option{-grecord-gcc-switches} for another
18437 way of storing compiler options into the object file.
18441 @cindex global offset table
18443 Generate position-independent code (PIC) suitable for use in a shared
18444 library, if supported for the target machine. Such code accesses all
18445 constant addresses through a global offset table (GOT)@. The dynamic
18446 loader resolves the GOT entries when the program starts (the dynamic
18447 loader is not part of GCC; it is part of the operating system). If
18448 the GOT size for the linked executable exceeds a machine-specific
18449 maximum size, you get an error message from the linker indicating that
18450 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
18451 instead. (These maximums are 8k on the SPARC and 32k
18452 on the m68k and RS/6000. The 386 has no such limit.)
18454 Position-independent code requires special support, and therefore works
18455 only on certain machines. For the 386, GCC supports PIC for System V
18456 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
18457 position-independent.
18459 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18464 If supported for the target machine, emit position-independent code,
18465 suitable for dynamic linking and avoiding any limit on the size of the
18466 global offset table. This option makes a difference on the m68k,
18467 PowerPC and SPARC@.
18469 Position-independent code requires special support, and therefore works
18470 only on certain machines.
18472 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18479 These options are similar to @option{-fpic} and @option{-fPIC}, but
18480 generated position independent code can be only linked into executables.
18481 Usually these options are used when @option{-pie} GCC option will be
18482 used during linking.
18484 @option{-fpie} and @option{-fPIE} both define the macros
18485 @code{__pie__} and @code{__PIE__}. The macros have the value 1
18486 for @option{-fpie} and 2 for @option{-fPIE}.
18488 @item -fno-jump-tables
18489 @opindex fno-jump-tables
18490 Do not use jump tables for switch statements even where it would be
18491 more efficient than other code generation strategies. This option is
18492 of use in conjunction with @option{-fpic} or @option{-fPIC} for
18493 building code which forms part of a dynamic linker and cannot
18494 reference the address of a jump table. On some targets, jump tables
18495 do not require a GOT and this option is not needed.
18497 @item -ffixed-@var{reg}
18499 Treat the register named @var{reg} as a fixed register; generated code
18500 should never refer to it (except perhaps as a stack pointer, frame
18501 pointer or in some other fixed role).
18503 @var{reg} must be the name of a register. The register names accepted
18504 are machine-specific and are defined in the @code{REGISTER_NAMES}
18505 macro in the machine description macro file.
18507 This flag does not have a negative form, because it specifies a
18510 @item -fcall-used-@var{reg}
18511 @opindex fcall-used
18512 Treat the register named @var{reg} as an allocable register that is
18513 clobbered by function calls. It may be allocated for temporaries or
18514 variables that do not live across a call. Functions compiled this way
18515 will not save and restore the register @var{reg}.
18517 It is an error to used this flag with the frame pointer or stack pointer.
18518 Use of this flag for other registers that have fixed pervasive roles in
18519 the machine's execution model will produce disastrous results.
18521 This flag does not have a negative form, because it specifies a
18524 @item -fcall-saved-@var{reg}
18525 @opindex fcall-saved
18526 Treat the register named @var{reg} as an allocable register saved by
18527 functions. It may be allocated even for temporaries or variables that
18528 live across a call. Functions compiled this way will save and restore
18529 the register @var{reg} if they use it.
18531 It is an error to used this flag with the frame pointer or stack pointer.
18532 Use of this flag for other registers that have fixed pervasive roles in
18533 the machine's execution model will produce disastrous results.
18535 A different sort of disaster will result from the use of this flag for
18536 a register in which function values may be returned.
18538 This flag does not have a negative form, because it specifies a
18541 @item -fpack-struct[=@var{n}]
18542 @opindex fpack-struct
18543 Without a value specified, pack all structure members together without
18544 holes. When a value is specified (which must be a small power of two), pack
18545 structure members according to this value, representing the maximum
18546 alignment (that is, objects with default alignment requirements larger than
18547 this will be output potentially unaligned at the next fitting location.
18549 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18550 code that is not binary compatible with code generated without that switch.
