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 -fconstexpr-depth=@var{n} -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-abi-version=@var{n} @gol
212 -fobjc-call-cxx-cdtors @gol
213 -fobjc-direct-dispatch @gol
214 -fobjc-exceptions @gol
217 -fobjc-std=objc1 @gol
218 -freplace-objc-classes @gol
221 -Wassign-intercept @gol
222 -Wno-protocol -Wselector @gol
223 -Wstrict-selector-match @gol
224 -Wundeclared-selector}
226 @item Language Independent Options
227 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
228 @gccoptlist{-fmessage-length=@var{n} @gol
229 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
230 -fno-diagnostics-show-option}
232 @item Warning Options
233 @xref{Warning Options,,Options to Request or Suppress Warnings}.
234 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
235 -pedantic-errors @gol
236 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
237 -Wno-attributes -Wno-builtin-macro-redefined @gol
238 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
239 -Wchar-subscripts -Wclobbered -Wcomment @gol
240 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol
241 -Wno-deprecated-declarations -Wdisabled-optimization @gol
242 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
243 -Wno-endif-labels -Werror -Werror=* @gol
244 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
245 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
246 -Wformat-security -Wformat-y2k @gol
247 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
248 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
249 -Winit-self -Winline @gol
250 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
251 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
252 -Wlogical-op -Wlong-long @gol
253 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
254 -Wmissing-format-attribute -Wmissing-include-dirs @gol
256 -Wno-multichar -Wnonnull -Wno-overflow @gol
257 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
258 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
259 -Wpointer-arith -Wno-pointer-to-int-cast @gol
260 -Wredundant-decls @gol
261 -Wreturn-type -Wsequence-point -Wshadow @gol
262 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
263 -Wstrict-aliasing -Wstrict-aliasing=n @gol
264 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
265 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
266 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
267 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
268 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
269 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
270 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value @gol
271 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
272 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
274 @item C and Objective-C-only Warning Options
275 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
276 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
277 -Wold-style-declaration -Wold-style-definition @gol
278 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
279 -Wdeclaration-after-statement -Wpointer-sign}
281 @item Debugging Options
282 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
283 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
284 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
285 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
286 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
287 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
288 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
289 -fdump-statistics @gol
291 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
295 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-nrv -fdump-tree-vect @gol
305 -fdump-tree-sink @gol
306 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
310 -ftree-vectorizer-verbose=@var{n} @gol
311 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
312 -fdump-final-insns=@var{file} @gol
313 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
314 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
315 -feliminate-unused-debug-symbols -femit-class-debug-always
316 -fdebug-types-section @gol
317 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
318 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
319 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
320 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
321 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
322 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
323 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
324 -gvms -gxcoff -gxcoff+ @gol
325 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
326 -fdebug-prefix-map=@var{old}=@var{new} @gol
327 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
328 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
329 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
330 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
331 -print-prog-name=@var{program} -print-search-dirs -Q @gol
332 -print-sysroot -print-sysroot-headers-suffix @gol
333 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
335 @item Optimization Options
336 @xref{Optimize Options,,Options that Control Optimization}.
337 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
338 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
339 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
340 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
341 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
342 -fcompare-elim -fcprop-registers -fcrossjumping @gol
343 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
344 -fcx-limited-range @gol
345 -fdata-sections -fdce -fdce -fdelayed-branch @gol
346 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
347 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
348 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
349 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
350 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
351 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
352 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
353 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
354 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
355 -fira-algorithm=@var{algorithm} @gol
356 -fira-region=@var{region} @gol
357 -fira-loop-pressure -fno-ira-share-save-slots @gol
358 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
359 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
360 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
361 -floop-parallelize-all -flto -flto-compression-level @gol
362 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
363 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
364 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
365 -fno-default-inline @gol
366 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
367 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
368 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
369 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
370 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
371 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
372 -fprefetch-loop-arrays @gol
373 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
374 -fprofile-generate=@var{path} @gol
375 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
376 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
377 -freorder-blocks-and-partition -freorder-functions @gol
378 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
379 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
380 -fsched-spec-load -fsched-spec-load-dangerous @gol
381 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
382 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
383 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
384 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
385 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
386 -fselective-scheduling -fselective-scheduling2 @gol
387 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
388 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
389 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
390 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
392 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
393 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
394 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
395 -ftree-loop-if-convert-stores -ftree-loop-im @gol
396 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
397 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
398 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
399 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
400 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
401 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
402 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
403 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
404 -fwhole-program -fwpa -fuse-linker-plugin @gol
405 --param @var{name}=@var{value}
406 -O -O0 -O1 -O2 -O3 -Os -Ofast}
408 @item Preprocessor Options
409 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
410 @gccoptlist{-A@var{question}=@var{answer} @gol
411 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
412 -C -dD -dI -dM -dN @gol
413 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
414 -idirafter @var{dir} @gol
415 -include @var{file} -imacros @var{file} @gol
416 -iprefix @var{file} -iwithprefix @var{dir} @gol
417 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
418 -imultilib @var{dir} -isysroot @var{dir} @gol
419 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
420 -P -fworking-directory -remap @gol
421 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
422 -Xpreprocessor @var{option}}
424 @item Assembler Option
425 @xref{Assembler Options,,Passing Options to the Assembler}.
426 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
429 @xref{Link Options,,Options for Linking}.
430 @gccoptlist{@var{object-file-name} -l@var{library} @gol
431 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
432 -s -static -static-libgcc -static-libstdc++ -shared @gol
433 -shared-libgcc -symbolic @gol
434 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
437 @item Directory Options
438 @xref{Directory Options,,Options for Directory Search}.
439 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
440 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
443 @item Machine Dependent Options
444 @xref{Submodel Options,,Hardware Models and Configurations}.
445 @c This list is ordered alphanumerically by subsection name.
446 @c Try and put the significant identifier (CPU or system) first,
447 @c so users have a clue at guessing where the ones they want will be.
450 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
451 -mabi=@var{name} @gol
452 -mapcs-stack-check -mno-apcs-stack-check @gol
453 -mapcs-float -mno-apcs-float @gol
454 -mapcs-reentrant -mno-apcs-reentrant @gol
455 -msched-prolog -mno-sched-prolog @gol
456 -mlittle-endian -mbig-endian -mwords-little-endian @gol
457 -mfloat-abi=@var{name} -mfpe @gol
458 -mfp16-format=@var{name}
459 -mthumb-interwork -mno-thumb-interwork @gol
460 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
461 -mstructure-size-boundary=@var{n} @gol
462 -mabort-on-noreturn @gol
463 -mlong-calls -mno-long-calls @gol
464 -msingle-pic-base -mno-single-pic-base @gol
465 -mpic-register=@var{reg} @gol
466 -mnop-fun-dllimport @gol
467 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
468 -mpoke-function-name @gol
470 -mtpcs-frame -mtpcs-leaf-frame @gol
471 -mcaller-super-interworking -mcallee-super-interworking @gol
473 -mword-relocations @gol
474 -mfix-cortex-m3-ldrd}
477 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
478 -mcall-prologues -mtiny-stack -mint8}
480 @emph{Blackfin Options}
481 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
482 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
483 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
484 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
485 -mno-id-shared-library -mshared-library-id=@var{n} @gol
486 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
487 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
488 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
492 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
493 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
494 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
495 -mstack-align -mdata-align -mconst-align @gol
496 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
497 -melf -maout -melinux -mlinux -sim -sim2 @gol
498 -mmul-bug-workaround -mno-mul-bug-workaround}
500 @emph{Darwin Options}
501 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
502 -arch_only -bind_at_load -bundle -bundle_loader @gol
503 -client_name -compatibility_version -current_version @gol
505 -dependency-file -dylib_file -dylinker_install_name @gol
506 -dynamic -dynamiclib -exported_symbols_list @gol
507 -filelist -flat_namespace -force_cpusubtype_ALL @gol
508 -force_flat_namespace -headerpad_max_install_names @gol
510 -image_base -init -install_name -keep_private_externs @gol
511 -multi_module -multiply_defined -multiply_defined_unused @gol
512 -noall_load -no_dead_strip_inits_and_terms @gol
513 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
514 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
515 -private_bundle -read_only_relocs -sectalign @gol
516 -sectobjectsymbols -whyload -seg1addr @gol
517 -sectcreate -sectobjectsymbols -sectorder @gol
518 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
519 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
520 -segprot -segs_read_only_addr -segs_read_write_addr @gol
521 -single_module -static -sub_library -sub_umbrella @gol
522 -twolevel_namespace -umbrella -undefined @gol
523 -unexported_symbols_list -weak_reference_mismatches @gol
524 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
525 -mkernel -mone-byte-bool}
527 @emph{DEC Alpha Options}
528 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
529 -mieee -mieee-with-inexact -mieee-conformant @gol
530 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
531 -mtrap-precision=@var{mode} -mbuild-constants @gol
532 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
533 -mbwx -mmax -mfix -mcix @gol
534 -mfloat-vax -mfloat-ieee @gol
535 -mexplicit-relocs -msmall-data -mlarge-data @gol
536 -msmall-text -mlarge-text @gol
537 -mmemory-latency=@var{time}}
539 @emph{DEC Alpha/VMS Options}
540 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
543 @gccoptlist{-msmall-model -mno-lsim}
546 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
547 -mhard-float -msoft-float @gol
548 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
549 -mdouble -mno-double @gol
550 -mmedia -mno-media -mmuladd -mno-muladd @gol
551 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
552 -mlinked-fp -mlong-calls -malign-labels @gol
553 -mlibrary-pic -macc-4 -macc-8 @gol
554 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
555 -moptimize-membar -mno-optimize-membar @gol
556 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
557 -mvliw-branch -mno-vliw-branch @gol
558 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
559 -mno-nested-cond-exec -mtomcat-stats @gol
563 @emph{GNU/Linux Options}
564 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
565 -tno-android-cc -tno-android-ld}
567 @emph{H8/300 Options}
568 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
571 @gccoptlist{-march=@var{architecture-type} @gol
572 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
573 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
574 -mfixed-range=@var{register-range} @gol
575 -mjump-in-delay -mlinker-opt -mlong-calls @gol
576 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
577 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
578 -mno-jump-in-delay -mno-long-load-store @gol
579 -mno-portable-runtime -mno-soft-float @gol
580 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
581 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
582 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
583 -munix=@var{unix-std} -nolibdld -static -threads}
585 @emph{i386 and x86-64 Options}
586 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
587 -mfpmath=@var{unit} @gol
588 -masm=@var{dialect} -mno-fancy-math-387 @gol
589 -mno-fp-ret-in-387 -msoft-float @gol
590 -mno-wide-multiply -mrtd -malign-double @gol
591 -mpreferred-stack-boundary=@var{num} @gol
592 -mincoming-stack-boundary=@var{num} @gol
593 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
594 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
595 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
596 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlwp @gol
597 -mthreads -mno-align-stringops -minline-all-stringops @gol
598 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
599 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
600 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
601 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
602 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
603 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
604 -mcmodel=@var{code-model} -mabi=@var{name} @gol
605 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
606 -msse2avx -mfentry -m8bit-idiv @gol
607 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
609 @emph{i386 and x86-64 Windows Options}
610 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
611 -mnop-fun-dllimport -mthread @gol
612 -municode -mwin32 -mwindows -fno-set-stack-executable}
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
671 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
672 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
673 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
674 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
675 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
678 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
679 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
680 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
681 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
684 @emph{MicroBlaze Options}
685 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
686 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
687 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
688 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
689 -mxl-mode-@var{app-model}}
692 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
693 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
694 -mips64 -mips64r2 @gol
695 -mips16 -mno-mips16 -mflip-mips16 @gol
696 -minterlink-mips16 -mno-interlink-mips16 @gol
697 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
698 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
699 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
700 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
701 -mfpu=@var{fpu-type} @gol
702 -msmartmips -mno-smartmips @gol
703 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
704 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
705 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
706 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
707 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
708 -membedded-data -mno-embedded-data @gol
709 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
710 -mcode-readable=@var{setting} @gol
711 -msplit-addresses -mno-split-addresses @gol
712 -mexplicit-relocs -mno-explicit-relocs @gol
713 -mcheck-zero-division -mno-check-zero-division @gol
714 -mdivide-traps -mdivide-breaks @gol
715 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
716 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
717 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
718 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
719 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
720 -mflush-func=@var{func} -mno-flush-func @gol
721 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
722 -mfp-exceptions -mno-fp-exceptions @gol
723 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
724 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
727 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
728 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
729 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
730 -mno-base-addresses -msingle-exit -mno-single-exit}
732 @emph{MN10300 Options}
733 @gccoptlist{-mmult-bug -mno-mult-bug @gol
734 -mno-am33 -mam33 -mam33-2 -mam34 @gol
735 -mtune=@var{cpu-type} @gol
736 -mreturn-pointer-on-d0 @gol
737 -mno-crt0 -mrelax -mliw -msetlb}
739 @emph{PDP-11 Options}
740 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
741 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
742 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
743 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
744 -mbranch-expensive -mbranch-cheap @gol
745 -munix-asm -mdec-asm}
747 @emph{picoChip Options}
748 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
749 -msymbol-as-address -mno-inefficient-warnings}
751 @emph{PowerPC Options}
752 See RS/6000 and PowerPC Options.
754 @emph{RS/6000 and PowerPC Options}
755 @gccoptlist{-mcpu=@var{cpu-type} @gol
756 -mtune=@var{cpu-type} @gol
757 -mcmodel=@var{code-model} @gol
758 -mpower -mno-power -mpower2 -mno-power2 @gol
759 -mpowerpc -mpowerpc64 -mno-powerpc @gol
760 -maltivec -mno-altivec @gol
761 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
762 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
763 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
764 -mfprnd -mno-fprnd @gol
765 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
766 -mnew-mnemonics -mold-mnemonics @gol
767 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
768 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
769 -malign-power -malign-natural @gol
770 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
771 -msingle-float -mdouble-float -msimple-fpu @gol
772 -mstring -mno-string -mupdate -mno-update @gol
773 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
774 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
775 -mstrict-align -mno-strict-align -mrelocatable @gol
776 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
777 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
778 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
779 -mprioritize-restricted-insns=@var{priority} @gol
780 -msched-costly-dep=@var{dependence_type} @gol
781 -minsert-sched-nops=@var{scheme} @gol
782 -mcall-sysv -mcall-netbsd @gol
783 -maix-struct-return -msvr4-struct-return @gol
784 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
785 -mblock-move-inline-limit=@var{num} @gol
786 -misel -mno-isel @gol
787 -misel=yes -misel=no @gol
789 -mspe=yes -mspe=no @gol
791 -mgen-cell-microcode -mwarn-cell-microcode @gol
792 -mvrsave -mno-vrsave @gol
793 -mmulhw -mno-mulhw @gol
794 -mdlmzb -mno-dlmzb @gol
795 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
796 -mprototype -mno-prototype @gol
797 -msim -mmvme -mads -myellowknife -memb -msdata @gol
798 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
799 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
800 -mno-recip-precision @gol
801 -mveclibabi=@var{type} -mfriz -mno-friz}
804 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
806 -mbig-endian-data -mlittle-endian-data @gol
809 -mas100-syntax -mno-as100-syntax@gol
811 -mmax-constant-size=@gol
813 -msave-acc-in-interrupts}
815 @emph{S/390 and zSeries Options}
816 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
817 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
818 -mlong-double-64 -mlong-double-128 @gol
819 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
820 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
821 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
822 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
823 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
826 @gccoptlist{-meb -mel @gol
830 -mscore5 -mscore5u -mscore7 -mscore7d}
833 @gccoptlist{-m1 -m2 -m2e @gol
834 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
836 -m4-nofpu -m4-single-only -m4-single -m4 @gol
837 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
838 -m5-64media -m5-64media-nofpu @gol
839 -m5-32media -m5-32media-nofpu @gol
840 -m5-compact -m5-compact-nofpu @gol
841 -mb -ml -mdalign -mrelax @gol
842 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
843 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
844 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
845 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
846 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
847 -maccumulate-outgoing-args -minvalid-symbols}
849 @emph{Solaris 2 Options}
850 @gccoptlist{-mimpure-text -mno-impure-text @gol
854 @gccoptlist{-mcpu=@var{cpu-type} @gol
855 -mtune=@var{cpu-type} @gol
856 -mcmodel=@var{code-model} @gol
857 -m32 -m64 -mapp-regs -mno-app-regs @gol
858 -mfaster-structs -mno-faster-structs @gol
859 -mfpu -mno-fpu -mhard-float -msoft-float @gol
860 -mhard-quad-float -msoft-quad-float @gol
862 -mstack-bias -mno-stack-bias @gol
863 -munaligned-doubles -mno-unaligned-doubles @gol
864 -mv8plus -mno-v8plus -mvis -mno-vis}
867 @gccoptlist{-mwarn-reloc -merror-reloc @gol
868 -msafe-dma -munsafe-dma @gol
870 -msmall-mem -mlarge-mem -mstdmain @gol
871 -mfixed-range=@var{register-range} @gol
873 -maddress-space-conversion -mno-address-space-conversion @gol
874 -mcache-size=@var{cache-size} @gol
875 -matomic-updates -mno-atomic-updates}
877 @emph{System V Options}
878 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
881 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
882 -mprolog-function -mno-prolog-function -mspace @gol
883 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
884 -mapp-regs -mno-app-regs @gol
885 -mdisable-callt -mno-disable-callt @gol
888 -mv850e1 -mv850es @gol
893 @gccoptlist{-mg -mgnu -munix}
895 @emph{VxWorks Options}
896 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
897 -Xbind-lazy -Xbind-now}
899 @emph{x86-64 Options}
900 See i386 and x86-64 Options.
902 @emph{Xstormy16 Options}
905 @emph{Xtensa Options}
906 @gccoptlist{-mconst16 -mno-const16 @gol
907 -mfused-madd -mno-fused-madd @gol
909 -mserialize-volatile -mno-serialize-volatile @gol
910 -mtext-section-literals -mno-text-section-literals @gol
911 -mtarget-align -mno-target-align @gol
912 -mlongcalls -mno-longcalls}
914 @emph{zSeries Options}
915 See S/390 and zSeries Options.
917 @item Code Generation Options
918 @xref{Code Gen Options,,Options for Code Generation Conventions}.
919 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
920 -ffixed-@var{reg} -fexceptions @gol
921 -fnon-call-exceptions -funwind-tables @gol
922 -fasynchronous-unwind-tables @gol
923 -finhibit-size-directive -finstrument-functions @gol
924 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
925 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
926 -fno-common -fno-ident @gol
927 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
928 -fno-jump-tables @gol
929 -frecord-gcc-switches @gol
930 -freg-struct-return -fshort-enums @gol
931 -fshort-double -fshort-wchar @gol
932 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
933 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
934 -fno-stack-limit -fsplit-stack @gol
935 -fleading-underscore -ftls-model=@var{model} @gol
936 -ftrapv -fwrapv -fbounds-check @gol
937 -fvisibility -fstrict-volatile-bitfields}
941 * Overall Options:: Controlling the kind of output:
942 an executable, object files, assembler files,
943 or preprocessed source.
944 * C Dialect Options:: Controlling the variant of C language compiled.
945 * C++ Dialect Options:: Variations on C++.
946 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
948 * Language Independent Options:: Controlling how diagnostics should be
950 * Warning Options:: How picky should the compiler be?
951 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
952 * Optimize Options:: How much optimization?
953 * Preprocessor Options:: Controlling header files and macro definitions.
954 Also, getting dependency information for Make.
955 * Assembler Options:: Passing options to the assembler.
956 * Link Options:: Specifying libraries and so on.
957 * Directory Options:: Where to find header files and libraries.
958 Where to find the compiler executable files.
959 * Spec Files:: How to pass switches to sub-processes.
960 * Target Options:: Running a cross-compiler, or an old version of GCC.
963 @node Overall Options
964 @section Options Controlling the Kind of Output
966 Compilation can involve up to four stages: preprocessing, compilation
967 proper, assembly and linking, always in that order. GCC is capable of
968 preprocessing and compiling several files either into several
969 assembler input files, or into one assembler input file; then each
970 assembler input file produces an object file, and linking combines all
971 the object files (those newly compiled, and those specified as input)
972 into an executable file.
974 @cindex file name suffix
975 For any given input file, the file name suffix determines what kind of
980 C source code which must be preprocessed.
983 C source code which should not be preprocessed.
986 C++ source code which should not be preprocessed.
989 Objective-C source code. Note that you must link with the @file{libobjc}
990 library to make an Objective-C program work.
993 Objective-C source code which should not be preprocessed.
997 Objective-C++ source code. Note that you must link with the @file{libobjc}
998 library to make an Objective-C++ program work. Note that @samp{.M} refers
999 to a literal capital M@.
1001 @item @var{file}.mii
1002 Objective-C++ source code which should not be preprocessed.
1005 C, C++, Objective-C or Objective-C++ header file to be turned into a
1006 precompiled header (default), or C, C++ header file to be turned into an
1007 Ada spec (via the @option{-fdump-ada-spec} switch).
1010 @itemx @var{file}.cp
1011 @itemx @var{file}.cxx
1012 @itemx @var{file}.cpp
1013 @itemx @var{file}.CPP
1014 @itemx @var{file}.c++
1016 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1017 the last two letters must both be literally @samp{x}. Likewise,
1018 @samp{.C} refers to a literal capital C@.
1022 Objective-C++ source code which must be preprocessed.
1024 @item @var{file}.mii
1025 Objective-C++ source code which should not be preprocessed.
1029 @itemx @var{file}.hp
1030 @itemx @var{file}.hxx
1031 @itemx @var{file}.hpp
1032 @itemx @var{file}.HPP
1033 @itemx @var{file}.h++
1034 @itemx @var{file}.tcc
1035 C++ header file to be turned into a precompiled header or Ada spec.
1038 @itemx @var{file}.for
1039 @itemx @var{file}.ftn
1040 Fixed form Fortran source code which should not be preprocessed.
1043 @itemx @var{file}.FOR
1044 @itemx @var{file}.fpp
1045 @itemx @var{file}.FPP
1046 @itemx @var{file}.FTN
1047 Fixed form Fortran source code which must be preprocessed (with the traditional
1050 @item @var{file}.f90
1051 @itemx @var{file}.f95
1052 @itemx @var{file}.f03
1053 @itemx @var{file}.f08
1054 Free form Fortran source code which should not be preprocessed.
1056 @item @var{file}.F90
1057 @itemx @var{file}.F95
1058 @itemx @var{file}.F03
1059 @itemx @var{file}.F08
1060 Free form Fortran source code which must be preprocessed (with the
1061 traditional preprocessor).
1066 @c FIXME: Descriptions of Java file types.
1072 @item @var{file}.ads
1073 Ada source code file which contains a library unit declaration (a
1074 declaration of a package, subprogram, or generic, or a generic
1075 instantiation), or a library unit renaming declaration (a package,
1076 generic, or subprogram renaming declaration). Such files are also
1079 @item @var{file}.adb
1080 Ada source code file containing a library unit body (a subprogram or
1081 package body). Such files are also called @dfn{bodies}.
1083 @c GCC also knows about some suffixes for languages not yet included:
1094 @itemx @var{file}.sx
1095 Assembler code which must be preprocessed.
1098 An object file to be fed straight into linking.
1099 Any file name with no recognized suffix is treated this way.
1103 You can specify the input language explicitly with the @option{-x} option:
1106 @item -x @var{language}
1107 Specify explicitly the @var{language} for the following input files
1108 (rather than letting the compiler choose a default based on the file
1109 name suffix). This option applies to all following input files until
1110 the next @option{-x} option. Possible values for @var{language} are:
1112 c c-header cpp-output
1113 c++ c++-header c++-cpp-output
1114 objective-c objective-c-header objective-c-cpp-output
1115 objective-c++ objective-c++-header objective-c++-cpp-output
1116 assembler assembler-with-cpp
1118 f77 f77-cpp-input f95 f95-cpp-input
1124 Turn off any specification of a language, so that subsequent files are
1125 handled according to their file name suffixes (as they are if @option{-x}
1126 has not been used at all).
1128 @item -pass-exit-codes
1129 @opindex pass-exit-codes
1130 Normally the @command{gcc} program will exit with the code of 1 if any
1131 phase of the compiler returns a non-success return code. If you specify
1132 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1133 numerically highest error produced by any phase that returned an error
1134 indication. The C, C++, and Fortran frontends return 4, if an internal
1135 compiler error is encountered.
1138 If you only want some of the stages of compilation, you can use
1139 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1140 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1141 @command{gcc} is to stop. Note that some combinations (for example,
1142 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1147 Compile or assemble the source files, but do not link. The linking
1148 stage simply is not done. The ultimate output is in the form of an
1149 object file for each source file.
1151 By default, the object file name for a source file is made by replacing
1152 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1154 Unrecognized input files, not requiring compilation or assembly, are
1159 Stop after the stage of compilation proper; do not assemble. The output
1160 is in the form of an assembler code file for each non-assembler input
1163 By default, the assembler file name for a source file is made by
1164 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1166 Input files that don't require compilation are ignored.
1170 Stop after the preprocessing stage; do not run the compiler proper. The
1171 output is in the form of preprocessed source code, which is sent to the
1174 Input files which don't require preprocessing are ignored.
1176 @cindex output file option
1179 Place output in file @var{file}. This applies regardless to whatever
1180 sort of output is being produced, whether it be an executable file,
1181 an object file, an assembler file or preprocessed C code.
1183 If @option{-o} is not specified, the default is to put an executable
1184 file in @file{a.out}, the object file for
1185 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1186 assembler file in @file{@var{source}.s}, a precompiled header file in
1187 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1192 Print (on standard error output) the commands executed to run the stages
1193 of compilation. Also print the version number of the compiler driver
1194 program and of the preprocessor and the compiler proper.
1198 Like @option{-v} except the commands are not executed and arguments
1199 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1200 This is useful for shell scripts to capture the driver-generated command lines.
1204 Use pipes rather than temporary files for communication between the
1205 various stages of compilation. This fails to work on some systems where
1206 the assembler is unable to read from a pipe; but the GNU assembler has
1211 Print (on the standard output) a description of the command line options
1212 understood by @command{gcc}. If the @option{-v} option is also specified
1213 then @option{--help} will also be passed on to the various processes
1214 invoked by @command{gcc}, so that they can display the command line options
1215 they accept. If the @option{-Wextra} option has also been specified
1216 (prior to the @option{--help} option), then command line options which
1217 have no documentation associated with them will also be displayed.
1220 @opindex target-help
1221 Print (on the standard output) a description of target-specific command
1222 line options for each tool. For some targets extra target-specific
1223 information may also be printed.
1225 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1226 Print (on the standard output) a description of the command line
1227 options understood by the compiler that fit into all specified classes
1228 and qualifiers. These are the supported classes:
1231 @item @samp{optimizers}
1232 This will display all of the optimization options supported by the
1235 @item @samp{warnings}
1236 This will display all of the options controlling warning messages
1237 produced by the compiler.
1240 This will display target-specific options. Unlike the
1241 @option{--target-help} option however, target-specific options of the
1242 linker and assembler will not be displayed. This is because those
1243 tools do not currently support the extended @option{--help=} syntax.
1246 This will display the values recognized by the @option{--param}
1249 @item @var{language}
1250 This will display the options supported for @var{language}, where
1251 @var{language} is the name of one of the languages supported in this
1255 This will display the options that are common to all languages.
1258 These are the supported qualifiers:
1261 @item @samp{undocumented}
1262 Display only those options which are undocumented.
1265 Display options which take an argument that appears after an equal
1266 sign in the same continuous piece of text, such as:
1267 @samp{--help=target}.
1269 @item @samp{separate}
1270 Display options which take an argument that appears as a separate word
1271 following the original option, such as: @samp{-o output-file}.
1274 Thus for example to display all the undocumented target-specific
1275 switches supported by the compiler the following can be used:
1278 --help=target,undocumented
1281 The sense of a qualifier can be inverted by prefixing it with the
1282 @samp{^} character, so for example to display all binary warning
1283 options (i.e., ones that are either on or off and that do not take an
1284 argument), which have a description the following can be used:
1287 --help=warnings,^joined,^undocumented
1290 The argument to @option{--help=} should not consist solely of inverted
1293 Combining several classes is possible, although this usually
1294 restricts the output by so much that there is nothing to display. One
1295 case where it does work however is when one of the classes is
1296 @var{target}. So for example to display all the target-specific
1297 optimization options the following can be used:
1300 --help=target,optimizers
1303 The @option{--help=} option can be repeated on the command line. Each
1304 successive use will display its requested class of options, skipping
1305 those that have already been displayed.
1307 If the @option{-Q} option appears on the command line before the
1308 @option{--help=} option, then the descriptive text displayed by
1309 @option{--help=} is changed. Instead of describing the displayed
1310 options, an indication is given as to whether the option is enabled,
1311 disabled or set to a specific value (assuming that the compiler
1312 knows this at the point where the @option{--help=} option is used).
