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
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} @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 -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-call-cxx-cdtors @gol
212 -fobjc-direct-dispatch @gol
213 -fobjc-exceptions @gol
215 -fobjc-std=objc1 @gol
216 -freplace-objc-classes @gol
219 -Wassign-intercept @gol
220 -Wno-protocol -Wselector @gol
221 -Wstrict-selector-match @gol
222 -Wundeclared-selector}
224 @item Language Independent Options
225 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
226 @gccoptlist{-fmessage-length=@var{n} @gol
227 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
228 -fdiagnostics-show-option}
230 @item Warning Options
231 @xref{Warning Options,,Options to Request or Suppress Warnings}.
232 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
233 -pedantic-errors @gol
234 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
235 -Wno-attributes -Wno-builtin-macro-redefined @gol
236 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
237 -Wchar-subscripts -Wclobbered -Wcomment @gol
238 -Wconversion -Wcoverage-mismatch -Wcpp -Wno-deprecated @gol
239 -Wno-deprecated-declarations -Wdisabled-optimization @gol
240 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
241 -Wno-endif-labels -Werror -Werror=* @gol
242 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
243 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
244 -Wformat-security -Wformat-y2k @gol
245 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
246 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
247 -Winit-self -Winline @gol
248 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
249 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
250 -Wlogical-op -Wlong-long @gol
251 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
252 -Wmissing-format-attribute -Wmissing-include-dirs @gol
254 -Wno-multichar -Wnonnull -Wno-overflow @gol
255 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
256 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
257 -Wpointer-arith -Wno-pointer-to-int-cast @gol
258 -Wredundant-decls @gol
259 -Wreturn-type -Wsequence-point -Wshadow @gol
260 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
261 -Wstrict-aliasing -Wstrict-aliasing=n @gol
262 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
263 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
264 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
265 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
266 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
267 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
268 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value @gol
269 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
270 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
272 @item C and Objective-C-only Warning Options
273 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
274 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
275 -Wold-style-declaration -Wold-style-definition @gol
276 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
277 -Wdeclaration-after-statement -Wpointer-sign}
279 @item Debugging Options
280 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
281 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
282 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
283 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
284 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
285 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
286 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
287 -fdump-statistics @gol
289 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
293 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-nrv -fdump-tree-vect @gol
303 -fdump-tree-sink @gol
304 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
308 -ftree-vectorizer-verbose=@var{n} @gol
309 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
310 -fdump-final-insns=@var{file} @gol
311 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
312 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
313 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
314 -fenable-icf-debug @gol
315 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
316 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
317 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
318 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
319 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
320 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
321 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
322 -gvms -gxcoff -gxcoff+ @gol
323 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
324 -fdebug-prefix-map=@var{old}=@var{new} @gol
325 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
326 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
327 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
328 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
329 -print-prog-name=@var{program} -print-search-dirs -Q @gol
330 -print-sysroot -print-sysroot-headers-suffix @gol
331 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
333 @item Optimization Options
334 @xref{Optimize Options,,Options that Control Optimization}.
335 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
336 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
337 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
338 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
339 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
340 -fcompare-elim -fcprop-registers -fcrossjumping @gol
341 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
342 -fcx-limited-range @gol
343 -fdata-sections -fdce -fdce -fdelayed-branch @gol
344 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
345 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
346 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
347 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
348 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
349 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
350 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
351 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
352 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
353 -fipa-struct-reorg -fira-algorithm=@var{algorithm} @gol
354 -fira-region=@var{region} @gol
355 -fira-loop-pressure -fno-ira-share-save-slots @gol
356 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
357 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
358 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
359 -floop-parallelize-all -flto -flto-compression-level
360 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
361 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
362 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
363 -fno-default-inline @gol
364 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
365 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
366 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
367 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
368 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
369 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
370 -fprefetch-loop-arrays @gol
371 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
372 -fprofile-generate=@var{path} @gol
373 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
374 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
375 -freorder-blocks-and-partition -freorder-functions @gol
376 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
377 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
378 -fsched-spec-load -fsched-spec-load-dangerous @gol
379 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
380 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
381 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
382 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
383 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
384 -fselective-scheduling -fselective-scheduling2 @gol
385 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
386 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
387 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
388 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
390 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
391 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
392 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
393 -ftree-loop-if-convert-memory-writes -ftree-loop-im @gol
394 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
395 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
396 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
397 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
398 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
399 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
400 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
401 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
402 -fwhole-program -fwpa -fuse-linker-plugin @gol
403 --param @var{name}=@var{value}
404 -O -O0 -O1 -O2 -O3 -Os -Ofast}
406 @item Preprocessor Options
407 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
408 @gccoptlist{-A@var{question}=@var{answer} @gol
409 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
410 -C -dD -dI -dM -dN @gol
411 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
412 -idirafter @var{dir} @gol
413 -include @var{file} -imacros @var{file} @gol
414 -iprefix @var{file} -iwithprefix @var{dir} @gol
415 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
416 -imultilib @var{dir} -isysroot @var{dir} @gol
417 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
418 -P -fworking-directory -remap @gol
419 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
420 -Xpreprocessor @var{option}}
422 @item Assembler Option
423 @xref{Assembler Options,,Passing Options to the Assembler}.
424 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
427 @xref{Link Options,,Options for Linking}.
428 @gccoptlist{@var{object-file-name} -l@var{library} @gol
429 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
430 -s -static -static-libgcc -static-libstdc++ -shared @gol
431 -shared-libgcc -symbolic @gol
432 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
435 @item Directory Options
436 @xref{Directory Options,,Options for Directory Search}.
437 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir}}
438 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-
441 @item Machine Dependent Options
442 @xref{Submodel Options,,Hardware Models and Configurations}.
443 @c This list is ordered alphanumerically by subsection name.
444 @c Try and put the significant identifier (CPU or system) first,
445 @c so users have a clue at guessing where the ones they want will be.
448 @gccoptlist{-EB -EL @gol
449 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
450 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
453 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
454 -mabi=@var{name} @gol
455 -mapcs-stack-check -mno-apcs-stack-check @gol
456 -mapcs-float -mno-apcs-float @gol
457 -mapcs-reentrant -mno-apcs-reentrant @gol
458 -msched-prolog -mno-sched-prolog @gol
459 -mlittle-endian -mbig-endian -mwords-little-endian @gol
460 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
461 -mfp16-format=@var{name}
462 -mthumb-interwork -mno-thumb-interwork @gol
463 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
464 -mstructure-size-boundary=@var{n} @gol
465 -mabort-on-noreturn @gol
466 -mlong-calls -mno-long-calls @gol
467 -msingle-pic-base -mno-single-pic-base @gol
468 -mpic-register=@var{reg} @gol
469 -mnop-fun-dllimport @gol
470 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
471 -mpoke-function-name @gol
473 -mtpcs-frame -mtpcs-leaf-frame @gol
474 -mcaller-super-interworking -mcallee-super-interworking @gol
476 -mword-relocations @gol
477 -mfix-cortex-m3-ldrd}
480 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
481 -mcall-prologues -mtiny-stack -mint8}
483 @emph{Blackfin Options}
484 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
485 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
486 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
487 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
488 -mno-id-shared-library -mshared-library-id=@var{n} @gol
489 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
490 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
491 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
495 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
496 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
497 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
498 -mstack-align -mdata-align -mconst-align @gol
499 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
500 -melf -maout -melinux -mlinux -sim -sim2 @gol
501 -mmul-bug-workaround -mno-mul-bug-workaround}
504 @gccoptlist{-mmac -mpush-args}
506 @emph{Darwin Options}
507 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
508 -arch_only -bind_at_load -bundle -bundle_loader @gol
509 -client_name -compatibility_version -current_version @gol
511 -dependency-file -dylib_file -dylinker_install_name @gol
512 -dynamic -dynamiclib -exported_symbols_list @gol
513 -filelist -flat_namespace -force_cpusubtype_ALL @gol
514 -force_flat_namespace -headerpad_max_install_names @gol
516 -image_base -init -install_name -keep_private_externs @gol
517 -multi_module -multiply_defined -multiply_defined_unused @gol
518 -noall_load -no_dead_strip_inits_and_terms @gol
519 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
520 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
521 -private_bundle -read_only_relocs -sectalign @gol
522 -sectobjectsymbols -whyload -seg1addr @gol
523 -sectcreate -sectobjectsymbols -sectorder @gol
524 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
525 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
526 -segprot -segs_read_only_addr -segs_read_write_addr @gol
527 -single_module -static -sub_library -sub_umbrella @gol
528 -twolevel_namespace -umbrella -undefined @gol
529 -unexported_symbols_list -weak_reference_mismatches @gol
530 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
531 -mkernel -mone-byte-bool}
533 @emph{DEC Alpha Options}
534 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
535 -mieee -mieee-with-inexact -mieee-conformant @gol
536 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
537 -mtrap-precision=@var{mode} -mbuild-constants @gol
538 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
539 -mbwx -mmax -mfix -mcix @gol
540 -mfloat-vax -mfloat-ieee @gol
541 -mexplicit-relocs -msmall-data -mlarge-data @gol
542 -msmall-text -mlarge-text @gol
543 -mmemory-latency=@var{time}}
545 @emph{DEC Alpha/VMS Options}
546 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
549 @gccoptlist{-msmall-model -mno-lsim}
552 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
553 -mhard-float -msoft-float @gol
554 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
555 -mdouble -mno-double @gol
556 -mmedia -mno-media -mmuladd -mno-muladd @gol
557 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
558 -mlinked-fp -mlong-calls -malign-labels @gol
559 -mlibrary-pic -macc-4 -macc-8 @gol
560 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
561 -moptimize-membar -mno-optimize-membar @gol
562 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
563 -mvliw-branch -mno-vliw-branch @gol
564 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
565 -mno-nested-cond-exec -mtomcat-stats @gol
569 @emph{GNU/Linux Options}
570 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
571 -tno-android-cc -tno-android-ld}
573 @emph{H8/300 Options}
574 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
577 @gccoptlist{-march=@var{architecture-type} @gol
578 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
579 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
580 -mfixed-range=@var{register-range} @gol
581 -mjump-in-delay -mlinker-opt -mlong-calls @gol
582 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
583 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
584 -mno-jump-in-delay -mno-long-load-store @gol
585 -mno-portable-runtime -mno-soft-float @gol
586 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
587 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
588 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
589 -munix=@var{unix-std} -nolibdld -static -threads}
591 @emph{i386 and x86-64 Options}
592 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
593 -mfpmath=@var{unit} @gol
594 -masm=@var{dialect} -mno-fancy-math-387 @gol
595 -mno-fp-ret-in-387 -msoft-float @gol
596 -mno-wide-multiply -mrtd -malign-double @gol
597 -mpreferred-stack-boundary=@var{num}
598 -mincoming-stack-boundary=@var{num} @gol
599 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
600 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
601 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
602 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlwp @gol
603 -mthreads -mno-align-stringops -minline-all-stringops @gol
604 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
605 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
606 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
607 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
608 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
609 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
610 -mcmodel=@var{code-model} -mabi=@var{name} @gol
611 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
612 -msse2avx -mfentry -m8bit-idiv}
615 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
616 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
617 -mconstant-gp -mauto-pic -mfused-madd @gol
618 -minline-float-divide-min-latency @gol
619 -minline-float-divide-max-throughput @gol
620 -mno-inline-float-divide @gol
621 -minline-int-divide-min-latency @gol
622 -minline-int-divide-max-throughput @gol
623 -mno-inline-int-divide @gol
624 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
625 -mno-inline-sqrt @gol
626 -mdwarf2-asm -mearly-stop-bits @gol
627 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
628 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
629 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
630 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
631 -msched-spec-ldc -msched-spec-control-ldc @gol
632 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
633 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
634 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
635 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
637 @emph{IA-64/VMS Options}
638 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
641 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
642 -msign-extend-enabled -muser-enabled}
644 @emph{M32R/D Options}
645 @gccoptlist{-m32r2 -m32rx -m32r @gol
647 -malign-loops -mno-align-loops @gol
648 -missue-rate=@var{number} @gol
649 -mbranch-cost=@var{number} @gol
650 -mmodel=@var{code-size-model-type} @gol
651 -msdata=@var{sdata-type} @gol
652 -mno-flush-func -mflush-func=@var{name} @gol
653 -mno-flush-trap -mflush-trap=@var{number} @gol
657 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
659 @emph{M680x0 Options}
660 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
661 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
662 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
663 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
664 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
665 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
666 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
667 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
670 @emph{M68hc1x Options}
671 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
672 -mauto-incdec -minmax -mlong-calls -mshort @gol
673 -msoft-reg-count=@var{count}}
676 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
677 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
678 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
679 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
680 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
683 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
684 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
685 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
686 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
689 @emph{MicroBlaze Options}
690 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
691 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
692 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
693 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
694 -mxl-mode-@var{app-model}}
697 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
698 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
699 -mips64 -mips64r2 @gol
700 -mips16 -mno-mips16 -mflip-mips16 @gol
701 -minterlink-mips16 -mno-interlink-mips16 @gol
702 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
703 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
704 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
705 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
706 -mfpu=@var{fpu-type} @gol
707 -msmartmips -mno-smartmips @gol
708 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
709 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
710 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
711 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
712 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
713 -membedded-data -mno-embedded-data @gol
714 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
715 -mcode-readable=@var{setting} @gol
716 -msplit-addresses -mno-split-addresses @gol
717 -mexplicit-relocs -mno-explicit-relocs @gol
718 -mcheck-zero-division -mno-check-zero-division @gol
719 -mdivide-traps -mdivide-breaks @gol
720 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
721 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
722 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
723 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
724 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
725 -mflush-func=@var{func} -mno-flush-func @gol
726 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
727 -mfp-exceptions -mno-fp-exceptions @gol
728 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
729 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
732 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
733 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
734 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
735 -mno-base-addresses -msingle-exit -mno-single-exit}
737 @emph{MN10300 Options}
738 @gccoptlist{-mmult-bug -mno-mult-bug @gol
739 -mno-am33 -mam33 -mam33-2 -mam34 @gol
740 -mtune=@var{cpu-type} @gol
741 -mreturn-pointer-on-d0 @gol
742 -mno-crt0 -mrelax -mliw}
744 @emph{PDP-11 Options}
745 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
746 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
747 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
748 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
749 -mbranch-expensive -mbranch-cheap @gol
750 -munix-asm -mdec-asm}
752 @emph{picoChip Options}
753 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
754 -msymbol-as-address -mno-inefficient-warnings}
756 @emph{PowerPC Options}
757 See RS/6000 and PowerPC Options.
759 @emph{RS/6000 and PowerPC Options}
760 @gccoptlist{-mcpu=@var{cpu-type} @gol
761 -mtune=@var{cpu-type} @gol
762 -mcmodel=@var{code-model} @gol
763 -mpower -mno-power -mpower2 -mno-power2 @gol
764 -mpowerpc -mpowerpc64 -mno-powerpc @gol
765 -maltivec -mno-altivec @gol
766 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
767 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
768 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
769 -mfprnd -mno-fprnd @gol
770 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
771 -mnew-mnemonics -mold-mnemonics @gol
772 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
773 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
774 -malign-power -malign-natural @gol
775 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
776 -msingle-float -mdouble-float -msimple-fpu @gol
777 -mstring -mno-string -mupdate -mno-update @gol
778 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
779 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
780 -mstrict-align -mno-strict-align -mrelocatable @gol
781 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
782 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
783 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
784 -mprioritize-restricted-insns=@var{priority} @gol
785 -msched-costly-dep=@var{dependence_type} @gol
786 -minsert-sched-nops=@var{scheme} @gol
787 -mcall-sysv -mcall-netbsd @gol
788 -maix-struct-return -msvr4-struct-return @gol
789 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
790 -mblock-move-inline-limit=@var{num} @gol
791 -misel -mno-isel @gol
792 -misel=yes -misel=no @gol
794 -mspe=yes -mspe=no @gol
796 -mgen-cell-microcode -mwarn-cell-microcode @gol
797 -mvrsave -mno-vrsave @gol
798 -mmulhw -mno-mulhw @gol
799 -mdlmzb -mno-dlmzb @gol
800 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
801 -mprototype -mno-prototype @gol
802 -msim -mmvme -mads -myellowknife -memb -msdata @gol
803 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
804 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision
805 -mno-recip-precision @gol
806 -mveclibabi=@var{type} -mfriz -mno-friz}
809 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
811 -mbig-endian-data -mlittle-endian-data @gol
814 -mas100-syntax -mno-as100-syntax@gol
816 -mmax-constant-size=@gol
818 -msave-acc-in-interrupts}
820 @emph{S/390 and zSeries Options}
821 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
822 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
823 -mlong-double-64 -mlong-double-128 @gol
824 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
825 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
826 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
827 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
828 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
831 @gccoptlist{-meb -mel @gol
835 -mscore5 -mscore5u -mscore7 -mscore7d}
838 @gccoptlist{-m1 -m2 -m2e @gol
839 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
841 -m4-nofpu -m4-single-only -m4-single -m4 @gol
842 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
843 -m5-64media -m5-64media-nofpu @gol
844 -m5-32media -m5-32media-nofpu @gol
845 -m5-compact -m5-compact-nofpu @gol
846 -mb -ml -mdalign -mrelax @gol
847 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
848 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
849 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
850 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
851 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
852 -maccumulate-outgoing-args -minvalid-symbols}
854 @emph{Solaris 2 Options}
855 @gccoptlist{-mimpure-text -mno-impure-text @gol
856 -threads -pthreads -pthread}
859 @gccoptlist{-mcpu=@var{cpu-type} @gol
860 -mtune=@var{cpu-type} @gol
861 -mcmodel=@var{code-model} @gol
862 -m32 -m64 -mapp-regs -mno-app-regs @gol
863 -mfaster-structs -mno-faster-structs @gol
864 -mfpu -mno-fpu -mhard-float -msoft-float @gol
865 -mhard-quad-float -msoft-quad-float @gol
867 -mstack-bias -mno-stack-bias @gol
868 -munaligned-doubles -mno-unaligned-doubles @gol
869 -mv8plus -mno-v8plus -mvis -mno-vis}
872 @gccoptlist{-mwarn-reloc -merror-reloc @gol
873 -msafe-dma -munsafe-dma @gol
875 -msmall-mem -mlarge-mem -mstdmain @gol
876 -mfixed-range=@var{register-range} @gol
878 -maddress-space-conversion -mno-address-space-conversion @gol
879 -mcache-size=@var{cache-size} @gol
880 -matomic-updates -mno-atomic-updates}
882 @emph{System V Options}
883 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
886 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
887 -mprolog-function -mno-prolog-function -mspace @gol
888 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
889 -mapp-regs -mno-app-regs @gol
890 -mdisable-callt -mno-disable-callt @gol
893 -mv850e1 -mv850es @gol
898 @gccoptlist{-mg -mgnu -munix}
900 @emph{VxWorks Options}
901 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
902 -Xbind-lazy -Xbind-now}
904 @emph{x86-64 Options}
905 See i386 and x86-64 Options.
907 @emph{i386 and x86-64 Windows Options}
908 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
909 -mnop-fun-dllimport -mthread @gol
910 -municode -mwin32 -mwindows -fno-set-stack-executable}
912 @emph{Xstormy16 Options}
915 @emph{Xtensa Options}
916 @gccoptlist{-mconst16 -mno-const16 @gol
917 -mfused-madd -mno-fused-madd @gol
919 -mserialize-volatile -mno-serialize-volatile @gol
920 -mtext-section-literals -mno-text-section-literals @gol
921 -mtarget-align -mno-target-align @gol
922 -mlongcalls -mno-longcalls}
924 @emph{zSeries Options}
925 See S/390 and zSeries Options.
927 @item Code Generation Options
928 @xref{Code Gen Options,,Options for Code Generation Conventions}.
929 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
930 -ffixed-@var{reg} -fexceptions @gol
931 -fnon-call-exceptions -funwind-tables @gol
932 -fasynchronous-unwind-tables @gol
933 -finhibit-size-directive -finstrument-functions @gol
934 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
935 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
936 -fno-common -fno-ident @gol
937 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
938 -fno-jump-tables @gol
939 -frecord-gcc-switches @gol
940 -freg-struct-return -fshort-enums @gol
941 -fshort-double -fshort-wchar @gol
942 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
943 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
944 -fno-stack-limit -fsplit-stack @gol
945 -fleading-underscore -ftls-model=@var{model} @gol
946 -ftrapv -fwrapv -fbounds-check @gol
947 -fvisibility -fstrict-volatile-bitfields}
951 * Overall Options:: Controlling the kind of output:
952 an executable, object files, assembler files,
953 or preprocessed source.
954 * C Dialect Options:: Controlling the variant of C language compiled.
955 * C++ Dialect Options:: Variations on C++.
956 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
958 * Language Independent Options:: Controlling how diagnostics should be
960 * Warning Options:: How picky should the compiler be?
961 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
962 * Optimize Options:: How much optimization?
963 * Preprocessor Options:: Controlling header files and macro definitions.
964 Also, getting dependency information for Make.
965 * Assembler Options:: Passing options to the assembler.
966 * Link Options:: Specifying libraries and so on.
967 * Directory Options:: Where to find header files and libraries.
968 Where to find the compiler executable files.
969 * Spec Files:: How to pass switches to sub-processes.
970 * Target Options:: Running a cross-compiler, or an old version of GCC.
973 @node Overall Options
974 @section Options Controlling the Kind of Output
976 Compilation can involve up to four stages: preprocessing, compilation
977 proper, assembly and linking, always in that order. GCC is capable of
978 preprocessing and compiling several files either into several
979 assembler input files, or into one assembler input file; then each
980 assembler input file produces an object file, and linking combines all
981 the object files (those newly compiled, and those specified as input)
982 into an executable file.
984 @cindex file name suffix
985 For any given input file, the file name suffix determines what kind of
990 C source code which must be preprocessed.
993 C source code which should not be preprocessed.
996 C++ source code which should not be preprocessed.
999 Objective-C source code. Note that you must link with the @file{libobjc}
1000 library to make an Objective-C program work.
1003 Objective-C source code which should not be preprocessed.
1007 Objective-C++ source code. Note that you must link with the @file{libobjc}
1008 library to make an Objective-C++ program work. Note that @samp{.M} refers
1009 to a literal capital M@.
1011 @item @var{file}.mii
1012 Objective-C++ source code which should not be preprocessed.
1015 C, C++, Objective-C or Objective-C++ header file to be turned into a
1016 precompiled header (default), or C, C++ header file to be turned into an
1017 Ada spec (via the @option{-fdump-ada-spec} switch).
1020 @itemx @var{file}.cp
1021 @itemx @var{file}.cxx
1022 @itemx @var{file}.cpp
1023 @itemx @var{file}.CPP
1024 @itemx @var{file}.c++
1026 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1027 the last two letters must both be literally @samp{x}. Likewise,
1028 @samp{.C} refers to a literal capital C@.
1032 Objective-C++ source code which must be preprocessed.
1034 @item @var{file}.mii
1035 Objective-C++ source code which should not be preprocessed.
1039 @itemx @var{file}.hp
1040 @itemx @var{file}.hxx
1041 @itemx @var{file}.hpp
1042 @itemx @var{file}.HPP
1043 @itemx @var{file}.h++
1044 @itemx @var{file}.tcc
1045 C++ header file to be turned into a precompiled header or Ada spec.
1048 @itemx @var{file}.for
1049 @itemx @var{file}.ftn
1050 Fixed form Fortran source code which should not be preprocessed.
1053 @itemx @var{file}.FOR
1054 @itemx @var{file}.fpp
1055 @itemx @var{file}.FPP
1056 @itemx @var{file}.FTN
1057 Fixed form Fortran source code which must be preprocessed (with the traditional
1060 @item @var{file}.f90
1061 @itemx @var{file}.f95
1062 @itemx @var{file}.f03
1063 @itemx @var{file}.f08
1064 Free form Fortran source code which should not be preprocessed.
1066 @item @var{file}.F90
1067 @itemx @var{file}.F95
1068 @itemx @var{file}.F03
1069 @itemx @var{file}.F08
1070 Free form Fortran source code which must be preprocessed (with the
1071 traditional preprocessor).
1076 @c FIXME: Descriptions of Java file types.
1082 @item @var{file}.ads
1083 Ada source code file which contains a library unit declaration (a
1084 declaration of a package, subprogram, or generic, or a generic
1085 instantiation), or a library unit renaming declaration (a package,
1086 generic, or subprogram renaming declaration). Such files are also
1089 @item @var{file}.adb
1090 Ada source code file containing a library unit body (a subprogram or
1091 package body). Such files are also called @dfn{bodies}.
1093 @c GCC also knows about some suffixes for languages not yet included:
1104 @itemx @var{file}.sx
1105 Assembler code which must be preprocessed.
1108 An object file to be fed straight into linking.
1109 Any file name with no recognized suffix is treated this way.
1113 You can specify the input language explicitly with the @option{-x} option:
1116 @item -x @var{language}
1117 Specify explicitly the @var{language} for the following input files
1118 (rather than letting the compiler choose a default based on the file
1119 name suffix). This option applies to all following input files until
1120 the next @option{-x} option. Possible values for @var{language} are:
1122 c c-header cpp-output
1123 c++ c++-header c++-cpp-output
1124 objective-c objective-c-header objective-c-cpp-output
1125 objective-c++ objective-c++-header objective-c++-cpp-output
1126 assembler assembler-with-cpp
1128 f77 f77-cpp-input f95 f95-cpp-input
1134 Turn off any specification of a language, so that subsequent files are
1135 handled according to their file name suffixes (as they are if @option{-x}
1136 has not been used at all).
1138 @item -pass-exit-codes
1139 @opindex pass-exit-codes
1140 Normally the @command{gcc} program will exit with the code of 1 if any
1141 phase of the compiler returns a non-success return code. If you specify
1142 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1143 numerically highest error produced by any phase that returned an error
1144 indication. The C, C++, and Fortran frontends return 4, if an internal
1145 compiler error is encountered.
1148 If you only want some of the stages of compilation, you can use
1149 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1150 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1151 @command{gcc} is to stop. Note that some combinations (for example,
1152 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1157 Compile or assemble the source files, but do not link. The linking
1158 stage simply is not done. The ultimate output is in the form of an
1159 object file for each source file.
1161 By default, the object file name for a source file is made by replacing
1162 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1164 Unrecognized input files, not requiring compilation or assembly, are
1169 Stop after the stage of compilation proper; do not assemble. The output
1170 is in the form of an assembler code file for each non-assembler input
1173 By default, the assembler file name for a source file is made by
1174 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1176 Input files that don't require compilation are ignored.
1180 Stop after the preprocessing stage; do not run the compiler proper. The
1181 output is in the form of preprocessed source code, which is sent to the
1184 Input files which don't require preprocessing are ignored.
1186 @cindex output file option
1189 Place output in file @var{file}. This applies regardless to whatever
1190 sort of output is being produced, whether it be an executable file,
1191 an object file, an assembler file or preprocessed C code.
1193 If @option{-o} is not specified, the default is to put an executable
1194 file in @file{a.out}, the object file for
1195 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1196 assembler file in @file{@var{source}.s}, a precompiled header file in
1197 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1202 Print (on standard error output) the commands executed to run the stages
1203 of compilation. Also print the version number of the compiler driver
1204 program and of the preprocessor and the compiler proper.
1208 Like @option{-v} except the commands are not executed and arguments
1209 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1210 This is useful for shell scripts to capture the driver-generated command lines.
1214 Use pipes rather than temporary files for communication between the
1215 various stages of compilation. This fails to work on some systems where
1216 the assembler is unable to read from a pipe; but the GNU assembler has
1221 Print (on the standard output) a description of the command line options
1222 understood by @command{gcc}. If the @option{-v} option is also specified
1223 then @option{--help} will also be passed on to the various processes
1224 invoked by @command{gcc}, so that they can display the command line options
1225 they accept. If the @option{-Wextra} option has also been specified
1226 (prior to the @option{--help} option), then command line options which
1227 have no documentation associated with them will also be displayed.
1230 @opindex target-help
1231 Print (on the standard output) a description of target-specific command
1232 line options for each tool. For some targets extra target-specific
1233 information may also be printed.
1235 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1236 Print (on the standard output) a description of the command line
1237 options understood by the compiler that fit into all specified classes
1238 and qualifiers. These are the supported classes:
1241 @item @samp{optimizers}
1242 This will display all of the optimization options supported by the
1245 @item @samp{warnings}
1246 This will display all of the options controlling warning messages
1247 produced by the compiler.
1250 This will display target-specific options. Unlike the
1251 @option{--target-help} option however, target-specific options of the
1252 linker and assembler will not be displayed. This is because those
1253 tools do not currently support the extended @option{--help=} syntax.
1256 This will display the values recognized by the @option{--param}
1259 @item @var{language}
1260 This will display the options supported for @var{language}, where
1261 @var{language} is the name of one of the languages supported in this
1265 This will display the options that are common to all languages.
1268 These are the supported qualifiers:
1271 @item @samp{undocumented}
1272 Display only those options which are undocumented.
1275 Display options which take an argument that appears after an equal
1276 sign in the same continuous piece of text, such as:
1277 @samp{--help=target}.
1279 @item @samp{separate}
1280 Display options which take an argument that appears as a separate word
1281 following the original option, such as: @samp{-o output-file}.
1284 Thus for example to display all the undocumented target-specific
1285 switches supported by the compiler the following can be used:
1288 --help=target,undocumented
1291 The sense of a qualifier can be inverted by prefixing it with the
1292 @samp{^} character, so for example to display all binary warning
1293 options (i.e., ones that are either on or off and that do not take an
1294 argument), which have a description the following can be used:
1297 --help=warnings,^joined,^undocumented
1300 The argument to @option{--help=} should not consist solely of inverted
1303 Combining several classes is possible, although this usually
1304 restricts the output by so much that there is nothing to display. One
1305 case where it does work however is when one of the classes is
1306 @var{target}. So for example to display all the target-specific
1307 optimization options the following can be used:
1310 --help=target,optimizers
1313 The @option{--help=} option can be repeated on the command line. Each
1314 successive use will display its requested class of options, skipping
1315 those that have already been displayed.
1317 If the @option{-Q} option appears on the command line before the
1318 @option{--help=} option, then the descriptive text displayed by
1319 @option{--help=} is changed. Instead of describing the displayed
1320 options, an indication is given as to whether the option is enabled,
1321 disabled or set to a specific value (assuming that the compiler
1322 knows this at the point where the @option{--help=} option is used).
1324 Here is a truncated example from the ARM port of @command{gcc}:
1327 % gcc -Q -mabi=2 --help=target -c
1328 The following options are target specific:
1330 -mabort-on-noreturn [disabled]
1334 The output is sensitive to the effects of previous command line
1335 options, so for example it is possible to find out which optimizations
1336 are enabled at @option{-O2} by using:
1339 -Q -O2 --help=optimizers
1342 Alternatively you can discover which binary optimizations are enabled
1343 by @option{-O3} by using:
1346 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1347 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1348 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1351 @item -no-canonical-prefixes
1352 @opindex no-canonical-prefixes
1353 Do not expand any symbolic links, resolve references to @samp{/../}
1354 or @samp{/./}, or make the path absolute when generating a relative
1359 Display the version number and copyrights of the invoked GCC@.
1363 Invoke all subcommands under a wrapper program. The name of the
1364 wrapper program and its parameters are passed as a comma separated
1368 gcc -c t.c -wrapper gdb,--args
1371 This will invoke all subprograms of @command{gcc} under
1372 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1373 @samp{gdb --args cc1 @dots{}}.
1375 @item -fplugin=@var{name}.so
1376 Load the plugin code in file @var{name}.so, assumed to be a
1377 shared object to be dlopen'd by the compiler. The base name of
1378 the shared object file is used to identify the plugin for the
1379 purposes of argument parsing (See
1380 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1381 Each plugin should define the callback functions specified in the
1384 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1385 Define an argument called @var{key} with a value of @var{value}
1386 for the plugin called @var{name}.
1388 @item -fdump-ada-spec@r{[}-slim@r{]}
1389 For C and C++ source and include files, generate corresponding Ada
1390 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1391 GNAT User's Guide}, which provides detailed documentation on this feature.
1393 @item -fdump-go-spec=@var{file}
1394 For input files in any language, generate corresponding Go
1395 declarations in @var{file}. This generates Go @code{const},
1396 @code{type}, @code{var}, and @code{func} declarations which may be a
1397 useful way to start writing a Go interface to code written in some
1400 @include @value{srcdir}/../libiberty/at-file.texi
1404 @section Compiling C++ Programs
1406 @cindex suffixes for C++ source
1407 @cindex C++ source file suffixes
1408 C++ source files conventionally use one of the suffixes @samp{.C},
1409 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1410 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1411 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1412 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1413 files with these names and compiles them as C++ programs even if you
1414 call the compiler the same way as for compiling C programs (usually
1415 with the name @command{gcc}).
1419 However, the use of @command{gcc} does not add the C++ library.
1420 @command{g++} is a program that calls GCC and treats @samp{.c},
1421 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1422 files unless @option{-x} is used, and automatically specifies linking
1423 against the C++ library. This program is also useful when
1424 precompiling a C header file with a @samp{.h} extension for use in C++
1425 compilations. On many systems, @command{g++} is also installed with
1426 the name @command{c++}.
1428 @cindex invoking @command{g++}
1429 When you compile C++ programs, you may specify many of the same
1430 command-line options that you use for compiling programs in any
1431 language; or command-line options meaningful for C and related
1432 languages; or options that are meaningful only for C++ programs.
1433 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1434 explanations of options for languages related to C@.
1435 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1436 explanations of options that are meaningful only for C++ programs.
1438 @node C Dialect Options
1439 @section Options Controlling C Dialect
1440 @cindex dialect options
1441 @cindex language dialect options
1442 @cindex options, dialect
1444 The following options control the dialect of C (or languages derived
1445 from C, such as C++, Objective-C and Objective-C++) that the compiler
1449 @cindex ANSI support
1453 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1454 equivalent to @samp{-std=c++98}.
1456 This turns off certain features of GCC that are incompatible with ISO
1457 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1458 such as the @code{asm} and @code{typeof} keywords, and
1459 predefined macros such as @code{unix} and @code{vax} that identify the
1460 type of system you are using. It also enables the undesirable and
1461 rarely used ISO trigraph feature. For the C compiler,
1462 it disables recognition of C++ style @samp{//} comments as well as
1463 the @code{inline} keyword.
1465 The alternate keywords @code{__asm__}, @code{__extension__},
1466 @code{__inline__} and @code{__typeof__} continue to work despite
1467 @option{-ansi}. You would not want to use them in an ISO C program, of
1468 course, but it is useful to put them in header files that might be included
1469 in compilations done with @option{-ansi}. Alternate predefined macros
1470 such as @code{__unix__} and @code{__vax__} are also available, with or
1471 without @option{-ansi}.
1473 The @option{-ansi} option does not cause non-ISO programs to be
1474 rejected gratuitously. For that, @option{-pedantic} is required in
1475 addition to @option{-ansi}. @xref{Warning Options}.
1477 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1478 option is used. Some header files may notice this macro and refrain
1479 from declaring certain functions or defining certain macros that the
1480 ISO standard doesn't call for; this is to avoid interfering with any
1481 programs that might use these names for other things.
1483 Functions that would normally be built in but do not have semantics
1484 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1485 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1486 built-in functions provided by GCC}, for details of the functions
1491 Determine the language standard. @xref{Standards,,Language Standards
1492 Supported by GCC}, for details of these standard versions. This option
1493 is currently only supported when compiling C or C++.
1495 The compiler can accept several base standards, such as @samp{c90} or
1496 @samp{c++98}, and GNU dialects of those standards, such as
1497 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1498 compiler will accept all programs following that standard and those
1499 using GNU extensions that do not contradict it. For example,
1500 @samp{-std=c90} turns off certain features of GCC that are
1501 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1502 keywords, but not other GNU extensions that do not have a meaning in
1503 ISO C90, such as omitting the middle term of a @code{?:}
1504 expression. On the other hand, by specifying a GNU dialect of a
1505 standard, all features the compiler support are enabled, even when
1506 those features change the meaning of the base standard and some
1507 strict-conforming programs may be rejected. The particular standard
1508 is used by @option{-pedantic} to identify which features are GNU
1509 extensions given that version of the standard. For example
1510 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1511 comments, while @samp{-std=gnu99 -pedantic} would not.
1513 A value for this option must be provided; possible values are
1519 Support all ISO C90 programs (certain GNU extensions that conflict
1520 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1522 @item iso9899:199409
1523 ISO C90 as modified in amendment 1.
1529 ISO C99. Note that this standard is not yet fully supported; see
1530 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1531 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1534 ISO C1X, the draft of the next revision of the ISO C standard.
1535 Support is limited and experimental and features enabled by this
1536 option may be changed or removed if changed in or removed from the
1541 GNU dialect of ISO C90 (including some C99 features). This
1542 is the default for C code.
1546 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1547 this will become the default. The name @samp{gnu9x} is deprecated.
1550 GNU dialect of ISO C1X. Support is limited and experimental and
1551 features enabled by this option may be changed or removed if changed
1552 in or removed from the standard draft.
1555 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1559 GNU dialect of @option{-std=c++98}. This is the default for
1563 The working draft of the upcoming ISO C++0x standard. This option
1564 enables experimental features that are likely to be included in
1565 C++0x. The working draft is constantly changing, and any feature that is
1566 enabled by this flag may be removed from future versions of GCC if it is
1567 not part of the C++0x standard.
1570 GNU dialect of @option{-std=c++0x}. This option enables
1571 experimental features that may be removed in future versions of GCC.
1574 @item -fgnu89-inline
1575 @opindex fgnu89-inline
1576 The option @option{-fgnu89-inline} tells GCC to use the traditional
1577 GNU semantics for @code{inline} functions when in C99 mode.
1578 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1579 is accepted and ignored by GCC versions 4.1.3 up to but not including
1580 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1581 C99 mode. Using this option is roughly equivalent to adding the
1582 @code{gnu_inline} function attribute to all inline functions
1583 (@pxref{Function Attributes}).
1585 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1586 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1587 specifies the default behavior). This option was first supported in
1588 GCC 4.3. This option is not supported in @option{-std=c90} or
1589 @option{-std=gnu90} mode.
1591 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1592 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1593 in effect for @code{inline} functions. @xref{Common Predefined
1594 Macros,,,cpp,The C Preprocessor}.
1596 @item -aux-info @var{filename}
1598 Output to the given filename prototyped declarations for all functions
1599 declared and/or defined in a translation unit, including those in header
1600 files. This option is silently ignored in any language other than C@.
1602 Besides declarations, the file indicates, in comments, the origin of
1603 each declaration (source file and line), whether the declaration was
1604 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1605 @samp{O} for old, respectively, in the first character after the line
1606 number and the colon), and whether it came from a declaration or a
1607 definition (@samp{C} or @samp{F}, respectively, in the following
1608 character). In the case of function definitions, a K&R-style list of
1609 arguments followed by their declarations is also provided, inside
1610 comments, after the declaration.
1614 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1615 keyword, so that code can use these words as identifiers. You can use
1616 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1617 instead. @option{-ansi} implies @option{-fno-asm}.
1619 In C++, this switch only affects the @code{typeof} keyword, since
1620 @code{asm} and @code{inline} are standard keywords. You may want to
1621 use the @option{-fno-gnu-keywords} flag instead, which has the same
1622 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1623 switch only affects the @code{asm} and @code{typeof} keywords, since
1624 @code{inline} is a standard keyword in ISO C99.
1627 @itemx -fno-builtin-@var{function}
1628 @opindex fno-builtin
1629 @cindex built-in functions
1630 Don't recognize built-in functions that do not begin with
1631 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1632 functions provided by GCC}, for details of the functions affected,
1633 including those which are not built-in functions when @option{-ansi} or
1634 @option{-std} options for strict ISO C conformance are used because they
1635 do not have an ISO standard meaning.
1637 GCC normally generates special code to handle certain built-in functions
1638 more efficiently; for instance, calls to @code{alloca} may become single
1639 instructions that adjust the stack directly, and calls to @code{memcpy}
1640 may become inline copy loops. The resulting code is often both smaller
1641 and faster, but since the function calls no longer appear as such, you
1642 cannot set a breakpoint on those calls, nor can you change the behavior
1643 of the functions by linking with a different library. In addition,
1644 when a function is recognized as a built-in function, GCC may use
1645 information about that function to warn about problems with calls to
1646 that function, or to generate more efficient code, even if the
1647 resulting code still contains calls to that function. For example,
1648 warnings are given with @option{-Wformat} for bad calls to
1649 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1650 known not to modify global memory.
1652 With the @option{-fno-builtin-@var{function}} option
1653 only the built-in function @var{function} is
1654 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1655 function is named that is not built-in in this version of GCC, this
1656 option is ignored. There is no corresponding
1657 @option{-fbuiltin-@var{function}} option; if you wish to enable
1658 built-in functions selectively when using @option{-fno-builtin} or
1659 @option{-ffreestanding}, you may define macros such as:
1662 #define abs(n) __builtin_abs ((n))
1663 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1668 @cindex hosted environment
1670 Assert that compilation takes place in a hosted environment. This implies
1671 @option{-fbuiltin}. A hosted environment is one in which the
1672 entire standard library is available, and in which @code{main} has a return
1673 type of @code{int}. Examples are nearly everything except a kernel.
1674 This is equivalent to @option{-fno-freestanding}.
1676 @item -ffreestanding
1677 @opindex ffreestanding
1678 @cindex hosted environment
1680 Assert that compilation takes place in a freestanding environment. This
1681 implies @option{-fno-builtin}. A freestanding environment
1682 is one in which the standard library may not exist, and program startup may
1683 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1684 This is equivalent to @option{-fno-hosted}.
1686 @xref{Standards,,Language Standards Supported by GCC}, for details of
1687 freestanding and hosted environments.
1691 @cindex OpenMP parallel
1692 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1693 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1694 compiler generates parallel code according to the OpenMP Application
1695 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1696 implies @option{-pthread}, and thus is only supported on targets that
1697 have support for @option{-pthread}.
1699 @item -fms-extensions
1700 @opindex fms-extensions
1701 Accept some non-standard constructs used in Microsoft header files.
1703 In C++ code, this allows member names in structures to be similar
1704 to previous types declarations.
1713 Some cases of unnamed fields in structures and unions are only
1714 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1715 fields within structs/unions}, for details.
1717 @item -fplan9-extensions
1718 Accept some non-standard constructs used in Plan 9 code.
1720 This enables @option{-fms-extensions}, permits passing pointers to
1721 structures with anonymous fields to functions which expect pointers to
1722 elements of the type of the field, and permits referring to anonymous
1723 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1724 struct/union fields within structs/unions}, for details. This is only
1725 supported for C, not C++.
1729 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1730 options for strict ISO C conformance) implies @option{-trigraphs}.
1732 @item -no-integrated-cpp
1733 @opindex no-integrated-cpp
1734 Performs a compilation in two passes: preprocessing and compiling. This
1735 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1736 @option{-B} option. The user supplied compilation step can then add in
1737 an additional preprocessing step after normal preprocessing but before
1738 compiling. The default is to use the integrated cpp (internal cpp)
1740 The semantics of this option will change if "cc1", "cc1plus", and
1741 "cc1obj" are merged.
1743 @cindex traditional C language
1744 @cindex C language, traditional
1746 @itemx -traditional-cpp
1747 @opindex traditional-cpp
1748 @opindex traditional
1749 Formerly, these options caused GCC to attempt to emulate a pre-standard
1750 C compiler. They are now only supported with the @option{-E} switch.
1751 The preprocessor continues to support a pre-standard mode. See the GNU
1752 CPP manual for details.
1754 @item -fcond-mismatch
1755 @opindex fcond-mismatch
1756 Allow conditional expressions with mismatched types in the second and
1757 third arguments. The value of such an expression is void. This option
1758 is not supported for C++.
1760 @item -flax-vector-conversions
1761 @opindex flax-vector-conversions
1762 Allow implicit conversions between vectors with differing numbers of
1763 elements and/or incompatible element types. This option should not be
1766 @item -funsigned-char
1767 @opindex funsigned-char
1768 Let the type @code{char} be unsigned, like @code{unsigned char}.
1770 Each kind of machine has a default for what @code{char} should
1771 be. It is either like @code{unsigned char} by default or like
1772 @code{signed char} by default.
1774 Ideally, a portable program should always use @code{signed char} or
1775 @code{unsigned char} when it depends on the signedness of an object.
1776 But many programs have been written to use plain @code{char} and
1777 expect it to be signed, or expect it to be unsigned, depending on the
1778 machines they were written for. This option, and its inverse, let you
1779 make such a program work with the opposite default.
1781 The type @code{char} is always a distinct type from each of
1782 @code{signed char} or @code{unsigned char}, even though its behavior
1783 is always just like one of those two.
1786 @opindex fsigned-char
1787 Let the type @code{char} be signed, like @code{signed char}.
1789 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1790 the negative form of @option{-funsigned-char}. Likewise, the option
1791 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1793 @item -fsigned-bitfields
1794 @itemx -funsigned-bitfields
1795 @itemx -fno-signed-bitfields
1796 @itemx -fno-unsigned-bitfields
1797 @opindex fsigned-bitfields
1798 @opindex funsigned-bitfields
1799 @opindex fno-signed-bitfields
1800 @opindex fno-unsigned-bitfields
1801 These options control whether a bit-field is signed or unsigned, when the
1802 declaration does not use either @code{signed} or @code{unsigned}. By
1803 default, such a bit-field is signed, because this is consistent: the
1804 basic integer types such as @code{int} are signed types.
1807 @node C++ Dialect Options
1808 @section Options Controlling C++ Dialect
1810 @cindex compiler options, C++
1811 @cindex C++ options, command line
1812 @cindex options, C++
1813 This section describes the command-line options that are only meaningful
1814 for C++ programs; but you can also use most of the GNU compiler options
1815 regardless of what language your program is in. For example, you
1816 might compile a file @code{firstClass.C} like this:
1819 g++ -g -frepo -O -c firstClass.C
1823 In this example, only @option{-frepo} is an option meant
1824 only for C++ programs; you can use the other options with any
1825 language supported by GCC@.
1827 Here is a list of options that are @emph{only} for compiling C++ programs:
1831 @item -fabi-version=@var{n}
1832 @opindex fabi-version
1833 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1834 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1835 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1836 the version that conforms most closely to the C++ ABI specification.
1837 Therefore, the ABI obtained using version 0 will change as ABI bugs
1840 The default is version 2.
1842 Version 3 corrects an error in mangling a constant address as a
1845 Version 4 implements a standard mangling for vector types.
1847 See also @option{-Wabi}.
1849 @item -fno-access-control
1850 @opindex fno-access-control
1851 Turn off all access checking. This switch is mainly useful for working
1852 around bugs in the access control code.
1856 Check that the pointer returned by @code{operator new} is non-null
1857 before attempting to modify the storage allocated. This check is
1858 normally unnecessary because the C++ standard specifies that
1859 @code{operator new} will only return @code{0} if it is declared
1860 @samp{throw()}, in which case the compiler will always check the
1861 return value even without this option. In all other cases, when
1862 @code{operator new} has a non-empty exception specification, memory
1863 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1864 @samp{new (nothrow)}.
1866 @item -fconserve-space
1867 @opindex fconserve-space
1868 Put uninitialized or runtime-initialized global variables into the
1869 common segment, as C does. This saves space in the executable at the
1870 cost of not diagnosing duplicate definitions. If you compile with this
1871 flag and your program mysteriously crashes after @code{main()} has
1872 completed, you may have an object that is being destroyed twice because
1873 two definitions were merged.
1875 This option is no longer useful on most targets, now that support has
1876 been added for putting variables into BSS without making them common.
1878 @item -fno-deduce-init-list
1879 @opindex fno-deduce-init-list
1880 Disable deduction of a template type parameter as
1881 std::initializer_list from a brace-enclosed initializer list, i.e.
1884 template <class T> auto forward(T t) -> decltype (realfn (t))
1891 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1895 This option is present because this deduction is an extension to the
1896 current specification in the C++0x working draft, and there was
1897 some concern about potential overload resolution problems.
1899 @item -ffriend-injection
1900 @opindex ffriend-injection
1901 Inject friend functions into the enclosing namespace, so that they are
1902 visible outside the scope of the class in which they are declared.
1903 Friend functions were documented to work this way in the old Annotated
1904 C++ Reference Manual, and versions of G++ before 4.1 always worked
1905 that way. However, in ISO C++ a friend function which is not declared
1906 in an enclosing scope can only be found using argument dependent
1907 lookup. This option causes friends to be injected as they were in
1910 This option is for compatibility, and may be removed in a future
1913 @item -fno-elide-constructors
1914 @opindex fno-elide-constructors
1915 The C++ standard allows an implementation to omit creating a temporary
1916 which is only used to initialize another object of the same type.
1917 Specifying this option disables that optimization, and forces G++ to
1918 call the copy constructor in all cases.
1920 @item -fno-enforce-eh-specs
1921 @opindex fno-enforce-eh-specs
1922 Don't generate code to check for violation of exception specifications
1923 at runtime. This option violates the C++ standard, but may be useful
1924 for reducing code size in production builds, much like defining
1925 @samp{NDEBUG}. This does not give user code permission to throw
1926 exceptions in violation of the exception specifications; the compiler
1927 will still optimize based on the specifications, so throwing an
1928 unexpected exception will result in undefined behavior.
1931 @itemx -fno-for-scope
1933 @opindex fno-for-scope
1934 If @option{-ffor-scope} is specified, the scope of variables declared in
1935 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1936 as specified by the C++ standard.
1937 If @option{-fno-for-scope} is specified, the scope of variables declared in
1938 a @i{for-init-statement} extends to the end of the enclosing scope,
1939 as was the case in old versions of G++, and other (traditional)
1940 implementations of C++.
1942 The default if neither flag is given to follow the standard,
1943 but to allow and give a warning for old-style code that would
1944 otherwise be invalid, or have different behavior.
1946 @item -fno-gnu-keywords
1947 @opindex fno-gnu-keywords
1948 Do not recognize @code{typeof} as a keyword, so that code can use this
1949 word as an identifier. You can use the keyword @code{__typeof__} instead.
1950 @option{-ansi} implies @option{-fno-gnu-keywords}.
1952 @item -fno-implicit-templates
1953 @opindex fno-implicit-templates
1954 Never emit code for non-inline templates which are instantiated
1955 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1956 @xref{Template Instantiation}, for more information.
1958 @item -fno-implicit-inline-templates
1959 @opindex fno-implicit-inline-templates
1960 Don't emit code for implicit instantiations of inline templates, either.
1961 The default is to handle inlines differently so that compiles with and
1962 without optimization will need the same set of explicit instantiations.
1964 @item -fno-implement-inlines
1965 @opindex fno-implement-inlines
1966 To save space, do not emit out-of-line copies of inline functions
1967 controlled by @samp{#pragma implementation}. This will cause linker
1968 errors if these functions are not inlined everywhere they are called.
1970 @item -fms-extensions
1971 @opindex fms-extensions
1972 Disable pedantic warnings about constructs used in MFC, such as implicit
1973 int and getting a pointer to member function via non-standard syntax.
1975 @item -fno-nonansi-builtins
1976 @opindex fno-nonansi-builtins
1977 Disable built-in declarations of functions that are not mandated by
1978 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1979 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1982 @opindex fnothrow-opt
1983 Treat a @code{throw()} exception specification as though it were a
1984 @code{noexcept} specification to reduce or eliminate the text size
1985 overhead relative to a function with no exception specification. If
1986 the function has local variables of types with non-trivial
1987 destructors, the exception specification will actually make the
1988 function smaller because the EH cleanups for those variables can be
1989 optimized away. The semantic effect is that an exception thrown out of
1990 a function with such an exception specification will result in a call
1991 to @code{terminate} rather than @code{unexpected}.
1993 @item -fno-operator-names
1994 @opindex fno-operator-names
1995 Do not treat the operator name keywords @code{and}, @code{bitand},
1996 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1997 synonyms as keywords.
1999 @item -fno-optional-diags
2000 @opindex fno-optional-diags
2001 Disable diagnostics that the standard says a compiler does not need to
2002 issue. Currently, the only such diagnostic issued by G++ is the one for
2003 a name having multiple meanings within a class.
2006 @opindex fpermissive
2007 Downgrade some diagnostics about nonconformant code from errors to
2008 warnings. Thus, using @option{-fpermissive} will allow some
2009 nonconforming code to compile.
2011 @item -fno-pretty-templates
2012 @opindex fno-pretty-templates
2013 When an error message refers to a specialization of a function
2014 template, the compiler will normally print the signature of the
2015 template followed by the template arguments and any typedefs or
2016 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2017 rather than @code{void f(int)}) so that it's clear which template is
2018 involved. When an error message refers to a specialization of a class
2019 template, the compiler will omit any template arguments which match
2020 the default template arguments for that template. If either of these
2021 behaviors make it harder to understand the error message rather than
2022 easier, using @option{-fno-pretty-templates} will disable them.
2026 Enable automatic template instantiation at link time. This option also
2027 implies @option{-fno-implicit-templates}. @xref{Template
2028 Instantiation}, for more information.
2032 Disable generation of information about every class with virtual
2033 functions for use by the C++ runtime type identification features
2034 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2035 of the language, you can save some space by using this flag. Note that
2036 exception handling uses the same information, but it will generate it as
2037 needed. The @samp{dynamic_cast} operator can still be used for casts that
2038 do not require runtime type information, i.e.@: casts to @code{void *} or to
2039 unambiguous base classes.
2043 Emit statistics about front-end processing at the end of the compilation.
2044 This information is generally only useful to the G++ development team.
2046 @item -fstrict-enums
2047 @opindex fstrict-enums
2048 Allow the compiler to optimize using the assumption that a value of
2049 enumeration type can only be one of the values of the enumeration (as
2050 defined in the C++ standard; basically, a value which can be
2051 represented in the minimum number of bits needed to represent all the
2052 enumerators). This assumption may not be valid if the program uses a
2053 cast to convert an arbitrary integer value to the enumeration type.
2055 @item -ftemplate-depth=@var{n}
2056 @opindex ftemplate-depth
2057 Set the maximum instantiation depth for template classes to @var{n}.
2058 A limit on the template instantiation depth is needed to detect
2059 endless recursions during template class instantiation. ANSI/ISO C++
2060 conforming programs must not rely on a maximum depth greater than 17
2061 (changed to 1024 in C++0x).
2063 @item -fno-threadsafe-statics
2064 @opindex fno-threadsafe-statics
2065 Do not emit the extra code to use the routines specified in the C++
2066 ABI for thread-safe initialization of local statics. You can use this
2067 option to reduce code size slightly in code that doesn't need to be
2070 @item -fuse-cxa-atexit
2071 @opindex fuse-cxa-atexit
2072 Register destructors for objects with static storage duration with the
2073 @code{__cxa_atexit} function rather than the @code{atexit} function.
2074 This option is required for fully standards-compliant handling of static
2075 destructors, but will only work if your C library supports
2076 @code{__cxa_atexit}.
2078 @item -fno-use-cxa-get-exception-ptr
2079 @opindex fno-use-cxa-get-exception-ptr
2080 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2081 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2082 if the runtime routine is not available.
2084 @item -fvisibility-inlines-hidden
2085 @opindex fvisibility-inlines-hidden
2086 This switch declares that the user does not attempt to compare
2087 pointers to inline methods where the addresses of the two functions
2088 were taken in different shared objects.
2090 The effect of this is that GCC may, effectively, mark inline methods with
2091 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2092 appear in the export table of a DSO and do not require a PLT indirection
2093 when used within the DSO@. Enabling this option can have a dramatic effect
2094 on load and link times of a DSO as it massively reduces the size of the
2095 dynamic export table when the library makes heavy use of templates.
2097 The behavior of this switch is not quite the same as marking the
2098 methods as hidden directly, because it does not affect static variables
2099 local to the function or cause the compiler to deduce that
2100 the function is defined in only one shared object.
2102 You may mark a method as having a visibility explicitly to negate the
2103 effect of the switch for that method. For example, if you do want to
2104 compare pointers to a particular inline method, you might mark it as
2105 having default visibility. Marking the enclosing class with explicit
2106 visibility will have no effect.
2108 Explicitly instantiated inline methods are unaffected by this option
2109 as their linkage might otherwise cross a shared library boundary.
2110 @xref{Template Instantiation}.
2112 @item -fvisibility-ms-compat
2113 @opindex fvisibility-ms-compat
2114 This flag attempts to use visibility settings to make GCC's C++
2115 linkage model compatible with that of Microsoft Visual Studio.
2117 The flag makes these changes to GCC's linkage model:
2121 It sets the default visibility to @code{hidden}, like
2122 @option{-fvisibility=hidden}.
2125 Types, but not their members, are not hidden by default.
2128 The One Definition Rule is relaxed for types without explicit
2129 visibility specifications which are defined in more than one different
2130 shared object: those declarations are permitted if they would have
2131 been permitted when this option was not used.
2134 In new code it is better to use @option{-fvisibility=hidden} and
2135 export those classes which are intended to be externally visible.
2136 Unfortunately it is possible for code to rely, perhaps accidentally,
2137 on the Visual Studio behavior.
2139 Among the consequences of these changes are that static data members
2140 of the same type with the same name but defined in different shared
2141 objects will be different, so changing one will not change the other;
2142 and that pointers to function members defined in different shared
2143 objects may not compare equal. When this flag is given, it is a
2144 violation of the ODR to define types with the same name differently.
2148 Do not use weak symbol support, even if it is provided by the linker.
2149 By default, G++ will use weak symbols if they are available. This
2150 option exists only for testing, and should not be used by end-users;
2151 it will result in inferior code and has no benefits. This option may
2152 be removed in a future release of G++.
2156 Do not search for header files in the standard directories specific to
2157 C++, but do still search the other standard directories. (This option
2158 is used when building the C++ library.)
2161 In addition, these optimization, warning, and code generation options
2162 have meanings only for C++ programs:
2165 @item -fno-default-inline
2166 @opindex fno-default-inline
2167 Do not assume @samp{inline} for functions defined inside a class scope.
2168 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2169 functions will have linkage like inline functions; they just won't be
2172 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2175 Warn when G++ generates code that is probably not compatible with the
2176 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2177 all such cases, there are probably some cases that are not warned about,
2178 even though G++ is generating incompatible code. There may also be
2179 cases where warnings are emitted even though the code that is generated
2182 You should rewrite your code to avoid these warnings if you are
2183 concerned about the fact that code generated by G++ may not be binary
2184 compatible with code generated by other compilers.
2186 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2191 A template with a non-type template parameter of reference type is
2192 mangled incorrectly:
2195 template <int &> struct S @{@};
2199 This is fixed in @option{-fabi-version=3}.
2202 SIMD vector types declared using @code{__attribute ((vector_size))} are
2203 mangled in a non-standard way that does not allow for overloading of
2204 functions taking vectors of different sizes.
2206 The mangling is changed in @option{-fabi-version=4}.
2209 The known incompatibilities in @option{-fabi-version=1} include:
2214 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2215 pack data into the same byte as a base class. For example:
2218 struct A @{ virtual void f(); int f1 : 1; @};
2219 struct B : public A @{ int f2 : 1; @};
2223 In this case, G++ will place @code{B::f2} into the same byte
2224 as@code{A::f1}; other compilers will not. You can avoid this problem
2225 by explicitly padding @code{A} so that its size is a multiple of the
2226 byte size on your platform; that will cause G++ and other compilers to
2227 layout @code{B} identically.
2230 Incorrect handling of tail-padding for virtual bases. G++ does not use
2231 tail padding when laying out virtual bases. For example:
2234 struct A @{ virtual void f(); char c1; @};
2235 struct B @{ B(); char c2; @};
2236 struct C : public A, public virtual B @{@};
2240 In this case, G++ will not place @code{B} into the tail-padding for
2241 @code{A}; other compilers will. You can avoid this problem by
2242 explicitly padding @code{A} so that its size is a multiple of its
2243 alignment (ignoring virtual base classes); that will cause G++ and other
2244 compilers to layout @code{C} identically.
2247 Incorrect handling of bit-fields with declared widths greater than that
2248 of their underlying types, when the bit-fields appear in a union. For
2252 union U @{ int i : 4096; @};
2256 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2257 union too small by the number of bits in an @code{int}.
2260 Empty classes can be placed at incorrect offsets. For example:
2270 struct C : public B, public A @{@};
2274 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2275 it should be placed at offset zero. G++ mistakenly believes that the
2276 @code{A} data member of @code{B} is already at offset zero.
2279 Names of template functions whose types involve @code{typename} or
2280 template template parameters can be mangled incorrectly.
2283 template <typename Q>
2284 void f(typename Q::X) @{@}
2286 template <template <typename> class Q>
2287 void f(typename Q<int>::X) @{@}
2291 Instantiations of these templates may be mangled incorrectly.
2295 It also warns psABI related changes. The known psABI changes at this
2301 For SYSV/x86-64, when passing union with long double, it is changed to
2302 pass in memory as specified in psABI. For example:
2312 @code{union U} will always be passed in memory.
2316 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2317 @opindex Wctor-dtor-privacy
2318 @opindex Wno-ctor-dtor-privacy
2319 Warn when a class seems unusable because all the constructors or
2320 destructors in that class are private, and it has neither friends nor
2321 public static member functions.
2323 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2325 @opindex Wno-noexcept
2326 Warn when a noexcept-expression evaluates to false because of a call
2327 to a function that does not have a non-throwing exception
2328 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2329 the compiler to never throw an exception.
2331 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2332 @opindex Wnon-virtual-dtor
2333 @opindex Wno-non-virtual-dtor
2334 Warn when a class has virtual functions and accessible non-virtual
2335 destructor, in which case it would be possible but unsafe to delete
2336 an instance of a derived class through a pointer to the base class.
2337 This warning is also enabled if -Weffc++ is specified.
2339 @item -Wreorder @r{(C++ and Objective-C++ only)}
2341 @opindex Wno-reorder
2342 @cindex reordering, warning
2343 @cindex warning for reordering of member initializers
2344 Warn when the order of member initializers given in the code does not
2345 match the order in which they must be executed. For instance:
2351 A(): j (0), i (1) @{ @}
2355 The compiler will rearrange the member initializers for @samp{i}
2356 and @samp{j} to match the declaration order of the members, emitting
2357 a warning to that effect. This warning is enabled by @option{-Wall}.
2360 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2363 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2366 Warn about violations of the following style guidelines from Scott Meyers'
2367 @cite{Effective C++} book:
2371 Item 11: Define a copy constructor and an assignment operator for classes
2372 with dynamically allocated memory.
2375 Item 12: Prefer initialization to assignment in constructors.
2378 Item 14: Make destructors virtual in base classes.
2381 Item 15: Have @code{operator=} return a reference to @code{*this}.
2384 Item 23: Don't try to return a reference when you must return an object.
2388 Also warn about violations of the following style guidelines from
2389 Scott Meyers' @cite{More Effective C++} book:
2393 Item 6: Distinguish between prefix and postfix forms of increment and
2394 decrement operators.
2397 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2401 When selecting this option, be aware that the standard library
2402 headers do not obey all of these guidelines; use @samp{grep -v}
2403 to filter out those warnings.
2405 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2406 @opindex Wstrict-null-sentinel
2407 @opindex Wno-strict-null-sentinel
2408 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2409 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2410 to @code{__null}. Although it is a null pointer constant not a null pointer,
2411 it is guaranteed to be of the same size as a pointer. But this use is
2412 not portable across different compilers.
2414 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2415 @opindex Wno-non-template-friend
2416 @opindex Wnon-template-friend
2417 Disable warnings when non-templatized friend functions are declared
2418 within a template. Since the advent of explicit template specification
2419 support in G++, if the name of the friend is an unqualified-id (i.e.,
2420 @samp{friend foo(int)}), the C++ language specification demands that the
2421 friend declare or define an ordinary, nontemplate function. (Section
2422 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2423 could be interpreted as a particular specialization of a templatized
2424 function. Because this non-conforming behavior is no longer the default
2425 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2426 check existing code for potential trouble spots and is on by default.
2427 This new compiler behavior can be turned off with
2428 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2429 but disables the helpful warning.
2431 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2432 @opindex Wold-style-cast
2433 @opindex Wno-old-style-cast
2434 Warn if an old-style (C-style) cast to a non-void type is used within
2435 a C++ program. The new-style casts (@samp{dynamic_cast},
2436 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2437 less vulnerable to unintended effects and much easier to search for.
2439 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2440 @opindex Woverloaded-virtual
2441 @opindex Wno-overloaded-virtual
2442 @cindex overloaded virtual function, warning
2443 @cindex warning for overloaded virtual function
2444 Warn when a function declaration hides virtual functions from a
2445 base class. For example, in:
2452 struct B: public A @{
2457 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2465 will fail to compile.
2467 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2468 @opindex Wno-pmf-conversions
2469 @opindex Wpmf-conversions
2470 Disable the diagnostic for converting a bound pointer to member function
2473 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2474 @opindex Wsign-promo
2475 @opindex Wno-sign-promo
2476 Warn when overload resolution chooses a promotion from unsigned or
2477 enumerated type to a signed type, over a conversion to an unsigned type of
2478 the same size. Previous versions of G++ would try to preserve
2479 unsignedness, but the standard mandates the current behavior.
2484 A& operator = (int);
2494 In this example, G++ will synthesize a default @samp{A& operator =
2495 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2498 @node Objective-C and Objective-C++ Dialect Options
2499 @section Options Controlling Objective-C and Objective-C++ Dialects
2501 @cindex compiler options, Objective-C and Objective-C++
2502 @cindex Objective-C and Objective-C++ options, command line
2503 @cindex options, Objective-C and Objective-C++
2504 (NOTE: This manual does not describe the Objective-C and Objective-C++
2505 languages themselves. See @xref{Standards,,Language Standards
2506 Supported by GCC}, for references.)
2508 This section describes the command-line options that are only meaningful
2509 for Objective-C and Objective-C++ programs, but you can also use most of
2510 the language-independent GNU compiler options.
2511 For example, you might compile a file @code{some_class.m} like this:
2514 gcc -g -fgnu-runtime -O -c some_class.m
2518 In this example, @option{-fgnu-runtime} is an option meant only for
2519 Objective-C and Objective-C++ programs; you can use the other options with
2520 any language supported by GCC@.
2522 Note that since Objective-C is an extension of the C language, Objective-C
2523 compilations may also use options specific to the C front-end (e.g.,
2524 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2525 C++-specific options (e.g., @option{-Wabi}).
2527 Here is a list of options that are @emph{only} for compiling Objective-C
2528 and Objective-C++ programs:
2531 @item -fconstant-string-class=@var{class-name}
2532 @opindex fconstant-string-class
2533 Use @var{class-name} as the name of the class to instantiate for each
2534 literal string specified with the syntax @code{@@"@dots{}"}. The default
2535 class name is @code{NXConstantString} if the GNU runtime is being used, and
2536 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2537 @option{-fconstant-cfstrings} option, if also present, will override the
2538 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2539 to be laid out as constant CoreFoundation strings.
2542 @opindex fgnu-runtime
2543 Generate object code compatible with the standard GNU Objective-C
2544 runtime. This is the default for most types of systems.
2546 @item -fnext-runtime
2547 @opindex fnext-runtime
2548 Generate output compatible with the NeXT runtime. This is the default
2549 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2550 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2553 @item -fno-nil-receivers
2554 @opindex fno-nil-receivers
2555 Assume that all Objective-C message dispatches (@code{[receiver
2556 message:arg]}) in this translation unit ensure that the receiver is
2557 not @code{nil}. This allows for more efficient entry points in the
2558 runtime to be used. Currently, this option is only available in
2559 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2561 @item -fobjc-call-cxx-cdtors
2562 @opindex fobjc-call-cxx-cdtors
2563 For each Objective-C class, check if any of its instance variables is a
2564 C++ object with a non-trivial default constructor. If so, synthesize a
2565 special @code{- (id) .cxx_construct} instance method that will run
2566 non-trivial default constructors on any such instance variables, in order,
2567 and then return @code{self}. Similarly, check if any instance variable
2568 is a C++ object with a non-trivial destructor, and if so, synthesize a
2569 special @code{- (void) .cxx_destruct} method that will run
2570 all such default destructors, in reverse order.
2572 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2573 methods thusly generated will only operate on instance variables
2574 declared in the current Objective-C class, and not those inherited
2575 from superclasses. It is the responsibility of the Objective-C
2576 runtime to invoke all such methods in an object's inheritance
2577 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2578 by the runtime immediately after a new object instance is allocated;
2579 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2580 before the runtime deallocates an object instance.
2582 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2583 support for invoking the @code{- (id) .cxx_construct} and
2584 @code{- (void) .cxx_destruct} methods.
2586 @item -fobjc-direct-dispatch
2587 @opindex fobjc-direct-dispatch
2588 Allow fast jumps to the message dispatcher. On Darwin this is
2589 accomplished via the comm page.
2591 @item -fobjc-exceptions
2592 @opindex fobjc-exceptions
2593 Enable syntactic support for structured exception handling in
2594 Objective-C, similar to what is offered by C++ and Java. This option
2595 is required to use the Objective-C keywords @code{@@try},
2596 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2597 @code{@@synchronized}. This option is available with both the GNU
2598 runtime and the NeXT runtime (but not available in conjunction with
2599 the NeXT runtime on Mac OS X 10.2 and earlier).
2603 Enable garbage collection (GC) in Objective-C and Objective-C++
2604 programs. This option is only available with the NeXT runtime; the
2605 GNU runtime has a different garbage collection implementation that
2606 does not require special compiler flags.
2608 @item -fobjc-std=objc1
2610 Conform to the language syntax of Objective-C 1.0, the language
2611 recognized by GCC 4.0. This only affects the Objective-C additions to
2612 the C/C++ language; it does not affect conformance to C/C++ standards,
2613 which is controlled by the separate C/C++ dialect option flags. When
2614 this option is used with the Objective-C or Objective-C++ compiler,
2615 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2616 This is useful if you need to make sure that your Objective-C code can
2617 be compiled with older versions of GCC.
2619 @item -freplace-objc-classes
2620 @opindex freplace-objc-classes
2621 Emit a special marker instructing @command{ld(1)} not to statically link in
2622 the resulting object file, and allow @command{dyld(1)} to load it in at
2623 run time instead. This is used in conjunction with the Fix-and-Continue
2624 debugging mode, where the object file in question may be recompiled and
2625 dynamically reloaded in the course of program execution, without the need
2626 to restart the program itself. Currently, Fix-and-Continue functionality
2627 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2632 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2633 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2634 compile time) with static class references that get initialized at load time,
2635 which improves run-time performance. Specifying the @option{-fzero-link} flag
2636 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2637 to be retained. This is useful in Zero-Link debugging mode, since it allows
2638 for individual class implementations to be modified during program execution.