18551 Additionally, it makes the code suboptimal.
18552 Use it to conform to a non-default application binary interface.
18554 @item -finstrument-functions
18555 @opindex finstrument-functions
18556 Generate instrumentation calls for entry and exit to functions. Just
18557 after function entry and just before function exit, the following
18558 profiling functions will be called with the address of the current
18559 function and its call site. (On some platforms,
18560 @code{__builtin_return_address} does not work beyond the current
18561 function, so the call site information may not be available to the
18562 profiling functions otherwise.)
18565 void __cyg_profile_func_enter (void *this_fn,
18567 void __cyg_profile_func_exit (void *this_fn,
18571 The first argument is the address of the start of the current function,
18572 which may be looked up exactly in the symbol table.
18574 This instrumentation is also done for functions expanded inline in other
18575 functions. The profiling calls will indicate where, conceptually, the
18576 inline function is entered and exited. This means that addressable
18577 versions of such functions must be available. If all your uses of a
18578 function are expanded inline, this may mean an additional expansion of
18579 code size. If you use @samp{extern inline} in your C code, an
18580 addressable version of such functions must be provided. (This is
18581 normally the case anyways, but if you get lucky and the optimizer always
18582 expands the functions inline, you might have gotten away without
18583 providing static copies.)
18585 A function may be given the attribute @code{no_instrument_function}, in
18586 which case this instrumentation will not be done. This can be used, for
18587 example, for the profiling functions listed above, high-priority
18588 interrupt routines, and any functions from which the profiling functions
18589 cannot safely be called (perhaps signal handlers, if the profiling
18590 routines generate output or allocate memory).
18592 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18593 @opindex finstrument-functions-exclude-file-list
18595 Set the list of functions that are excluded from instrumentation (see
18596 the description of @code{-finstrument-functions}). If the file that
18597 contains a function definition matches with one of @var{file}, then
18598 that function is not instrumented. The match is done on substrings:
18599 if the @var{file} parameter is a substring of the file name, it is
18600 considered to be a match.
18605 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18609 will exclude any inline function defined in files whose pathnames
18610 contain @code{/bits/stl} or @code{include/sys}.
18612 If, for some reason, you want to include letter @code{','} in one of
18613 @var{sym}, write @code{'\,'}. For example,
18614 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18615 (note the single quote surrounding the option).
18617 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18618 @opindex finstrument-functions-exclude-function-list
18620 This is similar to @code{-finstrument-functions-exclude-file-list},
18621 but this option sets the list of function names to be excluded from
18622 instrumentation. The function name to be matched is its user-visible
18623 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18624 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18625 match is done on substrings: if the @var{sym} parameter is a substring
18626 of the function name, it is considered to be a match. For C99 and C++
18627 extended identifiers, the function name must be given in UTF-8, not
18628 using universal character names.
18630 @item -fstack-check
18631 @opindex fstack-check
18632 Generate code to verify that you do not go beyond the boundary of the
18633 stack. You should specify this flag if you are running in an
18634 environment with multiple threads, but only rarely need to specify it in
18635 a single-threaded environment since stack overflow is automatically
18636 detected on nearly all systems if there is only one stack.
18638 Note that this switch does not actually cause checking to be done; the
18639 operating system or the language runtime must do that. The switch causes
18640 generation of code to ensure that they see the stack being extended.
18642 You can additionally specify a string parameter: @code{no} means no
18643 checking, @code{generic} means force the use of old-style checking,
18644 @code{specific} means use the best checking method and is equivalent
18645 to bare @option{-fstack-check}.
18647 Old-style checking is a generic mechanism that requires no specific
18648 target support in the compiler but comes with the following drawbacks:
18652 Modified allocation strategy for large objects: they will always be
18653 allocated dynamically if their size exceeds a fixed threshold.
18656 Fixed limit on the size of the static frame of functions: when it is
18657 topped by a particular function, stack checking is not reliable and
18658 a warning is issued by the compiler.
18661 Inefficiency: because of both the modified allocation strategy and the
18662 generic implementation, the performances of the code are hampered.
18665 Note that old-style stack checking is also the fallback method for
18666 @code{specific} if no target support has been added in the compiler.