1314 Here is a truncated example from the ARM port of @command{gcc}:
1317 % gcc -Q -mabi=2 --help=target -c
1318 The following options are target specific:
1320 -mabort-on-noreturn [disabled]
1324 The output is sensitive to the effects of previous command line
1325 options, so for example it is possible to find out which optimizations
1326 are enabled at @option{-O2} by using:
1329 -Q -O2 --help=optimizers
1332 Alternatively you can discover which binary optimizations are enabled
1333 by @option{-O3} by using:
1336 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1337 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1338 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1341 @item -no-canonical-prefixes
1342 @opindex no-canonical-prefixes
1343 Do not expand any symbolic links, resolve references to @samp{/../}
1344 or @samp{/./}, or make the path absolute when generating a relative
1349 Display the version number and copyrights of the invoked GCC@.
1353 Invoke all subcommands under a wrapper program. The name of the
1354 wrapper program and its parameters are passed as a comma separated
1358 gcc -c t.c -wrapper gdb,--args
1361 This will invoke all subprograms of @command{gcc} under
1362 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1363 @samp{gdb --args cc1 @dots{}}.
1365 @item -fplugin=@var{name}.so
1366 Load the plugin code in file @var{name}.so, assumed to be a
1367 shared object to be dlopen'd by the compiler. The base name of
1368 the shared object file is used to identify the plugin for the
1369 purposes of argument parsing (See
1370 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1371 Each plugin should define the callback functions specified in the
1374 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1375 Define an argument called @var{key} with a value of @var{value}
1376 for the plugin called @var{name}.
1378 @item -fdump-ada-spec@r{[}-slim@r{]}
1379 For C and C++ source and include files, generate corresponding Ada
1380 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1381 GNAT User's Guide}, which provides detailed documentation on this feature.
1383 @item -fdump-go-spec=@var{file}
1384 For input files in any language, generate corresponding Go
1385 declarations in @var{file}. This generates Go @code{const},
1386 @code{type}, @code{var}, and @code{func} declarations which may be a
1387 useful way to start writing a Go interface to code written in some
1390 @include @value{srcdir}/../libiberty/at-file.texi
1394 @section Compiling C++ Programs
1396 @cindex suffixes for C++ source
1397 @cindex C++ source file suffixes
1398 C++ source files conventionally use one of the suffixes @samp{.C},
1399 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1400 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1401 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1402 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1403 files with these names and compiles them as C++ programs even if you
1404 call the compiler the same way as for compiling C programs (usually
1405 with the name @command{gcc}).
1409 However, the use of @command{gcc} does not add the C++ library.
1410 @command{g++} is a program that calls GCC and treats @samp{.c},
1411 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1412 files unless @option{-x} is used, and automatically specifies linking
1413 against the C++ library. This program is also useful when
1414 precompiling a C header file with a @samp{.h} extension for use in C++
1415 compilations. On many systems, @command{g++} is also installed with
1416 the name @command{c++}.
1418 @cindex invoking @command{g++}
1419 When you compile C++ programs, you may specify many of the same
1420 command-line options that you use for compiling programs in any
1421 language; or command-line options meaningful for C and related
1422 languages; or options that are meaningful only for C++ programs.
1423 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1424 explanations of options for languages related to C@.
1425 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1426 explanations of options that are meaningful only for C++ programs.
1428 @node C Dialect Options
1429 @section Options Controlling C Dialect
1430 @cindex dialect options
1431 @cindex language dialect options
1432 @cindex options, dialect
1434 The following options control the dialect of C (or languages derived
1435 from C, such as C++, Objective-C and Objective-C++) that the compiler
1439 @cindex ANSI support
1443 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1444 equivalent to @samp{-std=c++98}.
1446 This turns off certain features of GCC that are incompatible with ISO
1447 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1448 such as the @code{asm} and @code{typeof} keywords, and
1449 predefined macros such as @code{unix} and @code{vax} that identify the
1450 type of system you are using. It also enables the undesirable and
1451 rarely used ISO trigraph feature. For the C compiler,
1452 it disables recognition of C++ style @samp{//} comments as well as
1453 the @code{inline} keyword.
1455 The alternate keywords @code{__asm__}, @code{__extension__},
1456 @code{__inline__} and @code{__typeof__} continue to work despite
1457 @option{-ansi}. You would not want to use them in an ISO C program, of
1458 course, but it is useful to put them in header files that might be included
1459 in compilations done with @option{-ansi}. Alternate predefined macros
1460 such as @code{__unix__} and @code{__vax__} are also available, with or
1461 without @option{-ansi}.
1463 The @option{-ansi} option does not cause non-ISO programs to be
1464 rejected gratuitously. For that, @option{-pedantic} is required in
1465 addition to @option{-ansi}. @xref{Warning Options}.
1467 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1468 option is used. Some header files may notice this macro and refrain
1469 from declaring certain functions or defining certain macros that the
1470 ISO standard doesn't call for; this is to avoid interfering with any
1471 programs that might use these names for other things.
1473 Functions that would normally be built in but do not have semantics
1474 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1475 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1476 built-in functions provided by GCC}, for details of the functions
1481 Determine the language standard. @xref{Standards,,Language Standards
1482 Supported by GCC}, for details of these standard versions. This option
1483 is currently only supported when compiling C or C++.
1485 The compiler can accept several base standards, such as @samp{c90} or
1486 @samp{c++98}, and GNU dialects of those standards, such as
1487 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1488 compiler will accept all programs following that standard and those
1489 using GNU extensions that do not contradict it. For example,
1490 @samp{-std=c90} turns off certain features of GCC that are
1491 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1492 keywords, but not other GNU extensions that do not have a meaning in
1493 ISO C90, such as omitting the middle term of a @code{?:}
1494 expression. On the other hand, by specifying a GNU dialect of a
1495 standard, all features the compiler support are enabled, even when
1496 those features change the meaning of the base standard and some
1497 strict-conforming programs may be rejected. The particular standard
1498 is used by @option{-pedantic} to identify which features are GNU
1499 extensions given that version of the standard. For example
1500 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1501 comments, while @samp{-std=gnu99 -pedantic} would not.
1503 A value for this option must be provided; possible values are
1509 Support all ISO C90 programs (certain GNU extensions that conflict
1510 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1512 @item iso9899:199409
1513 ISO C90 as modified in amendment 1.
1519 ISO C99. Note that this standard is not yet fully supported; see
1520 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1521 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1524 ISO C1X, the draft of the next revision of the ISO C standard.
1525 Support is limited and experimental and features enabled by this
1526 option may be changed or removed if changed in or removed from the
1531 GNU dialect of ISO C90 (including some C99 features). This
1532 is the default for C code.
1536 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1537 this will become the default. The name @samp{gnu9x} is deprecated.
1540 GNU dialect of ISO C1X. Support is limited and experimental and
1541 features enabled by this option may be changed or removed if changed
1542 in or removed from the standard draft.
1545 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1549 GNU dialect of @option{-std=c++98}. This is the default for
1553 The working draft of the upcoming ISO C++0x standard. This option
1554 enables experimental features that are likely to be included in
1555 C++0x. The working draft is constantly changing, and any feature that is
1556 enabled by this flag may be removed from future versions of GCC if it is
1557 not part of the C++0x standard.
1560 GNU dialect of @option{-std=c++0x}. This option enables
1561 experimental features that may be removed in future versions of GCC.
1564 @item -fgnu89-inline
1565 @opindex fgnu89-inline
1566 The option @option{-fgnu89-inline} tells GCC to use the traditional
1567 GNU semantics for @code{inline} functions when in C99 mode.
1568 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1569 is accepted and ignored by GCC versions 4.1.3 up to but not including
1570 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1571 C99 mode. Using this option is roughly equivalent to adding the
1572 @code{gnu_inline} function attribute to all inline functions
1573 (@pxref{Function Attributes}).
1575 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1576 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1577 specifies the default behavior). This option was first supported in
1578 GCC 4.3. This option is not supported in @option{-std=c90} or
1579 @option{-std=gnu90} mode.
1581 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1582 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1583 in effect for @code{inline} functions. @xref{Common Predefined
1584 Macros,,,cpp,The C Preprocessor}.
1586 @item -aux-info @var{filename}
1588 Output to the given filename prototyped declarations for all functions
1589 declared and/or defined in a translation unit, including those in header
1590 files. This option is silently ignored in any language other than C@.
1592 Besides declarations, the file indicates, in comments, the origin of
1593 each declaration (source file and line), whether the declaration was
1594 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1595 @samp{O} for old, respectively, in the first character after the line
1596 number and the colon), and whether it came from a declaration or a
1597 definition (@samp{C} or @samp{F}, respectively, in the following
1598 character). In the case of function definitions, a K&R-style list of
1599 arguments followed by their declarations is also provided, inside
1600 comments, after the declaration.
1604 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1605 keyword, so that code can use these words as identifiers. You can use
1606 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1607 instead. @option{-ansi} implies @option{-fno-asm}.
1609 In C++, this switch only affects the @code{typeof} keyword, since
1610 @code{asm} and @code{inline} are standard keywords. You may want to
1611 use the @option{-fno-gnu-keywords} flag instead, which has the same
1612 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1613 switch only affects the @code{asm} and @code{typeof} keywords, since
1614 @code{inline} is a standard keyword in ISO C99.
1617 @itemx -fno-builtin-@var{function}
1618 @opindex fno-builtin
1619 @cindex built-in functions
1620 Don't recognize built-in functions that do not begin with
1621 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1622 functions provided by GCC}, for details of the functions affected,
1623 including those which are not built-in functions when @option{-ansi} or
1624 @option{-std} options for strict ISO C conformance are used because they
1625 do not have an ISO standard meaning.
1627 GCC normally generates special code to handle certain built-in functions
1628 more efficiently; for instance, calls to @code{alloca} may become single
1629 instructions that adjust the stack directly, and calls to @code{memcpy}
1630 may become inline copy loops. The resulting code is often both smaller
1631 and faster, but since the function calls no longer appear as such, you
1632 cannot set a breakpoint on those calls, nor can you change the behavior
1633 of the functions by linking with a different library. In addition,
1634 when a function is recognized as a built-in function, GCC may use
1635 information about that function to warn about problems with calls to
1636 that function, or to generate more efficient code, even if the
1637 resulting code still contains calls to that function. For example,
1638 warnings are given with @option{-Wformat} for bad calls to
1639 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1640 known not to modify global memory.
1642 With the @option{-fno-builtin-@var{function}} option
1643 only the built-in function @var{function} is
1644 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1645 function is named that is not built-in in this version of GCC, this
1646 option is ignored. There is no corresponding
1647 @option{-fbuiltin-@var{function}} option; if you wish to enable
1648 built-in functions selectively when using @option{-fno-builtin} or
1649 @option{-ffreestanding}, you may define macros such as:
1652 #define abs(n) __builtin_abs ((n))
1653 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1658 @cindex hosted environment
1660 Assert that compilation takes place in a hosted environment. This implies
1661 @option{-fbuiltin}. A hosted environment is one in which the
1662 entire standard library is available, and in which @code{main} has a return
1663 type of @code{int}. Examples are nearly everything except a kernel.
1664 This is equivalent to @option{-fno-freestanding}.
1666 @item -ffreestanding
1667 @opindex ffreestanding
1668 @cindex hosted environment
1670 Assert that compilation takes place in a freestanding environment. This
1671 implies @option{-fno-builtin}. A freestanding environment
1672 is one in which the standard library may not exist, and program startup may
1673 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1674 This is equivalent to @option{-fno-hosted}.
1676 @xref{Standards,,Language Standards Supported by GCC}, for details of
1677 freestanding and hosted environments.
1681 @cindex OpenMP parallel
1682 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1683 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1684 compiler generates parallel code according to the OpenMP Application
1685 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1686 implies @option{-pthread}, and thus is only supported on targets that
1687 have support for @option{-pthread}.
1689 @item -fms-extensions
1690 @opindex fms-extensions
1691 Accept some non-standard constructs used in Microsoft header files.
1693 In C++ code, this allows member names in structures to be similar
1694 to previous types declarations.
1703 Some cases of unnamed fields in structures and unions are only
1704 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1705 fields within structs/unions}, for details.
1707 @item -fplan9-extensions
1708 Accept some non-standard constructs used in Plan 9 code.
1710 This enables @option{-fms-extensions}, permits passing pointers to
1711 structures with anonymous fields to functions which expect pointers to
1712 elements of the type of the field, and permits referring to anonymous
1713 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1714 struct/union fields within structs/unions}, for details. This is only
1715 supported for C, not C++.
1719 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1720 options for strict ISO C conformance) implies @option{-trigraphs}.
1722 @item -no-integrated-cpp
1723 @opindex no-integrated-cpp
1724 Performs a compilation in two passes: preprocessing and compiling. This
1725 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1726 @option{-B} option. The user supplied compilation step can then add in
1727 an additional preprocessing step after normal preprocessing but before
1728 compiling. The default is to use the integrated cpp (internal cpp)
1730 The semantics of this option will change if "cc1", "cc1plus", and
1731 "cc1obj" are merged.
1733 @cindex traditional C language
1734 @cindex C language, traditional
1736 @itemx -traditional-cpp
1737 @opindex traditional-cpp
1738 @opindex traditional
1739 Formerly, these options caused GCC to attempt to emulate a pre-standard
1740 C compiler. They are now only supported with the @option{-E} switch.
1741 The preprocessor continues to support a pre-standard mode. See the GNU
1742 CPP manual for details.
1744 @item -fcond-mismatch
1745 @opindex fcond-mismatch
1746 Allow conditional expressions with mismatched types in the second and
1747 third arguments. The value of such an expression is void. This option
1748 is not supported for C++.
1750 @item -flax-vector-conversions
1751 @opindex flax-vector-conversions
1752 Allow implicit conversions between vectors with differing numbers of
1753 elements and/or incompatible element types. This option should not be
1756 @item -funsigned-char
1757 @opindex funsigned-char
1758 Let the type @code{char} be unsigned, like @code{unsigned char}.
1760 Each kind of machine has a default for what @code{char} should
1761 be. It is either like @code{unsigned char} by default or like
1762 @code{signed char} by default.
1764 Ideally, a portable program should always use @code{signed char} or
1765 @code{unsigned char} when it depends on the signedness of an object.
1766 But many programs have been written to use plain @code{char} and
1767 expect it to be signed, or expect it to be unsigned, depending on the
1768 machines they were written for. This option, and its inverse, let you
1769 make such a program work with the opposite default.
1771 The type @code{char} is always a distinct type from each of
1772 @code{signed char} or @code{unsigned char}, even though its behavior
1773 is always just like one of those two.
1776 @opindex fsigned-char
1777 Let the type @code{char} be signed, like @code{signed char}.
1779 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1780 the negative form of @option{-funsigned-char}. Likewise, the option
1781 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1783 @item -fsigned-bitfields
1784 @itemx -funsigned-bitfields
1785 @itemx -fno-signed-bitfields
1786 @itemx -fno-unsigned-bitfields
1787 @opindex fsigned-bitfields
1788 @opindex funsigned-bitfields
1789 @opindex fno-signed-bitfields
1790 @opindex fno-unsigned-bitfields
1791 These options control whether a bit-field is signed or unsigned, when the
1792 declaration does not use either @code{signed} or @code{unsigned}. By
1793 default, such a bit-field is signed, because this is consistent: the
1794 basic integer types such as @code{int} are signed types.
1797 @node C++ Dialect Options
1798 @section Options Controlling C++ Dialect
1800 @cindex compiler options, C++
1801 @cindex C++ options, command line
1802 @cindex options, C++
1803 This section describes the command-line options that are only meaningful
1804 for C++ programs; but you can also use most of the GNU compiler options
1805 regardless of what language your program is in. For example, you
1806 might compile a file @code{firstClass.C} like this:
1809 g++ -g -frepo -O -c firstClass.C
1813 In this example, only @option{-frepo} is an option meant
1814 only for C++ programs; you can use the other options with any
1815 language supported by GCC@.
1817 Here is a list of options that are @emph{only} for compiling C++ programs:
1821 @item -fabi-version=@var{n}
1822 @opindex fabi-version
1823 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1824 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1825 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1826 the version that conforms most closely to the C++ ABI specification.
1827 Therefore, the ABI obtained using version 0 will change as ABI bugs
1830 The default is version 2.
1832 Version 3 corrects an error in mangling a constant address as a
1835 Version 4 implements a standard mangling for vector types.
1837 Version 5 corrects the mangling of attribute const/volatile on
1838 function pointer types, decltype of a plain decl, and use of a
1839 function parameter in the declaration of another parameter.
1841 See also @option{-Wabi}.
1843 @item -fno-access-control
1844 @opindex fno-access-control
1845 Turn off all access checking. This switch is mainly useful for working
1846 around bugs in the access control code.
1850 Check that the pointer returned by @code{operator new} is non-null
1851 before attempting to modify the storage allocated. This check is
1852 normally unnecessary because the C++ standard specifies that
1853 @code{operator new} will only return @code{0} if it is declared
1854 @samp{throw()}, in which case the compiler will always check the
1855 return value even without this option. In all other cases, when
1856 @code{operator new} has a non-empty exception specification, memory
1857 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1858 @samp{new (nothrow)}.
1860 @item -fconserve-space
1861 @opindex fconserve-space
1862 Put uninitialized or runtime-initialized global variables into the
1863 common segment, as C does. This saves space in the executable at the
1864 cost of not diagnosing duplicate definitions. If you compile with this
1865 flag and your program mysteriously crashes after @code{main()} has
1866 completed, you may have an object that is being destroyed twice because
1867 two definitions were merged.
1869 This option is no longer useful on most targets, now that support has
1870 been added for putting variables into BSS without making them common.
1872 @item -fconstexpr-depth=@var{n}
1873 @opindex fconstexpr-depth
1874 Set the maximum nested evaluation depth for C++0x constexpr functions
1875 to @var{n}. A limit is needed to detect endless recursion during
1876 constant expression evaluation. The minimum specified by the standard
1879 @item -fno-deduce-init-list
1880 @opindex fno-deduce-init-list
1881 Disable deduction of a template type parameter as
1882 std::initializer_list from a brace-enclosed initializer list, i.e.
1885 template <class T> auto forward(T t) -> decltype (realfn (t))
1892 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1896 This option is present because this deduction is an extension to the
1897 current specification in the C++0x working draft, and there was
1898 some concern about potential overload resolution problems.
1900 @item -ffriend-injection
1901 @opindex ffriend-injection
1902 Inject friend functions into the enclosing namespace, so that they are
1903 visible outside the scope of the class in which they are declared.
1904 Friend functions were documented to work this way in the old Annotated
1905 C++ Reference Manual, and versions of G++ before 4.1 always worked
1906 that way. However, in ISO C++ a friend function which is not declared
1907 in an enclosing scope can only be found using argument dependent
1908 lookup. This option causes friends to be injected as they were in
1911 This option is for compatibility, and may be removed in a future
1914 @item -fno-elide-constructors
1915 @opindex fno-elide-constructors
1916 The C++ standard allows an implementation to omit creating a temporary
1917 which is only used to initialize another object of the same type.
1918 Specifying this option disables that optimization, and forces G++ to
1919 call the copy constructor in all cases.
1921 @item -fno-enforce-eh-specs
1922 @opindex fno-enforce-eh-specs
1923 Don't generate code to check for violation of exception specifications
1924 at runtime. This option violates the C++ standard, but may be useful
1925 for reducing code size in production builds, much like defining
1926 @samp{NDEBUG}. This does not give user code permission to throw
1927 exceptions in violation of the exception specifications; the compiler
1928 will still optimize based on the specifications, so throwing an
1929 unexpected exception will result in undefined behavior.
1932 @itemx -fno-for-scope
1934 @opindex fno-for-scope
1935 If @option{-ffor-scope} is specified, the scope of variables declared in
1936 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1937 as specified by the C++ standard.
1938 If @option{-fno-for-scope} is specified, the scope of variables declared in
1939 a @i{for-init-statement} extends to the end of the enclosing scope,
1940 as was the case in old versions of G++, and other (traditional)
1941 implementations of C++.
1943 The default if neither flag is given to follow the standard,
1944 but to allow and give a warning for old-style code that would
1945 otherwise be invalid, or have different behavior.
1947 @item -fno-gnu-keywords
1948 @opindex fno-gnu-keywords
1949 Do not recognize @code{typeof} as a keyword, so that code can use this
1950 word as an identifier. You can use the keyword @code{__typeof__} instead.
1951 @option{-ansi} implies @option{-fno-gnu-keywords}.
1953 @item -fno-implicit-templates
1954 @opindex fno-implicit-templates
1955 Never emit code for non-inline templates which are instantiated
1956 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1957 @xref{Template Instantiation}, for more information.
1959 @item -fno-implicit-inline-templates
1960 @opindex fno-implicit-inline-templates
1961 Don't emit code for implicit instantiations of inline templates, either.
1962 The default is to handle inlines differently so that compiles with and
1963 without optimization will need the same set of explicit instantiations.
1965 @item -fno-implement-inlines
1966 @opindex fno-implement-inlines
1967 To save space, do not emit out-of-line copies of inline functions
1968 controlled by @samp{#pragma implementation}. This will cause linker
1969 errors if these functions are not inlined everywhere they are called.
1971 @item -fms-extensions
1972 @opindex fms-extensions
1973 Disable pedantic warnings about constructs used in MFC, such as implicit
1974 int and getting a pointer to member function via non-standard syntax.
1976 @item -fno-nonansi-builtins
1977 @opindex fno-nonansi-builtins
1978 Disable built-in declarations of functions that are not mandated by
1979 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1980 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1983 @opindex fnothrow-opt
1984 Treat a @code{throw()} exception specification as though it were a
1985 @code{noexcept} specification to reduce or eliminate the text size
1986 overhead relative to a function with no exception specification. If
1987 the function has local variables of types with non-trivial
1988 destructors, the exception specification will actually make the
1989 function smaller because the EH cleanups for those variables can be
1990 optimized away. The semantic effect is that an exception thrown out of
1991 a function with such an exception specification will result in a call
1992 to @code{terminate} rather than @code{unexpected}.
1994 @item -fno-operator-names
1995 @opindex fno-operator-names
1996 Do not treat the operator name keywords @code{and}, @code{bitand},
1997 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1998 synonyms as keywords.
2000 @item -fno-optional-diags
2001 @opindex fno-optional-diags
2002 Disable diagnostics that the standard says a compiler does not need to
2003 issue. Currently, the only such diagnostic issued by G++ is the one for
2004 a name having multiple meanings within a class.
2007 @opindex fpermissive
2008 Downgrade some diagnostics about nonconformant code from errors to
2009 warnings. Thus, using @option{-fpermissive} will allow some
2010 nonconforming code to compile.
2012 @item -fno-pretty-templates
2013 @opindex fno-pretty-templates
2014 When an error message refers to a specialization of a function
2015 template, the compiler will normally print the signature of the
2016 template followed by the template arguments and any typedefs or
2017 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2018 rather than @code{void f(int)}) so that it's clear which template is
2019 involved. When an error message refers to a specialization of a class
2020 template, the compiler will omit any template arguments which match
2021 the default template arguments for that template. If either of these
2022 behaviors make it harder to understand the error message rather than
2023 easier, using @option{-fno-pretty-templates} will disable them.
2027 Enable automatic template instantiation at link time. This option also
2028 implies @option{-fno-implicit-templates}. @xref{Template
2029 Instantiation}, for more information.
2033 Disable generation of information about every class with virtual
2034 functions for use by the C++ runtime type identification features
2035 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2036 of the language, you can save some space by using this flag. Note that
2037 exception handling uses the same information, but it will generate it as
2038 needed. The @samp{dynamic_cast} operator can still be used for casts that
2039 do not require runtime type information, i.e.@: casts to @code{void *} or to
2040 unambiguous base classes.
2044 Emit statistics about front-end processing at the end of the compilation.
2045 This information is generally only useful to the G++ development team.
2047 @item -fstrict-enums
2048 @opindex fstrict-enums
2049 Allow the compiler to optimize using the assumption that a value of
2050 enumeration type can only be one of the values of the enumeration (as
2051 defined in the C++ standard; basically, a value which can be
2052 represented in the minimum number of bits needed to represent all the
2053 enumerators). This assumption may not be valid if the program uses a
2054 cast to convert an arbitrary integer value to the enumeration type.
2056 @item -ftemplate-depth=@var{n}
2057 @opindex ftemplate-depth
2058 Set the maximum instantiation depth for template classes to @var{n}.
2059 A limit on the template instantiation depth is needed to detect
2060 endless recursions during template class instantiation. ANSI/ISO C++
2061 conforming programs must not rely on a maximum depth greater than 17
2062 (changed to 1024 in C++0x).
2064 @item -fno-threadsafe-statics
2065 @opindex fno-threadsafe-statics
2066 Do not emit the extra code to use the routines specified in the C++
2067 ABI for thread-safe initialization of local statics. You can use this
2068 option to reduce code size slightly in code that doesn't need to be
2071 @item -fuse-cxa-atexit
2072 @opindex fuse-cxa-atexit
2073 Register destructors for objects with static storage duration with the
2074 @code{__cxa_atexit} function rather than the @code{atexit} function.
2075 This option is required for fully standards-compliant handling of static
2076 destructors, but will only work if your C library supports
2077 @code{__cxa_atexit}.
2079 @item -fno-use-cxa-get-exception-ptr
2080 @opindex fno-use-cxa-get-exception-ptr
2081 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2082 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2083 if the runtime routine is not available.
2085 @item -fvisibility-inlines-hidden
2086 @opindex fvisibility-inlines-hidden
2087 This switch declares that the user does not attempt to compare
2088 pointers to inline methods where the addresses of the two functions
2089 were taken in different shared objects.
2091 The effect of this is that GCC may, effectively, mark inline methods with
2092 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2093 appear in the export table of a DSO and do not require a PLT indirection
2094 when used within the DSO@. Enabling this option can have a dramatic effect
2095 on load and link times of a DSO as it massively reduces the size of the
2096 dynamic export table when the library makes heavy use of templates.
2098 The behavior of this switch is not quite the same as marking the
2099 methods as hidden directly, because it does not affect static variables
2100 local to the function or cause the compiler to deduce that
2101 the function is defined in only one shared object.
2103 You may mark a method as having a visibility explicitly to negate the
2104 effect of the switch for that method. For example, if you do want to
2105 compare pointers to a particular inline method, you might mark it as
2106 having default visibility. Marking the enclosing class with explicit
2107 visibility will have no effect.
2109 Explicitly instantiated inline methods are unaffected by this option
2110 as their linkage might otherwise cross a shared library boundary.
2111 @xref{Template Instantiation}.
2113 @item -fvisibility-ms-compat
2114 @opindex fvisibility-ms-compat
2115 This flag attempts to use visibility settings to make GCC's C++
2116 linkage model compatible with that of Microsoft Visual Studio.
2118 The flag makes these changes to GCC's linkage model:
2122 It sets the default visibility to @code{hidden}, like
2123 @option{-fvisibility=hidden}.
2126 Types, but not their members, are not hidden by default.
2129 The One Definition Rule is relaxed for types without explicit
2130 visibility specifications which are defined in more than one different
2131 shared object: those declarations are permitted if they would have
2132 been permitted when this option was not used.
2135 In new code it is better to use @option{-fvisibility=hidden} and
2136 export those classes which are intended to be externally visible.
2137 Unfortunately it is possible for code to rely, perhaps accidentally,
2138 on the Visual Studio behavior.
2140 Among the consequences of these changes are that static data members
2141 of the same type with the same name but defined in different shared
2142 objects will be different, so changing one will not change the other;
2143 and that pointers to function members defined in different shared
2144 objects may not compare equal. When this flag is given, it is a
2145 violation of the ODR to define types with the same name differently.
2149 Do not use weak symbol support, even if it is provided by the linker.
2150 By default, G++ will use weak symbols if they are available. This
2151 option exists only for testing, and should not be used by end-users;
2152 it will result in inferior code and has no benefits. This option may
2153 be removed in a future release of G++.
2157 Do not search for header files in the standard directories specific to
2158 C++, but do still search the other standard directories. (This option
2159 is used when building the C++ library.)
2162 In addition, these optimization, warning, and code generation options
2163 have meanings only for C++ programs:
2166 @item -fno-default-inline
2167 @opindex fno-default-inline
2168 Do not assume @samp{inline} for functions defined inside a class scope.
2169 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2170 functions will have linkage like inline functions; they just won't be
2173 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2176 Warn when G++ generates code that is probably not compatible with the
2177 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2178 all such cases, there are probably some cases that are not warned about,
2179 even though G++ is generating incompatible code. There may also be
2180 cases where warnings are emitted even though the code that is generated
2183 You should rewrite your code to avoid these warnings if you are
2184 concerned about the fact that code generated by G++ may not be binary
2185 compatible with code generated by other compilers.
2187 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2192 A template with a non-type template parameter of reference type is
2193 mangled incorrectly:
2196 template <int &> struct S @{@};
2200 This is fixed in @option{-fabi-version=3}.
2203 SIMD vector types declared using @code{__attribute ((vector_size))} are
2204 mangled in a non-standard way that does not allow for overloading of
2205 functions taking vectors of different sizes.