2639 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2640 regardless of command line options.
2644 Dump interface declarations for all classes seen in the source file to a
2645 file named @file{@var{sourcename}.decl}.
2647 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2648 @opindex Wassign-intercept
2649 @opindex Wno-assign-intercept
2650 Warn whenever an Objective-C assignment is being intercepted by the
2653 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2654 @opindex Wno-protocol
2656 If a class is declared to implement a protocol, a warning is issued for
2657 every method in the protocol that is not implemented by the class. The
2658 default behavior is to issue a warning for every method not explicitly
2659 implemented in the class, even if a method implementation is inherited
2660 from the superclass. If you use the @option{-Wno-protocol} option, then
2661 methods inherited from the superclass are considered to be implemented,
2662 and no warning is issued for them.
2664 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2666 @opindex Wno-selector
2667 Warn if multiple methods of different types for the same selector are
2668 found during compilation. The check is performed on the list of methods
2669 in the final stage of compilation. Additionally, a check is performed
2670 for each selector appearing in a @code{@@selector(@dots{})}
2671 expression, and a corresponding method for that selector has been found
2672 during compilation. Because these checks scan the method table only at
2673 the end of compilation, these warnings are not produced if the final
2674 stage of compilation is not reached, for example because an error is
2675 found during compilation, or because the @option{-fsyntax-only} option is
2678 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2679 @opindex Wstrict-selector-match
2680 @opindex Wno-strict-selector-match
2681 Warn if multiple methods with differing argument and/or return types are
2682 found for a given selector when attempting to send a message using this
2683 selector to a receiver of type @code{id} or @code{Class}. When this flag
2684 is off (which is the default behavior), the compiler will omit such warnings
2685 if any differences found are confined to types which share the same size
2688 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2689 @opindex Wundeclared-selector
2690 @opindex Wno-undeclared-selector
2691 Warn if a @code{@@selector(@dots{})} expression referring to an
2692 undeclared selector is found. A selector is considered undeclared if no
2693 method with that name has been declared before the
2694 @code{@@selector(@dots{})} expression, either explicitly in an
2695 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2696 an @code{@@implementation} section. This option always performs its
2697 checks as soon as a @code{@@selector(@dots{})} expression is found,
2698 while @option{-Wselector} only performs its checks in the final stage of
2699 compilation. This also enforces the coding style convention
2700 that methods and selectors must be declared before being used.
2702 @item -print-objc-runtime-info
2703 @opindex print-objc-runtime-info
2704 Generate C header describing the largest structure that is passed by
2709 @node Language Independent Options
2710 @section Options to Control Diagnostic Messages Formatting
2711 @cindex options to control diagnostics formatting
2712 @cindex diagnostic messages
2713 @cindex message formatting
2715 Traditionally, diagnostic messages have been formatted irrespective of
2716 the output device's aspect (e.g.@: its width, @dots{}). The options described
2717 below can be used to control the diagnostic messages formatting
2718 algorithm, e.g.@: how many characters per line, how often source location
2719 information should be reported. Right now, only the C++ front end can
2720 honor these options. However it is expected, in the near future, that
2721 the remaining front ends would be able to digest them correctly.
2724 @item -fmessage-length=@var{n}
2725 @opindex fmessage-length
2726 Try to format error messages so that they fit on lines of about @var{n}
2727 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2728 the front ends supported by GCC@. If @var{n} is zero, then no
2729 line-wrapping will be done; each error message will appear on a single
2732 @opindex fdiagnostics-show-location
2733 @item -fdiagnostics-show-location=once
2734 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2735 reporter to emit @emph{once} source location information; that is, in
2736 case the message is too long to fit on a single physical line and has to
2737 be wrapped, the source location won't be emitted (as prefix) again,
2738 over and over, in subsequent continuation lines. This is the default
2741 @item -fdiagnostics-show-location=every-line
2742 Only meaningful in line-wrapping mode. Instructs the diagnostic
2743 messages reporter to emit the same source location information (as
2744 prefix) for physical lines that result from the process of breaking
2745 a message which is too long to fit on a single line.
2747 @item -fdiagnostics-show-option
2748 @opindex fdiagnostics-show-option
2749 This option instructs the diagnostic machinery to add text to each
2750 diagnostic emitted, which indicates which command line option directly
2751 controls that diagnostic, when such an option is known to the
2752 diagnostic machinery.
2754 @item -Wcoverage-mismatch
2755 @opindex Wcoverage-mismatch
2756 Warn if feedback profiles do not match when using the
2757 @option{-fprofile-use} option.
2758 If a source file was changed between @option{-fprofile-gen} and
2759 @option{-fprofile-use}, the files with the profile feedback can fail
2760 to match the source file and GCC can not use the profile feedback
2761 information. By default, this warning is enabled and is treated as an
2762 error. @option{-Wno-coverage-mismatch} can be used to disable the
2763 warning or @option{-Wno-error=coverage-mismatch} can be used to
2764 disable the error. Disable the error for this warning can result in
2765 poorly optimized code, so disabling the error is useful only in the
2766 case of very minor changes such as bug fixes to an existing code-base.
2767 Completely disabling the warning is not recommended.
2771 @node Warning Options
2772 @section Options to Request or Suppress Warnings
2773 @cindex options to control warnings
2774 @cindex warning messages
2775 @cindex messages, warning
2776 @cindex suppressing warnings
2778 Warnings are diagnostic messages that report constructions which
2779 are not inherently erroneous but which are risky or suggest there
2780 may have been an error.
2782 The following language-independent options do not enable specific
2783 warnings but control the kinds of diagnostics produced by GCC.
2786 @cindex syntax checking
2788 @opindex fsyntax-only
2789 Check the code for syntax errors, but don't do anything beyond that.
2791 @item -fmax-errors=@var{n}
2792 @opindex fmax-errors
2793 Limits the maximum number of error messages to @var{n}, at which point
2794 GCC bails out rather than attempting to continue processing the source
2795 code. If @var{n} is 0 (the default), there is no limit on the number
2796 of error messages produced. If @option{-Wfatal-errors} is also
2797 specified, then @option{-Wfatal-errors} takes precedence over this
2802 Inhibit all warning messages.
2807 Make all warnings into errors.
2812 Make the specified warning into an error. The specifier for a warning
2813 is appended, for example @option{-Werror=switch} turns the warnings
2814 controlled by @option{-Wswitch} into errors. This switch takes a
2815 negative form, to be used to negate @option{-Werror} for specific
2816 warnings, for example @option{-Wno-error=switch} makes
2817 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2818 is in effect. You can use the @option{-fdiagnostics-show-option}
2819 option to have each controllable warning amended with the option which
2820 controls it, to determine what to use with this option.
2822 Note that specifying @option{-Werror=}@var{foo} automatically implies
2823 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2826 @item -Wfatal-errors
2827 @opindex Wfatal-errors
2828 @opindex Wno-fatal-errors
2829 This option causes the compiler to abort compilation on the first error
2830 occurred rather than trying to keep going and printing further error
2835 You can request many specific warnings with options beginning
2836 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2837 implicit declarations. Each of these specific warning options also
2838 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2839 example, @option{-Wno-implicit}. This manual lists only one of the
2840 two forms, whichever is not the default. For further,
2841 language-specific options also refer to @ref{C++ Dialect Options} and
2842 @ref{Objective-C and Objective-C++ Dialect Options}.
2844 When an unrecognized warning option is requested (e.g.,
2845 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2846 that the option is not recognized. However, if the @option{-Wno-} form
2847 is used, the behavior is slightly different: No diagnostic will be
2848 produced for @option{-Wno-unknown-warning} unless other diagnostics
2849 are being produced. This allows the use of new @option{-Wno-} options
2850 with old compilers, but if something goes wrong, the compiler will
2851 warn that an unrecognized option was used.
2856 Issue all the warnings demanded by strict ISO C and ISO C++;
2857 reject all programs that use forbidden extensions, and some other
2858 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2859 version of the ISO C standard specified by any @option{-std} option used.
2861 Valid ISO C and ISO C++ programs should compile properly with or without
2862 this option (though a rare few will require @option{-ansi} or a
2863 @option{-std} option specifying the required version of ISO C)@. However,
2864 without this option, certain GNU extensions and traditional C and C++
2865 features are supported as well. With this option, they are rejected.
2867 @option{-pedantic} does not cause warning messages for use of the
2868 alternate keywords whose names begin and end with @samp{__}. Pedantic
2869 warnings are also disabled in the expression that follows
2870 @code{__extension__}. However, only system header files should use
2871 these escape routes; application programs should avoid them.
2872 @xref{Alternate Keywords}.
2874 Some users try to use @option{-pedantic} to check programs for strict ISO
2875 C conformance. They soon find that it does not do quite what they want:
2876 it finds some non-ISO practices, but not all---only those for which
2877 ISO C @emph{requires} a diagnostic, and some others for which
2878 diagnostics have been added.
2880 A feature to report any failure to conform to ISO C might be useful in
2881 some instances, but would require considerable additional work and would
2882 be quite different from @option{-pedantic}. We don't have plans to
2883 support such a feature in the near future.
2885 Where the standard specified with @option{-std} represents a GNU
2886 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2887 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2888 extended dialect is based. Warnings from @option{-pedantic} are given
2889 where they are required by the base standard. (It would not make sense
2890 for such warnings to be given only for features not in the specified GNU
2891 C dialect, since by definition the GNU dialects of C include all
2892 features the compiler supports with the given option, and there would be
2893 nothing to warn about.)
2895 @item -pedantic-errors
2896 @opindex pedantic-errors
2897 Like @option{-pedantic}, except that errors are produced rather than
2903 This enables all the warnings about constructions that some users
2904 consider questionable, and that are easy to avoid (or modify to
2905 prevent the warning), even in conjunction with macros. This also
2906 enables some language-specific warnings described in @ref{C++ Dialect
2907 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2909 @option{-Wall} turns on the following warning flags:
2911 @gccoptlist{-Waddress @gol
2912 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2914 -Wchar-subscripts @gol
2915 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2916 -Wimplicit-int @r{(C and Objective-C only)} @gol
2917 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2920 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2921 -Wmissing-braces @gol
2927 -Wsequence-point @gol
2928 -Wsign-compare @r{(only in C++)} @gol
2929 -Wstrict-aliasing @gol
2930 -Wstrict-overflow=1 @gol
2933 -Wuninitialized @gol
2934 -Wunknown-pragmas @gol
2935 -Wunused-function @gol
2938 -Wunused-variable @gol
2939 -Wvolatile-register-var @gol
2942 Note that some warning flags are not implied by @option{-Wall}. Some of
2943 them warn about constructions that users generally do not consider
2944 questionable, but which occasionally you might wish to check for;
2945 others warn about constructions that are necessary or hard to avoid in
2946 some cases, and there is no simple way to modify the code to suppress
2947 the warning. Some of them are enabled by @option{-Wextra} but many of
2948 them must be enabled individually.
2954 This enables some extra warning flags that are not enabled by
2955 @option{-Wall}. (This option used to be called @option{-W}. The older
2956 name is still supported, but the newer name is more descriptive.)
2958 @gccoptlist{-Wclobbered @gol
2960 -Wignored-qualifiers @gol
2961 -Wmissing-field-initializers @gol
2962 -Wmissing-parameter-type @r{(C only)} @gol
2963 -Wold-style-declaration @r{(C only)} @gol
2964 -Woverride-init @gol
2967 -Wuninitialized @gol
2968 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2969 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2972 The option @option{-Wextra} also prints warning messages for the
2978 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2979 @samp{>}, or @samp{>=}.
2982 (C++ only) An enumerator and a non-enumerator both appear in a
2983 conditional expression.
2986 (C++ only) Ambiguous virtual bases.
2989 (C++ only) Subscripting an array which has been declared @samp{register}.
2992 (C++ only) Taking the address of a variable which has been declared
2996 (C++ only) A base class is not initialized in a derived class' copy
3001 @item -Wchar-subscripts
3002 @opindex Wchar-subscripts
3003 @opindex Wno-char-subscripts
3004 Warn if an array subscript has type @code{char}. This is a common cause
3005 of error, as programmers often forget that this type is signed on some
3007 This warning is enabled by @option{-Wall}.
3011 @opindex Wno-comment
3012 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3013 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3014 This warning is enabled by @option{-Wall}.
3017 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3019 Suppress warning messages emitted by @code{#warning} directives.
3021 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3022 @opindex Wdouble-promotion
3023 @opindex Wno-double-promotion
3024 Give a warning when a value of type @code{float} is implicitly
3025 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3026 floating-point unit implement @code{float} in hardware, but emulate
3027 @code{double} in software. On such a machine, doing computations
3028 using @code{double} values is much more expensive because of the
3029 overhead required for software emulation.
3031 It is easy to accidentally do computations with @code{double} because
3032 floating-point literals are implicitly of type @code{double}. For
3036 float area(float radius)
3038 return 3.14159 * radius * radius;
3042 the compiler will perform the entire computation with @code{double}
3043 because the floating-point literal is a @code{double}.
3048 @opindex ffreestanding
3049 @opindex fno-builtin
3050 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3051 the arguments supplied have types appropriate to the format string
3052 specified, and that the conversions specified in the format string make
3053 sense. This includes standard functions, and others specified by format
3054 attributes (@pxref{Function Attributes}), in the @code{printf},
3055 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3056 not in the C standard) families (or other target-specific families).
3057 Which functions are checked without format attributes having been
3058 specified depends on the standard version selected, and such checks of
3059 functions without the attribute specified are disabled by
3060 @option{-ffreestanding} or @option{-fno-builtin}.
3062 The formats are checked against the format features supported by GNU
3063 libc version 2.2. These include all ISO C90 and C99 features, as well
3064 as features from the Single Unix Specification and some BSD and GNU
3065 extensions. Other library implementations may not support all these
3066 features; GCC does not support warning about features that go beyond a
3067 particular library's limitations. However, if @option{-pedantic} is used
3068 with @option{-Wformat}, warnings will be given about format features not
3069 in the selected standard version (but not for @code{strfmon} formats,
3070 since those are not in any version of the C standard). @xref{C Dialect
3071 Options,,Options Controlling C Dialect}.
3073 Since @option{-Wformat} also checks for null format arguments for
3074 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3076 @option{-Wformat} is included in @option{-Wall}. For more control over some
3077 aspects of format checking, the options @option{-Wformat-y2k},
3078 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3079 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3080 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3083 @opindex Wformat-y2k
3084 @opindex Wno-format-y2k
3085 If @option{-Wformat} is specified, also warn about @code{strftime}
3086 formats which may yield only a two-digit year.
3088 @item -Wno-format-contains-nul
3089 @opindex Wno-format-contains-nul
3090 @opindex Wformat-contains-nul
3091 If @option{-Wformat} is specified, do not warn about format strings that
3094 @item -Wno-format-extra-args
3095 @opindex Wno-format-extra-args
3096 @opindex Wformat-extra-args
3097 If @option{-Wformat} is specified, do not warn about excess arguments to a
3098 @code{printf} or @code{scanf} format function. The C standard specifies
3099 that such arguments are ignored.
3101 Where the unused arguments lie between used arguments that are
3102 specified with @samp{$} operand number specifications, normally
3103 warnings are still given, since the implementation could not know what
3104 type to pass to @code{va_arg} to skip the unused arguments. However,
3105 in the case of @code{scanf} formats, this option will suppress the
3106 warning if the unused arguments are all pointers, since the Single
3107 Unix Specification says that such unused arguments are allowed.
3109 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3110 @opindex Wno-format-zero-length
3111 @opindex Wformat-zero-length
3112 If @option{-Wformat} is specified, do not warn about zero-length formats.
3113 The C standard specifies that zero-length formats are allowed.
3115 @item -Wformat-nonliteral
3116 @opindex Wformat-nonliteral
3117 @opindex Wno-format-nonliteral
3118 If @option{-Wformat} is specified, also warn if the format string is not a
3119 string literal and so cannot be checked, unless the format function
3120 takes its format arguments as a @code{va_list}.
3122 @item -Wformat-security
3123 @opindex Wformat-security
3124 @opindex Wno-format-security
3125 If @option{-Wformat} is specified, also warn about uses of format
3126 functions that represent possible security problems. At present, this
3127 warns about calls to @code{printf} and @code{scanf} functions where the
3128 format string is not a string literal and there are no format arguments,
3129 as in @code{printf (foo);}. This may be a security hole if the format
3130 string came from untrusted input and contains @samp{%n}. (This is
3131 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3132 in future warnings may be added to @option{-Wformat-security} that are not
3133 included in @option{-Wformat-nonliteral}.)
3137 @opindex Wno-format=2
3138 Enable @option{-Wformat} plus format checks not included in
3139 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3140 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3142 @item -Wnonnull @r{(C and Objective-C only)}
3144 @opindex Wno-nonnull
3145 Warn about passing a null pointer for arguments marked as
3146 requiring a non-null value by the @code{nonnull} function attribute.
3148 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3149 can be disabled with the @option{-Wno-nonnull} option.
3151 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3153 @opindex Wno-init-self
3154 Warn about uninitialized variables which are initialized with themselves.
3155 Note this option can only be used with the @option{-Wuninitialized} option.
3157 For example, GCC will warn about @code{i} being uninitialized in the
3158 following snippet only when @option{-Winit-self} has been specified:
3169 @item -Wimplicit-int @r{(C and Objective-C only)}
3170 @opindex Wimplicit-int
3171 @opindex Wno-implicit-int
3172 Warn when a declaration does not specify a type.
3173 This warning is enabled by @option{-Wall}.
3175 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3176 @opindex Wimplicit-function-declaration
3177 @opindex Wno-implicit-function-declaration
3178 Give a warning whenever a function is used before being declared. In
3179 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3180 enabled by default and it is made into an error by
3181 @option{-pedantic-errors}. This warning is also enabled by
3184 @item -Wimplicit @r{(C and Objective-C only)}
3186 @opindex Wno-implicit
3187 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3188 This warning is enabled by @option{-Wall}.
3190 @item -Wignored-qualifiers @r{(C and C++ only)}
3191 @opindex Wignored-qualifiers
3192 @opindex Wno-ignored-qualifiers
3193 Warn if the return type of a function has a type qualifier
3194 such as @code{const}. For ISO C such a type qualifier has no effect,
3195 since the value returned by a function is not an lvalue.
3196 For C++, the warning is only emitted for scalar types or @code{void}.
3197 ISO C prohibits qualified @code{void} return types on function
3198 definitions, so such return types always receive a warning
3199 even without this option.
3201 This warning is also enabled by @option{-Wextra}.
3206 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3207 a function with external linkage, returning int, taking either zero
3208 arguments, two, or three arguments of appropriate types. This warning
3209 is enabled by default in C++ and is enabled by either @option{-Wall}
3210 or @option{-pedantic}.
3212 @item -Wmissing-braces
3213 @opindex Wmissing-braces
3214 @opindex Wno-missing-braces
3215 Warn if an aggregate or union initializer is not fully bracketed. In
3216 the following example, the initializer for @samp{a} is not fully
3217 bracketed, but that for @samp{b} is fully bracketed.
3220 int a[2][2] = @{ 0, 1, 2, 3 @};
3221 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3224 This warning is enabled by @option{-Wall}.
3226 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3227 @opindex Wmissing-include-dirs
3228 @opindex Wno-missing-include-dirs
3229 Warn if a user-supplied include directory does not exist.
3232 @opindex Wparentheses
3233 @opindex Wno-parentheses
3234 Warn if parentheses are omitted in certain contexts, such
3235 as when there is an assignment in a context where a truth value
3236 is expected, or when operators are nested whose precedence people
3237 often get confused about.
3239 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3240 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3241 interpretation from that of ordinary mathematical notation.
3243 Also warn about constructions where there may be confusion to which
3244 @code{if} statement an @code{else} branch belongs. Here is an example of
3259 In C/C++, every @code{else} branch belongs to the innermost possible
3260 @code{if} statement, which in this example is @code{if (b)}. This is
3261 often not what the programmer expected, as illustrated in the above
3262 example by indentation the programmer chose. When there is the
3263 potential for this confusion, GCC will issue a warning when this flag
3264 is specified. To eliminate the warning, add explicit braces around
3265 the innermost @code{if} statement so there is no way the @code{else}
3266 could belong to the enclosing @code{if}. The resulting code would
3283 Also warn for dangerous uses of the
3284 ?: with omitted middle operand GNU extension. When the condition
3285 in the ?: operator is a boolean expression the omitted value will
3286 be always 1. Often the user expects it to be a value computed
3287 inside the conditional expression instead.
3289 This warning is enabled by @option{-Wall}.
3291 @item -Wsequence-point
3292 @opindex Wsequence-point
3293 @opindex Wno-sequence-point
3294 Warn about code that may have undefined semantics because of violations
3295 of sequence point rules in the C and C++ standards.
3297 The C and C++ standards defines the order in which expressions in a C/C++
3298 program are evaluated in terms of @dfn{sequence points}, which represent
3299 a partial ordering between the execution of parts of the program: those
3300 executed before the sequence point, and those executed after it. These
3301 occur after the evaluation of a full expression (one which is not part
3302 of a larger expression), after the evaluation of the first operand of a
3303 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3304 function is called (but after the evaluation of its arguments and the
3305 expression denoting the called function), and in certain other places.
3306 Other than as expressed by the sequence point rules, the order of
3307 evaluation of subexpressions of an expression is not specified. All
3308 these rules describe only a partial order rather than a total order,
3309 since, for example, if two functions are called within one expression
3310 with no sequence point between them, the order in which the functions
3311 are called is not specified. However, the standards committee have
3312 ruled that function calls do not overlap.
3314 It is not specified when between sequence points modifications to the
3315 values of objects take effect. Programs whose behavior depends on this
3316 have undefined behavior; the C and C++ standards specify that ``Between
3317 the previous and next sequence point an object shall have its stored
3318 value modified at most once by the evaluation of an expression.
3319 Furthermore, the prior value shall be read only to determine the value
3320 to be stored.''. If a program breaks these rules, the results on any
3321 particular implementation are entirely unpredictable.
3323 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3324 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3325 diagnosed by this option, and it may give an occasional false positive
3326 result, but in general it has been found fairly effective at detecting
3327 this sort of problem in programs.
3329 The standard is worded confusingly, therefore there is some debate
3330 over the precise meaning of the sequence point rules in subtle cases.
3331 Links to discussions of the problem, including proposed formal
3332 definitions, may be found on the GCC readings page, at
3333 @uref{http://gcc.gnu.org/@/readings.html}.
3335 This warning is enabled by @option{-Wall} for C and C++.
3338 @opindex Wreturn-type
3339 @opindex Wno-return-type
3340 Warn whenever a function is defined with a return-type that defaults
3341 to @code{int}. Also warn about any @code{return} statement with no
3342 return-value in a function whose return-type is not @code{void}
3343 (falling off the end of the function body is considered returning
3344 without a value), and about a @code{return} statement with an
3345 expression in a function whose return-type is @code{void}.
3347 For C++, a function without return type always produces a diagnostic
3348 message, even when @option{-Wno-return-type} is specified. The only
3349 exceptions are @samp{main} and functions defined in system headers.
3351 This warning is enabled by @option{-Wall}.
3356 Warn whenever a @code{switch} statement has an index of enumerated type
3357 and lacks a @code{case} for one or more of the named codes of that
3358 enumeration. (The presence of a @code{default} label prevents this
3359 warning.) @code{case} labels outside the enumeration range also
3360 provoke warnings when this option is used (even if there is a
3361 @code{default} label).
3362 This warning is enabled by @option{-Wall}.
3364 @item -Wswitch-default
3365 @opindex Wswitch-default
3366 @opindex Wno-switch-default
3367 Warn whenever a @code{switch} statement does not have a @code{default}
3371 @opindex Wswitch-enum
3372 @opindex Wno-switch-enum
3373 Warn whenever a @code{switch} statement has an index of enumerated type
3374 and lacks a @code{case} for one or more of the named codes of that
3375 enumeration. @code{case} labels outside the enumeration range also
3376 provoke warnings when this option is used. The only difference
3377 between @option{-Wswitch} and this option is that this option gives a
3378 warning about an omitted enumeration code even if there is a
3379 @code{default} label.
3381 @item -Wsync-nand @r{(C and C++ only)}
3383 @opindex Wno-sync-nand
3384 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3385 built-in functions are used. These functions changed semantics in GCC 4.4.
3389 @opindex Wno-trigraphs
3390 Warn if any trigraphs are encountered that might change the meaning of
3391 the program (trigraphs within comments are not warned about).
3392 This warning is enabled by @option{-Wall}.
3394 @item -Wunused-but-set-parameter
3395 @opindex Wunused-but-set-parameter
3396 @opindex Wno-unused-but-set-parameter
3397 Warn whenever a function parameter is assigned to, but otherwise unused
3398 (aside from its declaration).
3400 To suppress this warning use the @samp{unused} attribute
3401 (@pxref{Variable Attributes}).
3403 This warning is also enabled by @option{-Wunused} together with
3406 @item -Wunused-but-set-variable
3407 @opindex Wunused-but-set-variable
3408 @opindex Wno-unused-but-set-variable
3409 Warn whenever a local variable is assigned to, but otherwise unused
3410 (aside from its declaration).
3411 This warning is enabled by @option{-Wall}.
3413 To suppress this warning use the @samp{unused} attribute
3414 (@pxref{Variable Attributes}).
3416 This warning is also enabled by @option{-Wunused}, which is enabled
3419 @item -Wunused-function
3420 @opindex Wunused-function
3421 @opindex Wno-unused-function
3422 Warn whenever a static function is declared but not defined or a
3423 non-inline static function is unused.
3424 This warning is enabled by @option{-Wall}.
3426 @item -Wunused-label
3427 @opindex Wunused-label
3428 @opindex Wno-unused-label
3429 Warn whenever a label is declared but not used.
3430 This warning is enabled by @option{-Wall}.
3432 To suppress this warning use the @samp{unused} attribute
3433 (@pxref{Variable Attributes}).
3435 @item -Wunused-parameter
3436 @opindex Wunused-parameter
3437 @opindex Wno-unused-parameter
3438 Warn whenever a function parameter is unused aside from its declaration.
3440 To suppress this warning use the @samp{unused} attribute
3441 (@pxref{Variable Attributes}).
3443 @item -Wno-unused-result
3444 @opindex Wunused-result
3445 @opindex Wno-unused-result
3446 Do not warn if a caller of a function marked with attribute
3447 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3448 its return value. The default is @option{-Wunused-result}.
3450 @item -Wunused-variable
3451 @opindex Wunused-variable
3452 @opindex Wno-unused-variable
3453 Warn whenever a local variable or non-constant static variable is unused
3454 aside from its declaration.
3455 This warning is enabled by @option{-Wall}.
3457 To suppress this warning use the @samp{unused} attribute
3458 (@pxref{Variable Attributes}).
3460 @item -Wunused-value
3461 @opindex Wunused-value
3462 @opindex Wno-unused-value
3463 Warn whenever a statement computes a result that is explicitly not
3464 used. To suppress this warning cast the unused expression to
3465 @samp{void}. This includes an expression-statement or the left-hand
3466 side of a comma expression that contains no side effects. For example,
3467 an expression such as @samp{x[i,j]} will cause a warning, while
3468 @samp{x[(void)i,j]} will not.
3470 This warning is enabled by @option{-Wall}.
3475 All the above @option{-Wunused} options combined.
3477 In order to get a warning about an unused function parameter, you must
3478 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3479 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3481 @item -Wuninitialized
3482 @opindex Wuninitialized
3483 @opindex Wno-uninitialized
3484 Warn if an automatic variable is used without first being initialized
3485 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3486 warn if a non-static reference or non-static @samp{const} member
3487 appears in a class without constructors.
3489 If you want to warn about code which uses the uninitialized value of the
3490 variable in its own initializer, use the @option{-Winit-self} option.
3492 These warnings occur for individual uninitialized or clobbered
3493 elements of structure, union or array variables as well as for
3494 variables which are uninitialized or clobbered as a whole. They do
3495 not occur for variables or elements declared @code{volatile}. Because
3496 these warnings depend on optimization, the exact variables or elements
3497 for which there are warnings will depend on the precise optimization
3498 options and version of GCC used.
3500 Note that there may be no warning about a variable that is used only
3501 to compute a value that itself is never used, because such
3502 computations may be deleted by data flow analysis before the warnings
3505 These warnings are made optional because GCC is not smart
3506 enough to see all the reasons why the code might be correct
3507 despite appearing to have an error. Here is one example of how
3528 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3529 always initialized, but GCC doesn't know this. Here is
3530 another common case:
3535 if (change_y) save_y = y, y = new_y;
3537 if (change_y) y = save_y;
3542 This has no bug because @code{save_y} is used only if it is set.
3544 @cindex @code{longjmp} warnings
3545 This option also warns when a non-volatile automatic variable might be
3546 changed by a call to @code{longjmp}. These warnings as well are possible
3547 only in optimizing compilation.
3549 The compiler sees only the calls to @code{setjmp}. It cannot know
3550 where @code{longjmp} will be called; in fact, a signal handler could
3551 call it at any point in the code. As a result, you may get a warning
3552 even when there is in fact no problem because @code{longjmp} cannot
3553 in fact be called at the place which would cause a problem.
3555 Some spurious warnings can be avoided if you declare all the functions
3556 you use that never return as @code{noreturn}. @xref{Function
3559 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3561 @item -Wunknown-pragmas
3562 @opindex Wunknown-pragmas
3563 @opindex Wno-unknown-pragmas
3564 @cindex warning for unknown pragmas
3565 @cindex unknown pragmas, warning
3566 @cindex pragmas, warning of unknown
3567 Warn when a #pragma directive is encountered which is not understood by
3568 GCC@. If this command line option is used, warnings will even be issued
3569 for unknown pragmas in system header files. This is not the case if
3570 the warnings were only enabled by the @option{-Wall} command line option.
3573 @opindex Wno-pragmas
3575 Do not warn about misuses of pragmas, such as incorrect parameters,
3576 invalid syntax, or conflicts between pragmas. See also
3577 @samp{-Wunknown-pragmas}.
3579 @item -Wstrict-aliasing
3580 @opindex Wstrict-aliasing
3581 @opindex Wno-strict-aliasing
3582 This option is only active when @option{-fstrict-aliasing} is active.
3583 It warns about code which might break the strict aliasing rules that the
3584 compiler is using for optimization. The warning does not catch all
3585 cases, but does attempt to catch the more common pitfalls. It is
3586 included in @option{-Wall}.
3587 It is equivalent to @option{-Wstrict-aliasing=3}
3589 @item -Wstrict-aliasing=n
3590 @opindex Wstrict-aliasing=n
3591 @opindex Wno-strict-aliasing=n
3592 This option is only active when @option{-fstrict-aliasing} is active.
3593 It warns about code which might break the strict aliasing rules that the
3594 compiler is using for optimization.
3595 Higher levels correspond to higher accuracy (fewer false positives).
3596 Higher levels also correspond to more effort, similar to the way -O works.
3597 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3600 Level 1: Most aggressive, quick, least accurate.
3601 Possibly useful when higher levels
3602 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3603 false negatives. However, it has many false positives.
3604 Warns for all pointer conversions between possibly incompatible types,
3605 even if never dereferenced. Runs in the frontend only.
3607 Level 2: Aggressive, quick, not too precise.
3608 May still have many false positives (not as many as level 1 though),
3609 and few false negatives (but possibly more than level 1).
3610 Unlike level 1, it only warns when an address is taken. Warns about
3611 incomplete types. Runs in the frontend only.
3613 Level 3 (default for @option{-Wstrict-aliasing}):
3614 Should have very few false positives and few false
3615 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3616 Takes care of the common pun+dereference pattern in the frontend:
3617 @code{*(int*)&some_float}.
3618 If optimization is enabled, it also runs in the backend, where it deals
3619 with multiple statement cases using flow-sensitive points-to information.
3620 Only warns when the converted pointer is dereferenced.
3621 Does not warn about incomplete types.
3623 @item -Wstrict-overflow
3624 @itemx -Wstrict-overflow=@var{n}
3625 @opindex Wstrict-overflow
3626 @opindex Wno-strict-overflow
3627 This option is only active when @option{-fstrict-overflow} is active.
3628 It warns about cases where the compiler optimizes based on the
3629 assumption that signed overflow does not occur. Note that it does not
3630 warn about all cases where the code might overflow: it only warns
3631 about cases where the compiler implements some optimization. Thus
3632 this warning depends on the optimization level.
3634 An optimization which assumes that signed overflow does not occur is
3635 perfectly safe if the values of the variables involved are such that
3636 overflow never does, in fact, occur. Therefore this warning can
3637 easily give a false positive: a warning about code which is not
3638 actually a problem. To help focus on important issues, several
3639 warning levels are defined. No warnings are issued for the use of
3640 undefined signed overflow when estimating how many iterations a loop
3641 will require, in particular when determining whether a loop will be
3645 @item -Wstrict-overflow=1
3646 Warn about cases which are both questionable and easy to avoid. For
3647 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3648 compiler will simplify this to @code{1}. This level of
3649 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3650 are not, and must be explicitly requested.
3652 @item -Wstrict-overflow=2
3653 Also warn about other cases where a comparison is simplified to a
3654 constant. For example: @code{abs (x) >= 0}. This can only be
3655 simplified when @option{-fstrict-overflow} is in effect, because
3656 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3657 zero. @option{-Wstrict-overflow} (with no level) is the same as
3658 @option{-Wstrict-overflow=2}.
3660 @item -Wstrict-overflow=3
3661 Also warn about other cases where a comparison is simplified. For
3662 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3664 @item -Wstrict-overflow=4
3665 Also warn about other simplifications not covered by the above cases.
3666 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3668 @item -Wstrict-overflow=5
3669 Also warn about cases where the compiler reduces the magnitude of a
3670 constant involved in a comparison. For example: @code{x + 2 > y} will
3671 be simplified to @code{x + 1 >= y}. This is reported only at the
3672 highest warning level because this simplification applies to many
3673 comparisons, so this warning level will give a very large number of
3677 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3678 @opindex Wsuggest-attribute=
3679 @opindex Wno-suggest-attribute=
3680 Warn for cases where adding an attribute may be beneficial. The
3681 attributes currently supported are listed below.
3684 @item -Wsuggest-attribute=pure
3685 @itemx -Wsuggest-attribute=const
3686 @itemx -Wsuggest-attribute=noreturn
3687 @opindex Wsuggest-attribute=pure
3688 @opindex Wno-suggest-attribute=pure
3689 @opindex Wsuggest-attribute=const
3690 @opindex Wno-suggest-attribute=const
3691 @opindex Wsuggest-attribute=noreturn
3692 @opindex Wno-suggest-attribute=noreturn
3694 Warn about functions which might be candidates for attributes
3695 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3696 functions visible in other compilation units or (in the case of @code{pure} and
3697 @code{const}) if it cannot prove that the function returns normally. A function
3698 returns normally if it doesn't contain an infinite loop nor returns abnormally
3699 by throwing, calling @code{abort()} or trapping. This analysis requires option
3700 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3701 higher. Higher optimization levels improve the accuracy of the analysis.
3704 @item -Warray-bounds
3705 @opindex Wno-array-bounds
3706 @opindex Warray-bounds
3707 This option is only active when @option{-ftree-vrp} is active
3708 (default for @option{-O2} and above). It warns about subscripts to arrays
3709 that are always out of bounds. This warning is enabled by @option{-Wall}.
3711 @item -Wno-div-by-zero
3712 @opindex Wno-div-by-zero
3713 @opindex Wdiv-by-zero
3714 Do not warn about compile-time integer division by zero. Floating point
3715 division by zero is not warned about, as it can be a legitimate way of
3716 obtaining infinities and NaNs.
3718 @item -Wsystem-headers
3719 @opindex Wsystem-headers
3720 @opindex Wno-system-headers
3721 @cindex warnings from system headers
3722 @cindex system headers, warnings from
3723 Print warning messages for constructs found in system header files.
3724 Warnings from system headers are normally suppressed, on the assumption
3725 that they usually do not indicate real problems and would only make the
3726 compiler output harder to read. Using this command line option tells
3727 GCC to emit warnings from system headers as if they occurred in user
3728 code. However, note that using @option{-Wall} in conjunction with this
3729 option will @emph{not} warn about unknown pragmas in system
3730 headers---for that, @option{-Wunknown-pragmas} must also be used.
3733 @opindex Wtrampolines
3734 @opindex Wno-trampolines
3735 Warn about trampolines generated for pointers to nested functions.
3737 A trampoline is a small piece of data or code that is created at run
3738 time on the stack when the address of a nested function is taken, and
3739 is used to call the nested function indirectly. For some targets, it
3740 is made up of data only and thus requires no special treatment. But,
3741 for most targets, it is made up of code and thus requires the stack
3742 to be made executable in order for the program to work properly.
3745 @opindex Wfloat-equal
3746 @opindex Wno-float-equal
3747 Warn if floating point values are used in equality comparisons.
3749 The idea behind this is that sometimes it is convenient (for the
3750 programmer) to consider floating-point values as approximations to
3751 infinitely precise real numbers. If you are doing this, then you need
3752 to compute (by analyzing the code, or in some other way) the maximum or
3753 likely maximum error that the computation introduces, and allow for it
3754 when performing comparisons (and when producing output, but that's a
3755 different problem). In particular, instead of testing for equality, you
3756 would check to see whether the two values have ranges that overlap; and
3757 this is done with the relational operators, so equality comparisons are
3760 @item -Wtraditional @r{(C and Objective-C only)}
3761 @opindex Wtraditional
3762 @opindex Wno-traditional
3763 Warn about certain constructs that behave differently in traditional and
3764 ISO C@. Also warn about ISO C constructs that have no traditional C
3765 equivalent, and/or problematic constructs which should be avoided.
3769 Macro parameters that appear within string literals in the macro body.
3770 In traditional C macro replacement takes place within string literals,
3771 but does not in ISO C@.
3774 In traditional C, some preprocessor directives did not exist.
3775 Traditional preprocessors would only consider a line to be a directive
3776 if the @samp{#} appeared in column 1 on the line. Therefore
3777 @option{-Wtraditional} warns about directives that traditional C
3778 understands but would ignore because the @samp{#} does not appear as the
3779 first character on the line. It also suggests you hide directives like
3780 @samp{#pragma} not understood by traditional C by indenting them. Some
3781 traditional implementations would not recognize @samp{#elif}, so it
3782 suggests avoiding it altogether.
3785 A function-like macro that appears without arguments.
3788 The unary plus operator.
3791 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3792 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3793 constants.) Note, these suffixes appear in macros defined in the system
3794 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3795 Use of these macros in user code might normally lead to spurious
3796 warnings, however GCC's integrated preprocessor has enough context to
3797 avoid warning in these cases.
3800 A function declared external in one block and then used after the end of
3804 A @code{switch} statement has an operand of type @code{long}.
3807 A non-@code{static} function declaration follows a @code{static} one.
3808 This construct is not accepted by some traditional C compilers.
3811 The ISO type of an integer constant has a different width or
3812 signedness from its traditional type. This warning is only issued if
3813 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3814 typically represent bit patterns, are not warned about.
3817 Usage of ISO string concatenation is detected.
3820 Initialization of automatic aggregates.
3823 Identifier conflicts with labels. Traditional C lacks a separate
3824 namespace for labels.
3827 Initialization of unions. If the initializer is zero, the warning is
3828 omitted. This is done under the assumption that the zero initializer in
3829 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3830 initializer warnings and relies on default initialization to zero in the
3834 Conversions by prototypes between fixed/floating point values and vice
3835 versa. The absence of these prototypes when compiling with traditional
3836 C would cause serious problems. This is a subset of the possible
3837 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3840 Use of ISO C style function definitions. This warning intentionally is
3841 @emph{not} issued for prototype declarations or variadic functions
3842 because these ISO C features will appear in your code when using
3843 libiberty's traditional C compatibility macros, @code{PARAMS} and
3844 @code{VPARAMS}. This warning is also bypassed for nested functions
3845 because that feature is already a GCC extension and thus not relevant to
3846 traditional C compatibility.
3849 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3850 @opindex Wtraditional-conversion
3851 @opindex Wno-traditional-conversion
3852 Warn if a prototype causes a type conversion that is different from what
3853 would happen to the same argument in the absence of a prototype. This
3854 includes conversions of fixed point to floating and vice versa, and
3855 conversions changing the width or signedness of a fixed point argument
3856 except when the same as the default promotion.
3858 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3859 @opindex Wdeclaration-after-statement
3860 @opindex Wno-declaration-after-statement
3861 Warn when a declaration is found after a statement in a block. This
3862 construct, known from C++, was introduced with ISO C99 and is by default
3863 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3864 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3869 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3871 @item -Wno-endif-labels
3872 @opindex Wno-endif-labels
3873 @opindex Wendif-labels
3874 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3879 Warn whenever a local variable or type declaration shadows another variable,
3880 parameter, type, or class member (in C++), or whenever a built-in function
3881 is shadowed. Note that in C++, the compiler will not warn if a local variable
3882 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3884 @item -Wlarger-than=@var{len}
3885 @opindex Wlarger-than=@var{len}
3886 @opindex Wlarger-than-@var{len}
3887 Warn whenever an object of larger than @var{len} bytes is defined.
3889 @item -Wframe-larger-than=@var{len}
3890 @opindex Wframe-larger-than
3891 Warn if the size of a function frame is larger than @var{len} bytes.
3892 The computation done to determine the stack frame size is approximate
3893 and not conservative.
3894 The actual requirements may be somewhat greater than @var{len}
3895 even if you do not get a warning. In addition, any space allocated
3896 via @code{alloca}, variable-length arrays, or related constructs
3897 is not included by the compiler when determining
3898 whether or not to issue a warning.
3900 @item -Wunsafe-loop-optimizations
3901 @opindex Wunsafe-loop-optimizations
3902 @opindex Wno-unsafe-loop-optimizations
3903 Warn if the loop cannot be optimized because the compiler could not
3904 assume anything on the bounds of the loop indices. With
3905 @option{-funsafe-loop-optimizations} warn if the compiler made
3908 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3909 @opindex Wno-pedantic-ms-format
3910 @opindex Wpedantic-ms-format
3911 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3912 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3913 depending on the MS runtime, when you are using the options @option{-Wformat}
3914 and @option{-pedantic} without gnu-extensions.
3916 @item -Wpointer-arith
3917 @opindex Wpointer-arith
3918 @opindex Wno-pointer-arith
3919 Warn about anything that depends on the ``size of'' a function type or
3920 of @code{void}. GNU C assigns these types a size of 1, for
3921 convenience in calculations with @code{void *} pointers and pointers
3922 to functions. In C++, warn also when an arithmetic operation involves
3923 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3926 @opindex Wtype-limits
3927 @opindex Wno-type-limits
3928 Warn if a comparison is always true or always false due to the limited
3929 range of the data type, but do not warn for constant expressions. For
3930 example, warn if an unsigned variable is compared against zero with
3931 @samp{<} or @samp{>=}. This warning is also enabled by
3934 @item -Wbad-function-cast @r{(C and Objective-C only)}
3935 @opindex Wbad-function-cast
3936 @opindex Wno-bad-function-cast
3937 Warn whenever a function call is cast to a non-matching type.
3938 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3940 @item -Wc++-compat @r{(C and Objective-C only)}
3941 Warn about ISO C constructs that are outside of the common subset of
3942 ISO C and ISO C++, e.g.@: request for implicit conversion from
3943 @code{void *} to a pointer to non-@code{void} type.
3945 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3946 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3947 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3948 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3952 @opindex Wno-cast-qual
3953 Warn whenever a pointer is cast so as to remove a type qualifier from
3954 the target type. For example, warn if a @code{const char *} is cast
3955 to an ordinary @code{char *}.
3957 Also warn when making a cast which introduces a type qualifier in an
3958 unsafe way. For example, casting @code{char **} to @code{const char **}
3959 is unsafe, as in this example:
3962 /* p is char ** value. */
3963 const char **q = (const char **) p;
3964 /* Assignment of readonly string to const char * is OK. */
3966 /* Now char** pointer points to read-only memory. */
3971 @opindex Wcast-align
3972 @opindex Wno-cast-align
3973 Warn whenever a pointer is cast such that the required alignment of the
3974 target is increased. For example, warn if a @code{char *} is cast to
3975 an @code{int *} on machines where integers can only be accessed at
3976 two- or four-byte boundaries.
3978 @item -Wwrite-strings
3979 @opindex Wwrite-strings
3980 @opindex Wno-write-strings
3981 When compiling C, give string constants the type @code{const
3982 char[@var{length}]} so that copying the address of one into a
3983 non-@code{const} @code{char *} pointer will get a warning. These
3984 warnings will help you find at compile time code that can try to write
3985 into a string constant, but only if you have been very careful about
3986 using @code{const} in declarations and prototypes. Otherwise, it will
3987 just be a nuisance. This is why we did not make @option{-Wall} request
3990 When compiling C++, warn about the deprecated conversion from string
3991 literals to @code{char *}. This warning is enabled by default for C++
3996 @opindex Wno-clobbered
3997 Warn for variables that might be changed by @samp{longjmp} or
3998 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4001 @opindex Wconversion
4002 @opindex Wno-conversion
4003 Warn for implicit conversions that may alter a value. This includes
4004 conversions between real and integer, like @code{abs (x)} when
4005 @code{x} is @code{double}; conversions between signed and unsigned,
4006 like @code{unsigned ui = -1}; and conversions to smaller types, like
4007 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4008 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4009 changed by the conversion like in @code{abs (2.0)}. Warnings about
4010 conversions between signed and unsigned integers can be disabled by
4011 using @option{-Wno-sign-conversion}.
4013 For C++, also warn for confusing overload resolution for user-defined
4014 conversions; and conversions that will never use a type conversion
4015 operator: conversions to @code{void}, the same type, a base class or a
4016 reference to them. Warnings about conversions between signed and
4017 unsigned integers are disabled by default in C++ unless
4018 @option{-Wsign-conversion} is explicitly enabled.
4020 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4021 @opindex Wconversion-null
4022 @opindex Wno-conversion-null
4023 Do not warn for conversions between @code{NULL} and non-pointer
4024 types. @option{-Wconversion-null} is enabled by default.
4027 @opindex Wempty-body
4028 @opindex Wno-empty-body
4029 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4030 while} statement. This warning is also enabled by @option{-Wextra}.
4032 @item -Wenum-compare
4033 @opindex Wenum-compare
4034 @opindex Wno-enum-compare
4035 Warn about a comparison between values of different enum types. In C++
4036 this warning is enabled by default. In C this warning is enabled by
4039 @item -Wjump-misses-init @r{(C, Objective-C only)}
4040 @opindex Wjump-misses-init
4041 @opindex Wno-jump-misses-init
4042 Warn if a @code{goto} statement or a @code{switch} statement jumps
4043 forward across the initialization of a variable, or jumps backward to a
4044 label after the variable has been initialized. This only warns about
4045 variables which are initialized when they are declared. This warning is
4046 only supported for C and Objective C; in C++ this sort of branch is an
4049 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4050 can be disabled with the @option{-Wno-jump-misses-init} option.
4052 @item -Wsign-compare
4053 @opindex Wsign-compare
4054 @opindex Wno-sign-compare
4055 @cindex warning for comparison of signed and unsigned values
4056 @cindex comparison of signed and unsigned values, warning
4057 @cindex signed and unsigned values, comparison warning
4058 Warn when a comparison between signed and unsigned values could produce
4059 an incorrect result when the signed value is converted to unsigned.
4060 This warning is also enabled by @option{-Wextra}; to get the other warnings
4061 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4063 @item -Wsign-conversion
4064 @opindex Wsign-conversion
4065 @opindex Wno-sign-conversion
4066 Warn for implicit conversions that may change the sign of an integer
4067 value, like assigning a signed integer expression to an unsigned
4068 integer variable. An explicit cast silences the warning. In C, this
4069 option is enabled also by @option{-Wconversion}.
4073 @opindex Wno-address
4074 Warn about suspicious uses of memory addresses. These include using
4075 the address of a function in a conditional expression, such as
4076 @code{void func(void); if (func)}, and comparisons against the memory
4077 address of a string literal, such as @code{if (x == "abc")}. Such
4078 uses typically indicate a programmer error: the address of a function
4079 always evaluates to true, so their use in a conditional usually
4080 indicate that the programmer forgot the parentheses in a function
4081 call; and comparisons against string literals result in unspecified
4082 behavior and are not portable in C, so they usually indicate that the
4083 programmer intended to use @code{strcmp}. This warning is enabled by
4087 @opindex Wlogical-op
4088 @opindex Wno-logical-op
4089 Warn about suspicious uses of logical operators in expressions.
4090 This includes using logical operators in contexts where a
4091 bit-wise operator is likely to be expected.
4093 @item -Waggregate-return
4094 @opindex Waggregate-return
4095 @opindex Wno-aggregate-return
4096 Warn if any functions that return structures or unions are defined or
4097 called. (In languages where you can return an array, this also elicits
4100 @item -Wno-attributes
4101 @opindex Wno-attributes
4102 @opindex Wattributes
4103 Do not warn if an unexpected @code{__attribute__} is used, such as
4104 unrecognized attributes, function attributes applied to variables,
4105 etc. This will not stop errors for incorrect use of supported
4108 @item -Wno-builtin-macro-redefined
4109 @opindex Wno-builtin-macro-redefined
4110 @opindex Wbuiltin-macro-redefined
4111 Do not warn if certain built-in macros are redefined. This suppresses
4112 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4113 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4115 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4116 @opindex Wstrict-prototypes
4117 @opindex Wno-strict-prototypes
4118 Warn if a function is declared or defined without specifying the
4119 argument types. (An old-style function definition is permitted without
4120 a warning if preceded by a declaration which specifies the argument
4123 @item -Wold-style-declaration @r{(C and Objective-C only)}
4124 @opindex Wold-style-declaration
4125 @opindex Wno-old-style-declaration
4126 Warn for obsolescent usages, according to the C Standard, in a
4127 declaration. For example, warn if storage-class specifiers like
4128 @code{static} are not the first things in a declaration. This warning
4129 is also enabled by @option{-Wextra}.
4131 @item -Wold-style-definition @r{(C and Objective-C only)}
4132 @opindex Wold-style-definition
4133 @opindex Wno-old-style-definition
4134 Warn if an old-style function definition is used. A warning is given
4135 even if there is a previous prototype.
4137 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4138 @opindex Wmissing-parameter-type
4139 @opindex Wno-missing-parameter-type
4140 A function parameter is declared without a type specifier in K&R-style
4147 This warning is also enabled by @option{-Wextra}.
4149 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4150 @opindex Wmissing-prototypes
4151 @opindex Wno-missing-prototypes
4152 Warn if a global function is defined without a previous prototype
4153 declaration. This warning is issued even if the definition itself
4154 provides a prototype. The aim is to detect global functions that fail
4155 to be declared in header files.
4157 @item -Wmissing-declarations
4158 @opindex Wmissing-declarations
4159 @opindex Wno-missing-declarations
4160 Warn if a global function is defined without a previous declaration.
4161 Do so even if the definition itself provides a prototype.
4162 Use this option to detect global functions that are not declared in
4163 header files. In C++, no warnings are issued for function templates,
4164 or for inline functions, or for functions in anonymous namespaces.
4166 @item -Wmissing-field-initializers
4167 @opindex Wmissing-field-initializers
4168 @opindex Wno-missing-field-initializers
4172 Warn if a structure's initializer has some fields missing. For
4173 example, the following code would cause such a warning, because
4174 @code{x.h} is implicitly zero:
4177 struct s @{ int f, g, h; @};
4178 struct s x = @{ 3, 4 @};
4181 This option does not warn about designated initializers, so the following
4182 modification would not trigger a warning:
4185 struct s @{ int f, g, h; @};
4186 struct s x = @{ .f = 3, .g = 4 @};
4189 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4190 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4192 @item -Wmissing-format-attribute
4193 @opindex Wmissing-format-attribute
4194 @opindex Wno-missing-format-attribute
4197 Warn about function pointers which might be candidates for @code{format}
4198 attributes. Note these are only possible candidates, not absolute ones.
4199 GCC will guess that function pointers with @code{format} attributes that
4200 are used in assignment, initialization, parameter passing or return
4201 statements should have a corresponding @code{format} attribute in the
4202 resulting type. I.e.@: the left-hand side of the assignment or
4203 initialization, the type of the parameter variable, or the return type
4204 of the containing function respectively should also have a @code{format}
4205 attribute to avoid the warning.
4207 GCC will also warn about function definitions which might be
4208 candidates for @code{format} attributes. Again, these are only
4209 possible candidates. GCC will guess that @code{format} attributes
4210 might be appropriate for any function that calls a function like
4211 @code{vprintf} or @code{vscanf}, but this might not always be the
4212 case, and some functions for which @code{format} attributes are
4213 appropriate may not be detected.
4215 @item -Wno-multichar
4216 @opindex Wno-multichar
4218 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4219 Usually they indicate a typo in the user's code, as they have
4220 implementation-defined values, and should not be used in portable code.
4222 @item -Wnormalized=<none|id|nfc|nfkc>
4223 @opindex Wnormalized=
4226 @cindex character set, input normalization
4227 In ISO C and ISO C++, two identifiers are different if they are
4228 different sequences of characters. However, sometimes when characters
4229 outside the basic ASCII character set are used, you can have two
4230 different character sequences that look the same. To avoid confusion,
4231 the ISO 10646 standard sets out some @dfn{normalization rules} which
4232 when applied ensure that two sequences that look the same are turned into
4233 the same sequence. GCC can warn you if you are using identifiers which
4234 have not been normalized; this option controls that warning.
4236 There are four levels of warning that GCC supports. The default is
4237 @option{-Wnormalized=nfc}, which warns about any identifier which is
4238 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4239 recommended form for most uses.
4241 Unfortunately, there are some characters which ISO C and ISO C++ allow
4242 in identifiers that when turned into NFC aren't allowable as
4243 identifiers. That is, there's no way to use these symbols in portable
4244 ISO C or C++ and have all your identifiers in NFC@.
4245 @option{-Wnormalized=id} suppresses the warning for these characters.
4246 It is hoped that future versions of the standards involved will correct
4247 this, which is why this option is not the default.
4249 You can switch the warning off for all characters by writing
4250 @option{-Wnormalized=none}. You would only want to do this if you
4251 were using some other normalization scheme (like ``D''), because
4252 otherwise you can easily create bugs that are literally impossible to see.
4254 Some characters in ISO 10646 have distinct meanings but look identical
4255 in some fonts or display methodologies, especially once formatting has
4256 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4257 LETTER N'', will display just like a regular @code{n} which has been
4258 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4259 normalization scheme to convert all these into a standard form as
4260 well, and GCC will warn if your code is not in NFKC if you use
4261 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4262 about every identifier that contains the letter O because it might be
4263 confused with the digit 0, and so is not the default, but may be
4264 useful as a local coding convention if the programming environment is
4265 unable to be fixed to display these characters distinctly.
4267 @item -Wno-deprecated
4268 @opindex Wno-deprecated
4269 @opindex Wdeprecated
4270 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4272 @item -Wno-deprecated-declarations
4273 @opindex Wno-deprecated-declarations
4274 @opindex Wdeprecated-declarations
4275 Do not warn about uses of functions (@pxref{Function Attributes}),
4276 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4277 Attributes}) marked as deprecated by using the @code{deprecated}
4281 @opindex Wno-overflow
4283 Do not warn about compile-time overflow in constant expressions.
4285 @item -Woverride-init @r{(C and Objective-C only)}
4286 @opindex Woverride-init
4287 @opindex Wno-override-init
4291 Warn if an initialized field without side effects is overridden when
4292 using designated initializers (@pxref{Designated Inits, , Designated
4295 This warning is included in @option{-Wextra}. To get other
4296 @option{-Wextra} warnings without this one, use @samp{-Wextra
4297 -Wno-override-init}.
4302 Warn if a structure is given the packed attribute, but the packed
4303 attribute has no effect on the layout or size of the structure.
4304 Such structures may be mis-aligned for little benefit. For
4305 instance, in this code, the variable @code{f.x} in @code{struct bar}
4306 will be misaligned even though @code{struct bar} does not itself
4307 have the packed attribute:
4314 @} __attribute__((packed));
4322 @item -Wpacked-bitfield-compat
4323 @opindex Wpacked-bitfield-compat
4324 @opindex Wno-packed-bitfield-compat
4325 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4326 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4327 the change can lead to differences in the structure layout. GCC
4328 informs you when the offset of such a field has changed in GCC 4.4.
4329 For example there is no longer a 4-bit padding between field @code{a}
4330 and @code{b} in this structure:
4337 @} __attribute__ ((packed));
4340 This warning is enabled by default. Use
4341 @option{-Wno-packed-bitfield-compat} to disable this warning.
4346 Warn if padding is included in a structure, either to align an element
4347 of the structure or to align the whole structure. Sometimes when this
4348 happens it is possible to rearrange the fields of the structure to
4349 reduce the padding and so make the structure smaller.
4351 @item -Wredundant-decls
4352 @opindex Wredundant-decls
4353 @opindex Wno-redundant-decls
4354 Warn if anything is declared more than once in the same scope, even in
4355 cases where multiple declaration is valid and changes nothing.
4357 @item -Wnested-externs @r{(C and Objective-C only)}
4358 @opindex Wnested-externs
4359 @opindex Wno-nested-externs
4360 Warn if an @code{extern} declaration is encountered within a function.
4365 Warn if a function can not be inlined and it was declared as inline.
4366 Even with this option, the compiler will not warn about failures to
4367 inline functions declared in system headers.
4369 The compiler uses a variety of heuristics to determine whether or not
4370 to inline a function. For example, the compiler takes into account
4371 the size of the function being inlined and the amount of inlining
4372 that has already been done in the current function. Therefore,
4373 seemingly insignificant changes in the source program can cause the
4374 warnings produced by @option{-Winline} to appear or disappear.
4376 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4377 @opindex Wno-invalid-offsetof
4378 @opindex Winvalid-offsetof
4379 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4380 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4381 to a non-POD type is undefined. In existing C++ implementations,
4382 however, @samp{offsetof} typically gives meaningful results even when
4383 applied to certain kinds of non-POD types. (Such as a simple
4384 @samp{struct} that fails to be a POD type only by virtue of having a
4385 constructor.) This flag is for users who are aware that they are
4386 writing nonportable code and who have deliberately chosen to ignore the
4389 The restrictions on @samp{offsetof} may be relaxed in a future version
4390 of the C++ standard.
4392 @item -Wno-int-to-pointer-cast
4393 @opindex Wno-int-to-pointer-cast
4394 @opindex Wint-to-pointer-cast
4395 Suppress warnings from casts to pointer type of an integer of a
4396 different size. In C++, casting to a pointer type of smaller size is
4397 an error. @option{Wint-to-pointer-cast} is enabled by default.
4400 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4401 @opindex Wno-pointer-to-int-cast
4402 @opindex Wpointer-to-int-cast
4403 Suppress warnings from casts from a pointer to an integer type of a
4407 @opindex Winvalid-pch
4408 @opindex Wno-invalid-pch
4409 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4410 the search path but can't be used.
4414 @opindex Wno-long-long
4415 Warn if @samp{long long} type is used. This is enabled by either
4416 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4417 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4419 @item -Wvariadic-macros
4420 @opindex Wvariadic-macros
4421 @opindex Wno-variadic-macros
4422 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4423 alternate syntax when in pedantic ISO C99 mode. This is default.
4424 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4429 Warn if variable length array is used in the code.
4430 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4431 the variable length array.
4433 @item -Wvolatile-register-var
4434 @opindex Wvolatile-register-var
4435 @opindex Wno-volatile-register-var
4436 Warn if a register variable is declared volatile. The volatile
4437 modifier does not inhibit all optimizations that may eliminate reads
4438 and/or writes to register variables. This warning is enabled by
4441 @item -Wdisabled-optimization
4442 @opindex Wdisabled-optimization
4443 @opindex Wno-disabled-optimization
4444 Warn if a requested optimization pass is disabled. This warning does
4445 not generally indicate that there is anything wrong with your code; it
4446 merely indicates that GCC's optimizers were unable to handle the code
4447 effectively. Often, the problem is that your code is too big or too
4448 complex; GCC will refuse to optimize programs when the optimization
4449 itself is likely to take inordinate amounts of time.
4451 @item -Wpointer-sign @r{(C and Objective-C only)}
4452 @opindex Wpointer-sign
4453 @opindex Wno-pointer-sign
4454 Warn for pointer argument passing or assignment with different signedness.
4455 This option is only supported for C and Objective-C@. It is implied by
4456 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4457 @option{-Wno-pointer-sign}.
4459 @item -Wstack-protector
4460 @opindex Wstack-protector
4461 @opindex Wno-stack-protector
4462 This option is only active when @option{-fstack-protector} is active. It
4463 warns about functions that will not be protected against stack smashing.
4466 @opindex Wno-mudflap
4467 Suppress warnings about constructs that cannot be instrumented by
4470 @item -Woverlength-strings
4471 @opindex Woverlength-strings
4472 @opindex Wno-overlength-strings
4473 Warn about string constants which are longer than the ``minimum
4474 maximum'' length specified in the C standard. Modern compilers
4475 generally allow string constants which are much longer than the
4476 standard's minimum limit, but very portable programs should avoid
4477 using longer strings.
4479 The limit applies @emph{after} string constant concatenation, and does
4480 not count the trailing NUL@. In C90, the limit was 509 characters; in
4481 C99, it was raised to 4095. C++98 does not specify a normative
4482 minimum maximum, so we do not diagnose overlength strings in C++@.
4484 This option is implied by @option{-pedantic}, and can be disabled with
4485 @option{-Wno-overlength-strings}.
4487 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4488 @opindex Wunsuffixed-float-constants
4490 GCC will issue a warning for any floating constant that does not have
4491 a suffix. When used together with @option{-Wsystem-headers} it will
4492 warn about such constants in system header files. This can be useful
4493 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4494 from the decimal floating-point extension to C99.
4497 @node Debugging Options
4498 @section Options for Debugging Your Program or GCC
4499 @cindex options, debugging
4500 @cindex debugging information options
4502 GCC has various special options that are used for debugging
4503 either your program or GCC:
4508 Produce debugging information in the operating system's native format
4509 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4512 On most systems that use stabs format, @option{-g} enables use of extra
4513 debugging information that only GDB can use; this extra information
4514 makes debugging work better in GDB but will probably make other debuggers
4516 refuse to read the program. If you want to control for certain whether
4517 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4518 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4520 GCC allows you to use @option{-g} with
4521 @option{-O}. The shortcuts taken by optimized code may occasionally
4522 produce surprising results: some variables you declared may not exist
4523 at all; flow of control may briefly move where you did not expect it;
4524 some statements may not be executed because they compute constant
4525 results or their values were already at hand; some statements may
4526 execute in different places because they were moved out of loops.
4528 Nevertheless it proves possible to debug optimized output. This makes
4529 it reasonable to use the optimizer for programs that might have bugs.
4531 The following options are useful when GCC is generated with the
4532 capability for more than one debugging format.
4536 Produce debugging information for use by GDB@. This means to use the
4537 most expressive format available (DWARF 2, stabs, or the native format
4538 if neither of those are supported), including GDB extensions if at all
4543 Produce debugging information in stabs format (if that is supported),
4544 without GDB extensions. This is the format used by DBX on most BSD
4545 systems. On MIPS, Alpha and System V Release 4 systems this option
4546 produces stabs debugging output which is not understood by DBX or SDB@.
4547 On System V Release 4 systems this option requires the GNU assembler.
4549 @item -feliminate-unused-debug-symbols
4550 @opindex feliminate-unused-debug-symbols
4551 Produce debugging information in stabs format (if that is supported),
4552 for only symbols that are actually used.
4554 @item -femit-class-debug-always
4555 Instead of emitting debugging information for a C++ class in only one
4556 object file, emit it in all object files using the class. This option
4557 should be used only with debuggers that are unable to handle the way GCC
4558 normally emits debugging information for classes because using this
4559 option will increase the size of debugging information by as much as a
4564 Produce debugging information in stabs format (if that is supported),
4565 using GNU extensions understood only by the GNU debugger (GDB)@. The
4566 use of these extensions is likely to make other debuggers crash or
4567 refuse to read the program.
4571 Produce debugging information in COFF format (if that is supported).
4572 This is the format used by SDB on most System V systems prior to
4577 Produce debugging information in XCOFF format (if that is supported).
4578 This is the format used by the DBX debugger on IBM RS/6000 systems.
4582 Produce debugging information in XCOFF format (if that is supported),
4583 using GNU extensions understood only by the GNU debugger (GDB)@. The
4584 use of these extensions is likely to make other debuggers crash or
4585 refuse to read the program, and may cause assemblers other than the GNU
4586 assembler (GAS) to fail with an error.
4588 @item -gdwarf-@var{version}
4589 @opindex gdwarf-@var{version}
4590 Produce debugging information in DWARF format (if that is
4591 supported). This is the format used by DBX on IRIX 6. The value
4592 of @var{version} may be either 2, 3 or 4; the default version is 2.
4594 Note that with DWARF version 2 some ports require, and will always
4595 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4597 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4598 for maximum benefit.
4600 @item -gstrict-dwarf
4601 @opindex gstrict-dwarf
4602 Disallow using extensions of later DWARF standard version than selected
4603 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4604 DWARF extensions from later standard versions is allowed.
4606 @item -gno-strict-dwarf
4607 @opindex gno-strict-dwarf
4608 Allow using extensions of later DWARF standard version than selected with
4609 @option{-gdwarf-@var{version}}.
4613 Produce debugging information in VMS debug format (if that is
4614 supported). This is the format used by DEBUG on VMS systems.
4617 @itemx -ggdb@var{level}
4618 @itemx -gstabs@var{level}
4619 @itemx -gcoff@var{level}
4620 @itemx -gxcoff@var{level}
4621 @itemx -gvms@var{level}
4622 Request debugging information and also use @var{level} to specify how
4623 much information. The default level is 2.
4625 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4628 Level 1 produces minimal information, enough for making backtraces in
4629 parts of the program that you don't plan to debug. This includes
4630 descriptions of functions and external variables, but no information
4631 about local variables and no line numbers.
4633 Level 3 includes extra information, such as all the macro definitions
4634 present in the program. Some debuggers support macro expansion when
4635 you use @option{-g3}.
4637 @option{-gdwarf-2} does not accept a concatenated debug level, because
4638 GCC used to support an option @option{-gdwarf} that meant to generate
4639 debug information in version 1 of the DWARF format (which is very
4640 different from version 2), and it would have been too confusing. That
4641 debug format is long obsolete, but the option cannot be changed now.
4642 Instead use an additional @option{-g@var{level}} option to change the
4643 debug level for DWARF.
4647 Turn off generation of debug info, if leaving out this option would have
4648 generated it, or turn it on at level 2 otherwise. The position of this
4649 argument in the command line does not matter, it takes effect after all
4650 other options are processed, and it does so only once, no matter how
4651 many times it is given. This is mainly intended to be used with
4652 @option{-fcompare-debug}.
4654 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4655 @opindex fdump-final-insns
4656 Dump the final internal representation (RTL) to @var{file}. If the
4657 optional argument is omitted (or if @var{file} is @code{.}), the name
4658 of the dump file will be determined by appending @code{.gkd} to the
4659 compilation output file name.
4661 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4662 @opindex fcompare-debug
4663 @opindex fno-compare-debug
4664 If no error occurs during compilation, run the compiler a second time,
4665 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4666 passed to the second compilation. Dump the final internal
4667 representation in both compilations, and print an error if they differ.
4669 If the equal sign is omitted, the default @option{-gtoggle} is used.
4671 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4672 and nonzero, implicitly enables @option{-fcompare-debug}. If
4673 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4674 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4677 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4678 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4679 of the final representation and the second compilation, preventing even
4680 @env{GCC_COMPARE_DEBUG} from taking effect.
4682 To verify full coverage during @option{-fcompare-debug} testing, set
4683 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4684 which GCC will reject as an invalid option in any actual compilation
4685 (rather than preprocessing, assembly or linking). To get just a
4686 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4687 not overridden} will do.
4689 @item -fcompare-debug-second
4690 @opindex fcompare-debug-second
4691 This option is implicitly passed to the compiler for the second
4692 compilation requested by @option{-fcompare-debug}, along with options to
4693 silence warnings, and omitting other options that would cause
4694 side-effect compiler outputs to files or to the standard output. Dump
4695 files and preserved temporary files are renamed so as to contain the
4696 @code{.gk} additional extension during the second compilation, to avoid
4697 overwriting those generated by the first.
4699 When this option is passed to the compiler driver, it causes the
4700 @emph{first} compilation to be skipped, which makes it useful for little
4701 other than debugging the compiler proper.
4703 @item -feliminate-dwarf2-dups
4704 @opindex feliminate-dwarf2-dups
4705 Compress DWARF2 debugging information by eliminating duplicated
4706 information about each symbol. This option only makes sense when
4707 generating DWARF2 debugging information with @option{-gdwarf-2}.
4709 @item -femit-struct-debug-baseonly
4710 Emit debug information for struct-like types
4711 only when the base name of the compilation source file
4712 matches the base name of file in which the struct was defined.
4714 This option substantially reduces the size of debugging information,
4715 but at significant potential loss in type information to the debugger.
4716 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4717 See @option{-femit-struct-debug-detailed} for more detailed control.
4719 This option works only with DWARF 2.
4721 @item -femit-struct-debug-reduced
4722 Emit debug information for struct-like types
4723 only when the base name of the compilation source file
4724 matches the base name of file in which the type was defined,
4725 unless the struct is a template or defined in a system header.
4727 This option significantly reduces the size of debugging information,
4728 with some potential loss in type information to the debugger.
4729 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4730 See @option{-femit-struct-debug-detailed} for more detailed control.
4732 This option works only with DWARF 2.
4734 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4735 Specify the struct-like types
4736 for which the compiler will generate debug information.
4737 The intent is to reduce duplicate struct debug information
4738 between different object files within the same program.
4740 This option is a detailed version of
4741 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4742 which will serve for most needs.
4744 A specification has the syntax@*
4745 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4747 The optional first word limits the specification to
4748 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4749 A struct type is used directly when it is the type of a variable, member.
4750 Indirect uses arise through pointers to structs.
4751 That is, when use of an incomplete struct would be legal, the use is indirect.
4753 @samp{struct one direct; struct two * indirect;}.
4755 The optional second word limits the specification to
4756 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4757 Generic structs are a bit complicated to explain.
4758 For C++, these are non-explicit specializations of template classes,
4759 or non-template classes within the above.