18668 @item -fstack-limit-register=@var{reg}
18669 @itemx -fstack-limit-symbol=@var{sym}
18670 @itemx -fno-stack-limit
18671 @opindex fstack-limit-register
18672 @opindex fstack-limit-symbol
18673 @opindex fno-stack-limit
18674 Generate code to ensure that the stack does not grow beyond a certain value,
18675 either the value of a register or the address of a symbol. If the stack
18676 would grow beyond the value, a signal is raised. For most targets,
18677 the signal is raised before the stack overruns the boundary, so
18678 it is possible to catch the signal without taking special precautions.
18680 For instance, if the stack starts at absolute address @samp{0x80000000}
18681 and grows downwards, you can use the flags
18682 @option{-fstack-limit-symbol=__stack_limit} and
18683 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18684 of 128KB@. Note that this may only work with the GNU linker.
18686 @item -fsplit-stack
18687 @opindex fsplit-stack
18688 Generate code to automatically split the stack before it overflows.
18689 The resulting program has a discontiguous stack which can only
18690 overflow if the program is unable to allocate any more memory. This
18691 is most useful when running threaded programs, as it is no longer
18692 necessary to calculate a good stack size to use for each thread. This
18693 is currently only implemented for the i386 and x86_64 backends running
18696 When code compiled with @option{-fsplit-stack} calls code compiled
18697 without @option{-fsplit-stack}, there may not be much stack space
18698 available for the latter code to run. If compiling all code,
18699 including library code, with @option{-fsplit-stack} is not an option,
18700 then the linker can fix up these calls so that the code compiled
18701 without @option{-fsplit-stack} always has a large stack. Support for
18702 this is implemented in the gold linker in GNU binutils release 2.21
18705 @item -fleading-underscore
18706 @opindex fleading-underscore
18707 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18708 change the way C symbols are represented in the object file. One use
18709 is to help link with legacy assembly code.
18711 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18712 generate code that is not binary compatible with code generated without that
18713 switch. Use it to conform to a non-default application binary interface.
18714 Not all targets provide complete support for this switch.
18716 @item -ftls-model=@var{model}
18717 @opindex ftls-model
18718 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18719 The @var{model} argument should be one of @code{global-dynamic},
18720 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18722 The default without @option{-fpic} is @code{initial-exec}; with
18723 @option{-fpic} the default is @code{global-dynamic}.
18725 @item -fvisibility=@var{default|internal|hidden|protected}
18726 @opindex fvisibility
18727 Set the default ELF image symbol visibility to the specified option---all
18728 symbols will be marked with this unless overridden within the code.
18729 Using this feature can very substantially improve linking and
18730 load times of shared object libraries, produce more optimized
18731 code, provide near-perfect API export and prevent symbol clashes.
18732 It is @strong{strongly} recommended that you use this in any shared objects
18735 Despite the nomenclature, @code{default} always means public; i.e.,
18736 available to be linked against from outside the shared object.
18737 @code{protected} and @code{internal} are pretty useless in real-world
18738 usage so the only other commonly used option will be @code{hidden}.
18739 The default if @option{-fvisibility} isn't specified is
18740 @code{default}, i.e., make every
18741 symbol public---this causes the same behavior as previous versions of
18744 A good explanation of the benefits offered by ensuring ELF
18745 symbols have the correct visibility is given by ``How To Write
18746 Shared Libraries'' by Ulrich Drepper (which can be found at
18747 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18748 solution made possible by this option to marking things hidden when
18749 the default is public is to make the default hidden and mark things
18750 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18751 and @code{__attribute__ ((visibility("default")))} instead of
18752 @code{__declspec(dllexport)} you get almost identical semantics with
18753 identical syntax. This is a great boon to those working with
18754 cross-platform projects.
18756 For those adding visibility support to existing code, you may find
18757 @samp{#pragma GCC visibility} of use. This works by you enclosing
18758 the declarations you wish to set visibility for with (for example)
18759 @samp{#pragma GCC visibility push(hidden)} and
18760 @samp{#pragma GCC visibility pop}.