2207 The mangling is changed in @option{-fabi-version=4}.
2210 The known incompatibilities in @option{-fabi-version=1} include:
2215 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2216 pack data into the same byte as a base class. For example:
2219 struct A @{ virtual void f(); int f1 : 1; @};
2220 struct B : public A @{ int f2 : 1; @};
2224 In this case, G++ will place @code{B::f2} into the same byte
2225 as@code{A::f1}; other compilers will not. You can avoid this problem
2226 by explicitly padding @code{A} so that its size is a multiple of the
2227 byte size on your platform; that will cause G++ and other compilers to
2228 layout @code{B} identically.
2231 Incorrect handling of tail-padding for virtual bases. G++ does not use
2232 tail padding when laying out virtual bases. For example:
2235 struct A @{ virtual void f(); char c1; @};
2236 struct B @{ B(); char c2; @};
2237 struct C : public A, public virtual B @{@};
2241 In this case, G++ will not place @code{B} into the tail-padding for
2242 @code{A}; other compilers will. You can avoid this problem by
2243 explicitly padding @code{A} so that its size is a multiple of its
2244 alignment (ignoring virtual base classes); that will cause G++ and other
2245 compilers to layout @code{C} identically.
2248 Incorrect handling of bit-fields with declared widths greater than that
2249 of their underlying types, when the bit-fields appear in a union. For
2253 union U @{ int i : 4096; @};
2257 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2258 union too small by the number of bits in an @code{int}.
2261 Empty classes can be placed at incorrect offsets. For example:
2271 struct C : public B, public A @{@};
2275 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2276 it should be placed at offset zero. G++ mistakenly believes that the
2277 @code{A} data member of @code{B} is already at offset zero.
2280 Names of template functions whose types involve @code{typename} or
2281 template template parameters can be mangled incorrectly.
2284 template <typename Q>
2285 void f(typename Q::X) @{@}
2287 template <template <typename> class Q>
2288 void f(typename Q<int>::X) @{@}
2292 Instantiations of these templates may be mangled incorrectly.
2296 It also warns psABI related changes. The known psABI changes at this
2302 For SYSV/x86-64, when passing union with long double, it is changed to
2303 pass in memory as specified in psABI. For example:
2313 @code{union U} will always be passed in memory.
2317 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2318 @opindex Wctor-dtor-privacy
2319 @opindex Wno-ctor-dtor-privacy
2320 Warn when a class seems unusable because all the constructors or
2321 destructors in that class are private, and it has neither friends nor
2322 public static member functions.
2324 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2326 @opindex Wno-noexcept
2327 Warn when a noexcept-expression evaluates to false because of a call
2328 to a function that does not have a non-throwing exception
2329 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2330 the compiler to never throw an exception.
2332 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2333 @opindex Wnon-virtual-dtor
2334 @opindex Wno-non-virtual-dtor
2335 Warn when a class has virtual functions and accessible non-virtual
2336 destructor, in which case it would be possible but unsafe to delete
2337 an instance of a derived class through a pointer to the base class.
2338 This warning is also enabled if -Weffc++ is specified.
2340 @item -Wreorder @r{(C++ and Objective-C++ only)}
2342 @opindex Wno-reorder
2343 @cindex reordering, warning
2344 @cindex warning for reordering of member initializers
2345 Warn when the order of member initializers given in the code does not
2346 match the order in which they must be executed. For instance:
2352 A(): j (0), i (1) @{ @}
2356 The compiler will rearrange the member initializers for @samp{i}
2357 and @samp{j} to match the declaration order of the members, emitting
2358 a warning to that effect. This warning is enabled by @option{-Wall}.
2361 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2364 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2367 Warn about violations of the following style guidelines from Scott Meyers'
2368 @cite{Effective C++} book:
2372 Item 11: Define a copy constructor and an assignment operator for classes
2373 with dynamically allocated memory.
2376 Item 12: Prefer initialization to assignment in constructors.
2379 Item 14: Make destructors virtual in base classes.
2382 Item 15: Have @code{operator=} return a reference to @code{*this}.
2385 Item 23: Don't try to return a reference when you must return an object.
2389 Also warn about violations of the following style guidelines from
2390 Scott Meyers' @cite{More Effective C++} book:
2394 Item 6: Distinguish between prefix and postfix forms of increment and
2395 decrement operators.
2398 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2402 When selecting this option, be aware that the standard library
2403 headers do not obey all of these guidelines; use @samp{grep -v}
2404 to filter out those warnings.
2406 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2407 @opindex Wstrict-null-sentinel
2408 @opindex Wno-strict-null-sentinel
2409 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2410 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2411 to @code{__null}. Although it is a null pointer constant not a null pointer,
2412 it is guaranteed to be of the same size as a pointer. But this use is
2413 not portable across different compilers.
2415 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2416 @opindex Wno-non-template-friend
2417 @opindex Wnon-template-friend
2418 Disable warnings when non-templatized friend functions are declared
2419 within a template. Since the advent of explicit template specification
2420 support in G++, if the name of the friend is an unqualified-id (i.e.,
2421 @samp{friend foo(int)}), the C++ language specification demands that the
2422 friend declare or define an ordinary, nontemplate function. (Section
2423 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2424 could be interpreted as a particular specialization of a templatized
2425 function. Because this non-conforming behavior is no longer the default
2426 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2427 check existing code for potential trouble spots and is on by default.
2428 This new compiler behavior can be turned off with
2429 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2430 but disables the helpful warning.
2432 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2433 @opindex Wold-style-cast
2434 @opindex Wno-old-style-cast
2435 Warn if an old-style (C-style) cast to a non-void type is used within
2436 a C++ program. The new-style casts (@samp{dynamic_cast},
2437 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2438 less vulnerable to unintended effects and much easier to search for.
2440 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2441 @opindex Woverloaded-virtual
2442 @opindex Wno-overloaded-virtual
2443 @cindex overloaded virtual function, warning
2444 @cindex warning for overloaded virtual function
2445 Warn when a function declaration hides virtual functions from a
2446 base class. For example, in:
2453 struct B: public A @{
2458 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2466 will fail to compile.
2468 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2469 @opindex Wno-pmf-conversions
2470 @opindex Wpmf-conversions
2471 Disable the diagnostic for converting a bound pointer to member function
2474 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2475 @opindex Wsign-promo
2476 @opindex Wno-sign-promo
2477 Warn when overload resolution chooses a promotion from unsigned or
2478 enumerated type to a signed type, over a conversion to an unsigned type of
2479 the same size. Previous versions of G++ would try to preserve
2480 unsignedness, but the standard mandates the current behavior.
2485 A& operator = (int);
2495 In this example, G++ will synthesize a default @samp{A& operator =
2496 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2499 @node Objective-C and Objective-C++ Dialect Options
2500 @section Options Controlling Objective-C and Objective-C++ Dialects
2502 @cindex compiler options, Objective-C and Objective-C++
2503 @cindex Objective-C and Objective-C++ options, command line
2504 @cindex options, Objective-C and Objective-C++
2505 (NOTE: This manual does not describe the Objective-C and Objective-C++
2506 languages themselves. @xref{Standards,,Language Standards
2507 Supported by GCC}, for references.)
2509 This section describes the command-line options that are only meaningful
2510 for Objective-C and Objective-C++ programs, but you can also use most of
2511 the language-independent GNU compiler options.
2512 For example, you might compile a file @code{some_class.m} like this:
2515 gcc -g -fgnu-runtime -O -c some_class.m
2519 In this example, @option{-fgnu-runtime} is an option meant only for
2520 Objective-C and Objective-C++ programs; you can use the other options with
2521 any language supported by GCC@.
2523 Note that since Objective-C is an extension of the C language, Objective-C
2524 compilations may also use options specific to the C front-end (e.g.,
2525 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2526 C++-specific options (e.g., @option{-Wabi}).
2528 Here is a list of options that are @emph{only} for compiling Objective-C
2529 and Objective-C++ programs:
2532 @item -fconstant-string-class=@var{class-name}
2533 @opindex fconstant-string-class
2534 Use @var{class-name} as the name of the class to instantiate for each
2535 literal string specified with the syntax @code{@@"@dots{}"}. The default
2536 class name is @code{NXConstantString} if the GNU runtime is being used, and
2537 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2538 @option{-fconstant-cfstrings} option, if also present, will override the
2539 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2540 to be laid out as constant CoreFoundation strings.
2543 @opindex fgnu-runtime
2544 Generate object code compatible with the standard GNU Objective-C
2545 runtime. This is the default for most types of systems.
2547 @item -fnext-runtime
2548 @opindex fnext-runtime
2549 Generate output compatible with the NeXT runtime. This is the default
2550 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2551 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2554 @item -fno-nil-receivers
2555 @opindex fno-nil-receivers
2556 Assume that all Objective-C message dispatches (@code{[receiver
2557 message:arg]}) in this translation unit ensure that the receiver is
2558 not @code{nil}. This allows for more efficient entry points in the
2559 runtime to be used. This option is only available in conjunction with
2560 the NeXT runtime and ABI version 0 or 1.
2562 @item -fobjc-abi-version=@var{n}
2563 @opindex fobjc-abi-version
2564 Use version @var{n} of the Objective-C ABI for the selected runtime.
2565 This option is currently supported only for the NeXT runtime. In that
2566 case, Version 0 is the traditional (32-bit) ABI without support for
2567 properties and other Objective-C 2.0 additions. Version 1 is the
2568 traditional (32-bit) ABI with support for properties and other
2569 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2570 nothing is specified, the default is Version 0 on 32-bit target
2571 machines, and Version 2 on 64-bit target machines.
2573 @item -fobjc-call-cxx-cdtors
2574 @opindex fobjc-call-cxx-cdtors
2575 For each Objective-C class, check if any of its instance variables is a
2576 C++ object with a non-trivial default constructor. If so, synthesize a
2577 special @code{- (id) .cxx_construct} instance method that will run
2578 non-trivial default constructors on any such instance variables, in order,
2579 and then return @code{self}. Similarly, check if any instance variable
2580 is a C++ object with a non-trivial destructor, and if so, synthesize a
2581 special @code{- (void) .cxx_destruct} method that will run
2582 all such default destructors, in reverse order.
2584 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2585 methods thusly generated will only operate on instance variables
2586 declared in the current Objective-C class, and not those inherited
2587 from superclasses. It is the responsibility of the Objective-C
2588 runtime to invoke all such methods in an object's inheritance
2589 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2590 by the runtime immediately after a new object instance is allocated;
2591 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2592 before the runtime deallocates an object instance.
2594 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2595 support for invoking the @code{- (id) .cxx_construct} and
2596 @code{- (void) .cxx_destruct} methods.
2598 @item -fobjc-direct-dispatch
2599 @opindex fobjc-direct-dispatch
2600 Allow fast jumps to the message dispatcher. On Darwin this is
2601 accomplished via the comm page.
2603 @item -fobjc-exceptions
2604 @opindex fobjc-exceptions
2605 Enable syntactic support for structured exception handling in
2606 Objective-C, similar to what is offered by C++ and Java. This option
2607 is required to use the Objective-C keywords @code{@@try},
2608 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2609 @code{@@synchronized}. This option is available with both the GNU
2610 runtime and the NeXT runtime (but not available in conjunction with
2611 the NeXT runtime on Mac OS X 10.2 and earlier).
2615 Enable garbage collection (GC) in Objective-C and Objective-C++
2616 programs. This option is only available with the NeXT runtime; the
2617 GNU runtime has a different garbage collection implementation that
2618 does not require special compiler flags.
2620 @item -fobjc-nilcheck
2621 @opindex fobjc-nilcheck
2622 For the NeXT runtime with version 2 of the ABI, check for a nil
2623 receiver in method invocations before doing the actual method call.
2624 This is the default and can be disabled using
2625 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2626 checked for nil in this way no matter what this flag is set to.
2627 Currently this flag does nothing when the GNU runtime, or an older
2628 version of the NeXT runtime ABI, is used.
2630 @item -fobjc-std=objc1
2632 Conform to the language syntax of Objective-C 1.0, the language
2633 recognized by GCC 4.0. This only affects the Objective-C additions to
2634 the C/C++ language; it does not affect conformance to C/C++ standards,
2635 which is controlled by the separate C/C++ dialect option flags. When
2636 this option is used with the Objective-C or Objective-C++ compiler,
2637 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2638 This is useful if you need to make sure that your Objective-C code can
2639 be compiled with older versions of GCC.
2641 @item -freplace-objc-classes
2642 @opindex freplace-objc-classes
2643 Emit a special marker instructing @command{ld(1)} not to statically link in
2644 the resulting object file, and allow @command{dyld(1)} to load it in at
2645 run time instead. This is used in conjunction with the Fix-and-Continue
2646 debugging mode, where the object file in question may be recompiled and
2647 dynamically reloaded in the course of program execution, without the need
2648 to restart the program itself. Currently, Fix-and-Continue functionality
2649 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2654 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2655 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2656 compile time) with static class references that get initialized at load time,
2657 which improves run-time performance. Specifying the @option{-fzero-link} flag
2658 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2659 to be retained. This is useful in Zero-Link debugging mode, since it allows
2660 for individual class implementations to be modified during program execution.
2661 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2662 regardless of command line options.
2666 Dump interface declarations for all classes seen in the source file to a
2667 file named @file{@var{sourcename}.decl}.
2669 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2670 @opindex Wassign-intercept
2671 @opindex Wno-assign-intercept
2672 Warn whenever an Objective-C assignment is being intercepted by the
2675 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2676 @opindex Wno-protocol
2678 If a class is declared to implement a protocol, a warning is issued for
2679 every method in the protocol that is not implemented by the class. The
2680 default behavior is to issue a warning for every method not explicitly
2681 implemented in the class, even if a method implementation is inherited
2682 from the superclass. If you use the @option{-Wno-protocol} option, then
2683 methods inherited from the superclass are considered to be implemented,
2684 and no warning is issued for them.
2686 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2688 @opindex Wno-selector
2689 Warn if multiple methods of different types for the same selector are
2690 found during compilation. The check is performed on the list of methods
2691 in the final stage of compilation. Additionally, a check is performed
2692 for each selector appearing in a @code{@@selector(@dots{})}
2693 expression, and a corresponding method for that selector has been found
2694 during compilation. Because these checks scan the method table only at
2695 the end of compilation, these warnings are not produced if the final
2696 stage of compilation is not reached, for example because an error is
2697 found during compilation, or because the @option{-fsyntax-only} option is
2700 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2701 @opindex Wstrict-selector-match
2702 @opindex Wno-strict-selector-match
2703 Warn if multiple methods with differing argument and/or return types are
2704 found for a given selector when attempting to send a message using this
2705 selector to a receiver of type @code{id} or @code{Class}. When this flag
2706 is off (which is the default behavior), the compiler will omit such warnings
2707 if any differences found are confined to types which share the same size
2710 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2711 @opindex Wundeclared-selector
2712 @opindex Wno-undeclared-selector
2713 Warn if a @code{@@selector(@dots{})} expression referring to an
2714 undeclared selector is found. A selector is considered undeclared if no
2715 method with that name has been declared before the
2716 @code{@@selector(@dots{})} expression, either explicitly in an
2717 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2718 an @code{@@implementation} section. This option always performs its
2719 checks as soon as a @code{@@selector(@dots{})} expression is found,
2720 while @option{-Wselector} only performs its checks in the final stage of
2721 compilation. This also enforces the coding style convention
2722 that methods and selectors must be declared before being used.
2724 @item -print-objc-runtime-info
2725 @opindex print-objc-runtime-info
2726 Generate C header describing the largest structure that is passed by
2731 @node Language Independent Options
2732 @section Options to Control Diagnostic Messages Formatting
2733 @cindex options to control diagnostics formatting
2734 @cindex diagnostic messages
2735 @cindex message formatting
2737 Traditionally, diagnostic messages have been formatted irrespective of
2738 the output device's aspect (e.g.@: its width, @dots{}). The options described
2739 below can be used to control the diagnostic messages formatting
2740 algorithm, e.g.@: how many characters per line, how often source location
2741 information should be reported. Right now, only the C++ front end can
2742 honor these options. However it is expected, in the near future, that
2743 the remaining front ends would be able to digest them correctly.
2746 @item -fmessage-length=@var{n}
2747 @opindex fmessage-length
2748 Try to format error messages so that they fit on lines of about @var{n}
2749 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2750 the front ends supported by GCC@. If @var{n} is zero, then no
2751 line-wrapping will be done; each error message will appear on a single
2754 @opindex fdiagnostics-show-location
2755 @item -fdiagnostics-show-location=once
2756 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2757 reporter to emit @emph{once} source location information; that is, in
2758 case the message is too long to fit on a single physical line and has to
2759 be wrapped, the source location won't be emitted (as prefix) again,
2760 over and over, in subsequent continuation lines. This is the default
2763 @item -fdiagnostics-show-location=every-line
2764 Only meaningful in line-wrapping mode. Instructs the diagnostic
2765 messages reporter to emit the same source location information (as
2766 prefix) for physical lines that result from the process of breaking
2767 a message which is too long to fit on a single line.
2769 @item -fno-diagnostics-show-option
2770 @opindex fno-diagnostics-show-option
2771 @opindex fdiagnostics-show-option
2772 By default, each diagnostic emitted includes text which indicates the
2773 command line option that directly controls the diagnostic (if such an
2774 option is known to the diagnostic machinery). Specifying the
2775 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2777 @item -Wcoverage-mismatch
2778 @opindex Wcoverage-mismatch
2779 Warn if feedback profiles do not match when using the
2780 @option{-fprofile-use} option.
2781 If a source file was changed between @option{-fprofile-gen} and
2782 @option{-fprofile-use}, the files with the profile feedback can fail
2783 to match the source file and GCC can not use the profile feedback
2784 information. By default, this warning is enabled and is treated as an
2785 error. @option{-Wno-coverage-mismatch} can be used to disable the
2786 warning or @option{-Wno-error=coverage-mismatch} can be used to
2787 disable the error. Disable the error for this warning can result in
2788 poorly optimized code, so disabling the error is useful only in the
2789 case of very minor changes such as bug fixes to an existing code-base.
2790 Completely disabling the warning is not recommended.
2794 @node Warning Options
2795 @section Options to Request or Suppress Warnings
2796 @cindex options to control warnings
2797 @cindex warning messages
2798 @cindex messages, warning
2799 @cindex suppressing warnings
2801 Warnings are diagnostic messages that report constructions which
2802 are not inherently erroneous but which are risky or suggest there
2803 may have been an error.
2805 The following language-independent options do not enable specific
2806 warnings but control the kinds of diagnostics produced by GCC.
2809 @cindex syntax checking
2811 @opindex fsyntax-only
2812 Check the code for syntax errors, but don't do anything beyond that.
2814 @item -fmax-errors=@var{n}
2815 @opindex fmax-errors
2816 Limits the maximum number of error messages to @var{n}, at which point
2817 GCC bails out rather than attempting to continue processing the source
2818 code. If @var{n} is 0 (the default), there is no limit on the number
2819 of error messages produced. If @option{-Wfatal-errors} is also
2820 specified, then @option{-Wfatal-errors} takes precedence over this
2825 Inhibit all warning messages.
2830 Make all warnings into errors.
2835 Make the specified warning into an error. The specifier for a warning
2836 is appended, for example @option{-Werror=switch} turns the warnings
2837 controlled by @option{-Wswitch} into errors. This switch takes a
2838 negative form, to be used to negate @option{-Werror} for specific
2839 warnings, for example @option{-Wno-error=switch} makes
2840 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2843 The warning message for each controllable warning includes the
2844 option which controls the warning. That option can then be used with
2845 @option{-Werror=} and @option{-Wno-error=} as described above.
2846 (Printing of the option in the warning message can be disabled using the
2847 @option{-fno-diagnostics-show-option} flag.)
2849 Note that specifying @option{-Werror=}@var{foo} automatically implies
2850 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2853 @item -Wfatal-errors
2854 @opindex Wfatal-errors
2855 @opindex Wno-fatal-errors
2856 This option causes the compiler to abort compilation on the first error
2857 occurred rather than trying to keep going and printing further error
2862 You can request many specific warnings with options beginning
2863 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2864 implicit declarations. Each of these specific warning options also
2865 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2866 example, @option{-Wno-implicit}. This manual lists only one of the
2867 two forms, whichever is not the default. For further,
2868 language-specific options also refer to @ref{C++ Dialect Options} and
2869 @ref{Objective-C and Objective-C++ Dialect Options}.
2871 When an unrecognized warning option is requested (e.g.,
2872 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2873 that the option is not recognized. However, if the @option{-Wno-} form
2874 is used, the behavior is slightly different: No diagnostic will be
2875 produced for @option{-Wno-unknown-warning} unless other diagnostics
2876 are being produced. This allows the use of new @option{-Wno-} options
2877 with old compilers, but if something goes wrong, the compiler will
2878 warn that an unrecognized option was used.
2883 Issue all the warnings demanded by strict ISO C and ISO C++;
2884 reject all programs that use forbidden extensions, and some other
2885 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2886 version of the ISO C standard specified by any @option{-std} option used.
2888 Valid ISO C and ISO C++ programs should compile properly with or without
2889 this option (though a rare few will require @option{-ansi} or a
2890 @option{-std} option specifying the required version of ISO C)@. However,
2891 without this option, certain GNU extensions and traditional C and C++
2892 features are supported as well. With this option, they are rejected.
2894 @option{-pedantic} does not cause warning messages for use of the
2895 alternate keywords whose names begin and end with @samp{__}. Pedantic
2896 warnings are also disabled in the expression that follows
2897 @code{__extension__}. However, only system header files should use
2898 these escape routes; application programs should avoid them.
2899 @xref{Alternate Keywords}.
2901 Some users try to use @option{-pedantic} to check programs for strict ISO
2902 C conformance. They soon find that it does not do quite what they want:
2903 it finds some non-ISO practices, but not all---only those for which
2904 ISO C @emph{requires} a diagnostic, and some others for which
2905 diagnostics have been added.
2907 A feature to report any failure to conform to ISO C might be useful in
2908 some instances, but would require considerable additional work and would
2909 be quite different from @option{-pedantic}. We don't have plans to
2910 support such a feature in the near future.
2912 Where the standard specified with @option{-std} represents a GNU
2913 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2914 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2915 extended dialect is based. Warnings from @option{-pedantic} are given
2916 where they are required by the base standard. (It would not make sense
2917 for such warnings to be given only for features not in the specified GNU
2918 C dialect, since by definition the GNU dialects of C include all
2919 features the compiler supports with the given option, and there would be
2920 nothing to warn about.)
2922 @item -pedantic-errors
2923 @opindex pedantic-errors
2924 Like @option{-pedantic}, except that errors are produced rather than
2930 This enables all the warnings about constructions that some users
2931 consider questionable, and that are easy to avoid (or modify to
2932 prevent the warning), even in conjunction with macros. This also
2933 enables some language-specific warnings described in @ref{C++ Dialect
2934 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2936 @option{-Wall} turns on the following warning flags:
2938 @gccoptlist{-Waddress @gol
2939 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2941 -Wchar-subscripts @gol
2942 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2943 -Wimplicit-int @r{(C and Objective-C only)} @gol
2944 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2947 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2948 -Wmissing-braces @gol
2954 -Wsequence-point @gol
2955 -Wsign-compare @r{(only in C++)} @gol
2956 -Wstrict-aliasing @gol
2957 -Wstrict-overflow=1 @gol
2960 -Wuninitialized @gol
2961 -Wunknown-pragmas @gol
2962 -Wunused-function @gol
2965 -Wunused-variable @gol
2966 -Wvolatile-register-var @gol
2969 Note that some warning flags are not implied by @option{-Wall}. Some of
2970 them warn about constructions that users generally do not consider
2971 questionable, but which occasionally you might wish to check for;
2972 others warn about constructions that are necessary or hard to avoid in
2973 some cases, and there is no simple way to modify the code to suppress
2974 the warning. Some of them are enabled by @option{-Wextra} but many of
2975 them must be enabled individually.
2981 This enables some extra warning flags that are not enabled by
2982 @option{-Wall}. (This option used to be called @option{-W}. The older
2983 name is still supported, but the newer name is more descriptive.)
2985 @gccoptlist{-Wclobbered @gol
2987 -Wignored-qualifiers @gol
2988 -Wmissing-field-initializers @gol
2989 -Wmissing-parameter-type @r{(C only)} @gol
2990 -Wold-style-declaration @r{(C only)} @gol
2991 -Woverride-init @gol
2994 -Wuninitialized @gol
2995 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2996 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2999 The option @option{-Wextra} also prints warning messages for the
3005 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3006 @samp{>}, or @samp{>=}.
3009 (C++ only) An enumerator and a non-enumerator both appear in a
3010 conditional expression.
3013 (C++ only) Ambiguous virtual bases.
3016 (C++ only) Subscripting an array which has been declared @samp{register}.
3019 (C++ only) Taking the address of a variable which has been declared
3023 (C++ only) A base class is not initialized in a derived class' copy
3028 @item -Wchar-subscripts
3029 @opindex Wchar-subscripts
3030 @opindex Wno-char-subscripts
3031 Warn if an array subscript has type @code{char}. This is a common cause
3032 of error, as programmers often forget that this type is signed on some
3034 This warning is enabled by @option{-Wall}.
3038 @opindex Wno-comment
3039 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3040 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3041 This warning is enabled by @option{-Wall}.
3044 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3046 Suppress warning messages emitted by @code{#warning} directives.
3048 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3049 @opindex Wdouble-promotion
3050 @opindex Wno-double-promotion
3051 Give a warning when a value of type @code{float} is implicitly
3052 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3053 floating-point unit implement @code{float} in hardware, but emulate
3054 @code{double} in software. On such a machine, doing computations
3055 using @code{double} values is much more expensive because of the
3056 overhead required for software emulation.
3058 It is easy to accidentally do computations with @code{double} because
3059 floating-point literals are implicitly of type @code{double}. For
3063 float area(float radius)
3065 return 3.14159 * radius * radius;
3069 the compiler will perform the entire computation with @code{double}
3070 because the floating-point literal is a @code{double}.
3075 @opindex ffreestanding
3076 @opindex fno-builtin
3077 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3078 the arguments supplied have types appropriate to the format string
3079 specified, and that the conversions specified in the format string make
3080 sense. This includes standard functions, and others specified by format
3081 attributes (@pxref{Function Attributes}), in the @code{printf},
3082 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3083 not in the C standard) families (or other target-specific families).
3084 Which functions are checked without format attributes having been
3085 specified depends on the standard version selected, and such checks of
3086 functions without the attribute specified are disabled by
3087 @option{-ffreestanding} or @option{-fno-builtin}.
3089 The formats are checked against the format features supported by GNU
3090 libc version 2.2. These include all ISO C90 and C99 features, as well
3091 as features from the Single Unix Specification and some BSD and GNU
3092 extensions. Other library implementations may not support all these
3093 features; GCC does not support warning about features that go beyond a
3094 particular library's limitations. However, if @option{-pedantic} is used
3095 with @option{-Wformat}, warnings will be given about format features not
3096 in the selected standard version (but not for @code{strfmon} formats,
3097 since those are not in any version of the C standard). @xref{C Dialect
3098 Options,,Options Controlling C Dialect}.
3100 Since @option{-Wformat} also checks for null format arguments for
3101 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3103 @option{-Wformat} is included in @option{-Wall}. For more control over some
3104 aspects of format checking, the options @option{-Wformat-y2k},
3105 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3106 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3107 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3110 @opindex Wformat-y2k
3111 @opindex Wno-format-y2k
3112 If @option{-Wformat} is specified, also warn about @code{strftime}
3113 formats which may yield only a two-digit year.
3115 @item -Wno-format-contains-nul
3116 @opindex Wno-format-contains-nul
3117 @opindex Wformat-contains-nul
3118 If @option{-Wformat} is specified, do not warn about format strings that
3121 @item -Wno-format-extra-args
3122 @opindex Wno-format-extra-args
3123 @opindex Wformat-extra-args
3124 If @option{-Wformat} is specified, do not warn about excess arguments to a
3125 @code{printf} or @code{scanf} format function. The C standard specifies
3126 that such arguments are ignored.
3128 Where the unused arguments lie between used arguments that are
3129 specified with @samp{$} operand number specifications, normally
3130 warnings are still given, since the implementation could not know what
3131 type to pass to @code{va_arg} to skip the unused arguments. However,
3132 in the case of @code{scanf} formats, this option will suppress the
3133 warning if the unused arguments are all pointers, since the Single
3134 Unix Specification says that such unused arguments are allowed.
3136 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3137 @opindex Wno-format-zero-length
3138 @opindex Wformat-zero-length
3139 If @option{-Wformat} is specified, do not warn about zero-length formats.
3140 The C standard specifies that zero-length formats are allowed.
3142 @item -Wformat-nonliteral
3143 @opindex Wformat-nonliteral
3144 @opindex Wno-format-nonliteral
3145 If @option{-Wformat} is specified, also warn if the format string is not a
3146 string literal and so cannot be checked, unless the format function
3147 takes its format arguments as a @code{va_list}.