4760 Other programming languages have generics,
4761 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4763 The third word specifies the source files for those
4764 structs for which the compiler will emit debug information.
4765 The values @samp{none} and @samp{any} have the normal meaning.
4766 The value @samp{base} means that
4767 the base of name of the file in which the type declaration appears
4768 must match the base of the name of the main compilation file.
4769 In practice, this means that
4770 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4771 but types declared in other header will not.
4772 The value @samp{sys} means those types satisfying @samp{base}
4773 or declared in system or compiler headers.
4775 You may need to experiment to determine the best settings for your application.
4777 The default is @samp{-femit-struct-debug-detailed=all}.
4779 This option works only with DWARF 2.
4781 @item -fenable-icf-debug
4782 @opindex fenable-icf-debug
4783 Generate additional debug information to support identical code folding (ICF).
4784 This option only works with DWARF version 2 or higher.
4786 @item -fno-merge-debug-strings
4787 @opindex fmerge-debug-strings
4788 @opindex fno-merge-debug-strings
4789 Direct the linker to not merge together strings in the debugging
4790 information which are identical in different object files. Merging is
4791 not supported by all assemblers or linkers. Merging decreases the size
4792 of the debug information in the output file at the cost of increasing
4793 link processing time. Merging is enabled by default.
4795 @item -fdebug-prefix-map=@var{old}=@var{new}
4796 @opindex fdebug-prefix-map
4797 When compiling files in directory @file{@var{old}}, record debugging
4798 information describing them as in @file{@var{new}} instead.
4800 @item -fno-dwarf2-cfi-asm
4801 @opindex fdwarf2-cfi-asm
4802 @opindex fno-dwarf2-cfi-asm
4803 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4804 instead of using GAS @code{.cfi_*} directives.
4806 @cindex @command{prof}
4809 Generate extra code to write profile information suitable for the
4810 analysis program @command{prof}. You must use this option when compiling
4811 the source files you want data about, and you must also use it when
4814 @cindex @command{gprof}
4817 Generate extra code to write profile information suitable for the
4818 analysis program @command{gprof}. You must use this option when compiling
4819 the source files you want data about, and you must also use it when
4824 Makes the compiler print out each function name as it is compiled, and
4825 print some statistics about each pass when it finishes.
4828 @opindex ftime-report
4829 Makes the compiler print some statistics about the time consumed by each
4830 pass when it finishes.
4833 @opindex fmem-report
4834 Makes the compiler print some statistics about permanent memory
4835 allocation when it finishes.
4837 @item -fpre-ipa-mem-report
4838 @opindex fpre-ipa-mem-report
4839 @item -fpost-ipa-mem-report
4840 @opindex fpost-ipa-mem-report
4841 Makes the compiler print some statistics about permanent memory
4842 allocation before or after interprocedural optimization.
4845 @opindex fstack-usage
4846 Makes the compiler output stack usage information for the program, on a
4847 per-function basis. The filename for the dump is made by appending
4848 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4849 the output file, if explicitly specified and it is not an executable,
4850 otherwise it is the basename of the source file. An entry is made up
4855 The name of the function.
4859 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4862 The qualifier @code{static} means that the function manipulates the stack
4863 statically: a fixed number of bytes are allocated for the frame on function
4864 entry and released on function exit; no stack adjustments are otherwise made
4865 in the function. The second field is this fixed number of bytes.
4867 The qualifier @code{dynamic} means that the function manipulates the stack
4868 dynamically: in addition to the static allocation described above, stack
4869 adjustments are made in the body of the function, for example to push/pop
4870 arguments around function calls. If the qualifier @code{bounded} is also
4871 present, the amount of these adjustments is bounded at compile-time and
4872 the second field is an upper bound of the total amount of stack used by
4873 the function. If it is not present, the amount of these adjustments is
4874 not bounded at compile-time and the second field only represents the
4877 @item -fprofile-arcs
4878 @opindex fprofile-arcs
4879 Add code so that program flow @dfn{arcs} are instrumented. During
4880 execution the program records how many times each branch and call is
4881 executed and how many times it is taken or returns. When the compiled
4882 program exits it saves this data to a file called
4883 @file{@var{auxname}.gcda} for each source file. The data may be used for
4884 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4885 test coverage analysis (@option{-ftest-coverage}). Each object file's
4886 @var{auxname} is generated from the name of the output file, if
4887 explicitly specified and it is not the final executable, otherwise it is
4888 the basename of the source file. In both cases any suffix is removed
4889 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4890 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4891 @xref{Cross-profiling}.
4893 @cindex @command{gcov}
4897 This option is used to compile and link code instrumented for coverage
4898 analysis. The option is a synonym for @option{-fprofile-arcs}
4899 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4900 linking). See the documentation for those options for more details.
4905 Compile the source files with @option{-fprofile-arcs} plus optimization
4906 and code generation options. For test coverage analysis, use the
4907 additional @option{-ftest-coverage} option. You do not need to profile
4908 every source file in a program.
4911 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4912 (the latter implies the former).
4915 Run the program on a representative workload to generate the arc profile
4916 information. This may be repeated any number of times. You can run
4917 concurrent instances of your program, and provided that the file system
4918 supports locking, the data files will be correctly updated. Also
4919 @code{fork} calls are detected and correctly handled (double counting
4923 For profile-directed optimizations, compile the source files again with
4924 the same optimization and code generation options plus
4925 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4926 Control Optimization}).
4929 For test coverage analysis, use @command{gcov} to produce human readable
4930 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4931 @command{gcov} documentation for further information.
4935 With @option{-fprofile-arcs}, for each function of your program GCC
4936 creates a program flow graph, then finds a spanning tree for the graph.
4937 Only arcs that are not on the spanning tree have to be instrumented: the
4938 compiler adds code to count the number of times that these arcs are
4939 executed. When an arc is the only exit or only entrance to a block, the
4940 instrumentation code can be added to the block; otherwise, a new basic
4941 block must be created to hold the instrumentation code.
4944 @item -ftest-coverage
4945 @opindex ftest-coverage
4946 Produce a notes file that the @command{gcov} code-coverage utility
4947 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4948 show program coverage. Each source file's note file is called
4949 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4950 above for a description of @var{auxname} and instructions on how to
4951 generate test coverage data. Coverage data will match the source files
4952 more closely, if you do not optimize.
4954 @item -fdbg-cnt-list
4955 @opindex fdbg-cnt-list
4956 Print the name and the counter upper bound for all debug counters.
4958 @item -fdbg-cnt=@var{counter-value-list}
4960 Set the internal debug counter upper bound. @var{counter-value-list}
4961 is a comma-separated list of @var{name}:@var{value} pairs
4962 which sets the upper bound of each debug counter @var{name} to @var{value}.
4963 All debug counters have the initial upper bound of @var{UINT_MAX},
4964 thus dbg_cnt() returns true always unless the upper bound is set by this option.
4965 e.g. With -fdbg-cnt=dce:10,tail_call:0
4966 dbg_cnt(dce) will return true only for first 10 invocations
4967 and dbg_cnt(tail_call) will return false always.
4969 @item -d@var{letters}
4970 @itemx -fdump-rtl-@var{pass}
4972 Says to make debugging dumps during compilation at times specified by
4973 @var{letters}. This is used for debugging the RTL-based passes of the
4974 compiler. The file names for most of the dumps are made by appending
4975 a pass number and a word to the @var{dumpname}, and the files are
4976 created in the directory of the output file. Note that the pass
4977 number is computed statically as passes get registered into the pass
4978 manager. Thus the numbering is not related to the dynamic order of
4979 execution of passes. In particular, a pass installed by a plugin
4980 could have a number over 200 even if it executed quite early.
4981 @var{dumpname} is generated from the name of the output file, if
4982 explicitly specified and it is not an executable, otherwise it is the
4983 basename of the source file. These switches may have different effects
4984 when @option{-E} is used for preprocessing.
4986 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4987 @option{-d} option @var{letters}. Here are the possible
4988 letters for use in @var{pass} and @var{letters}, and their meanings:
4992 @item -fdump-rtl-alignments
4993 @opindex fdump-rtl-alignments
4994 Dump after branch alignments have been computed.
4996 @item -fdump-rtl-asmcons
4997 @opindex fdump-rtl-asmcons
4998 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5000 @item -fdump-rtl-auto_inc_dec
5001 @opindex fdump-rtl-auto_inc_dec
5002 Dump after auto-inc-dec discovery. This pass is only run on
5003 architectures that have auto inc or auto dec instructions.
5005 @item -fdump-rtl-barriers
5006 @opindex fdump-rtl-barriers
5007 Dump after cleaning up the barrier instructions.
5009 @item -fdump-rtl-bbpart
5010 @opindex fdump-rtl-bbpart
5011 Dump after partitioning hot and cold basic blocks.
5013 @item -fdump-rtl-bbro
5014 @opindex fdump-rtl-bbro
5015 Dump after block reordering.
5017 @item -fdump-rtl-btl1
5018 @itemx -fdump-rtl-btl2
5019 @opindex fdump-rtl-btl2
5020 @opindex fdump-rtl-btl2
5021 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5022 after the two branch
5023 target load optimization passes.
5025 @item -fdump-rtl-bypass
5026 @opindex fdump-rtl-bypass
5027 Dump after jump bypassing and control flow optimizations.
5029 @item -fdump-rtl-combine
5030 @opindex fdump-rtl-combine
5031 Dump after the RTL instruction combination pass.
5033 @item -fdump-rtl-compgotos
5034 @opindex fdump-rtl-compgotos
5035 Dump after duplicating the computed gotos.
5037 @item -fdump-rtl-ce1
5038 @itemx -fdump-rtl-ce2
5039 @itemx -fdump-rtl-ce3
5040 @opindex fdump-rtl-ce1
5041 @opindex fdump-rtl-ce2
5042 @opindex fdump-rtl-ce3
5043 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5044 @option{-fdump-rtl-ce3} enable dumping after the three
5045 if conversion passes.
5047 @itemx -fdump-rtl-cprop_hardreg
5048 @opindex fdump-rtl-cprop_hardreg
5049 Dump after hard register copy propagation.
5051 @itemx -fdump-rtl-csa
5052 @opindex fdump-rtl-csa
5053 Dump after combining stack adjustments.
5055 @item -fdump-rtl-cse1
5056 @itemx -fdump-rtl-cse2
5057 @opindex fdump-rtl-cse1
5058 @opindex fdump-rtl-cse2
5059 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5060 the two common sub-expression elimination passes.
5062 @itemx -fdump-rtl-dce
5063 @opindex fdump-rtl-dce
5064 Dump after the standalone dead code elimination passes.
5066 @itemx -fdump-rtl-dbr
5067 @opindex fdump-rtl-dbr
5068 Dump after delayed branch scheduling.
5070 @item -fdump-rtl-dce1
5071 @itemx -fdump-rtl-dce2
5072 @opindex fdump-rtl-dce1
5073 @opindex fdump-rtl-dce2
5074 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5075 the two dead store elimination passes.
5078 @opindex fdump-rtl-eh
5079 Dump after finalization of EH handling code.
5081 @item -fdump-rtl-eh_ranges
5082 @opindex fdump-rtl-eh_ranges
5083 Dump after conversion of EH handling range regions.
5085 @item -fdump-rtl-expand
5086 @opindex fdump-rtl-expand
5087 Dump after RTL generation.
5089 @item -fdump-rtl-fwprop1
5090 @itemx -fdump-rtl-fwprop2
5091 @opindex fdump-rtl-fwprop1
5092 @opindex fdump-rtl-fwprop2
5093 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5094 dumping after the two forward propagation passes.
5096 @item -fdump-rtl-gcse1
5097 @itemx -fdump-rtl-gcse2
5098 @opindex fdump-rtl-gcse1
5099 @opindex fdump-rtl-gcse2
5100 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5101 after global common subexpression elimination.
5103 @item -fdump-rtl-init-regs
5104 @opindex fdump-rtl-init-regs
5105 Dump after the initialization of the registers.
5107 @item -fdump-rtl-initvals
5108 @opindex fdump-rtl-initvals
5109 Dump after the computation of the initial value sets.
5111 @itemx -fdump-rtl-into_cfglayout
5112 @opindex fdump-rtl-into_cfglayout
5113 Dump after converting to cfglayout mode.
5115 @item -fdump-rtl-ira
5116 @opindex fdump-rtl-ira
5117 Dump after iterated register allocation.
5119 @item -fdump-rtl-jump
5120 @opindex fdump-rtl-jump
5121 Dump after the second jump optimization.
5123 @item -fdump-rtl-loop2
5124 @opindex fdump-rtl-loop2
5125 @option{-fdump-rtl-loop2} enables dumping after the rtl
5126 loop optimization passes.
5128 @item -fdump-rtl-mach
5129 @opindex fdump-rtl-mach
5130 Dump after performing the machine dependent reorganization pass, if that
5133 @item -fdump-rtl-mode_sw
5134 @opindex fdump-rtl-mode_sw
5135 Dump after removing redundant mode switches.
5137 @item -fdump-rtl-rnreg
5138 @opindex fdump-rtl-rnreg
5139 Dump after register renumbering.
5141 @itemx -fdump-rtl-outof_cfglayout
5142 @opindex fdump-rtl-outof_cfglayout
5143 Dump after converting from cfglayout mode.
5145 @item -fdump-rtl-peephole2
5146 @opindex fdump-rtl-peephole2
5147 Dump after the peephole pass.
5149 @item -fdump-rtl-postreload
5150 @opindex fdump-rtl-postreload
5151 Dump after post-reload optimizations.
5153 @itemx -fdump-rtl-pro_and_epilogue
5154 @opindex fdump-rtl-pro_and_epilogue
5155 Dump after generating the function pro and epilogues.
5157 @item -fdump-rtl-regmove
5158 @opindex fdump-rtl-regmove
5159 Dump after the register move pass.
5161 @item -fdump-rtl-sched1
5162 @itemx -fdump-rtl-sched2
5163 @opindex fdump-rtl-sched1
5164 @opindex fdump-rtl-sched2
5165 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5166 after the basic block scheduling passes.
5168 @item -fdump-rtl-see
5169 @opindex fdump-rtl-see
5170 Dump after sign extension elimination.
5172 @item -fdump-rtl-seqabstr
5173 @opindex fdump-rtl-seqabstr
5174 Dump after common sequence discovery.
5176 @item -fdump-rtl-shorten
5177 @opindex fdump-rtl-shorten
5178 Dump after shortening branches.
5180 @item -fdump-rtl-sibling
5181 @opindex fdump-rtl-sibling
5182 Dump after sibling call optimizations.
5184 @item -fdump-rtl-split1
5185 @itemx -fdump-rtl-split2
5186 @itemx -fdump-rtl-split3
5187 @itemx -fdump-rtl-split4
5188 @itemx -fdump-rtl-split5
5189 @opindex fdump-rtl-split1
5190 @opindex fdump-rtl-split2
5191 @opindex fdump-rtl-split3
5192 @opindex fdump-rtl-split4
5193 @opindex fdump-rtl-split5
5194 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5195 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5196 @option{-fdump-rtl-split5} enable dumping after five rounds of
5197 instruction splitting.
5199 @item -fdump-rtl-sms
5200 @opindex fdump-rtl-sms
5201 Dump after modulo scheduling. This pass is only run on some
5204 @item -fdump-rtl-stack
5205 @opindex fdump-rtl-stack
5206 Dump after conversion from GCC's "flat register file" registers to the
5207 x87's stack-like registers. This pass is only run on x86 variants.
5209 @item -fdump-rtl-subreg1
5210 @itemx -fdump-rtl-subreg2
5211 @opindex fdump-rtl-subreg1
5212 @opindex fdump-rtl-subreg2
5213 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5214 the two subreg expansion passes.
5216 @item -fdump-rtl-unshare
5217 @opindex fdump-rtl-unshare
5218 Dump after all rtl has been unshared.
5220 @item -fdump-rtl-vartrack
5221 @opindex fdump-rtl-vartrack
5222 Dump after variable tracking.
5224 @item -fdump-rtl-vregs
5225 @opindex fdump-rtl-vregs
5226 Dump after converting virtual registers to hard registers.
5228 @item -fdump-rtl-web
5229 @opindex fdump-rtl-web
5230 Dump after live range splitting.
5232 @item -fdump-rtl-regclass
5233 @itemx -fdump-rtl-subregs_of_mode_init
5234 @itemx -fdump-rtl-subregs_of_mode_finish
5235 @itemx -fdump-rtl-dfinit
5236 @itemx -fdump-rtl-dfinish
5237 @opindex fdump-rtl-regclass
5238 @opindex fdump-rtl-subregs_of_mode_init
5239 @opindex fdump-rtl-subregs_of_mode_finish
5240 @opindex fdump-rtl-dfinit
5241 @opindex fdump-rtl-dfinish
5242 These dumps are defined but always produce empty files.
5244 @item -fdump-rtl-all
5245 @opindex fdump-rtl-all
5246 Produce all the dumps listed above.
5250 Annotate the assembler output with miscellaneous debugging information.
5254 Dump all macro definitions, at the end of preprocessing, in addition to
5259 Produce a core dump whenever an error occurs.
5263 Print statistics on memory usage, at the end of the run, to
5268 Annotate the assembler output with a comment indicating which
5269 pattern and alternative was used. The length of each instruction is
5274 Dump the RTL in the assembler output as a comment before each instruction.
5275 Also turns on @option{-dp} annotation.
5279 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5280 dump a representation of the control flow graph suitable for viewing with VCG
5281 to @file{@var{file}.@var{pass}.vcg}.
5285 Just generate RTL for a function instead of compiling it. Usually used
5286 with @option{-fdump-rtl-expand}.
5290 @opindex fdump-noaddr
5291 When doing debugging dumps, suppress address output. This makes it more
5292 feasible to use diff on debugging dumps for compiler invocations with
5293 different compiler binaries and/or different
5294 text / bss / data / heap / stack / dso start locations.
5296 @item -fdump-unnumbered
5297 @opindex fdump-unnumbered
5298 When doing debugging dumps, suppress instruction numbers and address output.
5299 This makes it more feasible to use diff on debugging dumps for compiler
5300 invocations with different options, in particular with and without
5303 @item -fdump-unnumbered-links
5304 @opindex fdump-unnumbered-links
5305 When doing debugging dumps (see @option{-d} option above), suppress
5306 instruction numbers for the links to the previous and next instructions
5309 @item -fdump-translation-unit @r{(C++ only)}
5310 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5311 @opindex fdump-translation-unit
5312 Dump a representation of the tree structure for the entire translation
5313 unit to a file. The file name is made by appending @file{.tu} to the
5314 source file name, and the file is created in the same directory as the
5315 output file. If the @samp{-@var{options}} form is used, @var{options}
5316 controls the details of the dump as described for the
5317 @option{-fdump-tree} options.
5319 @item -fdump-class-hierarchy @r{(C++ only)}
5320 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5321 @opindex fdump-class-hierarchy
5322 Dump a representation of each class's hierarchy and virtual function
5323 table layout to a file. The file name is made by appending
5324 @file{.class} to the source file name, and the file is created in the
5325 same directory as the output file. If the @samp{-@var{options}} form
5326 is used, @var{options} controls the details of the dump as described
5327 for the @option{-fdump-tree} options.
5329 @item -fdump-ipa-@var{switch}
5331 Control the dumping at various stages of inter-procedural analysis
5332 language tree to a file. The file name is generated by appending a
5333 switch specific suffix to the source file name, and the file is created
5334 in the same directory as the output file. The following dumps are
5339 Enables all inter-procedural analysis dumps.
5342 Dumps information about call-graph optimization, unused function removal,
5343 and inlining decisions.
5346 Dump after function inlining.
5350 @item -fdump-statistics-@var{option}
5351 @opindex fdump-statistics
5352 Enable and control dumping of pass statistics in a separate file. The
5353 file name is generated by appending a suffix ending in
5354 @samp{.statistics} to the source file name, and the file is created in
5355 the same directory as the output file. If the @samp{-@var{option}}
5356 form is used, @samp{-stats} will cause counters to be summed over the
5357 whole compilation unit while @samp{-details} will dump every event as
5358 the passes generate them. The default with no option is to sum
5359 counters for each function compiled.
5361 @item -fdump-tree-@var{switch}
5362 @itemx -fdump-tree-@var{switch}-@var{options}
5364 Control the dumping at various stages of processing the intermediate
5365 language tree to a file. The file name is generated by appending a
5366 switch specific suffix to the source file name, and the file is
5367 created in the same directory as the output file. If the
5368 @samp{-@var{options}} form is used, @var{options} is a list of
5369 @samp{-} separated options that control the details of the dump. Not
5370 all options are applicable to all dumps, those which are not
5371 meaningful will be ignored. The following options are available
5375 Print the address of each node. Usually this is not meaningful as it
5376 changes according to the environment and source file. Its primary use
5377 is for tying up a dump file with a debug environment.
5379 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5380 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5381 use working backward from mangled names in the assembly file.
5383 Inhibit dumping of members of a scope or body of a function merely
5384 because that scope has been reached. Only dump such items when they
5385 are directly reachable by some other path. When dumping pretty-printed
5386 trees, this option inhibits dumping the bodies of control structures.
5388 Print a raw representation of the tree. By default, trees are
5389 pretty-printed into a C-like representation.
5391 Enable more detailed dumps (not honored by every dump option).
5393 Enable dumping various statistics about the pass (not honored by every dump
5396 Enable showing basic block boundaries (disabled in raw dumps).
5398 Enable showing virtual operands for every statement.
5400 Enable showing line numbers for statements.
5402 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5404 Enable showing the tree dump for each statement.
5406 Enable showing the EH region number holding each statement.
5408 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5409 and @option{lineno}.
5412 The following tree dumps are possible:
5416 @opindex fdump-tree-original
5417 Dump before any tree based optimization, to @file{@var{file}.original}.
5420 @opindex fdump-tree-optimized
5421 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5424 @opindex fdump-tree-gimple
5425 Dump each function before and after the gimplification pass to a file. The
5426 file name is made by appending @file{.gimple} to the source file name.
5429 @opindex fdump-tree-cfg
5430 Dump the control flow graph of each function to a file. The file name is
5431 made by appending @file{.cfg} to the source file name.
5434 @opindex fdump-tree-vcg
5435 Dump the control flow graph of each function to a file in VCG format. The
5436 file name is made by appending @file{.vcg} to the source file name. Note
5437 that if the file contains more than one function, the generated file cannot
5438 be used directly by VCG@. You will need to cut and paste each function's
5439 graph into its own separate file first.
5442 @opindex fdump-tree-ch
5443 Dump each function after copying loop headers. The file name is made by
5444 appending @file{.ch} to the source file name.
5447 @opindex fdump-tree-ssa
5448 Dump SSA related information to a file. The file name is made by appending
5449 @file{.ssa} to the source file name.
5452 @opindex fdump-tree-alias
5453 Dump aliasing information for each function. The file name is made by
5454 appending @file{.alias} to the source file name.
5457 @opindex fdump-tree-ccp
5458 Dump each function after CCP@. The file name is made by appending
5459 @file{.ccp} to the source file name.
5462 @opindex fdump-tree-storeccp
5463 Dump each function after STORE-CCP@. The file name is made by appending
5464 @file{.storeccp} to the source file name.
5467 @opindex fdump-tree-pre
5468 Dump trees after partial redundancy elimination. The file name is made
5469 by appending @file{.pre} to the source file name.
5472 @opindex fdump-tree-fre
5473 Dump trees after full redundancy elimination. The file name is made
5474 by appending @file{.fre} to the source file name.
5477 @opindex fdump-tree-copyprop
5478 Dump trees after copy propagation. The file name is made
5479 by appending @file{.copyprop} to the source file name.
5481 @item store_copyprop
5482 @opindex fdump-tree-store_copyprop
5483 Dump trees after store copy-propagation. The file name is made
5484 by appending @file{.store_copyprop} to the source file name.
5487 @opindex fdump-tree-dce
5488 Dump each function after dead code elimination. The file name is made by
5489 appending @file{.dce} to the source file name.
5492 @opindex fdump-tree-mudflap
5493 Dump each function after adding mudflap instrumentation. The file name is
5494 made by appending @file{.mudflap} to the source file name.
5497 @opindex fdump-tree-sra
5498 Dump each function after performing scalar replacement of aggregates. The
5499 file name is made by appending @file{.sra} to the source file name.
5502 @opindex fdump-tree-sink
5503 Dump each function after performing code sinking. The file name is made
5504 by appending @file{.sink} to the source file name.
5507 @opindex fdump-tree-dom
5508 Dump each function after applying dominator tree optimizations. The file
5509 name is made by appending @file{.dom} to the source file name.
5512 @opindex fdump-tree-dse
5513 Dump each function after applying dead store elimination. The file
5514 name is made by appending @file{.dse} to the source file name.
5517 @opindex fdump-tree-phiopt
5518 Dump each function after optimizing PHI nodes into straightline code. The file
5519 name is made by appending @file{.phiopt} to the source file name.
5522 @opindex fdump-tree-forwprop
5523 Dump each function after forward propagating single use variables. The file
5524 name is made by appending @file{.forwprop} to the source file name.
5527 @opindex fdump-tree-copyrename
5528 Dump each function after applying the copy rename optimization. The file
5529 name is made by appending @file{.copyrename} to the source file name.
5532 @opindex fdump-tree-nrv
5533 Dump each function after applying the named return value optimization on
5534 generic trees. The file name is made by appending @file{.nrv} to the source
5538 @opindex fdump-tree-vect
5539 Dump each function after applying vectorization of loops. The file name is
5540 made by appending @file{.vect} to the source file name.
5543 @opindex fdump-tree-slp
5544 Dump each function after applying vectorization of basic blocks. The file name
5545 is made by appending @file{.slp} to the source file name.
5548 @opindex fdump-tree-vrp
5549 Dump each function after Value Range Propagation (VRP). The file name
5550 is made by appending @file{.vrp} to the source file name.
5553 @opindex fdump-tree-all
5554 Enable all the available tree dumps with the flags provided in this option.
5557 @item -ftree-vectorizer-verbose=@var{n}
5558 @opindex ftree-vectorizer-verbose
5559 This option controls the amount of debugging output the vectorizer prints.
5560 This information is written to standard error, unless
5561 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5562 in which case it is output to the usual dump listing file, @file{.vect}.
5563 For @var{n}=0 no diagnostic information is reported.
5564 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5565 and the total number of loops that got vectorized.
5566 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5567 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5568 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5569 level that @option{-fdump-tree-vect-stats} uses.
5570 Higher verbosity levels mean either more information dumped for each
5571 reported loop, or same amount of information reported for more loops:
5572 if @var{n}=3, vectorizer cost model information is reported.
5573 If @var{n}=4, alignment related information is added to the reports.
5574 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5575 memory access-patterns) is added to the reports.
5576 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5577 that did not pass the first analysis phase (i.e., may not be countable, or
5578 may have complicated control-flow).
5579 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5580 If @var{n}=8, SLP related information is added to the reports.
5581 For @var{n}=9, all the information the vectorizer generates during its
5582 analysis and transformation is reported. This is the same verbosity level
5583 that @option{-fdump-tree-vect-details} uses.
5585 @item -frandom-seed=@var{string}
5586 @opindex frandom-seed
5587 This option provides a seed that GCC uses when it would otherwise use
5588 random numbers. It is used to generate certain symbol names
5589 that have to be different in every compiled file. It is also used to
5590 place unique stamps in coverage data files and the object files that
5591 produce them. You can use the @option{-frandom-seed} option to produce
5592 reproducibly identical object files.
5594 The @var{string} should be different for every file you compile.
5596 @item -fsched-verbose=@var{n}
5597 @opindex fsched-verbose
5598 On targets that use instruction scheduling, this option controls the
5599 amount of debugging output the scheduler prints. This information is
5600 written to standard error, unless @option{-fdump-rtl-sched1} or
5601 @option{-fdump-rtl-sched2} is specified, in which case it is output
5602 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5603 respectively. However for @var{n} greater than nine, the output is
5604 always printed to standard error.
5606 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5607 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5608 For @var{n} greater than one, it also output basic block probabilities,
5609 detailed ready list information and unit/insn info. For @var{n} greater
5610 than two, it includes RTL at abort point, control-flow and regions info.
5611 And for @var{n} over four, @option{-fsched-verbose} also includes
5615 @itemx -save-temps=cwd
5617 Store the usual ``temporary'' intermediate files permanently; place them
5618 in the current directory and name them based on the source file. Thus,
5619 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5620 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5621 preprocessed @file{foo.i} output file even though the compiler now
5622 normally uses an integrated preprocessor.
5624 When used in combination with the @option{-x} command line option,
5625 @option{-save-temps} is sensible enough to avoid over writing an
5626 input source file with the same extension as an intermediate file.
5627 The corresponding intermediate file may be obtained by renaming the
5628 source file before using @option{-save-temps}.
5630 If you invoke GCC in parallel, compiling several different source
5631 files that share a common base name in different subdirectories or the
5632 same source file compiled for multiple output destinations, it is
5633 likely that the different parallel compilers will interfere with each
5634 other, and overwrite the temporary files. For instance:
5637 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5638 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5641 may result in @file{foo.i} and @file{foo.o} being written to
5642 simultaneously by both compilers.
5644 @item -save-temps=obj
5645 @opindex save-temps=obj
5646 Store the usual ``temporary'' intermediate files permanently. If the
5647 @option{-o} option is used, the temporary files are based on the
5648 object file. If the @option{-o} option is not used, the
5649 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5654 gcc -save-temps=obj -c foo.c
5655 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5656 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5659 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5660 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5661 @file{dir2/yfoobar.o}.
5663 @item -time@r{[}=@var{file}@r{]}
5665 Report the CPU time taken by each subprocess in the compilation
5666 sequence. For C source files, this is the compiler proper and assembler
5667 (plus the linker if linking is done).
5669 Without the specification of an output file, the output looks like this:
5676 The first number on each line is the ``user time'', that is time spent
5677 executing the program itself. The second number is ``system time'',
5678 time spent executing operating system routines on behalf of the program.
5679 Both numbers are in seconds.
5681 With the specification of an output file, the output is appended to the
5682 named file, and it looks like this:
5685 0.12 0.01 cc1 @var{options}
5686 0.00 0.01 as @var{options}
5689 The ``user time'' and the ``system time'' are moved before the program
5690 name, and the options passed to the program are displayed, so that one
5691 can later tell what file was being compiled, and with which options.
5693 @item -fvar-tracking
5694 @opindex fvar-tracking
5695 Run variable tracking pass. It computes where variables are stored at each
5696 position in code. Better debugging information is then generated
5697 (if the debugging information format supports this information).
5699 It is enabled by default when compiling with optimization (@option{-Os},
5700 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5701 the debug info format supports it.
5703 @item -fvar-tracking-assignments
5704 @opindex fvar-tracking-assignments
5705 @opindex fno-var-tracking-assignments
5706 Annotate assignments to user variables early in the compilation and
5707 attempt to carry the annotations over throughout the compilation all the
5708 way to the end, in an attempt to improve debug information while
5709 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5711 It can be enabled even if var-tracking is disabled, in which case
5712 annotations will be created and maintained, but discarded at the end.
5714 @item -fvar-tracking-assignments-toggle
5715 @opindex fvar-tracking-assignments-toggle
5716 @opindex fno-var-tracking-assignments-toggle
5717 Toggle @option{-fvar-tracking-assignments}, in the same way that
5718 @option{-gtoggle} toggles @option{-g}.
5720 @item -print-file-name=@var{library}
5721 @opindex print-file-name
5722 Print the full absolute name of the library file @var{library} that
5723 would be used when linking---and don't do anything else. With this
5724 option, GCC does not compile or link anything; it just prints the
5727 @item -print-multi-directory
5728 @opindex print-multi-directory
5729 Print the directory name corresponding to the multilib selected by any
5730 other switches present in the command line. This directory is supposed
5731 to exist in @env{GCC_EXEC_PREFIX}.
5733 @item -print-multi-lib
5734 @opindex print-multi-lib
5735 Print the mapping from multilib directory names to compiler switches
5736 that enable them. The directory name is separated from the switches by
5737 @samp{;}, and each switch starts with an @samp{@@} instead of the
5738 @samp{-}, without spaces between multiple switches. This is supposed to
5739 ease shell-processing.
5741 @item -print-multi-os-directory
5742 @opindex print-multi-os-directory
5743 Print the path to OS libraries for the selected
5744 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5745 present in the @file{lib} subdirectory and no multilibs are used, this is
5746 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5747 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5748 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5749 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5751 @item -print-prog-name=@var{program}
5752 @opindex print-prog-name
5753 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5755 @item -print-libgcc-file-name
5756 @opindex print-libgcc-file-name
5757 Same as @option{-print-file-name=libgcc.a}.
5759 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5760 but you do want to link with @file{libgcc.a}. You can do
5763 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5766 @item -print-search-dirs
5767 @opindex print-search-dirs
5768 Print the name of the configured installation directory and a list of
5769 program and library directories @command{gcc} will search---and don't do anything else.
5771 This is useful when @command{gcc} prints the error message
5772 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5773 To resolve this you either need to put @file{cpp0} and the other compiler
5774 components where @command{gcc} expects to find them, or you can set the environment
5775 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5776 Don't forget the trailing @samp{/}.
5777 @xref{Environment Variables}.
5779 @item -print-sysroot
5780 @opindex print-sysroot
5781 Print the target sysroot directory that will be used during
5782 compilation. This is the target sysroot specified either at configure
5783 time or using the @option{--sysroot} option, possibly with an extra
5784 suffix that depends on compilation options. If no target sysroot is
5785 specified, the option prints nothing.
5787 @item -print-sysroot-headers-suffix
5788 @opindex print-sysroot-headers-suffix
5789 Print the suffix added to the target sysroot when searching for
5790 headers, or give an error if the compiler is not configured with such
5791 a suffix---and don't do anything else.
5794 @opindex dumpmachine
5795 Print the compiler's target machine (for example,
5796 @samp{i686-pc-linux-gnu})---and don't do anything else.
5799 @opindex dumpversion
5800 Print the compiler version (for example, @samp{3.0})---and don't do
5805 Print the compiler's built-in specs---and don't do anything else. (This
5806 is used when GCC itself is being built.) @xref{Spec Files}.
5808 @item -feliminate-unused-debug-types
5809 @opindex feliminate-unused-debug-types
5810 Normally, when producing DWARF2 output, GCC will emit debugging
5811 information for all types declared in a compilation
5812 unit, regardless of whether or not they are actually used
5813 in that compilation unit. Sometimes this is useful, such as
5814 if, in the debugger, you want to cast a value to a type that is
5815 not actually used in your program (but is declared). More often,
5816 however, this results in a significant amount of wasted space.
5817 With this option, GCC will avoid producing debug symbol output
5818 for types that are nowhere used in the source file being compiled.
5821 @node Optimize Options
5822 @section Options That Control Optimization
5823 @cindex optimize options
5824 @cindex options, optimization
5826 These options control various sorts of optimizations.
5828 Without any optimization option, the compiler's goal is to reduce the
5829 cost of compilation and to make debugging produce the expected
5830 results. Statements are independent: if you stop the program with a
5831 breakpoint between statements, you can then assign a new value to any
5832 variable or change the program counter to any other statement in the
5833 function and get exactly the results you would expect from the source
5836 Turning on optimization flags makes the compiler attempt to improve
5837 the performance and/or code size at the expense of compilation time
5838 and possibly the ability to debug the program.
5840 The compiler performs optimization based on the knowledge it has of the
5841 program. Compiling multiple files at once to a single output file mode allows
5842 the compiler to use information gained from all of the files when compiling
5845 Not all optimizations are controlled directly by a flag. Only
5846 optimizations that have a flag are listed in this section.
5848 Most optimizations are only enabled if an @option{-O} level is set on
5849 the command line. Otherwise they are disabled, even if individual
5850 optimization flags are specified.
5852 Depending on the target and how GCC was configured, a slightly different
5853 set of optimizations may be enabled at each @option{-O} level than
5854 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5855 to find out the exact set of optimizations that are enabled at each level.
5856 @xref{Overall Options}, for examples.
5863 Optimize. Optimizing compilation takes somewhat more time, and a lot
5864 more memory for a large function.
5866 With @option{-O}, the compiler tries to reduce code size and execution
5867 time, without performing any optimizations that take a great deal of
5870 @option{-O} turns on the following optimization flags:
5874 -fcprop-registers @gol
5877 -fdelayed-branch @gol
5879 -fguess-branch-probability @gol
5880 -fif-conversion2 @gol
5881 -fif-conversion @gol
5882 -fipa-pure-const @gol
5884 -fipa-reference @gol
5886 -fsplit-wide-types @gol
5888 -ftree-builtin-call-dce @gol
5891 -ftree-copyrename @gol
5893 -ftree-dominator-opts @gol
5895 -ftree-forwprop @gol
5903 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5904 where doing so does not interfere with debugging.
5908 Optimize even more. GCC performs nearly all supported optimizations
5909 that do not involve a space-speed tradeoff.
5910 As compared to @option{-O}, this option increases both compilation time
5911 and the performance of the generated code.
5913 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5914 also turns on the following optimization flags:
5915 @gccoptlist{-fthread-jumps @gol
5916 -falign-functions -falign-jumps @gol
5917 -falign-loops -falign-labels @gol
5920 -fcse-follow-jumps -fcse-skip-blocks @gol
5921 -fdelete-null-pointer-checks @gol
5923 -fexpensive-optimizations @gol
5924 -fgcse -fgcse-lm @gol
5925 -finline-small-functions @gol
5926 -findirect-inlining @gol
5928 -foptimize-sibling-calls @gol
5929 -fpartial-inlining @gol
5932 -freorder-blocks -freorder-functions @gol
5933 -frerun-cse-after-loop @gol
5934 -fsched-interblock -fsched-spec @gol
5935 -fschedule-insns -fschedule-insns2 @gol
5936 -fstrict-aliasing -fstrict-overflow @gol
5937 -ftree-switch-conversion @gol
5941 Please note the warning under @option{-fgcse} about
5942 invoking @option{-O2} on programs that use computed gotos.
5946 Optimize yet more. @option{-O3} turns on all optimizations specified
5947 by @option{-O2} and also turns on the @option{-finline-functions},
5948 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5949 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
5950 @option{-fipa-cp-clone} options.
5954 Reduce compilation time and make debugging produce the expected
5955 results. This is the default.
5959 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5960 do not typically increase code size. It also performs further
5961 optimizations designed to reduce code size.
5963 @option{-Os} disables the following optimization flags:
5964 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5965 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5966 -fprefetch-loop-arrays -ftree-vect-loop-version}
5970 Disregard strict standards compliance. @option{-Ofast} enables all
5971 @option{-O3} optimizations. It also enables optimizations that are not
5972 valid for all standard compliant programs.
5973 It turns on @option{-ffast-math}.
5975 If you use multiple @option{-O} options, with or without level numbers,
5976 the last such option is the one that is effective.
5979 Options of the form @option{-f@var{flag}} specify machine-independent
5980 flags. Most flags have both positive and negative forms; the negative
5981 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5982 below, only one of the forms is listed---the one you typically will
5983 use. You can figure out the other form by either removing @samp{no-}
5986 The following options control specific optimizations. They are either
5987 activated by @option{-O} options or are related to ones that are. You
5988 can use the following flags in the rare cases when ``fine-tuning'' of
5989 optimizations to be performed is desired.
5992 @item -fno-default-inline
5993 @opindex fno-default-inline
5994 Do not make member functions inline by default merely because they are
5995 defined inside the class scope (C++ only). Otherwise, when you specify
5996 @w{@option{-O}}, member functions defined inside class scope are compiled
5997 inline by default; i.e., you don't need to add @samp{inline} in front of
5998 the member function name.
6000 @item -fno-defer-pop
6001 @opindex fno-defer-pop
6002 Always pop the arguments to each function call as soon as that function
6003 returns. For machines which must pop arguments after a function call,
6004 the compiler normally lets arguments accumulate on the stack for several
6005 function calls and pops them all at once.
6007 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6009 @item -fforward-propagate
6010 @opindex fforward-propagate
6011 Perform a forward propagation pass on RTL@. The pass tries to combine two
6012 instructions and checks if the result can be simplified. If loop unrolling
6013 is active, two passes are performed and the second is scheduled after
6016 This option is enabled by default at optimization levels @option{-O},
6017 @option{-O2}, @option{-O3}, @option{-Os}.
6019 @item -ffp-contract=@var{style}
6020 @opindex ffp-contract
6021 @option{-ffp-contract=off} disables floating-point expression contraction.
6022 @option{-ffp-contract=fast} enables floating-point expression contraction
6023 such as forming of fused multiply-add operations if the target has
6024 native support for them.
6025 @option{-ffp-contract=on} enables floating-point expression contraction
6026 if allowed by the language standard. This is currently not implemented
6027 and treated equal to @option{-ffp-contract=off}.
6029 The default is @option{-ffp-contract=fast}.
6031 @item -fomit-frame-pointer
6032 @opindex fomit-frame-pointer
6033 Don't keep the frame pointer in a register for functions that
6034 don't need one. This avoids the instructions to save, set up and
6035 restore frame pointers; it also makes an extra register available
6036 in many functions. @strong{It also makes debugging impossible on
6039 On some machines, such as the VAX, this flag has no effect, because
6040 the standard calling sequence automatically handles the frame pointer
6041 and nothing is saved by pretending it doesn't exist. The
6042 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6043 whether a target machine supports this flag. @xref{Registers,,Register
6044 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6046 Starting with GCC version 4.6, the default setting (when not optimizing for
6047 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6048 @option{-fomit-frame-pointer}. The default can be reverted to
6049 @option{-fno-omit-frame-pointer} by configuring GCC with the
6050 @option{--enable-frame-pointer} configure option.
6052 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6054 @item -foptimize-sibling-calls
6055 @opindex foptimize-sibling-calls
6056 Optimize sibling and tail recursive calls.
6058 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6062 Don't pay attention to the @code{inline} keyword. Normally this option
6063 is used to keep the compiler from expanding any functions inline.
6064 Note that if you are not optimizing, no functions can be expanded inline.
6066 @item -finline-small-functions
6067 @opindex finline-small-functions
6068 Integrate functions into their callers when their body is smaller than expected
6069 function call code (so overall size of program gets smaller). The compiler
6070 heuristically decides which functions are simple enough to be worth integrating
6073 Enabled at level @option{-O2}.
6075 @item -findirect-inlining
6076 @opindex findirect-inlining
6077 Inline also indirect calls that are discovered to be known at compile
6078 time thanks to previous inlining. This option has any effect only
6079 when inlining itself is turned on by the @option{-finline-functions}
6080 or @option{-finline-small-functions} options.
6082 Enabled at level @option{-O2}.
6084 @item -finline-functions
6085 @opindex finline-functions
6086 Integrate all simple functions into their callers. The compiler
6087 heuristically decides which functions are simple enough to be worth
6088 integrating in this way.
6090 If all calls to a given function are integrated, and the function is
6091 declared @code{static}, then the function is normally not output as
6092 assembler code in its own right.
6094 Enabled at level @option{-O3}.
6096 @item -finline-functions-called-once
6097 @opindex finline-functions-called-once
6098 Consider all @code{static} functions called once for inlining into their
6099 caller even if they are not marked @code{inline}. If a call to a given
6100 function is integrated, then the function is not output as assembler code
6103 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6105 @item -fearly-inlining
6106 @opindex fearly-inlining
6107 Inline functions marked by @code{always_inline} and functions whose body seems
6108 smaller than the function call overhead early before doing
6109 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6110 makes profiling significantly cheaper and usually inlining faster on programs
6111 having large chains of nested wrapper functions.
6117 Perform interprocedural scalar replacement of aggregates, removal of
6118 unused parameters and replacement of parameters passed by reference
6119 by parameters passed by value.
6121 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6123 @item -finline-limit=@var{n}
6124 @opindex finline-limit
6125 By default, GCC limits the size of functions that can be inlined. This flag
6126 allows coarse control of this limit. @var{n} is the size of functions that
6127 can be inlined in number of pseudo instructions.
6129 Inlining is actually controlled by a number of parameters, which may be
6130 specified individually by using @option{--param @var{name}=@var{value}}.
6131 The @option{-finline-limit=@var{n}} option sets some of these parameters
6135 @item max-inline-insns-single
6136 is set to @var{n}/2.
6137 @item max-inline-insns-auto
6138 is set to @var{n}/2.
6141 See below for a documentation of the individual
6142 parameters controlling inlining and for the defaults of these parameters.
6144 @emph{Note:} there may be no value to @option{-finline-limit} that results
6145 in default behavior.
6147 @emph{Note:} pseudo instruction represents, in this particular context, an
6148 abstract measurement of function's size. In no way does it represent a count
6149 of assembly instructions and as such its exact meaning might change from one
6150 release to an another.
6152 @item -fno-keep-inline-dllexport
6153 @opindex -fno-keep-inline-dllexport
6154 This is a more fine-grained version of @option{-fkeep-inline-functions},
6155 which applies only to functions that are declared using the @code{dllexport}
6156 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6159 @item -fkeep-inline-functions
6160 @opindex fkeep-inline-functions
6161 In C, emit @code{static} functions that are declared @code{inline}
6162 into the object file, even if the function has been inlined into all
6163 of its callers. This switch does not affect functions using the
6164 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6165 inline functions into the object file.
6167 @item -fkeep-static-consts
6168 @opindex fkeep-static-consts
6169 Emit variables declared @code{static const} when optimization isn't turned
6170 on, even if the variables aren't referenced.
6172 GCC enables this option by default. If you want to force the compiler to
6173 check if the variable was referenced, regardless of whether or not
6174 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6176 @item -fmerge-constants
6177 @opindex fmerge-constants
6178 Attempt to merge identical constants (string constants and floating point
6179 constants) across compilation units.
6181 This option is the default for optimized compilation if the assembler and
6182 linker support it. Use @option{-fno-merge-constants} to inhibit this
6185 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6187 @item -fmerge-all-constants
6188 @opindex fmerge-all-constants
6189 Attempt to merge identical constants and identical variables.
6191 This option implies @option{-fmerge-constants}. In addition to
6192 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6193 arrays or initialized constant variables with integral or floating point
6194 types. Languages like C or C++ require each variable, including multiple
6195 instances of the same variable in recursive calls, to have distinct locations,
6196 so using this option will result in non-conforming
6199 @item -fmodulo-sched
6200 @opindex fmodulo-sched
6201 Perform swing modulo scheduling immediately before the first scheduling
6202 pass. This pass looks at innermost loops and reorders their
6203 instructions by overlapping different iterations.
6205 @item -fmodulo-sched-allow-regmoves
6206 @opindex fmodulo-sched-allow-regmoves
6207 Perform more aggressive SMS based modulo scheduling with register moves
6208 allowed. By setting this flag certain anti-dependences edges will be
6209 deleted which will trigger the generation of reg-moves based on the
6210 life-range analysis. This option is effective only with
6211 @option{-fmodulo-sched} enabled.
6213 @item -fno-branch-count-reg
6214 @opindex fno-branch-count-reg
6215 Do not use ``decrement and branch'' instructions on a count register,
6216 but instead generate a sequence of instructions that decrement a
6217 register, compare it against zero, then branch based upon the result.
6218 This option is only meaningful on architectures that support such
6219 instructions, which include x86, PowerPC, IA-64 and S/390.
6221 The default is @option{-fbranch-count-reg}.
6223 @item -fno-function-cse
6224 @opindex fno-function-cse
6225 Do not put function addresses in registers; make each instruction that
6226 calls a constant function contain the function's address explicitly.
6228 This option results in less efficient code, but some strange hacks
6229 that alter the assembler output may be confused by the optimizations
6230 performed when this option is not used.
6232 The default is @option{-ffunction-cse}
6234 @item -fno-zero-initialized-in-bss
6235 @opindex fno-zero-initialized-in-bss
6236 If the target supports a BSS section, GCC by default puts variables that
6237 are initialized to zero into BSS@. This can save space in the resulting
6240 This option turns off this behavior because some programs explicitly
6241 rely on variables going to the data section. E.g., so that the
6242 resulting executable can find the beginning of that section and/or make
6243 assumptions based on that.
6245 The default is @option{-fzero-initialized-in-bss}.
6247 @item -fmudflap -fmudflapth -fmudflapir
6251 @cindex bounds checking
6253 For front-ends that support it (C and C++), instrument all risky
6254 pointer/array dereferencing operations, some standard library
6255 string/heap functions, and some other associated constructs with
6256 range/validity tests. Modules so instrumented should be immune to
6257 buffer overflows, invalid heap use, and some other classes of C/C++
6258 programming errors. The instrumentation relies on a separate runtime
6259 library (@file{libmudflap}), which will be linked into a program if
6260 @option{-fmudflap} is given at link time. Run-time behavior of the
6261 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6262 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6265 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6266 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6267 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6268 instrumentation should ignore pointer reads. This produces less
6269 instrumentation (and therefore faster execution) and still provides
6270 some protection against outright memory corrupting writes, but allows
6271 erroneously read data to propagate within a program.
6273 @item -fthread-jumps
6274 @opindex fthread-jumps
6275 Perform optimizations where we check to see if a jump branches to a
6276 location where another comparison subsumed by the first is found. If
6277 so, the first branch is redirected to either the destination of the
6278 second branch or a point immediately following it, depending on whether
6279 the condition is known to be true or false.
6281 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6283 @item -fsplit-wide-types
6284 @opindex fsplit-wide-types
6285 When using a type that occupies multiple registers, such as @code{long
6286 long} on a 32-bit system, split the registers apart and allocate them
6287 independently. This normally generates better code for those types,
6288 but may make debugging more difficult.
6290 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6293 @item -fcse-follow-jumps
6294 @opindex fcse-follow-jumps
6295 In common subexpression elimination (CSE), scan through jump instructions
6296 when the target of the jump is not reached by any other path. For
6297 example, when CSE encounters an @code{if} statement with an
6298 @code{else} clause, CSE will follow the jump when the condition
6301 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6303 @item -fcse-skip-blocks
6304 @opindex fcse-skip-blocks
6305 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6306 follow jumps which conditionally skip over blocks. When CSE
6307 encounters a simple @code{if} statement with no else clause,
6308 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6309 body of the @code{if}.
6311 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6313 @item -frerun-cse-after-loop
6314 @opindex frerun-cse-after-loop
6315 Re-run common subexpression elimination after loop optimizations has been
6318 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6322 Perform a global common subexpression elimination pass.
6323 This pass also performs global constant and copy propagation.
6325 @emph{Note:} When compiling a program using computed gotos, a GCC
6326 extension, you may get better runtime performance if you disable
6327 the global common subexpression elimination pass by adding
6328 @option{-fno-gcse} to the command line.
6330 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6334 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6335 attempt to move loads which are only killed by stores into themselves. This
6336 allows a loop containing a load/store sequence to be changed to a load outside
6337 the loop, and a copy/store within the loop.
6339 Enabled by default when gcse is enabled.
6343 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6344 global common subexpression elimination. This pass will attempt to move
6345 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6346 loops containing a load/store sequence can be changed to a load before
6347 the loop and a store after the loop.
6349 Not enabled at any optimization level.
6353 When @option{-fgcse-las} is enabled, the global common subexpression
6354 elimination pass eliminates redundant loads that come after stores to the
6355 same memory location (both partial and full redundancies).
6357 Not enabled at any optimization level.
6359 @item -fgcse-after-reload
6360 @opindex fgcse-after-reload
6361 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6362 pass is performed after reload. The purpose of this pass is to cleanup
6365 @item -funsafe-loop-optimizations
6366 @opindex funsafe-loop-optimizations
6367 If given, the loop optimizer will assume that loop indices do not
6368 overflow, and that the loops with nontrivial exit condition are not
6369 infinite. This enables a wider range of loop optimizations even if
6370 the loop optimizer itself cannot prove that these assumptions are valid.
6371 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6372 if it finds this kind of loop.
6374 @item -fcrossjumping
6375 @opindex fcrossjumping
6376 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6377 resulting code may or may not perform better than without cross-jumping.
6379 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6381 @item -fauto-inc-dec
6382 @opindex fauto-inc-dec
6383 Combine increments or decrements of addresses with memory accesses.
6384 This pass is always skipped on architectures that do not have
6385 instructions to support this. Enabled by default at @option{-O} and
6386 higher on architectures that support this.
6390 Perform dead code elimination (DCE) on RTL@.
6391 Enabled by default at @option{-O} and higher.
6395 Perform dead store elimination (DSE) on RTL@.
6396 Enabled by default at @option{-O} and higher.
6398 @item -fif-conversion
6399 @opindex fif-conversion
6400 Attempt to transform conditional jumps into branch-less equivalents. This
6401 include use of conditional moves, min, max, set flags and abs instructions, and
6402 some tricks doable by standard arithmetics. The use of conditional execution
6403 on chips where it is available is controlled by @code{if-conversion2}.
6405 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6407 @item -fif-conversion2
6408 @opindex fif-conversion2
6409 Use conditional execution (where available) to transform conditional jumps into
6410 branch-less equivalents.
6412 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6414 @item -fdelete-null-pointer-checks
6415 @opindex fdelete-null-pointer-checks
6416 Assume that programs cannot safely dereference null pointers, and that
6417 no code or data element resides there. This enables simple constant
6418 folding optimizations at all optimization levels. In addition, other
6419 optimization passes in GCC use this flag to control global dataflow
6420 analyses that eliminate useless checks for null pointers; these assume
6421 that if a pointer is checked after it has already been dereferenced,
6424 Note however that in some environments this assumption is not true.
6425 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6426 for programs which depend on that behavior.
6428 Some targets, especially embedded ones, disable this option at all levels.
6429 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6430 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6431 are enabled independently at different optimization levels.
6433 @item -fdevirtualize
6434 @opindex fdevirtualize
6435 Attempt to convert calls to virtual functions to direct calls. This
6436 is done both within a procedure and interprocedurally as part of
6437 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6438 propagation (@option{-fipa-cp}).
6439 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6441 @item -fexpensive-optimizations
6442 @opindex fexpensive-optimizations
6443 Perform a number of minor optimizations that are relatively expensive.
6445 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6447 @item -foptimize-register-move
6449 @opindex foptimize-register-move
6451 Attempt to reassign register numbers in move instructions and as
6452 operands of other simple instructions in order to maximize the amount of
6453 register tying. This is especially helpful on machines with two-operand
6456 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6459 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6461 @item -fira-algorithm=@var{algorithm}
6462 Use specified coloring algorithm for the integrated register
6463 allocator. The @var{algorithm} argument should be @code{priority} or
6464 @code{CB}. The first algorithm specifies Chow's priority coloring,
6465 the second one specifies Chaitin-Briggs coloring. The second
6466 algorithm can be unimplemented for some architectures. If it is
6467 implemented, it is the default because Chaitin-Briggs coloring as a
6468 rule generates a better code.
6470 @item -fira-region=@var{region}
6471 Use specified regions for the integrated register allocator. The
6472 @var{region} argument should be one of @code{all}, @code{mixed}, or
6473 @code{one}. The first value means using all loops as register
6474 allocation regions, the second value which is the default means using
6475 all loops except for loops with small register pressure as the
6476 regions, and third one means using all function as a single region.
6477 The first value can give best result for machines with small size and
6478 irregular register set, the third one results in faster and generates
6479 decent code and the smallest size code, and the default value usually
6480 give the best results in most cases and for most architectures.
6482 @item -fira-loop-pressure
6483 @opindex fira-loop-pressure
6484 Use IRA to evaluate register pressure in loops for decision to move
6485 loop invariants. Usage of this option usually results in generation
6486 of faster and smaller code on machines with big register files (>= 32
6487 registers) but it can slow compiler down.
6489 This option is enabled at level @option{-O3} for some targets.
6491 @item -fno-ira-share-save-slots
6492 @opindex fno-ira-share-save-slots
6493 Switch off sharing stack slots used for saving call used hard
6494 registers living through a call. Each hard register will get a
6495 separate stack slot and as a result function stack frame will be
6498 @item -fno-ira-share-spill-slots
6499 @opindex fno-ira-share-spill-slots
6500 Switch off sharing stack slots allocated for pseudo-registers. Each
6501 pseudo-register which did not get a hard register will get a separate
6502 stack slot and as a result function stack frame will be bigger.
6504 @item -fira-verbose=@var{n}
6505 @opindex fira-verbose
6506 Set up how verbose dump file for the integrated register allocator
6507 will be. Default value is 5. If the value is greater or equal to 10,
6508 the dump file will be stderr as if the value were @var{n} minus 10.
6510 @item -fdelayed-branch
6511 @opindex fdelayed-branch
6512 If supported for the target machine, attempt to reorder instructions
6513 to exploit instruction slots available after delayed branch
6516 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6518 @item -fschedule-insns
6519 @opindex fschedule-insns
6520 If supported for the target machine, attempt to reorder instructions to
6521 eliminate execution stalls due to required data being unavailable. This
6522 helps machines that have slow floating point or memory load instructions
6523 by allowing other instructions to be issued until the result of the load
6524 or floating point instruction is required.
6526 Enabled at levels @option{-O2}, @option{-O3}.
6528 @item -fschedule-insns2
6529 @opindex fschedule-insns2
6530 Similar to @option{-fschedule-insns}, but requests an additional pass of
6531 instruction scheduling after register allocation has been done. This is
6532 especially useful on machines with a relatively small number of
6533 registers and where memory load instructions take more than one cycle.
6535 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6537 @item -fno-sched-interblock
6538 @opindex fno-sched-interblock
6539 Don't schedule instructions across basic blocks. This is normally
6540 enabled by default when scheduling before register allocation, i.e.@:
6541 with @option{-fschedule-insns} or at @option{-O2} or higher.
6543 @item -fno-sched-spec
6544 @opindex fno-sched-spec
6545 Don't allow speculative motion of non-load instructions. This is normally
6546 enabled by default when scheduling before register allocation, i.e.@:
6547 with @option{-fschedule-insns} or at @option{-O2} or higher.
6549 @item -fsched-pressure
6550 @opindex fsched-pressure
6551 Enable register pressure sensitive insn scheduling before the register
6552 allocation. This only makes sense when scheduling before register
6553 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6554 @option{-O2} or higher. Usage of this option can improve the
6555 generated code and decrease its size by preventing register pressure
6556 increase above the number of available hard registers and as a
6557 consequence register spills in the register allocation.
6559 @item -fsched-spec-load
6560 @opindex fsched-spec-load
6561 Allow speculative motion of some load instructions. This only makes
6562 sense when scheduling before register allocation, i.e.@: with
6563 @option{-fschedule-insns} or at @option{-O2} or higher.
6565 @item -fsched-spec-load-dangerous
6566 @opindex fsched-spec-load-dangerous
6567 Allow speculative motion of more load instructions. This only makes
6568 sense when scheduling before register allocation, i.e.@: with
6569 @option{-fschedule-insns} or at @option{-O2} or higher.
6571 @item -fsched-stalled-insns
6572 @itemx -fsched-stalled-insns=@var{n}
6573 @opindex fsched-stalled-insns
6574 Define how many insns (if any) can be moved prematurely from the queue
6575 of stalled insns into the ready list, during the second scheduling pass.
6576 @option{-fno-sched-stalled-insns} means that no insns will be moved
6577 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6578 on how many queued insns can be moved prematurely.
6579 @option{-fsched-stalled-insns} without a value is equivalent to
6580 @option{-fsched-stalled-insns=1}.
6582 @item -fsched-stalled-insns-dep
6583 @itemx -fsched-stalled-insns-dep=@var{n}
6584 @opindex fsched-stalled-insns-dep
6585 Define how many insn groups (cycles) will be examined for a dependency
6586 on a stalled insn that is candidate for premature removal from the queue
6587 of stalled insns. This has an effect only during the second scheduling pass,
6588 and only if @option{-fsched-stalled-insns} is used.
6589 @option{-fno-sched-stalled-insns-dep} is equivalent to
6590 @option{-fsched-stalled-insns-dep=0}.
6591 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6592 @option{-fsched-stalled-insns-dep=1}.
6594 @item -fsched2-use-superblocks
6595 @opindex fsched2-use-superblocks
6596 When scheduling after register allocation, do use superblock scheduling
6597 algorithm. Superblock scheduling allows motion across basic block boundaries
6598 resulting on faster schedules. This option is experimental, as not all machine
6599 descriptions used by GCC model the CPU closely enough to avoid unreliable
6600 results from the algorithm.
6602 This only makes sense when scheduling after register allocation, i.e.@: with
6603 @option{-fschedule-insns2} or at @option{-O2} or higher.
6605 @item -fsched-group-heuristic
6606 @opindex fsched-group-heuristic
6607 Enable the group heuristic in the scheduler. This heuristic favors
6608 the instruction that belongs to a schedule group. This is enabled
6609 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6610 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6612 @item -fsched-critical-path-heuristic
6613 @opindex fsched-critical-path-heuristic
6614 Enable the critical-path heuristic in the scheduler. This heuristic favors
6615 instructions on the critical path. This is enabled by default when
6616 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6617 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6619 @item -fsched-spec-insn-heuristic
6620 @opindex fsched-spec-insn-heuristic
6621 Enable the speculative instruction heuristic in the scheduler. This
6622 heuristic favors speculative instructions with greater dependency weakness.
6623 This is enabled by default when scheduling is enabled, i.e.@:
6624 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6625 or at @option{-O2} or higher.
6627 @item -fsched-rank-heuristic
6628 @opindex fsched-rank-heuristic
6629 Enable the rank heuristic in the scheduler. This heuristic favors
6630 the instruction belonging to a basic block with greater size or frequency.
6631 This is enabled by default when scheduling is enabled, i.e.@:
6632 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6633 at @option{-O2} or higher.
6635 @item -fsched-last-insn-heuristic
6636 @opindex fsched-last-insn-heuristic
6637 Enable the last-instruction heuristic in the scheduler. This heuristic
6638 favors the instruction that is less dependent on the last instruction
6639 scheduled. This is enabled by default when scheduling is enabled,
6640 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6641 at @option{-O2} or higher.
6643 @item -fsched-dep-count-heuristic
6644 @opindex fsched-dep-count-heuristic
6645 Enable the dependent-count heuristic in the scheduler. This heuristic
6646 favors the instruction that has more instructions depending on it.
6647 This is enabled by default when scheduling is enabled, i.e.@:
6648 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6649 at @option{-O2} or higher.
6651 @item -freschedule-modulo-scheduled-loops
6652 @opindex freschedule-modulo-scheduled-loops
6653 The modulo scheduling comes before the traditional scheduling, if a loop
6654 was modulo scheduled we may want to prevent the later scheduling passes
6655 from changing its schedule, we use this option to control that.
6657 @item -fselective-scheduling
6658 @opindex fselective-scheduling
6659 Schedule instructions using selective scheduling algorithm. Selective
6660 scheduling runs instead of the first scheduler pass.
6662 @item -fselective-scheduling2
6663 @opindex fselective-scheduling2
6664 Schedule instructions using selective scheduling algorithm. Selective
6665 scheduling runs instead of the second scheduler pass.
6667 @item -fsel-sched-pipelining
6668 @opindex fsel-sched-pipelining
6669 Enable software pipelining of innermost loops during selective scheduling.
6670 This option has no effect until one of @option{-fselective-scheduling} or
6671 @option{-fselective-scheduling2} is turned on.
6673 @item -fsel-sched-pipelining-outer-loops
6674 @opindex fsel-sched-pipelining-outer-loops
6675 When pipelining loops during selective scheduling, also pipeline outer loops.
6676 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6678 @item -fcaller-saves
6679 @opindex fcaller-saves
6680 Enable values to be allocated in registers that will be clobbered by
6681 function calls, by emitting extra instructions to save and restore the
6682 registers around such calls. Such allocation is done only when it
6683 seems to result in better code than would otherwise be produced.
6685 This option is always enabled by default on certain machines, usually
6686 those which have no call-preserved registers to use instead.
6688 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6690 @item -fcombine-stack-adjustments
6691 @opindex fcombine-stack-adjustments
6692 Tracks stack adjustments (pushes and pops) and stack memory references
6693 and then tries to find ways to combine them.
6695 Enabled by default at @option{-O1} and higher.
6697 @item -fconserve-stack
6698 @opindex fconserve-stack
6699 Attempt to minimize stack usage. The compiler will attempt to use less
6700 stack space, even if that makes the program slower. This option
6701 implies setting the @option{large-stack-frame} parameter to 100
6702 and the @option{large-stack-frame-growth} parameter to 400.
6704 @item -ftree-reassoc
6705 @opindex ftree-reassoc
6706 Perform reassociation on trees. This flag is enabled by default
6707 at @option{-O} and higher.
6711 Perform partial redundancy elimination (PRE) on trees. This flag is
6712 enabled by default at @option{-O2} and @option{-O3}.
6714 @item -ftree-forwprop
6715 @opindex ftree-forwprop
6716 Perform forward propagation on trees. This flag is enabled by default
6717 at @option{-O} and higher.
6721 Perform full redundancy elimination (FRE) on trees. The difference
6722 between FRE and PRE is that FRE only considers expressions
6723 that are computed on all paths leading to the redundant computation.
6724 This analysis is faster than PRE, though it exposes fewer redundancies.
6725 This flag is enabled by default at @option{-O} and higher.
6727 @item -ftree-phiprop
6728 @opindex ftree-phiprop
6729 Perform hoisting of loads from conditional pointers on trees. This
6730 pass is enabled by default at @option{-O} and higher.
6732 @item -ftree-copy-prop
6733 @opindex ftree-copy-prop
6734 Perform copy propagation on trees. This pass eliminates unnecessary
6735 copy operations. This flag is enabled by default at @option{-O} and
6738 @item -fipa-pure-const
6739 @opindex fipa-pure-const
6740 Discover which functions are pure or constant.
6741 Enabled by default at @option{-O} and higher.
6743 @item -fipa-reference
6744 @opindex fipa-reference
6745 Discover which static variables do not escape cannot escape the
6747 Enabled by default at @option{-O} and higher.
6749 @item -fipa-struct-reorg
6750 @opindex fipa-struct-reorg
6751 Perform structure reorganization optimization, that change C-like structures
6752 layout in order to better utilize spatial locality. This transformation is
6753 affective for programs containing arrays of structures. Available in two
6754 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6755 or static (which uses built-in heuristics). It works only in whole program
6756 mode, so it requires @option{-fwhole-program} to be
6757 enabled. Structures considered @samp{cold} by this transformation are not
6758 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6760 With this flag, the program debug info reflects a new structure layout.
6764 Perform interprocedural pointer analysis and interprocedural modification
6765 and reference analysis. This option can cause excessive memory and
6766 compile-time usage on large compilation units. It is not enabled by
6767 default at any optimization level.
6770 @opindex fipa-profile
6771 Perform interprocedural profile propagation. The functions called only from
6772 cold functions are marked as cold. Also functions executed once (such as
6773 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6774 functions and loop less parts of functions executed once are then optimized for
6776 Enabled by default at @option{-O} and higher.
6780 Perform interprocedural constant propagation.
6781 This optimization analyzes the program to determine when values passed
6782 to functions are constants and then optimizes accordingly.
6783 This optimization can substantially increase performance
6784 if the application has constants passed to functions.
6785 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6787 @item -fipa-cp-clone
6788 @opindex fipa-cp-clone
6789 Perform function cloning to make interprocedural constant propagation stronger.
6790 When enabled, interprocedural constant propagation will perform function cloning
6791 when externally visible function can be called with constant arguments.
6792 Because this optimization can create multiple copies of functions,
6793 it may significantly increase code size
6794 (see @option{--param ipcp-unit-growth=@var{value}}).
6795 This flag is enabled by default at @option{-O3}.
6797 @item -fipa-matrix-reorg
6798 @opindex fipa-matrix-reorg
6799 Perform matrix flattening and transposing.
6800 Matrix flattening tries to replace an @math{m}-dimensional matrix
6801 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6802 This reduces the level of indirection needed for accessing the elements
6803 of the matrix. The second optimization is matrix transposing that
6804 attempts to change the order of the matrix's dimensions in order to
6805 improve cache locality.
6806 Both optimizations need the @option{-fwhole-program} flag.
6807 Transposing is enabled only if profiling information is available.
6811 Perform forward store motion on trees. This flag is
6812 enabled by default at @option{-O} and higher.
6814 @item -ftree-bit-ccp
6815 @opindex ftree-bit-ccp
6816 Perform sparse conditional bit constant propagation on trees and propagate
6817 pointer alignment information.
6818 This pass only operates on local scalar variables and is enabled by default
6819 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
6823 Perform sparse conditional constant propagation (CCP) on trees. This
6824 pass only operates on local scalar variables and is enabled by default
6825 at @option{-O} and higher.
6827 @item -ftree-switch-conversion
6828 Perform conversion of simple initializations in a switch to
6829 initializations from a scalar array. This flag is enabled by default
6830 at @option{-O2} and higher.
6834 Perform dead code elimination (DCE) on trees. This flag is enabled by
6835 default at @option{-O} and higher.
6837 @item -ftree-builtin-call-dce
6838 @opindex ftree-builtin-call-dce
6839 Perform conditional dead code elimination (DCE) for calls to builtin functions
6840 that may set @code{errno} but are otherwise side-effect free. This flag is
6841 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6844 @item -ftree-dominator-opts
6845 @opindex ftree-dominator-opts
6846 Perform a variety of simple scalar cleanups (constant/copy
6847 propagation, redundancy elimination, range propagation and expression
6848 simplification) based on a dominator tree traversal. This also
6849 performs jump threading (to reduce jumps to jumps). This flag is
6850 enabled by default at @option{-O} and higher.
6854 Perform dead store elimination (DSE) on trees. A dead store is a store into
6855 a memory location which will later be overwritten by another store without
6856 any intervening loads. In this case the earlier store can be deleted. This
6857 flag is enabled by default at @option{-O} and higher.
6861 Perform loop header copying on trees. This is beneficial since it increases
6862 effectiveness of code motion optimizations. It also saves one jump. This flag
6863 is enabled by default at @option{-O} and higher. It is not enabled
6864 for @option{-Os}, since it usually increases code size.
6866 @item -ftree-loop-optimize
6867 @opindex ftree-loop-optimize
6868 Perform loop optimizations on trees. This flag is enabled by default
6869 at @option{-O} and higher.
6871 @item -ftree-loop-linear
6872 @opindex ftree-loop-linear
6873 Perform loop interchange transformations on tree. Same as
6874 @option{-floop-interchange}. To use this code transformation, GCC has
6875 to be configured with @option{--with-ppl} and @option{--with-cloog} to
6876 enable the Graphite loop transformation infrastructure.