18761 Bear in mind that symbol visibility should be viewed @strong{as
18762 part of the API interface contract} and thus all new code should
18763 always specify visibility when it is not the default; i.e., declarations
18764 only for use within the local DSO should @strong{always} be marked explicitly
18765 as hidden as so to avoid PLT indirection overheads---making this
18766 abundantly clear also aids readability and self-documentation of the code.
18767 Note that due to ISO C++ specification requirements, operator new and
18768 operator delete must always be of default visibility.
18770 Be aware that headers from outside your project, in particular system
18771 headers and headers from any other library you use, may not be
18772 expecting to be compiled with visibility other than the default. You
18773 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18774 before including any such headers.
18776 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18777 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18778 no modifications. However, this means that calls to @samp{extern}
18779 functions with no explicit visibility will use the PLT, so it is more
18780 effective to use @samp{__attribute ((visibility))} and/or
18781 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18782 declarations should be treated as hidden.
18784 Note that @samp{-fvisibility} does affect C++ vague linkage
18785 entities. This means that, for instance, an exception class that will
18786 be thrown between DSOs must be explicitly marked with default
18787 visibility so that the @samp{type_info} nodes will be unified between
18790 An overview of these techniques, their benefits and how to use them
18791 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18793 @item -fstrict-volatile-bitfields
18794 @opindex fstrict-volatile-bitfields
18795 This option should be used if accesses to volatile bitfields (or other
18796 structure fields, although the compiler usually honors those types
18797 anyway) should use a single access of the width of the
18798 field's type, aligned to a natural alignment if possible. For
18799 example, targets with memory-mapped peripheral registers might require
18800 all such accesses to be 16 bits wide; with this flag the user could
18801 declare all peripheral bitfields as ``unsigned short'' (assuming short
18802 is 16 bits on these targets) to force GCC to use 16 bit accesses
18803 instead of, perhaps, a more efficient 32 bit access.
18805 If this option is disabled, the compiler will use the most efficient
18806 instruction. In the previous example, that might be a 32-bit load
18807 instruction, even though that will access bytes that do not contain
18808 any portion of the bitfield, or memory-mapped registers unrelated to
18809 the one being updated.
18811 If the target requires strict alignment, and honoring the field
18812 type would require violating this alignment, a warning is issued.
18813 If the field has @code{packed} attribute, the access is done without
18814 honoring the field type. If the field doesn't have @code{packed}
18815 attribute, the access is done honoring the field type. In both cases,
18816 GCC assumes that the user knows something about the target hardware
18817 that it is unaware of.
18819 The default value of this option is determined by the application binary
18820 interface for the target processor.
18826 @node Environment Variables
18827 @section Environment Variables Affecting GCC
18828 @cindex environment variables
18830 @c man begin ENVIRONMENT
18831 This section describes several environment variables that affect how GCC
18832 operates. Some of them work by specifying directories or prefixes to use
18833 when searching for various kinds of files. Some are used to specify other
18834 aspects of the compilation environment.
18836 Note that you can also specify places to search using options such as
18837 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18838 take precedence over places specified using environment variables, which
18839 in turn take precedence over those specified by the configuration of GCC@.
18840 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18841 GNU Compiler Collection (GCC) Internals}.
18846 @c @itemx LC_COLLATE
18848 @c @itemx LC_MONETARY
18849 @c @itemx LC_NUMERIC
18854 @c @findex LC_COLLATE
18855 @findex LC_MESSAGES
18856 @c @findex LC_MONETARY
18857 @c @findex LC_NUMERIC
18861 These environment variables control the way that GCC uses
18862 localization information that allow GCC to work with different
18863 national conventions. GCC inspects the locale categories
18864 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18865 so. These locale categories can be set to any value supported by your
18866 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18867 Kingdom encoded in UTF-8.
18869 The @env{LC_CTYPE} environment variable specifies character
18870 classification. GCC uses it to determine the character boundaries in
18871 a string; this is needed for some multibyte encodings that contain quote
18872 and escape characters that would otherwise be interpreted as a string
18875 The @env{LC_MESSAGES} environment variable specifies the language to
18876 use in diagnostic messages.