3149 @item -Wformat-security
3150 @opindex Wformat-security
3151 @opindex Wno-format-security
3152 If @option{-Wformat} is specified, also warn about uses of format
3153 functions that represent possible security problems. At present, this
3154 warns about calls to @code{printf} and @code{scanf} functions where the
3155 format string is not a string literal and there are no format arguments,
3156 as in @code{printf (foo);}. This may be a security hole if the format
3157 string came from untrusted input and contains @samp{%n}. (This is
3158 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3159 in future warnings may be added to @option{-Wformat-security} that are not
3160 included in @option{-Wformat-nonliteral}.)
3164 @opindex Wno-format=2
3165 Enable @option{-Wformat} plus format checks not included in
3166 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3167 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3169 @item -Wnonnull @r{(C and Objective-C only)}
3171 @opindex Wno-nonnull
3172 Warn about passing a null pointer for arguments marked as
3173 requiring a non-null value by the @code{nonnull} function attribute.
3175 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3176 can be disabled with the @option{-Wno-nonnull} option.
3178 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3180 @opindex Wno-init-self
3181 Warn about uninitialized variables which are initialized with themselves.
3182 Note this option can only be used with the @option{-Wuninitialized} option.
3184 For example, GCC will warn about @code{i} being uninitialized in the
3185 following snippet only when @option{-Winit-self} has been specified:
3196 @item -Wimplicit-int @r{(C and Objective-C only)}
3197 @opindex Wimplicit-int
3198 @opindex Wno-implicit-int
3199 Warn when a declaration does not specify a type.
3200 This warning is enabled by @option{-Wall}.
3202 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3203 @opindex Wimplicit-function-declaration
3204 @opindex Wno-implicit-function-declaration
3205 Give a warning whenever a function is used before being declared. In
3206 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3207 enabled by default and it is made into an error by
3208 @option{-pedantic-errors}. This warning is also enabled by
3211 @item -Wimplicit @r{(C and Objective-C only)}
3213 @opindex Wno-implicit
3214 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3215 This warning is enabled by @option{-Wall}.
3217 @item -Wignored-qualifiers @r{(C and C++ only)}
3218 @opindex Wignored-qualifiers
3219 @opindex Wno-ignored-qualifiers
3220 Warn if the return type of a function has a type qualifier
3221 such as @code{const}. For ISO C such a type qualifier has no effect,
3222 since the value returned by a function is not an lvalue.
3223 For C++, the warning is only emitted for scalar types or @code{void}.
3224 ISO C prohibits qualified @code{void} return types on function
3225 definitions, so such return types always receive a warning
3226 even without this option.
3228 This warning is also enabled by @option{-Wextra}.
3233 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3234 a function with external linkage, returning int, taking either zero
3235 arguments, two, or three arguments of appropriate types. This warning
3236 is enabled by default in C++ and is enabled by either @option{-Wall}
3237 or @option{-pedantic}.
3239 @item -Wmissing-braces
3240 @opindex Wmissing-braces
3241 @opindex Wno-missing-braces
3242 Warn if an aggregate or union initializer is not fully bracketed. In
3243 the following example, the initializer for @samp{a} is not fully
3244 bracketed, but that for @samp{b} is fully bracketed.
3247 int a[2][2] = @{ 0, 1, 2, 3 @};
3248 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3251 This warning is enabled by @option{-Wall}.
3253 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3254 @opindex Wmissing-include-dirs
3255 @opindex Wno-missing-include-dirs
3256 Warn if a user-supplied include directory does not exist.
3259 @opindex Wparentheses
3260 @opindex Wno-parentheses
3261 Warn if parentheses are omitted in certain contexts, such
3262 as when there is an assignment in a context where a truth value
3263 is expected, or when operators are nested whose precedence people
3264 often get confused about.
3266 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3267 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3268 interpretation from that of ordinary mathematical notation.
3270 Also warn about constructions where there may be confusion to which
3271 @code{if} statement an @code{else} branch belongs. Here is an example of
3286 In C/C++, every @code{else} branch belongs to the innermost possible
3287 @code{if} statement, which in this example is @code{if (b)}. This is
3288 often not what the programmer expected, as illustrated in the above
3289 example by indentation the programmer chose. When there is the
3290 potential for this confusion, GCC will issue a warning when this flag
3291 is specified. To eliminate the warning, add explicit braces around
3292 the innermost @code{if} statement so there is no way the @code{else}
3293 could belong to the enclosing @code{if}. The resulting code would
3310 Also warn for dangerous uses of the
3311 ?: with omitted middle operand GNU extension. When the condition
3312 in the ?: operator is a boolean expression the omitted value will
3313 be always 1. Often the user expects it to be a value computed
3314 inside the conditional expression instead.
3316 This warning is enabled by @option{-Wall}.
3318 @item -Wsequence-point
3319 @opindex Wsequence-point
3320 @opindex Wno-sequence-point
3321 Warn about code that may have undefined semantics because of violations
3322 of sequence point rules in the C and C++ standards.
3324 The C and C++ standards defines the order in which expressions in a C/C++
3325 program are evaluated in terms of @dfn{sequence points}, which represent
3326 a partial ordering between the execution of parts of the program: those
3327 executed before the sequence point, and those executed after it. These
3328 occur after the evaluation of a full expression (one which is not part
3329 of a larger expression), after the evaluation of the first operand of a
3330 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3331 function is called (but after the evaluation of its arguments and the
3332 expression denoting the called function), and in certain other places.
3333 Other than as expressed by the sequence point rules, the order of
3334 evaluation of subexpressions of an expression is not specified. All
3335 these rules describe only a partial order rather than a total order,
3336 since, for example, if two functions are called within one expression
3337 with no sequence point between them, the order in which the functions
3338 are called is not specified. However, the standards committee have
3339 ruled that function calls do not overlap.
3341 It is not specified when between sequence points modifications to the
3342 values of objects take effect. Programs whose behavior depends on this
3343 have undefined behavior; the C and C++ standards specify that ``Between
3344 the previous and next sequence point an object shall have its stored
3345 value modified at most once by the evaluation of an expression.
3346 Furthermore, the prior value shall be read only to determine the value
3347 to be stored.''. If a program breaks these rules, the results on any
3348 particular implementation are entirely unpredictable.
3350 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3351 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3352 diagnosed by this option, and it may give an occasional false positive
3353 result, but in general it has been found fairly effective at detecting
3354 this sort of problem in programs.
3356 The standard is worded confusingly, therefore there is some debate
3357 over the precise meaning of the sequence point rules in subtle cases.
3358 Links to discussions of the problem, including proposed formal
3359 definitions, may be found on the GCC readings page, at
3360 @uref{http://gcc.gnu.org/@/readings.html}.
3362 This warning is enabled by @option{-Wall} for C and C++.
3365 @opindex Wreturn-type
3366 @opindex Wno-return-type
3367 Warn whenever a function is defined with a return-type that defaults
3368 to @code{int}. Also warn about any @code{return} statement with no
3369 return-value in a function whose return-type is not @code{void}
3370 (falling off the end of the function body is considered returning
3371 without a value), and about a @code{return} statement with an
3372 expression in a function whose return-type is @code{void}.
3374 For C++, a function without return type always produces a diagnostic
3375 message, even when @option{-Wno-return-type} is specified. The only
3376 exceptions are @samp{main} and functions defined in system headers.
3378 This warning is enabled by @option{-Wall}.
3383 Warn whenever a @code{switch} statement has an index of enumerated type
3384 and lacks a @code{case} for one or more of the named codes of that
3385 enumeration. (The presence of a @code{default} label prevents this
3386 warning.) @code{case} labels outside the enumeration range also
3387 provoke warnings when this option is used (even if there is a
3388 @code{default} label).
3389 This warning is enabled by @option{-Wall}.
3391 @item -Wswitch-default
3392 @opindex Wswitch-default
3393 @opindex Wno-switch-default
3394 Warn whenever a @code{switch} statement does not have a @code{default}
3398 @opindex Wswitch-enum
3399 @opindex Wno-switch-enum
3400 Warn whenever a @code{switch} statement has an index of enumerated type
3401 and lacks a @code{case} for one or more of the named codes of that
3402 enumeration. @code{case} labels outside the enumeration range also
3403 provoke warnings when this option is used. The only difference
3404 between @option{-Wswitch} and this option is that this option gives a
3405 warning about an omitted enumeration code even if there is a
3406 @code{default} label.
3408 @item -Wsync-nand @r{(C and C++ only)}
3410 @opindex Wno-sync-nand
3411 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3412 built-in functions are used. These functions changed semantics in GCC 4.4.
3416 @opindex Wno-trigraphs
3417 Warn if any trigraphs are encountered that might change the meaning of
3418 the program (trigraphs within comments are not warned about).
3419 This warning is enabled by @option{-Wall}.
3421 @item -Wunused-but-set-parameter
3422 @opindex Wunused-but-set-parameter
3423 @opindex Wno-unused-but-set-parameter
3424 Warn whenever a function parameter is assigned to, but otherwise unused
3425 (aside from its declaration).
3427 To suppress this warning use the @samp{unused} attribute
3428 (@pxref{Variable Attributes}).
3430 This warning is also enabled by @option{-Wunused} together with
3433 @item -Wunused-but-set-variable
3434 @opindex Wunused-but-set-variable
3435 @opindex Wno-unused-but-set-variable
3436 Warn whenever a local variable is assigned to, but otherwise unused
3437 (aside from its declaration).
3438 This warning is enabled by @option{-Wall}.
3440 To suppress this warning use the @samp{unused} attribute
3441 (@pxref{Variable Attributes}).
3443 This warning is also enabled by @option{-Wunused}, which is enabled
3446 @item -Wunused-function
3447 @opindex Wunused-function
3448 @opindex Wno-unused-function
3449 Warn whenever a static function is declared but not defined or a
3450 non-inline static function is unused.
3451 This warning is enabled by @option{-Wall}.
3453 @item -Wunused-label
3454 @opindex Wunused-label
3455 @opindex Wno-unused-label
3456 Warn whenever a label is declared but not used.
3457 This warning is enabled by @option{-Wall}.
3459 To suppress this warning use the @samp{unused} attribute
3460 (@pxref{Variable Attributes}).
3462 @item -Wunused-parameter
3463 @opindex Wunused-parameter
3464 @opindex Wno-unused-parameter
3465 Warn whenever a function parameter is unused aside from its declaration.
3467 To suppress this warning use the @samp{unused} attribute
3468 (@pxref{Variable Attributes}).
3470 @item -Wno-unused-result
3471 @opindex Wunused-result
3472 @opindex Wno-unused-result
3473 Do not warn if a caller of a function marked with attribute
3474 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3475 its return value. The default is @option{-Wunused-result}.
3477 @item -Wunused-variable
3478 @opindex Wunused-variable
3479 @opindex Wno-unused-variable
3480 Warn whenever a local variable or non-constant static variable is unused
3481 aside from its declaration.
3482 This warning is enabled by @option{-Wall}.
3484 To suppress this warning use the @samp{unused} attribute
3485 (@pxref{Variable Attributes}).
3487 @item -Wunused-value
3488 @opindex Wunused-value
3489 @opindex Wno-unused-value
3490 Warn whenever a statement computes a result that is explicitly not
3491 used. To suppress this warning cast the unused expression to
3492 @samp{void}. This includes an expression-statement or the left-hand
3493 side of a comma expression that contains no side effects. For example,
3494 an expression such as @samp{x[i,j]} will cause a warning, while
3495 @samp{x[(void)i,j]} will not.
3497 This warning is enabled by @option{-Wall}.
3502 All the above @option{-Wunused} options combined.
3504 In order to get a warning about an unused function parameter, you must
3505 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3506 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3508 @item -Wuninitialized
3509 @opindex Wuninitialized
3510 @opindex Wno-uninitialized
3511 Warn if an automatic variable is used without first being initialized
3512 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3513 warn if a non-static reference or non-static @samp{const} member
3514 appears in a class without constructors.
3516 If you want to warn about code which uses the uninitialized value of the
3517 variable in its own initializer, use the @option{-Winit-self} option.
3519 These warnings occur for individual uninitialized or clobbered
3520 elements of structure, union or array variables as well as for
3521 variables which are uninitialized or clobbered as a whole. They do
3522 not occur for variables or elements declared @code{volatile}. Because
3523 these warnings depend on optimization, the exact variables or elements
3524 for which there are warnings will depend on the precise optimization
3525 options and version of GCC used.
3527 Note that there may be no warning about a variable that is used only
3528 to compute a value that itself is never used, because such
3529 computations may be deleted by data flow analysis before the warnings
3532 These warnings are made optional because GCC is not smart
3533 enough to see all the reasons why the code might be correct
3534 despite appearing to have an error. Here is one example of how
3555 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3556 always initialized, but GCC doesn't know this. Here is
3557 another common case:
3562 if (change_y) save_y = y, y = new_y;
3564 if (change_y) y = save_y;
3569 This has no bug because @code{save_y} is used only if it is set.
3571 @cindex @code{longjmp} warnings
3572 This option also warns when a non-volatile automatic variable might be
3573 changed by a call to @code{longjmp}. These warnings as well are possible
3574 only in optimizing compilation.
3576 The compiler sees only the calls to @code{setjmp}. It cannot know
3577 where @code{longjmp} will be called; in fact, a signal handler could
3578 call it at any point in the code. As a result, you may get a warning
3579 even when there is in fact no problem because @code{longjmp} cannot
3580 in fact be called at the place which would cause a problem.
3582 Some spurious warnings can be avoided if you declare all the functions
3583 you use that never return as @code{noreturn}. @xref{Function
3586 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3588 @item -Wunknown-pragmas
3589 @opindex Wunknown-pragmas
3590 @opindex Wno-unknown-pragmas
3591 @cindex warning for unknown pragmas
3592 @cindex unknown pragmas, warning
3593 @cindex pragmas, warning of unknown
3594 Warn when a #pragma directive is encountered which is not understood by
3595 GCC@. If this command line option is used, warnings will even be issued
3596 for unknown pragmas in system header files. This is not the case if
3597 the warnings were only enabled by the @option{-Wall} command line option.
3600 @opindex Wno-pragmas
3602 Do not warn about misuses of pragmas, such as incorrect parameters,
3603 invalid syntax, or conflicts between pragmas. See also
3604 @samp{-Wunknown-pragmas}.
3606 @item -Wstrict-aliasing
3607 @opindex Wstrict-aliasing
3608 @opindex Wno-strict-aliasing
3609 This option is only active when @option{-fstrict-aliasing} is active.
3610 It warns about code which might break the strict aliasing rules that the
3611 compiler is using for optimization. The warning does not catch all
3612 cases, but does attempt to catch the more common pitfalls. It is
3613 included in @option{-Wall}.
3614 It is equivalent to @option{-Wstrict-aliasing=3}
3616 @item -Wstrict-aliasing=n
3617 @opindex Wstrict-aliasing=n
3618 @opindex Wno-strict-aliasing=n
3619 This option is only active when @option{-fstrict-aliasing} is active.
3620 It warns about code which might break the strict aliasing rules that the
3621 compiler is using for optimization.
3622 Higher levels correspond to higher accuracy (fewer false positives).
3623 Higher levels also correspond to more effort, similar to the way -O works.
3624 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3627 Level 1: Most aggressive, quick, least accurate.
3628 Possibly useful when higher levels
3629 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3630 false negatives. However, it has many false positives.
3631 Warns for all pointer conversions between possibly incompatible types,
3632 even if never dereferenced. Runs in the frontend only.
3634 Level 2: Aggressive, quick, not too precise.
3635 May still have many false positives (not as many as level 1 though),
3636 and few false negatives (but possibly more than level 1).
3637 Unlike level 1, it only warns when an address is taken. Warns about
3638 incomplete types. Runs in the frontend only.
3640 Level 3 (default for @option{-Wstrict-aliasing}):
3641 Should have very few false positives and few false
3642 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3643 Takes care of the common pun+dereference pattern in the frontend:
3644 @code{*(int*)&some_float}.
3645 If optimization is enabled, it also runs in the backend, where it deals
3646 with multiple statement cases using flow-sensitive points-to information.
3647 Only warns when the converted pointer is dereferenced.
3648 Does not warn about incomplete types.
3650 @item -Wstrict-overflow
3651 @itemx -Wstrict-overflow=@var{n}
3652 @opindex Wstrict-overflow
3653 @opindex Wno-strict-overflow
3654 This option is only active when @option{-fstrict-overflow} is active.
3655 It warns about cases where the compiler optimizes based on the
3656 assumption that signed overflow does not occur. Note that it does not
3657 warn about all cases where the code might overflow: it only warns
3658 about cases where the compiler implements some optimization. Thus
3659 this warning depends on the optimization level.
3661 An optimization which assumes that signed overflow does not occur is
3662 perfectly safe if the values of the variables involved are such that
3663 overflow never does, in fact, occur. Therefore this warning can
3664 easily give a false positive: a warning about code which is not
3665 actually a problem. To help focus on important issues, several
3666 warning levels are defined. No warnings are issued for the use of
3667 undefined signed overflow when estimating how many iterations a loop
3668 will require, in particular when determining whether a loop will be
3672 @item -Wstrict-overflow=1
3673 Warn about cases which are both questionable and easy to avoid. For
3674 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3675 compiler will simplify this to @code{1}. This level of
3676 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3677 are not, and must be explicitly requested.
3679 @item -Wstrict-overflow=2
3680 Also warn about other cases where a comparison is simplified to a
3681 constant. For example: @code{abs (x) >= 0}. This can only be
3682 simplified when @option{-fstrict-overflow} is in effect, because
3683 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3684 zero. @option{-Wstrict-overflow} (with no level) is the same as
3685 @option{-Wstrict-overflow=2}.
3687 @item -Wstrict-overflow=3
3688 Also warn about other cases where a comparison is simplified. For
3689 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3691 @item -Wstrict-overflow=4
3692 Also warn about other simplifications not covered by the above cases.
3693 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3695 @item -Wstrict-overflow=5
3696 Also warn about cases where the compiler reduces the magnitude of a
3697 constant involved in a comparison. For example: @code{x + 2 > y} will
3698 be simplified to @code{x + 1 >= y}. This is reported only at the
3699 highest warning level because this simplification applies to many
3700 comparisons, so this warning level will give a very large number of
3704 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3705 @opindex Wsuggest-attribute=
3706 @opindex Wno-suggest-attribute=
3707 Warn for cases where adding an attribute may be beneficial. The
3708 attributes currently supported are listed below.
3711 @item -Wsuggest-attribute=pure
3712 @itemx -Wsuggest-attribute=const
3713 @itemx -Wsuggest-attribute=noreturn
3714 @opindex Wsuggest-attribute=pure
3715 @opindex Wno-suggest-attribute=pure
3716 @opindex Wsuggest-attribute=const
3717 @opindex Wno-suggest-attribute=const
3718 @opindex Wsuggest-attribute=noreturn
3719 @opindex Wno-suggest-attribute=noreturn
3721 Warn about functions which might be candidates for attributes
3722 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3723 functions visible in other compilation units or (in the case of @code{pure} and
3724 @code{const}) if it cannot prove that the function returns normally. A function
3725 returns normally if it doesn't contain an infinite loop nor returns abnormally
3726 by throwing, calling @code{abort()} or trapping. This analysis requires option
3727 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3728 higher. Higher optimization levels improve the accuracy of the analysis.
3731 @item -Warray-bounds
3732 @opindex Wno-array-bounds
3733 @opindex Warray-bounds
3734 This option is only active when @option{-ftree-vrp} is active
3735 (default for @option{-O2} and above). It warns about subscripts to arrays
3736 that are always out of bounds. This warning is enabled by @option{-Wall}.
3738 @item -Wno-div-by-zero
3739 @opindex Wno-div-by-zero
3740 @opindex Wdiv-by-zero
3741 Do not warn about compile-time integer division by zero. Floating point
3742 division by zero is not warned about, as it can be a legitimate way of
3743 obtaining infinities and NaNs.
3745 @item -Wsystem-headers
3746 @opindex Wsystem-headers
3747 @opindex Wno-system-headers
3748 @cindex warnings from system headers
3749 @cindex system headers, warnings from
3750 Print warning messages for constructs found in system header files.
3751 Warnings from system headers are normally suppressed, on the assumption
3752 that they usually do not indicate real problems and would only make the
3753 compiler output harder to read. Using this command line option tells
3754 GCC to emit warnings from system headers as if they occurred in user
3755 code. However, note that using @option{-Wall} in conjunction with this
3756 option will @emph{not} warn about unknown pragmas in system
3757 headers---for that, @option{-Wunknown-pragmas} must also be used.
3760 @opindex Wtrampolines
3761 @opindex Wno-trampolines
3762 Warn about trampolines generated for pointers to nested functions.
3764 A trampoline is a small piece of data or code that is created at run
3765 time on the stack when the address of a nested function is taken, and
3766 is used to call the nested function indirectly. For some targets, it
3767 is made up of data only and thus requires no special treatment. But,
3768 for most targets, it is made up of code and thus requires the stack
3769 to be made executable in order for the program to work properly.
3772 @opindex Wfloat-equal
3773 @opindex Wno-float-equal
3774 Warn if floating point values are used in equality comparisons.
3776 The idea behind this is that sometimes it is convenient (for the
3777 programmer) to consider floating-point values as approximations to
3778 infinitely precise real numbers. If you are doing this, then you need
3779 to compute (by analyzing the code, or in some other way) the maximum or
3780 likely maximum error that the computation introduces, and allow for it
3781 when performing comparisons (and when producing output, but that's a
3782 different problem). In particular, instead of testing for equality, you
3783 would check to see whether the two values have ranges that overlap; and
3784 this is done with the relational operators, so equality comparisons are
3787 @item -Wtraditional @r{(C and Objective-C only)}
3788 @opindex Wtraditional
3789 @opindex Wno-traditional
3790 Warn about certain constructs that behave differently in traditional and
3791 ISO C@. Also warn about ISO C constructs that have no traditional C
3792 equivalent, and/or problematic constructs which should be avoided.
3796 Macro parameters that appear within string literals in the macro body.
3797 In traditional C macro replacement takes place within string literals,
3798 but does not in ISO C@.
3801 In traditional C, some preprocessor directives did not exist.
3802 Traditional preprocessors would only consider a line to be a directive
3803 if the @samp{#} appeared in column 1 on the line. Therefore
3804 @option{-Wtraditional} warns about directives that traditional C
3805 understands but would ignore because the @samp{#} does not appear as the
3806 first character on the line. It also suggests you hide directives like
3807 @samp{#pragma} not understood by traditional C by indenting them. Some
3808 traditional implementations would not recognize @samp{#elif}, so it
3809 suggests avoiding it altogether.
3812 A function-like macro that appears without arguments.
3815 The unary plus operator.
3818 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3819 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3820 constants.) Note, these suffixes appear in macros defined in the system
3821 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3822 Use of these macros in user code might normally lead to spurious
3823 warnings, however GCC's integrated preprocessor has enough context to
3824 avoid warning in these cases.
3827 A function declared external in one block and then used after the end of
3831 A @code{switch} statement has an operand of type @code{long}.
3834 A non-@code{static} function declaration follows a @code{static} one.
3835 This construct is not accepted by some traditional C compilers.
3838 The ISO type of an integer constant has a different width or
3839 signedness from its traditional type. This warning is only issued if
3840 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3841 typically represent bit patterns, are not warned about.
3844 Usage of ISO string concatenation is detected.
3847 Initialization of automatic aggregates.
3850 Identifier conflicts with labels. Traditional C lacks a separate
3851 namespace for labels.
3854 Initialization of unions. If the initializer is zero, the warning is
3855 omitted. This is done under the assumption that the zero initializer in
3856 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3857 initializer warnings and relies on default initialization to zero in the
3861 Conversions by prototypes between fixed/floating point values and vice
3862 versa. The absence of these prototypes when compiling with traditional
3863 C would cause serious problems. This is a subset of the possible
3864 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3867 Use of ISO C style function definitions. This warning intentionally is
3868 @emph{not} issued for prototype declarations or variadic functions
3869 because these ISO C features will appear in your code when using
3870 libiberty's traditional C compatibility macros, @code{PARAMS} and
3871 @code{VPARAMS}. This warning is also bypassed for nested functions
3872 because that feature is already a GCC extension and thus not relevant to
3873 traditional C compatibility.
3876 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3877 @opindex Wtraditional-conversion
3878 @opindex Wno-traditional-conversion
3879 Warn if a prototype causes a type conversion that is different from what
3880 would happen to the same argument in the absence of a prototype. This
3881 includes conversions of fixed point to floating and vice versa, and
3882 conversions changing the width or signedness of a fixed point argument
3883 except when the same as the default promotion.
3885 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3886 @opindex Wdeclaration-after-statement
3887 @opindex Wno-declaration-after-statement
3888 Warn when a declaration is found after a statement in a block. This
3889 construct, known from C++, was introduced with ISO C99 and is by default
3890 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3891 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3896 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3898 @item -Wno-endif-labels
3899 @opindex Wno-endif-labels
3900 @opindex Wendif-labels
3901 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3906 Warn whenever a local variable or type declaration shadows another variable,
3907 parameter, type, or class member (in C++), or whenever a built-in function
3908 is shadowed. Note that in C++, the compiler will not warn if a local variable
3909 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3911 @item -Wlarger-than=@var{len}
3912 @opindex Wlarger-than=@var{len}
3913 @opindex Wlarger-than-@var{len}
3914 Warn whenever an object of larger than @var{len} bytes is defined.
3916 @item -Wframe-larger-than=@var{len}
3917 @opindex Wframe-larger-than
3918 Warn if the size of a function frame is larger than @var{len} bytes.
3919 The computation done to determine the stack frame size is approximate
3920 and not conservative.
3921 The actual requirements may be somewhat greater than @var{len}
3922 even if you do not get a warning. In addition, any space allocated
3923 via @code{alloca}, variable-length arrays, or related constructs
3924 is not included by the compiler when determining
3925 whether or not to issue a warning.
3927 @item -Wunsafe-loop-optimizations
3928 @opindex Wunsafe-loop-optimizations
3929 @opindex Wno-unsafe-loop-optimizations
3930 Warn if the loop cannot be optimized because the compiler could not
3931 assume anything on the bounds of the loop indices. With
3932 @option{-funsafe-loop-optimizations} warn if the compiler made
3935 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3936 @opindex Wno-pedantic-ms-format
3937 @opindex Wpedantic-ms-format
3938 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3939 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3940 depending on the MS runtime, when you are using the options @option{-Wformat}
3941 and @option{-pedantic} without gnu-extensions.
3943 @item -Wpointer-arith
3944 @opindex Wpointer-arith
3945 @opindex Wno-pointer-arith
3946 Warn about anything that depends on the ``size of'' a function type or
3947 of @code{void}. GNU C assigns these types a size of 1, for
3948 convenience in calculations with @code{void *} pointers and pointers
3949 to functions. In C++, warn also when an arithmetic operation involves
3950 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3953 @opindex Wtype-limits
3954 @opindex Wno-type-limits
3955 Warn if a comparison is always true or always false due to the limited
3956 range of the data type, but do not warn for constant expressions. For
3957 example, warn if an unsigned variable is compared against zero with
3958 @samp{<} or @samp{>=}. This warning is also enabled by
3961 @item -Wbad-function-cast @r{(C and Objective-C only)}
3962 @opindex Wbad-function-cast
3963 @opindex Wno-bad-function-cast
3964 Warn whenever a function call is cast to a non-matching type.
3965 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3967 @item -Wc++-compat @r{(C and Objective-C only)}
3968 Warn about ISO C constructs that are outside of the common subset of
3969 ISO C and ISO C++, e.g.@: request for implicit conversion from
3970 @code{void *} to a pointer to non-@code{void} type.
3972 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3973 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3974 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3975 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3979 @opindex Wno-cast-qual
3980 Warn whenever a pointer is cast so as to remove a type qualifier from
3981 the target type. For example, warn if a @code{const char *} is cast
3982 to an ordinary @code{char *}.
3984 Also warn when making a cast which introduces a type qualifier in an
3985 unsafe way. For example, casting @code{char **} to @code{const char **}
3986 is unsafe, as in this example:
3989 /* p is char ** value. */
3990 const char **q = (const char **) p;
3991 /* Assignment of readonly string to const char * is OK. */
3993 /* Now char** pointer points to read-only memory. */
3998 @opindex Wcast-align
3999 @opindex Wno-cast-align
4000 Warn whenever a pointer is cast such that the required alignment of the
4001 target is increased. For example, warn if a @code{char *} is cast to
4002 an @code{int *} on machines where integers can only be accessed at
4003 two- or four-byte boundaries.