6878 @item -floop-interchange
6879 @opindex floop-interchange
6880 Perform loop interchange transformations on loops. Interchanging two
6881 nested loops switches the inner and outer loops. For example, given a
6886 A(J, I) = A(J, I) * C
6890 loop interchange will transform the loop as if the user had written:
6894 A(J, I) = A(J, I) * C
6898 which can be beneficial when @code{N} is larger than the caches,
6899 because in Fortran, the elements of an array are stored in memory
6900 contiguously by column, and the original loop iterates over rows,
6901 potentially creating at each access a cache miss. This optimization
6902 applies to all the languages supported by GCC and is not limited to
6903 Fortran. To use this code transformation, GCC has to be configured
6904 with @option{--with-ppl} and @option{--with-cloog} to enable the
6905 Graphite loop transformation infrastructure.
6907 @item -floop-strip-mine
6908 @opindex floop-strip-mine
6909 Perform loop strip mining transformations on loops. Strip mining
6910 splits a loop into two nested loops. The outer loop has strides
6911 equal to the strip size and the inner loop has strides of the
6912 original loop within a strip. The strip length can be changed
6913 using the @option{loop-block-tile-size} parameter. For example,
6920 loop strip mining will transform the loop as if the user had written:
6923 DO I = II, min (II + 50, N)
6928 This optimization applies to all the languages supported by GCC and is
6929 not limited to Fortran. To use this code transformation, GCC has to
6930 be configured with @option{--with-ppl} and @option{--with-cloog} to
6931 enable the Graphite loop transformation infrastructure.
6934 @opindex floop-block
6935 Perform loop blocking transformations on loops. Blocking strip mines
6936 each loop in the loop nest such that the memory accesses of the
6937 element loops fit inside caches. The strip length can be changed
6938 using the @option{loop-block-tile-size} parameter. For example, given
6943 A(J, I) = B(I) + C(J)
6947 loop blocking will transform the loop as if the user had written:
6951 DO I = II, min (II + 50, N)
6952 DO J = JJ, min (JJ + 50, M)
6953 A(J, I) = B(I) + C(J)
6959 which can be beneficial when @code{M} is larger than the caches,
6960 because the innermost loop will iterate over a smaller amount of data
6961 that can be kept in the caches. This optimization applies to all the
6962 languages supported by GCC and is not limited to Fortran. To use this
6963 code transformation, GCC has to be configured with @option{--with-ppl}
6964 and @option{--with-cloog} to enable the Graphite loop transformation
6967 @item -fgraphite-identity
6968 @opindex fgraphite-identity
6969 Enable the identity transformation for graphite. For every SCoP we generate
6970 the polyhedral representation and transform it back to gimple. Using
6971 @option{-fgraphite-identity} we can check the costs or benefits of the
6972 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6973 are also performed by the code generator CLooG, like index splitting and
6974 dead code elimination in loops.
6976 @item -floop-flatten
6977 @opindex floop-flatten
6978 Removes the loop nesting structure: transforms the loop nest into a
6979 single loop. This transformation can be useful to vectorize all the
6980 levels of the loop nest.
6982 @item -floop-parallelize-all
6983 @opindex floop-parallelize-all
6984 Use the Graphite data dependence analysis to identify loops that can
6985 be parallelized. Parallelize all the loops that can be analyzed to
6986 not contain loop carried dependences without checking that it is
6987 profitable to parallelize the loops.
6989 @item -fcheck-data-deps
6990 @opindex fcheck-data-deps
6991 Compare the results of several data dependence analyzers. This option
6992 is used for debugging the data dependence analyzers.
6994 @item -ftree-loop-if-convert
6995 Attempt to transform conditional jumps in the innermost loops to
6996 branch-less equivalents. The intent is to remove control-flow from
6997 the innermost loops in order to improve the ability of the
6998 vectorization pass to handle these loops. This is enabled by default
6999 if vectorization is enabled.
7001 @item -ftree-loop-if-convert-stores
7002 Attempt to also if-convert conditional jumps containing memory writes.
7003 This transformation can be unsafe for multi-threaded programs as it
7004 transforms conditional memory writes into unconditional memory writes.
7007 for (i = 0; i < N; i++)
7011 would be transformed to
7013 for (i = 0; i < N; i++)
7014 A[i] = cond ? expr : A[i];
7016 potentially producing data races.
7018 @item -ftree-loop-distribution
7019 Perform loop distribution. This flag can improve cache performance on
7020 big loop bodies and allow further loop optimizations, like
7021 parallelization or vectorization, to take place. For example, the loop
7038 @item -ftree-loop-distribute-patterns
7039 Perform loop distribution of patterns that can be code generated with
7040 calls to a library. This flag is enabled by default at @option{-O3}.
7042 This pass distributes the initialization loops and generates a call to
7043 memset zero. For example, the loop
7059 and the initialization loop is transformed into a call to memset zero.
7061 @item -ftree-loop-im
7062 @opindex ftree-loop-im
7063 Perform loop invariant motion on trees. This pass moves only invariants that
7064 would be hard to handle at RTL level (function calls, operations that expand to
7065 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7066 operands of conditions that are invariant out of the loop, so that we can use
7067 just trivial invariantness analysis in loop unswitching. The pass also includes
7070 @item -ftree-loop-ivcanon
7071 @opindex ftree-loop-ivcanon
7072 Create a canonical counter for number of iterations in the loop for that
7073 determining number of iterations requires complicated analysis. Later
7074 optimizations then may determine the number easily. Useful especially
7075 in connection with unrolling.
7079 Perform induction variable optimizations (strength reduction, induction
7080 variable merging and induction variable elimination) on trees.
7082 @item -ftree-parallelize-loops=n
7083 @opindex ftree-parallelize-loops
7084 Parallelize loops, i.e., split their iteration space to run in n threads.
7085 This is only possible for loops whose iterations are independent
7086 and can be arbitrarily reordered. The optimization is only
7087 profitable on multiprocessor machines, for loops that are CPU-intensive,
7088 rather than constrained e.g.@: by memory bandwidth. This option
7089 implies @option{-pthread}, and thus is only supported on targets
7090 that have support for @option{-pthread}.
7094 Perform function-local points-to analysis on trees. This flag is
7095 enabled by default at @option{-O} and higher.
7099 Perform scalar replacement of aggregates. This pass replaces structure
7100 references with scalars to prevent committing structures to memory too
7101 early. This flag is enabled by default at @option{-O} and higher.
7103 @item -ftree-copyrename
7104 @opindex ftree-copyrename
7105 Perform copy renaming on trees. This pass attempts to rename compiler
7106 temporaries to other variables at copy locations, usually resulting in
7107 variable names which more closely resemble the original variables. This flag
7108 is enabled by default at @option{-O} and higher.
7112 Perform temporary expression replacement during the SSA->normal phase. Single
7113 use/single def temporaries are replaced at their use location with their
7114 defining expression. This results in non-GIMPLE code, but gives the expanders
7115 much more complex trees to work on resulting in better RTL generation. This is
7116 enabled by default at @option{-O} and higher.
7118 @item -ftree-vectorize
7119 @opindex ftree-vectorize
7120 Perform loop vectorization on trees. This flag is enabled by default at
7123 @item -ftree-slp-vectorize
7124 @opindex ftree-slp-vectorize
7125 Perform basic block vectorization on trees. This flag is enabled by default at
7126 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7128 @item -ftree-vect-loop-version
7129 @opindex ftree-vect-loop-version
7130 Perform loop versioning when doing loop vectorization on trees. When a loop
7131 appears to be vectorizable except that data alignment or data dependence cannot
7132 be determined at compile time then vectorized and non-vectorized versions of
7133 the loop are generated along with runtime checks for alignment or dependence
7134 to control which version is executed. This option is enabled by default
7135 except at level @option{-Os} where it is disabled.
7137 @item -fvect-cost-model
7138 @opindex fvect-cost-model
7139 Enable cost model for vectorization.
7143 Perform Value Range Propagation on trees. This is similar to the
7144 constant propagation pass, but instead of values, ranges of values are
7145 propagated. This allows the optimizers to remove unnecessary range
7146 checks like array bound checks and null pointer checks. This is
7147 enabled by default at @option{-O2} and higher. Null pointer check
7148 elimination is only done if @option{-fdelete-null-pointer-checks} is
7153 Perform tail duplication to enlarge superblock size. This transformation
7154 simplifies the control flow of the function allowing other optimizations to do
7157 @item -funroll-loops
7158 @opindex funroll-loops
7159 Unroll loops whose number of iterations can be determined at compile
7160 time or upon entry to the loop. @option{-funroll-loops} implies
7161 @option{-frerun-cse-after-loop}. This option makes code larger,
7162 and may or may not make it run faster.
7164 @item -funroll-all-loops
7165 @opindex funroll-all-loops
7166 Unroll all loops, even if their number of iterations is uncertain when
7167 the loop is entered. This usually makes programs run more slowly.
7168 @option{-funroll-all-loops} implies the same options as
7169 @option{-funroll-loops},
7171 @item -fsplit-ivs-in-unroller
7172 @opindex fsplit-ivs-in-unroller
7173 Enables expressing of values of induction variables in later iterations
7174 of the unrolled loop using the value in the first iteration. This breaks
7175 long dependency chains, thus improving efficiency of the scheduling passes.
7177 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7178 same effect. However in cases the loop body is more complicated than
7179 a single basic block, this is not reliable. It also does not work at all
7180 on some of the architectures due to restrictions in the CSE pass.
7182 This optimization is enabled by default.
7184 @item -fvariable-expansion-in-unroller
7185 @opindex fvariable-expansion-in-unroller
7186 With this option, the compiler will create multiple copies of some
7187 local variables when unrolling a loop which can result in superior code.
7189 @item -fpartial-inlining
7190 @opindex fpartial-inlining
7191 Inline parts of functions. This option has any effect only
7192 when inlining itself is turned on by the @option{-finline-functions}
7193 or @option{-finline-small-functions} options.
7195 Enabled at level @option{-O2}.
7197 @item -fpredictive-commoning
7198 @opindex fpredictive-commoning
7199 Perform predictive commoning optimization, i.e., reusing computations
7200 (especially memory loads and stores) performed in previous
7201 iterations of loops.
7203 This option is enabled at level @option{-O3}.
7205 @item -fprefetch-loop-arrays
7206 @opindex fprefetch-loop-arrays
7207 If supported by the target machine, generate instructions to prefetch
7208 memory to improve the performance of loops that access large arrays.
7210 This option may generate better or worse code; results are highly
7211 dependent on the structure of loops within the source code.
7213 Disabled at level @option{-Os}.
7216 @itemx -fno-peephole2
7217 @opindex fno-peephole
7218 @opindex fno-peephole2
7219 Disable any machine-specific peephole optimizations. The difference
7220 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7221 are implemented in the compiler; some targets use one, some use the
7222 other, a few use both.
7224 @option{-fpeephole} is enabled by default.
7225 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7227 @item -fno-guess-branch-probability
7228 @opindex fno-guess-branch-probability
7229 Do not guess branch probabilities using heuristics.
7231 GCC will use heuristics to guess branch probabilities if they are
7232 not provided by profiling feedback (@option{-fprofile-arcs}). These
7233 heuristics are based on the control flow graph. If some branch probabilities
7234 are specified by @samp{__builtin_expect}, then the heuristics will be
7235 used to guess branch probabilities for the rest of the control flow graph,
7236 taking the @samp{__builtin_expect} info into account. The interactions
7237 between the heuristics and @samp{__builtin_expect} can be complex, and in
7238 some cases, it may be useful to disable the heuristics so that the effects
7239 of @samp{__builtin_expect} are easier to understand.
7241 The default is @option{-fguess-branch-probability} at levels
7242 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7244 @item -freorder-blocks
7245 @opindex freorder-blocks
7246 Reorder basic blocks in the compiled function in order to reduce number of
7247 taken branches and improve code locality.
7249 Enabled at levels @option{-O2}, @option{-O3}.
7251 @item -freorder-blocks-and-partition
7252 @opindex freorder-blocks-and-partition
7253 In addition to reordering basic blocks in the compiled function, in order
7254 to reduce number of taken branches, partitions hot and cold basic blocks
7255 into separate sections of the assembly and .o files, to improve
7256 paging and cache locality performance.
7258 This optimization is automatically turned off in the presence of
7259 exception handling, for linkonce sections, for functions with a user-defined
7260 section attribute and on any architecture that does not support named
7263 @item -freorder-functions
7264 @opindex freorder-functions
7265 Reorder functions in the object file in order to
7266 improve code locality. This is implemented by using special
7267 subsections @code{.text.hot} for most frequently executed functions and
7268 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7269 the linker so object file format must support named sections and linker must
7270 place them in a reasonable way.
7272 Also profile feedback must be available in to make this option effective. See
7273 @option{-fprofile-arcs} for details.
7275 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7277 @item -fstrict-aliasing
7278 @opindex fstrict-aliasing
7279 Allow the compiler to assume the strictest aliasing rules applicable to
7280 the language being compiled. For C (and C++), this activates
7281 optimizations based on the type of expressions. In particular, an
7282 object of one type is assumed never to reside at the same address as an
7283 object of a different type, unless the types are almost the same. For
7284 example, an @code{unsigned int} can alias an @code{int}, but not a
7285 @code{void*} or a @code{double}. A character type may alias any other
7288 @anchor{Type-punning}Pay special attention to code like this:
7301 The practice of reading from a different union member than the one most
7302 recently written to (called ``type-punning'') is common. Even with
7303 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7304 is accessed through the union type. So, the code above will work as
7305 expected. @xref{Structures unions enumerations and bit-fields
7306 implementation}. However, this code might not:
7317 Similarly, access by taking the address, casting the resulting pointer
7318 and dereferencing the result has undefined behavior, even if the cast
7319 uses a union type, e.g.:
7323 return ((union a_union *) &d)->i;
7327 The @option{-fstrict-aliasing} option is enabled at levels
7328 @option{-O2}, @option{-O3}, @option{-Os}.
7330 @item -fstrict-overflow
7331 @opindex fstrict-overflow
7332 Allow the compiler to assume strict signed overflow rules, depending
7333 on the language being compiled. For C (and C++) this means that
7334 overflow when doing arithmetic with signed numbers is undefined, which
7335 means that the compiler may assume that it will not happen. This
7336 permits various optimizations. For example, the compiler will assume
7337 that an expression like @code{i + 10 > i} will always be true for
7338 signed @code{i}. This assumption is only valid if signed overflow is
7339 undefined, as the expression is false if @code{i + 10} overflows when
7340 using twos complement arithmetic. When this option is in effect any
7341 attempt to determine whether an operation on signed numbers will
7342 overflow must be written carefully to not actually involve overflow.
7344 This option also allows the compiler to assume strict pointer
7345 semantics: given a pointer to an object, if adding an offset to that
7346 pointer does not produce a pointer to the same object, the addition is
7347 undefined. This permits the compiler to conclude that @code{p + u >
7348 p} is always true for a pointer @code{p} and unsigned integer
7349 @code{u}. This assumption is only valid because pointer wraparound is
7350 undefined, as the expression is false if @code{p + u} overflows using
7351 twos complement arithmetic.
7353 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7354 that integer signed overflow is fully defined: it wraps. When
7355 @option{-fwrapv} is used, there is no difference between
7356 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7357 integers. With @option{-fwrapv} certain types of overflow are
7358 permitted. For example, if the compiler gets an overflow when doing
7359 arithmetic on constants, the overflowed value can still be used with
7360 @option{-fwrapv}, but not otherwise.
7362 The @option{-fstrict-overflow} option is enabled at levels
7363 @option{-O2}, @option{-O3}, @option{-Os}.
7365 @item -falign-functions
7366 @itemx -falign-functions=@var{n}
7367 @opindex falign-functions
7368 Align the start of functions to the next power-of-two greater than
7369 @var{n}, skipping up to @var{n} bytes. For instance,
7370 @option{-falign-functions=32} aligns functions to the next 32-byte
7371 boundary, but @option{-falign-functions=24} would align to the next
7372 32-byte boundary only if this can be done by skipping 23 bytes or less.
7374 @option{-fno-align-functions} and @option{-falign-functions=1} are
7375 equivalent and mean that functions will not be aligned.
7377 Some assemblers only support this flag when @var{n} is a power of two;
7378 in that case, it is rounded up.
7380 If @var{n} is not specified or is zero, use a machine-dependent default.
7382 Enabled at levels @option{-O2}, @option{-O3}.
7384 @item -falign-labels
7385 @itemx -falign-labels=@var{n}
7386 @opindex falign-labels
7387 Align all branch targets to a power-of-two boundary, skipping up to
7388 @var{n} bytes like @option{-falign-functions}. This option can easily
7389 make code slower, because it must insert dummy operations for when the
7390 branch target is reached in the usual flow of the code.
7392 @option{-fno-align-labels} and @option{-falign-labels=1} are
7393 equivalent and mean that labels will not be aligned.
7395 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7396 are greater than this value, then their values are used instead.
7398 If @var{n} is not specified or is zero, use a machine-dependent default
7399 which is very likely to be @samp{1}, meaning no alignment.
7401 Enabled at levels @option{-O2}, @option{-O3}.
7404 @itemx -falign-loops=@var{n}
7405 @opindex falign-loops
7406 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7407 like @option{-falign-functions}. The hope is that the loop will be
7408 executed many times, which will make up for any execution of the dummy
7411 @option{-fno-align-loops} and @option{-falign-loops=1} are
7412 equivalent and mean that loops will not be aligned.
7414 If @var{n} is not specified or is zero, use a machine-dependent default.
7416 Enabled at levels @option{-O2}, @option{-O3}.
7419 @itemx -falign-jumps=@var{n}
7420 @opindex falign-jumps
7421 Align branch targets to a power-of-two boundary, for branch targets
7422 where the targets can only be reached by jumping, skipping up to @var{n}
7423 bytes like @option{-falign-functions}. In this case, no dummy operations
7426 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7427 equivalent and mean that loops will not be aligned.
7429 If @var{n} is not specified or is zero, use a machine-dependent default.
7431 Enabled at levels @option{-O2}, @option{-O3}.
7433 @item -funit-at-a-time
7434 @opindex funit-at-a-time
7435 This option is left for compatibility reasons. @option{-funit-at-a-time}
7436 has no effect, while @option{-fno-unit-at-a-time} implies
7437 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7441 @item -fno-toplevel-reorder
7442 @opindex fno-toplevel-reorder
7443 Do not reorder top-level functions, variables, and @code{asm}
7444 statements. Output them in the same order that they appear in the
7445 input file. When this option is used, unreferenced static variables
7446 will not be removed. This option is intended to support existing code
7447 which relies on a particular ordering. For new code, it is better to
7450 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7451 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7456 Constructs webs as commonly used for register allocation purposes and assign
7457 each web individual pseudo register. This allows the register allocation pass
7458 to operate on pseudos directly, but also strengthens several other optimization
7459 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7460 however, make debugging impossible, since variables will no longer stay in a
7463 Enabled by default with @option{-funroll-loops}.
7465 @item -fwhole-program
7466 @opindex fwhole-program
7467 Assume that the current compilation unit represents the whole program being
7468 compiled. All public functions and variables with the exception of @code{main}
7469 and those merged by attribute @code{externally_visible} become static functions
7470 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.
7471 While this option is equivalent to proper use of the @code{static} keyword for
7472 programs consisting of a single file, in combination with option
7473 @option{-flto} this flag can be used to
7474 compile many smaller scale programs since the functions and variables become
7475 local for the whole combined compilation unit, not for the single source file
7478 This option implies @option{-fwhole-file} for Fortran programs.
7480 @item -flto[=@var{n}]
7482 This option runs the standard link-time optimizer. When invoked
7483 with source code, it generates GIMPLE (one of GCC's internal
7484 representations) and writes it to special ELF sections in the object
7485 file. When the object files are linked together, all the function
7486 bodies are read from these ELF sections and instantiated as if they
7487 had been part of the same translation unit.
7489 To use the link-timer optimizer, @option{-flto} needs to be specified at
7490 compile time and during the final link. For example,
7493 gcc -c -O2 -flto foo.c
7494 gcc -c -O2 -flto bar.c
7495 gcc -o myprog -flto -O2 foo.o bar.o
7498 The first two invocations to GCC will save a bytecode representation
7499 of GIMPLE into special ELF sections inside @file{foo.o} and
7500 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7501 @file{foo.o} and @file{bar.o}, merge the two files into a single
7502 internal image, and compile the result as usual. Since both
7503 @file{foo.o} and @file{bar.o} are merged into a single image, this
7504 causes all the inter-procedural analyses and optimizations in GCC to
7505 work across the two files as if they were a single one. This means,
7506 for example, that the inliner will be able to inline functions in
7507 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7509 Another (simpler) way to enable link-time optimization is,
7512 gcc -o myprog -flto -O2 foo.c bar.c
7515 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7516 merge them together into a single GIMPLE representation and optimize
7517 them as usual to produce @file{myprog}.
7519 The only important thing to keep in mind is that to enable link-time
7520 optimizations the @option{-flto} flag needs to be passed to both the
7521 compile and the link commands.
7523 To make whole program optimization effective, it is necessary to make
7524 certain whole program assumptions. The compiler needs to know
7525 what functions and variables can be accessed by libraries and runtime
7526 outside of the link time optimized unit. When supported by the linker,
7527 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7528 compiler information about used and externally visible symbols. When
7529 the linker plugin is not available, @option{-fwhole-program} should be
7530 used to allow the compiler to make these assumptions, which will lead
7531 to more aggressive optimization decisions.
7533 Note that when a file is compiled with @option{-flto}, the generated
7534 object file will be larger than a regular object file because it will
7535 contain GIMPLE bytecodes and the usual final code. This means that
7536 object files with LTO information can be linked as a normal object
7537 file. So, in the previous example, if the final link is done with
7540 gcc -o myprog foo.o bar.o
7543 The only difference will be that no inter-procedural optimizations
7544 will be applied to produce @file{myprog}. The two object files
7545 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7548 Additionally, the optimization flags used to compile individual files
7549 are not necessarily related to those used at link-time. For instance,
7552 gcc -c -O0 -flto foo.c
7553 gcc -c -O0 -flto bar.c
7554 gcc -o myprog -flto -O3 foo.o bar.o
7557 This will produce individual object files with unoptimized assembler
7558 code, but the resulting binary @file{myprog} will be optimized at
7559 @option{-O3}. Now, if the final binary is generated without
7560 @option{-flto}, then @file{myprog} will not be optimized.
7562 When producing the final binary with @option{-flto}, GCC will only
7563 apply link-time optimizations to those files that contain bytecode.
7564 Therefore, you can mix and match object files and libraries with
7565 GIMPLE bytecodes and final object code. GCC will automatically select
7566 which files to optimize in LTO mode and which files to link without
7569 There are some code generation flags that GCC will preserve when
7570 generating bytecodes, as they need to be used during the final link
7571 stage. Currently, the following options are saved into the GIMPLE
7572 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7573 @option{-m} target flags.
7575 At link time, these options are read-in and reapplied. Note that the
7576 current implementation makes no attempt at recognizing conflicting
7577 values for these options. If two or more files have a conflicting
7578 value (e.g., one file is compiled with @option{-fPIC} and another
7579 isn't), the compiler will simply use the last value read from the
7580 bytecode files. It is recommended, then, that all the files
7581 participating in the same link be compiled with the same options.
7583 Another feature of LTO is that it is possible to apply interprocedural
7584 optimizations on files written in different languages. This requires
7585 some support in the language front end. Currently, the C, C++ and
7586 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7587 something like this should work
7592 gfortran -c -flto baz.f90
7593 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7596 Notice that the final link is done with @command{g++} to get the C++
7597 runtime libraries and @option{-lgfortran} is added to get the Fortran
7598 runtime libraries. In general, when mixing languages in LTO mode, you
7599 should use the same link command used when mixing languages in a
7600 regular (non-LTO) compilation. This means that if your build process
7601 was mixing languages before, all you need to add is @option{-flto} to
7602 all the compile and link commands.
7604 If LTO encounters objects with C linkage declared with incompatible
7605 types in separate translation units to be linked together (undefined
7606 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7607 issued. The behavior is still undefined at runtime.
7609 If object files containing GIMPLE bytecode are stored in a library archive, say
7610 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7611 are using a linker with linker plugin support. To enable this feature, use
7612 the flag @option{-fuse-linker-plugin} at link-time:
7615 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7618 With the linker plugin enabled, the linker will extract the needed
7619 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7620 to make them part of the aggregated GIMPLE image to be optimized.
7622 If you are not using a linker with linker plugin support and/or do not
7623 enable linker plugin then the objects inside @file{libfoo.a}
7624 will be extracted and linked as usual, but they will not participate
7625 in the LTO optimization process.
7627 Link time optimizations do not require the presence of the whole program to
7628 operate. If the program does not require any symbols to be exported, it is
7629 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7630 the interprocedural optimizers to use more aggressive assumptions which may
7631 lead to improved optimization opportunities.
7632 Use of @option{-fwhole-program} is not needed when linker plugin is
7633 active (see @option{-fuse-linker-plugin}).
7635 Regarding portability: the current implementation of LTO makes no
7636 attempt at generating bytecode that can be ported between different
7637 types of hosts. The bytecode files are versioned and there is a
7638 strict version check, so bytecode files generated in one version of
7639 GCC will not work with an older/newer version of GCC.
7641 Link time optimization does not play well with generating debugging
7642 information. Combining @option{-flto} with
7643 @option{-g} is currently experimental and expected to produce wrong
7646 If you specify the optional @var{n}, the optimization and code
7647 generation done at link time is executed in parallel using @var{n}
7648 parallel jobs by utilizing an installed @command{make} program. The
7649 environment variable @env{MAKE} may be used to override the program
7650 used. The default value for @var{n} is 1.
7652 You can also specify @option{-flto=jobserver} to use GNU make's
7653 job server mode to determine the number of parallel jobs. This
7654 is useful when the Makefile calling GCC is already executing in parallel.
7655 The parent Makefile will need a @samp{+} prepended to the command recipe
7656 for this to work. This will likely only work if @env{MAKE} is
7659 This option is disabled by default.
7661 @item -flto-partition=@var{alg}
7662 @opindex flto-partition
7663 Specify the partitioning algorithm used by the link time optimizer.
7664 The value is either @code{1to1} to specify a partitioning mirroring
7665 the original source files or @code{balanced} to specify partitioning
7666 into equally sized chunks (whenever possible). Specifying @code{none}
7667 as an algorithm disables partitioning and streaming completely. The
7668 default value is @code{balanced}.
7670 @item -flto-compression-level=@var{n}
7671 This option specifies the level of compression used for intermediate
7672 language written to LTO object files, and is only meaningful in
7673 conjunction with LTO mode (@option{-flto}). Valid
7674 values are 0 (no compression) to 9 (maximum compression). Values
7675 outside this range are clamped to either 0 or 9. If the option is not
7676 given, a default balanced compression setting is used.
7679 Prints a report with internal details on the workings of the link-time
7680 optimizer. The contents of this report vary from version to version,
7681 it is meant to be useful to GCC developers when processing object
7682 files in LTO mode (via @option{-flto}).
7684 Disabled by default.
7686 @item -fuse-linker-plugin
7687 Enables the use of linker plugin during link time optimization. This option
7688 relies on the linker plugin support in linker that is available in @code{gold}
7689 or in GNU ld 2.21.51 or newer..
7691 This option enables the extraction of object files with GIMPLE bytecode out of
7692 library archives. This improves the quality of optimization by exposing more
7693 code the the link time optimizer. This information specify what symbols
7694 can be accessed externally (by non-LTO object or during dynamic linking).
7695 Resulting code quality improvements on binaries (and shared libraries that do
7696 use hidden visibility) is similar to @code{-fwhole-program}. See
7697 @option{-flto} for a description on the effect of this flag and how to use it.
7699 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7700 with linker supporting plugins (GNU ld or @code{gold}).
7702 @item -fcompare-elim
7703 @opindex fcompare-elim
7704 After register allocation and post-register allocation instruction splitting,
7705 identify arithmetic instructions that compute processor flags similar to a
7706 comparison operation based on that arithmetic. If possible, eliminate the
7707 explicit comparison operation.
7709 This pass only applies to certain targets that cannot explicitly represent
7710 the comparison operation before register allocation is complete.
7712 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7714 @item -fcprop-registers
7715 @opindex fcprop-registers
7716 After register allocation and post-register allocation instruction splitting,
7717 we perform a copy-propagation pass to try to reduce scheduling dependencies
7718 and occasionally eliminate the copy.
7720 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7722 @item -fprofile-correction
7723 @opindex fprofile-correction
7724 Profiles collected using an instrumented binary for multi-threaded programs may
7725 be inconsistent due to missed counter updates. When this option is specified,
7726 GCC will use heuristics to correct or smooth out such inconsistencies. By
7727 default, GCC will emit an error message when an inconsistent profile is detected.
7729 @item -fprofile-dir=@var{path}
7730 @opindex fprofile-dir
7732 Set the directory to search for the profile data files in to @var{path}.
7733 This option affects only the profile data generated by
7734 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7735 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7736 and its related options.
7737 By default, GCC will use the current directory as @var{path}, thus the
7738 profile data file will appear in the same directory as the object file.
7740 @item -fprofile-generate
7741 @itemx -fprofile-generate=@var{path}
7742 @opindex fprofile-generate
7744 Enable options usually used for instrumenting application to produce
7745 profile useful for later recompilation with profile feedback based
7746 optimization. You must use @option{-fprofile-generate} both when
7747 compiling and when linking your program.
7749 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7751 If @var{path} is specified, GCC will look at the @var{path} to find
7752 the profile feedback data files. See @option{-fprofile-dir}.
7755 @itemx -fprofile-use=@var{path}
7756 @opindex fprofile-use
7757 Enable profile feedback directed optimizations, and optimizations
7758 generally profitable only with profile feedback available.
7760 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7761 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7763 By default, GCC emits an error message if the feedback profiles do not
7764 match the source code. This error can be turned into a warning by using
7765 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7768 If @var{path} is specified, GCC will look at the @var{path} to find
7769 the profile feedback data files. See @option{-fprofile-dir}.
7772 The following options control compiler behavior regarding floating
7773 point arithmetic. These options trade off between speed and
7774 correctness. All must be specifically enabled.
7778 @opindex ffloat-store
7779 Do not store floating point variables in registers, and inhibit other
7780 options that might change whether a floating point value is taken from a
7783 @cindex floating point precision
7784 This option prevents undesirable excess precision on machines such as
7785 the 68000 where the floating registers (of the 68881) keep more
7786 precision than a @code{double} is supposed to have. Similarly for the
7787 x86 architecture. For most programs, the excess precision does only
7788 good, but a few programs rely on the precise definition of IEEE floating
7789 point. Use @option{-ffloat-store} for such programs, after modifying
7790 them to store all pertinent intermediate computations into variables.
7792 @item -fexcess-precision=@var{style}
7793 @opindex fexcess-precision
7794 This option allows further control over excess precision on machines
7795 where floating-point registers have more precision than the IEEE
7796 @code{float} and @code{double} types and the processor does not
7797 support operations rounding to those types. By default,
7798 @option{-fexcess-precision=fast} is in effect; this means that
7799 operations are carried out in the precision of the registers and that
7800 it is unpredictable when rounding to the types specified in the source
7801 code takes place. When compiling C, if
7802 @option{-fexcess-precision=standard} is specified then excess
7803 precision will follow the rules specified in ISO C99; in particular,
7804 both casts and assignments cause values to be rounded to their
7805 semantic types (whereas @option{-ffloat-store} only affects
7806 assignments). This option is enabled by default for C if a strict
7807 conformance option such as @option{-std=c99} is used.
7810 @option{-fexcess-precision=standard} is not implemented for languages
7811 other than C, and has no effect if
7812 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7813 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7814 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7815 semantics apply without excess precision, and in the latter, rounding
7820 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7821 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7822 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7824 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7826 This option is not turned on by any @option{-O} option besides
7827 @option{-Ofast} since it can result in incorrect output for programs
7828 which depend on an exact implementation of IEEE or ISO rules/specifications
7829 for math functions. It may, however, yield faster code for programs
7830 that do not require the guarantees of these specifications.
7832 @item -fno-math-errno
7833 @opindex fno-math-errno
7834 Do not set ERRNO after calling math functions that are executed
7835 with a single instruction, e.g., sqrt. A program that relies on
7836 IEEE exceptions for math error handling may want to use this flag
7837 for speed while maintaining IEEE arithmetic compatibility.
7839 This option is not turned on by any @option{-O} option since
7840 it can result in incorrect output for programs which depend on
7841 an exact implementation of IEEE or ISO rules/specifications for
7842 math functions. It may, however, yield faster code for programs
7843 that do not require the guarantees of these specifications.
7845 The default is @option{-fmath-errno}.
7847 On Darwin systems, the math library never sets @code{errno}. There is
7848 therefore no reason for the compiler to consider the possibility that
7849 it might, and @option{-fno-math-errno} is the default.
7851 @item -funsafe-math-optimizations
7852 @opindex funsafe-math-optimizations
7854 Allow optimizations for floating-point arithmetic that (a) assume
7855 that arguments and results are valid and (b) may violate IEEE or
7856 ANSI standards. When used at link-time, it may include libraries
7857 or startup files that change the default FPU control word or other
7858 similar optimizations.
7860 This option is not turned on by any @option{-O} option since
7861 it can result in incorrect output for programs which depend on
7862 an exact implementation of IEEE or ISO rules/specifications for
7863 math functions. It may, however, yield faster code for programs
7864 that do not require the guarantees of these specifications.
7865 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7866 @option{-fassociative-math} and @option{-freciprocal-math}.
7868 The default is @option{-fno-unsafe-math-optimizations}.
7870 @item -fassociative-math
7871 @opindex fassociative-math
7873 Allow re-association of operands in series of floating-point operations.
7874 This violates the ISO C and C++ language standard by possibly changing
7875 computation result. NOTE: re-ordering may change the sign of zero as
7876 well as ignore NaNs and inhibit or create underflow or overflow (and
7877 thus cannot be used on a code which relies on rounding behavior like
7878 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7879 and thus may not be used when ordered comparisons are required.
7880 This option requires that both @option{-fno-signed-zeros} and
7881 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7882 much sense with @option{-frounding-math}. For Fortran the option
7883 is automatically enabled when both @option{-fno-signed-zeros} and
7884 @option{-fno-trapping-math} are in effect.
7886 The default is @option{-fno-associative-math}.
7888 @item -freciprocal-math
7889 @opindex freciprocal-math
7891 Allow the reciprocal of a value to be used instead of dividing by
7892 the value if this enables optimizations. For example @code{x / y}
7893 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7894 is subject to common subexpression elimination. Note that this loses
7895 precision and increases the number of flops operating on the value.
7897 The default is @option{-fno-reciprocal-math}.
7899 @item -ffinite-math-only
7900 @opindex ffinite-math-only
7901 Allow optimizations for floating-point arithmetic that assume
7902 that arguments and results are not NaNs or +-Infs.
7904 This option is not turned on by any @option{-O} option since
7905 it can result in incorrect output for programs which depend on
7906 an exact implementation of IEEE or ISO rules/specifications for
7907 math functions. It may, however, yield faster code for programs
7908 that do not require the guarantees of these specifications.
7910 The default is @option{-fno-finite-math-only}.
7912 @item -fno-signed-zeros
7913 @opindex fno-signed-zeros
7914 Allow optimizations for floating point arithmetic that ignore the
7915 signedness of zero. IEEE arithmetic specifies the behavior of
7916 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7917 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7918 This option implies that the sign of a zero result isn't significant.
7920 The default is @option{-fsigned-zeros}.
7922 @item -fno-trapping-math
7923 @opindex fno-trapping-math
7924 Compile code assuming that floating-point operations cannot generate
7925 user-visible traps. These traps include division by zero, overflow,
7926 underflow, inexact result and invalid operation. This option requires
7927 that @option{-fno-signaling-nans} be in effect. Setting this option may
7928 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7930 This option should never be turned on by any @option{-O} option since
7931 it can result in incorrect output for programs which depend on
7932 an exact implementation of IEEE or ISO rules/specifications for
7935 The default is @option{-ftrapping-math}.
7937 @item -frounding-math
7938 @opindex frounding-math
7939 Disable transformations and optimizations that assume default floating
7940 point rounding behavior. This is round-to-zero for all floating point
7941 to integer conversions, and round-to-nearest for all other arithmetic
7942 truncations. This option should be specified for programs that change
7943 the FP rounding mode dynamically, or that may be executed with a
7944 non-default rounding mode. This option disables constant folding of
7945 floating point expressions at compile-time (which may be affected by
7946 rounding mode) and arithmetic transformations that are unsafe in the
7947 presence of sign-dependent rounding modes.
7949 The default is @option{-fno-rounding-math}.
7951 This option is experimental and does not currently guarantee to
7952 disable all GCC optimizations that are affected by rounding mode.
7953 Future versions of GCC may provide finer control of this setting
7954 using C99's @code{FENV_ACCESS} pragma. This command line option
7955 will be used to specify the default state for @code{FENV_ACCESS}.
7957 @item -fsignaling-nans
7958 @opindex fsignaling-nans
7959 Compile code assuming that IEEE signaling NaNs may generate user-visible
7960 traps during floating-point operations. Setting this option disables
7961 optimizations that may change the number of exceptions visible with
7962 signaling NaNs. This option implies @option{-ftrapping-math}.
7964 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7967 The default is @option{-fno-signaling-nans}.
7969 This option is experimental and does not currently guarantee to
7970 disable all GCC optimizations that affect signaling NaN behavior.
7972 @item -fsingle-precision-constant
7973 @opindex fsingle-precision-constant
7974 Treat floating point constant as single precision constant instead of
7975 implicitly converting it to double precision constant.
7977 @item -fcx-limited-range
7978 @opindex fcx-limited-range
7979 When enabled, this option states that a range reduction step is not
7980 needed when performing complex division. Also, there is no checking
7981 whether the result of a complex multiplication or division is @code{NaN
7982 + I*NaN}, with an attempt to rescue the situation in that case. The
7983 default is @option{-fno-cx-limited-range}, but is enabled by
7984 @option{-ffast-math}.
7986 This option controls the default setting of the ISO C99
7987 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7990 @item -fcx-fortran-rules
7991 @opindex fcx-fortran-rules
7992 Complex multiplication and division follow Fortran rules. Range
7993 reduction is done as part of complex division, but there is no checking
7994 whether the result of a complex multiplication or division is @code{NaN
7995 + I*NaN}, with an attempt to rescue the situation in that case.
7997 The default is @option{-fno-cx-fortran-rules}.
8001 The following options control optimizations that may improve
8002 performance, but are not enabled by any @option{-O} options. This
8003 section includes experimental options that may produce broken code.
8006 @item -fbranch-probabilities
8007 @opindex fbranch-probabilities
8008 After running a program compiled with @option{-fprofile-arcs}
8009 (@pxref{Debugging Options,, Options for Debugging Your Program or
8010 @command{gcc}}), you can compile it a second time using
8011 @option{-fbranch-probabilities}, to improve optimizations based on
8012 the number of times each branch was taken. When the program
8013 compiled with @option{-fprofile-arcs} exits it saves arc execution
8014 counts to a file called @file{@var{sourcename}.gcda} for each source
8015 file. The information in this data file is very dependent on the
8016 structure of the generated code, so you must use the same source code
8017 and the same optimization options for both compilations.
8019 With @option{-fbranch-probabilities}, GCC puts a
8020 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8021 These can be used to improve optimization. Currently, they are only
8022 used in one place: in @file{reorg.c}, instead of guessing which path a
8023 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8024 exactly determine which path is taken more often.
8026 @item -fprofile-values
8027 @opindex fprofile-values
8028 If combined with @option{-fprofile-arcs}, it adds code so that some
8029 data about values of expressions in the program is gathered.
8031 With @option{-fbranch-probabilities}, it reads back the data gathered
8032 from profiling values of expressions for usage in optimizations.
8034 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8038 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8039 a code to gather information about values of expressions.
8041 With @option{-fbranch-probabilities}, it reads back the data gathered
8042 and actually performs the optimizations based on them.
8043 Currently the optimizations include specialization of division operation
8044 using the knowledge about the value of the denominator.
8046 @item -frename-registers
8047 @opindex frename-registers
8048 Attempt to avoid false dependencies in scheduled code by making use
8049 of registers left over after register allocation. This optimization
8050 will most benefit processors with lots of registers. Depending on the
8051 debug information format adopted by the target, however, it can
8052 make debugging impossible, since variables will no longer stay in
8053 a ``home register''.
8055 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8059 Perform tail duplication to enlarge superblock size. This transformation
8060 simplifies the control flow of the function allowing other optimizations to do
8063 Enabled with @option{-fprofile-use}.
8065 @item -funroll-loops
8066 @opindex funroll-loops
8067 Unroll loops whose number of iterations can be determined at compile time or
8068 upon entry to the loop. @option{-funroll-loops} implies
8069 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8070 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8071 small constant number of iterations). This option makes code larger, and may
8072 or may not make it run faster.
8074 Enabled with @option{-fprofile-use}.
8076 @item -funroll-all-loops
8077 @opindex funroll-all-loops
8078 Unroll all loops, even if their number of iterations is uncertain when
8079 the loop is entered. This usually makes programs run more slowly.
8080 @option{-funroll-all-loops} implies the same options as
8081 @option{-funroll-loops}.
8084 @opindex fpeel-loops
8085 Peels the loops for that there is enough information that they do not
8086 roll much (from profile feedback). It also turns on complete loop peeling
8087 (i.e.@: complete removal of loops with small constant number of iterations).
8089 Enabled with @option{-fprofile-use}.
8091 @item -fmove-loop-invariants
8092 @opindex fmove-loop-invariants
8093 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8094 at level @option{-O1}
8096 @item -funswitch-loops
8097 @opindex funswitch-loops
8098 Move branches with loop invariant conditions out of the loop, with duplicates
8099 of the loop on both branches (modified according to result of the condition).
8101 @item -ffunction-sections
8102 @itemx -fdata-sections
8103 @opindex ffunction-sections
8104 @opindex fdata-sections
8105 Place each function or data item into its own section in the output
8106 file if the target supports arbitrary sections. The name of the
8107 function or the name of the data item determines the section's name
8110 Use these options on systems where the linker can perform optimizations
8111 to improve locality of reference in the instruction space. Most systems
8112 using the ELF object format and SPARC processors running Solaris 2 have
8113 linkers with such optimizations. AIX may have these optimizations in
8116 Only use these options when there are significant benefits from doing
8117 so. When you specify these options, the assembler and linker will
8118 create larger object and executable files and will also be slower.
8119 You will not be able to use @code{gprof} on all systems if you
8120 specify this option and you may have problems with debugging if
8121 you specify both this option and @option{-g}.
8123 @item -fbranch-target-load-optimize
8124 @opindex fbranch-target-load-optimize
8125 Perform branch target register load optimization before prologue / epilogue
8127 The use of target registers can typically be exposed only during reload,
8128 thus hoisting loads out of loops and doing inter-block scheduling needs
8129 a separate optimization pass.
8131 @item -fbranch-target-load-optimize2
8132 @opindex fbranch-target-load-optimize2
8133 Perform branch target register load optimization after prologue / epilogue
8136 @item -fbtr-bb-exclusive
8137 @opindex fbtr-bb-exclusive
8138 When performing branch target register load optimization, don't reuse
8139 branch target registers in within any basic block.
8141 @item -fstack-protector
8142 @opindex fstack-protector
8143 Emit extra code to check for buffer overflows, such as stack smashing
8144 attacks. This is done by adding a guard variable to functions with
8145 vulnerable objects. This includes functions that call alloca, and
8146 functions with buffers larger than 8 bytes. The guards are initialized
8147 when a function is entered and then checked when the function exits.
8148 If a guard check fails, an error message is printed and the program exits.
8150 @item -fstack-protector-all
8151 @opindex fstack-protector-all
8152 Like @option{-fstack-protector} except that all functions are protected.
8154 @item -fsection-anchors
8155 @opindex fsection-anchors
8156 Try to reduce the number of symbolic address calculations by using
8157 shared ``anchor'' symbols to address nearby objects. This transformation
8158 can help to reduce the number of GOT entries and GOT accesses on some
8161 For example, the implementation of the following function @code{foo}:
8165 int foo (void) @{ return a + b + c; @}
8168 would usually calculate the addresses of all three variables, but if you
8169 compile it with @option{-fsection-anchors}, it will access the variables
8170 from a common anchor point instead. The effect is similar to the
8171 following pseudocode (which isn't valid C):
8176 register int *xr = &x;
8177 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8181 Not all targets support this option.
8183 @item --param @var{name}=@var{value}
8185 In some places, GCC uses various constants to control the amount of
8186 optimization that is done. For example, GCC will not inline functions
8187 that contain more that a certain number of instructions. You can
8188 control some of these constants on the command-line using the
8189 @option{--param} option.
8191 The names of specific parameters, and the meaning of the values, are
8192 tied to the internals of the compiler, and are subject to change
8193 without notice in future releases.
8195 In each case, the @var{value} is an integer. The allowable choices for
8196 @var{name} are given in the following table:
8199 @item struct-reorg-cold-struct-ratio
8200 The threshold ratio (as a percentage) between a structure frequency
8201 and the frequency of the hottest structure in the program. This parameter
8202 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
8203 We say that if the ratio of a structure frequency, calculated by profiling,
8204 to the hottest structure frequency in the program is less than this
8205 parameter, then structure reorganization is not applied to this structure.
8208 @item predictable-branch-outcome
8209 When branch is predicted to be taken with probability lower than this threshold
8210 (in percent), then it is considered well predictable. The default is 10.
8212 @item max-crossjump-edges
8213 The maximum number of incoming edges to consider for crossjumping.
8214 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8215 the number of edges incoming to each block. Increasing values mean
8216 more aggressive optimization, making the compile time increase with
8217 probably small improvement in executable size.
8219 @item min-crossjump-insns
8220 The minimum number of instructions which must be matched at the end
8221 of two blocks before crossjumping will be performed on them. This
8222 value is ignored in the case where all instructions in the block being
8223 crossjumped from are matched. The default value is 5.
8225 @item max-grow-copy-bb-insns
8226 The maximum code size expansion factor when copying basic blocks
8227 instead of jumping. The expansion is relative to a jump instruction.
8228 The default value is 8.
8230 @item max-goto-duplication-insns
8231 The maximum number of instructions to duplicate to a block that jumps
8232 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8233 passes, GCC factors computed gotos early in the compilation process,
8234 and unfactors them as late as possible. Only computed jumps at the
8235 end of a basic blocks with no more than max-goto-duplication-insns are
8236 unfactored. The default value is 8.
8238 @item max-delay-slot-insn-search
8239 The maximum number of instructions to consider when looking for an
8240 instruction to fill a delay slot. If more than this arbitrary number of
8241 instructions is searched, the time savings from filling the delay slot
8242 will be minimal so stop searching. Increasing values mean more
8243 aggressive optimization, making the compile time increase with probably
8244 small improvement in executable run time.
8246 @item max-delay-slot-live-search
8247 When trying to fill delay slots, the maximum number of instructions to
8248 consider when searching for a block with valid live register
8249 information. Increasing this arbitrarily chosen value means more
8250 aggressive optimization, increasing the compile time. This parameter
8251 should be removed when the delay slot code is rewritten to maintain the
8254 @item max-gcse-memory
8255 The approximate maximum amount of memory that will be allocated in
8256 order to perform the global common subexpression elimination
8257 optimization. If more memory than specified is required, the
8258 optimization will not be done.
8260 @item max-gcse-insertion-ratio
8261 If the ratio of expression insertions to deletions is larger than this value
8262 for any expression, then RTL PRE will insert or remove the expression and thus
8263 leave partially redundant computations in the instruction stream. The default value is 20.
8265 @item max-pending-list-length
8266 The maximum number of pending dependencies scheduling will allow
8267 before flushing the current state and starting over. Large functions
8268 with few branches or calls can create excessively large lists which
8269 needlessly consume memory and resources.
8271 @item max-inline-insns-single
8272 Several parameters control the tree inliner used in gcc.
8273 This number sets the maximum number of instructions (counted in GCC's
8274 internal representation) in a single function that the tree inliner
8275 will consider for inlining. This only affects functions declared
8276 inline and methods implemented in a class declaration (C++).
8277 The default value is 300.
8279 @item max-inline-insns-auto
8280 When you use @option{-finline-functions} (included in @option{-O3}),
8281 a lot of functions that would otherwise not be considered for inlining
8282 by the compiler will be investigated. To those functions, a different
8283 (more restrictive) limit compared to functions declared inline can
8285 The default value is 40.
8287 @item large-function-insns
8288 The limit specifying really large functions. For functions larger than this
8289 limit after inlining, inlining is constrained by
8290 @option{--param large-function-growth}. This parameter is useful primarily
8291 to avoid extreme compilation time caused by non-linear algorithms used by the
8293 The default value is 2700.
8295 @item large-function-growth
8296 Specifies maximal growth of large function caused by inlining in percents.
8297 The default value is 100 which limits large function growth to 2.0 times
8300 @item large-unit-insns
8301 The limit specifying large translation unit. Growth caused by inlining of
8302 units larger than this limit is limited by @option{--param inline-unit-growth}.
8303 For small units this might be too tight (consider unit consisting of function A
8304 that is inline and B that just calls A three time. If B is small relative to
8305 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8306 large units consisting of small inlineable functions however the overall unit
8307 growth limit is needed to avoid exponential explosion of code size. Thus for
8308 smaller units, the size is increased to @option{--param large-unit-insns}
8309 before applying @option{--param inline-unit-growth}. The default is 10000
8311 @item inline-unit-growth
8312 Specifies maximal overall growth of the compilation unit caused by inlining.
8313 The default value is 30 which limits unit growth to 1.3 times the original
8316 @item ipcp-unit-growth
8317 Specifies maximal overall growth of the compilation unit caused by
8318 interprocedural constant propagation. The default value is 10 which limits
8319 unit growth to 1.1 times the original size.
8321 @item large-stack-frame
8322 The limit specifying large stack frames. While inlining the algorithm is trying
8323 to not grow past this limit too much. Default value is 256 bytes.
8325 @item large-stack-frame-growth
8326 Specifies maximal growth of large stack frames caused by inlining in percents.
8327 The default value is 1000 which limits large stack frame growth to 11 times
8330 @item max-inline-insns-recursive
8331 @itemx max-inline-insns-recursive-auto
8332 Specifies maximum number of instructions out-of-line copy of self recursive inline
8333 function can grow into by performing recursive inlining.
8335 For functions declared inline @option{--param max-inline-insns-recursive} is
8336 taken into account. For function not declared inline, recursive inlining
8337 happens only when @option{-finline-functions} (included in @option{-O3}) is
8338 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8339 default value is 450.
8341 @item max-inline-recursive-depth
8342 @itemx max-inline-recursive-depth-auto
8343 Specifies maximum recursion depth used by the recursive inlining.
8345 For functions declared inline @option{--param max-inline-recursive-depth} is
8346 taken into account. For function not declared inline, recursive inlining
8347 happens only when @option{-finline-functions} (included in @option{-O3}) is
8348 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8351 @item min-inline-recursive-probability
8352 Recursive inlining is profitable only for function having deep recursion
8353 in average and can hurt for function having little recursion depth by
8354 increasing the prologue size or complexity of function body to other
8357 When profile feedback is available (see @option{-fprofile-generate}) the actual
8358 recursion depth can be guessed from probability that function will recurse via
8359 given call expression. This parameter limits inlining only to call expression
8360 whose probability exceeds given threshold (in percents). The default value is
8363 @item early-inlining-insns
8364 Specify growth that early inliner can make. In effect it increases amount of
8365 inlining for code having large abstraction penalty. The default value is 10.
8367 @item max-early-inliner-iterations
8368 @itemx max-early-inliner-iterations
8369 Limit of iterations of early inliner. This basically bounds number of nested
8370 indirect calls early inliner can resolve. Deeper chains are still handled by
8373 @item comdat-sharing-probability
8374 @itemx comdat-sharing-probability
8375 Probability (in percent) that C++ inline function with comdat visibility
8376 will be shared across multiple compilation units. The default value is 20.
8378 @item min-vect-loop-bound
8379 The minimum number of iterations under which a loop will not get vectorized
8380 when @option{-ftree-vectorize} is used. The number of iterations after
8381 vectorization needs to be greater than the value specified by this option
8382 to allow vectorization. The default value is 0.
8384 @item gcse-cost-distance-ratio
8385 Scaling factor in calculation of maximum distance an expression
8386 can be moved by GCSE optimizations. This is currently supported only in the
8387 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8388 will be with simple expressions, i.e., the expressions which have cost
8389 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8390 hoisting of simple expressions. The default value is 10.
8392 @item gcse-unrestricted-cost
8393 Cost, roughly measured as the cost of a single typical machine
8394 instruction, at which GCSE optimizations will not constrain
8395 the distance an expression can travel. This is currently
8396 supported only in the code hoisting pass. The lesser the cost,
8397 the more aggressive code hoisting will be. Specifying 0 will
8398 allow all expressions to travel unrestricted distances.
8399 The default value is 3.
8401 @item max-hoist-depth
8402 The depth of search in the dominator tree for expressions to hoist.
8403 This is used to avoid quadratic behavior in hoisting algorithm.
8404 The value of 0 will avoid limiting the search, but may slow down compilation
8405 of huge functions. The default value is 30.
8407 @item max-unrolled-insns
8408 The maximum number of instructions that a loop should have if that loop
8409 is unrolled, and if the loop is unrolled, it determines how many times
8410 the loop code is unrolled.
8412 @item max-average-unrolled-insns
8413 The maximum number of instructions biased by probabilities of their execution
8414 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8415 it determines how many times the loop code is unrolled.
8417 @item max-unroll-times
8418 The maximum number of unrollings of a single loop.
8420 @item max-peeled-insns
8421 The maximum number of instructions that a loop should have if that loop
8422 is peeled, and if the loop is peeled, it determines how many times
8423 the loop code is peeled.
8425 @item max-peel-times
8426 The maximum number of peelings of a single loop.
8428 @item max-completely-peeled-insns
8429 The maximum number of insns of a completely peeled loop.
8431 @item max-completely-peel-times
8432 The maximum number of iterations of a loop to be suitable for complete peeling.
8434 @item max-completely-peel-loop-nest-depth
8435 The maximum depth of a loop nest suitable for complete peeling.
8437 @item max-unswitch-insns
8438 The maximum number of insns of an unswitched loop.
8440 @item max-unswitch-level
8441 The maximum number of branches unswitched in a single loop.
8444 The minimum cost of an expensive expression in the loop invariant motion.
8446 @item iv-consider-all-candidates-bound
8447 Bound on number of candidates for induction variables below that
8448 all candidates are considered for each use in induction variable
8449 optimizations. Only the most relevant candidates are considered
8450 if there are more candidates, to avoid quadratic time complexity.
8452 @item iv-max-considered-uses
8453 The induction variable optimizations give up on loops that contain more
8454 induction variable uses.
8456 @item iv-always-prune-cand-set-bound
8457 If number of candidates in the set is smaller than this value,
8458 we always try to remove unnecessary ivs from the set during its
8459 optimization when a new iv is added to the set.
8461 @item scev-max-expr-size
8462 Bound on size of expressions used in the scalar evolutions analyzer.
8463 Large expressions slow the analyzer.
8465 @item scev-max-expr-complexity
8466 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8467 Complex expressions slow the analyzer.
8469 @item omega-max-vars
8470 The maximum number of variables in an Omega constraint system.
8471 The default value is 128.
8473 @item omega-max-geqs
8474 The maximum number of inequalities in an Omega constraint system.
8475 The default value is 256.
8478 The maximum number of equalities in an Omega constraint system.
8479 The default value is 128.
8481 @item omega-max-wild-cards
8482 The maximum number of wildcard variables that the Omega solver will
8483 be able to insert. The default value is 18.
8485 @item omega-hash-table-size
8486 The size of the hash table in the Omega solver. The default value is
8489 @item omega-max-keys
8490 The maximal number of keys used by the Omega solver. The default
8493 @item omega-eliminate-redundant-constraints
8494 When set to 1, use expensive methods to eliminate all redundant
8495 constraints. The default value is 0.
8497 @item vect-max-version-for-alignment-checks
8498 The maximum number of runtime checks that can be performed when
8499 doing loop versioning for alignment in the vectorizer. See option
8500 ftree-vect-loop-version for more information.
8502 @item vect-max-version-for-alias-checks
8503 The maximum number of runtime checks that can be performed when
8504 doing loop versioning for alias in the vectorizer. See option
8505 ftree-vect-loop-version for more information.
8507 @item max-iterations-to-track
8509 The maximum number of iterations of a loop the brute force algorithm
8510 for analysis of # of iterations of the loop tries to evaluate.
8512 @item hot-bb-count-fraction
8513 Select fraction of the maximal count of repetitions of basic block in program
8514 given basic block needs to have to be considered hot.
8516 @item hot-bb-frequency-fraction
8517 Select fraction of the entry block frequency of executions of basic block in
8518 function given basic block needs to have to be considered hot
8520 @item max-predicted-iterations
8521 The maximum number of loop iterations we predict statically. This is useful
8522 in cases where function contain single loop with known bound and other loop
8523 with unknown. We predict the known number of iterations correctly, while
8524 the unknown number of iterations average to roughly 10. This means that the
8525 loop without bounds would appear artificially cold relative to the other one.
8527 @item align-threshold
8529 Select fraction of the maximal frequency of executions of basic block in
8530 function given basic block will get aligned.
8532 @item align-loop-iterations
8534 A loop expected to iterate at lest the selected number of iterations will get
8537 @item tracer-dynamic-coverage
8538 @itemx tracer-dynamic-coverage-feedback
8540 This value is used to limit superblock formation once the given percentage of
8541 executed instructions is covered. This limits unnecessary code size
8544 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8545 feedback is available. The real profiles (as opposed to statically estimated
8546 ones) are much less balanced allowing the threshold to be larger value.
8548 @item tracer-max-code-growth
8549 Stop tail duplication once code growth has reached given percentage. This is
8550 rather hokey argument, as most of the duplicates will be eliminated later in
8551 cross jumping, so it may be set to much higher values than is the desired code
8554 @item tracer-min-branch-ratio
8556 Stop reverse growth when the reverse probability of best edge is less than this
8557 threshold (in percent).
8559 @item tracer-min-branch-ratio
8560 @itemx tracer-min-branch-ratio-feedback
8562 Stop forward growth if the best edge do have probability lower than this
8565 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8566 compilation for profile feedback and one for compilation without. The value
8567 for compilation with profile feedback needs to be more conservative (higher) in
8568 order to make tracer effective.
8570 @item max-cse-path-length
8572 Maximum number of basic blocks on path that cse considers. The default is 10.
8575 The maximum instructions CSE process before flushing. The default is 1000.
8577 @item ggc-min-expand
8579 GCC uses a garbage collector to manage its own memory allocation. This
8580 parameter specifies the minimum percentage by which the garbage
8581 collector's heap should be allowed to expand between collections.
8582 Tuning this may improve compilation speed; it has no effect on code
8585 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8586 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8587 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8588 GCC is not able to calculate RAM on a particular platform, the lower
8589 bound of 30% is used. Setting this parameter and
8590 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8591 every opportunity. This is extremely slow, but can be useful for
8594 @item ggc-min-heapsize
8596 Minimum size of the garbage collector's heap before it begins bothering
8597 to collect garbage. The first collection occurs after the heap expands
8598 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8599 tuning this may improve compilation speed, and has no effect on code
8602 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8603 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8604 with a lower bound of 4096 (four megabytes) and an upper bound of
8605 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8606 particular platform, the lower bound is used. Setting this parameter
8607 very large effectively disables garbage collection. Setting this
8608 parameter and @option{ggc-min-expand} to zero causes a full collection
8609 to occur at every opportunity.
8611 @item max-reload-search-insns
8612 The maximum number of instruction reload should look backward for equivalent
8613 register. Increasing values mean more aggressive optimization, making the
8614 compile time increase with probably slightly better performance. The default
8617 @item max-cselib-memory-locations
8618 The maximum number of memory locations cselib should take into account.
8619 Increasing values mean more aggressive optimization, making the compile time
8620 increase with probably slightly better performance. The default value is 500.
8622 @item reorder-blocks-duplicate
8623 @itemx reorder-blocks-duplicate-feedback
8625 Used by basic block reordering pass to decide whether to use unconditional
8626 branch or duplicate the code on its destination. Code is duplicated when its
8627 estimated size is smaller than this value multiplied by the estimated size of
8628 unconditional jump in the hot spots of the program.
8630 The @option{reorder-block-duplicate-feedback} is used only when profile
8631 feedback is available and may be set to higher values than
8632 @option{reorder-block-duplicate} since information about the hot spots is more
8635 @item max-sched-ready-insns
8636 The maximum number of instructions ready to be issued the scheduler should
8637 consider at any given time during the first scheduling pass. Increasing
8638 values mean more thorough searches, making the compilation time increase
8639 with probably little benefit. The default value is 100.
8641 @item max-sched-region-blocks
8642 The maximum number of blocks in a region to be considered for
8643 interblock scheduling. The default value is 10.
8645 @item max-pipeline-region-blocks
8646 The maximum number of blocks in a region to be considered for
8647 pipelining in the selective scheduler. The default value is 15.
8649 @item max-sched-region-insns
8650 The maximum number of insns in a region to be considered for
8651 interblock scheduling. The default value is 100.
8653 @item max-pipeline-region-insns
8654 The maximum number of insns in a region to be considered for
8655 pipelining in the selective scheduler. The default value is 200.
8658 The minimum probability (in percents) of reaching a source block
8659 for interblock speculative scheduling. The default value is 40.
8661 @item max-sched-extend-regions-iters
8662 The maximum number of iterations through CFG to extend regions.
8663 0 - disable region extension,
8664 N - do at most N iterations.
8665 The default value is 0.
8667 @item max-sched-insn-conflict-delay
8668 The maximum conflict delay for an insn to be considered for speculative motion.
8669 The default value is 3.
8671 @item sched-spec-prob-cutoff
8672 The minimal probability of speculation success (in percents), so that
8673 speculative insn will be scheduled.
8674 The default value is 40.
8676 @item sched-mem-true-dep-cost
8677 Minimal distance (in CPU cycles) between store and load targeting same
8678 memory locations. The default value is 1.
8680 @item selsched-max-lookahead
8681 The maximum size of the lookahead window of selective scheduling. It is a
8682 depth of search for available instructions.
8683 The default value is 50.
8685 @item selsched-max-sched-times
8686 The maximum number of times that an instruction will be scheduled during
8687 selective scheduling. This is the limit on the number of iterations
8688 through which the instruction may be pipelined. The default value is 2.
8690 @item selsched-max-insns-to-rename
8691 The maximum number of best instructions in the ready list that are considered
8692 for renaming in the selective scheduler. The default value is 2.
8694 @item max-last-value-rtl
8695 The maximum size measured as number of RTLs that can be recorded in an expression
8696 in combiner for a pseudo register as last known value of that register. The default
8699 @item integer-share-limit
8700 Small integer constants can use a shared data structure, reducing the
8701 compiler's memory usage and increasing its speed. This sets the maximum
8702 value of a shared integer constant. The default value is 256.
8704 @item min-virtual-mappings
8705 Specifies the minimum number of virtual mappings in the incremental
8706 SSA updater that should be registered to trigger the virtual mappings
8707 heuristic defined by virtual-mappings-ratio. The default value is
8710 @item virtual-mappings-ratio
8711 If the number of virtual mappings is virtual-mappings-ratio bigger
8712 than the number of virtual symbols to be updated, then the incremental
8713 SSA updater switches to a full update for those symbols. The default
8716 @item ssp-buffer-size
8717 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8718 protection when @option{-fstack-protection} is used.
8720 @item max-jump-thread-duplication-stmts
8721 Maximum number of statements allowed in a block that needs to be
8722 duplicated when threading jumps.
8724 @item max-fields-for-field-sensitive
8725 Maximum number of fields in a structure we will treat in
8726 a field sensitive manner during pointer analysis. The default is zero
8727 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8729 @item prefetch-latency
8730 Estimate on average number of instructions that are executed before
8731 prefetch finishes. The distance we prefetch ahead is proportional
8732 to this constant. Increasing this number may also lead to less
8733 streams being prefetched (see @option{simultaneous-prefetches}).
8735 @item simultaneous-prefetches
8736 Maximum number of prefetches that can run at the same time.
8738 @item l1-cache-line-size
8739 The size of cache line in L1 cache, in bytes.
8742 The size of L1 cache, in kilobytes.
8745 The size of L2 cache, in kilobytes.
8747 @item min-insn-to-prefetch-ratio
8748 The minimum ratio between the number of instructions and the
8749 number of prefetches to enable prefetching in a loop.
8751 @item prefetch-min-insn-to-mem-ratio
8752 The minimum ratio between the number of instructions and the
8753 number of memory references to enable prefetching in a loop.
8755 @item use-canonical-types
8756 Whether the compiler should use the ``canonical'' type system. By
8757 default, this should always be 1, which uses a more efficient internal
8758 mechanism for comparing types in C++ and Objective-C++. However, if
8759 bugs in the canonical type system are causing compilation failures,
8760 set this value to 0 to disable canonical types.
8762 @item switch-conversion-max-branch-ratio
8763 Switch initialization conversion will refuse to create arrays that are
8764 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8765 branches in the switch.
8767 @item max-partial-antic-length
8768 Maximum length of the partial antic set computed during the tree
8769 partial redundancy elimination optimization (@option{-ftree-pre}) when
8770 optimizing at @option{-O3} and above. For some sorts of source code
8771 the enhanced partial redundancy elimination optimization can run away,
8772 consuming all of the memory available on the host machine. This
8773 parameter sets a limit on the length of the sets that are computed,
8774 which prevents the runaway behavior. Setting a value of 0 for
8775 this parameter will allow an unlimited set length.
8777 @item sccvn-max-scc-size
8778 Maximum size of a strongly connected component (SCC) during SCCVN
8779 processing. If this limit is hit, SCCVN processing for the whole
8780 function will not be done and optimizations depending on it will
8781 be disabled. The default maximum SCC size is 10000.
8783 @item ira-max-loops-num
8784 IRA uses a regional register allocation by default. If a function
8785 contains loops more than number given by the parameter, only at most
8786 given number of the most frequently executed loops will form regions
8787 for the regional register allocation. The default value of the
8790 @item ira-max-conflict-table-size
8791 Although IRA uses a sophisticated algorithm of compression conflict
8792 table, the table can be still big for huge functions. If the conflict
8793 table for a function could be more than size in MB given by the
8794 parameter, the conflict table is not built and faster, simpler, and
8795 lower quality register allocation algorithm will be used. The
8796 algorithm do not use pseudo-register conflicts. The default value of
8797 the parameter is 2000.
8799 @item ira-loop-reserved-regs
8800 IRA can be used to evaluate more accurate register pressure in loops
8801 for decision to move loop invariants (see @option{-O3}). The number
8802 of available registers reserved for some other purposes is described
8803 by this parameter. The default value of the parameter is 2 which is
8804 minimal number of registers needed for execution of typical
8805 instruction. This value is the best found from numerous experiments.
8807 @item loop-invariant-max-bbs-in-loop
8808 Loop invariant motion can be very expensive, both in compile time and
8809 in amount of needed compile time memory, with very large loops. Loops
8810 with more basic blocks than this parameter won't have loop invariant
8811 motion optimization performed on them. The default value of the
8812 parameter is 1000 for -O1 and 10000 for -O2 and above.
8814 @item max-vartrack-size
8815 Sets a maximum number of hash table slots to use during variable
8816 tracking dataflow analysis of any function. If this limit is exceeded
8817 with variable tracking at assignments enabled, analysis for that
8818 function is retried without it, after removing all debug insns from
8819 the function. If the limit is exceeded even without debug insns, var
8820 tracking analysis is completely disabled for the function. Setting
8821 the parameter to zero makes it unlimited.
8823 @item min-nondebug-insn-uid
8824 Use uids starting at this parameter for nondebug insns. The range below
8825 the parameter is reserved exclusively for debug insns created by
8826 @option{-fvar-tracking-assignments}, but debug insns may get
8827 (non-overlapping) uids above it if the reserved range is exhausted.
8829 @item ipa-sra-ptr-growth-factor
8830 IPA-SRA will replace a pointer to an aggregate with one or more new
8831 parameters only when their cumulative size is less or equal to
8832 @option{ipa-sra-ptr-growth-factor} times the size of the original
8835 @item graphite-max-nb-scop-params
8836 To avoid exponential effects in the Graphite loop transforms, the
8837 number of parameters in a Static Control Part (SCoP) is bounded. The
8838 default value is 10 parameters. A variable whose value is unknown at
8839 compile time and defined outside a SCoP is a parameter of the SCoP.
8841 @item graphite-max-bbs-per-function
8842 To avoid exponential effects in the detection of SCoPs, the size of
8843 the functions analyzed by Graphite is bounded. The default value is
8846 @item loop-block-tile-size
8847 Loop blocking or strip mining transforms, enabled with
8848 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8849 loop in the loop nest by a given number of iterations. The strip
8850 length can be changed using the @option{loop-block-tile-size}
8851 parameter. The default value is 51 iterations.
8853 @item devirt-type-list-size
8854 IPA-CP attempts to track all possible types passed to a function's
8855 parameter in order to perform devirtualization.
8856 @option{devirt-type-list-size} is the maximum number of types it
8857 stores per a single formal parameter of a function.
8859 @item lto-partitions
8860 Specify desired number of partitions produced during WHOPR compilation.
8861 The number of partitions should exceed the number of CPUs used for compilation.
8862 The default value is 32.
8864 @item lto-minpartition
8865 Size of minimal partition for WHOPR (in estimated instructions).
8866 This prevents expenses of splitting very small programs into too many
8869 @item cxx-max-namespaces-for-diagnostic-help
8870 The maximum number of namespaces to consult for suggestions when C++
8871 name lookup fails for an identifier. The default is 1000.
8876 @node Preprocessor Options
8877 @section Options Controlling the Preprocessor
8878 @cindex preprocessor options
8879 @cindex options, preprocessor
8881 These options control the C preprocessor, which is run on each C source
8882 file before actual compilation.
8884 If you use the @option{-E} option, nothing is done except preprocessing.
8885 Some of these options make sense only together with @option{-E} because
8886 they cause the preprocessor output to be unsuitable for actual
8890 @item -Wp,@var{option}
8892 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8893 and pass @var{option} directly through to the preprocessor. If
8894 @var{option} contains commas, it is split into multiple options at the
8895 commas. However, many options are modified, translated or interpreted
8896 by the compiler driver before being passed to the preprocessor, and
8897 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8898 interface is undocumented and subject to change, so whenever possible
8899 you should avoid using @option{-Wp} and let the driver handle the
8902 @item -Xpreprocessor @var{option}
8903 @opindex Xpreprocessor
8904 Pass @var{option} as an option to the preprocessor. You can use this to
8905 supply system-specific preprocessor options which GCC does not know how to
8908 If you want to pass an option that takes an argument, you must use
8909 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8912 @include cppopts.texi
8914 @node Assembler Options
8915 @section Passing Options to the Assembler
8917 @c prevent bad page break with this line
8918 You can pass options to the assembler.
8921 @item -Wa,@var{option}
8923 Pass @var{option} as an option to the assembler. If @var{option}
8924 contains commas, it is split into multiple options at the commas.