18878 If the @env{LC_ALL} environment variable is set, it overrides the value
18879 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18880 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18881 environment variable. If none of these variables are set, GCC
18882 defaults to traditional C English behavior.
18886 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18887 files. GCC uses temporary files to hold the output of one stage of
18888 compilation which is to be used as input to the next stage: for example,
18889 the output of the preprocessor, which is the input to the compiler
18892 @item GCC_COMPARE_DEBUG
18893 @findex GCC_COMPARE_DEBUG
18894 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
18895 @option{-fcompare-debug} to the compiler driver. See the documentation
18896 of this option for more details.
18898 @item GCC_EXEC_PREFIX
18899 @findex GCC_EXEC_PREFIX
18900 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18901 names of the subprograms executed by the compiler. No slash is added
18902 when this prefix is combined with the name of a subprogram, but you can
18903 specify a prefix that ends with a slash if you wish.
18905 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18906 an appropriate prefix to use based on the pathname it was invoked with.
18908 If GCC cannot find the subprogram using the specified prefix, it
18909 tries looking in the usual places for the subprogram.
18911 The default value of @env{GCC_EXEC_PREFIX} is
18912 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18913 the installed compiler. In many cases @var{prefix} is the value
18914 of @code{prefix} when you ran the @file{configure} script.
18916 Other prefixes specified with @option{-B} take precedence over this prefix.
18918 This prefix is also used for finding files such as @file{crt0.o} that are
18921 In addition, the prefix is used in an unusual way in finding the
18922 directories to search for header files. For each of the standard
18923 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18924 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18925 replacing that beginning with the specified prefix to produce an
18926 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18927 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18928 These alternate directories are searched first; the standard directories
18929 come next. If a standard directory begins with the configured
18930 @var{prefix} then the value of @var{prefix} is replaced by
18931 @env{GCC_EXEC_PREFIX} when looking for header files.
18933 @item COMPILER_PATH
18934 @findex COMPILER_PATH
18935 The value of @env{COMPILER_PATH} is a colon-separated list of
18936 directories, much like @env{PATH}. GCC tries the directories thus
18937 specified when searching for subprograms, if it can't find the
18938 subprograms using @env{GCC_EXEC_PREFIX}.
18941 @findex LIBRARY_PATH
18942 The value of @env{LIBRARY_PATH} is a colon-separated list of
18943 directories, much like @env{PATH}. When configured as a native compiler,
18944 GCC tries the directories thus specified when searching for special
18945 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18946 using GCC also uses these directories when searching for ordinary
18947 libraries for the @option{-l} option (but directories specified with
18948 @option{-L} come first).
18952 @cindex locale definition
18953 This variable is used to pass locale information to the compiler. One way in
18954 which this information is used is to determine the character set to be used
18955 when character literals, string literals and comments are parsed in C and C++.
18956 When the compiler is configured to allow multibyte characters,
18957 the following values for @env{LANG} are recognized:
18961 Recognize JIS characters.
18963 Recognize SJIS characters.
18965 Recognize EUCJP characters.
18968 If @env{LANG} is not defined, or if it has some other value, then the
18969 compiler will use mblen and mbtowc as defined by the default locale to
18970 recognize and translate multibyte characters.
18974 Some additional environments variables affect the behavior of the
18977 @include cppenv.texi
18981 @node Precompiled Headers
18982 @section Using Precompiled Headers
18983 @cindex precompiled headers
18984 @cindex speed of compilation
18986 Often large projects have many header files that are included in every
18987 source file. The time the compiler takes to process these header files
18988 over and over again can account for nearly all of the time required to
18989 build the project. To make builds faster, GCC allows users to
18990 `precompile' a header file; then, if builds can use the precompiled
18991 header file they will be much faster.
18993 To create a precompiled header file, simply compile it as you would any
18994 other file, if necessary using the @option{-x} option to make the driver
18995 treat it as a C or C++ header file. You will probably want to use a
18996 tool like @command{make} to keep the precompiled header up-to-date when
18997 the headers it contains change.