4005 @item -Wwrite-strings
4006 @opindex Wwrite-strings
4007 @opindex Wno-write-strings
4008 When compiling C, give string constants the type @code{const
4009 char[@var{length}]} so that copying the address of one into a
4010 non-@code{const} @code{char *} pointer will get a warning. These
4011 warnings will help you find at compile time code that can try to write
4012 into a string constant, but only if you have been very careful about
4013 using @code{const} in declarations and prototypes. Otherwise, it will
4014 just be a nuisance. This is why we did not make @option{-Wall} request
4017 When compiling C++, warn about the deprecated conversion from string
4018 literals to @code{char *}. This warning is enabled by default for C++
4023 @opindex Wno-clobbered
4024 Warn for variables that might be changed by @samp{longjmp} or
4025 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4028 @opindex Wconversion
4029 @opindex Wno-conversion
4030 Warn for implicit conversions that may alter a value. This includes
4031 conversions between real and integer, like @code{abs (x)} when
4032 @code{x} is @code{double}; conversions between signed and unsigned,
4033 like @code{unsigned ui = -1}; and conversions to smaller types, like
4034 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4035 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4036 changed by the conversion like in @code{abs (2.0)}. Warnings about
4037 conversions between signed and unsigned integers can be disabled by
4038 using @option{-Wno-sign-conversion}.
4040 For C++, also warn for confusing overload resolution for user-defined
4041 conversions; and conversions that will never use a type conversion
4042 operator: conversions to @code{void}, the same type, a base class or a
4043 reference to them. Warnings about conversions between signed and
4044 unsigned integers are disabled by default in C++ unless
4045 @option{-Wsign-conversion} is explicitly enabled.
4047 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4048 @opindex Wconversion-null
4049 @opindex Wno-conversion-null
4050 Do not warn for conversions between @code{NULL} and non-pointer
4051 types. @option{-Wconversion-null} is enabled by default.
4054 @opindex Wempty-body
4055 @opindex Wno-empty-body
4056 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4057 while} statement. This warning is also enabled by @option{-Wextra}.
4059 @item -Wenum-compare
4060 @opindex Wenum-compare
4061 @opindex Wno-enum-compare
4062 Warn about a comparison between values of different enum types. In C++
4063 this warning is enabled by default. In C this warning is enabled by
4066 @item -Wjump-misses-init @r{(C, Objective-C only)}
4067 @opindex Wjump-misses-init
4068 @opindex Wno-jump-misses-init
4069 Warn if a @code{goto} statement or a @code{switch} statement jumps
4070 forward across the initialization of a variable, or jumps backward to a
4071 label after the variable has been initialized. This only warns about
4072 variables which are initialized when they are declared. This warning is
4073 only supported for C and Objective C; in C++ this sort of branch is an
4076 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4077 can be disabled with the @option{-Wno-jump-misses-init} option.
4079 @item -Wsign-compare
4080 @opindex Wsign-compare
4081 @opindex Wno-sign-compare
4082 @cindex warning for comparison of signed and unsigned values
4083 @cindex comparison of signed and unsigned values, warning
4084 @cindex signed and unsigned values, comparison warning
4085 Warn when a comparison between signed and unsigned values could produce
4086 an incorrect result when the signed value is converted to unsigned.
4087 This warning is also enabled by @option{-Wextra}; to get the other warnings
4088 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4090 @item -Wsign-conversion
4091 @opindex Wsign-conversion
4092 @opindex Wno-sign-conversion
4093 Warn for implicit conversions that may change the sign of an integer
4094 value, like assigning a signed integer expression to an unsigned
4095 integer variable. An explicit cast silences the warning. In C, this
4096 option is enabled also by @option{-Wconversion}.
4100 @opindex Wno-address
4101 Warn about suspicious uses of memory addresses. These include using
4102 the address of a function in a conditional expression, such as
4103 @code{void func(void); if (func)}, and comparisons against the memory
4104 address of a string literal, such as @code{if (x == "abc")}. Such
4105 uses typically indicate a programmer error: the address of a function
4106 always evaluates to true, so their use in a conditional usually
4107 indicate that the programmer forgot the parentheses in a function
4108 call; and comparisons against string literals result in unspecified
4109 behavior and are not portable in C, so they usually indicate that the
4110 programmer intended to use @code{strcmp}. This warning is enabled by
4114 @opindex Wlogical-op
4115 @opindex Wno-logical-op
4116 Warn about suspicious uses of logical operators in expressions.
4117 This includes using logical operators in contexts where a
4118 bit-wise operator is likely to be expected.
4120 @item -Waggregate-return
4121 @opindex Waggregate-return
4122 @opindex Wno-aggregate-return
4123 Warn if any functions that return structures or unions are defined or
4124 called. (In languages where you can return an array, this also elicits
4127 @item -Wno-attributes
4128 @opindex Wno-attributes
4129 @opindex Wattributes
4130 Do not warn if an unexpected @code{__attribute__} is used, such as
4131 unrecognized attributes, function attributes applied to variables,
4132 etc. This will not stop errors for incorrect use of supported
4135 @item -Wno-builtin-macro-redefined
4136 @opindex Wno-builtin-macro-redefined
4137 @opindex Wbuiltin-macro-redefined
4138 Do not warn if certain built-in macros are redefined. This suppresses
4139 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4140 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4142 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4143 @opindex Wstrict-prototypes
4144 @opindex Wno-strict-prototypes
4145 Warn if a function is declared or defined without specifying the
4146 argument types. (An old-style function definition is permitted without
4147 a warning if preceded by a declaration which specifies the argument
4150 @item -Wold-style-declaration @r{(C and Objective-C only)}
4151 @opindex Wold-style-declaration
4152 @opindex Wno-old-style-declaration
4153 Warn for obsolescent usages, according to the C Standard, in a
4154 declaration. For example, warn if storage-class specifiers like
4155 @code{static} are not the first things in a declaration. This warning
4156 is also enabled by @option{-Wextra}.
4158 @item -Wold-style-definition @r{(C and Objective-C only)}
4159 @opindex Wold-style-definition
4160 @opindex Wno-old-style-definition
4161 Warn if an old-style function definition is used. A warning is given
4162 even if there is a previous prototype.
4164 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4165 @opindex Wmissing-parameter-type
4166 @opindex Wno-missing-parameter-type
4167 A function parameter is declared without a type specifier in K&R-style
4174 This warning is also enabled by @option{-Wextra}.
4176 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4177 @opindex Wmissing-prototypes
4178 @opindex Wno-missing-prototypes
4179 Warn if a global function is defined without a previous prototype
4180 declaration. This warning is issued even if the definition itself
4181 provides a prototype. The aim is to detect global functions that fail
4182 to be declared in header files.
4184 @item -Wmissing-declarations
4185 @opindex Wmissing-declarations
4186 @opindex Wno-missing-declarations
4187 Warn if a global function is defined without a previous declaration.
4188 Do so even if the definition itself provides a prototype.
4189 Use this option to detect global functions that are not declared in
4190 header files. In C++, no warnings are issued for function templates,
4191 or for inline functions, or for functions in anonymous namespaces.
4193 @item -Wmissing-field-initializers
4194 @opindex Wmissing-field-initializers
4195 @opindex Wno-missing-field-initializers
4199 Warn if a structure's initializer has some fields missing. For
4200 example, the following code would cause such a warning, because
4201 @code{x.h} is implicitly zero:
4204 struct s @{ int f, g, h; @};
4205 struct s x = @{ 3, 4 @};
4208 This option does not warn about designated initializers, so the following
4209 modification would not trigger a warning:
4212 struct s @{ int f, g, h; @};
4213 struct s x = @{ .f = 3, .g = 4 @};
4216 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4217 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4219 @item -Wmissing-format-attribute
4220 @opindex Wmissing-format-attribute
4221 @opindex Wno-missing-format-attribute
4224 Warn about function pointers which might be candidates for @code{format}
4225 attributes. Note these are only possible candidates, not absolute ones.
4226 GCC will guess that function pointers with @code{format} attributes that
4227 are used in assignment, initialization, parameter passing or return
4228 statements should have a corresponding @code{format} attribute in the
4229 resulting type. I.e.@: the left-hand side of the assignment or
4230 initialization, the type of the parameter variable, or the return type
4231 of the containing function respectively should also have a @code{format}
4232 attribute to avoid the warning.
4234 GCC will also warn about function definitions which might be
4235 candidates for @code{format} attributes. Again, these are only
4236 possible candidates. GCC will guess that @code{format} attributes
4237 might be appropriate for any function that calls a function like
4238 @code{vprintf} or @code{vscanf}, but this might not always be the
4239 case, and some functions for which @code{format} attributes are
4240 appropriate may not be detected.
4242 @item -Wno-multichar
4243 @opindex Wno-multichar
4245 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4246 Usually they indicate a typo in the user's code, as they have
4247 implementation-defined values, and should not be used in portable code.
4249 @item -Wnormalized=<none|id|nfc|nfkc>
4250 @opindex Wnormalized=
4253 @cindex character set, input normalization
4254 In ISO C and ISO C++, two identifiers are different if they are
4255 different sequences of characters. However, sometimes when characters
4256 outside the basic ASCII character set are used, you can have two
4257 different character sequences that look the same. To avoid confusion,
4258 the ISO 10646 standard sets out some @dfn{normalization rules} which
4259 when applied ensure that two sequences that look the same are turned into
4260 the same sequence. GCC can warn you if you are using identifiers which
4261 have not been normalized; this option controls that warning.
4263 There are four levels of warning that GCC supports. The default is
4264 @option{-Wnormalized=nfc}, which warns about any identifier which is
4265 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4266 recommended form for most uses.
4268 Unfortunately, there are some characters which ISO C and ISO C++ allow
4269 in identifiers that when turned into NFC aren't allowable as
4270 identifiers. That is, there's no way to use these symbols in portable
4271 ISO C or C++ and have all your identifiers in NFC@.
4272 @option{-Wnormalized=id} suppresses the warning for these characters.
4273 It is hoped that future versions of the standards involved will correct
4274 this, which is why this option is not the default.
4276 You can switch the warning off for all characters by writing
4277 @option{-Wnormalized=none}. You would only want to do this if you
4278 were using some other normalization scheme (like ``D''), because
4279 otherwise you can easily create bugs that are literally impossible to see.
4281 Some characters in ISO 10646 have distinct meanings but look identical
4282 in some fonts or display methodologies, especially once formatting has
4283 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4284 LETTER N'', will display just like a regular @code{n} which has been
4285 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4286 normalization scheme to convert all these into a standard form as
4287 well, and GCC will warn if your code is not in NFKC if you use
4288 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4289 about every identifier that contains the letter O because it might be
4290 confused with the digit 0, and so is not the default, but may be
4291 useful as a local coding convention if the programming environment is
4292 unable to be fixed to display these characters distinctly.
4294 @item -Wno-deprecated
4295 @opindex Wno-deprecated
4296 @opindex Wdeprecated
4297 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4299 @item -Wno-deprecated-declarations
4300 @opindex Wno-deprecated-declarations
4301 @opindex Wdeprecated-declarations
4302 Do not warn about uses of functions (@pxref{Function Attributes}),
4303 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4304 Attributes}) marked as deprecated by using the @code{deprecated}
4308 @opindex Wno-overflow
4310 Do not warn about compile-time overflow in constant expressions.
4312 @item -Woverride-init @r{(C and Objective-C only)}
4313 @opindex Woverride-init
4314 @opindex Wno-override-init
4318 Warn if an initialized field without side effects is overridden when
4319 using designated initializers (@pxref{Designated Inits, , Designated
4322 This warning is included in @option{-Wextra}. To get other
4323 @option{-Wextra} warnings without this one, use @samp{-Wextra
4324 -Wno-override-init}.
4329 Warn if a structure is given the packed attribute, but the packed
4330 attribute has no effect on the layout or size of the structure.
4331 Such structures may be mis-aligned for little benefit. For
4332 instance, in this code, the variable @code{f.x} in @code{struct bar}
4333 will be misaligned even though @code{struct bar} does not itself
4334 have the packed attribute:
4341 @} __attribute__((packed));
4349 @item -Wpacked-bitfield-compat
4350 @opindex Wpacked-bitfield-compat
4351 @opindex Wno-packed-bitfield-compat
4352 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4353 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4354 the change can lead to differences in the structure layout. GCC
4355 informs you when the offset of such a field has changed in GCC 4.4.
4356 For example there is no longer a 4-bit padding between field @code{a}
4357 and @code{b} in this structure:
4364 @} __attribute__ ((packed));
4367 This warning is enabled by default. Use
4368 @option{-Wno-packed-bitfield-compat} to disable this warning.
4373 Warn if padding is included in a structure, either to align an element
4374 of the structure or to align the whole structure. Sometimes when this
4375 happens it is possible to rearrange the fields of the structure to
4376 reduce the padding and so make the structure smaller.
4378 @item -Wredundant-decls
4379 @opindex Wredundant-decls
4380 @opindex Wno-redundant-decls
4381 Warn if anything is declared more than once in the same scope, even in
4382 cases where multiple declaration is valid and changes nothing.
4384 @item -Wnested-externs @r{(C and Objective-C only)}
4385 @opindex Wnested-externs
4386 @opindex Wno-nested-externs
4387 Warn if an @code{extern} declaration is encountered within a function.
4392 Warn if a function can not be inlined and it was declared as inline.
4393 Even with this option, the compiler will not warn about failures to
4394 inline functions declared in system headers.
4396 The compiler uses a variety of heuristics to determine whether or not
4397 to inline a function. For example, the compiler takes into account
4398 the size of the function being inlined and the amount of inlining
4399 that has already been done in the current function. Therefore,
4400 seemingly insignificant changes in the source program can cause the
4401 warnings produced by @option{-Winline} to appear or disappear.
4403 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4404 @opindex Wno-invalid-offsetof
4405 @opindex Winvalid-offsetof
4406 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4407 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4408 to a non-POD type is undefined. In existing C++ implementations,
4409 however, @samp{offsetof} typically gives meaningful results even when
4410 applied to certain kinds of non-POD types. (Such as a simple
4411 @samp{struct} that fails to be a POD type only by virtue of having a
4412 constructor.) This flag is for users who are aware that they are
4413 writing nonportable code and who have deliberately chosen to ignore the
4416 The restrictions on @samp{offsetof} may be relaxed in a future version
4417 of the C++ standard.
4419 @item -Wno-int-to-pointer-cast
4420 @opindex Wno-int-to-pointer-cast
4421 @opindex Wint-to-pointer-cast
4422 Suppress warnings from casts to pointer type of an integer of a
4423 different size. In C++, casting to a pointer type of smaller size is
4424 an error. @option{Wint-to-pointer-cast} is enabled by default.
4427 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4428 @opindex Wno-pointer-to-int-cast
4429 @opindex Wpointer-to-int-cast
4430 Suppress warnings from casts from a pointer to an integer type of a
4434 @opindex Winvalid-pch
4435 @opindex Wno-invalid-pch
4436 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4437 the search path but can't be used.
4441 @opindex Wno-long-long
4442 Warn if @samp{long long} type is used. This is enabled by either
4443 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4444 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4446 @item -Wvariadic-macros
4447 @opindex Wvariadic-macros
4448 @opindex Wno-variadic-macros
4449 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4450 alternate syntax when in pedantic ISO C99 mode. This is default.
4451 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4456 Warn if variable length array is used in the code.
4457 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4458 the variable length array.
4460 @item -Wvolatile-register-var
4461 @opindex Wvolatile-register-var
4462 @opindex Wno-volatile-register-var
4463 Warn if a register variable is declared volatile. The volatile
4464 modifier does not inhibit all optimizations that may eliminate reads
4465 and/or writes to register variables. This warning is enabled by
4468 @item -Wdisabled-optimization
4469 @opindex Wdisabled-optimization
4470 @opindex Wno-disabled-optimization
4471 Warn if a requested optimization pass is disabled. This warning does
4472 not generally indicate that there is anything wrong with your code; it
4473 merely indicates that GCC's optimizers were unable to handle the code
4474 effectively. Often, the problem is that your code is too big or too
4475 complex; GCC will refuse to optimize programs when the optimization
4476 itself is likely to take inordinate amounts of time.
4478 @item -Wpointer-sign @r{(C and Objective-C only)}
4479 @opindex Wpointer-sign
4480 @opindex Wno-pointer-sign
4481 Warn for pointer argument passing or assignment with different signedness.
4482 This option is only supported for C and Objective-C@. It is implied by
4483 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4484 @option{-Wno-pointer-sign}.
4486 @item -Wstack-protector
4487 @opindex Wstack-protector
4488 @opindex Wno-stack-protector
4489 This option is only active when @option{-fstack-protector} is active. It
4490 warns about functions that will not be protected against stack smashing.
4493 @opindex Wno-mudflap
4494 Suppress warnings about constructs that cannot be instrumented by
4497 @item -Woverlength-strings
4498 @opindex Woverlength-strings
4499 @opindex Wno-overlength-strings
4500 Warn about string constants which are longer than the ``minimum
4501 maximum'' length specified in the C standard. Modern compilers
4502 generally allow string constants which are much longer than the
4503 standard's minimum limit, but very portable programs should avoid
4504 using longer strings.
4506 The limit applies @emph{after} string constant concatenation, and does
4507 not count the trailing NUL@. In C90, the limit was 509 characters; in
4508 C99, it was raised to 4095. C++98 does not specify a normative
4509 minimum maximum, so we do not diagnose overlength strings in C++@.
4511 This option is implied by @option{-pedantic}, and can be disabled with
4512 @option{-Wno-overlength-strings}.
4514 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4515 @opindex Wunsuffixed-float-constants
4517 GCC will issue a warning for any floating constant that does not have
4518 a suffix. When used together with @option{-Wsystem-headers} it will
4519 warn about such constants in system header files. This can be useful
4520 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4521 from the decimal floating-point extension to C99.
4524 @node Debugging Options
4525 @section Options for Debugging Your Program or GCC
4526 @cindex options, debugging
4527 @cindex debugging information options
4529 GCC has various special options that are used for debugging
4530 either your program or GCC:
4535 Produce debugging information in the operating system's native format
4536 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4539 On most systems that use stabs format, @option{-g} enables use of extra
4540 debugging information that only GDB can use; this extra information
4541 makes debugging work better in GDB but will probably make other debuggers
4543 refuse to read the program. If you want to control for certain whether
4544 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4545 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4547 GCC allows you to use @option{-g} with
4548 @option{-O}. The shortcuts taken by optimized code may occasionally
4549 produce surprising results: some variables you declared may not exist
4550 at all; flow of control may briefly move where you did not expect it;
4551 some statements may not be executed because they compute constant
4552 results or their values were already at hand; some statements may
4553 execute in different places because they were moved out of loops.
4555 Nevertheless it proves possible to debug optimized output. This makes
4556 it reasonable to use the optimizer for programs that might have bugs.
4558 The following options are useful when GCC is generated with the
4559 capability for more than one debugging format.
4563 Produce debugging information for use by GDB@. This means to use the
4564 most expressive format available (DWARF 2, stabs, or the native format
4565 if neither of those are supported), including GDB extensions if at all
4570 Produce debugging information in stabs format (if that is supported),
4571 without GDB extensions. This is the format used by DBX on most BSD
4572 systems. On MIPS, Alpha and System V Release 4 systems this option
4573 produces stabs debugging output which is not understood by DBX or SDB@.
4574 On System V Release 4 systems this option requires the GNU assembler.
4576 @item -feliminate-unused-debug-symbols
4577 @opindex feliminate-unused-debug-symbols
4578 Produce debugging information in stabs format (if that is supported),
4579 for only symbols that are actually used.
4581 @item -femit-class-debug-always
4582 Instead of emitting debugging information for a C++ class in only one
4583 object file, emit it in all object files using the class. This option
4584 should be used only with debuggers that are unable to handle the way GCC
4585 normally emits debugging information for classes because using this
4586 option will increase the size of debugging information by as much as a
4589 @item -fno-debug-types-section
4590 @opindex fno-types-section
4591 @opindex ftypes-section
4592 By default when using Dwarf v4 or higher type DIEs will be put into
4593 their own .debug_types section instead of making them part of the
4594 .debug_info section. It is more efficient to put them in a separate
4595 comdat sections since the linker will then be able to remove duplicates.
4596 But not all dwarf consumers support .debug_types sections yet.
4600 Produce debugging information in stabs format (if that is supported),
4601 using GNU extensions understood only by the GNU debugger (GDB)@. The
4602 use of these extensions is likely to make other debuggers crash or
4603 refuse to read the program.
4607 Produce debugging information in COFF format (if that is supported).
4608 This is the format used by SDB on most System V systems prior to
4613 Produce debugging information in XCOFF format (if that is supported).
4614 This is the format used by the DBX debugger on IBM RS/6000 systems.
4618 Produce debugging information in XCOFF format (if that is supported),
4619 using GNU extensions understood only by the GNU debugger (GDB)@. The
4620 use of these extensions is likely to make other debuggers crash or
4621 refuse to read the program, and may cause assemblers other than the GNU
4622 assembler (GAS) to fail with an error.
4624 @item -gdwarf-@var{version}
4625 @opindex gdwarf-@var{version}
4626 Produce debugging information in DWARF format (if that is
4627 supported). This is the format used by DBX on IRIX 6. The value
4628 of @var{version} may be either 2, 3 or 4; the default version is 2.
4630 Note that with DWARF version 2 some ports require, and will always
4631 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4633 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4634 for maximum benefit.
4636 @item -gstrict-dwarf
4637 @opindex gstrict-dwarf
4638 Disallow using extensions of later DWARF standard version than selected
4639 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4640 DWARF extensions from later standard versions is allowed.
4642 @item -gno-strict-dwarf
4643 @opindex gno-strict-dwarf
4644 Allow using extensions of later DWARF standard version than selected with
4645 @option{-gdwarf-@var{version}}.
4649 Produce debugging information in VMS debug format (if that is
4650 supported). This is the format used by DEBUG on VMS systems.
4653 @itemx -ggdb@var{level}
4654 @itemx -gstabs@var{level}
4655 @itemx -gcoff@var{level}
4656 @itemx -gxcoff@var{level}
4657 @itemx -gvms@var{level}
4658 Request debugging information and also use @var{level} to specify how
4659 much information. The default level is 2.
4661 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4664 Level 1 produces minimal information, enough for making backtraces in
4665 parts of the program that you don't plan to debug. This includes
4666 descriptions of functions and external variables, but no information
4667 about local variables and no line numbers.
4669 Level 3 includes extra information, such as all the macro definitions
4670 present in the program. Some debuggers support macro expansion when
4671 you use @option{-g3}.
4673 @option{-gdwarf-2} does not accept a concatenated debug level, because
4674 GCC used to support an option @option{-gdwarf} that meant to generate
4675 debug information in version 1 of the DWARF format (which is very
4676 different from version 2), and it would have been too confusing. That
4677 debug format is long obsolete, but the option cannot be changed now.
4678 Instead use an additional @option{-g@var{level}} option to change the
4679 debug level for DWARF.
4683 Turn off generation of debug info, if leaving out this option would have
4684 generated it, or turn it on at level 2 otherwise. The position of this
4685 argument in the command line does not matter, it takes effect after all
4686 other options are processed, and it does so only once, no matter how
4687 many times it is given. This is mainly intended to be used with
4688 @option{-fcompare-debug}.
4690 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4691 @opindex fdump-final-insns
4692 Dump the final internal representation (RTL) to @var{file}. If the
4693 optional argument is omitted (or if @var{file} is @code{.}), the name
4694 of the dump file will be determined by appending @code{.gkd} to the
4695 compilation output file name.
4697 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4698 @opindex fcompare-debug
4699 @opindex fno-compare-debug
4700 If no error occurs during compilation, run the compiler a second time,
4701 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4702 passed to the second compilation. Dump the final internal
4703 representation in both compilations, and print an error if they differ.
4705 If the equal sign is omitted, the default @option{-gtoggle} is used.
4707 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4708 and nonzero, implicitly enables @option{-fcompare-debug}. If
4709 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4710 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4713 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4714 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4715 of the final representation and the second compilation, preventing even
4716 @env{GCC_COMPARE_DEBUG} from taking effect.
4718 To verify full coverage during @option{-fcompare-debug} testing, set
4719 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4720 which GCC will reject as an invalid option in any actual compilation
4721 (rather than preprocessing, assembly or linking). To get just a
4722 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4723 not overridden} will do.
4725 @item -fcompare-debug-second
4726 @opindex fcompare-debug-second
4727 This option is implicitly passed to the compiler for the second
4728 compilation requested by @option{-fcompare-debug}, along with options to
4729 silence warnings, and omitting other options that would cause
4730 side-effect compiler outputs to files or to the standard output. Dump
4731 files and preserved temporary files are renamed so as to contain the
4732 @code{.gk} additional extension during the second compilation, to avoid
4733 overwriting those generated by the first.
4735 When this option is passed to the compiler driver, it causes the
4736 @emph{first} compilation to be skipped, which makes it useful for little
4737 other than debugging the compiler proper.
4739 @item -feliminate-dwarf2-dups
4740 @opindex feliminate-dwarf2-dups
4741 Compress DWARF2 debugging information by eliminating duplicated
4742 information about each symbol. This option only makes sense when
4743 generating DWARF2 debugging information with @option{-gdwarf-2}.
4745 @item -femit-struct-debug-baseonly
4746 Emit debug information for struct-like types
4747 only when the base name of the compilation source file
4748 matches the base name of file in which the struct was defined.
4750 This option substantially reduces the size of debugging information,
4751 but at significant potential loss in type information to the debugger.
4752 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4753 See @option{-femit-struct-debug-detailed} for more detailed control.
4755 This option works only with DWARF 2.
4757 @item -femit-struct-debug-reduced
4758 Emit debug information for struct-like types
4759 only when the base name of the compilation source file
4760 matches the base name of file in which the type was defined,
4761 unless the struct is a template or defined in a system header.
4763 This option significantly reduces the size of debugging information,
4764 with some potential loss in type information to the debugger.
4765 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4766 See @option{-femit-struct-debug-detailed} for more detailed control.
4768 This option works only with DWARF 2.
4770 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4771 Specify the struct-like types
4772 for which the compiler will generate debug information.
4773 The intent is to reduce duplicate struct debug information
4774 between different object files within the same program.
4776 This option is a detailed version of
4777 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4778 which will serve for most needs.
4780 A specification has the syntax@*
4781 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4783 The optional first word limits the specification to
4784 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4785 A struct type is used directly when it is the type of a variable, member.
4786 Indirect uses arise through pointers to structs.
4787 That is, when use of an incomplete struct would be legal, the use is indirect.
4789 @samp{struct one direct; struct two * indirect;}.
4791 The optional second word limits the specification to
4792 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4793 Generic structs are a bit complicated to explain.
4794 For C++, these are non-explicit specializations of template classes,
4795 or non-template classes within the above.
4796 Other programming languages have generics,
4797 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4799 The third word specifies the source files for those
4800 structs for which the compiler will emit debug information.
4801 The values @samp{none} and @samp{any} have the normal meaning.
4802 The value @samp{base} means that
4803 the base of name of the file in which the type declaration appears
4804 must match the base of the name of the main compilation file.
4805 In practice, this means that
4806 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4807 but types declared in other header will not.
4808 The value @samp{sys} means those types satisfying @samp{base}
4809 or declared in system or compiler headers.
4811 You may need to experiment to determine the best settings for your application.
4813 The default is @samp{-femit-struct-debug-detailed=all}.
4815 This option works only with DWARF 2.
4817 @item -fno-merge-debug-strings
4818 @opindex fmerge-debug-strings
4819 @opindex fno-merge-debug-strings
4820 Direct the linker to not merge together strings in the debugging
4821 information which are identical in different object files. Merging is
4822 not supported by all assemblers or linkers. Merging decreases the size
4823 of the debug information in the output file at the cost of increasing
4824 link processing time. Merging is enabled by default.
4826 @item -fdebug-prefix-map=@var{old}=@var{new}
4827 @opindex fdebug-prefix-map
4828 When compiling files in directory @file{@var{old}}, record debugging
4829 information describing them as in @file{@var{new}} instead.
4831 @item -fno-dwarf2-cfi-asm
4832 @opindex fdwarf2-cfi-asm
4833 @opindex fno-dwarf2-cfi-asm
4834 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4835 instead of using GAS @code{.cfi_*} directives.
4837 @cindex @command{prof}
4840 Generate extra code to write profile information suitable for the
4841 analysis program @command{prof}. You must use this option when compiling
4842 the source files you want data about, and you must also use it when
4845 @cindex @command{gprof}
4848 Generate extra code to write profile information suitable for the
4849 analysis program @command{gprof}. You must use this option when compiling
4850 the source files you want data about, and you must also use it when
4855 Makes the compiler print out each function name as it is compiled, and
4856 print some statistics about each pass when it finishes.
4859 @opindex ftime-report
4860 Makes the compiler print some statistics about the time consumed by each
4861 pass when it finishes.
4864 @opindex fmem-report
4865 Makes the compiler print some statistics about permanent memory
4866 allocation when it finishes.