8926 @item -Xassembler @var{option}
8928 Pass @var{option} as an option to the assembler. You can use this to
8929 supply system-specific assembler options which GCC does not know how to
8932 If you want to pass an option that takes an argument, you must use
8933 @option{-Xassembler} twice, once for the option and once for the argument.
8938 @section Options for Linking
8939 @cindex link options
8940 @cindex options, linking
8942 These options come into play when the compiler links object files into
8943 an executable output file. They are meaningless if the compiler is
8944 not doing a link step.
8948 @item @var{object-file-name}
8949 A file name that does not end in a special recognized suffix is
8950 considered to name an object file or library. (Object files are
8951 distinguished from libraries by the linker according to the file
8952 contents.) If linking is done, these object files are used as input
8961 If any of these options is used, then the linker is not run, and
8962 object file names should not be used as arguments. @xref{Overall
8966 @item -l@var{library}
8967 @itemx -l @var{library}
8969 Search the library named @var{library} when linking. (The second
8970 alternative with the library as a separate argument is only for
8971 POSIX compliance and is not recommended.)
8973 It makes a difference where in the command you write this option; the
8974 linker searches and processes libraries and object files in the order they
8975 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8976 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8977 to functions in @samp{z}, those functions may not be loaded.
8979 The linker searches a standard list of directories for the library,
8980 which is actually a file named @file{lib@var{library}.a}. The linker
8981 then uses this file as if it had been specified precisely by name.
8983 The directories searched include several standard system directories
8984 plus any that you specify with @option{-L}.
8986 Normally the files found this way are library files---archive files
8987 whose members are object files. The linker handles an archive file by
8988 scanning through it for members which define symbols that have so far
8989 been referenced but not defined. But if the file that is found is an
8990 ordinary object file, it is linked in the usual fashion. The only
8991 difference between using an @option{-l} option and specifying a file name
8992 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8993 and searches several directories.
8997 You need this special case of the @option{-l} option in order to
8998 link an Objective-C or Objective-C++ program.
9001 @opindex nostartfiles
9002 Do not use the standard system startup files when linking.
9003 The standard system libraries are used normally, unless @option{-nostdlib}
9004 or @option{-nodefaultlibs} is used.
9006 @item -nodefaultlibs
9007 @opindex nodefaultlibs
9008 Do not use the standard system libraries when linking.
9009 Only the libraries you specify will be passed to the linker, options
9010 specifying linkage of the system libraries, such as @code{-static-libgcc}
9011 or @code{-shared-libgcc}, will be ignored.
9012 The standard startup files are used normally, unless @option{-nostartfiles}
9013 is used. The compiler may generate calls to @code{memcmp},
9014 @code{memset}, @code{memcpy} and @code{memmove}.
9015 These entries are usually resolved by entries in
9016 libc. These entry points should be supplied through some other
9017 mechanism when this option is specified.
9021 Do not use the standard system startup files or libraries when linking.
9022 No startup files and only the libraries you specify will be passed to
9023 the linker, options specifying linkage of the system libraries, such as
9024 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9025 The compiler may generate calls to @code{memcmp}, @code{memset},
9026 @code{memcpy} and @code{memmove}.
9027 These entries are usually resolved by entries in
9028 libc. These entry points should be supplied through some other
9029 mechanism when this option is specified.
9031 @cindex @option{-lgcc}, use with @option{-nostdlib}
9032 @cindex @option{-nostdlib} and unresolved references
9033 @cindex unresolved references and @option{-nostdlib}
9034 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9035 @cindex @option{-nodefaultlibs} and unresolved references
9036 @cindex unresolved references and @option{-nodefaultlibs}
9037 One of the standard libraries bypassed by @option{-nostdlib} and
9038 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9039 that GCC uses to overcome shortcomings of particular machines, or special
9040 needs for some languages.
9041 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9042 Collection (GCC) Internals},
9043 for more discussion of @file{libgcc.a}.)
9044 In most cases, you need @file{libgcc.a} even when you want to avoid
9045 other standard libraries. In other words, when you specify @option{-nostdlib}
9046 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9047 This ensures that you have no unresolved references to internal GCC
9048 library subroutines. (For example, @samp{__main}, used to ensure C++
9049 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9050 GNU Compiler Collection (GCC) Internals}.)
9054 Produce a position independent executable on targets which support it.
9055 For predictable results, you must also specify the same set of options
9056 that were used to generate code (@option{-fpie}, @option{-fPIE},
9057 or model suboptions) when you specify this option.
9061 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9062 that support it. This instructs the linker to add all symbols, not
9063 only used ones, to the dynamic symbol table. This option is needed
9064 for some uses of @code{dlopen} or to allow obtaining backtraces
9065 from within a program.
9069 Remove all symbol table and relocation information from the executable.
9073 On systems that support dynamic linking, this prevents linking with the shared
9074 libraries. On other systems, this option has no effect.
9078 Produce a shared object which can then be linked with other objects to
9079 form an executable. Not all systems support this option. For predictable
9080 results, you must also specify the same set of options that were used to
9081 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9082 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9083 needs to build supplementary stub code for constructors to work. On
9084 multi-libbed systems, @samp{gcc -shared} must select the correct support
9085 libraries to link against. Failing to supply the correct flags may lead
9086 to subtle defects. Supplying them in cases where they are not necessary
9089 @item -shared-libgcc
9090 @itemx -static-libgcc
9091 @opindex shared-libgcc
9092 @opindex static-libgcc
9093 On systems that provide @file{libgcc} as a shared library, these options
9094 force the use of either the shared or static version respectively.
9095 If no shared version of @file{libgcc} was built when the compiler was
9096 configured, these options have no effect.
9098 There are several situations in which an application should use the
9099 shared @file{libgcc} instead of the static version. The most common
9100 of these is when the application wishes to throw and catch exceptions
9101 across different shared libraries. In that case, each of the libraries
9102 as well as the application itself should use the shared @file{libgcc}.
9104 Therefore, the G++ and GCJ drivers automatically add
9105 @option{-shared-libgcc} whenever you build a shared library or a main
9106 executable, because C++ and Java programs typically use exceptions, so
9107 this is the right thing to do.
9109 If, instead, you use the GCC driver to create shared libraries, you may
9110 find that they will not always be linked with the shared @file{libgcc}.
9111 If GCC finds, at its configuration time, that you have a non-GNU linker
9112 or a GNU linker that does not support option @option{--eh-frame-hdr},
9113 it will link the shared version of @file{libgcc} into shared libraries
9114 by default. Otherwise, it will take advantage of the linker and optimize
9115 away the linking with the shared version of @file{libgcc}, linking with
9116 the static version of libgcc by default. This allows exceptions to
9117 propagate through such shared libraries, without incurring relocation
9118 costs at library load time.
9120 However, if a library or main executable is supposed to throw or catch
9121 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9122 for the languages used in the program, or using the option
9123 @option{-shared-libgcc}, such that it is linked with the shared
9126 @item -static-libstdc++
9127 When the @command{g++} program is used to link a C++ program, it will
9128 normally automatically link against @option{libstdc++}. If
9129 @file{libstdc++} is available as a shared library, and the
9130 @option{-static} option is not used, then this will link against the
9131 shared version of @file{libstdc++}. That is normally fine. However, it
9132 is sometimes useful to freeze the version of @file{libstdc++} used by
9133 the program without going all the way to a fully static link. The
9134 @option{-static-libstdc++} option directs the @command{g++} driver to
9135 link @file{libstdc++} statically, without necessarily linking other
9136 libraries statically.
9140 Bind references to global symbols when building a shared object. Warn
9141 about any unresolved references (unless overridden by the link editor
9142 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9145 @item -T @var{script}
9147 @cindex linker script
9148 Use @var{script} as the linker script. This option is supported by most
9149 systems using the GNU linker. On some targets, such as bare-board
9150 targets without an operating system, the @option{-T} option may be required
9151 when linking to avoid references to undefined symbols.
9153 @item -Xlinker @var{option}
9155 Pass @var{option} as an option to the linker. You can use this to
9156 supply system-specific linker options which GCC does not know how to
9159 If you want to pass an option that takes a separate argument, you must use
9160 @option{-Xlinker} twice, once for the option and once for the argument.
9161 For example, to pass @option{-assert definitions}, you must write
9162 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9163 @option{-Xlinker "-assert definitions"}, because this passes the entire
9164 string as a single argument, which is not what the linker expects.
9166 When using the GNU linker, it is usually more convenient to pass
9167 arguments to linker options using the @option{@var{option}=@var{value}}
9168 syntax than as separate arguments. For example, you can specify
9169 @samp{-Xlinker -Map=output.map} rather than
9170 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9171 this syntax for command-line options.
9173 @item -Wl,@var{option}
9175 Pass @var{option} as an option to the linker. If @var{option} contains
9176 commas, it is split into multiple options at the commas. You can use this
9177 syntax to pass an argument to the option.
9178 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9179 linker. When using the GNU linker, you can also get the same effect with
9180 @samp{-Wl,-Map=output.map}.
9182 @item -u @var{symbol}
9184 Pretend the symbol @var{symbol} is undefined, to force linking of
9185 library modules to define it. You can use @option{-u} multiple times with
9186 different symbols to force loading of additional library modules.
9189 @node Directory Options
9190 @section Options for Directory Search
9191 @cindex directory options
9192 @cindex options, directory search
9195 These options specify directories to search for header files, for
9196 libraries and for parts of the compiler:
9201 Add the directory @var{dir} to the head of the list of directories to be
9202 searched for header files. This can be used to override a system header
9203 file, substituting your own version, since these directories are
9204 searched before the system header file directories. However, you should
9205 not use this option to add directories that contain vendor-supplied
9206 system header files (use @option{-isystem} for that). If you use more than
9207 one @option{-I} option, the directories are scanned in left-to-right
9208 order; the standard system directories come after.
9210 If a standard system include directory, or a directory specified with
9211 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9212 option will be ignored. The directory will still be searched but as a
9213 system directory at its normal position in the system include chain.
9214 This is to ensure that GCC's procedure to fix buggy system headers and
9215 the ordering for the include_next directive are not inadvertently changed.
9216 If you really need to change the search order for system directories,
9217 use the @option{-nostdinc} and/or @option{-isystem} options.
9219 @item -iplugindir=@var{dir}
9220 Set the directory to search for plugins which are passed
9221 by @option{-fplugin=@var{name}} instead of
9222 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9223 to be used by the user, but only passed by the driver.
9225 @item -iquote@var{dir}
9227 Add the directory @var{dir} to the head of the list of directories to
9228 be searched for header files only for the case of @samp{#include
9229 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9230 otherwise just like @option{-I}.
9234 Add directory @var{dir} to the list of directories to be searched
9237 @item -B@var{prefix}
9239 This option specifies where to find the executables, libraries,
9240 include files, and data files of the compiler itself.
9242 The compiler driver program runs one or more of the subprograms
9243 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9244 @var{prefix} as a prefix for each program it tries to run, both with and
9245 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9247 For each subprogram to be run, the compiler driver first tries the
9248 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9249 was not specified, the driver tries two standard prefixes, which are
9250 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9251 those results in a file name that is found, the unmodified program
9252 name is searched for using the directories specified in your
9253 @env{PATH} environment variable.
9255 The compiler will check to see if the path provided by the @option{-B}
9256 refers to a directory, and if necessary it will add a directory
9257 separator character at the end of the path.
9259 @option{-B} prefixes that effectively specify directory names also apply
9260 to libraries in the linker, because the compiler translates these
9261 options into @option{-L} options for the linker. They also apply to
9262 includes files in the preprocessor, because the compiler translates these
9263 options into @option{-isystem} options for the preprocessor. In this case,
9264 the compiler appends @samp{include} to the prefix.
9266 The run-time support file @file{libgcc.a} can also be searched for using
9267 the @option{-B} prefix, if needed. If it is not found there, the two
9268 standard prefixes above are tried, and that is all. The file is left
9269 out of the link if it is not found by those means.
9271 Another way to specify a prefix much like the @option{-B} prefix is to use
9272 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9275 As a special kludge, if the path provided by @option{-B} is
9276 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9277 9, then it will be replaced by @file{[dir/]include}. This is to help
9278 with boot-strapping the compiler.
9280 @item -specs=@var{file}
9282 Process @var{file} after the compiler reads in the standard @file{specs}
9283 file, in order to override the defaults that the @file{gcc} driver
9284 program uses when determining what switches to pass to @file{cc1},
9285 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9286 @option{-specs=@var{file}} can be specified on the command line, and they
9287 are processed in order, from left to right.
9289 @item --sysroot=@var{dir}
9291 Use @var{dir} as the logical root directory for headers and libraries.
9292 For example, if the compiler would normally search for headers in
9293 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9294 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9296 If you use both this option and the @option{-isysroot} option, then
9297 the @option{--sysroot} option will apply to libraries, but the
9298 @option{-isysroot} option will apply to header files.
9300 The GNU linker (beginning with version 2.16) has the necessary support
9301 for this option. If your linker does not support this option, the
9302 header file aspect of @option{--sysroot} will still work, but the
9303 library aspect will not.
9307 This option has been deprecated. Please use @option{-iquote} instead for
9308 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9309 Any directories you specify with @option{-I} options before the @option{-I-}
9310 option are searched only for the case of @samp{#include "@var{file}"};
9311 they are not searched for @samp{#include <@var{file}>}.
9313 If additional directories are specified with @option{-I} options after
9314 the @option{-I-}, these directories are searched for all @samp{#include}
9315 directives. (Ordinarily @emph{all} @option{-I} directories are used
9318 In addition, the @option{-I-} option inhibits the use of the current
9319 directory (where the current input file came from) as the first search
9320 directory for @samp{#include "@var{file}"}. There is no way to
9321 override this effect of @option{-I-}. With @option{-I.} you can specify
9322 searching the directory which was current when the compiler was
9323 invoked. That is not exactly the same as what the preprocessor does
9324 by default, but it is often satisfactory.
9326 @option{-I-} does not inhibit the use of the standard system directories
9327 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9334 @section Specifying subprocesses and the switches to pass to them
9337 @command{gcc} is a driver program. It performs its job by invoking a
9338 sequence of other programs to do the work of compiling, assembling and
9339 linking. GCC interprets its command-line parameters and uses these to
9340 deduce which programs it should invoke, and which command-line options
9341 it ought to place on their command lines. This behavior is controlled
9342 by @dfn{spec strings}. In most cases there is one spec string for each
9343 program that GCC can invoke, but a few programs have multiple spec
9344 strings to control their behavior. The spec strings built into GCC can
9345 be overridden by using the @option{-specs=} command-line switch to specify
9348 @dfn{Spec files} are plaintext files that are used to construct spec
9349 strings. They consist of a sequence of directives separated by blank
9350 lines. The type of directive is determined by the first non-whitespace
9351 character on the line and it can be one of the following:
9354 @item %@var{command}
9355 Issues a @var{command} to the spec file processor. The commands that can
9359 @item %include <@var{file}>
9360 @cindex @code{%include}
9361 Search for @var{file} and insert its text at the current point in the
9364 @item %include_noerr <@var{file}>
9365 @cindex @code{%include_noerr}
9366 Just like @samp{%include}, but do not generate an error message if the include
9367 file cannot be found.
9369 @item %rename @var{old_name} @var{new_name}
9370 @cindex @code{%rename}
9371 Rename the spec string @var{old_name} to @var{new_name}.
9375 @item *[@var{spec_name}]:
9376 This tells the compiler to create, override or delete the named spec
9377 string. All lines after this directive up to the next directive or
9378 blank line are considered to be the text for the spec string. If this
9379 results in an empty string then the spec will be deleted. (Or, if the
9380 spec did not exist, then nothing will happened.) Otherwise, if the spec
9381 does not currently exist a new spec will be created. If the spec does
9382 exist then its contents will be overridden by the text of this
9383 directive, unless the first character of that text is the @samp{+}
9384 character, in which case the text will be appended to the spec.
9386 @item [@var{suffix}]:
9387 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9388 and up to the next directive or blank line are considered to make up the
9389 spec string for the indicated suffix. When the compiler encounters an
9390 input file with the named suffix, it will processes the spec string in
9391 order to work out how to compile that file. For example:
9398 This says that any input file whose name ends in @samp{.ZZ} should be
9399 passed to the program @samp{z-compile}, which should be invoked with the
9400 command-line switch @option{-input} and with the result of performing the
9401 @samp{%i} substitution. (See below.)
9403 As an alternative to providing a spec string, the text that follows a
9404 suffix directive can be one of the following:
9407 @item @@@var{language}
9408 This says that the suffix is an alias for a known @var{language}. This is
9409 similar to using the @option{-x} command-line switch to GCC to specify a
9410 language explicitly. For example:
9417 Says that .ZZ files are, in fact, C++ source files.
9420 This causes an error messages saying:
9423 @var{name} compiler not installed on this system.
9427 GCC already has an extensive list of suffixes built into it.
9428 This directive will add an entry to the end of the list of suffixes, but
9429 since the list is searched from the end backwards, it is effectively
9430 possible to override earlier entries using this technique.
9434 GCC has the following spec strings built into it. Spec files can
9435 override these strings or create their own. Note that individual
9436 targets can also add their own spec strings to this list.
9439 asm Options to pass to the assembler
9440 asm_final Options to pass to the assembler post-processor
9441 cpp Options to pass to the C preprocessor
9442 cc1 Options to pass to the C compiler
9443 cc1plus Options to pass to the C++ compiler
9444 endfile Object files to include at the end of the link
9445 link Options to pass to the linker
9446 lib Libraries to include on the command line to the linker
9447 libgcc Decides which GCC support library to pass to the linker
9448 linker Sets the name of the linker
9449 predefines Defines to be passed to the C preprocessor
9450 signed_char Defines to pass to CPP to say whether @code{char} is signed
9452 startfile Object files to include at the start of the link
9455 Here is a small example of a spec file:
9461 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9464 This example renames the spec called @samp{lib} to @samp{old_lib} and
9465 then overrides the previous definition of @samp{lib} with a new one.
9466 The new definition adds in some extra command-line options before
9467 including the text of the old definition.
9469 @dfn{Spec strings} are a list of command-line options to be passed to their
9470 corresponding program. In addition, the spec strings can contain
9471 @samp{%}-prefixed sequences to substitute variable text or to
9472 conditionally insert text into the command line. Using these constructs
9473 it is possible to generate quite complex command lines.
9475 Here is a table of all defined @samp{%}-sequences for spec
9476 strings. Note that spaces are not generated automatically around the
9477 results of expanding these sequences. Therefore you can concatenate them
9478 together or combine them with constant text in a single argument.
9482 Substitute one @samp{%} into the program name or argument.
9485 Substitute the name of the input file being processed.
9488 Substitute the basename of the input file being processed.
9489 This is the substring up to (and not including) the last period
9490 and not including the directory.
9493 This is the same as @samp{%b}, but include the file suffix (text after
9497 Marks the argument containing or following the @samp{%d} as a
9498 temporary file name, so that that file will be deleted if GCC exits
9499 successfully. Unlike @samp{%g}, this contributes no text to the
9502 @item %g@var{suffix}
9503 Substitute a file name that has suffix @var{suffix} and is chosen
9504 once per compilation, and mark the argument in the same way as
9505 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9506 name is now chosen in a way that is hard to predict even when previously
9507 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9508 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9509 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9510 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9511 was simply substituted with a file name chosen once per compilation,
9512 without regard to any appended suffix (which was therefore treated
9513 just like ordinary text), making such attacks more likely to succeed.
9515 @item %u@var{suffix}
9516 Like @samp{%g}, but generates a new temporary file name even if
9517 @samp{%u@var{suffix}} was already seen.
9519 @item %U@var{suffix}
9520 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9521 new one if there is no such last file name. In the absence of any
9522 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9523 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9524 would involve the generation of two distinct file names, one
9525 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9526 simply substituted with a file name chosen for the previous @samp{%u},
9527 without regard to any appended suffix.
9529 @item %j@var{suffix}
9530 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9531 writable, and if save-temps is off; otherwise, substitute the name
9532 of a temporary file, just like @samp{%u}. This temporary file is not
9533 meant for communication between processes, but rather as a junk
9536 @item %|@var{suffix}
9537 @itemx %m@var{suffix}
9538 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9539 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9540 all. These are the two most common ways to instruct a program that it
9541 should read from standard input or write to standard output. If you
9542 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9543 construct: see for example @file{f/lang-specs.h}.
9545 @item %.@var{SUFFIX}
9546 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9547 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9548 terminated by the next space or %.
9551 Marks the argument containing or following the @samp{%w} as the
9552 designated output file of this compilation. This puts the argument
9553 into the sequence of arguments that @samp{%o} will substitute later.
9556 Substitutes the names of all the output files, with spaces
9557 automatically placed around them. You should write spaces
9558 around the @samp{%o} as well or the results are undefined.
9559 @samp{%o} is for use in the specs for running the linker.
9560 Input files whose names have no recognized suffix are not compiled
9561 at all, but they are included among the output files, so they will
9565 Substitutes the suffix for object files. Note that this is
9566 handled specially when it immediately follows @samp{%g, %u, or %U},
9567 because of the need for those to form complete file names. The
9568 handling is such that @samp{%O} is treated exactly as if it had already
9569 been substituted, except that @samp{%g, %u, and %U} do not currently
9570 support additional @var{suffix} characters following @samp{%O} as they would
9571 following, for example, @samp{.o}.
9574 Substitutes the standard macro predefinitions for the
9575 current target machine. Use this when running @code{cpp}.
9578 Like @samp{%p}, but puts @samp{__} before and after the name of each
9579 predefined macro, except for macros that start with @samp{__} or with
9580 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9584 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9585 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9586 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9587 and @option{-imultilib} as necessary.
9590 Current argument is the name of a library or startup file of some sort.
9591 Search for that file in a standard list of directories and substitute
9592 the full name found. The current working directory is included in the
9593 list of directories scanned.
9596 Current argument is the name of a linker script. Search for that file
9597 in the current list of directories to scan for libraries. If the file
9598 is located insert a @option{--script} option into the command line
9599 followed by the full path name found. If the file is not found then
9600 generate an error message. Note: the current working directory is not
9604 Print @var{str} as an error message. @var{str} is terminated by a newline.
9605 Use this when inconsistent options are detected.
9608 Substitute the contents of spec string @var{name} at this point.
9611 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9613 @item %x@{@var{option}@}
9614 Accumulate an option for @samp{%X}.
9617 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9621 Output the accumulated assembler options specified by @option{-Wa}.
9624 Output the accumulated preprocessor options specified by @option{-Wp}.
9627 Process the @code{asm} spec. This is used to compute the
9628 switches to be passed to the assembler.
9631 Process the @code{asm_final} spec. This is a spec string for
9632 passing switches to an assembler post-processor, if such a program is
9636 Process the @code{link} spec. This is the spec for computing the
9637 command line passed to the linker. Typically it will make use of the
9638 @samp{%L %G %S %D and %E} sequences.
9641 Dump out a @option{-L} option for each directory that GCC believes might
9642 contain startup files. If the target supports multilibs then the
9643 current multilib directory will be prepended to each of these paths.
9646 Process the @code{lib} spec. This is a spec string for deciding which
9647 libraries should be included on the command line to the linker.
9650 Process the @code{libgcc} spec. This is a spec string for deciding
9651 which GCC support library should be included on the command line to the linker.
9654 Process the @code{startfile} spec. This is a spec for deciding which
9655 object files should be the first ones passed to the linker. Typically
9656 this might be a file named @file{crt0.o}.
9659 Process the @code{endfile} spec. This is a spec string that specifies
9660 the last object files that will be passed to the linker.
9663 Process the @code{cpp} spec. This is used to construct the arguments
9664 to be passed to the C preprocessor.
9667 Process the @code{cc1} spec. This is used to construct the options to be
9668 passed to the actual C compiler (@samp{cc1}).
9671 Process the @code{cc1plus} spec. This is used to construct the options to be
9672 passed to the actual C++ compiler (@samp{cc1plus}).
9675 Substitute the variable part of a matched option. See below.
9676 Note that each comma in the substituted string is replaced by
9680 Remove all occurrences of @code{-S} from the command line. Note---this
9681 command is position dependent. @samp{%} commands in the spec string
9682 before this one will see @code{-S}, @samp{%} commands in the spec string
9683 after this one will not.
9685 @item %:@var{function}(@var{args})
9686 Call the named function @var{function}, passing it @var{args}.
9687 @var{args} is first processed as a nested spec string, then split
9688 into an argument vector in the usual fashion. The function returns
9689 a string which is processed as if it had appeared literally as part
9690 of the current spec.
9692 The following built-in spec functions are provided:
9696 The @code{getenv} spec function takes two arguments: an environment
9697 variable name and a string. If the environment variable is not
9698 defined, a fatal error is issued. Otherwise, the return value is the
9699 value of the environment variable concatenated with the string. For
9700 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9703 %:getenv(TOPDIR /include)
9706 expands to @file{/path/to/top/include}.
9708 @item @code{if-exists}
9709 The @code{if-exists} spec function takes one argument, an absolute
9710 pathname to a file. If the file exists, @code{if-exists} returns the
9711 pathname. Here is a small example of its usage:
9715 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9718 @item @code{if-exists-else}
9719 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9720 spec function, except that it takes two arguments. The first argument is
9721 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9722 returns the pathname. If it does not exist, it returns the second argument.
9723 This way, @code{if-exists-else} can be used to select one file or another,
9724 based on the existence of the first. Here is a small example of its usage:
9728 crt0%O%s %:if-exists(crti%O%s) \
9729 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9732 @item @code{replace-outfile}
9733 The @code{replace-outfile} spec function takes two arguments. It looks for the
9734 first argument in the outfiles array and replaces it with the second argument. Here
9735 is a small example of its usage:
9738 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9741 @item @code{remove-outfile}
9742 The @code{remove-outfile} spec function takes one argument. It looks for the
9743 first argument in the outfiles array and removes it. Here is a small example
9747 %:remove-outfile(-lm)
9750 @item @code{pass-through-libs}
9751 The @code{pass-through-libs} spec function takes any number of arguments. It
9752 finds any @option{-l} options and any non-options ending in ".a" (which it
9753 assumes are the names of linker input library archive files) and returns a
9754 result containing all the found arguments each prepended by
9755 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9756 intended to be passed to the LTO linker plugin.
9759 %:pass-through-libs(%G %L %G)
9762 @item @code{print-asm-header}
9763 The @code{print-asm-header} function takes no arguments and simply
9764 prints a banner like:
9770 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9773 It is used to separate compiler options from assembler options
9774 in the @option{--target-help} output.
9778 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9779 If that switch was not specified, this substitutes nothing. Note that
9780 the leading dash is omitted when specifying this option, and it is
9781 automatically inserted if the substitution is performed. Thus the spec
9782 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9783 and would output the command line option @option{-foo}.
9785 @item %W@{@code{S}@}
9786 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9789 @item %@{@code{S}*@}
9790 Substitutes all the switches specified to GCC whose names start
9791 with @code{-S}, but which also take an argument. This is used for
9792 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9793 GCC considers @option{-o foo} as being
9794 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9795 text, including the space. Thus two arguments would be generated.
9797 @item %@{@code{S}*&@code{T}*@}
9798 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9799 (the order of @code{S} and @code{T} in the spec is not significant).
9800 There can be any number of ampersand-separated variables; for each the
9801 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9803 @item %@{@code{S}:@code{X}@}
9804 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9806 @item %@{!@code{S}:@code{X}@}
9807 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9809 @item %@{@code{S}*:@code{X}@}
9810 Substitutes @code{X} if one or more switches whose names start with
9811 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9812 once, no matter how many such switches appeared. However, if @code{%*}
9813 appears somewhere in @code{X}, then @code{X} will be substituted once
9814 for each matching switch, with the @code{%*} replaced by the part of
9815 that switch that matched the @code{*}.
9817 @item %@{.@code{S}:@code{X}@}
9818 Substitutes @code{X}, if processing a file with suffix @code{S}.
9820 @item %@{!.@code{S}:@code{X}@}
9821 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9823 @item %@{,@code{S}:@code{X}@}
9824 Substitutes @code{X}, if processing a file for language @code{S}.
9826 @item %@{!,@code{S}:@code{X}@}
9827 Substitutes @code{X}, if not processing a file for language @code{S}.
9829 @item %@{@code{S}|@code{P}:@code{X}@}
9830 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9831 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9832 @code{*} sequences as well, although they have a stronger binding than
9833 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9834 alternatives must be starred, and only the first matching alternative
9837 For example, a spec string like this:
9840 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9843 will output the following command-line options from the following input
9844 command-line options:
9849 -d fred.c -foo -baz -boggle
9850 -d jim.d -bar -baz -boggle
9853 @item %@{S:X; T:Y; :D@}
9855 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9856 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9857 be as many clauses as you need. This may be combined with @code{.},
9858 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9863 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9864 construct may contain other nested @samp{%} constructs or spaces, or
9865 even newlines. They are processed as usual, as described above.
9866 Trailing white space in @code{X} is ignored. White space may also
9867 appear anywhere on the left side of the colon in these constructs,
9868 except between @code{.} or @code{*} and the corresponding word.
9870 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9871 handled specifically in these constructs. If another value of
9872 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9873 @option{-W} switch is found later in the command line, the earlier
9874 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9875 just one letter, which passes all matching options.
9877 The character @samp{|} at the beginning of the predicate text is used to
9878 indicate that a command should be piped to the following command, but
9879 only if @option{-pipe} is specified.
9881 It is built into GCC which switches take arguments and which do not.
9882 (You might think it would be useful to generalize this to allow each
9883 compiler's spec to say which switches take arguments. But this cannot
9884 be done in a consistent fashion. GCC cannot even decide which input
9885 files have been specified without knowing which switches take arguments,
9886 and it must know which input files to compile in order to tell which
9889 GCC also knows implicitly that arguments starting in @option{-l} are to be
9890 treated as compiler output files, and passed to the linker in their
9891 proper position among the other output files.
9893 @c man begin OPTIONS
9895 @node Target Options
9896 @section Specifying Target Machine and Compiler Version
9897 @cindex target options
9898 @cindex cross compiling
9899 @cindex specifying machine version
9900 @cindex specifying compiler version and target machine
9901 @cindex compiler version, specifying
9902 @cindex target machine, specifying
9904 The usual way to run GCC is to run the executable called @command{gcc}, or
9905 @command{@var{machine}-gcc} when cross-compiling, or
9906 @command{@var{machine}-gcc-@var{version}} to run a version other than the
9907 one that was installed last.
9909 @node Submodel Options
9910 @section Hardware Models and Configurations
9911 @cindex submodel options
9912 @cindex specifying hardware config
9913 @cindex hardware models and configurations, specifying
9914 @cindex machine dependent options
9916 Each target machine types can have its own
9917 special options, starting with @samp{-m}, to choose among various
9918 hardware models or configurations---for example, 68010 vs 68020,
9919 floating coprocessor or none. A single installed version of the
9920 compiler can compile for any model or configuration, according to the
9923 Some configurations of the compiler also support additional special
9924 options, usually for compatibility with other compilers on the same
9927 @c This list is ordered alphanumerically by subsection name.
9928 @c It should be the same order and spelling as these options are listed
9929 @c in Machine Dependent Options
9935 * Blackfin Options::
9939 * DEC Alpha Options::
9940 * DEC Alpha/VMS Options::
9943 * GNU/Linux Options::
9946 * i386 and x86-64 Options::
9947 * i386 and x86-64 Windows Options::
9949 * IA-64/VMS Options::
9957 * MicroBlaze Options::
9962 * picoChip Options::
9964 * RS/6000 and PowerPC Options::
9966 * S/390 and zSeries Options::
9969 * Solaris 2 Options::
9972 * System V Options::
9977 * Xstormy16 Options::
9983 @subsection ARC Options
9986 These options are defined for ARC implementations:
9991 Compile code for little endian mode. This is the default.
9995 Compile code for big endian mode.
9998 @opindex mmangle-cpu
9999 Prepend the name of the cpu to all public symbol names.
10000 In multiple-processor systems, there are many ARC variants with different
10001 instruction and register set characteristics. This flag prevents code
10002 compiled for one cpu to be linked with code compiled for another.
10003 No facility exists for handling variants that are ``almost identical''.
10004 This is an all or nothing option.
10006 @item -mcpu=@var{cpu}
10008 Compile code for ARC variant @var{cpu}.
10009 Which variants are supported depend on the configuration.
10010 All variants support @option{-mcpu=base}, this is the default.
10012 @item -mtext=@var{text-section}
10013 @itemx -mdata=@var{data-section}
10014 @itemx -mrodata=@var{readonly-data-section}
10018 Put functions, data, and readonly data in @var{text-section},
10019 @var{data-section}, and @var{readonly-data-section} respectively
10020 by default. This can be overridden with the @code{section} attribute.
10021 @xref{Variable Attributes}.
10026 @subsection ARM Options
10027 @cindex ARM options
10029 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10033 @item -mabi=@var{name}
10035 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10036 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10039 @opindex mapcs-frame
10040 Generate a stack frame that is compliant with the ARM Procedure Call
10041 Standard for all functions, even if this is not strictly necessary for
10042 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10043 with this option will cause the stack frames not to be generated for
10044 leaf functions. The default is @option{-mno-apcs-frame}.
10048 This is a synonym for @option{-mapcs-frame}.
10051 @c not currently implemented
10052 @item -mapcs-stack-check
10053 @opindex mapcs-stack-check
10054 Generate code to check the amount of stack space available upon entry to
10055 every function (that actually uses some stack space). If there is
10056 insufficient space available then either the function
10057 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10058 called, depending upon the amount of stack space required. The run time
10059 system is required to provide these functions. The default is
10060 @option{-mno-apcs-stack-check}, since this produces smaller code.
10062 @c not currently implemented
10064 @opindex mapcs-float
10065 Pass floating point arguments using the float point registers. This is
10066 one of the variants of the APCS@. This option is recommended if the
10067 target hardware has a floating point unit or if a lot of floating point
10068 arithmetic is going to be performed by the code. The default is
10069 @option{-mno-apcs-float}, since integer only code is slightly increased in
10070 size if @option{-mapcs-float} is used.
10072 @c not currently implemented
10073 @item -mapcs-reentrant
10074 @opindex mapcs-reentrant
10075 Generate reentrant, position independent code. The default is
10076 @option{-mno-apcs-reentrant}.
10079 @item -mthumb-interwork
10080 @opindex mthumb-interwork
10081 Generate code which supports calling between the ARM and Thumb
10082 instruction sets. Without this option the two instruction sets cannot
10083 be reliably used inside one program. The default is
10084 @option{-mno-thumb-interwork}, since slightly larger code is generated
10085 when @option{-mthumb-interwork} is specified.
10087 @item -mno-sched-prolog
10088 @opindex mno-sched-prolog
10089 Prevent the reordering of instructions in the function prolog, or the
10090 merging of those instruction with the instructions in the function's
10091 body. This means that all functions will start with a recognizable set
10092 of instructions (or in fact one of a choice from a small set of
10093 different function prologues), and this information can be used to
10094 locate the start if functions inside an executable piece of code. The
10095 default is @option{-msched-prolog}.
10097 @item -mfloat-abi=@var{name}
10098 @opindex mfloat-abi
10099 Specifies which floating-point ABI to use. Permissible values
10100 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10102 Specifying @samp{soft} causes GCC to generate output containing
10103 library calls for floating-point operations.
10104 @samp{softfp} allows the generation of code using hardware floating-point
10105 instructions, but still uses the soft-float calling conventions.
10106 @samp{hard} allows generation of floating-point instructions
10107 and uses FPU-specific calling conventions.
10109 The default depends on the specific target configuration. Note that
10110 the hard-float and soft-float ABIs are not link-compatible; you must
10111 compile your entire program with the same ABI, and link with a
10112 compatible set of libraries.
10115 @opindex mhard-float
10116 Equivalent to @option{-mfloat-abi=hard}.
10119 @opindex msoft-float
10120 Equivalent to @option{-mfloat-abi=soft}.
10122 @item -mlittle-endian
10123 @opindex mlittle-endian
10124 Generate code for a processor running in little-endian mode. This is
10125 the default for all standard configurations.
10128 @opindex mbig-endian
10129 Generate code for a processor running in big-endian mode; the default is
10130 to compile code for a little-endian processor.
10132 @item -mwords-little-endian
10133 @opindex mwords-little-endian
10134 This option only applies when generating code for big-endian processors.
10135 Generate code for a little-endian word order but a big-endian byte
10136 order. That is, a byte order of the form @samp{32107654}. Note: this
10137 option should only be used if you require compatibility with code for
10138 big-endian ARM processors generated by versions of the compiler prior to
10141 @item -mcpu=@var{name}
10143 This specifies the name of the target ARM processor. GCC uses this name
10144 to determine what kind of instructions it can emit when generating
10145 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10146 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10147 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10148 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10149 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10151 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10152 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10153 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10154 @samp{strongarm1110},
10155 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10156 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10157 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10158 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10159 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10160 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10161 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10162 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10163 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
10166 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10168 @item -mtune=@var{name}
10170 This option is very similar to the @option{-mcpu=} option, except that
10171 instead of specifying the actual target processor type, and hence
10172 restricting which instructions can be used, it specifies that GCC should
10173 tune the performance of the code as if the target were of the type
10174 specified in this option, but still choosing the instructions that it
10175 will generate based on the cpu specified by a @option{-mcpu=} option.
10176 For some ARM implementations better performance can be obtained by using
10179 @item -march=@var{name}
10181 This specifies the name of the target ARM architecture. GCC uses this
10182 name to determine what kind of instructions it can emit when generating
10183 assembly code. This option can be used in conjunction with or instead
10184 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10185 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10186 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10187 @samp{armv6}, @samp{armv6j},
10188 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10189 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10190 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10192 @item -mfpu=@var{name}
10193 @itemx -mfpe=@var{number}
10194 @itemx -mfp=@var{number}
10198 This specifies what floating point hardware (or hardware emulation) is
10199 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10200 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10201 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10202 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10203 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10204 @option{-mfp} and @option{-mfpe} are synonyms for
10205 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10208 If @option{-msoft-float} is specified this specifies the format of
10209 floating point values.
10211 If the selected floating-point hardware includes the NEON extension
10212 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10213 operations will not be used by GCC's auto-vectorization pass unless
10214 @option{-funsafe-math-optimizations} is also specified. This is
10215 because NEON hardware does not fully implement the IEEE 754 standard for
10216 floating-point arithmetic (in particular denormal values are treated as
10217 zero), so the use of NEON instructions may lead to a loss of precision.
10219 @item -mfp16-format=@var{name}
10220 @opindex mfp16-format
10221 Specify the format of the @code{__fp16} half-precision floating-point type.
10222 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10223 the default is @samp{none}, in which case the @code{__fp16} type is not
10224 defined. @xref{Half-Precision}, for more information.
10226 @item -mstructure-size-boundary=@var{n}
10227 @opindex mstructure-size-boundary
10228 The size of all structures and unions will be rounded up to a multiple
10229 of the number of bits set by this option. Permissible values are 8, 32
10230 and 64. The default value varies for different toolchains. For the COFF
10231 targeted toolchain the default value is 8. A value of 64 is only allowed
10232 if the underlying ABI supports it.
10234 Specifying the larger number can produce faster, more efficient code, but
10235 can also increase the size of the program. Different values are potentially
10236 incompatible. Code compiled with one value cannot necessarily expect to
10237 work with code or libraries compiled with another value, if they exchange
10238 information using structures or unions.
10240 @item -mabort-on-noreturn
10241 @opindex mabort-on-noreturn
10242 Generate a call to the function @code{abort} at the end of a
10243 @code{noreturn} function. It will be executed if the function tries to
10247 @itemx -mno-long-calls
10248 @opindex mlong-calls
10249 @opindex mno-long-calls
10250 Tells the compiler to perform function calls by first loading the
10251 address of the function into a register and then performing a subroutine
10252 call on this register. This switch is needed if the target function
10253 will lie outside of the 64 megabyte addressing range of the offset based
10254 version of subroutine call instruction.
10256 Even if this switch is enabled, not all function calls will be turned
10257 into long calls. The heuristic is that static functions, functions
10258 which have the @samp{short-call} attribute, functions that are inside
10259 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10260 definitions have already been compiled within the current compilation
10261 unit, will not be turned into long calls. The exception to this rule is
10262 that weak function definitions, functions with the @samp{long-call}
10263 attribute or the @samp{section} attribute, and functions that are within
10264 the scope of a @samp{#pragma long_calls} directive, will always be
10265 turned into long calls.
10267 This feature is not enabled by default. Specifying
10268 @option{-mno-long-calls} will restore the default behavior, as will
10269 placing the function calls within the scope of a @samp{#pragma
10270 long_calls_off} directive. Note these switches have no effect on how
10271 the compiler generates code to handle function calls via function
10274 @item -msingle-pic-base
10275 @opindex msingle-pic-base
10276 Treat the register used for PIC addressing as read-only, rather than
10277 loading it in the prologue for each function. The run-time system is
10278 responsible for initializing this register with an appropriate value
10279 before execution begins.
10281 @item -mpic-register=@var{reg}
10282 @opindex mpic-register
10283 Specify the register to be used for PIC addressing. The default is R10
10284 unless stack-checking is enabled, when R9 is used.
10286 @item -mcirrus-fix-invalid-insns
10287 @opindex mcirrus-fix-invalid-insns
10288 @opindex mno-cirrus-fix-invalid-insns
10289 Insert NOPs into the instruction stream to in order to work around
10290 problems with invalid Maverick instruction combinations. This option
10291 is only valid if the @option{-mcpu=ep9312} option has been used to
10292 enable generation of instructions for the Cirrus Maverick floating
10293 point co-processor. This option is not enabled by default, since the
10294 problem is only present in older Maverick implementations. The default
10295 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10298 @item -mpoke-function-name
10299 @opindex mpoke-function-name
10300 Write the name of each function into the text section, directly
10301 preceding the function prologue. The generated code is similar to this:
10305 .ascii "arm_poke_function_name", 0
10308 .word 0xff000000 + (t1 - t0)
10309 arm_poke_function_name
10311 stmfd sp!, @{fp, ip, lr, pc@}
10315 When performing a stack backtrace, code can inspect the value of
10316 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10317 location @code{pc - 12} and the top 8 bits are set, then we know that
10318 there is a function name embedded immediately preceding this location
10319 and has length @code{((pc[-3]) & 0xff000000)}.
10323 Generate code for the Thumb instruction set. The default is to
10324 use the 32-bit ARM instruction set.
10325 This option automatically enables either 16-bit Thumb-1 or
10326 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10327 and @option{-march=@var{name}} options. This option is not passed to the
10328 assembler. If you want to force assembler files to be interpreted as Thumb code,
10329 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10330 option directly to the assembler by prefixing it with @option{-Wa}.
10333 @opindex mtpcs-frame
10334 Generate a stack frame that is compliant with the Thumb Procedure Call
10335 Standard for all non-leaf functions. (A leaf function is one that does
10336 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10338 @item -mtpcs-leaf-frame
10339 @opindex mtpcs-leaf-frame
10340 Generate a stack frame that is compliant with the Thumb Procedure Call
10341 Standard for all leaf functions. (A leaf function is one that does
10342 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10344 @item -mcallee-super-interworking
10345 @opindex mcallee-super-interworking
10346 Gives all externally visible functions in the file being compiled an ARM
10347 instruction set header which switches to Thumb mode before executing the
10348 rest of the function. This allows these functions to be called from
10349 non-interworking code. This option is not valid in AAPCS configurations
10350 because interworking is enabled by default.
10352 @item -mcaller-super-interworking
10353 @opindex mcaller-super-interworking
10354 Allows calls via function pointers (including virtual functions) to
10355 execute correctly regardless of whether the target code has been
10356 compiled for interworking or not. There is a small overhead in the cost
10357 of executing a function pointer if this option is enabled. This option
10358 is not valid in AAPCS configurations because interworking is enabled
10361 @item -mtp=@var{name}
10363 Specify the access model for the thread local storage pointer. The valid
10364 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10365 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10366 (supported in the arm6k architecture), and @option{auto}, which uses the
10367 best available method for the selected processor. The default setting is
10370 @item -mword-relocations
10371 @opindex mword-relocations
10372 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10373 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10374 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10377 @item -mfix-cortex-m3-ldrd
10378 @opindex mfix-cortex-m3-ldrd
10379 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10380 with overlapping destination and base registers are used. This option avoids
10381 generating these instructions. This option is enabled by default when
10382 @option{-mcpu=cortex-m3} is specified.
10387 @subsection AVR Options
10388 @cindex AVR Options
10390 These options are defined for AVR implementations:
10393 @item -mmcu=@var{mcu}
10395 Specify ATMEL AVR instruction set or MCU type.
10397 Instruction set avr1 is for the minimal AVR core, not supported by the C
10398 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10399 attiny11, attiny12, attiny15, attiny28).
10401 Instruction set avr2 (default) is for the classic AVR core with up to
10402 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10403 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10404 at90c8534, at90s8535).
10406 Instruction set avr3 is for the classic AVR core with up to 128K program
10407 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10409 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10410 memory space (MCU types: atmega8, atmega83, atmega85).
10412 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10413 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10414 atmega64, atmega128, at43usb355, at94k).
10416 @item -mno-interrupts
10417 @opindex mno-interrupts
10418 Generated code is not compatible with hardware interrupts.
10419 Code size will be smaller.
10421 @item -mcall-prologues
10422 @opindex mcall-prologues
10423 Functions prologues/epilogues expanded as call to appropriate
10424 subroutines. Code size will be smaller.
10427 @opindex mtiny-stack
10428 Change only the low 8 bits of the stack pointer.
10432 Assume int to be 8 bit integer. This affects the sizes of all types: A
10433 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10434 and long long will be 4 bytes. Please note that this option does not
10435 comply to the C standards, but it will provide you with smaller code
10439 @node Blackfin Options
10440 @subsection Blackfin Options
10441 @cindex Blackfin Options
10444 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10446 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10447 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10448 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10449 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10450 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10451 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10452 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10454 The optional @var{sirevision} specifies the silicon revision of the target
10455 Blackfin processor. Any workarounds available for the targeted silicon revision
10456 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10457 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10458 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10459 hexadecimal digits representing the major and minor numbers in the silicon
10460 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10461 is not defined. If @var{sirevision} is @samp{any}, the
10462 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10463 If this optional @var{sirevision} is not used, GCC assumes the latest known
10464 silicon revision of the targeted Blackfin processor.
10466 Support for @samp{bf561} is incomplete. For @samp{bf561},
10467 Only the processor macro is defined.
10468 Without this option, @samp{bf532} is used as the processor by default.
10469 The corresponding predefined processor macros for @var{cpu} is to
10470 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10471 provided by libgloss to be linked in if @option{-msim} is not given.
10475 Specifies that the program will be run on the simulator. This causes
10476 the simulator BSP provided by libgloss to be linked in. This option
10477 has effect only for @samp{bfin-elf} toolchain.
10478 Certain other options, such as @option{-mid-shared-library} and
10479 @option{-mfdpic}, imply @option{-msim}.
10481 @item -momit-leaf-frame-pointer
10482 @opindex momit-leaf-frame-pointer
10483 Don't keep the frame pointer in a register for leaf functions. This
10484 avoids the instructions to save, set up and restore frame pointers and
10485 makes an extra register available in leaf functions. The option
10486 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10487 which might make debugging harder.
10489 @item -mspecld-anomaly
10490 @opindex mspecld-anomaly
10491 When enabled, the compiler will ensure that the generated code does not
10492 contain speculative loads after jump instructions. If this option is used,
10493 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10495 @item -mno-specld-anomaly
10496 @opindex mno-specld-anomaly
10497 Don't generate extra code to prevent speculative loads from occurring.
10499 @item -mcsync-anomaly
10500 @opindex mcsync-anomaly
10501 When enabled, the compiler will ensure that the generated code does not
10502 contain CSYNC or SSYNC instructions too soon after conditional branches.
10503 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10505 @item -mno-csync-anomaly
10506 @opindex mno-csync-anomaly
10507 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10508 occurring too soon after a conditional branch.
10512 When enabled, the compiler is free to take advantage of the knowledge that
10513 the entire program fits into the low 64k of memory.
10516 @opindex mno-low-64k
10517 Assume that the program is arbitrarily large. This is the default.
10519 @item -mstack-check-l1
10520 @opindex mstack-check-l1
10521 Do stack checking using information placed into L1 scratchpad memory by the
10524 @item -mid-shared-library
10525 @opindex mid-shared-library
10526 Generate code that supports shared libraries via the library ID method.
10527 This allows for execute in place and shared libraries in an environment
10528 without virtual memory management. This option implies @option{-fPIC}.
10529 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10531 @item -mno-id-shared-library
10532 @opindex mno-id-shared-library
10533 Generate code that doesn't assume ID based shared libraries are being used.
10534 This is the default.
10536 @item -mleaf-id-shared-library
10537 @opindex mleaf-id-shared-library
10538 Generate code that supports shared libraries via the library ID method,
10539 but assumes that this library or executable won't link against any other
10540 ID shared libraries. That allows the compiler to use faster code for jumps
10543 @item -mno-leaf-id-shared-library
10544 @opindex mno-leaf-id-shared-library
10545 Do not assume that the code being compiled won't link against any ID shared
10546 libraries. Slower code will be generated for jump and call insns.
10548 @item -mshared-library-id=n
10549 @opindex mshared-library-id
10550 Specified the identification number of the ID based shared library being
10551 compiled. Specifying a value of 0 will generate more compact code, specifying
10552 other values will force the allocation of that number to the current
10553 library but is no more space or time efficient than omitting this option.
10557 Generate code that allows the data segment to be located in a different
10558 area of memory from the text segment. This allows for execute in place in
10559 an environment without virtual memory management by eliminating relocations
10560 against the text section.
10562 @item -mno-sep-data
10563 @opindex mno-sep-data
10564 Generate code that assumes that the data segment follows the text segment.
10565 This is the default.
10568 @itemx -mno-long-calls
10569 @opindex mlong-calls
10570 @opindex mno-long-calls
10571 Tells the compiler to perform function calls by first loading the
10572 address of the function into a register and then performing a subroutine
10573 call on this register. This switch is needed if the target function
10574 will lie outside of the 24 bit addressing range of the offset based
10575 version of subroutine call instruction.
10577 This feature is not enabled by default. Specifying
10578 @option{-mno-long-calls} will restore the default behavior. Note these
10579 switches have no effect on how the compiler generates code to handle
10580 function calls via function pointers.
10584 Link with the fast floating-point library. This library relaxes some of
10585 the IEEE floating-point standard's rules for checking inputs against
10586 Not-a-Number (NAN), in the interest of performance.
10589 @opindex minline-plt
10590 Enable inlining of PLT entries in function calls to functions that are
10591 not known to bind locally. It has no effect without @option{-mfdpic}.
10594 @opindex mmulticore
10595 Build standalone application for multicore Blackfin processor. Proper
10596 start files and link scripts will be used to support multicore.
10597 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10598 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10599 @option{-mcorea} or @option{-mcoreb}. If it's used without
10600 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10601 programming model is used. In this model, the main function of Core B
10602 should be named as coreb_main. If it's used with @option{-mcorea} or
10603 @option{-mcoreb}, one application per core programming model is used.
10604 If this option is not used, single core application programming
10609 Build standalone application for Core A of BF561 when using
10610 one application per core programming model. Proper start files
10611 and link scripts will be used to support Core A. This option
10612 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10616 Build standalone application for Core B of BF561 when using
10617 one application per core programming model. Proper start files
10618 and link scripts will be used to support Core B. This option
10619 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10620 should be used instead of main. It must be used with
10621 @option{-mmulticore}.
10625 Build standalone application for SDRAM. Proper start files and
10626 link scripts will be used to put the application into SDRAM.
10627 Loader should initialize SDRAM before loading the application
10628 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10632 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10633 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10634 are enabled; for standalone applications the default is off.
10638 @subsection CRIS Options
10639 @cindex CRIS Options
10641 These options are defined specifically for the CRIS ports.
10644 @item -march=@var{architecture-type}
10645 @itemx -mcpu=@var{architecture-type}
10648 Generate code for the specified architecture. The choices for
10649 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10650 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10651 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10654 @item -mtune=@var{architecture-type}
10656 Tune to @var{architecture-type} everything applicable about the generated
10657 code, except for the ABI and the set of available instructions. The
10658 choices for @var{architecture-type} are the same as for
10659 @option{-march=@var{architecture-type}}.
10661 @item -mmax-stack-frame=@var{n}
10662 @opindex mmax-stack-frame
10663 Warn when the stack frame of a function exceeds @var{n} bytes.
10669 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10670 @option{-march=v3} and @option{-march=v8} respectively.
10672 @item -mmul-bug-workaround
10673 @itemx -mno-mul-bug-workaround
10674 @opindex mmul-bug-workaround
10675 @opindex mno-mul-bug-workaround
10676 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10677 models where it applies. This option is active by default.
10681 Enable CRIS-specific verbose debug-related information in the assembly
10682 code. This option also has the effect to turn off the @samp{#NO_APP}
10683 formatted-code indicator to the assembler at the beginning of the
10688 Do not use condition-code results from previous instruction; always emit
10689 compare and test instructions before use of condition codes.
10691 @item -mno-side-effects
10692 @opindex mno-side-effects
10693 Do not emit instructions with side-effects in addressing modes other than
10696 @item -mstack-align
10697 @itemx -mno-stack-align
10698 @itemx -mdata-align
10699 @itemx -mno-data-align
10700 @itemx -mconst-align
10701 @itemx -mno-const-align
10702 @opindex mstack-align
10703 @opindex mno-stack-align
10704 @opindex mdata-align
10705 @opindex mno-data-align
10706 @opindex mconst-align
10707 @opindex mno-const-align
10708 These options (no-options) arranges (eliminate arrangements) for the
10709 stack-frame, individual data and constants to be aligned for the maximum
10710 single data access size for the chosen CPU model. The default is to
10711 arrange for 32-bit alignment. ABI details such as structure layout are
10712 not affected by these options.
10720 Similar to the stack- data- and const-align options above, these options
10721 arrange for stack-frame, writable data and constants to all be 32-bit,
10722 16-bit or 8-bit aligned. The default is 32-bit alignment.
10724 @item -mno-prologue-epilogue
10725 @itemx -mprologue-epilogue
10726 @opindex mno-prologue-epilogue
10727 @opindex mprologue-epilogue
10728 With @option{-mno-prologue-epilogue}, the normal function prologue and
10729 epilogue that sets up the stack-frame are omitted and no return
10730 instructions or return sequences are generated in the code. Use this
10731 option only together with visual inspection of the compiled code: no
10732 warnings or errors are generated when call-saved registers must be saved,
10733 or storage for local variable needs to be allocated.
10737 @opindex mno-gotplt
10739 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10740 instruction sequences that load addresses for functions from the PLT part
10741 of the GOT rather than (traditional on other architectures) calls to the
10742 PLT@. The default is @option{-mgotplt}.
10746 Legacy no-op option only recognized with the cris-axis-elf and
10747 cris-axis-linux-gnu targets.
10751 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10755 This option, recognized for the cris-axis-elf arranges
10756 to link with input-output functions from a simulator library. Code,
10757 initialized data and zero-initialized data are allocated consecutively.
10761 Like @option{-sim}, but pass linker options to locate initialized data at
10762 0x40000000 and zero-initialized data at 0x80000000.
10766 @subsection CRX Options
10767 @cindex CRX Options
10769 These options are defined specifically for the CRX ports.
10775 Enable the use of multiply-accumulate instructions. Disabled by default.
10778 @opindex mpush-args
10779 Push instructions will be used to pass outgoing arguments when functions
10780 are called. Enabled by default.
10783 @node Darwin Options
10784 @subsection Darwin Options
10785 @cindex Darwin options
10787 These options are defined for all architectures running the Darwin operating
10790 FSF GCC on Darwin does not create ``fat'' object files; it will create
10791 an object file for the single architecture that it was built to
10792 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10793 @option{-arch} options are used; it does so by running the compiler or
10794 linker multiple times and joining the results together with
10797 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10798 @samp{i686}) is determined by the flags that specify the ISA
10799 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10800 @option{-force_cpusubtype_ALL} option can be used to override this.
10802 The Darwin tools vary in their behavior when presented with an ISA
10803 mismatch. The assembler, @file{as}, will only permit instructions to
10804 be used that are valid for the subtype of the file it is generating,
10805 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10806 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10807 and print an error if asked to create a shared library with a less
10808 restrictive subtype than its input files (for instance, trying to put
10809 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10810 for executables, @file{ld}, will quietly give the executable the most
10811 restrictive subtype of any of its input files.
10816 Add the framework directory @var{dir} to the head of the list of
10817 directories to be searched for header files. These directories are
10818 interleaved with those specified by @option{-I} options and are
10819 scanned in a left-to-right order.
10821 A framework directory is a directory with frameworks in it. A
10822 framework is a directory with a @samp{"Headers"} and/or
10823 @samp{"PrivateHeaders"} directory contained directly in it that ends
10824 in @samp{".framework"}. The name of a framework is the name of this
10825 directory excluding the @samp{".framework"}. Headers associated with
10826 the framework are found in one of those two directories, with
10827 @samp{"Headers"} being searched first. A subframework is a framework
10828 directory that is in a framework's @samp{"Frameworks"} directory.
10829 Includes of subframework headers can only appear in a header of a
10830 framework that contains the subframework, or in a sibling subframework
10831 header. Two subframeworks are siblings if they occur in the same
10832 framework. A subframework should not have the same name as a
10833 framework, a warning will be issued if this is violated. Currently a
10834 subframework cannot have subframeworks, in the future, the mechanism
10835 may be extended to support this. The standard frameworks can be found
10836 in @samp{"/System/Library/Frameworks"} and
10837 @samp{"/Library/Frameworks"}. An example include looks like
10838 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10839 the name of the framework and header.h is found in the
10840 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10842 @item -iframework@var{dir}
10843 @opindex iframework
10844 Like @option{-F} except the directory is a treated as a system
10845 directory. The main difference between this @option{-iframework} and
10846 @option{-F} is that with @option{-iframework} the compiler does not
10847 warn about constructs contained within header files found via
10848 @var{dir}. This option is valid only for the C family of languages.
10852 Emit debugging information for symbols that are used. For STABS
10853 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10854 This is by default ON@.
10858 Emit debugging information for all symbols and types.
10860 @item -mmacosx-version-min=@var{version}
10861 The earliest version of MacOS X that this executable will run on
10862 is @var{version}. Typical values of @var{version} include @code{10.1},
10863 @code{10.2}, and @code{10.3.9}.
10865 If the compiler was built to use the system's headers by default,
10866 then the default for this option is the system version on which the
10867 compiler is running, otherwise the default is to make choices which
10868 are compatible with as many systems and code bases as possible.
10872 Enable kernel development mode. The @option{-mkernel} option sets
10873 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10874 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10875 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10876 applicable. This mode also sets @option{-mno-altivec},
10877 @option{-msoft-float}, @option{-fno-builtin} and
10878 @option{-mlong-branch} for PowerPC targets.
10880 @item -mone-byte-bool
10881 @opindex mone-byte-bool
10882 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10883 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10884 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10885 option has no effect on x86.
10887 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10888 to generate code that is not binary compatible with code generated
10889 without that switch. Using this switch may require recompiling all
10890 other modules in a program, including system libraries. Use this
10891 switch to conform to a non-default data model.
10893 @item -mfix-and-continue
10894 @itemx -ffix-and-continue
10895 @itemx -findirect-data
10896 @opindex mfix-and-continue
10897 @opindex ffix-and-continue
10898 @opindex findirect-data
10899 Generate code suitable for fast turn around development. Needed to
10900 enable gdb to dynamically load @code{.o} files into already running
10901 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10902 are provided for backwards compatibility.
10906 Loads all members of static archive libraries.
10907 See man ld(1) for more information.
10909 @item -arch_errors_fatal
10910 @opindex arch_errors_fatal
10911 Cause the errors having to do with files that have the wrong architecture
10914 @item -bind_at_load
10915 @opindex bind_at_load
10916 Causes the output file to be marked such that the dynamic linker will
10917 bind all undefined references when the file is loaded or launched.
10921 Produce a Mach-o bundle format file.
10922 See man ld(1) for more information.
10924 @item -bundle_loader @var{executable}
10925 @opindex bundle_loader
10926 This option specifies the @var{executable} that will be loading the build
10927 output file being linked. See man ld(1) for more information.
10930 @opindex dynamiclib
10931 When passed this option, GCC will produce a dynamic library instead of
10932 an executable when linking, using the Darwin @file{libtool} command.
10934 @item -force_cpusubtype_ALL
10935 @opindex force_cpusubtype_ALL
10936 This causes GCC's output file to have the @var{ALL} subtype, instead of
10937 one controlled by the @option{-mcpu} or @option{-march} option.
10939 @item -allowable_client @var{client_name}
10940 @itemx -client_name
10941 @itemx -compatibility_version
10942 @itemx -current_version
10944 @itemx -dependency-file
10946 @itemx -dylinker_install_name
10948 @itemx -exported_symbols_list
10951 @itemx -flat_namespace
10952 @itemx -force_flat_namespace
10953 @itemx -headerpad_max_install_names
10956 @itemx -install_name
10957 @itemx -keep_private_externs
10958 @itemx -multi_module
10959 @itemx -multiply_defined
10960 @itemx -multiply_defined_unused
10963 @itemx -no_dead_strip_inits_and_terms
10964 @itemx -nofixprebinding
10965 @itemx -nomultidefs
10967 @itemx -noseglinkedit
10968 @itemx -pagezero_size
10970 @itemx -prebind_all_twolevel_modules
10971 @itemx -private_bundle
10973 @itemx -read_only_relocs
10975 @itemx -sectobjectsymbols
10979 @itemx -sectobjectsymbols
10982 @itemx -segs_read_only_addr
10984 @itemx -segs_read_write_addr
10985 @itemx -seg_addr_table
10986 @itemx -seg_addr_table_filename
10987 @itemx -seglinkedit
10989 @itemx -segs_read_only_addr
10990 @itemx -segs_read_write_addr
10991 @itemx -single_module
10993 @itemx -sub_library
10995 @itemx -sub_umbrella
10996 @itemx -twolevel_namespace
10999 @itemx -unexported_symbols_list
11000 @itemx -weak_reference_mismatches
11001 @itemx -whatsloaded
11002 @opindex allowable_client
11003 @opindex client_name
11004 @opindex compatibility_version
11005 @opindex current_version
11006 @opindex dead_strip
11007 @opindex dependency-file
11008 @opindex dylib_file
11009 @opindex dylinker_install_name
11011 @opindex exported_symbols_list
11013 @opindex flat_namespace
11014 @opindex force_flat_namespace
11015 @opindex headerpad_max_install_names
11016 @opindex image_base
11018 @opindex install_name
11019 @opindex keep_private_externs
11020 @opindex multi_module
11021 @opindex multiply_defined
11022 @opindex multiply_defined_unused
11023 @opindex noall_load
11024 @opindex no_dead_strip_inits_and_terms
11025 @opindex nofixprebinding
11026 @opindex nomultidefs
11028 @opindex noseglinkedit
11029 @opindex pagezero_size
11031 @opindex prebind_all_twolevel_modules
11032 @opindex private_bundle
11033 @opindex read_only_relocs
11035 @opindex sectobjectsymbols
11038 @opindex sectcreate
11039 @opindex sectobjectsymbols
11042 @opindex segs_read_only_addr
11043 @opindex segs_read_write_addr
11044 @opindex seg_addr_table
11045 @opindex seg_addr_table_filename
11046 @opindex seglinkedit
11048 @opindex segs_read_only_addr
11049 @opindex segs_read_write_addr
11050 @opindex single_module
11052 @opindex sub_library
11053 @opindex sub_umbrella
11054 @opindex twolevel_namespace
11057 @opindex unexported_symbols_list
11058 @opindex weak_reference_mismatches
11059 @opindex whatsloaded
11060 These options are passed to the Darwin linker. The Darwin linker man page
11061 describes them in detail.
11064 @node DEC Alpha Options
11065 @subsection DEC Alpha Options
11067 These @samp{-m} options are defined for the DEC Alpha implementations:
11070 @item -mno-soft-float
11071 @itemx -msoft-float
11072 @opindex mno-soft-float
11073 @opindex msoft-float
11074 Use (do not use) the hardware floating-point instructions for
11075 floating-point operations. When @option{-msoft-float} is specified,
11076 functions in @file{libgcc.a} will be used to perform floating-point
11077 operations. Unless they are replaced by routines that emulate the
11078 floating-point operations, or compiled in such a way as to call such
11079 emulations routines, these routines will issue floating-point
11080 operations. If you are compiling for an Alpha without floating-point
11081 operations, you must ensure that the library is built so as not to call
11084 Note that Alpha implementations without floating-point operations are
11085 required to have floating-point registers.
11088 @itemx -mno-fp-regs
11090 @opindex mno-fp-regs
11091 Generate code that uses (does not use) the floating-point register set.
11092 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11093 register set is not used, floating point operands are passed in integer
11094 registers as if they were integers and floating-point results are passed
11095 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11096 so any function with a floating-point argument or return value called by code
11097 compiled with @option{-mno-fp-regs} must also be compiled with that
11100 A typical use of this option is building a kernel that does not use,
11101 and hence need not save and restore, any floating-point registers.
11105 The Alpha architecture implements floating-point hardware optimized for
11106 maximum performance. It is mostly compliant with the IEEE floating
11107 point standard. However, for full compliance, software assistance is
11108 required. This option generates code fully IEEE compliant code
11109 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11110 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11111 defined during compilation. The resulting code is less efficient but is
11112 able to correctly support denormalized numbers and exceptional IEEE
11113 values such as not-a-number and plus/minus infinity. Other Alpha
11114 compilers call this option @option{-ieee_with_no_inexact}.
11116 @item -mieee-with-inexact
11117 @opindex mieee-with-inexact
11118 This is like @option{-mieee} except the generated code also maintains
11119 the IEEE @var{inexact-flag}. Turning on this option causes the
11120 generated code to implement fully-compliant IEEE math. In addition to
11121 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11122 macro. On some Alpha implementations the resulting code may execute
11123 significantly slower than the code generated by default. Since there is
11124 very little code that depends on the @var{inexact-flag}, you should
11125 normally not specify this option. Other Alpha compilers call this
11126 option @option{-ieee_with_inexact}.
11128 @item -mfp-trap-mode=@var{trap-mode}
11129 @opindex mfp-trap-mode
11130 This option controls what floating-point related traps are enabled.
11131 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11132 The trap mode can be set to one of four values:
11136 This is the default (normal) setting. The only traps that are enabled
11137 are the ones that cannot be disabled in software (e.g., division by zero
11141 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11145 Like @samp{u}, but the instructions are marked to be safe for software
11146 completion (see Alpha architecture manual for details).
11149 Like @samp{su}, but inexact traps are enabled as well.
11152 @item -mfp-rounding-mode=@var{rounding-mode}
11153 @opindex mfp-rounding-mode
11154 Selects the IEEE rounding mode. Other Alpha compilers call this option
11155 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11160 Normal IEEE rounding mode. Floating point numbers are rounded towards
11161 the nearest machine number or towards the even machine number in case
11165 Round towards minus infinity.
11168 Chopped rounding mode. Floating point numbers are rounded towards zero.
11171 Dynamic rounding mode. A field in the floating point control register
11172 (@var{fpcr}, see Alpha architecture reference manual) controls the
11173 rounding mode in effect. The C library initializes this register for
11174 rounding towards plus infinity. Thus, unless your program modifies the
11175 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11178 @item -mtrap-precision=@var{trap-precision}
11179 @opindex mtrap-precision
11180 In the Alpha architecture, floating point traps are imprecise. This
11181 means without software assistance it is impossible to recover from a
11182 floating trap and program execution normally needs to be terminated.
11183 GCC can generate code that can assist operating system trap handlers
11184 in determining the exact location that caused a floating point trap.
11185 Depending on the requirements of an application, different levels of
11186 precisions can be selected:
11190 Program precision. This option is the default and means a trap handler
11191 can only identify which program caused a floating point exception.
11194 Function precision. The trap handler can determine the function that
11195 caused a floating point exception.
11198 Instruction precision. The trap handler can determine the exact
11199 instruction that caused a floating point exception.
11202 Other Alpha compilers provide the equivalent options called
11203 @option{-scope_safe} and @option{-resumption_safe}.
11205 @item -mieee-conformant
11206 @opindex mieee-conformant
11207 This option marks the generated code as IEEE conformant. You must not
11208 use this option unless you also specify @option{-mtrap-precision=i} and either
11209 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11210 is to emit the line @samp{.eflag 48} in the function prologue of the
11211 generated assembly file. Under DEC Unix, this has the effect that
11212 IEEE-conformant math library routines will be linked in.
11214 @item -mbuild-constants
11215 @opindex mbuild-constants
11216 Normally GCC examines a 32- or 64-bit integer constant to
11217 see if it can construct it from smaller constants in two or three
11218 instructions. If it cannot, it will output the constant as a literal and
11219 generate code to load it from the data segment at runtime.
11221 Use this option to require GCC to construct @emph{all} integer constants
11222 using code, even if it takes more instructions (the maximum is six).
11224 You would typically use this option to build a shared library dynamic
11225 loader. Itself a shared library, it must relocate itself in memory
11226 before it can find the variables and constants in its own data segment.
11232 Select whether to generate code to be assembled by the vendor-supplied
11233 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11251 Indicate whether GCC should generate code to use the optional BWX,
11252 CIX, FIX and MAX instruction sets. The default is to use the instruction
11253 sets supported by the CPU type specified via @option{-mcpu=} option or that
11254 of the CPU on which GCC was built if none was specified.
11257 @itemx -mfloat-ieee
11258 @opindex mfloat-vax
11259 @opindex mfloat-ieee
11260 Generate code that uses (does not use) VAX F and G floating point
11261 arithmetic instead of IEEE single and double precision.
11263 @item -mexplicit-relocs
11264 @itemx -mno-explicit-relocs
11265 @opindex mexplicit-relocs
11266 @opindex mno-explicit-relocs
11267 Older Alpha assemblers provided no way to generate symbol relocations
11268 except via assembler macros. Use of these macros does not allow
11269 optimal instruction scheduling. GNU binutils as of version 2.12
11270 supports a new syntax that allows the compiler to explicitly mark
11271 which relocations should apply to which instructions. This option
11272 is mostly useful for debugging, as GCC detects the capabilities of
11273 the assembler when it is built and sets the default accordingly.