18999 A precompiled header file will be searched for when @code{#include} is
19000 seen in the compilation. As it searches for the included file
19001 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
19002 compiler looks for a precompiled header in each directory just before it
19003 looks for the include file in that directory. The name searched for is
19004 the name specified in the @code{#include} with @samp{.gch} appended. If
19005 the precompiled header file can't be used, it is ignored.
19007 For instance, if you have @code{#include "all.h"}, and you have
19008 @file{all.h.gch} in the same directory as @file{all.h}, then the
19009 precompiled header file will be used if possible, and the original
19010 header will be used otherwise.
19012 Alternatively, you might decide to put the precompiled header file in a
19013 directory and use @option{-I} to ensure that directory is searched
19014 before (or instead of) the directory containing the original header.
19015 Then, if you want to check that the precompiled header file is always
19016 used, you can put a file of the same name as the original header in this
19017 directory containing an @code{#error} command.
19019 This also works with @option{-include}. So yet another way to use
19020 precompiled headers, good for projects not designed with precompiled
19021 header files in mind, is to simply take most of the header files used by
19022 a project, include them from another header file, precompile that header
19023 file, and @option{-include} the precompiled header. If the header files
19024 have guards against multiple inclusion, they will be skipped because
19025 they've already been included (in the precompiled header).
19027 If you need to precompile the same header file for different
19028 languages, targets, or compiler options, you can instead make a
19029 @emph{directory} named like @file{all.h.gch}, and put each precompiled
19030 header in the directory, perhaps using @option{-o}. It doesn't matter
19031 what you call the files in the directory, every precompiled header in
19032 the directory will be considered. The first precompiled header
19033 encountered in the directory that is valid for this compilation will
19034 be used; they're searched in no particular order.
19036 There are many other possibilities, limited only by your imagination,
19037 good sense, and the constraints of your build system.
19039 A precompiled header file can be used only when these conditions apply:
19043 Only one precompiled header can be used in a particular compilation.
19046 A precompiled header can't be used once the first C token is seen. You
19047 can have preprocessor directives before a precompiled header; you can
19048 even include a precompiled header from inside another header, so long as
19049 there are no C tokens before the @code{#include}.
19052 The precompiled header file must be produced for the same language as
19053 the current compilation. You can't use a C precompiled header for a C++
19057 The precompiled header file must have been produced by the same compiler
19058 binary as the current compilation is using.
19061 Any macros defined before the precompiled header is included must
19062 either be defined in the same way as when the precompiled header was
19063 generated, or must not affect the precompiled header, which usually
19064 means that they don't appear in the precompiled header at all.
19066 The @option{-D} option is one way to define a macro before a
19067 precompiled header is included; using a @code{#define} can also do it.
19068 There are also some options that define macros implicitly, like
19069 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
19072 @item If debugging information is output when using the precompiled
19073 header, using @option{-g} or similar, the same kind of debugging information
19074 must have been output when building the precompiled header. However,
19075 a precompiled header built using @option{-g} can be used in a compilation
19076 when no debugging information is being output.
19078 @item The same @option{-m} options must generally be used when building
19079 and using the precompiled header. @xref{Submodel Options},
19080 for any cases where this rule is relaxed.
19082 @item Each of the following options must be the same when building and using
19083 the precompiled header:
19085 @gccoptlist{-fexceptions}
19088 Some other command-line options starting with @option{-f},
19089 @option{-p}, or @option{-O} must be defined in the same way as when
19090 the precompiled header was generated. At present, it's not clear
19091 which options are safe to change and which are not; the safest choice
19092 is to use exactly the same options when generating and using the
19093 precompiled header. The following are known to be safe:
19095 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
19096 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
19097 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
19102 For all of these except the last, the compiler will automatically
19103 ignore the precompiled header if the conditions aren't met. If you
19104 find an option combination that doesn't work and doesn't cause the
19105 precompiled header to be ignored, please consider filing a bug report,
19108 If you do use differing options when generating and using the
19109 precompiled header, the actual behavior will be a mixture of the
19110 behavior for the options. For instance, if you use @option{-g} to
19111 generate the precompiled header but not when using it, you may or may
19112 not get debugging information for routines in the precompiled header.