4868 @item -fpre-ipa-mem-report
4869 @opindex fpre-ipa-mem-report
4870 @item -fpost-ipa-mem-report
4871 @opindex fpost-ipa-mem-report
4872 Makes the compiler print some statistics about permanent memory
4873 allocation before or after interprocedural optimization.
4876 @opindex fstack-usage
4877 Makes the compiler output stack usage information for the program, on a
4878 per-function basis. The filename for the dump is made by appending
4879 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4880 the output file, if explicitly specified and it is not an executable,
4881 otherwise it is the basename of the source file. An entry is made up
4886 The name of the function.
4890 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4893 The qualifier @code{static} means that the function manipulates the stack
4894 statically: a fixed number of bytes are allocated for the frame on function
4895 entry and released on function exit; no stack adjustments are otherwise made
4896 in the function. The second field is this fixed number of bytes.
4898 The qualifier @code{dynamic} means that the function manipulates the stack
4899 dynamically: in addition to the static allocation described above, stack
4900 adjustments are made in the body of the function, for example to push/pop
4901 arguments around function calls. If the qualifier @code{bounded} is also
4902 present, the amount of these adjustments is bounded at compile-time and
4903 the second field is an upper bound of the total amount of stack used by
4904 the function. If it is not present, the amount of these adjustments is
4905 not bounded at compile-time and the second field only represents the
4908 @item -fprofile-arcs
4909 @opindex fprofile-arcs
4910 Add code so that program flow @dfn{arcs} are instrumented. During
4911 execution the program records how many times each branch and call is
4912 executed and how many times it is taken or returns. When the compiled
4913 program exits it saves this data to a file called
4914 @file{@var{auxname}.gcda} for each source file. The data may be used for
4915 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4916 test coverage analysis (@option{-ftest-coverage}). Each object file's
4917 @var{auxname} is generated from the name of the output file, if
4918 explicitly specified and it is not the final executable, otherwise it is
4919 the basename of the source file. In both cases any suffix is removed
4920 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4921 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4922 @xref{Cross-profiling}.
4924 @cindex @command{gcov}
4928 This option is used to compile and link code instrumented for coverage
4929 analysis. The option is a synonym for @option{-fprofile-arcs}
4930 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4931 linking). See the documentation for those options for more details.
4936 Compile the source files with @option{-fprofile-arcs} plus optimization
4937 and code generation options. For test coverage analysis, use the
4938 additional @option{-ftest-coverage} option. You do not need to profile
4939 every source file in a program.
4942 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4943 (the latter implies the former).
4946 Run the program on a representative workload to generate the arc profile
4947 information. This may be repeated any number of times. You can run
4948 concurrent instances of your program, and provided that the file system
4949 supports locking, the data files will be correctly updated. Also
4950 @code{fork} calls are detected and correctly handled (double counting
4954 For profile-directed optimizations, compile the source files again with
4955 the same optimization and code generation options plus
4956 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4957 Control Optimization}).
4960 For test coverage analysis, use @command{gcov} to produce human readable
4961 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4962 @command{gcov} documentation for further information.
4966 With @option{-fprofile-arcs}, for each function of your program GCC
4967 creates a program flow graph, then finds a spanning tree for the graph.
4968 Only arcs that are not on the spanning tree have to be instrumented: the
4969 compiler adds code to count the number of times that these arcs are
4970 executed. When an arc is the only exit or only entrance to a block, the
4971 instrumentation code can be added to the block; otherwise, a new basic
4972 block must be created to hold the instrumentation code.
4975 @item -ftest-coverage
4976 @opindex ftest-coverage
4977 Produce a notes file that the @command{gcov} code-coverage utility
4978 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4979 show program coverage. Each source file's note file is called
4980 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4981 above for a description of @var{auxname} and instructions on how to
4982 generate test coverage data. Coverage data will match the source files
4983 more closely, if you do not optimize.
4985 @item -fdbg-cnt-list
4986 @opindex fdbg-cnt-list
4987 Print the name and the counter upper bound for all debug counters.
4989 @item -fdbg-cnt=@var{counter-value-list}
4991 Set the internal debug counter upper bound. @var{counter-value-list}
4992 is a comma-separated list of @var{name}:@var{value} pairs
4993 which sets the upper bound of each debug counter @var{name} to @var{value}.
4994 All debug counters have the initial upper bound of @var{UINT_MAX},
4995 thus dbg_cnt() returns true always unless the upper bound is set by this option.
4996 e.g. With -fdbg-cnt=dce:10,tail_call:0
4997 dbg_cnt(dce) will return true only for first 10 invocations
4998 and dbg_cnt(tail_call) will return false always.
5000 @item -d@var{letters}
5001 @itemx -fdump-rtl-@var{pass}
5003 Says to make debugging dumps during compilation at times specified by
5004 @var{letters}. This is used for debugging the RTL-based passes of the
5005 compiler. The file names for most of the dumps are made by appending
5006 a pass number and a word to the @var{dumpname}, and the files are
5007 created in the directory of the output file. Note that the pass
5008 number is computed statically as passes get registered into the pass
5009 manager. Thus the numbering is not related to the dynamic order of
5010 execution of passes. In particular, a pass installed by a plugin
5011 could have a number over 200 even if it executed quite early.
5012 @var{dumpname} is generated from the name of the output file, if
5013 explicitly specified and it is not an executable, otherwise it is the
5014 basename of the source file. These switches may have different effects
5015 when @option{-E} is used for preprocessing.
5017 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5018 @option{-d} option @var{letters}. Here are the possible
5019 letters for use in @var{pass} and @var{letters}, and their meanings:
5023 @item -fdump-rtl-alignments
5024 @opindex fdump-rtl-alignments
5025 Dump after branch alignments have been computed.
5027 @item -fdump-rtl-asmcons
5028 @opindex fdump-rtl-asmcons
5029 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5031 @item -fdump-rtl-auto_inc_dec
5032 @opindex fdump-rtl-auto_inc_dec
5033 Dump after auto-inc-dec discovery. This pass is only run on
5034 architectures that have auto inc or auto dec instructions.
5036 @item -fdump-rtl-barriers
5037 @opindex fdump-rtl-barriers
5038 Dump after cleaning up the barrier instructions.
5040 @item -fdump-rtl-bbpart
5041 @opindex fdump-rtl-bbpart
5042 Dump after partitioning hot and cold basic blocks.
5044 @item -fdump-rtl-bbro
5045 @opindex fdump-rtl-bbro
5046 Dump after block reordering.
5048 @item -fdump-rtl-btl1
5049 @itemx -fdump-rtl-btl2
5050 @opindex fdump-rtl-btl2
5051 @opindex fdump-rtl-btl2
5052 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5053 after the two branch
5054 target load optimization passes.
5056 @item -fdump-rtl-bypass
5057 @opindex fdump-rtl-bypass
5058 Dump after jump bypassing and control flow optimizations.
5060 @item -fdump-rtl-combine
5061 @opindex fdump-rtl-combine
5062 Dump after the RTL instruction combination pass.
5064 @item -fdump-rtl-compgotos
5065 @opindex fdump-rtl-compgotos
5066 Dump after duplicating the computed gotos.
5068 @item -fdump-rtl-ce1
5069 @itemx -fdump-rtl-ce2
5070 @itemx -fdump-rtl-ce3
5071 @opindex fdump-rtl-ce1
5072 @opindex fdump-rtl-ce2
5073 @opindex fdump-rtl-ce3
5074 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5075 @option{-fdump-rtl-ce3} enable dumping after the three
5076 if conversion passes.
5078 @itemx -fdump-rtl-cprop_hardreg
5079 @opindex fdump-rtl-cprop_hardreg
5080 Dump after hard register copy propagation.
5082 @itemx -fdump-rtl-csa
5083 @opindex fdump-rtl-csa
5084 Dump after combining stack adjustments.
5086 @item -fdump-rtl-cse1
5087 @itemx -fdump-rtl-cse2
5088 @opindex fdump-rtl-cse1
5089 @opindex fdump-rtl-cse2
5090 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5091 the two common sub-expression elimination passes.
5093 @itemx -fdump-rtl-dce
5094 @opindex fdump-rtl-dce
5095 Dump after the standalone dead code elimination passes.
5097 @itemx -fdump-rtl-dbr
5098 @opindex fdump-rtl-dbr
5099 Dump after delayed branch scheduling.
5101 @item -fdump-rtl-dce1
5102 @itemx -fdump-rtl-dce2
5103 @opindex fdump-rtl-dce1
5104 @opindex fdump-rtl-dce2
5105 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5106 the two dead store elimination passes.
5109 @opindex fdump-rtl-eh
5110 Dump after finalization of EH handling code.
5112 @item -fdump-rtl-eh_ranges
5113 @opindex fdump-rtl-eh_ranges
5114 Dump after conversion of EH handling range regions.
5116 @item -fdump-rtl-expand
5117 @opindex fdump-rtl-expand
5118 Dump after RTL generation.
5120 @item -fdump-rtl-fwprop1
5121 @itemx -fdump-rtl-fwprop2
5122 @opindex fdump-rtl-fwprop1
5123 @opindex fdump-rtl-fwprop2
5124 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5125 dumping after the two forward propagation passes.
5127 @item -fdump-rtl-gcse1
5128 @itemx -fdump-rtl-gcse2
5129 @opindex fdump-rtl-gcse1
5130 @opindex fdump-rtl-gcse2
5131 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5132 after global common subexpression elimination.
5134 @item -fdump-rtl-init-regs
5135 @opindex fdump-rtl-init-regs
5136 Dump after the initialization of the registers.
5138 @item -fdump-rtl-initvals
5139 @opindex fdump-rtl-initvals
5140 Dump after the computation of the initial value sets.
5142 @itemx -fdump-rtl-into_cfglayout
5143 @opindex fdump-rtl-into_cfglayout
5144 Dump after converting to cfglayout mode.
5146 @item -fdump-rtl-ira
5147 @opindex fdump-rtl-ira
5148 Dump after iterated register allocation.
5150 @item -fdump-rtl-jump
5151 @opindex fdump-rtl-jump
5152 Dump after the second jump optimization.
5154 @item -fdump-rtl-loop2
5155 @opindex fdump-rtl-loop2
5156 @option{-fdump-rtl-loop2} enables dumping after the rtl
5157 loop optimization passes.
5159 @item -fdump-rtl-mach
5160 @opindex fdump-rtl-mach
5161 Dump after performing the machine dependent reorganization pass, if that
5164 @item -fdump-rtl-mode_sw
5165 @opindex fdump-rtl-mode_sw
5166 Dump after removing redundant mode switches.
5168 @item -fdump-rtl-rnreg
5169 @opindex fdump-rtl-rnreg
5170 Dump after register renumbering.
5172 @itemx -fdump-rtl-outof_cfglayout
5173 @opindex fdump-rtl-outof_cfglayout
5174 Dump after converting from cfglayout mode.
5176 @item -fdump-rtl-peephole2
5177 @opindex fdump-rtl-peephole2
5178 Dump after the peephole pass.
5180 @item -fdump-rtl-postreload
5181 @opindex fdump-rtl-postreload
5182 Dump after post-reload optimizations.
5184 @itemx -fdump-rtl-pro_and_epilogue
5185 @opindex fdump-rtl-pro_and_epilogue
5186 Dump after generating the function pro and epilogues.
5188 @item -fdump-rtl-regmove
5189 @opindex fdump-rtl-regmove
5190 Dump after the register move pass.
5192 @item -fdump-rtl-sched1
5193 @itemx -fdump-rtl-sched2
5194 @opindex fdump-rtl-sched1
5195 @opindex fdump-rtl-sched2
5196 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5197 after the basic block scheduling passes.
5199 @item -fdump-rtl-see
5200 @opindex fdump-rtl-see
5201 Dump after sign extension elimination.
5203 @item -fdump-rtl-seqabstr
5204 @opindex fdump-rtl-seqabstr
5205 Dump after common sequence discovery.
5207 @item -fdump-rtl-shorten
5208 @opindex fdump-rtl-shorten
5209 Dump after shortening branches.
5211 @item -fdump-rtl-sibling
5212 @opindex fdump-rtl-sibling
5213 Dump after sibling call optimizations.
5215 @item -fdump-rtl-split1
5216 @itemx -fdump-rtl-split2
5217 @itemx -fdump-rtl-split3
5218 @itemx -fdump-rtl-split4
5219 @itemx -fdump-rtl-split5
5220 @opindex fdump-rtl-split1
5221 @opindex fdump-rtl-split2
5222 @opindex fdump-rtl-split3
5223 @opindex fdump-rtl-split4
5224 @opindex fdump-rtl-split5
5225 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5226 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5227 @option{-fdump-rtl-split5} enable dumping after five rounds of
5228 instruction splitting.
5230 @item -fdump-rtl-sms
5231 @opindex fdump-rtl-sms
5232 Dump after modulo scheduling. This pass is only run on some
5235 @item -fdump-rtl-stack
5236 @opindex fdump-rtl-stack
5237 Dump after conversion from GCC's "flat register file" registers to the
5238 x87's stack-like registers. This pass is only run on x86 variants.
5240 @item -fdump-rtl-subreg1
5241 @itemx -fdump-rtl-subreg2
5242 @opindex fdump-rtl-subreg1
5243 @opindex fdump-rtl-subreg2
5244 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5245 the two subreg expansion passes.
5247 @item -fdump-rtl-unshare
5248 @opindex fdump-rtl-unshare
5249 Dump after all rtl has been unshared.
5251 @item -fdump-rtl-vartrack
5252 @opindex fdump-rtl-vartrack
5253 Dump after variable tracking.
5255 @item -fdump-rtl-vregs
5256 @opindex fdump-rtl-vregs
5257 Dump after converting virtual registers to hard registers.
5259 @item -fdump-rtl-web
5260 @opindex fdump-rtl-web
5261 Dump after live range splitting.
5263 @item -fdump-rtl-regclass
5264 @itemx -fdump-rtl-subregs_of_mode_init
5265 @itemx -fdump-rtl-subregs_of_mode_finish
5266 @itemx -fdump-rtl-dfinit
5267 @itemx -fdump-rtl-dfinish
5268 @opindex fdump-rtl-regclass
5269 @opindex fdump-rtl-subregs_of_mode_init
5270 @opindex fdump-rtl-subregs_of_mode_finish
5271 @opindex fdump-rtl-dfinit
5272 @opindex fdump-rtl-dfinish
5273 These dumps are defined but always produce empty files.
5275 @item -fdump-rtl-all
5276 @opindex fdump-rtl-all
5277 Produce all the dumps listed above.
5281 Annotate the assembler output with miscellaneous debugging information.
5285 Dump all macro definitions, at the end of preprocessing, in addition to
5290 Produce a core dump whenever an error occurs.
5294 Print statistics on memory usage, at the end of the run, to
5299 Annotate the assembler output with a comment indicating which
5300 pattern and alternative was used. The length of each instruction is
5305 Dump the RTL in the assembler output as a comment before each instruction.
5306 Also turns on @option{-dp} annotation.
5310 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5311 dump a representation of the control flow graph suitable for viewing with VCG
5312 to @file{@var{file}.@var{pass}.vcg}.
5316 Just generate RTL for a function instead of compiling it. Usually used
5317 with @option{-fdump-rtl-expand}.
5321 @opindex fdump-noaddr
5322 When doing debugging dumps, suppress address output. This makes it more
5323 feasible to use diff on debugging dumps for compiler invocations with
5324 different compiler binaries and/or different
5325 text / bss / data / heap / stack / dso start locations.
5327 @item -fdump-unnumbered
5328 @opindex fdump-unnumbered
5329 When doing debugging dumps, suppress instruction numbers and address output.
5330 This makes it more feasible to use diff on debugging dumps for compiler
5331 invocations with different options, in particular with and without
5334 @item -fdump-unnumbered-links
5335 @opindex fdump-unnumbered-links
5336 When doing debugging dumps (see @option{-d} option above), suppress
5337 instruction numbers for the links to the previous and next instructions
5340 @item -fdump-translation-unit @r{(C++ only)}
5341 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5342 @opindex fdump-translation-unit
5343 Dump a representation of the tree structure for the entire translation
5344 unit to a file. The file name is made by appending @file{.tu} to the
5345 source file name, and the file is created in the same directory as the
5346 output file. If the @samp{-@var{options}} form is used, @var{options}
5347 controls the details of the dump as described for the
5348 @option{-fdump-tree} options.
5350 @item -fdump-class-hierarchy @r{(C++ only)}
5351 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5352 @opindex fdump-class-hierarchy
5353 Dump a representation of each class's hierarchy and virtual function
5354 table layout to a file. The file name is made by appending
5355 @file{.class} to the source file name, and the file is created in the
5356 same directory as the output file. If the @samp{-@var{options}} form
5357 is used, @var{options} controls the details of the dump as described
5358 for the @option{-fdump-tree} options.
5360 @item -fdump-ipa-@var{switch}
5362 Control the dumping at various stages of inter-procedural analysis
5363 language tree to a file. The file name is generated by appending a
5364 switch specific suffix to the source file name, and the file is created
5365 in the same directory as the output file. The following dumps are
5370 Enables all inter-procedural analysis dumps.
5373 Dumps information about call-graph optimization, unused function removal,
5374 and inlining decisions.
5377 Dump after function inlining.
5381 @item -fdump-statistics-@var{option}
5382 @opindex fdump-statistics
5383 Enable and control dumping of pass statistics in a separate file. The
5384 file name is generated by appending a suffix ending in
5385 @samp{.statistics} to the source file name, and the file is created in
5386 the same directory as the output file. If the @samp{-@var{option}}
5387 form is used, @samp{-stats} will cause counters to be summed over the
5388 whole compilation unit while @samp{-details} will dump every event as
5389 the passes generate them. The default with no option is to sum
5390 counters for each function compiled.
5392 @item -fdump-tree-@var{switch}
5393 @itemx -fdump-tree-@var{switch}-@var{options}
5395 Control the dumping at various stages of processing the intermediate
5396 language tree to a file. The file name is generated by appending a
5397 switch specific suffix to the source file name, and the file is
5398 created in the same directory as the output file. If the
5399 @samp{-@var{options}} form is used, @var{options} is a list of
5400 @samp{-} separated options that control the details of the dump. Not
5401 all options are applicable to all dumps, those which are not
5402 meaningful will be ignored. The following options are available
5406 Print the address of each node. Usually this is not meaningful as it
5407 changes according to the environment and source file. Its primary use
5408 is for tying up a dump file with a debug environment.
5410 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5411 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5412 use working backward from mangled names in the assembly file.
5414 Inhibit dumping of members of a scope or body of a function merely
5415 because that scope has been reached. Only dump such items when they
5416 are directly reachable by some other path. When dumping pretty-printed
5417 trees, this option inhibits dumping the bodies of control structures.
5419 Print a raw representation of the tree. By default, trees are
5420 pretty-printed into a C-like representation.
5422 Enable more detailed dumps (not honored by every dump option).
5424 Enable dumping various statistics about the pass (not honored by every dump
5427 Enable showing basic block boundaries (disabled in raw dumps).
5429 Enable showing virtual operands for every statement.
5431 Enable showing line numbers for statements.
5433 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5435 Enable showing the tree dump for each statement.
5437 Enable showing the EH region number holding each statement.
5439 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5440 and @option{lineno}.
5443 The following tree dumps are possible:
5447 @opindex fdump-tree-original
5448 Dump before any tree based optimization, to @file{@var{file}.original}.
5451 @opindex fdump-tree-optimized
5452 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5455 @opindex fdump-tree-gimple
5456 Dump each function before and after the gimplification pass to a file. The
5457 file name is made by appending @file{.gimple} to the source file name.
5460 @opindex fdump-tree-cfg
5461 Dump the control flow graph of each function to a file. The file name is
5462 made by appending @file{.cfg} to the source file name.
5465 @opindex fdump-tree-vcg
5466 Dump the control flow graph of each function to a file in VCG format. The
5467 file name is made by appending @file{.vcg} to the source file name. Note
5468 that if the file contains more than one function, the generated file cannot
5469 be used directly by VCG@. You will need to cut and paste each function's
5470 graph into its own separate file first.
5473 @opindex fdump-tree-ch
5474 Dump each function after copying loop headers. The file name is made by
5475 appending @file{.ch} to the source file name.
5478 @opindex fdump-tree-ssa
5479 Dump SSA related information to a file. The file name is made by appending
5480 @file{.ssa} to the source file name.
5483 @opindex fdump-tree-alias
5484 Dump aliasing information for each function. The file name is made by
5485 appending @file{.alias} to the source file name.
5488 @opindex fdump-tree-ccp
5489 Dump each function after CCP@. The file name is made by appending
5490 @file{.ccp} to the source file name.
5493 @opindex fdump-tree-storeccp
5494 Dump each function after STORE-CCP@. The file name is made by appending
5495 @file{.storeccp} to the source file name.
5498 @opindex fdump-tree-pre
5499 Dump trees after partial redundancy elimination. The file name is made
5500 by appending @file{.pre} to the source file name.
5503 @opindex fdump-tree-fre
5504 Dump trees after full redundancy elimination. The file name is made
5505 by appending @file{.fre} to the source file name.
5508 @opindex fdump-tree-copyprop
5509 Dump trees after copy propagation. The file name is made
5510 by appending @file{.copyprop} to the source file name.
5512 @item store_copyprop
5513 @opindex fdump-tree-store_copyprop
5514 Dump trees after store copy-propagation. The file name is made
5515 by appending @file{.store_copyprop} to the source file name.
5518 @opindex fdump-tree-dce
5519 Dump each function after dead code elimination. The file name is made by
5520 appending @file{.dce} to the source file name.
5523 @opindex fdump-tree-mudflap
5524 Dump each function after adding mudflap instrumentation. The file name is
5525 made by appending @file{.mudflap} to the source file name.
5528 @opindex fdump-tree-sra
5529 Dump each function after performing scalar replacement of aggregates. The
5530 file name is made by appending @file{.sra} to the source file name.
5533 @opindex fdump-tree-sink
5534 Dump each function after performing code sinking. The file name is made
5535 by appending @file{.sink} to the source file name.
5538 @opindex fdump-tree-dom
5539 Dump each function after applying dominator tree optimizations. The file
5540 name is made by appending @file{.dom} to the source file name.
5543 @opindex fdump-tree-dse
5544 Dump each function after applying dead store elimination. The file
5545 name is made by appending @file{.dse} to the source file name.
5548 @opindex fdump-tree-phiopt
5549 Dump each function after optimizing PHI nodes into straightline code. The file
5550 name is made by appending @file{.phiopt} to the source file name.
5553 @opindex fdump-tree-forwprop
5554 Dump each function after forward propagating single use variables. The file
5555 name is made by appending @file{.forwprop} to the source file name.
5558 @opindex fdump-tree-copyrename
5559 Dump each function after applying the copy rename optimization. The file
5560 name is made by appending @file{.copyrename} to the source file name.
5563 @opindex fdump-tree-nrv
5564 Dump each function after applying the named return value optimization on
5565 generic trees. The file name is made by appending @file{.nrv} to the source
5569 @opindex fdump-tree-vect
5570 Dump each function after applying vectorization of loops. The file name is
5571 made by appending @file{.vect} to the source file name.
5574 @opindex fdump-tree-slp
5575 Dump each function after applying vectorization of basic blocks. The file name
5576 is made by appending @file{.slp} to the source file name.
5579 @opindex fdump-tree-vrp
5580 Dump each function after Value Range Propagation (VRP). The file name
5581 is made by appending @file{.vrp} to the source file name.
5584 @opindex fdump-tree-all
5585 Enable all the available tree dumps with the flags provided in this option.
5588 @item -ftree-vectorizer-verbose=@var{n}
5589 @opindex ftree-vectorizer-verbose
5590 This option controls the amount of debugging output the vectorizer prints.
5591 This information is written to standard error, unless
5592 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5593 in which case it is output to the usual dump listing file, @file{.vect}.
5594 For @var{n}=0 no diagnostic information is reported.
5595 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5596 and the total number of loops that got vectorized.
5597 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5598 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5599 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5600 level that @option{-fdump-tree-vect-stats} uses.
5601 Higher verbosity levels mean either more information dumped for each
5602 reported loop, or same amount of information reported for more loops:
5603 if @var{n}=3, vectorizer cost model information is reported.
5604 If @var{n}=4, alignment related information is added to the reports.
5605 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5606 memory access-patterns) is added to the reports.
5607 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5608 that did not pass the first analysis phase (i.e., may not be countable, or
5609 may have complicated control-flow).
5610 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5611 If @var{n}=8, SLP related information is added to the reports.
5612 For @var{n}=9, all the information the vectorizer generates during its
5613 analysis and transformation is reported. This is the same verbosity level
5614 that @option{-fdump-tree-vect-details} uses.
5616 @item -frandom-seed=@var{string}
5617 @opindex frandom-seed
5618 This option provides a seed that GCC uses when it would otherwise use
5619 random numbers. It is used to generate certain symbol names
5620 that have to be different in every compiled file. It is also used to
5621 place unique stamps in coverage data files and the object files that
5622 produce them. You can use the @option{-frandom-seed} option to produce
5623 reproducibly identical object files.
5625 The @var{string} should be different for every file you compile.
5627 @item -fsched-verbose=@var{n}
5628 @opindex fsched-verbose
5629 On targets that use instruction scheduling, this option controls the
5630 amount of debugging output the scheduler prints. This information is
5631 written to standard error, unless @option{-fdump-rtl-sched1} or
5632 @option{-fdump-rtl-sched2} is specified, in which case it is output
5633 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5634 respectively. However for @var{n} greater than nine, the output is
5635 always printed to standard error.
5637 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5638 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5639 For @var{n} greater than one, it also output basic block probabilities,
5640 detailed ready list information and unit/insn info. For @var{n} greater
5641 than two, it includes RTL at abort point, control-flow and regions info.
5642 And for @var{n} over four, @option{-fsched-verbose} also includes
5646 @itemx -save-temps=cwd
5648 Store the usual ``temporary'' intermediate files permanently; place them
5649 in the current directory and name them based on the source file. Thus,
5650 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5651 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5652 preprocessed @file{foo.i} output file even though the compiler now
5653 normally uses an integrated preprocessor.
5655 When used in combination with the @option{-x} command line option,
5656 @option{-save-temps} is sensible enough to avoid over writing an
5657 input source file with the same extension as an intermediate file.
5658 The corresponding intermediate file may be obtained by renaming the
5659 source file before using @option{-save-temps}.
5661 If you invoke GCC in parallel, compiling several different source
5662 files that share a common base name in different subdirectories or the
5663 same source file compiled for multiple output destinations, it is
5664 likely that the different parallel compilers will interfere with each
5665 other, and overwrite the temporary files. For instance:
5668 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5669 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5672 may result in @file{foo.i} and @file{foo.o} being written to
5673 simultaneously by both compilers.
5675 @item -save-temps=obj
5676 @opindex save-temps=obj
5677 Store the usual ``temporary'' intermediate files permanently. If the
5678 @option{-o} option is used, the temporary files are based on the
5679 object file. If the @option{-o} option is not used, the
5680 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5685 gcc -save-temps=obj -c foo.c
5686 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5687 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5690 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5691 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5692 @file{dir2/yfoobar.o}.
5694 @item -time@r{[}=@var{file}@r{]}
5696 Report the CPU time taken by each subprocess in the compilation
5697 sequence. For C source files, this is the compiler proper and assembler
5698 (plus the linker if linking is done).
5700 Without the specification of an output file, the output looks like this:
5707 The first number on each line is the ``user time'', that is time spent
5708 executing the program itself. The second number is ``system time'',
5709 time spent executing operating system routines on behalf of the program.
5710 Both numbers are in seconds.
5712 With the specification of an output file, the output is appended to the
5713 named file, and it looks like this:
5716 0.12 0.01 cc1 @var{options}
5717 0.00 0.01 as @var{options}
5720 The ``user time'' and the ``system time'' are moved before the program
5721 name, and the options passed to the program are displayed, so that one
5722 can later tell what file was being compiled, and with which options.
5724 @item -fvar-tracking
5725 @opindex fvar-tracking
5726 Run variable tracking pass. It computes where variables are stored at each
5727 position in code. Better debugging information is then generated
5728 (if the debugging information format supports this information).
5730 It is enabled by default when compiling with optimization (@option{-Os},
5731 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5732 the debug info format supports it.
5734 @item -fvar-tracking-assignments
5735 @opindex fvar-tracking-assignments
5736 @opindex fno-var-tracking-assignments
5737 Annotate assignments to user variables early in the compilation and
5738 attempt to carry the annotations over throughout the compilation all the
5739 way to the end, in an attempt to improve debug information while
5740 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5742 It can be enabled even if var-tracking is disabled, in which case
5743 annotations will be created and maintained, but discarded at the end.
5745 @item -fvar-tracking-assignments-toggle
5746 @opindex fvar-tracking-assignments-toggle
5747 @opindex fno-var-tracking-assignments-toggle
5748 Toggle @option{-fvar-tracking-assignments}, in the same way that
5749 @option{-gtoggle} toggles @option{-g}.