11276 @itemx -mlarge-data
11277 @opindex msmall-data
11278 @opindex mlarge-data
11279 When @option{-mexplicit-relocs} is in effect, static data is
11280 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11281 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11282 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11283 16-bit relocations off of the @code{$gp} register. This limits the
11284 size of the small data area to 64KB, but allows the variables to be
11285 directly accessed via a single instruction.
11287 The default is @option{-mlarge-data}. With this option the data area
11288 is limited to just below 2GB@. Programs that require more than 2GB of
11289 data must use @code{malloc} or @code{mmap} to allocate the data in the
11290 heap instead of in the program's data segment.
11292 When generating code for shared libraries, @option{-fpic} implies
11293 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11296 @itemx -mlarge-text
11297 @opindex msmall-text
11298 @opindex mlarge-text
11299 When @option{-msmall-text} is used, the compiler assumes that the
11300 code of the entire program (or shared library) fits in 4MB, and is
11301 thus reachable with a branch instruction. When @option{-msmall-data}
11302 is used, the compiler can assume that all local symbols share the
11303 same @code{$gp} value, and thus reduce the number of instructions
11304 required for a function call from 4 to 1.
11306 The default is @option{-mlarge-text}.
11308 @item -mcpu=@var{cpu_type}
11310 Set the instruction set and instruction scheduling parameters for
11311 machine type @var{cpu_type}. You can specify either the @samp{EV}
11312 style name or the corresponding chip number. GCC supports scheduling
11313 parameters for the EV4, EV5 and EV6 family of processors and will
11314 choose the default values for the instruction set from the processor
11315 you specify. If you do not specify a processor type, GCC will default
11316 to the processor on which the compiler was built.
11318 Supported values for @var{cpu_type} are
11324 Schedules as an EV4 and has no instruction set extensions.
11328 Schedules as an EV5 and has no instruction set extensions.
11332 Schedules as an EV5 and supports the BWX extension.
11337 Schedules as an EV5 and supports the BWX and MAX extensions.
11341 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11345 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11348 Native Linux/GNU toolchains also support the value @samp{native},
11349 which selects the best architecture option for the host processor.
11350 @option{-mcpu=native} has no effect if GCC does not recognize
11353 @item -mtune=@var{cpu_type}
11355 Set only the instruction scheduling parameters for machine type
11356 @var{cpu_type}. The instruction set is not changed.
11358 Native Linux/GNU toolchains also support the value @samp{native},
11359 which selects the best architecture option for the host processor.
11360 @option{-mtune=native} has no effect if GCC does not recognize
11363 @item -mmemory-latency=@var{time}
11364 @opindex mmemory-latency
11365 Sets the latency the scheduler should assume for typical memory
11366 references as seen by the application. This number is highly
11367 dependent on the memory access patterns used by the application
11368 and the size of the external cache on the machine.
11370 Valid options for @var{time} are
11374 A decimal number representing clock cycles.
11380 The compiler contains estimates of the number of clock cycles for
11381 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11382 (also called Dcache, Scache, and Bcache), as well as to main memory.
11383 Note that L3 is only valid for EV5.
11388 @node DEC Alpha/VMS Options
11389 @subsection DEC Alpha/VMS Options
11391 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11394 @item -mvms-return-codes
11395 @opindex mvms-return-codes
11396 Return VMS condition codes from main. The default is to return POSIX
11397 style condition (e.g.@: error) codes.
11399 @item -mdebug-main=@var{prefix}
11400 @opindex mdebug-main=@var{prefix}
11401 Flag the first routine whose name starts with @var{prefix} as the main
11402 routine for the debugger.
11406 Default to 64bit memory allocation routines.
11410 @subsection FR30 Options
11411 @cindex FR30 Options
11413 These options are defined specifically for the FR30 port.
11417 @item -msmall-model
11418 @opindex msmall-model
11419 Use the small address space model. This can produce smaller code, but
11420 it does assume that all symbolic values and addresses will fit into a
11425 Assume that run-time support has been provided and so there is no need
11426 to include the simulator library (@file{libsim.a}) on the linker
11432 @subsection FRV Options
11433 @cindex FRV Options
11439 Only use the first 32 general purpose registers.
11444 Use all 64 general purpose registers.
11449 Use only the first 32 floating point registers.
11454 Use all 64 floating point registers
11457 @opindex mhard-float
11459 Use hardware instructions for floating point operations.
11462 @opindex msoft-float
11464 Use library routines for floating point operations.
11469 Dynamically allocate condition code registers.
11474 Do not try to dynamically allocate condition code registers, only
11475 use @code{icc0} and @code{fcc0}.
11480 Change ABI to use double word insns.
11485 Do not use double word instructions.
11490 Use floating point double instructions.
11493 @opindex mno-double
11495 Do not use floating point double instructions.
11500 Use media instructions.
11505 Do not use media instructions.
11510 Use multiply and add/subtract instructions.
11513 @opindex mno-muladd
11515 Do not use multiply and add/subtract instructions.
11520 Select the FDPIC ABI, that uses function descriptors to represent
11521 pointers to functions. Without any PIC/PIE-related options, it
11522 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11523 assumes GOT entries and small data are within a 12-bit range from the
11524 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11525 are computed with 32 bits.
11526 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11529 @opindex minline-plt
11531 Enable inlining of PLT entries in function calls to functions that are
11532 not known to bind locally. It has no effect without @option{-mfdpic}.
11533 It's enabled by default if optimizing for speed and compiling for
11534 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11535 optimization option such as @option{-O3} or above is present in the
11541 Assume a large TLS segment when generating thread-local code.
11546 Do not assume a large TLS segment when generating thread-local code.
11551 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11552 that is known to be in read-only sections. It's enabled by default,
11553 except for @option{-fpic} or @option{-fpie}: even though it may help
11554 make the global offset table smaller, it trades 1 instruction for 4.
11555 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11556 one of which may be shared by multiple symbols, and it avoids the need
11557 for a GOT entry for the referenced symbol, so it's more likely to be a
11558 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11560 @item -multilib-library-pic
11561 @opindex multilib-library-pic
11563 Link with the (library, not FD) pic libraries. It's implied by
11564 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11565 @option{-fpic} without @option{-mfdpic}. You should never have to use
11569 @opindex mlinked-fp
11571 Follow the EABI requirement of always creating a frame pointer whenever
11572 a stack frame is allocated. This option is enabled by default and can
11573 be disabled with @option{-mno-linked-fp}.
11576 @opindex mlong-calls
11578 Use indirect addressing to call functions outside the current
11579 compilation unit. This allows the functions to be placed anywhere
11580 within the 32-bit address space.
11582 @item -malign-labels
11583 @opindex malign-labels
11585 Try to align labels to an 8-byte boundary by inserting nops into the
11586 previous packet. This option only has an effect when VLIW packing
11587 is enabled. It doesn't create new packets; it merely adds nops to
11590 @item -mlibrary-pic
11591 @opindex mlibrary-pic
11593 Generate position-independent EABI code.
11598 Use only the first four media accumulator registers.
11603 Use all eight media accumulator registers.
11608 Pack VLIW instructions.
11613 Do not pack VLIW instructions.
11616 @opindex mno-eflags
11618 Do not mark ABI switches in e_flags.
11621 @opindex mcond-move
11623 Enable the use of conditional-move instructions (default).
11625 This switch is mainly for debugging the compiler and will likely be removed
11626 in a future version.
11628 @item -mno-cond-move
11629 @opindex mno-cond-move
11631 Disable the use of conditional-move instructions.
11633 This switch is mainly for debugging the compiler and will likely be removed
11634 in a future version.
11639 Enable the use of conditional set instructions (default).
11641 This switch is mainly for debugging the compiler and will likely be removed
11642 in a future version.
11647 Disable the use of conditional set instructions.
11649 This switch is mainly for debugging the compiler and will likely be removed
11650 in a future version.
11653 @opindex mcond-exec
11655 Enable the use of conditional execution (default).
11657 This switch is mainly for debugging the compiler and will likely be removed
11658 in a future version.
11660 @item -mno-cond-exec
11661 @opindex mno-cond-exec
11663 Disable the use of conditional execution.
11665 This switch is mainly for debugging the compiler and will likely be removed
11666 in a future version.
11668 @item -mvliw-branch
11669 @opindex mvliw-branch
11671 Run a pass to pack branches into VLIW instructions (default).
11673 This switch is mainly for debugging the compiler and will likely be removed
11674 in a future version.
11676 @item -mno-vliw-branch
11677 @opindex mno-vliw-branch
11679 Do not run a pass to pack branches into VLIW instructions.
11681 This switch is mainly for debugging the compiler and will likely be removed
11682 in a future version.
11684 @item -mmulti-cond-exec
11685 @opindex mmulti-cond-exec
11687 Enable optimization of @code{&&} and @code{||} in conditional execution
11690 This switch is mainly for debugging the compiler and will likely be removed
11691 in a future version.
11693 @item -mno-multi-cond-exec
11694 @opindex mno-multi-cond-exec
11696 Disable optimization of @code{&&} and @code{||} in conditional execution.
11698 This switch is mainly for debugging the compiler and will likely be removed
11699 in a future version.
11701 @item -mnested-cond-exec
11702 @opindex mnested-cond-exec
11704 Enable nested conditional execution optimizations (default).
11706 This switch is mainly for debugging the compiler and will likely be removed
11707 in a future version.
11709 @item -mno-nested-cond-exec
11710 @opindex mno-nested-cond-exec
11712 Disable nested conditional execution optimizations.
11714 This switch is mainly for debugging the compiler and will likely be removed
11715 in a future version.
11717 @item -moptimize-membar
11718 @opindex moptimize-membar
11720 This switch removes redundant @code{membar} instructions from the
11721 compiler generated code. It is enabled by default.
11723 @item -mno-optimize-membar
11724 @opindex mno-optimize-membar
11726 This switch disables the automatic removal of redundant @code{membar}
11727 instructions from the generated code.
11729 @item -mtomcat-stats
11730 @opindex mtomcat-stats
11732 Cause gas to print out tomcat statistics.
11734 @item -mcpu=@var{cpu}
11737 Select the processor type for which to generate code. Possible values are
11738 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11739 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11743 @node GNU/Linux Options
11744 @subsection GNU/Linux Options
11746 These @samp{-m} options are defined for GNU/Linux targets:
11751 Use the GNU C library. This is the default except
11752 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11756 Use uClibc C library. This is the default on
11757 @samp{*-*-linux-*uclibc*} targets.
11761 Use Bionic C library. This is the default on
11762 @samp{*-*-linux-*android*} targets.
11766 Compile code compatible with Android platform. This is the default on
11767 @samp{*-*-linux-*android*} targets.
11769 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11770 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11771 this option makes the GCC driver pass Android-specific options to the linker.
11772 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11775 @item -tno-android-cc
11776 @opindex tno-android-cc
11777 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11778 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11779 @option{-fno-rtti} by default.
11781 @item -tno-android-ld
11782 @opindex tno-android-ld
11783 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11784 linking options to the linker.
11788 @node H8/300 Options
11789 @subsection H8/300 Options
11791 These @samp{-m} options are defined for the H8/300 implementations:
11796 Shorten some address references at link time, when possible; uses the
11797 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11798 ld, Using ld}, for a fuller description.
11802 Generate code for the H8/300H@.
11806 Generate code for the H8S@.
11810 Generate code for the H8S and H8/300H in the normal mode. This switch
11811 must be used either with @option{-mh} or @option{-ms}.
11815 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11819 Make @code{int} data 32 bits by default.
11822 @opindex malign-300
11823 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11824 The default for the H8/300H and H8S is to align longs and floats on 4
11826 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11827 This option has no effect on the H8/300.
11831 @subsection HPPA Options
11832 @cindex HPPA Options
11834 These @samp{-m} options are defined for the HPPA family of computers:
11837 @item -march=@var{architecture-type}
11839 Generate code for the specified architecture. The choices for
11840 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11841 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11842 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11843 architecture option for your machine. Code compiled for lower numbered
11844 architectures will run on higher numbered architectures, but not the
11847 @item -mpa-risc-1-0
11848 @itemx -mpa-risc-1-1
11849 @itemx -mpa-risc-2-0
11850 @opindex mpa-risc-1-0
11851 @opindex mpa-risc-1-1
11852 @opindex mpa-risc-2-0
11853 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11856 @opindex mbig-switch
11857 Generate code suitable for big switch tables. Use this option only if
11858 the assembler/linker complain about out of range branches within a switch
11861 @item -mjump-in-delay
11862 @opindex mjump-in-delay
11863 Fill delay slots of function calls with unconditional jump instructions
11864 by modifying the return pointer for the function call to be the target
11865 of the conditional jump.
11867 @item -mdisable-fpregs
11868 @opindex mdisable-fpregs
11869 Prevent floating point registers from being used in any manner. This is
11870 necessary for compiling kernels which perform lazy context switching of
11871 floating point registers. If you use this option and attempt to perform
11872 floating point operations, the compiler will abort.
11874 @item -mdisable-indexing
11875 @opindex mdisable-indexing
11876 Prevent the compiler from using indexing address modes. This avoids some
11877 rather obscure problems when compiling MIG generated code under MACH@.
11879 @item -mno-space-regs
11880 @opindex mno-space-regs
11881 Generate code that assumes the target has no space registers. This allows
11882 GCC to generate faster indirect calls and use unscaled index address modes.
11884 Such code is suitable for level 0 PA systems and kernels.
11886 @item -mfast-indirect-calls
11887 @opindex mfast-indirect-calls
11888 Generate code that assumes calls never cross space boundaries. This
11889 allows GCC to emit code which performs faster indirect calls.
11891 This option will not work in the presence of shared libraries or nested
11894 @item -mfixed-range=@var{register-range}
11895 @opindex mfixed-range
11896 Generate code treating the given register range as fixed registers.
11897 A fixed register is one that the register allocator can not use. This is
11898 useful when compiling kernel code. A register range is specified as
11899 two registers separated by a dash. Multiple register ranges can be
11900 specified separated by a comma.
11902 @item -mlong-load-store
11903 @opindex mlong-load-store
11904 Generate 3-instruction load and store sequences as sometimes required by
11905 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11908 @item -mportable-runtime
11909 @opindex mportable-runtime
11910 Use the portable calling conventions proposed by HP for ELF systems.
11914 Enable the use of assembler directives only GAS understands.
11916 @item -mschedule=@var{cpu-type}
11918 Schedule code according to the constraints for the machine type
11919 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11920 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11921 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11922 proper scheduling option for your machine. The default scheduling is
11926 @opindex mlinker-opt
11927 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11928 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11929 linkers in which they give bogus error messages when linking some programs.
11932 @opindex msoft-float
11933 Generate output containing library calls for floating point.
11934 @strong{Warning:} the requisite libraries are not available for all HPPA
11935 targets. Normally the facilities of the machine's usual C compiler are
11936 used, but this cannot be done directly in cross-compilation. You must make
11937 your own arrangements to provide suitable library functions for
11940 @option{-msoft-float} changes the calling convention in the output file;
11941 therefore, it is only useful if you compile @emph{all} of a program with
11942 this option. In particular, you need to compile @file{libgcc.a}, the
11943 library that comes with GCC, with @option{-msoft-float} in order for
11948 Generate the predefine, @code{_SIO}, for server IO@. The default is
11949 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11950 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11951 options are available under HP-UX and HI-UX@.
11955 Use GNU ld specific options. This passes @option{-shared} to ld when
11956 building a shared library. It is the default when GCC is configured,
11957 explicitly or implicitly, with the GNU linker. This option does not
11958 have any affect on which ld is called, it only changes what parameters
11959 are passed to that ld. The ld that is called is determined by the
11960 @option{--with-ld} configure option, GCC's program search path, and
11961 finally by the user's @env{PATH}. The linker used by GCC can be printed
11962 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11963 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11967 Use HP ld specific options. This passes @option{-b} to ld when building
11968 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11969 links. It is the default when GCC is configured, explicitly or
11970 implicitly, with the HP linker. This option does not have any affect on
11971 which ld is called, it only changes what parameters are passed to that
11972 ld. The ld that is called is determined by the @option{--with-ld}
11973 configure option, GCC's program search path, and finally by the user's
11974 @env{PATH}. The linker used by GCC can be printed using @samp{which
11975 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11976 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11979 @opindex mno-long-calls
11980 Generate code that uses long call sequences. This ensures that a call
11981 is always able to reach linker generated stubs. The default is to generate
11982 long calls only when the distance from the call site to the beginning
11983 of the function or translation unit, as the case may be, exceeds a
11984 predefined limit set by the branch type being used. The limits for
11985 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11986 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11989 Distances are measured from the beginning of functions when using the
11990 @option{-ffunction-sections} option, or when using the @option{-mgas}
11991 and @option{-mno-portable-runtime} options together under HP-UX with
11994 It is normally not desirable to use this option as it will degrade
11995 performance. However, it may be useful in large applications,
11996 particularly when partial linking is used to build the application.
11998 The types of long calls used depends on the capabilities of the
11999 assembler and linker, and the type of code being generated. The
12000 impact on systems that support long absolute calls, and long pic
12001 symbol-difference or pc-relative calls should be relatively small.
12002 However, an indirect call is used on 32-bit ELF systems in pic code
12003 and it is quite long.
12005 @item -munix=@var{unix-std}
12007 Generate compiler predefines and select a startfile for the specified
12008 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12009 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12010 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12011 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12012 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12015 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12016 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12017 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12018 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12019 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12020 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12022 It is @emph{important} to note that this option changes the interfaces
12023 for various library routines. It also affects the operational behavior
12024 of the C library. Thus, @emph{extreme} care is needed in using this
12027 Library code that is intended to operate with more than one UNIX
12028 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12029 as appropriate. Most GNU software doesn't provide this capability.
12033 Suppress the generation of link options to search libdld.sl when the
12034 @option{-static} option is specified on HP-UX 10 and later.
12038 The HP-UX implementation of setlocale in libc has a dependency on
12039 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12040 when the @option{-static} option is specified, special link options
12041 are needed to resolve this dependency.
12043 On HP-UX 10 and later, the GCC driver adds the necessary options to
12044 link with libdld.sl when the @option{-static} option is specified.
12045 This causes the resulting binary to be dynamic. On the 64-bit port,
12046 the linkers generate dynamic binaries by default in any case. The
12047 @option{-nolibdld} option can be used to prevent the GCC driver from
12048 adding these link options.
12052 Add support for multithreading with the @dfn{dce thread} library
12053 under HP-UX@. This option sets flags for both the preprocessor and
12057 @node i386 and x86-64 Options
12058 @subsection Intel 386 and AMD x86-64 Options
12059 @cindex i386 Options
12060 @cindex x86-64 Options
12061 @cindex Intel 386 Options
12062 @cindex AMD x86-64 Options
12064 These @samp{-m} options are defined for the i386 and x86-64 family of
12068 @item -mtune=@var{cpu-type}
12070 Tune to @var{cpu-type} everything applicable about the generated code, except
12071 for the ABI and the set of available instructions. The choices for
12072 @var{cpu-type} are:
12075 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12076 If you know the CPU on which your code will run, then you should use
12077 the corresponding @option{-mtune} option instead of
12078 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12079 of your application will have, then you should use this option.
12081 As new processors are deployed in the marketplace, the behavior of this
12082 option will change. Therefore, if you upgrade to a newer version of
12083 GCC, the code generated option will change to reflect the processors
12084 that were most common when that version of GCC was released.
12086 There is no @option{-march=generic} option because @option{-march}
12087 indicates the instruction set the compiler can use, and there is no
12088 generic instruction set applicable to all processors. In contrast,
12089 @option{-mtune} indicates the processor (or, in this case, collection of
12090 processors) for which the code is optimized.
12092 This selects the CPU to tune for at compilation time by determining
12093 the processor type of the compiling machine. Using @option{-mtune=native}
12094 will produce code optimized for the local machine under the constraints
12095 of the selected instruction set. Using @option{-march=native} will
12096 enable all instruction subsets supported by the local machine (hence
12097 the result might not run on different machines).
12099 Original Intel's i386 CPU@.
12101 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12102 @item i586, pentium
12103 Intel Pentium CPU with no MMX support.
12105 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12107 Intel PentiumPro CPU@.
12109 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12110 instruction set will be used, so the code will run on all i686 family chips.
12112 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12113 @item pentium3, pentium3m
12114 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12117 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12118 support. Used by Centrino notebooks.
12119 @item pentium4, pentium4m
12120 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12122 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12125 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12126 SSE2 and SSE3 instruction set support.
12128 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12129 instruction set support.
12131 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12132 and SSE4.2 instruction set support.
12134 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12135 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12137 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12138 instruction set support.
12140 AMD K6 CPU with MMX instruction set support.
12142 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12143 @item athlon, athlon-tbird
12144 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12146 @item athlon-4, athlon-xp, athlon-mp
12147 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12148 instruction set support.
12149 @item k8, opteron, athlon64, athlon-fx
12150 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12151 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12152 @item k8-sse3, opteron-sse3, athlon64-sse3
12153 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12154 @item amdfam10, barcelona
12155 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12156 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12157 instruction set extensions.)
12159 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12162 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12163 instruction set support.
12165 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12166 implemented for this chip.)
12168 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12169 implemented for this chip.)
12171 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12174 While picking a specific @var{cpu-type} will schedule things appropriately
12175 for that particular chip, the compiler will not generate any code that
12176 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12179 @item -march=@var{cpu-type}
12181 Generate instructions for the machine type @var{cpu-type}. The choices
12182 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12183 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12185 @item -mcpu=@var{cpu-type}
12187 A deprecated synonym for @option{-mtune}.
12189 @item -mfpmath=@var{unit}
12191 Generate floating point arithmetics for selected unit @var{unit}. The choices
12192 for @var{unit} are:
12196 Use the standard 387 floating point coprocessor present majority of chips and
12197 emulated otherwise. Code compiled with this option will run almost everywhere.
12198 The temporary results are computed in 80bit precision instead of precision
12199 specified by the type resulting in slightly different results compared to most
12200 of other chips. See @option{-ffloat-store} for more detailed description.
12202 This is the default choice for i386 compiler.
12205 Use scalar floating point instructions present in the SSE instruction set.
12206 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12207 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12208 instruction set supports only single precision arithmetics, thus the double and
12209 extended precision arithmetics is still done using 387. Later version, present
12210 only in Pentium4 and the future AMD x86-64 chips supports double precision
12213 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12214 or @option{-msse2} switches to enable SSE extensions and make this option
12215 effective. For the x86-64 compiler, these extensions are enabled by default.
12217 The resulting code should be considerably faster in the majority of cases and avoid
12218 the numerical instability problems of 387 code, but may break some existing
12219 code that expects temporaries to be 80bit.
12221 This is the default choice for the x86-64 compiler.
12226 Attempt to utilize both instruction sets at once. This effectively double the
12227 amount of available registers and on chips with separate execution units for
12228 387 and SSE the execution resources too. Use this option with care, as it is
12229 still experimental, because the GCC register allocator does not model separate
12230 functional units well resulting in instable performance.
12233 @item -masm=@var{dialect}
12234 @opindex masm=@var{dialect}
12235 Output asm instructions using selected @var{dialect}. Supported
12236 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12237 not support @samp{intel}.
12240 @itemx -mno-ieee-fp
12242 @opindex mno-ieee-fp
12243 Control whether or not the compiler uses IEEE floating point
12244 comparisons. These handle correctly the case where the result of a
12245 comparison is unordered.
12248 @opindex msoft-float
12249 Generate output containing library calls for floating point.
12250 @strong{Warning:} the requisite libraries are not part of GCC@.
12251 Normally the facilities of the machine's usual C compiler are used, but
12252 this can't be done directly in cross-compilation. You must make your
12253 own arrangements to provide suitable library functions for
12256 On machines where a function returns floating point results in the 80387
12257 register stack, some floating point opcodes may be emitted even if
12258 @option{-msoft-float} is used.
12260 @item -mno-fp-ret-in-387
12261 @opindex mno-fp-ret-in-387
12262 Do not use the FPU registers for return values of functions.
12264 The usual calling convention has functions return values of types
12265 @code{float} and @code{double} in an FPU register, even if there
12266 is no FPU@. The idea is that the operating system should emulate
12269 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12270 in ordinary CPU registers instead.
12272 @item -mno-fancy-math-387
12273 @opindex mno-fancy-math-387
12274 Some 387 emulators do not support the @code{sin}, @code{cos} and
12275 @code{sqrt} instructions for the 387. Specify this option to avoid
12276 generating those instructions. This option is the default on FreeBSD,
12277 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12278 indicates that the target cpu will always have an FPU and so the
12279 instruction will not need emulation. As of revision 2.6.1, these
12280 instructions are not generated unless you also use the
12281 @option{-funsafe-math-optimizations} switch.
12283 @item -malign-double
12284 @itemx -mno-align-double
12285 @opindex malign-double
12286 @opindex mno-align-double
12287 Control whether GCC aligns @code{double}, @code{long double}, and
12288 @code{long long} variables on a two word boundary or a one word
12289 boundary. Aligning @code{double} variables on a two word boundary will
12290 produce code that runs somewhat faster on a @samp{Pentium} at the
12291 expense of more memory.
12293 On x86-64, @option{-malign-double} is enabled by default.
12295 @strong{Warning:} if you use the @option{-malign-double} switch,
12296 structures containing the above types will be aligned differently than
12297 the published application binary interface specifications for the 386
12298 and will not be binary compatible with structures in code compiled
12299 without that switch.
12301 @item -m96bit-long-double
12302 @itemx -m128bit-long-double
12303 @opindex m96bit-long-double
12304 @opindex m128bit-long-double
12305 These switches control the size of @code{long double} type. The i386
12306 application binary interface specifies the size to be 96 bits,
12307 so @option{-m96bit-long-double} is the default in 32 bit mode.
12309 Modern architectures (Pentium and newer) would prefer @code{long double}
12310 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12311 conforming to the ABI, this would not be possible. So specifying a
12312 @option{-m128bit-long-double} will align @code{long double}
12313 to a 16 byte boundary by padding the @code{long double} with an additional
12316 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12317 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12319 Notice that neither of these options enable any extra precision over the x87
12320 standard of 80 bits for a @code{long double}.
12322 @strong{Warning:} if you override the default value for your target ABI, the
12323 structures and arrays containing @code{long double} variables will change
12324 their size as well as function calling convention for function taking
12325 @code{long double} will be modified. Hence they will not be binary
12326 compatible with arrays or structures in code compiled without that switch.
12328 @item -mlarge-data-threshold=@var{number}
12329 @opindex mlarge-data-threshold=@var{number}
12330 When @option{-mcmodel=medium} is specified, the data greater than
12331 @var{threshold} are placed in large data section. This value must be the
12332 same across all object linked into the binary and defaults to 65535.
12336 Use a different function-calling convention, in which functions that
12337 take a fixed number of arguments return with the @code{ret} @var{num}
12338 instruction, which pops their arguments while returning. This saves one
12339 instruction in the caller since there is no need to pop the arguments
12342 You can specify that an individual function is called with this calling
12343 sequence with the function attribute @samp{stdcall}. You can also
12344 override the @option{-mrtd} option by using the function attribute
12345 @samp{cdecl}. @xref{Function Attributes}.
12347 @strong{Warning:} this calling convention is incompatible with the one
12348 normally used on Unix, so you cannot use it if you need to call
12349 libraries compiled with the Unix compiler.
12351 Also, you must provide function prototypes for all functions that
12352 take variable numbers of arguments (including @code{printf});
12353 otherwise incorrect code will be generated for calls to those
12356 In addition, seriously incorrect code will result if you call a
12357 function with too many arguments. (Normally, extra arguments are
12358 harmlessly ignored.)
12360 @item -mregparm=@var{num}
12362 Control how many registers are used to pass integer arguments. By
12363 default, no registers are used to pass arguments, and at most 3
12364 registers can be used. You can control this behavior for a specific
12365 function by using the function attribute @samp{regparm}.
12366 @xref{Function Attributes}.
12368 @strong{Warning:} if you use this switch, and
12369 @var{num} is nonzero, then you must build all modules with the same
12370 value, including any libraries. This includes the system libraries and
12374 @opindex msseregparm
12375 Use SSE register passing conventions for float and double arguments
12376 and return values. You can control this behavior for a specific
12377 function by using the function attribute @samp{sseregparm}.
12378 @xref{Function Attributes}.
12380 @strong{Warning:} if you use this switch then you must build all
12381 modules with the same value, including any libraries. This includes
12382 the system libraries and startup modules.
12384 @item -mvect8-ret-in-mem
12385 @opindex mvect8-ret-in-mem
12386 Return 8-byte vectors in memory instead of MMX registers. This is the
12387 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12388 Studio compilers until version 12. Later compiler versions (starting
12389 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12390 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12391 you need to remain compatible with existing code produced by those
12392 previous compiler versions or older versions of GCC.
12401 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12402 is specified, the significands of results of floating-point operations are
12403 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12404 significands of results of floating-point operations to 53 bits (double
12405 precision) and @option{-mpc80} rounds the significands of results of
12406 floating-point operations to 64 bits (extended double precision), which is
12407 the default. When this option is used, floating-point operations in higher
12408 precisions are not available to the programmer without setting the FPU
12409 control word explicitly.
12411 Setting the rounding of floating-point operations to less than the default
12412 80 bits can speed some programs by 2% or more. Note that some mathematical
12413 libraries assume that extended precision (80 bit) floating-point operations
12414 are enabled by default; routines in such libraries could suffer significant
12415 loss of accuracy, typically through so-called "catastrophic cancellation",
12416 when this option is used to set the precision to less than extended precision.
12418 @item -mstackrealign
12419 @opindex mstackrealign
12420 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12421 option will generate an alternate prologue and epilogue that realigns the
12422 runtime stack if necessary. This supports mixing legacy codes that keep
12423 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12424 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12425 applicable to individual functions.
12427 @item -mpreferred-stack-boundary=@var{num}
12428 @opindex mpreferred-stack-boundary
12429 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12430 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12431 the default is 4 (16 bytes or 128 bits).
12433 @item -mincoming-stack-boundary=@var{num}
12434 @opindex mincoming-stack-boundary
12435 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12436 boundary. If @option{-mincoming-stack-boundary} is not specified,
12437 the one specified by @option{-mpreferred-stack-boundary} will be used.
12439 On Pentium and PentiumPro, @code{double} and @code{long double} values
12440 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12441 suffer significant run time performance penalties. On Pentium III, the
12442 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12443 properly if it is not 16 byte aligned.
12445 To ensure proper alignment of this values on the stack, the stack boundary
12446 must be as aligned as that required by any value stored on the stack.
12447 Further, every function must be generated such that it keeps the stack
12448 aligned. Thus calling a function compiled with a higher preferred
12449 stack boundary from a function compiled with a lower preferred stack
12450 boundary will most likely misalign the stack. It is recommended that
12451 libraries that use callbacks always use the default setting.
12453 This extra alignment does consume extra stack space, and generally
12454 increases code size. Code that is sensitive to stack space usage, such
12455 as embedded systems and operating system kernels, may want to reduce the
12456 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12483 @itemx -mno-fsgsbase
12513 These switches enable or disable the use of instructions in the MMX,
12514 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12515 F16C, SSE4A, FMA4, XOP, LWP, ABM, BMI, or 3DNow!@: extended instruction sets.
12516 These extensions are also available as built-in functions: see
12517 @ref{X86 Built-in Functions}, for details of the functions enabled and
12518 disabled by these switches.
12520 To have SSE/SSE2 instructions generated automatically from floating-point
12521 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12523 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12524 generates new AVX instructions or AVX equivalence for all SSEx instructions
12527 These options will enable GCC to use these extended instructions in
12528 generated code, even without @option{-mfpmath=sse}. Applications which
12529 perform runtime CPU detection must compile separate files for each
12530 supported architecture, using the appropriate flags. In particular,
12531 the file containing the CPU detection code should be compiled without
12535 @itemx -mno-fused-madd
12536 @opindex mfused-madd
12537 @opindex mno-fused-madd
12538 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12539 instructions. The default is to use these instructions.
12543 This option instructs GCC to emit a @code{cld} instruction in the prologue
12544 of functions that use string instructions. String instructions depend on
12545 the DF flag to select between autoincrement or autodecrement mode. While the
12546 ABI specifies the DF flag to be cleared on function entry, some operating
12547 systems violate this specification by not clearing the DF flag in their
12548 exception dispatchers. The exception handler can be invoked with the DF flag
12549 set which leads to wrong direction mode, when string instructions are used.
12550 This option can be enabled by default on 32-bit x86 targets by configuring
12551 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12552 instructions can be suppressed with the @option{-mno-cld} compiler option
12556 @opindex mvzeroupper
12557 This option instructs GCC to emit a @code{vzeroupper} instruction
12558 before a transfer of control flow out of the function to minimize
12559 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12564 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12565 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12566 data types. This is useful for high resolution counters that could be updated
12567 by multiple processors (or cores). This instruction is generated as part of
12568 atomic built-in functions: see @ref{Atomic Builtins} for details.
12572 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12573 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12574 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12575 SAHF are load and store instructions, respectively, for certain status flags.
12576 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12577 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12581 This option will enable GCC to use movbe instruction to implement
12582 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12586 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12587 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12588 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12592 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12593 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12594 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12595 variants) for single precision floating point arguments. These instructions
12596 are generated only when @option{-funsafe-math-optimizations} is enabled
12597 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12598 Note that while the throughput of the sequence is higher than the throughput
12599 of the non-reciprocal instruction, the precision of the sequence can be
12600 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12602 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12603 already with @option{-ffast-math} (or the above option combination), and
12604 doesn't need @option{-mrecip}.
12606 @item -mveclibabi=@var{type}
12607 @opindex mveclibabi
12608 Specifies the ABI type to use for vectorizing intrinsics using an
12609 external library. Supported types are @code{svml} for the Intel short
12610 vector math library and @code{acml} for the AMD math core library style
12611 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12612 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12613 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12614 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12615 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12616 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12617 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12618 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12619 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12620 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12621 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12622 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12623 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12624 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12625 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12626 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12627 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12628 compatible library will have to be specified at link time.
12630 @item -mabi=@var{name}
12632 Generate code for the specified calling convention. Permissible values
12633 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12634 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12635 ABI when targeting Windows. On all other systems, the default is the
12636 SYSV ABI. You can control this behavior for a specific function by
12637 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12638 @xref{Function Attributes}.
12641 @itemx -mno-push-args
12642 @opindex mpush-args
12643 @opindex mno-push-args
12644 Use PUSH operations to store outgoing parameters. This method is shorter
12645 and usually equally fast as method using SUB/MOV operations and is enabled
12646 by default. In some cases disabling it may improve performance because of
12647 improved scheduling and reduced dependencies.
12649 @item -maccumulate-outgoing-args
12650 @opindex maccumulate-outgoing-args
12651 If enabled, the maximum amount of space required for outgoing arguments will be
12652 computed in the function prologue. This is faster on most modern CPUs
12653 because of reduced dependencies, improved scheduling and reduced stack usage
12654 when preferred stack boundary is not equal to 2. The drawback is a notable
12655 increase in code size. This switch implies @option{-mno-push-args}.
12659 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12660 on thread-safe exception handling must compile and link all code with the
12661 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12662 @option{-D_MT}; when linking, it links in a special thread helper library
12663 @option{-lmingwthrd} which cleans up per thread exception handling data.
12665 @item -mno-align-stringops
12666 @opindex mno-align-stringops
12667 Do not align destination of inlined string operations. This switch reduces
12668 code size and improves performance in case the destination is already aligned,
12669 but GCC doesn't know about it.
12671 @item -minline-all-stringops
12672 @opindex minline-all-stringops
12673 By default GCC inlines string operations only when destination is known to be
12674 aligned at least to 4 byte boundary. This enables more inlining, increase code
12675 size, but may improve performance of code that depends on fast memcpy, strlen
12676 and memset for short lengths.
12678 @item -minline-stringops-dynamically
12679 @opindex minline-stringops-dynamically
12680 For string operation of unknown size, inline runtime checks so for small
12681 blocks inline code is used, while for large blocks library call is used.
12683 @item -mstringop-strategy=@var{alg}
12684 @opindex mstringop-strategy=@var{alg}
12685 Overwrite internal decision heuristic about particular algorithm to inline
12686 string operation with. The allowed values are @code{rep_byte},
12687 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12688 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12689 expanding inline loop, @code{libcall} for always expanding library call.
12691 @item -momit-leaf-frame-pointer
12692 @opindex momit-leaf-frame-pointer
12693 Don't keep the frame pointer in a register for leaf functions. This
12694 avoids the instructions to save, set up and restore frame pointers and
12695 makes an extra register available in leaf functions. The option
12696 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12697 which might make debugging harder.
12699 @item -mtls-direct-seg-refs
12700 @itemx -mno-tls-direct-seg-refs
12701 @opindex mtls-direct-seg-refs
12702 Controls whether TLS variables may be accessed with offsets from the
12703 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12704 or whether the thread base pointer must be added. Whether or not this
12705 is legal depends on the operating system, and whether it maps the
12706 segment to cover the entire TLS area.
12708 For systems that use GNU libc, the default is on.
12711 @itemx -mno-sse2avx
12713 Specify that the assembler should encode SSE instructions with VEX
12714 prefix. The option @option{-mavx} turns this on by default.
12719 If profiling is active @option{-pg} put the profiling
12720 counter call before prologue.
12721 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12722 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12725 @itemx -mno-8bit-idiv
12727 On some processors, like Intel Atom, 8bit unsigned integer divide is
12728 much faster than 32bit/64bit integer divide. This option will generate a
12729 runt-time check. If both dividend and divisor are within range of 0
12730 to 255, 8bit unsigned integer divide will be used instead of
12731 32bit/64bit integer divide.
12735 These @samp{-m} switches are supported in addition to the above
12736 on AMD x86-64 processors in 64-bit environments.
12743 Generate code for a 32-bit or 64-bit environment.
12744 The 32-bit environment sets int, long and pointer to 32 bits and
12745 generates code that runs on any i386 system.
12746 The 64-bit environment sets int to 32 bits and long and pointer
12747 to 64 bits and generates code for AMD's x86-64 architecture. For
12748 darwin only the -m64 option turns off the @option{-fno-pic} and
12749 @option{-mdynamic-no-pic} options.
12751 @item -mno-red-zone
12752 @opindex mno-red-zone
12753 Do not use a so called red zone for x86-64 code. The red zone is mandated
12754 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12755 stack pointer that will not be modified by signal or interrupt handlers
12756 and therefore can be used for temporary data without adjusting the stack
12757 pointer. The flag @option{-mno-red-zone} disables this red zone.
12759 @item -mcmodel=small
12760 @opindex mcmodel=small
12761 Generate code for the small code model: the program and its symbols must
12762 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12763 Programs can be statically or dynamically linked. This is the default
12766 @item -mcmodel=kernel
12767 @opindex mcmodel=kernel
12768 Generate code for the kernel code model. The kernel runs in the
12769 negative 2 GB of the address space.
12770 This model has to be used for Linux kernel code.
12772 @item -mcmodel=medium
12773 @opindex mcmodel=medium
12774 Generate code for the medium model: The program is linked in the lower 2
12775 GB of the address space. Small symbols are also placed there. Symbols
12776 with sizes larger than @option{-mlarge-data-threshold} are put into
12777 large data or bss sections and can be located above 2GB. Programs can
12778 be statically or dynamically linked.
12780 @item -mcmodel=large
12781 @opindex mcmodel=large
12782 Generate code for the large model: This model makes no assumptions
12783 about addresses and sizes of sections.
12786 @node IA-64 Options
12787 @subsection IA-64 Options
12788 @cindex IA-64 Options
12790 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12794 @opindex mbig-endian
12795 Generate code for a big endian target. This is the default for HP-UX@.
12797 @item -mlittle-endian
12798 @opindex mlittle-endian
12799 Generate code for a little endian target. This is the default for AIX5
12805 @opindex mno-gnu-as
12806 Generate (or don't) code for the GNU assembler. This is the default.
12807 @c Also, this is the default if the configure option @option{--with-gnu-as}
12813 @opindex mno-gnu-ld
12814 Generate (or don't) code for the GNU linker. This is the default.
12815 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12820 Generate code that does not use a global pointer register. The result
12821 is not position independent code, and violates the IA-64 ABI@.
12823 @item -mvolatile-asm-stop
12824 @itemx -mno-volatile-asm-stop
12825 @opindex mvolatile-asm-stop
12826 @opindex mno-volatile-asm-stop
12827 Generate (or don't) a stop bit immediately before and after volatile asm
12830 @item -mregister-names
12831 @itemx -mno-register-names
12832 @opindex mregister-names
12833 @opindex mno-register-names
12834 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12835 the stacked registers. This may make assembler output more readable.
12841 Disable (or enable) optimizations that use the small data section. This may
12842 be useful for working around optimizer bugs.
12844 @item -mconstant-gp
12845 @opindex mconstant-gp
12846 Generate code that uses a single constant global pointer value. This is
12847 useful when compiling kernel code.
12851 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12852 This is useful when compiling firmware code.
12854 @item -minline-float-divide-min-latency
12855 @opindex minline-float-divide-min-latency
12856 Generate code for inline divides of floating point values
12857 using the minimum latency algorithm.
12859 @item -minline-float-divide-max-throughput
12860 @opindex minline-float-divide-max-throughput
12861 Generate code for inline divides of floating point values
12862 using the maximum throughput algorithm.
12864 @item -mno-inline-float-divide
12865 @opindex mno-inline-float-divide
12866 Do not generate inline code for divides of floating point values.
12868 @item -minline-int-divide-min-latency
12869 @opindex minline-int-divide-min-latency
12870 Generate code for inline divides of integer values
12871 using the minimum latency algorithm.
12873 @item -minline-int-divide-max-throughput
12874 @opindex minline-int-divide-max-throughput
12875 Generate code for inline divides of integer values
12876 using the maximum throughput algorithm.
12878 @item -mno-inline-int-divide
12879 @opindex mno-inline-int-divide
12880 Do not generate inline code for divides of integer values.
12882 @item -minline-sqrt-min-latency
12883 @opindex minline-sqrt-min-latency
12884 Generate code for inline square roots
12885 using the minimum latency algorithm.
12887 @item -minline-sqrt-max-throughput
12888 @opindex minline-sqrt-max-throughput
12889 Generate code for inline square roots
12890 using the maximum throughput algorithm.
12892 @item -mno-inline-sqrt
12893 @opindex mno-inline-sqrt
12894 Do not generate inline code for sqrt.
12897 @itemx -mno-fused-madd
12898 @opindex mfused-madd
12899 @opindex mno-fused-madd
12900 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12901 instructions. The default is to use these instructions.
12903 @item -mno-dwarf2-asm
12904 @itemx -mdwarf2-asm
12905 @opindex mno-dwarf2-asm
12906 @opindex mdwarf2-asm
12907 Don't (or do) generate assembler code for the DWARF2 line number debugging
12908 info. This may be useful when not using the GNU assembler.
12910 @item -mearly-stop-bits
12911 @itemx -mno-early-stop-bits
12912 @opindex mearly-stop-bits
12913 @opindex mno-early-stop-bits
12914 Allow stop bits to be placed earlier than immediately preceding the
12915 instruction that triggered the stop bit. This can improve instruction
12916 scheduling, but does not always do so.
12918 @item -mfixed-range=@var{register-range}
12919 @opindex mfixed-range
12920 Generate code treating the given register range as fixed registers.
12921 A fixed register is one that the register allocator can not use. This is
12922 useful when compiling kernel code. A register range is specified as
12923 two registers separated by a dash. Multiple register ranges can be
12924 specified separated by a comma.
12926 @item -mtls-size=@var{tls-size}
12928 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12931 @item -mtune=@var{cpu-type}
12933 Tune the instruction scheduling for a particular CPU, Valid values are
12934 itanium, itanium1, merced, itanium2, and mckinley.
12940 Generate code for a 32-bit or 64-bit environment.
12941 The 32-bit environment sets int, long and pointer to 32 bits.
12942 The 64-bit environment sets int to 32 bits and long and pointer
12943 to 64 bits. These are HP-UX specific flags.
12945 @item -mno-sched-br-data-spec
12946 @itemx -msched-br-data-spec
12947 @opindex mno-sched-br-data-spec
12948 @opindex msched-br-data-spec
12949 (Dis/En)able data speculative scheduling before reload.
12950 This will result in generation of the ld.a instructions and
12951 the corresponding check instructions (ld.c / chk.a).
12952 The default is 'disable'.
12954 @item -msched-ar-data-spec
12955 @itemx -mno-sched-ar-data-spec
12956 @opindex msched-ar-data-spec
12957 @opindex mno-sched-ar-data-spec
12958 (En/Dis)able data speculative scheduling after reload.
12959 This will result in generation of the ld.a instructions and
12960 the corresponding check instructions (ld.c / chk.a).
12961 The default is 'enable'.
12963 @item -mno-sched-control-spec
12964 @itemx -msched-control-spec
12965 @opindex mno-sched-control-spec
12966 @opindex msched-control-spec
12967 (Dis/En)able control speculative scheduling. This feature is
12968 available only during region scheduling (i.e.@: before reload).
12969 This will result in generation of the ld.s instructions and
12970 the corresponding check instructions chk.s .
12971 The default is 'disable'.
12973 @item -msched-br-in-data-spec
12974 @itemx -mno-sched-br-in-data-spec
12975 @opindex msched-br-in-data-spec
12976 @opindex mno-sched-br-in-data-spec
12977 (En/Dis)able speculative scheduling of the instructions that
12978 are dependent on the data speculative loads before reload.
12979 This is effective only with @option{-msched-br-data-spec} enabled.
12980 The default is 'enable'.
12982 @item -msched-ar-in-data-spec
12983 @itemx -mno-sched-ar-in-data-spec
12984 @opindex msched-ar-in-data-spec
12985 @opindex mno-sched-ar-in-data-spec
12986 (En/Dis)able speculative scheduling of the instructions that
12987 are dependent on the data speculative loads after reload.
12988 This is effective only with @option{-msched-ar-data-spec} enabled.
12989 The default is 'enable'.
12991 @item -msched-in-control-spec
12992 @itemx -mno-sched-in-control-spec
12993 @opindex msched-in-control-spec
12994 @opindex mno-sched-in-control-spec
12995 (En/Dis)able speculative scheduling of the instructions that
12996 are dependent on the control speculative loads.
12997 This is effective only with @option{-msched-control-spec} enabled.
12998 The default is 'enable'.
13000 @item -mno-sched-prefer-non-data-spec-insns
13001 @itemx -msched-prefer-non-data-spec-insns
13002 @opindex mno-sched-prefer-non-data-spec-insns
13003 @opindex msched-prefer-non-data-spec-insns
13004 If enabled, data speculative instructions will be chosen for schedule
13005 only if there are no other choices at the moment. This will make
13006 the use of the data speculation much more conservative.
13007 The default is 'disable'.
13009 @item -mno-sched-prefer-non-control-spec-insns
13010 @itemx -msched-prefer-non-control-spec-insns
13011 @opindex mno-sched-prefer-non-control-spec-insns
13012 @opindex msched-prefer-non-control-spec-insns
13013 If enabled, control speculative instructions will be chosen for schedule
13014 only if there are no other choices at the moment. This will make
13015 the use of the control speculation much more conservative.
13016 The default is 'disable'.
13018 @item -mno-sched-count-spec-in-critical-path
13019 @itemx -msched-count-spec-in-critical-path
13020 @opindex mno-sched-count-spec-in-critical-path
13021 @opindex msched-count-spec-in-critical-path
13022 If enabled, speculative dependencies will be considered during
13023 computation of the instructions priorities. This will make the use of the
13024 speculation a bit more conservative.
13025 The default is 'disable'.
13027 @item -msched-spec-ldc
13028 @opindex msched-spec-ldc
13029 Use a simple data speculation check. This option is on by default.
13031 @item -msched-control-spec-ldc
13032 @opindex msched-spec-ldc
13033 Use a simple check for control speculation. This option is on by default.
13035 @item -msched-stop-bits-after-every-cycle
13036 @opindex msched-stop-bits-after-every-cycle
13037 Place a stop bit after every cycle when scheduling. This option is on
13040 @item -msched-fp-mem-deps-zero-cost
13041 @opindex msched-fp-mem-deps-zero-cost
13042 Assume that floating-point stores and loads are not likely to cause a conflict
13043 when placed into the same instruction group. This option is disabled by
13046 @item -msel-sched-dont-check-control-spec
13047 @opindex msel-sched-dont-check-control-spec
13048 Generate checks for control speculation in selective scheduling.
13049 This flag is disabled by default.
13051 @item -msched-max-memory-insns=@var{max-insns}
13052 @opindex msched-max-memory-insns
13053 Limit on the number of memory insns per instruction group, giving lower
13054 priority to subsequent memory insns attempting to schedule in the same
13055 instruction group. Frequently useful to prevent cache bank conflicts.
13056 The default value is 1.
13058 @item -msched-max-memory-insns-hard-limit
13059 @opindex msched-max-memory-insns-hard-limit
13060 Disallow more than `msched-max-memory-insns' in instruction group.
13061 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13062 when limit is reached but may still schedule memory operations.
13066 @node IA-64/VMS Options
13067 @subsection IA-64/VMS Options
13069 These @samp{-m} options are defined for the IA-64/VMS implementations:
13072 @item -mvms-return-codes
13073 @opindex mvms-return-codes
13074 Return VMS condition codes from main. The default is to return POSIX
13075 style condition (e.g.@ error) codes.
13077 @item -mdebug-main=@var{prefix}
13078 @opindex mdebug-main=@var{prefix}
13079 Flag the first routine whose name starts with @var{prefix} as the main
13080 routine for the debugger.
13084 Default to 64bit memory allocation routines.
13088 @subsection LM32 Options
13089 @cindex LM32 options
13091 These @option{-m} options are defined for the Lattice Mico32 architecture:
13094 @item -mbarrel-shift-enabled
13095 @opindex mbarrel-shift-enabled
13096 Enable barrel-shift instructions.
13098 @item -mdivide-enabled
13099 @opindex mdivide-enabled
13100 Enable divide and modulus instructions.
13102 @item -mmultiply-enabled
13103 @opindex multiply-enabled
13104 Enable multiply instructions.
13106 @item -msign-extend-enabled
13107 @opindex msign-extend-enabled
13108 Enable sign extend instructions.
13110 @item -muser-enabled
13111 @opindex muser-enabled
13112 Enable user-defined instructions.
13117 @subsection M32C Options
13118 @cindex M32C options
13121 @item -mcpu=@var{name}
13123 Select the CPU for which code is generated. @var{name} may be one of
13124 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13125 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13126 the M32C/80 series.
13130 Specifies that the program will be run on the simulator. This causes
13131 an alternate runtime library to be linked in which supports, for
13132 example, file I/O@. You must not use this option when generating
13133 programs that will run on real hardware; you must provide your own
13134 runtime library for whatever I/O functions are needed.
13136 @item -memregs=@var{number}
13138 Specifies the number of memory-based pseudo-registers GCC will use
13139 during code generation. These pseudo-registers will be used like real
13140 registers, so there is a tradeoff between GCC's ability to fit the
13141 code into available registers, and the performance penalty of using
13142 memory instead of registers. Note that all modules in a program must
13143 be compiled with the same value for this option. Because of that, you
13144 must not use this option with the default runtime libraries gcc
13149 @node M32R/D Options
13150 @subsection M32R/D Options
13151 @cindex M32R/D options
13153 These @option{-m} options are defined for Renesas M32R/D architectures:
13158 Generate code for the M32R/2@.
13162 Generate code for the M32R/X@.
13166 Generate code for the M32R@. This is the default.
13168 @item -mmodel=small
13169 @opindex mmodel=small
13170 Assume all objects live in the lower 16MB of memory (so that their addresses
13171 can be loaded with the @code{ld24} instruction), and assume all subroutines
13172 are reachable with the @code{bl} instruction.
13173 This is the default.
13175 The addressability of a particular object can be set with the
13176 @code{model} attribute.
13178 @item -mmodel=medium
13179 @opindex mmodel=medium
13180 Assume objects may be anywhere in the 32-bit address space (the compiler
13181 will generate @code{seth/add3} instructions to load their addresses), and
13182 assume all subroutines are reachable with the @code{bl} instruction.
13184 @item -mmodel=large
13185 @opindex mmodel=large
13186 Assume objects may be anywhere in the 32-bit address space (the compiler
13187 will generate @code{seth/add3} instructions to load their addresses), and
13188 assume subroutines may not be reachable with the @code{bl} instruction
13189 (the compiler will generate the much slower @code{seth/add3/jl}
13190 instruction sequence).
13193 @opindex msdata=none
13194 Disable use of the small data area. Variables will be put into
13195 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13196 @code{section} attribute has been specified).
13197 This is the default.
13199 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13200 Objects may be explicitly put in the small data area with the
13201 @code{section} attribute using one of these sections.
13203 @item -msdata=sdata
13204 @opindex msdata=sdata
13205 Put small global and static data in the small data area, but do not
13206 generate special code to reference them.
13209 @opindex msdata=use
13210 Put small global and static data in the small data area, and generate
13211 special instructions to reference them.
13215 @cindex smaller data references
13216 Put global and static objects less than or equal to @var{num} bytes
13217 into the small data or bss sections instead of the normal data or bss
13218 sections. The default value of @var{num} is 8.
13219 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13220 for this option to have any effect.
13222 All modules should be compiled with the same @option{-G @var{num}} value.
13223 Compiling with different values of @var{num} may or may not work; if it
13224 doesn't the linker will give an error message---incorrect code will not be
13229 Makes the M32R specific code in the compiler display some statistics
13230 that might help in debugging programs.
13232 @item -malign-loops
13233 @opindex malign-loops
13234 Align all loops to a 32-byte boundary.
13236 @item -mno-align-loops
13237 @opindex mno-align-loops
13238 Do not enforce a 32-byte alignment for loops. This is the default.
13240 @item -missue-rate=@var{number}
13241 @opindex missue-rate=@var{number}
13242 Issue @var{number} instructions per cycle. @var{number} can only be 1
13245 @item -mbranch-cost=@var{number}
13246 @opindex mbranch-cost=@var{number}
13247 @var{number} can only be 1 or 2. If it is 1 then branches will be
13248 preferred over conditional code, if it is 2, then the opposite will
13251 @item -mflush-trap=@var{number}
13252 @opindex mflush-trap=@var{number}
13253 Specifies the trap number to use to flush the cache. The default is
13254 12. Valid numbers are between 0 and 15 inclusive.
13256 @item -mno-flush-trap
13257 @opindex mno-flush-trap
13258 Specifies that the cache cannot be flushed by using a trap.
13260 @item -mflush-func=@var{name}
13261 @opindex mflush-func=@var{name}
13262 Specifies the name of the operating system function to call to flush
13263 the cache. The default is @emph{_flush_cache}, but a function call
13264 will only be used if a trap is not available.
13266 @item -mno-flush-func
13267 @opindex mno-flush-func
13268 Indicates that there is no OS function for flushing the cache.
13272 @node M680x0 Options
13273 @subsection M680x0 Options
13274 @cindex M680x0 options
13276 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13277 The default settings depend on which architecture was selected when
13278 the compiler was configured; the defaults for the most common choices
13282 @item -march=@var{arch}
13284 Generate code for a specific M680x0 or ColdFire instruction set
13285 architecture. Permissible values of @var{arch} for M680x0
13286 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13287 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13288 architectures are selected according to Freescale's ISA classification
13289 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13290 @samp{isab} and @samp{isac}.
13292 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13293 code for a ColdFire target. The @var{arch} in this macro is one of the
13294 @option{-march} arguments given above.
13296 When used together, @option{-march} and @option{-mtune} select code
13297 that runs on a family of similar processors but that is optimized
13298 for a particular microarchitecture.
13300 @item -mcpu=@var{cpu}
13302 Generate code for a specific M680x0 or ColdFire processor.
13303 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13304 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13305 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13306 below, which also classifies the CPUs into families:
13308 @multitable @columnfractions 0.20 0.80
13309 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13310 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13311 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13312 @item @samp{5206e} @tab @samp{5206e}
13313 @item @samp{5208} @tab @samp{5207} @samp{5208}
13314 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13315 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13316 @item @samp{5216} @tab @samp{5214} @samp{5216}
13317 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13318 @item @samp{5225} @tab @samp{5224} @samp{5225}
13319 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13320 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13321 @item @samp{5249} @tab @samp{5249}
13322 @item @samp{5250} @tab @samp{5250}
13323 @item @samp{5271} @tab @samp{5270} @samp{5271}
13324 @item @samp{5272} @tab @samp{5272}
13325 @item @samp{5275} @tab @samp{5274} @samp{5275}
13326 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13327 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13328 @item @samp{5307} @tab @samp{5307}
13329 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13330 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13331 @item @samp{5407} @tab @samp{5407}
13332 @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}
13335 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13336 @var{arch} is compatible with @var{cpu}. Other combinations of
13337 @option{-mcpu} and @option{-march} are rejected.
13339 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13340 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13341 where the value of @var{family} is given by the table above.
13343 @item -mtune=@var{tune}
13345 Tune the code for a particular microarchitecture, within the
13346 constraints set by @option{-march} and @option{-mcpu}.
13347 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13348 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13349 and @samp{cpu32}. The ColdFire microarchitectures
13350 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13352 You can also use @option{-mtune=68020-40} for code that needs
13353 to run relatively well on 68020, 68030 and 68040 targets.
13354 @option{-mtune=68020-60} is similar but includes 68060 targets
13355 as well. These two options select the same tuning decisions as
13356 @option{-m68020-40} and @option{-m68020-60} respectively.
13358 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13359 when tuning for 680x0 architecture @var{arch}. It also defines
13360 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13361 option is used. If gcc is tuning for a range of architectures,
13362 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13363 it defines the macros for every architecture in the range.
13365 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13366 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13367 of the arguments given above.
13373 Generate output for a 68000. This is the default
13374 when the compiler is configured for 68000-based systems.
13375 It is equivalent to @option{-march=68000}.
13377 Use this option for microcontrollers with a 68000 or EC000 core,
13378 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13382 Generate output for a 68010. This is the default
13383 when the compiler is configured for 68010-based systems.
13384 It is equivalent to @option{-march=68010}.
13390 Generate output for a 68020. This is the default
13391 when the compiler is configured for 68020-based systems.
13392 It is equivalent to @option{-march=68020}.
13396 Generate output for a 68030. This is the default when the compiler is
13397 configured for 68030-based systems. It is equivalent to
13398 @option{-march=68030}.
13402 Generate output for a 68040. This is the default when the compiler is
13403 configured for 68040-based systems. It is equivalent to
13404 @option{-march=68040}.
13406 This option inhibits the use of 68881/68882 instructions that have to be
13407 emulated by software on the 68040. Use this option if your 68040 does not
13408 have code to emulate those instructions.
13412 Generate output for a 68060. This is the default when the compiler is
13413 configured for 68060-based systems. It is equivalent to
13414 @option{-march=68060}.
13416 This option inhibits the use of 68020 and 68881/68882 instructions that
13417 have to be emulated by software on the 68060. Use this option if your 68060
13418 does not have code to emulate those instructions.
13422 Generate output for a CPU32. This is the default
13423 when the compiler is configured for CPU32-based systems.
13424 It is equivalent to @option{-march=cpu32}.
13426 Use this option for microcontrollers with a
13427 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13428 68336, 68340, 68341, 68349 and 68360.
13432 Generate output for a 520X ColdFire CPU@. This is the default
13433 when the compiler is configured for 520X-based systems.
13434 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13435 in favor of that option.
13437 Use this option for microcontroller with a 5200 core, including
13438 the MCF5202, MCF5203, MCF5204 and MCF5206.
13442 Generate output for a 5206e ColdFire CPU@. The option is now
13443 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13447 Generate output for a member of the ColdFire 528X family.
13448 The option is now deprecated in favor of the equivalent
13449 @option{-mcpu=528x}.
13453 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13454 in favor of the equivalent @option{-mcpu=5307}.
13458 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13459 in favor of the equivalent @option{-mcpu=5407}.
13463 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13464 This includes use of hardware floating point instructions.
13465 The option is equivalent to @option{-mcpu=547x}, and is now
13466 deprecated in favor of that option.
13470 Generate output for a 68040, without using any of the new instructions.
13471 This results in code which can run relatively efficiently on either a
13472 68020/68881 or a 68030 or a 68040. The generated code does use the
13473 68881 instructions that are emulated on the 68040.
13475 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13479 Generate output for a 68060, without using any of the new instructions.
13480 This results in code which can run relatively efficiently on either a
13481 68020/68881 or a 68030 or a 68040. The generated code does use the
13482 68881 instructions that are emulated on the 68060.
13484 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13488 @opindex mhard-float
13490 Generate floating-point instructions. This is the default for 68020
13491 and above, and for ColdFire devices that have an FPU@. It defines the
13492 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13493 on ColdFire targets.
13496 @opindex msoft-float
13497 Do not generate floating-point instructions; use library calls instead.
13498 This is the default for 68000, 68010, and 68832 targets. It is also
13499 the default for ColdFire devices that have no FPU.
13505 Generate (do not generate) ColdFire hardware divide and remainder
13506 instructions. If @option{-march} is used without @option{-mcpu},
13507 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13508 architectures. Otherwise, the default is taken from the target CPU
13509 (either the default CPU, or the one specified by @option{-mcpu}). For
13510 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13511 @option{-mcpu=5206e}.
13513 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13517 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13518 Additionally, parameters passed on the stack are also aligned to a
13519 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13523 Do not consider type @code{int} to be 16 bits wide. This is the default.
13526 @itemx -mno-bitfield
13527 @opindex mnobitfield
13528 @opindex mno-bitfield
13529 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13530 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13534 Do use the bit-field instructions. The @option{-m68020} option implies
13535 @option{-mbitfield}. This is the default if you use a configuration
13536 designed for a 68020.
13540 Use a different function-calling convention, in which functions
13541 that take a fixed number of arguments return with the @code{rtd}
13542 instruction, which pops their arguments while returning. This
13543 saves one instruction in the caller since there is no need to pop
13544 the arguments there.
13546 This calling convention is incompatible with the one normally
13547 used on Unix, so you cannot use it if you need to call libraries
13548 compiled with the Unix compiler.
13550 Also, you must provide function prototypes for all functions that
13551 take variable numbers of arguments (including @code{printf});
13552 otherwise incorrect code will be generated for calls to those
13555 In addition, seriously incorrect code will result if you call a
13556 function with too many arguments. (Normally, extra arguments are
13557 harmlessly ignored.)
13559 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13560 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13564 Do not use the calling conventions selected by @option{-mrtd}.
13565 This is the default.
13568 @itemx -mno-align-int
13569 @opindex malign-int
13570 @opindex mno-align-int
13571 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13572 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13573 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13574 Aligning variables on 32-bit boundaries produces code that runs somewhat
13575 faster on processors with 32-bit busses at the expense of more memory.
13577 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13578 align structures containing the above types differently than
13579 most published application binary interface specifications for the m68k.
13583 Use the pc-relative addressing mode of the 68000 directly, instead of
13584 using a global offset table. At present, this option implies @option{-fpic},
13585 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13586 not presently supported with @option{-mpcrel}, though this could be supported for
13587 68020 and higher processors.
13589 @item -mno-strict-align
13590 @itemx -mstrict-align
13591 @opindex mno-strict-align
13592 @opindex mstrict-align
13593 Do not (do) assume that unaligned memory references will be handled by
13597 Generate code that allows the data segment to be located in a different
13598 area of memory from the text segment. This allows for execute in place in
13599 an environment without virtual memory management. This option implies
13602 @item -mno-sep-data
13603 Generate code that assumes that the data segment follows the text segment.
13604 This is the default.
13606 @item -mid-shared-library
13607 Generate code that supports shared libraries via the library ID method.
13608 This allows for execute in place and shared libraries in an environment
13609 without virtual memory management. This option implies @option{-fPIC}.
13611 @item -mno-id-shared-library
13612 Generate code that doesn't assume ID based shared libraries are being used.
13613 This is the default.
13615 @item -mshared-library-id=n
13616 Specified the identification number of the ID based shared library being
13617 compiled. Specifying a value of 0 will generate more compact code, specifying
13618 other values will force the allocation of that number to the current
13619 library but is no more space or time efficient than omitting this option.
13625 When generating position-independent code for ColdFire, generate code
13626 that works if the GOT has more than 8192 entries. This code is
13627 larger and slower than code generated without this option. On M680x0
13628 processors, this option is not needed; @option{-fPIC} suffices.
13630 GCC normally uses a single instruction to load values from the GOT@.
13631 While this is relatively efficient, it only works if the GOT
13632 is smaller than about 64k. Anything larger causes the linker
13633 to report an error such as:
13635 @cindex relocation truncated to fit (ColdFire)
13637 relocation truncated to fit: R_68K_GOT16O foobar
13640 If this happens, you should recompile your code with @option{-mxgot}.
13641 It should then work with very large GOTs. However, code generated with
13642 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13643 the value of a global symbol.
13645 Note that some linkers, including newer versions of the GNU linker,
13646 can create multiple GOTs and sort GOT entries. If you have such a linker,
13647 you should only need to use @option{-mxgot} when compiling a single
13648 object file that accesses more than 8192 GOT entries. Very few do.
13650 These options have no effect unless GCC is generating
13651 position-independent code.
13655 @node M68hc1x Options
13656 @subsection M68hc1x Options
13657 @cindex M68hc1x options
13659 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13660 microcontrollers. The default values for these options depends on
13661 which style of microcontroller was selected when the compiler was configured;
13662 the defaults for the most common choices are given below.
13669 Generate output for a 68HC11. This is the default
13670 when the compiler is configured for 68HC11-based systems.
13676 Generate output for a 68HC12. This is the default
13677 when the compiler is configured for 68HC12-based systems.
13683 Generate output for a 68HCS12.
13685 @item -mauto-incdec
13686 @opindex mauto-incdec
13687 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13694 Enable the use of 68HC12 min and max instructions.
13697 @itemx -mno-long-calls
13698 @opindex mlong-calls
13699 @opindex mno-long-calls
13700 Treat all calls as being far away (near). If calls are assumed to be
13701 far away, the compiler will use the @code{call} instruction to
13702 call a function and the @code{rtc} instruction for returning.
13706 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13708 @item -msoft-reg-count=@var{count}
13709 @opindex msoft-reg-count
13710 Specify the number of pseudo-soft registers which are used for the
13711 code generation. The maximum number is 32. Using more pseudo-soft
13712 register may or may not result in better code depending on the program.
13713 The default is 4 for 68HC11 and 2 for 68HC12.
13717 @node MCore Options
13718 @subsection MCore Options
13719 @cindex MCore options
13721 These are the @samp{-m} options defined for the Motorola M*Core
13727 @itemx -mno-hardlit
13729 @opindex mno-hardlit
13730 Inline constants into the code stream if it can be done in two
13731 instructions or less.
13737 Use the divide instruction. (Enabled by default).
13739 @item -mrelax-immediate
13740 @itemx -mno-relax-immediate
13741 @opindex mrelax-immediate
13742 @opindex mno-relax-immediate
13743 Allow arbitrary sized immediates in bit operations.
13745 @item -mwide-bitfields
13746 @itemx -mno-wide-bitfields
13747 @opindex mwide-bitfields
13748 @opindex mno-wide-bitfields
13749 Always treat bit-fields as int-sized.
13751 @item -m4byte-functions
13752 @itemx -mno-4byte-functions
13753 @opindex m4byte-functions
13754 @opindex mno-4byte-functions
13755 Force all functions to be aligned to a four byte boundary.
13757 @item -mcallgraph-data
13758 @itemx -mno-callgraph-data
13759 @opindex mcallgraph-data
13760 @opindex mno-callgraph-data
13761 Emit callgraph information.
13764 @itemx -mno-slow-bytes
13765 @opindex mslow-bytes
13766 @opindex mno-slow-bytes
13767 Prefer word access when reading byte quantities.
13769 @item -mlittle-endian
13770 @itemx -mbig-endian
13771 @opindex mlittle-endian
13772 @opindex mbig-endian
13773 Generate code for a little endian target.
13779 Generate code for the 210 processor.
13783 Assume that run-time support has been provided and so omit the
13784 simulator library (@file{libsim.a)} from the linker command line.
13786 @item -mstack-increment=@var{size}
13787 @opindex mstack-increment
13788 Set the maximum amount for a single stack increment operation. Large
13789 values can increase the speed of programs which contain functions
13790 that need a large amount of stack space, but they can also trigger a
13791 segmentation fault if the stack is extended too much. The default
13797 @subsection MeP Options
13798 @cindex MeP options
13804 Enables the @code{abs} instruction, which is the absolute difference
13805 between two registers.
13809 Enables all the optional instructions - average, multiply, divide, bit
13810 operations, leading zero, absolute difference, min/max, clip, and
13816 Enables the @code{ave} instruction, which computes the average of two
13819 @item -mbased=@var{n}
13821 Variables of size @var{n} bytes or smaller will be placed in the
13822 @code{.based} section by default. Based variables use the @code{$tp}
13823 register as a base register, and there is a 128 byte limit to the
13824 @code{.based} section.
13828 Enables the bit operation instructions - bit test (@code{btstm}), set
13829 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13830 test-and-set (@code{tas}).
13832 @item -mc=@var{name}
13834 Selects which section constant data will be placed in. @var{name} may
13835 be @code{tiny}, @code{near}, or @code{far}.
13839 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13840 useful unless you also provide @code{-mminmax}.
13842 @item -mconfig=@var{name}
13844 Selects one of the build-in core configurations. Each MeP chip has
13845 one or more modules in it; each module has a core CPU and a variety of
13846 coprocessors, optional instructions, and peripherals. The
13847 @code{MeP-Integrator} tool, not part of GCC, provides these
13848 configurations through this option; using this option is the same as
13849 using all the corresponding command line options. The default
13850 configuration is @code{default}.
13854 Enables the coprocessor instructions. By default, this is a 32-bit
13855 coprocessor. Note that the coprocessor is normally enabled via the
13856 @code{-mconfig=} option.
13860 Enables the 32-bit coprocessor's instructions.
13864 Enables the 64-bit coprocessor's instructions.
13868 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13872 Causes constant variables to be placed in the @code{.near} section.
13876 Enables the @code{div} and @code{divu} instructions.
13880 Generate big-endian code.
13884 Generate little-endian code.
13886 @item -mio-volatile
13887 @opindex mio-volatile
13888 Tells the compiler that any variable marked with the @code{io}
13889 attribute is to be considered volatile.
13893 Causes variables to be assigned to the @code{.far} section by default.
13897 Enables the @code{leadz} (leading zero) instruction.
13901 Causes variables to be assigned to the @code{.near} section by default.
13905 Enables the @code{min} and @code{max} instructions.
13909 Enables the multiplication and multiply-accumulate instructions.
13913 Disables all the optional instructions enabled by @code{-mall-opts}.
13917 Enables the @code{repeat} and @code{erepeat} instructions, used for
13918 low-overhead looping.
13922 Causes all variables to default to the @code{.tiny} section. Note
13923 that there is a 65536 byte limit to this section. Accesses to these
13924 variables use the @code{%gp} base register.
13928 Enables the saturation instructions. Note that the compiler does not
13929 currently generate these itself, but this option is included for
13930 compatibility with other tools, like @code{as}.
13934 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13938 Link the simulator runtime libraries.
13942 Link the simulator runtime libraries, excluding built-in support
13943 for reset and exception vectors and tables.
13947 Causes all functions to default to the @code{.far} section. Without
13948 this option, functions default to the @code{.near} section.
13950 @item -mtiny=@var{n}
13952 Variables that are @var{n} bytes or smaller will be allocated to the
13953 @code{.tiny} section. These variables use the @code{$gp} base
13954 register. The default for this option is 4, but note that there's a
13955 65536 byte limit to the @code{.tiny} section.
13959 @node MicroBlaze Options
13960 @subsection MicroBlaze Options
13961 @cindex MicroBlaze Options
13966 @opindex msoft-float
13967 Use software emulation for floating point (default).
13970 @opindex mhard-float
13971 Use hardware floating point instructions.
13975 Do not optimize block moves, use @code{memcpy}.
13977 @item -mno-clearbss
13978 @opindex mno-clearbss
13979 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
13981 @item -mcpu=@var{cpu-type}
13983 Use features of and schedule code for given CPU.
13984 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
13985 where @var{X} is a major version, @var{YY} is the minor version, and
13986 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
13987 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
13989 @item -mxl-soft-mul
13990 @opindex mxl-soft-mul
13991 Use software multiply emulation (default).
13993 @item -mxl-soft-div
13994 @opindex mxl-soft-div
13995 Use software emulation for divides (default).
13997 @item -mxl-barrel-shift
13998 @opindex mxl-barrel-shift
13999 Use the hardware barrel shifter.
14001 @item -mxl-pattern-compare
14002 @opindex mxl-pattern-compare
14003 Use pattern compare instructions.
14005 @item -msmall-divides
14006 @opindex msmall-divides
14007 Use table lookup optimization for small signed integer divisions.
14009 @item -mxl-stack-check
14010 @opindex mxl-stack-check
14011 This option is deprecated. Use -fstack-check instead.
14014 @opindex mxl-gp-opt
14015 Use GP relative sdata/sbss sections.
14017 @item -mxl-multiply-high
14018 @opindex mxl-multiply-high
14019 Use multiply high instructions for high part of 32x32 multiply.
14021 @item -mxl-float-convert
14022 @opindex mxl-float-convert
14023 Use hardware floating point conversion instructions.
14025 @item -mxl-float-sqrt
14026 @opindex mxl-float-sqrt
14027 Use hardware floating point square root instruction.
14029 @item -mxl-mode-@var{app-model}
14030 Select application model @var{app-model}. Valid models are
14033 normal executable (default), uses startup code @file{crt0.o}.
14036 for use with Xilinx Microprocessor Debugger (XMD) based
14037 software intrusive debug agent called xmdstub. This uses startup file
14038 @file{crt1.o} and sets the start address of the program to be 0x800.
14041 for applications that are loaded using a bootloader.
14042 This model uses startup file @file{crt2.o} which does not contain a processor
14043 reset vector handler. This is suitable for transferring control on a
14044 processor reset to the bootloader rather than the application.
14047 for applications that do not require any of the
14048 MicroBlaze vectors. This option may be useful for applications running
14049 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14052 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14053 @option{-mxl-mode-@var{app-model}}.
14058 @subsection MIPS Options
14059 @cindex MIPS options
14065 Generate big-endian code.
14069 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14072 @item -march=@var{arch}
14074 Generate code that will run on @var{arch}, which can be the name of a
14075 generic MIPS ISA, or the name of a particular processor.
14077 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14078 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14079 The processor names are:
14080 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14081 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14082 @samp{5kc}, @samp{5kf},
14084 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14085 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14086 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14087 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14088 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14089 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14093 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14094 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14095 @samp{rm7000}, @samp{rm9000},
14096 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14099 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14100 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14102 The special value @samp{from-abi} selects the
14103 most compatible architecture for the selected ABI (that is,
14104 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14106 Native Linux/GNU toolchains also support the value @samp{native},
14107 which selects the best architecture option for the host processor.
14108 @option{-march=native} has no effect if GCC does not recognize
14111 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14112 (for example, @samp{-march=r2k}). Prefixes are optional, and
14113 @samp{vr} may be written @samp{r}.
14115 Names of the form @samp{@var{n}f2_1} refer to processors with
14116 FPUs clocked at half the rate of the core, names of the form
14117 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14118 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14119 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14120 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14121 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14122 accepted as synonyms for @samp{@var{n}f1_1}.
14124 GCC defines two macros based on the value of this option. The first
14125 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14126 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14127 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14128 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14129 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14131 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14132 above. In other words, it will have the full prefix and will not
14133 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14134 the macro names the resolved architecture (either @samp{"mips1"} or
14135 @samp{"mips3"}). It names the default architecture when no
14136 @option{-march} option is given.
14138 @item -mtune=@var{arch}
14140 Optimize for @var{arch}. Among other things, this option controls
14141 the way instructions are scheduled, and the perceived cost of arithmetic
14142 operations. The list of @var{arch} values is the same as for
14145 When this option is not used, GCC will optimize for the processor
14146 specified by @option{-march}. By using @option{-march} and
14147 @option{-mtune} together, it is possible to generate code that will
14148 run on a family of processors, but optimize the code for one
14149 particular member of that family.
14151 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14152 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14153 @samp{-march} ones described above.
14157 Equivalent to @samp{-march=mips1}.
14161 Equivalent to @samp{-march=mips2}.
14165 Equivalent to @samp{-march=mips3}.
14169 Equivalent to @samp{-march=mips4}.
14173 Equivalent to @samp{-march=mips32}.
14177 Equivalent to @samp{-march=mips32r2}.
14181 Equivalent to @samp{-march=mips64}.
14185 Equivalent to @samp{-march=mips64r2}.
14190 @opindex mno-mips16
14191 Generate (do not generate) MIPS16 code. If GCC is targetting a
14192 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14194 MIPS16 code generation can also be controlled on a per-function basis
14195 by means of @code{mips16} and @code{nomips16} attributes.
14196 @xref{Function Attributes}, for more information.
14198 @item -mflip-mips16
14199 @opindex mflip-mips16
14200 Generate MIPS16 code on alternating functions. This option is provided
14201 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14202 not intended for ordinary use in compiling user code.
14204 @item -minterlink-mips16
14205 @itemx -mno-interlink-mips16
14206 @opindex minterlink-mips16
14207 @opindex mno-interlink-mips16
14208 Require (do not require) that non-MIPS16 code be link-compatible with
14211 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14212 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14213 therefore disables direct jumps unless GCC knows that the target of the
14214 jump is not MIPS16.
14226 Generate code for the given ABI@.
14228 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14229 generates 64-bit code when you select a 64-bit architecture, but you
14230 can use @option{-mgp32} to get 32-bit code instead.
14232 For information about the O64 ABI, see
14233 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14235 GCC supports a variant of the o32 ABI in which floating-point registers
14236 are 64 rather than 32 bits wide. You can select this combination with
14237 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14238 and @samp{mfhc1} instructions and is therefore only supported for
14239 MIPS32R2 processors.
14241 The register assignments for arguments and return values remain the
14242 same, but each scalar value is passed in a single 64-bit register
14243 rather than a pair of 32-bit registers. For example, scalar
14244 floating-point values are returned in @samp{$f0} only, not a
14245 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14246 remains the same, but all 64 bits are saved.
14249 @itemx -mno-abicalls
14251 @opindex mno-abicalls
14252 Generate (do not generate) code that is suitable for SVR4-style
14253 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14258 Generate (do not generate) code that is fully position-independent,
14259 and that can therefore be linked into shared libraries. This option
14260 only affects @option{-mabicalls}.
14262 All @option{-mabicalls} code has traditionally been position-independent,
14263 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14264 as an extension, the GNU toolchain allows executables to use absolute
14265 accesses for locally-binding symbols. It can also use shorter GP
14266 initialization sequences and generate direct calls to locally-defined
14267 functions. This mode is selected by @option{-mno-shared}.
14269 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14270 objects that can only be linked by the GNU linker. However, the option
14271 does not affect the ABI of the final executable; it only affects the ABI
14272 of relocatable objects. Using @option{-mno-shared} will generally make
14273 executables both smaller and quicker.
14275 @option{-mshared} is the default.
14281 Assume (do not assume) that the static and dynamic linkers
14282 support PLTs and copy relocations. This option only affects
14283 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14284 has no effect without @samp{-msym32}.
14286 You can make @option{-mplt} the default by configuring
14287 GCC with @option{--with-mips-plt}. The default is
14288 @option{-mno-plt} otherwise.
14294 Lift (do not lift) the usual restrictions on the size of the global
14297 GCC normally uses a single instruction to load values from the GOT@.
14298 While this is relatively efficient, it will only work if the GOT
14299 is smaller than about 64k. Anything larger will cause the linker
14300 to report an error such as:
14302 @cindex relocation truncated to fit (MIPS)
14304 relocation truncated to fit: R_MIPS_GOT16 foobar
14307 If this happens, you should recompile your code with @option{-mxgot}.
14308 It should then work with very large GOTs, although it will also be
14309 less efficient, since it will take three instructions to fetch the
14310 value of a global symbol.
14312 Note that some linkers can create multiple GOTs. If you have such a
14313 linker, you should only need to use @option{-mxgot} when a single object
14314 file accesses more than 64k's worth of GOT entries. Very few do.
14316 These options have no effect unless GCC is generating position
14321 Assume that general-purpose registers are 32 bits wide.
14325 Assume that general-purpose registers are 64 bits wide.
14329 Assume that floating-point registers are 32 bits wide.
14333 Assume that floating-point registers are 64 bits wide.
14336 @opindex mhard-float
14337 Use floating-point coprocessor instructions.
14340 @opindex msoft-float
14341 Do not use floating-point coprocessor instructions. Implement
14342 floating-point calculations using library calls instead.
14344 @item -msingle-float
14345 @opindex msingle-float
14346 Assume that the floating-point coprocessor only supports single-precision
14349 @item -mdouble-float
14350 @opindex mdouble-float
14351 Assume that the floating-point coprocessor supports double-precision
14352 operations. This is the default.
14358 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14359 implement atomic memory built-in functions. When neither option is
14360 specified, GCC will use the instructions if the target architecture
14363 @option{-mllsc} is useful if the runtime environment can emulate the
14364 instructions and @option{-mno-llsc} can be useful when compiling for
14365 nonstandard ISAs. You can make either option the default by
14366 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14367 respectively. @option{--with-llsc} is the default for some
14368 configurations; see the installation documentation for details.
14374 Use (do not use) revision 1 of the MIPS DSP ASE@.
14375 @xref{MIPS DSP Built-in Functions}. This option defines the
14376 preprocessor macro @samp{__mips_dsp}. It also defines
14377 @samp{__mips_dsp_rev} to 1.
14383 Use (do not use) revision 2 of the MIPS DSP ASE@.
14384 @xref{MIPS DSP Built-in Functions}. This option defines the
14385 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14386 It also defines @samp{__mips_dsp_rev} to 2.
14389 @itemx -mno-smartmips
14390 @opindex msmartmips
14391 @opindex mno-smartmips
14392 Use (do not use) the MIPS SmartMIPS ASE.
14394 @item -mpaired-single
14395 @itemx -mno-paired-single
14396 @opindex mpaired-single
14397 @opindex mno-paired-single
14398 Use (do not use) paired-single floating-point instructions.
14399 @xref{MIPS Paired-Single Support}. This option requires
14400 hardware floating-point support to be enabled.
14406 Use (do not use) MIPS Digital Media Extension instructions.
14407 This option can only be used when generating 64-bit code and requires
14408 hardware floating-point support to be enabled.
14413 @opindex mno-mips3d
14414 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14415 The option @option{-mips3d} implies @option{-mpaired-single}.
14421 Use (do not use) MT Multithreading instructions.
14425 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14426 an explanation of the default and the way that the pointer size is
14431 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14433 The default size of @code{int}s, @code{long}s and pointers depends on
14434 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14435 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14436 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14437 or the same size as integer registers, whichever is smaller.
14443 Assume (do not assume) that all symbols have 32-bit values, regardless
14444 of the selected ABI@. This option is useful in combination with
14445 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14446 to generate shorter and faster references to symbolic addresses.
14450 Put definitions of externally-visible data in a small data section
14451 if that data is no bigger than @var{num} bytes. GCC can then access
14452 the data more efficiently; see @option{-mgpopt} for details.
14454 The default @option{-G} option depends on the configuration.
14456 @item -mlocal-sdata
14457 @itemx -mno-local-sdata
14458 @opindex mlocal-sdata
14459 @opindex mno-local-sdata
14460 Extend (do not extend) the @option{-G} behavior to local data too,
14461 such as to static variables in C@. @option{-mlocal-sdata} is the
14462 default for all configurations.
14464 If the linker complains that an application is using too much small data,
14465 you might want to try rebuilding the less performance-critical parts with
14466 @option{-mno-local-sdata}. You might also want to build large
14467 libraries with @option{-mno-local-sdata}, so that the libraries leave
14468 more room for the main program.
14470 @item -mextern-sdata
14471 @itemx -mno-extern-sdata
14472 @opindex mextern-sdata
14473 @opindex mno-extern-sdata
14474 Assume (do not assume) that externally-defined data will be in
14475 a small data section if that data is within the @option{-G} limit.
14476 @option{-mextern-sdata} is the default for all configurations.
14478 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14479 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14480 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14481 is placed in a small data section. If @var{Var} is defined by another
14482 module, you must either compile that module with a high-enough
14483 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14484 definition. If @var{Var} is common, you must link the application
14485 with a high-enough @option{-G} setting.
14487 The easiest way of satisfying these restrictions is to compile
14488 and link every module with the same @option{-G} option. However,
14489 you may wish to build a library that supports several different
14490 small data limits. You can do this by compiling the library with
14491 the highest supported @option{-G} setting and additionally using
14492 @option{-mno-extern-sdata} to stop the library from making assumptions
14493 about externally-defined data.
14499 Use (do not use) GP-relative accesses for symbols that are known to be
14500 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14501 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14504 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14505 might not hold the value of @code{_gp}. For example, if the code is
14506 part of a library that might be used in a boot monitor, programs that
14507 call boot monitor routines will pass an unknown value in @code{$gp}.
14508 (In such situations, the boot monitor itself would usually be compiled
14509 with @option{-G0}.)
14511 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14512 @option{-mno-extern-sdata}.
14514 @item -membedded-data
14515 @itemx -mno-embedded-data
14516 @opindex membedded-data
14517 @opindex mno-embedded-data
14518 Allocate variables to the read-only data section first if possible, then
14519 next in the small data section if possible, otherwise in data. This gives
14520 slightly slower code than the default, but reduces the amount of RAM required
14521 when executing, and thus may be preferred for some embedded systems.
14523 @item -muninit-const-in-rodata
14524 @itemx -mno-uninit-const-in-rodata
14525 @opindex muninit-const-in-rodata
14526 @opindex mno-uninit-const-in-rodata
14527 Put uninitialized @code{const} variables in the read-only data section.
14528 This option is only meaningful in conjunction with @option{-membedded-data}.
14530 @item -mcode-readable=@var{setting}
14531 @opindex mcode-readable
14532 Specify whether GCC may generate code that reads from executable sections.
14533 There are three possible settings:
14536 @item -mcode-readable=yes
14537 Instructions may freely access executable sections. This is the
14540 @item -mcode-readable=pcrel
14541 MIPS16 PC-relative load instructions can access executable sections,
14542 but other instructions must not do so. This option is useful on 4KSc
14543 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14544 It is also useful on processors that can be configured to have a dual
14545 instruction/data SRAM interface and that, like the M4K, automatically
14546 redirect PC-relative loads to the instruction RAM.
14548 @item -mcode-readable=no
14549 Instructions must not access executable sections. This option can be
14550 useful on targets that are configured to have a dual instruction/data
14551 SRAM interface but that (unlike the M4K) do not automatically redirect
14552 PC-relative loads to the instruction RAM.
14555 @item -msplit-addresses
14556 @itemx -mno-split-addresses
14557 @opindex msplit-addresses
14558 @opindex mno-split-addresses
14559 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14560 relocation operators. This option has been superseded by
14561 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14563 @item -mexplicit-relocs
14564 @itemx -mno-explicit-relocs
14565 @opindex mexplicit-relocs
14566 @opindex mno-explicit-relocs
14567 Use (do not use) assembler relocation operators when dealing with symbolic
14568 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14569 is to use assembler macros instead.
14571 @option{-mexplicit-relocs} is the default if GCC was configured
14572 to use an assembler that supports relocation operators.
14574 @item -mcheck-zero-division
14575 @itemx -mno-check-zero-division
14576 @opindex mcheck-zero-division
14577 @opindex mno-check-zero-division
14578 Trap (do not trap) on integer division by zero.
14580 The default is @option{-mcheck-zero-division}.
14582 @item -mdivide-traps
14583 @itemx -mdivide-breaks
14584 @opindex mdivide-traps
14585 @opindex mdivide-breaks
14586 MIPS systems check for division by zero by generating either a
14587 conditional trap or a break instruction. Using traps results in
14588 smaller code, but is only supported on MIPS II and later. Also, some
14589 versions of the Linux kernel have a bug that prevents trap from
14590 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14591 allow conditional traps on architectures that support them and
14592 @option{-mdivide-breaks} to force the use of breaks.
14594 The default is usually @option{-mdivide-traps}, but this can be
14595 overridden at configure time using @option{--with-divide=breaks}.
14596 Divide-by-zero checks can be completely disabled using
14597 @option{-mno-check-zero-division}.
14602 @opindex mno-memcpy
14603 Force (do not force) the use of @code{memcpy()} for non-trivial block
14604 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14605 most constant-sized copies.
14608 @itemx -mno-long-calls
14609 @opindex mlong-calls
14610 @opindex mno-long-calls
14611 Disable (do not disable) use of the @code{jal} instruction. Calling
14612 functions using @code{jal} is more efficient but requires the caller
14613 and callee to be in the same 256 megabyte segment.
14615 This option has no effect on abicalls code. The default is
14616 @option{-mno-long-calls}.
14622 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14623 instructions, as provided by the R4650 ISA@.
14626 @itemx -mno-fused-madd
14627 @opindex mfused-madd
14628 @opindex mno-fused-madd
14629 Enable (disable) use of the floating point multiply-accumulate
14630 instructions, when they are available. The default is
14631 @option{-mfused-madd}.
14633 When multiply-accumulate instructions are used, the intermediate
14634 product is calculated to infinite precision and is not subject to
14635 the FCSR Flush to Zero bit. This may be undesirable in some
14640 Tell the MIPS assembler to not run its preprocessor over user
14641 assembler files (with a @samp{.s} suffix) when assembling them.
14644 @itemx -mno-fix-r4000
14645 @opindex mfix-r4000
14646 @opindex mno-fix-r4000
14647 Work around certain R4000 CPU errata:
14650 A double-word or a variable shift may give an incorrect result if executed
14651 immediately after starting an integer division.
14653 A double-word or a variable shift may give an incorrect result if executed
14654 while an integer multiplication is in progress.
14656 An integer division may give an incorrect result if started in a delay slot
14657 of a taken branch or a jump.
14661 @itemx -mno-fix-r4400
14662 @opindex mfix-r4400
14663 @opindex mno-fix-r4400
14664 Work around certain R4400 CPU errata:
14667 A double-word or a variable shift may give an incorrect result if executed
14668 immediately after starting an integer division.
14672 @itemx -mno-fix-r10000
14673 @opindex mfix-r10000
14674 @opindex mno-fix-r10000
14675 Work around certain R10000 errata:
14678 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14679 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14682 This option can only be used if the target architecture supports
14683 branch-likely instructions. @option{-mfix-r10000} is the default when
14684 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14688 @itemx -mno-fix-vr4120
14689 @opindex mfix-vr4120
14690 Work around certain VR4120 errata:
14693 @code{dmultu} does not always produce the correct result.
14695 @code{div} and @code{ddiv} do not always produce the correct result if one
14696 of the operands is negative.
14698 The workarounds for the division errata rely on special functions in
14699 @file{libgcc.a}. At present, these functions are only provided by
14700 the @code{mips64vr*-elf} configurations.
14702 Other VR4120 errata require a nop to be inserted between certain pairs of
14703 instructions. These errata are handled by the assembler, not by GCC itself.
14706 @opindex mfix-vr4130
14707 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14708 workarounds are implemented by the assembler rather than by GCC,
14709 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14710 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14711 instructions are available instead.
14714 @itemx -mno-fix-sb1
14716 Work around certain SB-1 CPU core errata.
14717 (This flag currently works around the SB-1 revision 2
14718 ``F1'' and ``F2'' floating point errata.)
14720 @item -mr10k-cache-barrier=@var{setting}
14721 @opindex mr10k-cache-barrier
14722 Specify whether GCC should insert cache barriers to avoid the
14723 side-effects of speculation on R10K processors.
14725 In common with many processors, the R10K tries to predict the outcome
14726 of a conditional branch and speculatively executes instructions from
14727 the ``taken'' branch. It later aborts these instructions if the
14728 predicted outcome was wrong. However, on the R10K, even aborted
14729 instructions can have side effects.
14731 This problem only affects kernel stores and, depending on the system,
14732 kernel loads. As an example, a speculatively-executed store may load
14733 the target memory into cache and mark the cache line as dirty, even if
14734 the store itself is later aborted. If a DMA operation writes to the
14735 same area of memory before the ``dirty'' line is flushed, the cached
14736 data will overwrite the DMA-ed data. See the R10K processor manual
14737 for a full description, including other potential problems.
14739 One workaround is to insert cache barrier instructions before every memory
14740 access that might be speculatively executed and that might have side
14741 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14742 controls GCC's implementation of this workaround. It assumes that
14743 aborted accesses to any byte in the following regions will not have
14748 the memory occupied by the current function's stack frame;
14751 the memory occupied by an incoming stack argument;
14754 the memory occupied by an object with a link-time-constant address.
14757 It is the kernel's responsibility to ensure that speculative
14758 accesses to these regions are indeed safe.
14760 If the input program contains a function declaration such as:
14766 then the implementation of @code{foo} must allow @code{j foo} and
14767 @code{jal foo} to be executed speculatively. GCC honors this
14768 restriction for functions it compiles itself. It expects non-GCC
14769 functions (such as hand-written assembly code) to do the same.
14771 The option has three forms:
14774 @item -mr10k-cache-barrier=load-store
14775 Insert a cache barrier before a load or store that might be
14776 speculatively executed and that might have side effects even
14779 @item -mr10k-cache-barrier=store
14780 Insert a cache barrier before a store that might be speculatively
14781 executed and that might have side effects even if aborted.
14783 @item -mr10k-cache-barrier=none
14784 Disable the insertion of cache barriers. This is the default setting.
14787 @item -mflush-func=@var{func}
14788 @itemx -mno-flush-func
14789 @opindex mflush-func
14790 Specifies the function to call to flush the I and D caches, or to not
14791 call any such function. If called, the function must take the same
14792 arguments as the common @code{_flush_func()}, that is, the address of the
14793 memory range for which the cache is being flushed, the size of the
14794 memory range, and the number 3 (to flush both caches). The default
14795 depends on the target GCC was configured for, but commonly is either
14796 @samp{_flush_func} or @samp{__cpu_flush}.
14798 @item mbranch-cost=@var{num}
14799 @opindex mbranch-cost
14800 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14801 This cost is only a heuristic and is not guaranteed to produce
14802 consistent results across releases. A zero cost redundantly selects
14803 the default, which is based on the @option{-mtune} setting.
14805 @item -mbranch-likely
14806 @itemx -mno-branch-likely
14807 @opindex mbranch-likely
14808 @opindex mno-branch-likely
14809 Enable or disable use of Branch Likely instructions, regardless of the
14810 default for the selected architecture. By default, Branch Likely
14811 instructions may be generated if they are supported by the selected
14812 architecture. An exception is for the MIPS32 and MIPS64 architectures
14813 and processors which implement those architectures; for those, Branch
14814 Likely instructions will not be generated by default because the MIPS32
14815 and MIPS64 architectures specifically deprecate their use.
14817 @item -mfp-exceptions
14818 @itemx -mno-fp-exceptions
14819 @opindex mfp-exceptions
14820 Specifies whether FP exceptions are enabled. This affects how we schedule
14821 FP instructions for some processors. The default is that FP exceptions are
14824 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14825 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14828 @item -mvr4130-align
14829 @itemx -mno-vr4130-align
14830 @opindex mvr4130-align
14831 The VR4130 pipeline is two-way superscalar, but can only issue two
14832 instructions together if the first one is 8-byte aligned. When this
14833 option is enabled, GCC will align pairs of instructions that it
14834 thinks should execute in parallel.
14836 This option only has an effect when optimizing for the VR4130.
14837 It normally makes code faster, but at the expense of making it bigger.
14838 It is enabled by default at optimization level @option{-O3}.
14843 Enable (disable) generation of @code{synci} instructions on
14844 architectures that support it. The @code{synci} instructions (if
14845 enabled) will be generated when @code{__builtin___clear_cache()} is
14848 This option defaults to @code{-mno-synci}, but the default can be
14849 overridden by configuring with @code{--with-synci}.
14851 When compiling code for single processor systems, it is generally safe
14852 to use @code{synci}. However, on many multi-core (SMP) systems, it
14853 will not invalidate the instruction caches on all cores and may lead
14854 to undefined behavior.
14856 @item -mrelax-pic-calls
14857 @itemx -mno-relax-pic-calls
14858 @opindex mrelax-pic-calls
14859 Try to turn PIC calls that are normally dispatched via register
14860 @code{$25} into direct calls. This is only possible if the linker can
14861 resolve the destination at link-time and if the destination is within
14862 range for a direct call.
14864 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14865 an assembler and a linker that supports the @code{.reloc} assembly
14866 directive and @code{-mexplicit-relocs} is in effect. With
14867 @code{-mno-explicit-relocs}, this optimization can be performed by the
14868 assembler and the linker alone without help from the compiler.
14870 @item -mmcount-ra-address
14871 @itemx -mno-mcount-ra-address
14872 @opindex mmcount-ra-address
14873 @opindex mno-mcount-ra-address
14874 Emit (do not emit) code that allows @code{_mcount} to modify the
14875 calling function's return address. When enabled, this option extends
14876 the usual @code{_mcount} interface with a new @var{ra-address}
14877 parameter, which has type @code{intptr_t *} and is passed in register
14878 @code{$12}. @code{_mcount} can then modify the return address by
14879 doing both of the following:
14882 Returning the new address in register @code{$31}.
14884 Storing the new address in @code{*@var{ra-address}},
14885 if @var{ra-address} is nonnull.
14888 The default is @option{-mno-mcount-ra-address}.
14893 @subsection MMIX Options
14894 @cindex MMIX Options
14896 These options are defined for the MMIX:
14900 @itemx -mno-libfuncs
14902 @opindex mno-libfuncs
14903 Specify that intrinsic library functions are being compiled, passing all
14904 values in registers, no matter the size.
14907 @itemx -mno-epsilon
14909 @opindex mno-epsilon
14910 Generate floating-point comparison instructions that compare with respect
14911 to the @code{rE} epsilon register.
14913 @item -mabi=mmixware
14915 @opindex mabi=mmixware
14917 Generate code that passes function parameters and return values that (in
14918 the called function) are seen as registers @code{$0} and up, as opposed to
14919 the GNU ABI which uses global registers @code{$231} and up.
14921 @item -mzero-extend
14922 @itemx -mno-zero-extend
14923 @opindex mzero-extend
14924 @opindex mno-zero-extend
14925 When reading data from memory in sizes shorter than 64 bits, use (do not
14926 use) zero-extending load instructions by default, rather than
14927 sign-extending ones.
14930 @itemx -mno-knuthdiv
14932 @opindex mno-knuthdiv
14933 Make the result of a division yielding a remainder have the same sign as
14934 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14935 remainder follows the sign of the dividend. Both methods are
14936 arithmetically valid, the latter being almost exclusively used.
14938 @item -mtoplevel-symbols
14939 @itemx -mno-toplevel-symbols
14940 @opindex mtoplevel-symbols
14941 @opindex mno-toplevel-symbols
14942 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14943 code can be used with the @code{PREFIX} assembly directive.
14947 Generate an executable in the ELF format, rather than the default
14948 @samp{mmo} format used by the @command{mmix} simulator.
14950 @item -mbranch-predict
14951 @itemx -mno-branch-predict
14952 @opindex mbranch-predict
14953 @opindex mno-branch-predict
14954 Use (do not use) the probable-branch instructions, when static branch
14955 prediction indicates a probable branch.
14957 @item -mbase-addresses
14958 @itemx -mno-base-addresses
14959 @opindex mbase-addresses
14960 @opindex mno-base-addresses
14961 Generate (do not generate) code that uses @emph{base addresses}. Using a
14962 base address automatically generates a request (handled by the assembler
14963 and the linker) for a constant to be set up in a global register. The
14964 register is used for one or more base address requests within the range 0
14965 to 255 from the value held in the register. The generally leads to short
14966 and fast code, but the number of different data items that can be
14967 addressed is limited. This means that a program that uses lots of static
14968 data may require @option{-mno-base-addresses}.
14970 @item -msingle-exit
14971 @itemx -mno-single-exit
14972 @opindex msingle-exit
14973 @opindex mno-single-exit
14974 Force (do not force) generated code to have a single exit point in each
14978 @node MN10300 Options
14979 @subsection MN10300 Options
14980 @cindex MN10300 options
14982 These @option{-m} options are defined for Matsushita MN10300 architectures:
14987 Generate code to avoid bugs in the multiply instructions for the MN10300
14988 processors. This is the default.
14990 @item -mno-mult-bug
14991 @opindex mno-mult-bug
14992 Do not generate code to avoid bugs in the multiply instructions for the
14993 MN10300 processors.
14997 Generate code which uses features specific to the AM33 processor.
15001 Do not generate code which uses features specific to the AM33 processor. This
15006 Generate code which uses features specific to the AM33/2.0 processor.
15010 Generate code which uses features specific to the AM34 processor.
15012 @item -mtune=@var{cpu-type}
15014 Use the timing characteristics of the indicated CPU type when
15015 scheduling instructions. This does not change the targeted processor
15016 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15017 @samp{am33-2} or @samp{am34}.
15019 @item -mreturn-pointer-on-d0
15020 @opindex mreturn-pointer-on-d0
15021 When generating a function which returns a pointer, return the pointer
15022 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15023 only in a0, and attempts to call such functions without a prototype
15024 would result in errors. Note that this option is on by default; use
15025 @option{-mno-return-pointer-on-d0} to disable it.
15029 Do not link in the C run-time initialization object file.
15033 Indicate to the linker that it should perform a relaxation optimization pass
15034 to shorten branches, calls and absolute memory addresses. This option only
15035 has an effect when used on the command line for the final link step.
15037 This option makes symbolic debugging impossible.
15041 Allow the compiler to generate @emph{Long Instruction Word}
15042 instructions if the target is the @samp{AM33} or later. This is the
15043 default. This option defines the preprocessor macro @samp{__LIW__}.
15047 Do not allow the compiler to generate @emph{Long Instruction Word}
15048 instructions. This option defines the preprocessor macro
15053 @node PDP-11 Options
15054 @subsection PDP-11 Options
15055 @cindex PDP-11 Options
15057 These options are defined for the PDP-11:
15062 Use hardware FPP floating point. This is the default. (FIS floating
15063 point on the PDP-11/40 is not supported.)
15066 @opindex msoft-float
15067 Do not use hardware floating point.
15071 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15075 Return floating-point results in memory. This is the default.
15079 Generate code for a PDP-11/40.
15083 Generate code for a PDP-11/45. This is the default.
15087 Generate code for a PDP-11/10.
15089 @item -mbcopy-builtin
15090 @opindex mbcopy-builtin
15091 Use inline @code{movmemhi} patterns for copying memory. This is the
15096 Do not use inline @code{movmemhi} patterns for copying memory.
15102 Use 16-bit @code{int}. This is the default.
15108 Use 32-bit @code{int}.
15111 @itemx -mno-float32
15113 @opindex mno-float32
15114 Use 64-bit @code{float}. This is the default.
15117 @itemx -mno-float64
15119 @opindex mno-float64
15120 Use 32-bit @code{float}.
15124 Use @code{abshi2} pattern. This is the default.
15128 Do not use @code{abshi2} pattern.
15130 @item -mbranch-expensive
15131 @opindex mbranch-expensive
15132 Pretend that branches are expensive. This is for experimenting with
15133 code generation only.
15135 @item -mbranch-cheap
15136 @opindex mbranch-cheap
15137 Do not pretend that branches are expensive. This is the default.
15141 Use Unix assembler syntax. This is the default when configured for
15142 @samp{pdp11-*-bsd}.
15146 Use DEC assembler syntax. This is the default when configured for any
15147 PDP-11 target other than @samp{pdp11-*-bsd}.
15150 @node picoChip Options
15151 @subsection picoChip Options
15152 @cindex picoChip options
15154 These @samp{-m} options are defined for picoChip implementations:
15158 @item -mae=@var{ae_type}
15160 Set the instruction set, register set, and instruction scheduling
15161 parameters for array element type @var{ae_type}. Supported values
15162 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15164 @option{-mae=ANY} selects a completely generic AE type. Code
15165 generated with this option will run on any of the other AE types. The
15166 code will not be as efficient as it would be if compiled for a specific
15167 AE type, and some types of operation (e.g., multiplication) will not
15168 work properly on all types of AE.
15170 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15171 for compiled code, and is the default.
15173 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15174 option may suffer from poor performance of byte (char) manipulation,
15175 since the DSP AE does not provide hardware support for byte load/stores.
15177 @item -msymbol-as-address
15178 Enable the compiler to directly use a symbol name as an address in a
15179 load/store instruction, without first loading it into a
15180 register. Typically, the use of this option will generate larger
15181 programs, which run faster than when the option isn't used. However, the
15182 results vary from program to program, so it is left as a user option,
15183 rather than being permanently enabled.
15185 @item -mno-inefficient-warnings
15186 Disables warnings about the generation of inefficient code. These
15187 warnings can be generated, for example, when compiling code which
15188 performs byte-level memory operations on the MAC AE type. The MAC AE has
15189 no hardware support for byte-level memory operations, so all byte
15190 load/stores must be synthesized from word load/store operations. This is
15191 inefficient and a warning will be generated indicating to the programmer
15192 that they should rewrite the code to avoid byte operations, or to target
15193 an AE type which has the necessary hardware support. This option enables
15194 the warning to be turned off.
15198 @node PowerPC Options
15199 @subsection PowerPC Options
15200 @cindex PowerPC options
15202 These are listed under @xref{RS/6000 and PowerPC Options}.
15204 @node RS/6000 and PowerPC Options
15205 @subsection IBM RS/6000 and PowerPC Options
15206 @cindex RS/6000 and PowerPC Options
15207 @cindex IBM RS/6000 and PowerPC Options
15209 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15216 @itemx -mno-powerpc
15217 @itemx -mpowerpc-gpopt
15218 @itemx -mno-powerpc-gpopt
15219 @itemx -mpowerpc-gfxopt
15220 @itemx -mno-powerpc-gfxopt
15223 @itemx -mno-powerpc64
15227 @itemx -mno-popcntb
15229 @itemx -mno-popcntd
15238 @itemx -mno-hard-dfp
15242 @opindex mno-power2
15244 @opindex mno-powerpc
15245 @opindex mpowerpc-gpopt
15246 @opindex mno-powerpc-gpopt
15247 @opindex mpowerpc-gfxopt
15248 @opindex mno-powerpc-gfxopt
15249 @opindex mpowerpc64
15250 @opindex mno-powerpc64
15254 @opindex mno-popcntb
15256 @opindex mno-popcntd
15262 @opindex mno-mfpgpr
15264 @opindex mno-hard-dfp
15265 GCC supports two related instruction set architectures for the
15266 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15267 instructions supported by the @samp{rios} chip set used in the original
15268 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15269 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15270 the IBM 4xx, 6xx, and follow-on microprocessors.
15272 Neither architecture is a subset of the other. However there is a
15273 large common subset of instructions supported by both. An MQ
15274 register is included in processors supporting the POWER architecture.
15276 You use these options to specify which instructions are available on the
15277 processor you are using. The default value of these options is
15278 determined when configuring GCC@. Specifying the
15279 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15280 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15281 rather than the options listed above.
15283 The @option{-mpower} option allows GCC to generate instructions that
15284 are found only in the POWER architecture and to use the MQ register.
15285 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15286 to generate instructions that are present in the POWER2 architecture but
15287 not the original POWER architecture.
15289 The @option{-mpowerpc} option allows GCC to generate instructions that
15290 are found only in the 32-bit subset of the PowerPC architecture.
15291 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15292 GCC to use the optional PowerPC architecture instructions in the
15293 General Purpose group, including floating-point square root. Specifying
15294 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15295 use the optional PowerPC architecture instructions in the Graphics
15296 group, including floating-point select.
15298 The @option{-mmfcrf} option allows GCC to generate the move from
15299 condition register field instruction implemented on the POWER4
15300 processor and other processors that support the PowerPC V2.01
15302 The @option{-mpopcntb} option allows GCC to generate the popcount and
15303 double precision FP reciprocal estimate instruction implemented on the
15304 POWER5 processor and other processors that support the PowerPC V2.02
15306 The @option{-mpopcntd} option allows GCC to generate the popcount
15307 instruction implemented on the POWER7 processor and other processors
15308 that support the PowerPC V2.06 architecture.
15309 The @option{-mfprnd} option allows GCC to generate the FP round to
15310 integer instructions implemented on the POWER5+ processor and other
15311 processors that support the PowerPC V2.03 architecture.
15312 The @option{-mcmpb} option allows GCC to generate the compare bytes
15313 instruction implemented on the POWER6 processor and other processors
15314 that support the PowerPC V2.05 architecture.
15315 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15316 general purpose register instructions implemented on the POWER6X
15317 processor and other processors that support the extended PowerPC V2.05
15319 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15320 point instructions implemented on some POWER processors.
15322 The @option{-mpowerpc64} option allows GCC to generate the additional
15323 64-bit instructions that are found in the full PowerPC64 architecture
15324 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15325 @option{-mno-powerpc64}.
15327 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15328 will use only the instructions in the common subset of both
15329 architectures plus some special AIX common-mode calls, and will not use
15330 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15331 permits GCC to use any instruction from either architecture and to
15332 allow use of the MQ register; specify this for the Motorola MPC601.
15334 @item -mnew-mnemonics
15335 @itemx -mold-mnemonics
15336 @opindex mnew-mnemonics
15337 @opindex mold-mnemonics
15338 Select which mnemonics to use in the generated assembler code. With
15339 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15340 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15341 assembler mnemonics defined for the POWER architecture. Instructions
15342 defined in only one architecture have only one mnemonic; GCC uses that
15343 mnemonic irrespective of which of these options is specified.
15345 GCC defaults to the mnemonics appropriate for the architecture in
15346 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15347 value of these option. Unless you are building a cross-compiler, you
15348 should normally not specify either @option{-mnew-mnemonics} or
15349 @option{-mold-mnemonics}, but should instead accept the default.
15351 @item -mcpu=@var{cpu_type}
15353 Set architecture type, register usage, choice of mnemonics, and
15354 instruction scheduling parameters for machine type @var{cpu_type}.
15355 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15356 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15357 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15358 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15359 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15360 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15361 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15362 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15363 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15364 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15365 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15367 @option{-mcpu=common} selects a completely generic processor. Code
15368 generated under this option will run on any POWER or PowerPC processor.
15369 GCC will use only the instructions in the common subset of both
15370 architectures, and will not use the MQ register. GCC assumes a generic
15371 processor model for scheduling purposes.
15373 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15374 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15375 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15376 types, with an appropriate, generic processor model assumed for
15377 scheduling purposes.
15379 The other options specify a specific processor. Code generated under
15380 those options will run best on that processor, and may not run at all on
15383 The @option{-mcpu} options automatically enable or disable the
15386 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15387 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15388 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15389 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15391 The particular options set for any particular CPU will vary between
15392 compiler versions, depending on what setting seems to produce optimal
15393 code for that CPU; it doesn't necessarily reflect the actual hardware's
15394 capabilities. If you wish to set an individual option to a particular
15395 value, you may specify it after the @option{-mcpu} option, like
15396 @samp{-mcpu=970 -mno-altivec}.
15398 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15399 not enabled or disabled by the @option{-mcpu} option at present because
15400 AIX does not have full support for these options. You may still
15401 enable or disable them individually if you're sure it'll work in your
15404 @item -mtune=@var{cpu_type}
15406 Set the instruction scheduling parameters for machine type
15407 @var{cpu_type}, but do not set the architecture type, register usage, or
15408 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15409 values for @var{cpu_type} are used for @option{-mtune} as for
15410 @option{-mcpu}. If both are specified, the code generated will use the
15411 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15412 scheduling parameters set by @option{-mtune}.
15414 @item -mcmodel=small
15415 @opindex mcmodel=small
15416 Generate PowerPC64 code for the small model: The TOC is limited to
15419 @item -mcmodel=medium
15420 @opindex mcmodel=medium
15421 Generate PowerPC64 code for the medium model: The TOC and other static
15422 data may be up to a total of 4G in size.
15424 @item -mcmodel=large
15425 @opindex mcmodel=large
15426 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15427 in size. Other data and code is only limited by the 64-bit address
15431 @itemx -mno-altivec
15433 @opindex mno-altivec
15434 Generate code that uses (does not use) AltiVec instructions, and also
15435 enable the use of built-in functions that allow more direct access to
15436 the AltiVec instruction set. You may also need to set
15437 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15443 @opindex mno-vrsave
15444 Generate VRSAVE instructions when generating AltiVec code.
15446 @item -mgen-cell-microcode
15447 @opindex mgen-cell-microcode
15448 Generate Cell microcode instructions
15450 @item -mwarn-cell-microcode
15451 @opindex mwarn-cell-microcode
15452 Warning when a Cell microcode instruction is going to emitted. An example
15453 of a Cell microcode instruction is a variable shift.
15456 @opindex msecure-plt
15457 Generate code that allows ld and ld.so to build executables and shared
15458 libraries with non-exec .plt and .got sections. This is a PowerPC
15459 32-bit SYSV ABI option.
15463 Generate code that uses a BSS .plt section that ld.so fills in, and
15464 requires .plt and .got sections that are both writable and executable.
15465 This is a PowerPC 32-bit SYSV ABI option.
15471 This switch enables or disables the generation of ISEL instructions.
15473 @item -misel=@var{yes/no}
15474 This switch has been deprecated. Use @option{-misel} and
15475 @option{-mno-isel} instead.
15481 This switch enables or disables the generation of SPE simd
15487 @opindex mno-paired
15488 This switch enables or disables the generation of PAIRED simd
15491 @item -mspe=@var{yes/no}
15492 This option has been deprecated. Use @option{-mspe} and
15493 @option{-mno-spe} instead.
15499 Generate code that uses (does not use) vector/scalar (VSX)
15500 instructions, and also enable the use of built-in functions that allow
15501 more direct access to the VSX instruction set.
15503 @item -mfloat-gprs=@var{yes/single/double/no}
15504 @itemx -mfloat-gprs
15505 @opindex mfloat-gprs
15506 This switch enables or disables the generation of floating point
15507 operations on the general purpose registers for architectures that
15510 The argument @var{yes} or @var{single} enables the use of
15511 single-precision floating point operations.
15513 The argument @var{double} enables the use of single and
15514 double-precision floating point operations.
15516 The argument @var{no} disables floating point operations on the
15517 general purpose registers.
15519 This option is currently only available on the MPC854x.
15525 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15526 targets (including GNU/Linux). The 32-bit environment sets int, long
15527 and pointer to 32 bits and generates code that runs on any PowerPC
15528 variant. The 64-bit environment sets int to 32 bits and long and
15529 pointer to 64 bits, and generates code for PowerPC64, as for
15530 @option{-mpowerpc64}.
15533 @itemx -mno-fp-in-toc
15534 @itemx -mno-sum-in-toc
15535 @itemx -mminimal-toc
15537 @opindex mno-fp-in-toc
15538 @opindex mno-sum-in-toc
15539 @opindex mminimal-toc
15540 Modify generation of the TOC (Table Of Contents), which is created for
15541 every executable file. The @option{-mfull-toc} option is selected by
15542 default. In that case, GCC will allocate at least one TOC entry for
15543 each unique non-automatic variable reference in your program. GCC
15544 will also place floating-point constants in the TOC@. However, only
15545 16,384 entries are available in the TOC@.
15547 If you receive a linker error message that saying you have overflowed
15548 the available TOC space, you can reduce the amount of TOC space used
15549 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15550 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15551 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15552 generate code to calculate the sum of an address and a constant at
15553 run-time instead of putting that sum into the TOC@. You may specify one
15554 or both of these options. Each causes GCC to produce very slightly
15555 slower and larger code at the expense of conserving TOC space.
15557 If you still run out of space in the TOC even when you specify both of
15558 these options, specify @option{-mminimal-toc} instead. This option causes
15559 GCC to make only one TOC entry for every file. When you specify this
15560 option, GCC will produce code that is slower and larger but which
15561 uses extremely little TOC space. You may wish to use this option
15562 only on files that contain less frequently executed code.
15568 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15569 @code{long} type, and the infrastructure needed to support them.
15570 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15571 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15572 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15575 @itemx -mno-xl-compat
15576 @opindex mxl-compat
15577 @opindex mno-xl-compat
15578 Produce code that conforms more closely to IBM XL compiler semantics
15579 when using AIX-compatible ABI@. Pass floating-point arguments to
15580 prototyped functions beyond the register save area (RSA) on the stack
15581 in addition to argument FPRs. Do not assume that most significant
15582 double in 128-bit long double value is properly rounded when comparing
15583 values and converting to double. Use XL symbol names for long double
15586 The AIX calling convention was extended but not initially documented to
15587 handle an obscure K&R C case of calling a function that takes the
15588 address of its arguments with fewer arguments than declared. IBM XL
15589 compilers access floating point arguments which do not fit in the
15590 RSA from the stack when a subroutine is compiled without
15591 optimization. Because always storing floating-point arguments on the
15592 stack is inefficient and rarely needed, this option is not enabled by
15593 default and only is necessary when calling subroutines compiled by IBM
15594 XL compilers without optimization.
15598 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15599 application written to use message passing with special startup code to
15600 enable the application to run. The system must have PE installed in the
15601 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15602 must be overridden with the @option{-specs=} option to specify the
15603 appropriate directory location. The Parallel Environment does not
15604 support threads, so the @option{-mpe} option and the @option{-pthread}
15605 option are incompatible.
15607 @item -malign-natural
15608 @itemx -malign-power
15609 @opindex malign-natural
15610 @opindex malign-power
15611 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15612 @option{-malign-natural} overrides the ABI-defined alignment of larger
15613 types, such as floating-point doubles, on their natural size-based boundary.
15614 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15615 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15617 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15621 @itemx -mhard-float
15622 @opindex msoft-float
15623 @opindex mhard-float
15624 Generate code that does not use (uses) the floating-point register set.
15625 Software floating point emulation is provided if you use the
15626 @option{-msoft-float} option, and pass the option to GCC when linking.
15628 @item -msingle-float
15629 @itemx -mdouble-float
15630 @opindex msingle-float
15631 @opindex mdouble-float
15632 Generate code for single or double-precision floating point operations.
15633 @option{-mdouble-float} implies @option{-msingle-float}.
15636 @opindex msimple-fpu
15637 Do not generate sqrt and div instructions for hardware floating point unit.
15641 Specify type of floating point unit. Valid values are @var{sp_lite}
15642 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15643 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15644 and @var{dp_full} (equivalent to -mdouble-float).
15647 @opindex mxilinx-fpu
15648 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15651 @itemx -mno-multiple
15653 @opindex mno-multiple
15654 Generate code that uses (does not use) the load multiple word
15655 instructions and the store multiple word instructions. These
15656 instructions are generated by default on POWER systems, and not
15657 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15658 endian PowerPC systems, since those instructions do not work when the
15659 processor is in little endian mode. The exceptions are PPC740 and
15660 PPC750 which permit the instructions usage in little endian mode.
15665 @opindex mno-string
15666 Generate code that uses (does not use) the load string instructions
15667 and the store string word instructions to save multiple registers and
15668 do small block moves. These instructions are generated by default on
15669 POWER systems, and not generated on PowerPC systems. Do not use
15670 @option{-mstring} on little endian PowerPC systems, since those
15671 instructions do not work when the processor is in little endian mode.
15672 The exceptions are PPC740 and PPC750 which permit the instructions
15673 usage in little endian mode.
15678 @opindex mno-update
15679 Generate code that uses (does not use) the load or store instructions
15680 that update the base register to the address of the calculated memory
15681 location. These instructions are generated by default. If you use
15682 @option{-mno-update}, there is a small window between the time that the
15683 stack pointer is updated and the address of the previous frame is
15684 stored, which means code that walks the stack frame across interrupts or
15685 signals may get corrupted data.
15687 @item -mavoid-indexed-addresses
15688 @itemx -mno-avoid-indexed-addresses
15689 @opindex mavoid-indexed-addresses
15690 @opindex mno-avoid-indexed-addresses
15691 Generate code that tries to avoid (not avoid) the use of indexed load
15692 or store instructions. These instructions can incur a performance
15693 penalty on Power6 processors in certain situations, such as when
15694 stepping through large arrays that cross a 16M boundary. This option
15695 is enabled by default when targetting Power6 and disabled otherwise.
15698 @itemx -mno-fused-madd
15699 @opindex mfused-madd
15700 @opindex mno-fused-madd
15701 Generate code that uses (does not use) the floating point multiply and
15702 accumulate instructions. These instructions are generated by default
15703 if hardware floating point is used. The machine dependent
15704 @option{-mfused-madd} option is now mapped to the machine independent
15705 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15706 mapped to @option{-ffp-contract=off}.
15712 Generate code that uses (does not use) the half-word multiply and
15713 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15714 These instructions are generated by default when targetting those
15721 Generate code that uses (does not use) the string-search @samp{dlmzb}
15722 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15723 generated by default when targetting those processors.
15725 @item -mno-bit-align
15727 @opindex mno-bit-align
15728 @opindex mbit-align
15729 On System V.4 and embedded PowerPC systems do not (do) force structures
15730 and unions that contain bit-fields to be aligned to the base type of the
15733 For example, by default a structure containing nothing but 8
15734 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15735 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15736 the structure would be aligned to a 1 byte boundary and be one byte in
15739 @item -mno-strict-align
15740 @itemx -mstrict-align
15741 @opindex mno-strict-align
15742 @opindex mstrict-align
15743 On System V.4 and embedded PowerPC systems do not (do) assume that
15744 unaligned memory references will be handled by the system.
15746 @item -mrelocatable
15747 @itemx -mno-relocatable
15748 @opindex mrelocatable
15749 @opindex mno-relocatable
15750 Generate code that allows (does not allow) a static executable to be
15751 relocated to a different address at runtime. A simple embedded
15752 PowerPC system loader should relocate the entire contents of
15753 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15754 a table of 32-bit addresses generated by this option. For this to
15755 work, all objects linked together must be compiled with
15756 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15757 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15759 @item -mrelocatable-lib
15760 @itemx -mno-relocatable-lib
15761 @opindex mrelocatable-lib
15762 @opindex mno-relocatable-lib
15763 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15764 @code{.fixup} section to allow static executables to be relocated at
15765 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15766 alignment of @option{-mrelocatable}. Objects compiled with
15767 @option{-mrelocatable-lib} may be linked with objects compiled with
15768 any combination of the @option{-mrelocatable} options.
15774 On System V.4 and embedded PowerPC systems do not (do) assume that
15775 register 2 contains a pointer to a global area pointing to the addresses
15776 used in the program.
15779 @itemx -mlittle-endian
15781 @opindex mlittle-endian
15782 On System V.4 and embedded PowerPC systems compile code for the
15783 processor in little endian mode. The @option{-mlittle-endian} option is
15784 the same as @option{-mlittle}.
15787 @itemx -mbig-endian
15789 @opindex mbig-endian
15790 On System V.4 and embedded PowerPC systems compile code for the
15791 processor in big endian mode. The @option{-mbig-endian} option is
15792 the same as @option{-mbig}.
15794 @item -mdynamic-no-pic
15795 @opindex mdynamic-no-pic
15796 On Darwin and Mac OS X systems, compile code so that it is not
15797 relocatable, but that its external references are relocatable. The
15798 resulting code is suitable for applications, but not shared
15801 @item -msingle-pic-base
15802 @opindex msingle-pic-base
15803 Treat the register used for PIC addressing as read-only, rather than
15804 loading it in the prologue for each function. The run-time system is
15805 responsible for initializing this register with an appropriate value
15806 before execution begins.
15808 @item -mprioritize-restricted-insns=@var{priority}
15809 @opindex mprioritize-restricted-insns
15810 This option controls the priority that is assigned to
15811 dispatch-slot restricted instructions during the second scheduling
15812 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15813 @var{no/highest/second-highest} priority to dispatch slot restricted
15816 @item -msched-costly-dep=@var{dependence_type}
15817 @opindex msched-costly-dep
15818 This option controls which dependences are considered costly
15819 by the target during instruction scheduling. The argument
15820 @var{dependence_type} takes one of the following values:
15821 @var{no}: no dependence is costly,
15822 @var{all}: all dependences are costly,
15823 @var{true_store_to_load}: a true dependence from store to load is costly,
15824 @var{store_to_load}: any dependence from store to load is costly,
15825 @var{number}: any dependence which latency >= @var{number} is costly.
15827 @item -minsert-sched-nops=@var{scheme}
15828 @opindex minsert-sched-nops
15829 This option controls which nop insertion scheme will be used during
15830 the second scheduling pass. The argument @var{scheme} takes one of the
15832 @var{no}: Don't insert nops.
15833 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15834 according to the scheduler's grouping.
15835 @var{regroup_exact}: Insert nops to force costly dependent insns into
15836 separate groups. Insert exactly as many nops as needed to force an insn
15837 to a new group, according to the estimated processor grouping.
15838 @var{number}: Insert nops to force costly dependent insns into
15839 separate groups. Insert @var{number} nops to force an insn to a new group.
15842 @opindex mcall-sysv
15843 On System V.4 and embedded PowerPC systems compile code using calling
15844 conventions that adheres to the March 1995 draft of the System V
15845 Application Binary Interface, PowerPC processor supplement. This is the
15846 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15848 @item -mcall-sysv-eabi
15850 @opindex mcall-sysv-eabi
15851 @opindex mcall-eabi
15852 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15854 @item -mcall-sysv-noeabi
15855 @opindex mcall-sysv-noeabi
15856 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15858 @item -mcall-aixdesc
15860 On System V.4 and embedded PowerPC systems compile code for the AIX
15864 @opindex mcall-linux
15865 On System V.4 and embedded PowerPC systems compile code for the
15866 Linux-based GNU system.
15870 On System V.4 and embedded PowerPC systems compile code for the
15871 Hurd-based GNU system.
15873 @item -mcall-freebsd
15874 @opindex mcall-freebsd
15875 On System V.4 and embedded PowerPC systems compile code for the
15876 FreeBSD operating system.
15878 @item -mcall-netbsd
15879 @opindex mcall-netbsd
15880 On System V.4 and embedded PowerPC systems compile code for the
15881 NetBSD operating system.
15883 @item -mcall-openbsd
15884 @opindex mcall-netbsd
15885 On System V.4 and embedded PowerPC systems compile code for the
15886 OpenBSD operating system.
15888 @item -maix-struct-return
15889 @opindex maix-struct-return
15890 Return all structures in memory (as specified by the AIX ABI)@.
15892 @item -msvr4-struct-return
15893 @opindex msvr4-struct-return
15894 Return structures smaller than 8 bytes in registers (as specified by the
15897 @item -mabi=@var{abi-type}
15899 Extend the current ABI with a particular extension, or remove such extension.
15900 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15901 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15905 Extend the current ABI with SPE ABI extensions. This does not change
15906 the default ABI, instead it adds the SPE ABI extensions to the current
15910 @opindex mabi=no-spe
15911 Disable Booke SPE ABI extensions for the current ABI@.
15913 @item -mabi=ibmlongdouble
15914 @opindex mabi=ibmlongdouble
15915 Change the current ABI to use IBM extended precision long double.
15916 This is a PowerPC 32-bit SYSV ABI option.
15918 @item -mabi=ieeelongdouble
15919 @opindex mabi=ieeelongdouble
15920 Change the current ABI to use IEEE extended precision long double.
15921 This is a PowerPC 32-bit Linux ABI option.
15924 @itemx -mno-prototype
15925 @opindex mprototype
15926 @opindex mno-prototype
15927 On System V.4 and embedded PowerPC systems assume that all calls to
15928 variable argument functions are properly prototyped. Otherwise, the
15929 compiler must insert an instruction before every non prototyped call to
15930 set or clear bit 6 of the condition code register (@var{CR}) to
15931 indicate whether floating point values were passed in the floating point
15932 registers in case the function takes a variable arguments. With
15933 @option{-mprototype}, only calls to prototyped variable argument functions
15934 will set or clear the bit.
15938 On embedded PowerPC systems, assume that the startup module is called
15939 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15940 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15945 On embedded PowerPC systems, assume that the startup module is called
15946 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15951 On embedded PowerPC systems, assume that the startup module is called
15952 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15955 @item -myellowknife
15956 @opindex myellowknife
15957 On embedded PowerPC systems, assume that the startup module is called
15958 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15963 On System V.4 and embedded PowerPC systems, specify that you are
15964 compiling for a VxWorks system.
15968 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15969 header to indicate that @samp{eabi} extended relocations are used.
15975 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15976 Embedded Applications Binary Interface (eabi) which is a set of
15977 modifications to the System V.4 specifications. Selecting @option{-meabi}
15978 means that the stack is aligned to an 8 byte boundary, a function
15979 @code{__eabi} is called to from @code{main} to set up the eabi
15980 environment, and the @option{-msdata} option can use both @code{r2} and
15981 @code{r13} to point to two separate small data areas. Selecting
15982 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15983 do not call an initialization function from @code{main}, and the
15984 @option{-msdata} option will only use @code{r13} to point to a single
15985 small data area. The @option{-meabi} option is on by default if you
15986 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15989 @opindex msdata=eabi
15990 On System V.4 and embedded PowerPC systems, put small initialized
15991 @code{const} global and static data in the @samp{.sdata2} section, which
15992 is pointed to by register @code{r2}. Put small initialized
15993 non-@code{const} global and static data in the @samp{.sdata} section,
15994 which is pointed to by register @code{r13}. Put small uninitialized
15995 global and static data in the @samp{.sbss} section, which is adjacent to
15996 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15997 incompatible with the @option{-mrelocatable} option. The
15998 @option{-msdata=eabi} option also sets the @option{-memb} option.
16001 @opindex msdata=sysv
16002 On System V.4 and embedded PowerPC systems, put small global and static
16003 data in the @samp{.sdata} section, which is pointed to by register
16004 @code{r13}. Put small uninitialized global and static data in the
16005 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16006 The @option{-msdata=sysv} option is incompatible with the
16007 @option{-mrelocatable} option.
16009 @item -msdata=default
16011 @opindex msdata=default
16013 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16014 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16015 same as @option{-msdata=sysv}.
16018 @opindex msdata=data
16019 On System V.4 and embedded PowerPC systems, put small global
16020 data in the @samp{.sdata} section. Put small uninitialized global
16021 data in the @samp{.sbss} section. Do not use register @code{r13}
16022 to address small data however. This is the default behavior unless
16023 other @option{-msdata} options are used.
16027 @opindex msdata=none
16029 On embedded PowerPC systems, put all initialized global and static data
16030 in the @samp{.data} section, and all uninitialized data in the
16031 @samp{.bss} section.
16033 @item -mblock-move-inline-limit=@var{num}
16034 @opindex mblock-move-inline-limit
16035 Inline all block moves (such as calls to @code{memcpy} or structure
16036 copies) less than or equal to @var{num} bytes. The minimum value for
16037 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16038 targets. The default value is target-specific.
16042 @cindex smaller data references (PowerPC)
16043 @cindex .sdata/.sdata2 references (PowerPC)
16044 On embedded PowerPC systems, put global and static items less than or
16045 equal to @var{num} bytes into the small data or bss sections instead of
16046 the normal data or bss section. By default, @var{num} is 8. The
16047 @option{-G @var{num}} switch is also passed to the linker.
16048 All modules should be compiled with the same @option{-G @var{num}} value.
16051 @itemx -mno-regnames
16053 @opindex mno-regnames
16054 On System V.4 and embedded PowerPC systems do (do not) emit register
16055 names in the assembly language output using symbolic forms.
16058 @itemx -mno-longcall
16060 @opindex mno-longcall
16061 By default assume that all calls are far away so that a longer more
16062 expensive calling sequence is required. This is required for calls
16063 further than 32 megabytes (33,554,432 bytes) from the current location.
16064 A short call will be generated if the compiler knows
16065 the call cannot be that far away. This setting can be overridden by
16066 the @code{shortcall} function attribute, or by @code{#pragma
16069 Some linkers are capable of detecting out-of-range calls and generating
16070 glue code on the fly. On these systems, long calls are unnecessary and
16071 generate slower code. As of this writing, the AIX linker can do this,
16072 as can the GNU linker for PowerPC/64. It is planned to add this feature
16073 to the GNU linker for 32-bit PowerPC systems as well.
16075 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16076 callee, L42'', plus a ``branch island'' (glue code). The two target
16077 addresses represent the callee and the ``branch island''. The
16078 Darwin/PPC linker will prefer the first address and generate a ``bl
16079 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16080 otherwise, the linker will generate ``bl L42'' to call the ``branch
16081 island''. The ``branch island'' is appended to the body of the
16082 calling function; it computes the full 32-bit address of the callee
16085 On Mach-O (Darwin) systems, this option directs the compiler emit to
16086 the glue for every direct call, and the Darwin linker decides whether
16087 to use or discard it.
16089 In the future, we may cause GCC to ignore all longcall specifications
16090 when the linker is known to generate glue.
16092 @item -mtls-markers
16093 @itemx -mno-tls-markers
16094 @opindex mtls-markers
16095 @opindex mno-tls-markers
16096 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16097 specifying the function argument. The relocation allows ld to
16098 reliably associate function call with argument setup instructions for
16099 TLS optimization, which in turn allows gcc to better schedule the
16104 Adds support for multithreading with the @dfn{pthreads} library.
16105 This option sets flags for both the preprocessor and linker.
16110 This option will enable GCC to use the reciprocal estimate and
16111 reciprocal square root estimate instructions with additional
16112 Newton-Raphson steps to increase precision instead of doing a divide or
16113 square root and divide for floating point arguments. You should use
16114 the @option{-ffast-math} option when using @option{-mrecip} (or at
16115 least @option{-funsafe-math-optimizations},
16116 @option{-finite-math-only}, @option{-freciprocal-math} and
16117 @option{-fno-trapping-math}). Note that while the throughput of the
16118 sequence is generally higher than the throughput of the non-reciprocal
16119 instruction, the precision of the sequence can be decreased by up to 2
16120 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16123 @item -mrecip=@var{opt}
16124 @opindex mrecip=opt
16125 This option allows to control which reciprocal estimate instructions
16126 may be used. @var{opt} is a comma separated list of options, that may
16127 be preceded by a @code{!} to invert the option:
16128 @code{all}: enable all estimate instructions,
16129 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16130 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16131 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16132 @code{divf}: enable the single precision reciprocal approximation instructions;
16133 @code{divd}: enable the double precision reciprocal approximation instructions;
16134 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16135 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16136 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16138 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16139 all of the reciprocal estimate instructions, except for the
16140 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16141 which handle the double precision reciprocal square root calculations.
16143 @item -mrecip-precision
16144 @itemx -mno-recip-precision
16145 @opindex mrecip-precision
16146 Assume (do not assume) that the reciprocal estimate instructions
16147 provide higher precision estimates than is mandated by the powerpc
16148 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16149 automatically selects @option{-mrecip-precision}. The double
16150 precision square root estimate instructions are not generated by
16151 default on low precision machines, since they do not provide an
16152 estimate that converges after three steps.
16154 @item -mveclibabi=@var{type}
16155 @opindex mveclibabi
16156 Specifies the ABI type to use for vectorizing intrinsics using an
16157 external library. The only type supported at present is @code{mass},
16158 which specifies to use IBM's Mathematical Acceleration Subsystem
16159 (MASS) libraries for vectorizing intrinsics using external libraries.
16160 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16161 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16162 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16163 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16164 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16165 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16166 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16167 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16168 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16169 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16170 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16171 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16172 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16173 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16174 for power7. Both @option{-ftree-vectorize} and
16175 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16176 libraries will have to be specified at link time.
16181 Generate (do not generate) the @code{friz} instruction when the
16182 @option{-funsafe-math-optimizations} option is used to optimize
16183 rounding a floating point value to 64-bit integer and back to floating
16184 point. The @code{friz} instruction does not return the same value if
16185 the floating point number is too large to fit in an integer.
16189 @subsection RX Options
16192 These command line options are defined for RX targets:
16195 @item -m64bit-doubles
16196 @itemx -m32bit-doubles
16197 @opindex m64bit-doubles
16198 @opindex m32bit-doubles
16199 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16200 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16201 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16202 works on 32-bit values, which is why the default is
16203 @option{-m32bit-doubles}.
16209 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16210 floating point hardware. The default is enabled for the @var{RX600}
16211 series and disabled for the @var{RX200} series.
16213 Floating point instructions will only be generated for 32-bit floating
16214 point values however, so if the @option{-m64bit-doubles} option is in
16215 use then the FPU hardware will not be used for doubles.
16217 @emph{Note} If the @option{-fpu} option is enabled then
16218 @option{-funsafe-math-optimizations} is also enabled automatically.
16219 This is because the RX FPU instructions are themselves unsafe.
16221 @item -mcpu=@var{name}
16223 Selects the type of RX CPU to be targeted. Currently three types are
16224 supported, the generic @var{RX600} and @var{RX200} series hardware and
16225 the specific @var{RX610} cpu. The default is @var{RX600}.
16227 The only difference between @var{RX600} and @var{RX610} is that the
16228 @var{RX610} does not support the @code{MVTIPL} instruction.
16230 The @var{RX200} series does not have a hardware floating point unit
16231 and so @option{-nofpu} is enabled by default when this type is
16234 @item -mbig-endian-data
16235 @itemx -mlittle-endian-data
16236 @opindex mbig-endian-data
16237 @opindex mlittle-endian-data
16238 Store data (but not code) in the big-endian format. The default is
16239 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16242 @item -msmall-data-limit=@var{N}
16243 @opindex msmall-data-limit
16244 Specifies the maximum size in bytes of global and static variables
16245 which can be placed into the small data area. Using the small data
16246 area can lead to smaller and faster code, but the size of area is
16247 limited and it is up to the programmer to ensure that the area does
16248 not overflow. Also when the small data area is used one of the RX's
16249 registers (@code{r13}) is reserved for use pointing to this area, so
16250 it is no longer available for use by the compiler. This could result
16251 in slower and/or larger code if variables which once could have been
16252 held in @code{r13} are now pushed onto the stack.
16254 Note, common variables (variables which have not been initialised) and
16255 constants are not placed into the small data area as they are assigned
16256 to other sections in the output executable.
16258 The default value is zero, which disables this feature. Note, this
16259 feature is not enabled by default with higher optimization levels
16260 (@option{-O2} etc) because of the potentially detrimental effects of
16261 reserving register @code{r13}. It is up to the programmer to
16262 experiment and discover whether this feature is of benefit to their
16269 Use the simulator runtime. The default is to use the libgloss board
16272 @item -mas100-syntax
16273 @itemx -mno-as100-syntax
16274 @opindex mas100-syntax
16275 @opindex mno-as100-syntax
16276 When generating assembler output use a syntax that is compatible with
16277 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16278 assembler but it has some restrictions so generating it is not the
16281 @item -mmax-constant-size=@var{N}
16282 @opindex mmax-constant-size
16283 Specifies the maximum size, in bytes, of a constant that can be used as
16284 an operand in a RX instruction. Although the RX instruction set does
16285 allow constants of up to 4 bytes in length to be used in instructions,
16286 a longer value equates to a longer instruction. Thus in some
16287 circumstances it can be beneficial to restrict the size of constants
16288 that are used in instructions. Constants that are too big are instead
16289 placed into a constant pool and referenced via register indirection.
16291 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16292 or 4 means that constants of any size are allowed.
16296 Enable linker relaxation. Linker relaxation is a process whereby the
16297 linker will attempt to reduce the size of a program by finding shorter
16298 versions of various instructions. Disabled by default.
16300 @item -mint-register=@var{N}
16301 @opindex mint-register
16302 Specify the number of registers to reserve for fast interrupt handler
16303 functions. The value @var{N} can be between 0 and 4. A value of 1
16304 means that register @code{r13} will be reserved for the exclusive use
16305 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16306 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16307 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16308 A value of 0, the default, does not reserve any registers.
16310 @item -msave-acc-in-interrupts
16311 @opindex msave-acc-in-interrupts
16312 Specifies that interrupt handler functions should preserve the
16313 accumulator register. This is only necessary if normal code might use
16314 the accumulator register, for example because it performs 64-bit
16315 multiplications. The default is to ignore the accumulator as this
16316 makes the interrupt handlers faster.
16320 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16321 has special significance to the RX port when used with the
16322 @code{interrupt} function attribute. This attribute indicates a
16323 function intended to process fast interrupts. GCC will will ensure
16324 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16325 and/or @code{r13} and only provided that the normal use of the
16326 corresponding registers have been restricted via the
16327 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16330 @node S/390 and zSeries Options
16331 @subsection S/390 and zSeries Options
16332 @cindex S/390 and zSeries Options
16334 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16338 @itemx -msoft-float
16339 @opindex mhard-float
16340 @opindex msoft-float
16341 Use (do not use) the hardware floating-point instructions and registers
16342 for floating-point operations. When @option{-msoft-float} is specified,
16343 functions in @file{libgcc.a} will be used to perform floating-point
16344 operations. When @option{-mhard-float} is specified, the compiler
16345 generates IEEE floating-point instructions. This is the default.