5751 @item -print-file-name=@var{library}
5752 @opindex print-file-name
5753 Print the full absolute name of the library file @var{library} that
5754 would be used when linking---and don't do anything else. With this
5755 option, GCC does not compile or link anything; it just prints the
5758 @item -print-multi-directory
5759 @opindex print-multi-directory
5760 Print the directory name corresponding to the multilib selected by any
5761 other switches present in the command line. This directory is supposed
5762 to exist in @env{GCC_EXEC_PREFIX}.
5764 @item -print-multi-lib
5765 @opindex print-multi-lib
5766 Print the mapping from multilib directory names to compiler switches
5767 that enable them. The directory name is separated from the switches by
5768 @samp{;}, and each switch starts with an @samp{@@} instead of the
5769 @samp{-}, without spaces between multiple switches. This is supposed to
5770 ease shell-processing.
5772 @item -print-multi-os-directory
5773 @opindex print-multi-os-directory
5774 Print the path to OS libraries for the selected
5775 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5776 present in the @file{lib} subdirectory and no multilibs are used, this is
5777 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5778 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5779 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5780 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5782 @item -print-prog-name=@var{program}
5783 @opindex print-prog-name
5784 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5786 @item -print-libgcc-file-name
5787 @opindex print-libgcc-file-name
5788 Same as @option{-print-file-name=libgcc.a}.
5790 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5791 but you do want to link with @file{libgcc.a}. You can do
5794 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5797 @item -print-search-dirs
5798 @opindex print-search-dirs
5799 Print the name of the configured installation directory and a list of
5800 program and library directories @command{gcc} will search---and don't do anything else.
5802 This is useful when @command{gcc} prints the error message
5803 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5804 To resolve this you either need to put @file{cpp0} and the other compiler
5805 components where @command{gcc} expects to find them, or you can set the environment
5806 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5807 Don't forget the trailing @samp{/}.
5808 @xref{Environment Variables}.
5810 @item -print-sysroot
5811 @opindex print-sysroot
5812 Print the target sysroot directory that will be used during
5813 compilation. This is the target sysroot specified either at configure
5814 time or using the @option{--sysroot} option, possibly with an extra
5815 suffix that depends on compilation options. If no target sysroot is
5816 specified, the option prints nothing.
5818 @item -print-sysroot-headers-suffix
5819 @opindex print-sysroot-headers-suffix
5820 Print the suffix added to the target sysroot when searching for
5821 headers, or give an error if the compiler is not configured with such
5822 a suffix---and don't do anything else.
5825 @opindex dumpmachine
5826 Print the compiler's target machine (for example,
5827 @samp{i686-pc-linux-gnu})---and don't do anything else.
5830 @opindex dumpversion
5831 Print the compiler version (for example, @samp{3.0})---and don't do
5836 Print the compiler's built-in specs---and don't do anything else. (This
5837 is used when GCC itself is being built.) @xref{Spec Files}.
5839 @item -feliminate-unused-debug-types
5840 @opindex feliminate-unused-debug-types
5841 Normally, when producing DWARF2 output, GCC will emit debugging
5842 information for all types declared in a compilation
5843 unit, regardless of whether or not they are actually used
5844 in that compilation unit. Sometimes this is useful, such as
5845 if, in the debugger, you want to cast a value to a type that is
5846 not actually used in your program (but is declared). More often,
5847 however, this results in a significant amount of wasted space.
5848 With this option, GCC will avoid producing debug symbol output
5849 for types that are nowhere used in the source file being compiled.
5852 @node Optimize Options
5853 @section Options That Control Optimization
5854 @cindex optimize options
5855 @cindex options, optimization
5857 These options control various sorts of optimizations.
5859 Without any optimization option, the compiler's goal is to reduce the
5860 cost of compilation and to make debugging produce the expected
5861 results. Statements are independent: if you stop the program with a
5862 breakpoint between statements, you can then assign a new value to any
5863 variable or change the program counter to any other statement in the
5864 function and get exactly the results you would expect from the source
5867 Turning on optimization flags makes the compiler attempt to improve
5868 the performance and/or code size at the expense of compilation time
5869 and possibly the ability to debug the program.
5871 The compiler performs optimization based on the knowledge it has of the
5872 program. Compiling multiple files at once to a single output file mode allows
5873 the compiler to use information gained from all of the files when compiling
5876 Not all optimizations are controlled directly by a flag. Only
5877 optimizations that have a flag are listed in this section.
5879 Most optimizations are only enabled if an @option{-O} level is set on
5880 the command line. Otherwise they are disabled, even if individual
5881 optimization flags are specified.
5883 Depending on the target and how GCC was configured, a slightly different
5884 set of optimizations may be enabled at each @option{-O} level than
5885 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5886 to find out the exact set of optimizations that are enabled at each level.
5887 @xref{Overall Options}, for examples.
5894 Optimize. Optimizing compilation takes somewhat more time, and a lot
5895 more memory for a large function.
5897 With @option{-O}, the compiler tries to reduce code size and execution
5898 time, without performing any optimizations that take a great deal of
5901 @option{-O} turns on the following optimization flags:
5905 -fcprop-registers @gol
5908 -fdelayed-branch @gol
5910 -fguess-branch-probability @gol
5911 -fif-conversion2 @gol
5912 -fif-conversion @gol
5913 -fipa-pure-const @gol
5915 -fipa-reference @gol
5917 -fsplit-wide-types @gol
5919 -ftree-builtin-call-dce @gol
5922 -ftree-copyrename @gol
5924 -ftree-dominator-opts @gol
5926 -ftree-forwprop @gol
5934 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5935 where doing so does not interfere with debugging.
5939 Optimize even more. GCC performs nearly all supported optimizations
5940 that do not involve a space-speed tradeoff.
5941 As compared to @option{-O}, this option increases both compilation time
5942 and the performance of the generated code.
5944 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5945 also turns on the following optimization flags:
5946 @gccoptlist{-fthread-jumps @gol
5947 -falign-functions -falign-jumps @gol
5948 -falign-loops -falign-labels @gol
5951 -fcse-follow-jumps -fcse-skip-blocks @gol
5952 -fdelete-null-pointer-checks @gol
5954 -fexpensive-optimizations @gol
5955 -fgcse -fgcse-lm @gol
5956 -finline-small-functions @gol
5957 -findirect-inlining @gol
5959 -foptimize-sibling-calls @gol
5960 -fpartial-inlining @gol
5963 -freorder-blocks -freorder-functions @gol
5964 -frerun-cse-after-loop @gol
5965 -fsched-interblock -fsched-spec @gol
5966 -fschedule-insns -fschedule-insns2 @gol
5967 -fstrict-aliasing -fstrict-overflow @gol
5968 -ftree-switch-conversion @gol
5972 Please note the warning under @option{-fgcse} about
5973 invoking @option{-O2} on programs that use computed gotos.
5977 Optimize yet more. @option{-O3} turns on all optimizations specified
5978 by @option{-O2} and also turns on the @option{-finline-functions},
5979 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5980 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
5981 @option{-fipa-cp-clone} options.
5985 Reduce compilation time and make debugging produce the expected
5986 results. This is the default.
5990 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5991 do not typically increase code size. It also performs further
5992 optimizations designed to reduce code size.
5994 @option{-Os} disables the following optimization flags:
5995 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5996 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5997 -fprefetch-loop-arrays -ftree-vect-loop-version}
6001 Disregard strict standards compliance. @option{-Ofast} enables all
6002 @option{-O3} optimizations. It also enables optimizations that are not
6003 valid for all standard compliant programs.
6004 It turns on @option{-ffast-math}.
6006 If you use multiple @option{-O} options, with or without level numbers,
6007 the last such option is the one that is effective.
6010 Options of the form @option{-f@var{flag}} specify machine-independent
6011 flags. Most flags have both positive and negative forms; the negative
6012 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6013 below, only one of the forms is listed---the one you typically will
6014 use. You can figure out the other form by either removing @samp{no-}
6017 The following options control specific optimizations. They are either
6018 activated by @option{-O} options or are related to ones that are. You
6019 can use the following flags in the rare cases when ``fine-tuning'' of
6020 optimizations to be performed is desired.
6023 @item -fno-default-inline
6024 @opindex fno-default-inline
6025 Do not make member functions inline by default merely because they are
6026 defined inside the class scope (C++ only). Otherwise, when you specify
6027 @w{@option{-O}}, member functions defined inside class scope are compiled
6028 inline by default; i.e., you don't need to add @samp{inline} in front of
6029 the member function name.
6031 @item -fno-defer-pop
6032 @opindex fno-defer-pop
6033 Always pop the arguments to each function call as soon as that function
6034 returns. For machines which must pop arguments after a function call,
6035 the compiler normally lets arguments accumulate on the stack for several
6036 function calls and pops them all at once.
6038 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6040 @item -fforward-propagate
6041 @opindex fforward-propagate
6042 Perform a forward propagation pass on RTL@. The pass tries to combine two
6043 instructions and checks if the result can be simplified. If loop unrolling
6044 is active, two passes are performed and the second is scheduled after
6047 This option is enabled by default at optimization levels @option{-O},
6048 @option{-O2}, @option{-O3}, @option{-Os}.
6050 @item -ffp-contract=@var{style}
6051 @opindex ffp-contract
6052 @option{-ffp-contract=off} disables floating-point expression contraction.
6053 @option{-ffp-contract=fast} enables floating-point expression contraction
6054 such as forming of fused multiply-add operations if the target has
6055 native support for them.
6056 @option{-ffp-contract=on} enables floating-point expression contraction
6057 if allowed by the language standard. This is currently not implemented
6058 and treated equal to @option{-ffp-contract=off}.
6060 The default is @option{-ffp-contract=fast}.
6062 @item -fomit-frame-pointer
6063 @opindex fomit-frame-pointer
6064 Don't keep the frame pointer in a register for functions that
6065 don't need one. This avoids the instructions to save, set up and
6066 restore frame pointers; it also makes an extra register available
6067 in many functions. @strong{It also makes debugging impossible on
6070 On some machines, such as the VAX, this flag has no effect, because
6071 the standard calling sequence automatically handles the frame pointer
6072 and nothing is saved by pretending it doesn't exist. The
6073 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6074 whether a target machine supports this flag. @xref{Registers,,Register
6075 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6077 Starting with GCC version 4.6, the default setting (when not optimizing for
6078 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6079 @option{-fomit-frame-pointer}. The default can be reverted to
6080 @option{-fno-omit-frame-pointer} by configuring GCC with the
6081 @option{--enable-frame-pointer} configure option.
6083 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6085 @item -foptimize-sibling-calls
6086 @opindex foptimize-sibling-calls
6087 Optimize sibling and tail recursive calls.
6089 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6093 Don't pay attention to the @code{inline} keyword. Normally this option
6094 is used to keep the compiler from expanding any functions inline.
6095 Note that if you are not optimizing, no functions can be expanded inline.
6097 @item -finline-small-functions
6098 @opindex finline-small-functions
6099 Integrate functions into their callers when their body is smaller than expected
6100 function call code (so overall size of program gets smaller). The compiler
6101 heuristically decides which functions are simple enough to be worth integrating
6104 Enabled at level @option{-O2}.
6106 @item -findirect-inlining
6107 @opindex findirect-inlining
6108 Inline also indirect calls that are discovered to be known at compile
6109 time thanks to previous inlining. This option has any effect only
6110 when inlining itself is turned on by the @option{-finline-functions}
6111 or @option{-finline-small-functions} options.
6113 Enabled at level @option{-O2}.
6115 @item -finline-functions
6116 @opindex finline-functions
6117 Integrate all simple functions into their callers. The compiler
6118 heuristically decides which functions are simple enough to be worth
6119 integrating in this way.
6121 If all calls to a given function are integrated, and the function is
6122 declared @code{static}, then the function is normally not output as
6123 assembler code in its own right.
6125 Enabled at level @option{-O3}.
6127 @item -finline-functions-called-once
6128 @opindex finline-functions-called-once
6129 Consider all @code{static} functions called once for inlining into their
6130 caller even if they are not marked @code{inline}. If a call to a given
6131 function is integrated, then the function is not output as assembler code
6134 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6136 @item -fearly-inlining
6137 @opindex fearly-inlining
6138 Inline functions marked by @code{always_inline} and functions whose body seems
6139 smaller than the function call overhead early before doing
6140 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6141 makes profiling significantly cheaper and usually inlining faster on programs
6142 having large chains of nested wrapper functions.
6148 Perform interprocedural scalar replacement of aggregates, removal of
6149 unused parameters and replacement of parameters passed by reference
6150 by parameters passed by value.
6152 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6154 @item -finline-limit=@var{n}
6155 @opindex finline-limit
6156 By default, GCC limits the size of functions that can be inlined. This flag
6157 allows coarse control of this limit. @var{n} is the size of functions that
6158 can be inlined in number of pseudo instructions.
6160 Inlining is actually controlled by a number of parameters, which may be
6161 specified individually by using @option{--param @var{name}=@var{value}}.
6162 The @option{-finline-limit=@var{n}} option sets some of these parameters
6166 @item max-inline-insns-single
6167 is set to @var{n}/2.
6168 @item max-inline-insns-auto
6169 is set to @var{n}/2.
6172 See below for a documentation of the individual
6173 parameters controlling inlining and for the defaults of these parameters.
6175 @emph{Note:} there may be no value to @option{-finline-limit} that results
6176 in default behavior.
6178 @emph{Note:} pseudo instruction represents, in this particular context, an
6179 abstract measurement of function's size. In no way does it represent a count
6180 of assembly instructions and as such its exact meaning might change from one
6181 release to an another.
6183 @item -fno-keep-inline-dllexport
6184 @opindex -fno-keep-inline-dllexport
6185 This is a more fine-grained version of @option{-fkeep-inline-functions},
6186 which applies only to functions that are declared using the @code{dllexport}
6187 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6190 @item -fkeep-inline-functions
6191 @opindex fkeep-inline-functions
6192 In C, emit @code{static} functions that are declared @code{inline}
6193 into the object file, even if the function has been inlined into all
6194 of its callers. This switch does not affect functions using the
6195 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6196 inline functions into the object file.
6198 @item -fkeep-static-consts
6199 @opindex fkeep-static-consts
6200 Emit variables declared @code{static const} when optimization isn't turned
6201 on, even if the variables aren't referenced.
6203 GCC enables this option by default. If you want to force the compiler to
6204 check if the variable was referenced, regardless of whether or not
6205 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6207 @item -fmerge-constants
6208 @opindex fmerge-constants
6209 Attempt to merge identical constants (string constants and floating point
6210 constants) across compilation units.
6212 This option is the default for optimized compilation if the assembler and
6213 linker support it. Use @option{-fno-merge-constants} to inhibit this
6216 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6218 @item -fmerge-all-constants
6219 @opindex fmerge-all-constants
6220 Attempt to merge identical constants and identical variables.
6222 This option implies @option{-fmerge-constants}. In addition to
6223 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6224 arrays or initialized constant variables with integral or floating point
6225 types. Languages like C or C++ require each variable, including multiple
6226 instances of the same variable in recursive calls, to have distinct locations,
6227 so using this option will result in non-conforming
6230 @item -fmodulo-sched
6231 @opindex fmodulo-sched
6232 Perform swing modulo scheduling immediately before the first scheduling
6233 pass. This pass looks at innermost loops and reorders their
6234 instructions by overlapping different iterations.
6236 @item -fmodulo-sched-allow-regmoves
6237 @opindex fmodulo-sched-allow-regmoves
6238 Perform more aggressive SMS based modulo scheduling with register moves
6239 allowed. By setting this flag certain anti-dependences edges will be
6240 deleted which will trigger the generation of reg-moves based on the
6241 life-range analysis. This option is effective only with
6242 @option{-fmodulo-sched} enabled.
6244 @item -fno-branch-count-reg
6245 @opindex fno-branch-count-reg
6246 Do not use ``decrement and branch'' instructions on a count register,
6247 but instead generate a sequence of instructions that decrement a
6248 register, compare it against zero, then branch based upon the result.
6249 This option is only meaningful on architectures that support such
6250 instructions, which include x86, PowerPC, IA-64 and S/390.
6252 The default is @option{-fbranch-count-reg}.
6254 @item -fno-function-cse
6255 @opindex fno-function-cse
6256 Do not put function addresses in registers; make each instruction that
6257 calls a constant function contain the function's address explicitly.
6259 This option results in less efficient code, but some strange hacks
6260 that alter the assembler output may be confused by the optimizations
6261 performed when this option is not used.
6263 The default is @option{-ffunction-cse}
6265 @item -fno-zero-initialized-in-bss
6266 @opindex fno-zero-initialized-in-bss
6267 If the target supports a BSS section, GCC by default puts variables that
6268 are initialized to zero into BSS@. This can save space in the resulting
6271 This option turns off this behavior because some programs explicitly
6272 rely on variables going to the data section. E.g., so that the
6273 resulting executable can find the beginning of that section and/or make
6274 assumptions based on that.
6276 The default is @option{-fzero-initialized-in-bss}.
6278 @item -fmudflap -fmudflapth -fmudflapir
6282 @cindex bounds checking
6284 For front-ends that support it (C and C++), instrument all risky
6285 pointer/array dereferencing operations, some standard library
6286 string/heap functions, and some other associated constructs with
6287 range/validity tests. Modules so instrumented should be immune to
6288 buffer overflows, invalid heap use, and some other classes of C/C++
6289 programming errors. The instrumentation relies on a separate runtime
6290 library (@file{libmudflap}), which will be linked into a program if
6291 @option{-fmudflap} is given at link time. Run-time behavior of the
6292 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6293 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6296 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6297 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6298 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6299 instrumentation should ignore pointer reads. This produces less
6300 instrumentation (and therefore faster execution) and still provides
6301 some protection against outright memory corrupting writes, but allows
6302 erroneously read data to propagate within a program.
6304 @item -fthread-jumps
6305 @opindex fthread-jumps
6306 Perform optimizations where we check to see if a jump branches to a
6307 location where another comparison subsumed by the first is found. If
6308 so, the first branch is redirected to either the destination of the
6309 second branch or a point immediately following it, depending on whether
6310 the condition is known to be true or false.
6312 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6314 @item -fsplit-wide-types
6315 @opindex fsplit-wide-types
6316 When using a type that occupies multiple registers, such as @code{long
6317 long} on a 32-bit system, split the registers apart and allocate them
6318 independently. This normally generates better code for those types,
6319 but may make debugging more difficult.
6321 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6324 @item -fcse-follow-jumps
6325 @opindex fcse-follow-jumps
6326 In common subexpression elimination (CSE), scan through jump instructions
6327 when the target of the jump is not reached by any other path. For
6328 example, when CSE encounters an @code{if} statement with an
6329 @code{else} clause, CSE will follow the jump when the condition
6332 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6334 @item -fcse-skip-blocks
6335 @opindex fcse-skip-blocks
6336 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6337 follow jumps which conditionally skip over blocks. When CSE
6338 encounters a simple @code{if} statement with no else clause,
6339 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6340 body of the @code{if}.
6342 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6344 @item -frerun-cse-after-loop
6345 @opindex frerun-cse-after-loop
6346 Re-run common subexpression elimination after loop optimizations has been
6349 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6353 Perform a global common subexpression elimination pass.
6354 This pass also performs global constant and copy propagation.
6356 @emph{Note:} When compiling a program using computed gotos, a GCC
6357 extension, you may get better runtime performance if you disable
6358 the global common subexpression elimination pass by adding
6359 @option{-fno-gcse} to the command line.
6361 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6365 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6366 attempt to move loads which are only killed by stores into themselves. This
6367 allows a loop containing a load/store sequence to be changed to a load outside
6368 the loop, and a copy/store within the loop.
6370 Enabled by default when gcse is enabled.
6374 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6375 global common subexpression elimination. This pass will attempt to move
6376 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6377 loops containing a load/store sequence can be changed to a load before
6378 the loop and a store after the loop.
6380 Not enabled at any optimization level.
6384 When @option{-fgcse-las} is enabled, the global common subexpression
6385 elimination pass eliminates redundant loads that come after stores to the
6386 same memory location (both partial and full redundancies).
6388 Not enabled at any optimization level.
6390 @item -fgcse-after-reload
6391 @opindex fgcse-after-reload
6392 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6393 pass is performed after reload. The purpose of this pass is to cleanup
6396 @item -funsafe-loop-optimizations
6397 @opindex funsafe-loop-optimizations
6398 If given, the loop optimizer will assume that loop indices do not
6399 overflow, and that the loops with nontrivial exit condition are not
6400 infinite. This enables a wider range of loop optimizations even if
6401 the loop optimizer itself cannot prove that these assumptions are valid.
6402 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6403 if it finds this kind of loop.
6405 @item -fcrossjumping
6406 @opindex fcrossjumping
6407 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6408 resulting code may or may not perform better than without cross-jumping.
6410 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6412 @item -fauto-inc-dec
6413 @opindex fauto-inc-dec
6414 Combine increments or decrements of addresses with memory accesses.
6415 This pass is always skipped on architectures that do not have
6416 instructions to support this. Enabled by default at @option{-O} and
6417 higher on architectures that support this.
6421 Perform dead code elimination (DCE) on RTL@.
6422 Enabled by default at @option{-O} and higher.
6426 Perform dead store elimination (DSE) on RTL@.
6427 Enabled by default at @option{-O} and higher.
6429 @item -fif-conversion
6430 @opindex fif-conversion
6431 Attempt to transform conditional jumps into branch-less equivalents. This
6432 include use of conditional moves, min, max, set flags and abs instructions, and
6433 some tricks doable by standard arithmetics. The use of conditional execution
6434 on chips where it is available is controlled by @code{if-conversion2}.
6436 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6438 @item -fif-conversion2
6439 @opindex fif-conversion2
6440 Use conditional execution (where available) to transform conditional jumps into
6441 branch-less equivalents.
6443 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6445 @item -fdelete-null-pointer-checks
6446 @opindex fdelete-null-pointer-checks
6447 Assume that programs cannot safely dereference null pointers, and that
6448 no code or data element resides there. This enables simple constant
6449 folding optimizations at all optimization levels. In addition, other
6450 optimization passes in GCC use this flag to control global dataflow
6451 analyses that eliminate useless checks for null pointers; these assume
6452 that if a pointer is checked after it has already been dereferenced,
6455 Note however that in some environments this assumption is not true.
6456 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6457 for programs which depend on that behavior.
6459 Some targets, especially embedded ones, disable this option at all levels.
6460 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6461 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6462 are enabled independently at different optimization levels.
6464 @item -fdevirtualize
6465 @opindex fdevirtualize
6466 Attempt to convert calls to virtual functions to direct calls. This
6467 is done both within a procedure and interprocedurally as part of
6468 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6469 propagation (@option{-fipa-cp}).
6470 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6472 @item -fexpensive-optimizations
6473 @opindex fexpensive-optimizations
6474 Perform a number of minor optimizations that are relatively expensive.
6476 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6478 @item -foptimize-register-move
6480 @opindex foptimize-register-move
6482 Attempt to reassign register numbers in move instructions and as
6483 operands of other simple instructions in order to maximize the amount of
6484 register tying. This is especially helpful on machines with two-operand
6487 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6490 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6492 @item -fira-algorithm=@var{algorithm}
6493 Use specified coloring algorithm for the integrated register
6494 allocator. The @var{algorithm} argument should be @code{priority} or
6495 @code{CB}. The first algorithm specifies Chow's priority coloring,
6496 the second one specifies Chaitin-Briggs coloring. The second
6497 algorithm can be unimplemented for some architectures. If it is
6498 implemented, it is the default because Chaitin-Briggs coloring as a
6499 rule generates a better code.
6501 @item -fira-region=@var{region}
6502 Use specified regions for the integrated register allocator. The
6503 @var{region} argument should be one of @code{all}, @code{mixed}, or
6504 @code{one}. The first value means using all loops as register
6505 allocation regions, the second value which is the default means using
6506 all loops except for loops with small register pressure as the
6507 regions, and third one means using all function as a single region.
6508 The first value can give best result for machines with small size and
6509 irregular register set, the third one results in faster and generates
6510 decent code and the smallest size code, and the default value usually
6511 give the best results in most cases and for most architectures.
6513 @item -fira-loop-pressure
6514 @opindex fira-loop-pressure
6515 Use IRA to evaluate register pressure in loops for decision to move
6516 loop invariants. Usage of this option usually results in generation
6517 of faster and smaller code on machines with big register files (>= 32
6518 registers) but it can slow compiler down.
6520 This option is enabled at level @option{-O3} for some targets.
6522 @item -fno-ira-share-save-slots
6523 @opindex fno-ira-share-save-slots
6524 Switch off sharing stack slots used for saving call used hard
6525 registers living through a call. Each hard register will get a
6526 separate stack slot and as a result function stack frame will be
6529 @item -fno-ira-share-spill-slots
6530 @opindex fno-ira-share-spill-slots
6531 Switch off sharing stack slots allocated for pseudo-registers. Each
6532 pseudo-register which did not get a hard register will get a separate
6533 stack slot and as a result function stack frame will be bigger.
6535 @item -fira-verbose=@var{n}
6536 @opindex fira-verbose
6537 Set up how verbose dump file for the integrated register allocator
6538 will be. Default value is 5. If the value is greater or equal to 10,
6539 the dump file will be stderr as if the value were @var{n} minus 10.
6541 @item -fdelayed-branch
6542 @opindex fdelayed-branch
6543 If supported for the target machine, attempt to reorder instructions
6544 to exploit instruction slots available after delayed branch
6547 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6549 @item -fschedule-insns
6550 @opindex fschedule-insns
6551 If supported for the target machine, attempt to reorder instructions to
6552 eliminate execution stalls due to required data being unavailable. This
6553 helps machines that have slow floating point or memory load instructions
6554 by allowing other instructions to be issued until the result of the load
6555 or floating point instruction is required.
6557 Enabled at levels @option{-O2}, @option{-O3}.
6559 @item -fschedule-insns2
6560 @opindex fschedule-insns2
6561 Similar to @option{-fschedule-insns}, but requests an additional pass of
6562 instruction scheduling after register allocation has been done. This is
6563 especially useful on machines with a relatively small number of
6564 registers and where memory load instructions take more than one cycle.
6566 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6568 @item -fno-sched-interblock
6569 @opindex fno-sched-interblock
6570 Don't schedule instructions across basic blocks. This is normally
6571 enabled by default when scheduling before register allocation, i.e.@:
6572 with @option{-fschedule-insns} or at @option{-O2} or higher.
6574 @item -fno-sched-spec
6575 @opindex fno-sched-spec
6576 Don't allow speculative motion of non-load instructions. This is normally
6577 enabled by default when scheduling before register allocation, i.e.@:
6578 with @option{-fschedule-insns} or at @option{-O2} or higher.
6580 @item -fsched-pressure
6581 @opindex fsched-pressure
6582 Enable register pressure sensitive insn scheduling before the register
6583 allocation. This only makes sense when scheduling before register
6584 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6585 @option{-O2} or higher. Usage of this option can improve the
6586 generated code and decrease its size by preventing register pressure
6587 increase above the number of available hard registers and as a
6588 consequence register spills in the register allocation.
6590 @item -fsched-spec-load
6591 @opindex fsched-spec-load
6592 Allow speculative motion of some load instructions. This only makes
6593 sense when scheduling before register allocation, i.e.@: with
6594 @option{-fschedule-insns} or at @option{-O2} or higher.
6596 @item -fsched-spec-load-dangerous
6597 @opindex fsched-spec-load-dangerous
6598 Allow speculative motion of more load instructions. This only makes
6599 sense when scheduling before register allocation, i.e.@: with
6600 @option{-fschedule-insns} or at @option{-O2} or higher.
6602 @item -fsched-stalled-insns
6603 @itemx -fsched-stalled-insns=@var{n}
6604 @opindex fsched-stalled-insns
6605 Define how many insns (if any) can be moved prematurely from the queue
6606 of stalled insns into the ready list, during the second scheduling pass.
6607 @option{-fno-sched-stalled-insns} means that no insns will be moved
6608 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6609 on how many queued insns can be moved prematurely.
6610 @option{-fsched-stalled-insns} without a value is equivalent to
6611 @option{-fsched-stalled-insns=1}.
6613 @item -fsched-stalled-insns-dep
6614 @itemx -fsched-stalled-insns-dep=@var{n}
6615 @opindex fsched-stalled-insns-dep
6616 Define how many insn groups (cycles) will be examined for a dependency
6617 on a stalled insn that is candidate for premature removal from the queue
6618 of stalled insns. This has an effect only during the second scheduling pass,
6619 and only if @option{-fsched-stalled-insns} is used.
6620 @option{-fno-sched-stalled-insns-dep} is equivalent to
6621 @option{-fsched-stalled-insns-dep=0}.
6622 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6623 @option{-fsched-stalled-insns-dep=1}.
6625 @item -fsched2-use-superblocks
6626 @opindex fsched2-use-superblocks
6627 When scheduling after register allocation, do use superblock scheduling
6628 algorithm. Superblock scheduling allows motion across basic block boundaries
6629 resulting on faster schedules. This option is experimental, as not all machine
6630 descriptions used by GCC model the CPU closely enough to avoid unreliable
6631 results from the algorithm.