16348 @itemx -mno-hard-dfp
16350 @opindex mno-hard-dfp
16351 Use (do not use) the hardware decimal-floating-point instructions for
16352 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16353 specified, functions in @file{libgcc.a} will be used to perform
16354 decimal-floating-point operations. When @option{-mhard-dfp} is
16355 specified, the compiler generates decimal-floating-point hardware
16356 instructions. This is the default for @option{-march=z9-ec} or higher.
16358 @item -mlong-double-64
16359 @itemx -mlong-double-128
16360 @opindex mlong-double-64
16361 @opindex mlong-double-128
16362 These switches control the size of @code{long double} type. A size
16363 of 64bit makes the @code{long double} type equivalent to the @code{double}
16364 type. This is the default.
16367 @itemx -mno-backchain
16368 @opindex mbackchain
16369 @opindex mno-backchain
16370 Store (do not store) the address of the caller's frame as backchain pointer
16371 into the callee's stack frame.
16372 A backchain may be needed to allow debugging using tools that do not understand
16373 DWARF-2 call frame information.
16374 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16375 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16376 the backchain is placed into the topmost word of the 96/160 byte register
16379 In general, code compiled with @option{-mbackchain} is call-compatible with
16380 code compiled with @option{-mmo-backchain}; however, use of the backchain
16381 for debugging purposes usually requires that the whole binary is built with
16382 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16383 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16384 to build a linux kernel use @option{-msoft-float}.
16386 The default is to not maintain the backchain.
16388 @item -mpacked-stack
16389 @itemx -mno-packed-stack
16390 @opindex mpacked-stack
16391 @opindex mno-packed-stack
16392 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16393 specified, the compiler uses the all fields of the 96/160 byte register save
16394 area only for their default purpose; unused fields still take up stack space.
16395 When @option{-mpacked-stack} is specified, register save slots are densely
16396 packed at the top of the register save area; unused space is reused for other
16397 purposes, allowing for more efficient use of the available stack space.
16398 However, when @option{-mbackchain} is also in effect, the topmost word of
16399 the save area is always used to store the backchain, and the return address
16400 register is always saved two words below the backchain.
16402 As long as the stack frame backchain is not used, code generated with
16403 @option{-mpacked-stack} is call-compatible with code generated with
16404 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16405 S/390 or zSeries generated code that uses the stack frame backchain at run
16406 time, not just for debugging purposes. Such code is not call-compatible
16407 with code compiled with @option{-mpacked-stack}. Also, note that the
16408 combination of @option{-mbackchain},
16409 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16410 to build a linux kernel use @option{-msoft-float}.
16412 The default is to not use the packed stack layout.
16415 @itemx -mno-small-exec
16416 @opindex msmall-exec
16417 @opindex mno-small-exec
16418 Generate (or do not generate) code using the @code{bras} instruction
16419 to do subroutine calls.
16420 This only works reliably if the total executable size does not
16421 exceed 64k. The default is to use the @code{basr} instruction instead,
16422 which does not have this limitation.
16428 When @option{-m31} is specified, generate code compliant to the
16429 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16430 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16431 particular to generate 64-bit instructions. For the @samp{s390}
16432 targets, the default is @option{-m31}, while the @samp{s390x}
16433 targets default to @option{-m64}.
16439 When @option{-mzarch} is specified, generate code using the
16440 instructions available on z/Architecture.
16441 When @option{-mesa} is specified, generate code using the
16442 instructions available on ESA/390. Note that @option{-mesa} is
16443 not possible with @option{-m64}.
16444 When generating code compliant to the GNU/Linux for S/390 ABI,
16445 the default is @option{-mesa}. When generating code compliant
16446 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16452 Generate (or do not generate) code using the @code{mvcle} instruction
16453 to perform block moves. When @option{-mno-mvcle} is specified,
16454 use a @code{mvc} loop instead. This is the default unless optimizing for
16461 Print (or do not print) additional debug information when compiling.
16462 The default is to not print debug information.
16464 @item -march=@var{cpu-type}
16466 Generate code that will run on @var{cpu-type}, which is the name of a system
16467 representing a certain processor type. Possible values for
16468 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16469 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16470 When generating code using the instructions available on z/Architecture,
16471 the default is @option{-march=z900}. Otherwise, the default is
16472 @option{-march=g5}.
16474 @item -mtune=@var{cpu-type}
16476 Tune to @var{cpu-type} everything applicable about the generated code,
16477 except for the ABI and the set of available instructions.
16478 The list of @var{cpu-type} values is the same as for @option{-march}.
16479 The default is the value used for @option{-march}.
16482 @itemx -mno-tpf-trace
16483 @opindex mtpf-trace
16484 @opindex mno-tpf-trace
16485 Generate code that adds (does not add) in TPF OS specific branches to trace
16486 routines in the operating system. This option is off by default, even
16487 when compiling for the TPF OS@.
16490 @itemx -mno-fused-madd
16491 @opindex mfused-madd
16492 @opindex mno-fused-madd
16493 Generate code that uses (does not use) the floating point multiply and
16494 accumulate instructions. These instructions are generated by default if
16495 hardware floating point is used.
16497 @item -mwarn-framesize=@var{framesize}
16498 @opindex mwarn-framesize
16499 Emit a warning if the current function exceeds the given frame size. Because
16500 this is a compile time check it doesn't need to be a real problem when the program
16501 runs. It is intended to identify functions which most probably cause
16502 a stack overflow. It is useful to be used in an environment with limited stack
16503 size e.g.@: the linux kernel.
16505 @item -mwarn-dynamicstack
16506 @opindex mwarn-dynamicstack
16507 Emit a warning if the function calls alloca or uses dynamically
16508 sized arrays. This is generally a bad idea with a limited stack size.
16510 @item -mstack-guard=@var{stack-guard}
16511 @itemx -mstack-size=@var{stack-size}
16512 @opindex mstack-guard
16513 @opindex mstack-size
16514 If these options are provided the s390 back end emits additional instructions in
16515 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16516 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16517 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16518 the frame size of the compiled function is chosen.
16519 These options are intended to be used to help debugging stack overflow problems.
16520 The additionally emitted code causes only little overhead and hence can also be
16521 used in production like systems without greater performance degradation. The given
16522 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16523 @var{stack-guard} without exceeding 64k.
16524 In order to be efficient the extra code makes the assumption that the stack starts
16525 at an address aligned to the value given by @var{stack-size}.
16526 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16529 @node Score Options
16530 @subsection Score Options
16531 @cindex Score Options
16533 These options are defined for Score implementations:
16538 Compile code for big endian mode. This is the default.
16542 Compile code for little endian mode.
16546 Disable generate bcnz instruction.
16550 Enable generate unaligned load and store instruction.
16554 Enable the use of multiply-accumulate instructions. Disabled by default.
16558 Specify the SCORE5 as the target architecture.
16562 Specify the SCORE5U of the target architecture.
16566 Specify the SCORE7 as the target architecture. This is the default.
16570 Specify the SCORE7D as the target architecture.
16574 @subsection SH Options
16576 These @samp{-m} options are defined for the SH implementations:
16581 Generate code for the SH1.
16585 Generate code for the SH2.
16588 Generate code for the SH2e.
16592 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16593 that the floating-point unit is not used.
16595 @item -m2a-single-only
16596 @opindex m2a-single-only
16597 Generate code for the SH2a-FPU, in such a way that no double-precision
16598 floating point operations are used.
16601 @opindex m2a-single
16602 Generate code for the SH2a-FPU assuming the floating-point unit is in
16603 single-precision mode by default.
16607 Generate code for the SH2a-FPU assuming the floating-point unit is in
16608 double-precision mode by default.
16612 Generate code for the SH3.
16616 Generate code for the SH3e.
16620 Generate code for the SH4 without a floating-point unit.
16622 @item -m4-single-only
16623 @opindex m4-single-only
16624 Generate code for the SH4 with a floating-point unit that only
16625 supports single-precision arithmetic.
16629 Generate code for the SH4 assuming the floating-point unit is in
16630 single-precision mode by default.
16634 Generate code for the SH4.
16638 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16639 floating-point unit is not used.
16641 @item -m4a-single-only
16642 @opindex m4a-single-only
16643 Generate code for the SH4a, in such a way that no double-precision
16644 floating point operations are used.
16647 @opindex m4a-single
16648 Generate code for the SH4a assuming the floating-point unit is in
16649 single-precision mode by default.
16653 Generate code for the SH4a.
16657 Same as @option{-m4a-nofpu}, except that it implicitly passes
16658 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16659 instructions at the moment.
16663 Compile code for the processor in big endian mode.
16667 Compile code for the processor in little endian mode.
16671 Align doubles at 64-bit boundaries. Note that this changes the calling
16672 conventions, and thus some functions from the standard C library will
16673 not work unless you recompile it first with @option{-mdalign}.
16677 Shorten some address references at link time, when possible; uses the
16678 linker option @option{-relax}.
16682 Use 32-bit offsets in @code{switch} tables. The default is to use
16687 Enable the use of bit manipulation instructions on SH2A.
16691 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16692 alignment constraints.
16696 Comply with the calling conventions defined by Renesas.
16700 Comply with the calling conventions defined by Renesas.
16704 Comply with the calling conventions defined for GCC before the Renesas
16705 conventions were available. This option is the default for all
16706 targets of the SH toolchain except for @samp{sh-symbianelf}.
16709 @opindex mnomacsave
16710 Mark the @code{MAC} register as call-clobbered, even if
16711 @option{-mhitachi} is given.
16715 Increase IEEE-compliance of floating-point code.
16716 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16717 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16718 comparisons of NANs / infinities incurs extra overhead in every
16719 floating point comparison, therefore the default is set to
16720 @option{-ffinite-math-only}.
16722 @item -minline-ic_invalidate
16723 @opindex minline-ic_invalidate
16724 Inline code to invalidate instruction cache entries after setting up
16725 nested function trampolines.
16726 This option has no effect if -musermode is in effect and the selected
16727 code generation option (e.g. -m4) does not allow the use of the icbi
16729 If the selected code generation option does not allow the use of the icbi
16730 instruction, and -musermode is not in effect, the inlined code will
16731 manipulate the instruction cache address array directly with an associative
16732 write. This not only requires privileged mode, but it will also
16733 fail if the cache line had been mapped via the TLB and has become unmapped.
16737 Dump instruction size and location in the assembly code.
16740 @opindex mpadstruct
16741 This option is deprecated. It pads structures to multiple of 4 bytes,
16742 which is incompatible with the SH ABI@.
16746 Optimize for space instead of speed. Implied by @option{-Os}.
16749 @opindex mprefergot
16750 When generating position-independent code, emit function calls using
16751 the Global Offset Table instead of the Procedure Linkage Table.
16755 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16756 if the inlined code would not work in user mode.
16757 This is the default when the target is @code{sh-*-linux*}.
16759 @item -multcost=@var{number}
16760 @opindex multcost=@var{number}
16761 Set the cost to assume for a multiply insn.
16763 @item -mdiv=@var{strategy}
16764 @opindex mdiv=@var{strategy}
16765 Set the division strategy to use for SHmedia code. @var{strategy} must be
16766 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16767 inv:call2, inv:fp .
16768 "fp" performs the operation in floating point. This has a very high latency,
16769 but needs only a few instructions, so it might be a good choice if
16770 your code has enough easily exploitable ILP to allow the compiler to
16771 schedule the floating point instructions together with other instructions.
16772 Division by zero causes a floating point exception.
16773 "inv" uses integer operations to calculate the inverse of the divisor,
16774 and then multiplies the dividend with the inverse. This strategy allows
16775 cse and hoisting of the inverse calculation. Division by zero calculates
16776 an unspecified result, but does not trap.
16777 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16778 have been found, or if the entire operation has been hoisted to the same
16779 place, the last stages of the inverse calculation are intertwined with the
16780 final multiply to reduce the overall latency, at the expense of using a few
16781 more instructions, and thus offering fewer scheduling opportunities with
16783 "call" calls a library function that usually implements the inv:minlat
16785 This gives high code density for m5-*media-nofpu compilations.
16786 "call2" uses a different entry point of the same library function, where it
16787 assumes that a pointer to a lookup table has already been set up, which
16788 exposes the pointer load to cse / code hoisting optimizations.
16789 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16790 code generation, but if the code stays unoptimized, revert to the "call",
16791 "call2", or "fp" strategies, respectively. Note that the
16792 potentially-trapping side effect of division by zero is carried by a
16793 separate instruction, so it is possible that all the integer instructions
16794 are hoisted out, but the marker for the side effect stays where it is.
16795 A recombination to fp operations or a call is not possible in that case.
16796 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16797 that the inverse calculation was nor separated from the multiply, they speed
16798 up division where the dividend fits into 20 bits (plus sign where applicable),
16799 by inserting a test to skip a number of operations in this case; this test
16800 slows down the case of larger dividends. inv20u assumes the case of a such
16801 a small dividend to be unlikely, and inv20l assumes it to be likely.
16803 @item -maccumulate-outgoing-args
16804 @opindex maccumulate-outgoing-args
16805 Reserve space once for outgoing arguments in the function prologue rather
16806 than around each call. Generally beneficial for performance and size. Also
16807 needed for unwinding to avoid changing the stack frame around conditional code.
16809 @item -mdivsi3_libfunc=@var{name}
16810 @opindex mdivsi3_libfunc=@var{name}
16811 Set the name of the library function used for 32 bit signed division to
16812 @var{name}. This only affect the name used in the call and inv:call
16813 division strategies, and the compiler will still expect the same
16814 sets of input/output/clobbered registers as if this option was not present.
16816 @item -mfixed-range=@var{register-range}
16817 @opindex mfixed-range
16818 Generate code treating the given register range as fixed registers.
16819 A fixed register is one that the register allocator can not use. This is
16820 useful when compiling kernel code. A register range is specified as
16821 two registers separated by a dash. Multiple register ranges can be
16822 specified separated by a comma.
16824 @item -madjust-unroll
16825 @opindex madjust-unroll
16826 Throttle unrolling to avoid thrashing target registers.
16827 This option only has an effect if the gcc code base supports the
16828 TARGET_ADJUST_UNROLL_MAX target hook.
16830 @item -mindexed-addressing
16831 @opindex mindexed-addressing
16832 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16833 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16834 semantics for the indexed addressing mode. The architecture allows the
16835 implementation of processors with 64 bit MMU, which the OS could use to
16836 get 32 bit addressing, but since no current hardware implementation supports
16837 this or any other way to make the indexed addressing mode safe to use in
16838 the 32 bit ABI, the default is -mno-indexed-addressing.
16840 @item -mgettrcost=@var{number}
16841 @opindex mgettrcost=@var{number}
16842 Set the cost assumed for the gettr instruction to @var{number}.
16843 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16847 Assume pt* instructions won't trap. This will generally generate better
16848 scheduled code, but is unsafe on current hardware. The current architecture
16849 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16850 This has the unintentional effect of making it unsafe to schedule ptabs /
16851 ptrel before a branch, or hoist it out of a loop. For example,
16852 __do_global_ctors, a part of libgcc that runs constructors at program
16853 startup, calls functions in a list which is delimited by @minus{}1. With the
16854 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16855 That means that all the constructors will be run a bit quicker, but when
16856 the loop comes to the end of the list, the program crashes because ptabs
16857 loads @minus{}1 into a target register. Since this option is unsafe for any
16858 hardware implementing the current architecture specification, the default
16859 is -mno-pt-fixed. Unless the user specifies a specific cost with
16860 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16861 this deters register allocation using target registers for storing
16864 @item -minvalid-symbols
16865 @opindex minvalid-symbols
16866 Assume symbols might be invalid. Ordinary function symbols generated by
16867 the compiler will always be valid to load with movi/shori/ptabs or
16868 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16869 to generate symbols that will cause ptabs / ptrel to trap.
16870 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16871 It will then prevent cross-basic-block cse, hoisting and most scheduling
16872 of symbol loads. The default is @option{-mno-invalid-symbols}.
16875 @node Solaris 2 Options
16876 @subsection Solaris 2 Options
16877 @cindex Solaris 2 options
16879 These @samp{-m} options are supported on Solaris 2:
16882 @item -mimpure-text
16883 @opindex mimpure-text
16884 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16885 the compiler to not pass @option{-z text} to the linker when linking a
16886 shared object. Using this option, you can link position-dependent
16887 code into a shared object.
16889 @option{-mimpure-text} suppresses the ``relocations remain against
16890 allocatable but non-writable sections'' linker error message.
16891 However, the necessary relocations will trigger copy-on-write, and the
16892 shared object is not actually shared across processes. Instead of
16893 using @option{-mimpure-text}, you should compile all source code with
16894 @option{-fpic} or @option{-fPIC}.
16898 These switches are supported in addition to the above on Solaris 2:
16903 Add support for multithreading using the Solaris threads library. This
16904 option sets flags for both the preprocessor and linker. This option does
16905 not affect the thread safety of object code produced by the compiler or
16906 that of libraries supplied with it.
16910 Add support for multithreading using the POSIX threads library. This
16911 option sets flags for both the preprocessor and linker. This option does
16912 not affect the thread safety of object code produced by the compiler or
16913 that of libraries supplied with it.
16917 This is a synonym for @option{-pthreads}.
16920 @node SPARC Options
16921 @subsection SPARC Options
16922 @cindex SPARC options
16924 These @samp{-m} options are supported on the SPARC:
16927 @item -mno-app-regs
16929 @opindex mno-app-regs
16931 Specify @option{-mapp-regs} to generate output using the global registers
16932 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16935 To be fully SVR4 ABI compliant at the cost of some performance loss,
16936 specify @option{-mno-app-regs}. You should compile libraries and system
16937 software with this option.
16940 @itemx -mhard-float
16942 @opindex mhard-float
16943 Generate output containing floating point instructions. This is the
16947 @itemx -msoft-float
16949 @opindex msoft-float
16950 Generate output containing library calls for floating point.
16951 @strong{Warning:} the requisite libraries are not available for all SPARC
16952 targets. Normally the facilities of the machine's usual C compiler are
16953 used, but this cannot be done directly in cross-compilation. You must make
16954 your own arrangements to provide suitable library functions for
16955 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16956 @samp{sparclite-*-*} do provide software floating point support.
16958 @option{-msoft-float} changes the calling convention in the output file;
16959 therefore, it is only useful if you compile @emph{all} of a program with
16960 this option. In particular, you need to compile @file{libgcc.a}, the
16961 library that comes with GCC, with @option{-msoft-float} in order for
16964 @item -mhard-quad-float
16965 @opindex mhard-quad-float
16966 Generate output containing quad-word (long double) floating point
16969 @item -msoft-quad-float
16970 @opindex msoft-quad-float
16971 Generate output containing library calls for quad-word (long double)
16972 floating point instructions. The functions called are those specified
16973 in the SPARC ABI@. This is the default.
16975 As of this writing, there are no SPARC implementations that have hardware
16976 support for the quad-word floating point instructions. They all invoke
16977 a trap handler for one of these instructions, and then the trap handler
16978 emulates the effect of the instruction. Because of the trap handler overhead,
16979 this is much slower than calling the ABI library routines. Thus the
16980 @option{-msoft-quad-float} option is the default.
16982 @item -mno-unaligned-doubles
16983 @itemx -munaligned-doubles
16984 @opindex mno-unaligned-doubles
16985 @opindex munaligned-doubles
16986 Assume that doubles have 8 byte alignment. This is the default.
16988 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16989 alignment only if they are contained in another type, or if they have an
16990 absolute address. Otherwise, it assumes they have 4 byte alignment.
16991 Specifying this option avoids some rare compatibility problems with code
16992 generated by other compilers. It is not the default because it results
16993 in a performance loss, especially for floating point code.
16995 @item -mno-faster-structs
16996 @itemx -mfaster-structs
16997 @opindex mno-faster-structs
16998 @opindex mfaster-structs
16999 With @option{-mfaster-structs}, the compiler assumes that structures
17000 should have 8 byte alignment. This enables the use of pairs of
17001 @code{ldd} and @code{std} instructions for copies in structure
17002 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17003 However, the use of this changed alignment directly violates the SPARC
17004 ABI@. Thus, it's intended only for use on targets where the developer
17005 acknowledges that their resulting code will not be directly in line with
17006 the rules of the ABI@.
17008 @item -mcpu=@var{cpu_type}
17010 Set the instruction set, register set, and instruction scheduling parameters
17011 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17012 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17013 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17014 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17015 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
17017 Default instruction scheduling parameters are used for values that select
17018 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17019 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17021 Here is a list of each supported architecture and their supported
17026 v8: supersparc, hypersparc, leon
17027 sparclite: f930, f934, sparclite86x
17029 v9: ultrasparc, ultrasparc3, niagara, niagara2
17032 By default (unless configured otherwise), GCC generates code for the V7
17033 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17034 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17035 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17036 SPARCStation 1, 2, IPX etc.
17038 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17039 architecture. The only difference from V7 code is that the compiler emits
17040 the integer multiply and integer divide instructions which exist in SPARC-V8
17041 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17042 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17045 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17046 the SPARC architecture. This adds the integer multiply, integer divide step
17047 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17048 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17049 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17050 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17051 MB86934 chip, which is the more recent SPARClite with FPU@.
17053 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17054 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17055 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17056 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17057 optimizes it for the TEMIC SPARClet chip.
17059 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17060 architecture. This adds 64-bit integer and floating-point move instructions,
17061 3 additional floating-point condition code registers and conditional move
17062 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17063 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17064 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17065 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17066 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17067 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17068 additionally optimizes it for Sun UltraSPARC T2 chips.
17070 @item -mtune=@var{cpu_type}
17072 Set the instruction scheduling parameters for machine type
17073 @var{cpu_type}, but do not set the instruction set or register set that the
17074 option @option{-mcpu=@var{cpu_type}} would.
17076 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17077 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17078 that select a particular cpu implementation. Those are @samp{cypress},
17079 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17080 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17081 @samp{niagara}, and @samp{niagara2}.
17086 @opindex mno-v8plus
17087 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17088 difference from the V8 ABI is that the global and out registers are
17089 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17090 mode for all SPARC-V9 processors.
17096 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17097 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17100 These @samp{-m} options are supported in addition to the above
17101 on SPARC-V9 processors in 64-bit environments:
17104 @item -mlittle-endian
17105 @opindex mlittle-endian
17106 Generate code for a processor running in little-endian mode. It is only
17107 available for a few configurations and most notably not on Solaris and Linux.
17113 Generate code for a 32-bit or 64-bit environment.
17114 The 32-bit environment sets int, long and pointer to 32 bits.
17115 The 64-bit environment sets int to 32 bits and long and pointer
17118 @item -mcmodel=medlow
17119 @opindex mcmodel=medlow
17120 Generate code for the Medium/Low code model: 64-bit addresses, programs
17121 must be linked in the low 32 bits of memory. Programs can be statically
17122 or dynamically linked.
17124 @item -mcmodel=medmid
17125 @opindex mcmodel=medmid
17126 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17127 must be linked in the low 44 bits of memory, the text and data segments must
17128 be less than 2GB in size and the data segment must be located within 2GB of
17131 @item -mcmodel=medany
17132 @opindex mcmodel=medany
17133 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17134 may be linked anywhere in memory, the text and data segments must be less
17135 than 2GB in size and the data segment must be located within 2GB of the
17138 @item -mcmodel=embmedany
17139 @opindex mcmodel=embmedany
17140 Generate code for the Medium/Anywhere code model for embedded systems:
17141 64-bit addresses, the text and data segments must be less than 2GB in
17142 size, both starting anywhere in memory (determined at link time). The
17143 global register %g4 points to the base of the data segment. Programs
17144 are statically linked and PIC is not supported.
17147 @itemx -mno-stack-bias
17148 @opindex mstack-bias
17149 @opindex mno-stack-bias
17150 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17151 frame pointer if present, are offset by @minus{}2047 which must be added back
17152 when making stack frame references. This is the default in 64-bit mode.
17153 Otherwise, assume no such offset is present.
17157 @subsection SPU Options
17158 @cindex SPU options
17160 These @samp{-m} options are supported on the SPU:
17164 @itemx -merror-reloc
17165 @opindex mwarn-reloc
17166 @opindex merror-reloc
17168 The loader for SPU does not handle dynamic relocations. By default, GCC
17169 will give an error when it generates code that requires a dynamic
17170 relocation. @option{-mno-error-reloc} disables the error,
17171 @option{-mwarn-reloc} will generate a warning instead.
17174 @itemx -munsafe-dma
17176 @opindex munsafe-dma
17178 Instructions which initiate or test completion of DMA must not be
17179 reordered with respect to loads and stores of the memory which is being
17180 accessed. Users typically address this problem using the volatile
17181 keyword, but that can lead to inefficient code in places where the
17182 memory is known to not change. Rather than mark the memory as volatile
17183 we treat the DMA instructions as potentially effecting all memory. With
17184 @option{-munsafe-dma} users must use the volatile keyword to protect
17187 @item -mbranch-hints
17188 @opindex mbranch-hints
17190 By default, GCC will generate a branch hint instruction to avoid
17191 pipeline stalls for always taken or probably taken branches. A hint
17192 will not be generated closer than 8 instructions away from its branch.
17193 There is little reason to disable them, except for debugging purposes,
17194 or to make an object a little bit smaller.
17198 @opindex msmall-mem
17199 @opindex mlarge-mem
17201 By default, GCC generates code assuming that addresses are never larger
17202 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17203 a full 32 bit address.
17208 By default, GCC links against startup code that assumes the SPU-style
17209 main function interface (which has an unconventional parameter list).
17210 With @option{-mstdmain}, GCC will link your program against startup
17211 code that assumes a C99-style interface to @code{main}, including a
17212 local copy of @code{argv} strings.
17214 @item -mfixed-range=@var{register-range}
17215 @opindex mfixed-range
17216 Generate code treating the given register range as fixed registers.
17217 A fixed register is one that the register allocator can not use. This is
17218 useful when compiling kernel code. A register range is specified as
17219 two registers separated by a dash. Multiple register ranges can be
17220 specified separated by a comma.
17226 Compile code assuming that pointers to the PPU address space accessed
17227 via the @code{__ea} named address space qualifier are either 32 or 64
17228 bits wide. The default is 32 bits. As this is an ABI changing option,
17229 all object code in an executable must be compiled with the same setting.
17231 @item -maddress-space-conversion
17232 @itemx -mno-address-space-conversion
17233 @opindex maddress-space-conversion
17234 @opindex mno-address-space-conversion
17235 Allow/disallow treating the @code{__ea} address space as superset
17236 of the generic address space. This enables explicit type casts
17237 between @code{__ea} and generic pointer as well as implicit
17238 conversions of generic pointers to @code{__ea} pointers. The
17239 default is to allow address space pointer conversions.
17241 @item -mcache-size=@var{cache-size}
17242 @opindex mcache-size
17243 This option controls the version of libgcc that the compiler links to an
17244 executable and selects a software-managed cache for accessing variables
17245 in the @code{__ea} address space with a particular cache size. Possible
17246 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17247 and @samp{128}. The default cache size is 64KB.
17249 @item -matomic-updates
17250 @itemx -mno-atomic-updates
17251 @opindex matomic-updates
17252 @opindex mno-atomic-updates
17253 This option controls the version of libgcc that the compiler links to an
17254 executable and selects whether atomic updates to the software-managed
17255 cache of PPU-side variables are used. If you use atomic updates, changes
17256 to a PPU variable from SPU code using the @code{__ea} named address space
17257 qualifier will not interfere with changes to other PPU variables residing
17258 in the same cache line from PPU code. If you do not use atomic updates,
17259 such interference may occur; however, writing back cache lines will be
17260 more efficient. The default behavior is to use atomic updates.
17263 @itemx -mdual-nops=@var{n}
17264 @opindex mdual-nops
17265 By default, GCC will insert nops to increase dual issue when it expects
17266 it to increase performance. @var{n} can be a value from 0 to 10. A
17267 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17268 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17270 @item -mhint-max-nops=@var{n}
17271 @opindex mhint-max-nops
17272 Maximum number of nops to insert for a branch hint. A branch hint must
17273 be at least 8 instructions away from the branch it is effecting. GCC
17274 will insert up to @var{n} nops to enforce this, otherwise it will not
17275 generate the branch hint.
17277 @item -mhint-max-distance=@var{n}
17278 @opindex mhint-max-distance
17279 The encoding of the branch hint instruction limits the hint to be within
17280 256 instructions of the branch it is effecting. By default, GCC makes
17281 sure it is within 125.
17284 @opindex msafe-hints
17285 Work around a hardware bug which causes the SPU to stall indefinitely.
17286 By default, GCC will insert the @code{hbrp} instruction to make sure
17287 this stall won't happen.
17291 @node System V Options
17292 @subsection Options for System V
17294 These additional options are available on System V Release 4 for
17295 compatibility with other compilers on those systems:
17300 Create a shared object.
17301 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17305 Identify the versions of each tool used by the compiler, in a
17306 @code{.ident} assembler directive in the output.
17310 Refrain from adding @code{.ident} directives to the output file (this is
17313 @item -YP,@var{dirs}
17315 Search the directories @var{dirs}, and no others, for libraries
17316 specified with @option{-l}.
17318 @item -Ym,@var{dir}
17320 Look in the directory @var{dir} to find the M4 preprocessor.
17321 The assembler uses this option.
17322 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17323 @c the generic assembler that comes with Solaris takes just -Ym.
17327 @subsection V850 Options
17328 @cindex V850 Options
17330 These @samp{-m} options are defined for V850 implementations:
17334 @itemx -mno-long-calls
17335 @opindex mlong-calls
17336 @opindex mno-long-calls
17337 Treat all calls as being far away (near). If calls are assumed to be
17338 far away, the compiler will always load the functions address up into a
17339 register, and call indirect through the pointer.
17345 Do not optimize (do optimize) basic blocks that use the same index
17346 pointer 4 or more times to copy pointer into the @code{ep} register, and
17347 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17348 option is on by default if you optimize.
17350 @item -mno-prolog-function
17351 @itemx -mprolog-function
17352 @opindex mno-prolog-function
17353 @opindex mprolog-function
17354 Do not use (do use) external functions to save and restore registers
17355 at the prologue and epilogue of a function. The external functions
17356 are slower, but use less code space if more than one function saves
17357 the same number of registers. The @option{-mprolog-function} option
17358 is on by default if you optimize.
17362 Try to make the code as small as possible. At present, this just turns
17363 on the @option{-mep} and @option{-mprolog-function} options.
17365 @item -mtda=@var{n}
17367 Put static or global variables whose size is @var{n} bytes or less into
17368 the tiny data area that register @code{ep} points to. The tiny data
17369 area can hold up to 256 bytes in total (128 bytes for byte references).
17371 @item -msda=@var{n}
17373 Put static or global variables whose size is @var{n} bytes or less into
17374 the small data area that register @code{gp} points to. The small data
17375 area can hold up to 64 kilobytes.
17377 @item -mzda=@var{n}
17379 Put static or global variables whose size is @var{n} bytes or less into
17380 the first 32 kilobytes of memory.
17384 Specify that the target processor is the V850.
17387 @opindex mbig-switch
17388 Generate code suitable for big switch tables. Use this option only if
17389 the assembler/linker complain about out of range branches within a switch
17394 This option will cause r2 and r5 to be used in the code generated by
17395 the compiler. This setting is the default.
17397 @item -mno-app-regs
17398 @opindex mno-app-regs
17399 This option will cause r2 and r5 to be treated as fixed registers.
17403 Specify that the target processor is the V850E2V3. The preprocessor
17404 constants @samp{__v850e2v3__} will be defined if
17405 this option is used.
17409 Specify that the target processor is the V850E2. The preprocessor
17410 constants @samp{__v850e2__} will be defined if
17414 Specify that the target processor is the V850E1. The preprocessor
17415 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17419 Specify that the target processor is the V850ES. This is an alias for
17420 the @option{-mv850e1} option.
17424 Specify that the target processor is the V850E@. The preprocessor
17425 constant @samp{__v850e__} will be defined if this option is used.
17427 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17428 nor @option{-mv850e2} nor @option{-mv850e2v3}
17429 are defined then a default target processor will be chosen and the
17430 relevant @samp{__v850*__} preprocessor constant will be defined.
17432 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17433 defined, regardless of which processor variant is the target.
17435 @item -mdisable-callt
17436 @opindex mdisable-callt
17437 This option will suppress generation of the CALLT instruction for the
17438 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17439 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17444 @subsection VAX Options
17445 @cindex VAX options
17447 These @samp{-m} options are defined for the VAX:
17452 Do not output certain jump instructions (@code{aobleq} and so on)
17453 that the Unix assembler for the VAX cannot handle across long
17458 Do output those jump instructions, on the assumption that you
17459 will assemble with the GNU assembler.
17463 Output code for g-format floating point numbers instead of d-format.
17466 @node VxWorks Options
17467 @subsection VxWorks Options
17468 @cindex VxWorks Options
17470 The options in this section are defined for all VxWorks targets.
17471 Options specific to the target hardware are listed with the other
17472 options for that target.
17477 GCC can generate code for both VxWorks kernels and real time processes
17478 (RTPs). This option switches from the former to the latter. It also
17479 defines the preprocessor macro @code{__RTP__}.
17482 @opindex non-static
17483 Link an RTP executable against shared libraries rather than static
17484 libraries. The options @option{-static} and @option{-shared} can
17485 also be used for RTPs (@pxref{Link Options}); @option{-static}
17492 These options are passed down to the linker. They are defined for
17493 compatibility with Diab.
17496 @opindex Xbind-lazy
17497 Enable lazy binding of function calls. This option is equivalent to
17498 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17502 Disable lazy binding of function calls. This option is the default and
17503 is defined for compatibility with Diab.
17506 @node x86-64 Options
17507 @subsection x86-64 Options
17508 @cindex x86-64 options
17510 These are listed under @xref{i386 and x86-64 Options}.
17512 @node i386 and x86-64 Windows Options
17513 @subsection i386 and x86-64 Windows Options
17514 @cindex i386 and x86-64 Windows Options
17516 These additional options are available for Windows targets:
17521 This option is available for Cygwin and MinGW targets. It
17522 specifies that a console application is to be generated, by
17523 instructing the linker to set the PE header subsystem type
17524 required for console applications.
17525 This is the default behavior for Cygwin and MinGW targets.
17529 This option is available for Cygwin and MinGW targets. It
17530 specifies that a DLL - a dynamic link library - is to be
17531 generated, enabling the selection of the required runtime
17532 startup object and entry point.
17534 @item -mnop-fun-dllimport
17535 @opindex mnop-fun-dllimport
17536 This option is available for Cygwin and MinGW targets. It
17537 specifies that the dllimport attribute should be ignored.
17541 This option is available for MinGW targets. It specifies
17542 that MinGW-specific thread support is to be used.
17546 This option is available for mingw-w64 targets. It specifies
17547 that the UNICODE macro is getting pre-defined and that the
17548 unicode capable runtime startup code is chosen.
17552 This option is available for Cygwin and MinGW targets. It
17553 specifies that the typical Windows pre-defined macros are to
17554 be set in the pre-processor, but does not influence the choice
17555 of runtime library/startup code.
17559 This option is available for Cygwin and MinGW targets. It
17560 specifies that a GUI application is to be generated by
17561 instructing the linker to set the PE header subsystem type
17564 @item -fno-set-stack-executable
17565 @opindex fno-set-stack-executable
17566 This option is available for MinGW targets. It specifies that
17567 the executable flag for stack used by nested functions isn't
17568 set. This is necessary for binaries running in kernel mode of
17569 Windows, as there the user32 API, which is used to set executable
17570 privileges, isn't available.
17572 @item -mpe-aligned-commons
17573 @opindex mpe-aligned-commons
17574 This option is available for Cygwin and MinGW targets. It
17575 specifies that the GNU extension to the PE file format that
17576 permits the correct alignment of COMMON variables should be
17577 used when generating code. It will be enabled by default if
17578 GCC detects that the target assembler found during configuration
17579 supports the feature.
17582 See also under @ref{i386 and x86-64 Options} for standard options.
17584 @node Xstormy16 Options
17585 @subsection Xstormy16 Options
17586 @cindex Xstormy16 Options
17588 These options are defined for Xstormy16:
17593 Choose startup files and linker script suitable for the simulator.
17596 @node Xtensa Options
17597 @subsection Xtensa Options
17598 @cindex Xtensa Options
17600 These options are supported for Xtensa targets:
17604 @itemx -mno-const16
17606 @opindex mno-const16
17607 Enable or disable use of @code{CONST16} instructions for loading
17608 constant values. The @code{CONST16} instruction is currently not a
17609 standard option from Tensilica. When enabled, @code{CONST16}
17610 instructions are always used in place of the standard @code{L32R}
17611 instructions. The use of @code{CONST16} is enabled by default only if
17612 the @code{L32R} instruction is not available.
17615 @itemx -mno-fused-madd
17616 @opindex mfused-madd
17617 @opindex mno-fused-madd
17618 Enable or disable use of fused multiply/add and multiply/subtract
17619 instructions in the floating-point option. This has no effect if the
17620 floating-point option is not also enabled. Disabling fused multiply/add
17621 and multiply/subtract instructions forces the compiler to use separate
17622 instructions for the multiply and add/subtract operations. This may be
17623 desirable in some cases where strict IEEE 754-compliant results are
17624 required: the fused multiply add/subtract instructions do not round the
17625 intermediate result, thereby producing results with @emph{more} bits of
17626 precision than specified by the IEEE standard. Disabling fused multiply
17627 add/subtract instructions also ensures that the program output is not
17628 sensitive to the compiler's ability to combine multiply and add/subtract
17631 @item -mserialize-volatile
17632 @itemx -mno-serialize-volatile
17633 @opindex mserialize-volatile
17634 @opindex mno-serialize-volatile
17635 When this option is enabled, GCC inserts @code{MEMW} instructions before
17636 @code{volatile} memory references to guarantee sequential consistency.
17637 The default is @option{-mserialize-volatile}. Use
17638 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17640 @item -mforce-no-pic
17641 @opindex mforce-no-pic
17642 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17643 position-independent code (PIC), this option disables PIC for compiling
17646 @item -mtext-section-literals
17647 @itemx -mno-text-section-literals
17648 @opindex mtext-section-literals
17649 @opindex mno-text-section-literals
17650 Control the treatment of literal pools. The default is
17651 @option{-mno-text-section-literals}, which places literals in a separate
17652 section in the output file. This allows the literal pool to be placed
17653 in a data RAM/ROM, and it also allows the linker to combine literal
17654 pools from separate object files to remove redundant literals and
17655 improve code size. With @option{-mtext-section-literals}, the literals
17656 are interspersed in the text section in order to keep them as close as
17657 possible to their references. This may be necessary for large assembly
17660 @item -mtarget-align
17661 @itemx -mno-target-align
17662 @opindex mtarget-align
17663 @opindex mno-target-align
17664 When this option is enabled, GCC instructs the assembler to
17665 automatically align instructions to reduce branch penalties at the
17666 expense of some code density. The assembler attempts to widen density
17667 instructions to align branch targets and the instructions following call
17668 instructions. If there are not enough preceding safe density
17669 instructions to align a target, no widening will be performed. The
17670 default is @option{-mtarget-align}. These options do not affect the
17671 treatment of auto-aligned instructions like @code{LOOP}, which the
17672 assembler will always align, either by widening density instructions or
17673 by inserting no-op instructions.
17676 @itemx -mno-longcalls
17677 @opindex mlongcalls
17678 @opindex mno-longcalls
17679 When this option is enabled, GCC instructs the assembler to translate
17680 direct calls to indirect calls unless it can determine that the target
17681 of a direct call is in the range allowed by the call instruction. This
17682 translation typically occurs for calls to functions in other source
17683 files. Specifically, the assembler translates a direct @code{CALL}
17684 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17685 The default is @option{-mno-longcalls}. This option should be used in
17686 programs where the call target can potentially be out of range. This
17687 option is implemented in the assembler, not the compiler, so the
17688 assembly code generated by GCC will still show direct call
17689 instructions---look at the disassembled object code to see the actual
17690 instructions. Note that the assembler will use an indirect call for
17691 every cross-file call, not just those that really will be out of range.
17694 @node zSeries Options
17695 @subsection zSeries Options
17696 @cindex zSeries options
17698 These are listed under @xref{S/390 and zSeries Options}.
17700 @node Code Gen Options
17701 @section Options for Code Generation Conventions
17702 @cindex code generation conventions
17703 @cindex options, code generation
17704 @cindex run-time options
17706 These machine-independent options control the interface conventions
17707 used in code generation.
17709 Most of them have both positive and negative forms; the negative form
17710 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17711 one of the forms is listed---the one which is not the default. You
17712 can figure out the other form by either removing @samp{no-} or adding
17716 @item -fbounds-check
17717 @opindex fbounds-check
17718 For front-ends that support it, generate additional code to check that
17719 indices used to access arrays are within the declared range. This is
17720 currently only supported by the Java and Fortran front-ends, where
17721 this option defaults to true and false respectively.
17725 This option generates traps for signed overflow on addition, subtraction,
17726 multiplication operations.
17730 This option instructs the compiler to assume that signed arithmetic
17731 overflow of addition, subtraction and multiplication wraps around
17732 using twos-complement representation. This flag enables some optimizations
17733 and disables others. This option is enabled by default for the Java
17734 front-end, as required by the Java language specification.
17737 @opindex fexceptions
17738 Enable exception handling. Generates extra code needed to propagate
17739 exceptions. For some targets, this implies GCC will generate frame
17740 unwind information for all functions, which can produce significant data
17741 size overhead, although it does not affect execution. If you do not
17742 specify this option, GCC will enable it by default for languages like
17743 C++ which normally require exception handling, and disable it for
17744 languages like C that do not normally require it. However, you may need
17745 to enable this option when compiling C code that needs to interoperate
17746 properly with exception handlers written in C++. You may also wish to
17747 disable this option if you are compiling older C++ programs that don't
17748 use exception handling.
17750 @item -fnon-call-exceptions
17751 @opindex fnon-call-exceptions
17752 Generate code that allows trapping instructions to throw exceptions.
17753 Note that this requires platform-specific runtime support that does
17754 not exist everywhere. Moreover, it only allows @emph{trapping}
17755 instructions to throw exceptions, i.e.@: memory references or floating
17756 point instructions. It does not allow exceptions to be thrown from
17757 arbitrary signal handlers such as @code{SIGALRM}.
17759 @item -funwind-tables
17760 @opindex funwind-tables
17761 Similar to @option{-fexceptions}, except that it will just generate any needed
17762 static data, but will not affect the generated code in any other way.
17763 You will normally not enable this option; instead, a language processor
17764 that needs this handling would enable it on your behalf.
17766 @item -fasynchronous-unwind-tables
17767 @opindex fasynchronous-unwind-tables
17768 Generate unwind table in dwarf2 format, if supported by target machine. The
17769 table is exact at each instruction boundary, so it can be used for stack
17770 unwinding from asynchronous events (such as debugger or garbage collector).
17772 @item -fpcc-struct-return
17773 @opindex fpcc-struct-return
17774 Return ``short'' @code{struct} and @code{union} values in memory like
17775 longer ones, rather than in registers. This convention is less
17776 efficient, but it has the advantage of allowing intercallability between
17777 GCC-compiled files and files compiled with other compilers, particularly
17778 the Portable C Compiler (pcc).
17780 The precise convention for returning structures in memory depends
17781 on the target configuration macros.
17783 Short structures and unions are those whose size and alignment match
17784 that of some integer type.
17786 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17787 switch is not binary compatible with code compiled with the
17788 @option{-freg-struct-return} switch.
17789 Use it to conform to a non-default application binary interface.
17791 @item -freg-struct-return
17792 @opindex freg-struct-return
17793 Return @code{struct} and @code{union} values in registers when possible.
17794 This is more efficient for small structures than
17795 @option{-fpcc-struct-return}.
17797 If you specify neither @option{-fpcc-struct-return} nor
17798 @option{-freg-struct-return}, GCC defaults to whichever convention is
17799 standard for the target. If there is no standard convention, GCC
17800 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17801 the principal compiler. In those cases, we can choose the standard, and
17802 we chose the more efficient register return alternative.
17804 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17805 switch is not binary compatible with code compiled with the
17806 @option{-fpcc-struct-return} switch.
17807 Use it to conform to a non-default application binary interface.
17809 @item -fshort-enums
17810 @opindex fshort-enums
17811 Allocate to an @code{enum} type only as many bytes as it needs for the
17812 declared range of possible values. Specifically, the @code{enum} type
17813 will be equivalent to the smallest integer type which has enough room.
17815 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17816 code that is not binary compatible with code generated without that switch.
17817 Use it to conform to a non-default application binary interface.
17819 @item -fshort-double
17820 @opindex fshort-double
17821 Use the same size for @code{double} as for @code{float}.
17823 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17824 code that is not binary compatible with code generated without that switch.
17825 Use it to conform to a non-default application binary interface.
17827 @item -fshort-wchar
17828 @opindex fshort-wchar
17829 Override the underlying type for @samp{wchar_t} to be @samp{short
17830 unsigned int} instead of the default for the target. This option is
17831 useful for building programs to run under WINE@.
17833 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17834 code that is not binary compatible with code generated without that switch.
17835 Use it to conform to a non-default application binary interface.
17838 @opindex fno-common
17839 In C code, controls the placement of uninitialized global variables.
17840 Unix C compilers have traditionally permitted multiple definitions of
17841 such variables in different compilation units by placing the variables
17843 This is the behavior specified by @option{-fcommon}, and is the default
17844 for GCC on most targets.
17845 On the other hand, this behavior is not required by ISO C, and on some
17846 targets may carry a speed or code size penalty on variable references.
17847 The @option{-fno-common} option specifies that the compiler should place
17848 uninitialized global variables in the data section of the object file,
17849 rather than generating them as common blocks.
17850 This has the effect that if the same variable is declared
17851 (without @code{extern}) in two different compilations,
17852 you will get a multiple-definition error when you link them.
17853 In this case, you must compile with @option{-fcommon} instead.
17854 Compiling with @option{-fno-common} is useful on targets for which
17855 it provides better performance, or if you wish to verify that the
17856 program will work on other systems which always treat uninitialized
17857 variable declarations this way.
17861 Ignore the @samp{#ident} directive.
17863 @item -finhibit-size-directive
17864 @opindex finhibit-size-directive
17865 Don't output a @code{.size} assembler directive, or anything else that
17866 would cause trouble if the function is split in the middle, and the
17867 two halves are placed at locations far apart in memory. This option is
17868 used when compiling @file{crtstuff.c}; you should not need to use it
17871 @item -fverbose-asm
17872 @opindex fverbose-asm
17873 Put extra commentary information in the generated assembly code to
17874 make it more readable. This option is generally only of use to those
17875 who actually need to read the generated assembly code (perhaps while
17876 debugging the compiler itself).
17878 @option{-fno-verbose-asm}, the default, causes the
17879 extra information to be omitted and is useful when comparing two assembler
17882 @item -frecord-gcc-switches
17883 @opindex frecord-gcc-switches
17884 This switch causes the command line that was used to invoke the
17885 compiler to be recorded into the object file that is being created.
17886 This switch is only implemented on some targets and the exact format
17887 of the recording is target and binary file format dependent, but it
17888 usually takes the form of a section containing ASCII text. This
17889 switch is related to the @option{-fverbose-asm} switch, but that
17890 switch only records information in the assembler output file as
17891 comments, so it never reaches the object file.
17895 @cindex global offset table
17897 Generate position-independent code (PIC) suitable for use in a shared
17898 library, if supported for the target machine. Such code accesses all
17899 constant addresses through a global offset table (GOT)@. The dynamic
17900 loader resolves the GOT entries when the program starts (the dynamic
17901 loader is not part of GCC; it is part of the operating system). If
17902 the GOT size for the linked executable exceeds a machine-specific
17903 maximum size, you get an error message from the linker indicating that
17904 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17905 instead. (These maximums are 8k on the SPARC and 32k
17906 on the m68k and RS/6000. The 386 has no such limit.)
17908 Position-independent code requires special support, and therefore works
17909 only on certain machines. For the 386, GCC supports PIC for System V
17910 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17911 position-independent.
17913 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17918 If supported for the target machine, emit position-independent code,
17919 suitable for dynamic linking and avoiding any limit on the size of the
17920 global offset table. This option makes a difference on the m68k,
17921 PowerPC and SPARC@.
17923 Position-independent code requires special support, and therefore works
17924 only on certain machines.
17926 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17933 These options are similar to @option{-fpic} and @option{-fPIC}, but
17934 generated position independent code can be only linked into executables.
17935 Usually these options are used when @option{-pie} GCC option will be
17936 used during linking.
17938 @option{-fpie} and @option{-fPIE} both define the macros
17939 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17940 for @option{-fpie} and 2 for @option{-fPIE}.
17942 @item -fno-jump-tables
17943 @opindex fno-jump-tables
17944 Do not use jump tables for switch statements even where it would be
17945 more efficient than other code generation strategies. This option is
17946 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17947 building code which forms part of a dynamic linker and cannot
17948 reference the address of a jump table. On some targets, jump tables
17949 do not require a GOT and this option is not needed.
17951 @item -ffixed-@var{reg}
17953 Treat the register named @var{reg} as a fixed register; generated code
17954 should never refer to it (except perhaps as a stack pointer, frame
17955 pointer or in some other fixed role).
17957 @var{reg} must be the name of a register. The register names accepted
17958 are machine-specific and are defined in the @code{REGISTER_NAMES}
17959 macro in the machine description macro file.
17961 This flag does not have a negative form, because it specifies a
17964 @item -fcall-used-@var{reg}
17965 @opindex fcall-used
17966 Treat the register named @var{reg} as an allocable register that is
17967 clobbered by function calls. It may be allocated for temporaries or
17968 variables that do not live across a call. Functions compiled this way
17969 will not save and restore the register @var{reg}.
17971 It is an error to used this flag with the frame pointer or stack pointer.
17972 Use of this flag for other registers that have fixed pervasive roles in
17973 the machine's execution model will produce disastrous results.
17975 This flag does not have a negative form, because it specifies a
17978 @item -fcall-saved-@var{reg}
17979 @opindex fcall-saved
17980 Treat the register named @var{reg} as an allocable register saved by
17981 functions. It may be allocated even for temporaries or variables that
17982 live across a call. Functions compiled this way will save and restore
17983 the register @var{reg} if they use it.
17985 It is an error to used this flag with the frame pointer or stack pointer.
17986 Use of this flag for other registers that have fixed pervasive roles in
17987 the machine's execution model will produce disastrous results.
17989 A different sort of disaster will result from the use of this flag for
17990 a register in which function values may be returned.
17992 This flag does not have a negative form, because it specifies a
17995 @item -fpack-struct[=@var{n}]
17996 @opindex fpack-struct
17997 Without a value specified, pack all structure members together without
17998 holes. When a value is specified (which must be a small power of two), pack
17999 structure members according to this value, representing the maximum
18000 alignment (that is, objects with default alignment requirements larger than
18001 this will be output potentially unaligned at the next fitting location.
18003 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18004 code that is not binary compatible with code generated without that switch.
18005 Additionally, it makes the code suboptimal.
18006 Use it to conform to a non-default application binary interface.
18008 @item -finstrument-functions
18009 @opindex finstrument-functions
18010 Generate instrumentation calls for entry and exit to functions. Just
18011 after function entry and just before function exit, the following
18012 profiling functions will be called with the address of the current
18013 function and its call site. (On some platforms,
18014 @code{__builtin_return_address} does not work beyond the current
18015 function, so the call site information may not be available to the
18016 profiling functions otherwise.)
18019 void __cyg_profile_func_enter (void *this_fn,
18021 void __cyg_profile_func_exit (void *this_fn,
18025 The first argument is the address of the start of the current function,
18026 which may be looked up exactly in the symbol table.
18028 This instrumentation is also done for functions expanded inline in other
18029 functions. The profiling calls will indicate where, conceptually, the
18030 inline function is entered and exited. This means that addressable
18031 versions of such functions must be available. If all your uses of a
18032 function are expanded inline, this may mean an additional expansion of
18033 code size. If you use @samp{extern inline} in your C code, an
18034 addressable version of such functions must be provided. (This is
18035 normally the case anyways, but if you get lucky and the optimizer always
18036 expands the functions inline, you might have gotten away without
18037 providing static copies.)
18039 A function may be given the attribute @code{no_instrument_function}, in
18040 which case this instrumentation will not be done. This can be used, for
18041 example, for the profiling functions listed above, high-priority
18042 interrupt routines, and any functions from which the profiling functions
18043 cannot safely be called (perhaps signal handlers, if the profiling
18044 routines generate output or allocate memory).
18046 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18047 @opindex finstrument-functions-exclude-file-list
18049 Set the list of functions that are excluded from instrumentation (see
18050 the description of @code{-finstrument-functions}). If the file that
18051 contains a function definition matches with one of @var{file}, then
18052 that function is not instrumented. The match is done on substrings:
18053 if the @var{file} parameter is a substring of the file name, it is
18054 considered to be a match.
18059 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18063 will exclude any inline function defined in files whose pathnames
18064 contain @code{/bits/stl} or @code{include/sys}.
18066 If, for some reason, you want to include letter @code{','} in one of
18067 @var{sym}, write @code{'\,'}. For example,
18068 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18069 (note the single quote surrounding the option).
18071 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18072 @opindex finstrument-functions-exclude-function-list
18074 This is similar to @code{-finstrument-functions-exclude-file-list},
18075 but this option sets the list of function names to be excluded from
18076 instrumentation. The function name to be matched is its user-visible
18077 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18078 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18079 match is done on substrings: if the @var{sym} parameter is a substring
18080 of the function name, it is considered to be a match. For C99 and C++
18081 extended identifiers, the function name must be given in UTF-8, not
18082 using universal character names.
18084 @item -fstack-check
18085 @opindex fstack-check
18086 Generate code to verify that you do not go beyond the boundary of the
18087 stack. You should specify this flag if you are running in an
18088 environment with multiple threads, but only rarely need to specify it in
18089 a single-threaded environment since stack overflow is automatically
18090 detected on nearly all systems if there is only one stack.
18092 Note that this switch does not actually cause checking to be done; the
18093 operating system or the language runtime must do that. The switch causes
18094 generation of code to ensure that they see the stack being extended.
18096 You can additionally specify a string parameter: @code{no} means no
18097 checking, @code{generic} means force the use of old-style checking,
18098 @code{specific} means use the best checking method and is equivalent
18099 to bare @option{-fstack-check}.
18101 Old-style checking is a generic mechanism that requires no specific
18102 target support in the compiler but comes with the following drawbacks:
18106 Modified allocation strategy for large objects: they will always be
18107 allocated dynamically if their size exceeds a fixed threshold.
18110 Fixed limit on the size of the static frame of functions: when it is
18111 topped by a particular function, stack checking is not reliable and
18112 a warning is issued by the compiler.
18115 Inefficiency: because of both the modified allocation strategy and the
18116 generic implementation, the performances of the code are hampered.
18119 Note that old-style stack checking is also the fallback method for
18120 @code{specific} if no target support has been added in the compiler.
18122 @item -fstack-limit-register=@var{reg}
18123 @itemx -fstack-limit-symbol=@var{sym}
18124 @itemx -fno-stack-limit
18125 @opindex fstack-limit-register
18126 @opindex fstack-limit-symbol
18127 @opindex fno-stack-limit
18128 Generate code to ensure that the stack does not grow beyond a certain value,
18129 either the value of a register or the address of a symbol. If the stack
18130 would grow beyond the value, a signal is raised. For most targets,
18131 the signal is raised before the stack overruns the boundary, so
18132 it is possible to catch the signal without taking special precautions.
18134 For instance, if the stack starts at absolute address @samp{0x80000000}
18135 and grows downwards, you can use the flags
18136 @option{-fstack-limit-symbol=__stack_limit} and
18137 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18138 of 128KB@. Note that this may only work with the GNU linker.
18140 @item -fsplit-stack
18141 @opindex fsplit-stack
18142 Generate code to automatically split the stack before it overflows.
18143 The resulting program has a discontiguous stack which can only
18144 overflow if the program is unable to allocate any more memory. This
18145 is most useful when running threaded programs, as it is no longer
18146 necessary to calculate a good stack size to use for each thread. This
18147 is currently only implemented for the i386 and x86_64 backends running
18150 When code compiled with @option{-fsplit-stack} calls code compiled
18151 without @option{-fsplit-stack}, there may not be much stack space
18152 available for the latter code to run. If compiling all code,
18153 including library code, with @option{-fsplit-stack} is not an option,
18154 then the linker can fix up these calls so that the code compiled
18155 without @option{-fsplit-stack} always has a large stack. Support for
18156 this is implemented in the gold linker in GNU binutils release 2.21
18159 @item -fleading-underscore
18160 @opindex fleading-underscore
18161 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18162 change the way C symbols are represented in the object file. One use
18163 is to help link with legacy assembly code.
18165 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18166 generate code that is not binary compatible with code generated without that
18167 switch. Use it to conform to a non-default application binary interface.
18168 Not all targets provide complete support for this switch.
18170 @item -ftls-model=@var{model}
18171 @opindex ftls-model
18172 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18173 The @var{model} argument should be one of @code{global-dynamic},
18174 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18176 The default without @option{-fpic} is @code{initial-exec}; with
18177 @option{-fpic} the default is @code{global-dynamic}.
18179 @item -fvisibility=@var{default|internal|hidden|protected}
18180 @opindex fvisibility
18181 Set the default ELF image symbol visibility to the specified option---all
18182 symbols will be marked with this unless overridden within the code.
18183 Using this feature can very substantially improve linking and
18184 load times of shared object libraries, produce more optimized
18185 code, provide near-perfect API export and prevent symbol clashes.
18186 It is @strong{strongly} recommended that you use this in any shared objects
18189 Despite the nomenclature, @code{default} always means public; i.e.,
18190 available to be linked against from outside the shared object.
18191 @code{protected} and @code{internal} are pretty useless in real-world
18192 usage so the only other commonly used option will be @code{hidden}.
18193 The default if @option{-fvisibility} isn't specified is
18194 @code{default}, i.e., make every
18195 symbol public---this causes the same behavior as previous versions of
18198 A good explanation of the benefits offered by ensuring ELF
18199 symbols have the correct visibility is given by ``How To Write
18200 Shared Libraries'' by Ulrich Drepper (which can be found at
18201 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18202 solution made possible by this option to marking things hidden when
18203 the default is public is to make the default hidden and mark things
18204 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18205 and @code{__attribute__ ((visibility("default")))} instead of
18206 @code{__declspec(dllexport)} you get almost identical semantics with
18207 identical syntax. This is a great boon to those working with
18208 cross-platform projects.
18210 For those adding visibility support to existing code, you may find
18211 @samp{#pragma GCC visibility} of use. This works by you enclosing
18212 the declarations you wish to set visibility for with (for example)
18213 @samp{#pragma GCC visibility push(hidden)} and
18214 @samp{#pragma GCC visibility pop}.
18215 Bear in mind that symbol visibility should be viewed @strong{as
18216 part of the API interface contract} and thus all new code should
18217 always specify visibility when it is not the default; i.e., declarations
18218 only for use within the local DSO should @strong{always} be marked explicitly
18219 as hidden as so to avoid PLT indirection overheads---making this
18220 abundantly clear also aids readability and self-documentation of the code.
18221 Note that due to ISO C++ specification requirements, operator new and
18222 operator delete must always be of default visibility.
18224 Be aware that headers from outside your project, in particular system
18225 headers and headers from any other library you use, may not be
18226 expecting to be compiled with visibility other than the default. You
18227 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18228 before including any such headers.
18230 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18231 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18232 no modifications. However, this means that calls to @samp{extern}
18233 functions with no explicit visibility will use the PLT, so it is more
18234 effective to use @samp{__attribute ((visibility))} and/or
18235 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18236 declarations should be treated as hidden.
18238 Note that @samp{-fvisibility} does affect C++ vague linkage
18239 entities. This means that, for instance, an exception class that will
18240 be thrown between DSOs must be explicitly marked with default
18241 visibility so that the @samp{type_info} nodes will be unified between
18244 An overview of these techniques, their benefits and how to use them
18245 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18247 @item -fstrict-volatile-bitfields
18248 @opindex fstrict-volatile-bitfields
18249 This option should be used if accesses to volatile bitfields (or other
18250 structure fields, although the compiler usually honors those types
18251 anyway) should use a single access of the width of the
18252 field's type, aligned to a natural alignment if possible. For
18253 example, targets with memory-mapped peripheral registers might require
18254 all such accesses to be 16 bits wide; with this flag the user could
18255 declare all peripheral bitfields as ``unsigned short'' (assuming short
18256 is 16 bits on these targets) to force GCC to use 16 bit accesses
18257 instead of, perhaps, a more efficient 32 bit access.
18259 If this option is disabled, the compiler will use the most efficient
18260 instruction. In the previous example, that might be a 32-bit load
18261 instruction, even though that will access bytes that do not contain
18262 any portion of the bitfield, or memory-mapped registers unrelated to
18263 the one being updated.
18265 If the target requires strict alignment, and honoring the field
18266 type would require violating this alignment, a warning is issued.
18267 If the field has @code{packed} attribute, the access is done without
18268 honoring the field type. If the field doesn't have @code{packed}
18269 attribute, the access is done honoring the field type. In both cases,
18270 GCC assumes that the user knows something about the target hardware
18271 that it is unaware of.
18273 The default value of this option is determined by the application binary
18274 interface for the target processor.
18280 @node Environment Variables
18281 @section Environment Variables Affecting GCC
18282 @cindex environment variables
18284 @c man begin ENVIRONMENT
18285 This section describes several environment variables that affect how GCC
18286 operates. Some of them work by specifying directories or prefixes to use
18287 when searching for various kinds of files. Some are used to specify other
18288 aspects of the compilation environment.
18290 Note that you can also specify places to search using options such as
18291 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18292 take precedence over places specified using environment variables, which
18293 in turn take precedence over those specified by the configuration of GCC@.
18294 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18295 GNU Compiler Collection (GCC) Internals}.
18300 @c @itemx LC_COLLATE
18302 @c @itemx LC_MONETARY
18303 @c @itemx LC_NUMERIC
18308 @c @findex LC_COLLATE
18309 @findex LC_MESSAGES
18310 @c @findex LC_MONETARY
18311 @c @findex LC_NUMERIC
18315 These environment variables control the way that GCC uses
18316 localization information that allow GCC to work with different
18317 national conventions. GCC inspects the locale categories
18318 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18319 so. These locale categories can be set to any value supported by your
18320 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18321 Kingdom encoded in UTF-8.
18323 The @env{LC_CTYPE} environment variable specifies character
18324 classification. GCC uses it to determine the character boundaries in
18325 a string; this is needed for some multibyte encodings that contain quote
18326 and escape characters that would otherwise be interpreted as a string
18329 The @env{LC_MESSAGES} environment variable specifies the language to
18330 use in diagnostic messages.
18332 If the @env{LC_ALL} environment variable is set, it overrides the value
18333 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18334 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18335 environment variable. If none of these variables are set, GCC
18336 defaults to traditional C English behavior.
18340 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18341 files. GCC uses temporary files to hold the output of one stage of
18342 compilation which is to be used as input to the next stage: for example,
18343 the output of the preprocessor, which is the input to the compiler
18346 @item GCC_EXEC_PREFIX
18347 @findex GCC_EXEC_PREFIX
18348 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18349 names of the subprograms executed by the compiler. No slash is added
18350 when this prefix is combined with the name of a subprogram, but you can
18351 specify a prefix that ends with a slash if you wish.
18353 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18354 an appropriate prefix to use based on the pathname it was invoked with.
18356 If GCC cannot find the subprogram using the specified prefix, it
18357 tries looking in the usual places for the subprogram.
18359 The default value of @env{GCC_EXEC_PREFIX} is
18360 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18361 the installed compiler. In many cases @var{prefix} is the value
18362 of @code{prefix} when you ran the @file{configure} script.
18364 Other prefixes specified with @option{-B} take precedence over this prefix.
18366 This prefix is also used for finding files such as @file{crt0.o} that are
18369 In addition, the prefix is used in an unusual way in finding the
18370 directories to search for header files. For each of the standard
18371 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18372 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18373 replacing that beginning with the specified prefix to produce an
18374 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18375 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18376 These alternate directories are searched first; the standard directories
18377 come next. If a standard directory begins with the configured
18378 @var{prefix} then the value of @var{prefix} is replaced by
18379 @env{GCC_EXEC_PREFIX} when looking for header files.
18381 @item COMPILER_PATH
18382 @findex COMPILER_PATH
18383 The value of @env{COMPILER_PATH} is a colon-separated list of
18384 directories, much like @env{PATH}. GCC tries the directories thus
18385 specified when searching for subprograms, if it can't find the
18386 subprograms using @env{GCC_EXEC_PREFIX}.
18389 @findex LIBRARY_PATH
18390 The value of @env{LIBRARY_PATH} is a colon-separated list of
18391 directories, much like @env{PATH}. When configured as a native compiler,
18392 GCC tries the directories thus specified when searching for special
18393 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18394 using GCC also uses these directories when searching for ordinary
18395 libraries for the @option{-l} option (but directories specified with
18396 @option{-L} come first).
18400 @cindex locale definition
18401 This variable is used to pass locale information to the compiler. One way in
18402 which this information is used is to determine the character set to be used
18403 when character literals, string literals and comments are parsed in C and C++.
18404 When the compiler is configured to allow multibyte characters,
18405 the following values for @env{LANG} are recognized:
18409 Recognize JIS characters.
18411 Recognize SJIS characters.
18413 Recognize EUCJP characters.
18416 If @env{LANG} is not defined, or if it has some other value, then the
18417 compiler will use mblen and mbtowc as defined by the default locale to
18418 recognize and translate multibyte characters.
18422 Some additional environments variables affect the behavior of the
18425 @include cppenv.texi
18429 @node Precompiled Headers
18430 @section Using Precompiled Headers
18431 @cindex precompiled headers
18432 @cindex speed of compilation
18434 Often large projects have many header files that are included in every
18435 source file. The time the compiler takes to process these header files
18436 over and over again can account for nearly all of the time required to
18437 build the project. To make builds faster, GCC allows users to
18438 `precompile' a header file; then, if builds can use the precompiled
18439 header file they will be much faster.
18441 To create a precompiled header file, simply compile it as you would any
18442 other file, if necessary using the @option{-x} option to make the driver
18443 treat it as a C or C++ header file. You will probably want to use a
18444 tool like @command{make} to keep the precompiled header up-to-date when
18445 the headers it contains change.
18447 A precompiled header file will be searched for when @code{#include} is
18448 seen in the compilation. As it searches for the included file
18449 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18450 compiler looks for a precompiled header in each directory just before it
18451 looks for the include file in that directory. The name searched for is
18452 the name specified in the @code{#include} with @samp{.gch} appended. If
18453 the precompiled header file can't be used, it is ignored.
18455 For instance, if you have @code{#include "all.h"}, and you have
18456 @file{all.h.gch} in the same directory as @file{all.h}, then the
18457 precompiled header file will be used if possible, and the original
18458 header will be used otherwise.
18460 Alternatively, you might decide to put the precompiled header file in a
18461 directory and use @option{-I} to ensure that directory is searched
18462 before (or instead of) the directory containing the original header.
18463 Then, if you want to check that the precompiled header file is always
18464 used, you can put a file of the same name as the original header in this
18465 directory containing an @code{#error} command.
18467 This also works with @option{-include}. So yet another way to use
18468 precompiled headers, good for projects not designed with precompiled
18469 header files in mind, is to simply take most of the header files used by
18470 a project, include them from another header file, precompile that header
18471 file, and @option{-include} the precompiled header. If the header files
18472 have guards against multiple inclusion, they will be skipped because
18473 they've already been included (in the precompiled header).
18475 If you need to precompile the same header file for different
18476 languages, targets, or compiler options, you can instead make a
18477 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18478 header in the directory, perhaps using @option{-o}. It doesn't matter
18479 what you call the files in the directory, every precompiled header in
18480 the directory will be considered. The first precompiled header
18481 encountered in the directory that is valid for this compilation will
18482 be used; they're searched in no particular order.
18484 There are many other possibilities, limited only by your imagination,
18485 good sense, and the constraints of your build system.
18487 A precompiled header file can be used only when these conditions apply:
18491 Only one precompiled header can be used in a particular compilation.
18494 A precompiled header can't be used once the first C token is seen. You
18495 can have preprocessor directives before a precompiled header; you can
18496 even include a precompiled header from inside another header, so long as
18497 there are no C tokens before the @code{#include}.
18500 The precompiled header file must be produced for the same language as
18501 the current compilation. You can't use a C precompiled header for a C++
18505 The precompiled header file must have been produced by the same compiler
18506 binary as the current compilation is using.
18509 Any macros defined before the precompiled header is included must
18510 either be defined in the same way as when the precompiled header was
18511 generated, or must not affect the precompiled header, which usually
18512 means that they don't appear in the precompiled header at all.
18514 The @option{-D} option is one way to define a macro before a
18515 precompiled header is included; using a @code{#define} can also do it.
18516 There are also some options that define macros implicitly, like
18517 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18520 @item If debugging information is output when using the precompiled
18521 header, using @option{-g} or similar, the same kind of debugging information
18522 must have been output when building the precompiled header. However,
18523 a precompiled header built using @option{-g} can be used in a compilation
18524 when no debugging information is being output.
18526 @item The same @option{-m} options must generally be used when building
18527 and using the precompiled header. @xref{Submodel Options},
18528 for any cases where this rule is relaxed.
18530 @item Each of the following options must be the same when building and using
18531 the precompiled header:
18533 @gccoptlist{-fexceptions}
18536 Some other command-line options starting with @option{-f},
18537 @option{-p}, or @option{-O} must be defined in the same way as when
18538 the precompiled header was generated. At present, it's not clear
18539 which options are safe to change and which are not; the safest choice
18540 is to use exactly the same options when generating and using the
18541 precompiled header. The following are known to be safe:
18543 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18544 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18545 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18550 For all of these except the last, the compiler will automatically
18551 ignore the precompiled header if the conditions aren't met. If you
18552 find an option combination that doesn't work and doesn't cause the
18553 precompiled header to be ignored, please consider filing a bug report,
18556 If you do use differing options when generating and using the
18557 precompiled header, the actual behavior will be a mixture of the
18558 behavior for the options. For instance, if you use @option{-g} to
18559 generate the precompiled header but not when using it, you may or may
18560 not get debugging information for routines in the precompiled header.