6633 This only makes sense when scheduling after register allocation, i.e.@: with
6634 @option{-fschedule-insns2} or at @option{-O2} or higher.
6636 @item -fsched-group-heuristic
6637 @opindex fsched-group-heuristic
6638 Enable the group heuristic in the scheduler. This heuristic favors
6639 the instruction that belongs to a schedule group. This is enabled
6640 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6641 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6643 @item -fsched-critical-path-heuristic
6644 @opindex fsched-critical-path-heuristic
6645 Enable the critical-path heuristic in the scheduler. This heuristic favors
6646 instructions on the critical path. This is enabled by default when
6647 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6648 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6650 @item -fsched-spec-insn-heuristic
6651 @opindex fsched-spec-insn-heuristic
6652 Enable the speculative instruction heuristic in the scheduler. This
6653 heuristic favors speculative instructions with greater dependency weakness.
6654 This is enabled by default when scheduling is enabled, i.e.@:
6655 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6656 or at @option{-O2} or higher.
6658 @item -fsched-rank-heuristic
6659 @opindex fsched-rank-heuristic
6660 Enable the rank heuristic in the scheduler. This heuristic favors
6661 the instruction belonging to a basic block with greater size or frequency.
6662 This is enabled by default when scheduling is enabled, i.e.@:
6663 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6664 at @option{-O2} or higher.
6666 @item -fsched-last-insn-heuristic
6667 @opindex fsched-last-insn-heuristic
6668 Enable the last-instruction heuristic in the scheduler. This heuristic
6669 favors the instruction that is less dependent on the last instruction
6670 scheduled. This is enabled by default when scheduling is enabled,
6671 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6672 at @option{-O2} or higher.
6674 @item -fsched-dep-count-heuristic
6675 @opindex fsched-dep-count-heuristic
6676 Enable the dependent-count heuristic in the scheduler. This heuristic
6677 favors the instruction that has more instructions depending on it.
6678 This is enabled by default when scheduling is enabled, i.e.@:
6679 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6680 at @option{-O2} or higher.
6682 @item -freschedule-modulo-scheduled-loops
6683 @opindex freschedule-modulo-scheduled-loops
6684 The modulo scheduling comes before the traditional scheduling, if a loop
6685 was modulo scheduled we may want to prevent the later scheduling passes
6686 from changing its schedule, we use this option to control that.
6688 @item -fselective-scheduling
6689 @opindex fselective-scheduling
6690 Schedule instructions using selective scheduling algorithm. Selective
6691 scheduling runs instead of the first scheduler pass.
6693 @item -fselective-scheduling2
6694 @opindex fselective-scheduling2
6695 Schedule instructions using selective scheduling algorithm. Selective
6696 scheduling runs instead of the second scheduler pass.
6698 @item -fsel-sched-pipelining
6699 @opindex fsel-sched-pipelining
6700 Enable software pipelining of innermost loops during selective scheduling.
6701 This option has no effect until one of @option{-fselective-scheduling} or
6702 @option{-fselective-scheduling2} is turned on.
6704 @item -fsel-sched-pipelining-outer-loops
6705 @opindex fsel-sched-pipelining-outer-loops
6706 When pipelining loops during selective scheduling, also pipeline outer loops.
6707 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6709 @item -fcaller-saves
6710 @opindex fcaller-saves
6711 Enable values to be allocated in registers that will be clobbered by
6712 function calls, by emitting extra instructions to save and restore the
6713 registers around such calls. Such allocation is done only when it
6714 seems to result in better code than would otherwise be produced.
6716 This option is always enabled by default on certain machines, usually
6717 those which have no call-preserved registers to use instead.
6719 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6721 @item -fcombine-stack-adjustments
6722 @opindex fcombine-stack-adjustments
6723 Tracks stack adjustments (pushes and pops) and stack memory references
6724 and then tries to find ways to combine them.
6726 Enabled by default at @option{-O1} and higher.
6728 @item -fconserve-stack
6729 @opindex fconserve-stack
6730 Attempt to minimize stack usage. The compiler will attempt to use less
6731 stack space, even if that makes the program slower. This option
6732 implies setting the @option{large-stack-frame} parameter to 100
6733 and the @option{large-stack-frame-growth} parameter to 400.
6735 @item -ftree-reassoc
6736 @opindex ftree-reassoc
6737 Perform reassociation on trees. This flag is enabled by default
6738 at @option{-O} and higher.
6742 Perform partial redundancy elimination (PRE) on trees. This flag is
6743 enabled by default at @option{-O2} and @option{-O3}.
6745 @item -ftree-forwprop
6746 @opindex ftree-forwprop
6747 Perform forward propagation on trees. This flag is enabled by default
6748 at @option{-O} and higher.
6752 Perform full redundancy elimination (FRE) on trees. The difference
6753 between FRE and PRE is that FRE only considers expressions
6754 that are computed on all paths leading to the redundant computation.
6755 This analysis is faster than PRE, though it exposes fewer redundancies.
6756 This flag is enabled by default at @option{-O} and higher.
6758 @item -ftree-phiprop
6759 @opindex ftree-phiprop
6760 Perform hoisting of loads from conditional pointers on trees. This
6761 pass is enabled by default at @option{-O} and higher.
6763 @item -ftree-copy-prop
6764 @opindex ftree-copy-prop
6765 Perform copy propagation on trees. This pass eliminates unnecessary
6766 copy operations. This flag is enabled by default at @option{-O} and
6769 @item -fipa-pure-const
6770 @opindex fipa-pure-const
6771 Discover which functions are pure or constant.
6772 Enabled by default at @option{-O} and higher.
6774 @item -fipa-reference
6775 @opindex fipa-reference
6776 Discover which static variables do not escape cannot escape the
6778 Enabled by default at @option{-O} and higher.
6782 Perform interprocedural pointer analysis and interprocedural modification
6783 and reference analysis. This option can cause excessive memory and
6784 compile-time usage on large compilation units. It is not enabled by
6785 default at any optimization level.
6788 @opindex fipa-profile
6789 Perform interprocedural profile propagation. The functions called only from
6790 cold functions are marked as cold. Also functions executed once (such as
6791 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6792 functions and loop less parts of functions executed once are then optimized for
6794 Enabled by default at @option{-O} and higher.
6798 Perform interprocedural constant propagation.
6799 This optimization analyzes the program to determine when values passed
6800 to functions are constants and then optimizes accordingly.
6801 This optimization can substantially increase performance
6802 if the application has constants passed to functions.
6803 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6805 @item -fipa-cp-clone
6806 @opindex fipa-cp-clone
6807 Perform function cloning to make interprocedural constant propagation stronger.
6808 When enabled, interprocedural constant propagation will perform function cloning
6809 when externally visible function can be called with constant arguments.
6810 Because this optimization can create multiple copies of functions,
6811 it may significantly increase code size
6812 (see @option{--param ipcp-unit-growth=@var{value}}).
6813 This flag is enabled by default at @option{-O3}.
6815 @item -fipa-matrix-reorg
6816 @opindex fipa-matrix-reorg
6817 Perform matrix flattening and transposing.
6818 Matrix flattening tries to replace an @math{m}-dimensional matrix
6819 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6820 This reduces the level of indirection needed for accessing the elements
6821 of the matrix. The second optimization is matrix transposing that
6822 attempts to change the order of the matrix's dimensions in order to
6823 improve cache locality.
6824 Both optimizations need the @option{-fwhole-program} flag.
6825 Transposing is enabled only if profiling information is available.
6829 Perform forward store motion on trees. This flag is
6830 enabled by default at @option{-O} and higher.
6832 @item -ftree-bit-ccp
6833 @opindex ftree-bit-ccp
6834 Perform sparse conditional bit constant propagation on trees and propagate
6835 pointer alignment information.
6836 This pass only operates on local scalar variables and is enabled by default
6837 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
6841 Perform sparse conditional constant propagation (CCP) on trees. This
6842 pass only operates on local scalar variables and is enabled by default
6843 at @option{-O} and higher.
6845 @item -ftree-switch-conversion
6846 Perform conversion of simple initializations in a switch to
6847 initializations from a scalar array. This flag is enabled by default
6848 at @option{-O2} and higher.
6852 Perform dead code elimination (DCE) on trees. This flag is enabled by
6853 default at @option{-O} and higher.
6855 @item -ftree-builtin-call-dce
6856 @opindex ftree-builtin-call-dce
6857 Perform conditional dead code elimination (DCE) for calls to builtin functions
6858 that may set @code{errno} but are otherwise side-effect free. This flag is
6859 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6862 @item -ftree-dominator-opts
6863 @opindex ftree-dominator-opts
6864 Perform a variety of simple scalar cleanups (constant/copy
6865 propagation, redundancy elimination, range propagation and expression
6866 simplification) based on a dominator tree traversal. This also
6867 performs jump threading (to reduce jumps to jumps). This flag is
6868 enabled by default at @option{-O} and higher.
6872 Perform dead store elimination (DSE) on trees. A dead store is a store into
6873 a memory location which will later be overwritten by another store without
6874 any intervening loads. In this case the earlier store can be deleted. This
6875 flag is enabled by default at @option{-O} and higher.
6879 Perform loop header copying on trees. This is beneficial since it increases
6880 effectiveness of code motion optimizations. It also saves one jump. This flag
6881 is enabled by default at @option{-O} and higher. It is not enabled
6882 for @option{-Os}, since it usually increases code size.
6884 @item -ftree-loop-optimize
6885 @opindex ftree-loop-optimize
6886 Perform loop optimizations on trees. This flag is enabled by default
6887 at @option{-O} and higher.
6889 @item -ftree-loop-linear
6890 @opindex ftree-loop-linear
6891 Perform loop interchange transformations on tree. Same as
6892 @option{-floop-interchange}. To use this code transformation, GCC has
6893 to be configured with @option{--with-ppl} and @option{--with-cloog} to
6894 enable the Graphite loop transformation infrastructure.
6896 @item -floop-interchange
6897 @opindex floop-interchange
6898 Perform loop interchange transformations on loops. Interchanging two
6899 nested loops switches the inner and outer loops. For example, given a
6904 A(J, I) = A(J, I) * C
6908 loop interchange will transform the loop as if the user had written:
6912 A(J, I) = A(J, I) * C
6916 which can be beneficial when @code{N} is larger than the caches,
6917 because in Fortran, the elements of an array are stored in memory
6918 contiguously by column, and the original loop iterates over rows,
6919 potentially creating at each access a cache miss. This optimization
6920 applies to all the languages supported by GCC and is not limited to
6921 Fortran. To use this code transformation, GCC has to be configured
6922 with @option{--with-ppl} and @option{--with-cloog} to enable the
6923 Graphite loop transformation infrastructure.
6925 @item -floop-strip-mine
6926 @opindex floop-strip-mine
6927 Perform loop strip mining transformations on loops. Strip mining
6928 splits a loop into two nested loops. The outer loop has strides
6929 equal to the strip size and the inner loop has strides of the
6930 original loop within a strip. The strip length can be changed
6931 using the @option{loop-block-tile-size} parameter. For example,
6938 loop strip mining will transform the loop as if the user had written:
6941 DO I = II, min (II + 50, N)
6946 This optimization applies to all the languages supported by GCC and is
6947 not limited to Fortran. To use this code transformation, GCC has to
6948 be configured with @option{--with-ppl} and @option{--with-cloog} to
6949 enable the Graphite loop transformation infrastructure.
6952 @opindex floop-block
6953 Perform loop blocking transformations on loops. Blocking strip mines
6954 each loop in the loop nest such that the memory accesses of the
6955 element loops fit inside caches. The strip length can be changed
6956 using the @option{loop-block-tile-size} parameter. For example, given
6961 A(J, I) = B(I) + C(J)
6965 loop blocking will transform the loop as if the user had written:
6969 DO I = II, min (II + 50, N)
6970 DO J = JJ, min (JJ + 50, M)
6971 A(J, I) = B(I) + C(J)
6977 which can be beneficial when @code{M} is larger than the caches,
6978 because the innermost loop will iterate over a smaller amount of data
6979 that can be kept in the caches. This optimization applies to all the
6980 languages supported by GCC and is not limited to Fortran. To use this
6981 code transformation, GCC has to be configured with @option{--with-ppl}
6982 and @option{--with-cloog} to enable the Graphite loop transformation
6985 @item -fgraphite-identity
6986 @opindex fgraphite-identity
6987 Enable the identity transformation for graphite. For every SCoP we generate
6988 the polyhedral representation and transform it back to gimple. Using
6989 @option{-fgraphite-identity} we can check the costs or benefits of the
6990 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6991 are also performed by the code generator CLooG, like index splitting and
6992 dead code elimination in loops.
6994 @item -floop-flatten
6995 @opindex floop-flatten
6996 Removes the loop nesting structure: transforms the loop nest into a
6997 single loop. This transformation can be useful to vectorize all the
6998 levels of the loop nest.
7000 @item -floop-parallelize-all
7001 @opindex floop-parallelize-all
7002 Use the Graphite data dependence analysis to identify loops that can
7003 be parallelized. Parallelize all the loops that can be analyzed to
7004 not contain loop carried dependences without checking that it is
7005 profitable to parallelize the loops.
7007 @item -fcheck-data-deps
7008 @opindex fcheck-data-deps
7009 Compare the results of several data dependence analyzers. This option
7010 is used for debugging the data dependence analyzers.
7012 @item -ftree-loop-if-convert
7013 Attempt to transform conditional jumps in the innermost loops to
7014 branch-less equivalents. The intent is to remove control-flow from
7015 the innermost loops in order to improve the ability of the
7016 vectorization pass to handle these loops. This is enabled by default
7017 if vectorization is enabled.
7019 @item -ftree-loop-if-convert-stores
7020 Attempt to also if-convert conditional jumps containing memory writes.
7021 This transformation can be unsafe for multi-threaded programs as it
7022 transforms conditional memory writes into unconditional memory writes.
7025 for (i = 0; i < N; i++)
7029 would be transformed to
7031 for (i = 0; i < N; i++)
7032 A[i] = cond ? expr : A[i];
7034 potentially producing data races.
7036 @item -ftree-loop-distribution
7037 Perform loop distribution. This flag can improve cache performance on
7038 big loop bodies and allow further loop optimizations, like
7039 parallelization or vectorization, to take place. For example, the loop
7056 @item -ftree-loop-distribute-patterns
7057 Perform loop distribution of patterns that can be code generated with
7058 calls to a library. This flag is enabled by default at @option{-O3}.
7060 This pass distributes the initialization loops and generates a call to
7061 memset zero. For example, the loop
7077 and the initialization loop is transformed into a call to memset zero.
7079 @item -ftree-loop-im
7080 @opindex ftree-loop-im
7081 Perform loop invariant motion on trees. This pass moves only invariants that
7082 would be hard to handle at RTL level (function calls, operations that expand to
7083 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7084 operands of conditions that are invariant out of the loop, so that we can use
7085 just trivial invariantness analysis in loop unswitching. The pass also includes
7088 @item -ftree-loop-ivcanon
7089 @opindex ftree-loop-ivcanon
7090 Create a canonical counter for number of iterations in the loop for that
7091 determining number of iterations requires complicated analysis. Later
7092 optimizations then may determine the number easily. Useful especially
7093 in connection with unrolling.
7097 Perform induction variable optimizations (strength reduction, induction
7098 variable merging and induction variable elimination) on trees.
7100 @item -ftree-parallelize-loops=n
7101 @opindex ftree-parallelize-loops
7102 Parallelize loops, i.e., split their iteration space to run in n threads.
7103 This is only possible for loops whose iterations are independent
7104 and can be arbitrarily reordered. The optimization is only
7105 profitable on multiprocessor machines, for loops that are CPU-intensive,
7106 rather than constrained e.g.@: by memory bandwidth. This option
7107 implies @option{-pthread}, and thus is only supported on targets
7108 that have support for @option{-pthread}.
7112 Perform function-local points-to analysis on trees. This flag is
7113 enabled by default at @option{-O} and higher.
7117 Perform scalar replacement of aggregates. This pass replaces structure
7118 references with scalars to prevent committing structures to memory too
7119 early. This flag is enabled by default at @option{-O} and higher.
7121 @item -ftree-copyrename
7122 @opindex ftree-copyrename
7123 Perform copy renaming on trees. This pass attempts to rename compiler
7124 temporaries to other variables at copy locations, usually resulting in
7125 variable names which more closely resemble the original variables. This flag
7126 is enabled by default at @option{-O} and higher.
7130 Perform temporary expression replacement during the SSA->normal phase. Single
7131 use/single def temporaries are replaced at their use location with their
7132 defining expression. This results in non-GIMPLE code, but gives the expanders
7133 much more complex trees to work on resulting in better RTL generation. This is
7134 enabled by default at @option{-O} and higher.
7136 @item -ftree-vectorize
7137 @opindex ftree-vectorize
7138 Perform loop vectorization on trees. This flag is enabled by default at
7141 @item -ftree-slp-vectorize
7142 @opindex ftree-slp-vectorize
7143 Perform basic block vectorization on trees. This flag is enabled by default at
7144 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7146 @item -ftree-vect-loop-version
7147 @opindex ftree-vect-loop-version
7148 Perform loop versioning when doing loop vectorization on trees. When a loop
7149 appears to be vectorizable except that data alignment or data dependence cannot
7150 be determined at compile time then vectorized and non-vectorized versions of
7151 the loop are generated along with runtime checks for alignment or dependence
7152 to control which version is executed. This option is enabled by default
7153 except at level @option{-Os} where it is disabled.
7155 @item -fvect-cost-model
7156 @opindex fvect-cost-model
7157 Enable cost model for vectorization.
7161 Perform Value Range Propagation on trees. This is similar to the
7162 constant propagation pass, but instead of values, ranges of values are
7163 propagated. This allows the optimizers to remove unnecessary range
7164 checks like array bound checks and null pointer checks. This is
7165 enabled by default at @option{-O2} and higher. Null pointer check
7166 elimination is only done if @option{-fdelete-null-pointer-checks} is
7171 Perform tail duplication to enlarge superblock size. This transformation
7172 simplifies the control flow of the function allowing other optimizations to do
7175 @item -funroll-loops
7176 @opindex funroll-loops
7177 Unroll loops whose number of iterations can be determined at compile
7178 time or upon entry to the loop. @option{-funroll-loops} implies
7179 @option{-frerun-cse-after-loop}. This option makes code larger,
7180 and may or may not make it run faster.
7182 @item -funroll-all-loops
7183 @opindex funroll-all-loops
7184 Unroll all loops, even if their number of iterations is uncertain when
7185 the loop is entered. This usually makes programs run more slowly.
7186 @option{-funroll-all-loops} implies the same options as
7187 @option{-funroll-loops},
7189 @item -fsplit-ivs-in-unroller
7190 @opindex fsplit-ivs-in-unroller
7191 Enables expressing of values of induction variables in later iterations
7192 of the unrolled loop using the value in the first iteration. This breaks
7193 long dependency chains, thus improving efficiency of the scheduling passes.
7195 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7196 same effect. However in cases the loop body is more complicated than
7197 a single basic block, this is not reliable. It also does not work at all
7198 on some of the architectures due to restrictions in the CSE pass.
7200 This optimization is enabled by default.
7202 @item -fvariable-expansion-in-unroller
7203 @opindex fvariable-expansion-in-unroller
7204 With this option, the compiler will create multiple copies of some
7205 local variables when unrolling a loop which can result in superior code.
7207 @item -fpartial-inlining
7208 @opindex fpartial-inlining
7209 Inline parts of functions. This option has any effect only
7210 when inlining itself is turned on by the @option{-finline-functions}
7211 or @option{-finline-small-functions} options.
7213 Enabled at level @option{-O2}.
7215 @item -fpredictive-commoning
7216 @opindex fpredictive-commoning
7217 Perform predictive commoning optimization, i.e., reusing computations
7218 (especially memory loads and stores) performed in previous
7219 iterations of loops.
7221 This option is enabled at level @option{-O3}.
7223 @item -fprefetch-loop-arrays
7224 @opindex fprefetch-loop-arrays
7225 If supported by the target machine, generate instructions to prefetch
7226 memory to improve the performance of loops that access large arrays.
7228 This option may generate better or worse code; results are highly
7229 dependent on the structure of loops within the source code.
7231 Disabled at level @option{-Os}.
7234 @itemx -fno-peephole2
7235 @opindex fno-peephole
7236 @opindex fno-peephole2
7237 Disable any machine-specific peephole optimizations. The difference
7238 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7239 are implemented in the compiler; some targets use one, some use the
7240 other, a few use both.
7242 @option{-fpeephole} is enabled by default.
7243 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7245 @item -fno-guess-branch-probability
7246 @opindex fno-guess-branch-probability
7247 Do not guess branch probabilities using heuristics.
7249 GCC will use heuristics to guess branch probabilities if they are
7250 not provided by profiling feedback (@option{-fprofile-arcs}). These
7251 heuristics are based on the control flow graph. If some branch probabilities
7252 are specified by @samp{__builtin_expect}, then the heuristics will be
7253 used to guess branch probabilities for the rest of the control flow graph,
7254 taking the @samp{__builtin_expect} info into account. The interactions
7255 between the heuristics and @samp{__builtin_expect} can be complex, and in
7256 some cases, it may be useful to disable the heuristics so that the effects
7257 of @samp{__builtin_expect} are easier to understand.
7259 The default is @option{-fguess-branch-probability} at levels
7260 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7262 @item -freorder-blocks
7263 @opindex freorder-blocks
7264 Reorder basic blocks in the compiled function in order to reduce number of
7265 taken branches and improve code locality.
7267 Enabled at levels @option{-O2}, @option{-O3}.
7269 @item -freorder-blocks-and-partition
7270 @opindex freorder-blocks-and-partition
7271 In addition to reordering basic blocks in the compiled function, in order
7272 to reduce number of taken branches, partitions hot and cold basic blocks
7273 into separate sections of the assembly and .o files, to improve
7274 paging and cache locality performance.
7276 This optimization is automatically turned off in the presence of
7277 exception handling, for linkonce sections, for functions with a user-defined
7278 section attribute and on any architecture that does not support named
7281 @item -freorder-functions
7282 @opindex freorder-functions
7283 Reorder functions in the object file in order to
7284 improve code locality. This is implemented by using special
7285 subsections @code{.text.hot} for most frequently executed functions and
7286 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7287 the linker so object file format must support named sections and linker must
7288 place them in a reasonable way.
7290 Also profile feedback must be available in to make this option effective. See
7291 @option{-fprofile-arcs} for details.
7293 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7295 @item -fstrict-aliasing
7296 @opindex fstrict-aliasing
7297 Allow the compiler to assume the strictest aliasing rules applicable to
7298 the language being compiled. For C (and C++), this activates
7299 optimizations based on the type of expressions. In particular, an
7300 object of one type is assumed never to reside at the same address as an
7301 object of a different type, unless the types are almost the same. For
7302 example, an @code{unsigned int} can alias an @code{int}, but not a
7303 @code{void*} or a @code{double}. A character type may alias any other
7306 @anchor{Type-punning}Pay special attention to code like this:
7319 The practice of reading from a different union member than the one most
7320 recently written to (called ``type-punning'') is common. Even with
7321 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7322 is accessed through the union type. So, the code above will work as
7323 expected. @xref{Structures unions enumerations and bit-fields
7324 implementation}. However, this code might not:
7335 Similarly, access by taking the address, casting the resulting pointer
7336 and dereferencing the result has undefined behavior, even if the cast
7337 uses a union type, e.g.:
7341 return ((union a_union *) &d)->i;
7345 The @option{-fstrict-aliasing} option is enabled at levels
7346 @option{-O2}, @option{-O3}, @option{-Os}.
7348 @item -fstrict-overflow
7349 @opindex fstrict-overflow
7350 Allow the compiler to assume strict signed overflow rules, depending
7351 on the language being compiled. For C (and C++) this means that
7352 overflow when doing arithmetic with signed numbers is undefined, which
7353 means that the compiler may assume that it will not happen. This
7354 permits various optimizations. For example, the compiler will assume
7355 that an expression like @code{i + 10 > i} will always be true for
7356 signed @code{i}. This assumption is only valid if signed overflow is
7357 undefined, as the expression is false if @code{i + 10} overflows when
7358 using twos complement arithmetic. When this option is in effect any
7359 attempt to determine whether an operation on signed numbers will
7360 overflow must be written carefully to not actually involve overflow.
7362 This option also allows the compiler to assume strict pointer
7363 semantics: given a pointer to an object, if adding an offset to that
7364 pointer does not produce a pointer to the same object, the addition is
7365 undefined. This permits the compiler to conclude that @code{p + u >
7366 p} is always true for a pointer @code{p} and unsigned integer
7367 @code{u}. This assumption is only valid because pointer wraparound is
7368 undefined, as the expression is false if @code{p + u} overflows using
7369 twos complement arithmetic.
7371 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7372 that integer signed overflow is fully defined: it wraps. When
7373 @option{-fwrapv} is used, there is no difference between
7374 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7375 integers. With @option{-fwrapv} certain types of overflow are
7376 permitted. For example, if the compiler gets an overflow when doing
7377 arithmetic on constants, the overflowed value can still be used with
7378 @option{-fwrapv}, but not otherwise.
7380 The @option{-fstrict-overflow} option is enabled at levels
7381 @option{-O2}, @option{-O3}, @option{-Os}.
7383 @item -falign-functions
7384 @itemx -falign-functions=@var{n}
7385 @opindex falign-functions
7386 Align the start of functions to the next power-of-two greater than
7387 @var{n}, skipping up to @var{n} bytes. For instance,
7388 @option{-falign-functions=32} aligns functions to the next 32-byte
7389 boundary, but @option{-falign-functions=24} would align to the next
7390 32-byte boundary only if this can be done by skipping 23 bytes or less.
7392 @option{-fno-align-functions} and @option{-falign-functions=1} are
7393 equivalent and mean that functions will not be aligned.
7395 Some assemblers only support this flag when @var{n} is a power of two;
7396 in that case, it is rounded up.
7398 If @var{n} is not specified or is zero, use a machine-dependent default.
7400 Enabled at levels @option{-O2}, @option{-O3}.
7402 @item -falign-labels
7403 @itemx -falign-labels=@var{n}
7404 @opindex falign-labels
7405 Align all branch targets to a power-of-two boundary, skipping up to
7406 @var{n} bytes like @option{-falign-functions}. This option can easily
7407 make code slower, because it must insert dummy operations for when the
7408 branch target is reached in the usual flow of the code.
7410 @option{-fno-align-labels} and @option{-falign-labels=1} are
7411 equivalent and mean that labels will not be aligned.
7413 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7414 are greater than this value, then their values are used instead.
7416 If @var{n} is not specified or is zero, use a machine-dependent default
7417 which is very likely to be @samp{1}, meaning no alignment.
7419 Enabled at levels @option{-O2}, @option{-O3}.
7422 @itemx -falign-loops=@var{n}
7423 @opindex falign-loops
7424 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7425 like @option{-falign-functions}. The hope is that the loop will be
7426 executed many times, which will make up for any execution of the dummy
7429 @option{-fno-align-loops} and @option{-falign-loops=1} are
7430 equivalent and mean that loops will not be aligned.
7432 If @var{n} is not specified or is zero, use a machine-dependent default.
7434 Enabled at levels @option{-O2}, @option{-O3}.
7437 @itemx -falign-jumps=@var{n}
7438 @opindex falign-jumps
7439 Align branch targets to a power-of-two boundary, for branch targets
7440 where the targets can only be reached by jumping, skipping up to @var{n}
7441 bytes like @option{-falign-functions}. In this case, no dummy operations
7444 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7445 equivalent and mean that loops will not be aligned.
7447 If @var{n} is not specified or is zero, use a machine-dependent default.
7449 Enabled at levels @option{-O2}, @option{-O3}.
7451 @item -funit-at-a-time
7452 @opindex funit-at-a-time
7453 This option is left for compatibility reasons. @option{-funit-at-a-time}
7454 has no effect, while @option{-fno-unit-at-a-time} implies
7455 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7459 @item -fno-toplevel-reorder
7460 @opindex fno-toplevel-reorder
7461 Do not reorder top-level functions, variables, and @code{asm}
7462 statements. Output them in the same order that they appear in the
7463 input file. When this option is used, unreferenced static variables
7464 will not be removed. This option is intended to support existing code
7465 which relies on a particular ordering. For new code, it is better to
7468 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7469 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7474 Constructs webs as commonly used for register allocation purposes and assign
7475 each web individual pseudo register. This allows the register allocation pass
7476 to operate on pseudos directly, but also strengthens several other optimization
7477 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7478 however, make debugging impossible, since variables will no longer stay in a
7481 Enabled by default with @option{-funroll-loops}.
7483 @item -fwhole-program
7484 @opindex fwhole-program
7485 Assume that the current compilation unit represents the whole program being
7486 compiled. All public functions and variables with the exception of @code{main}
7487 and those merged by attribute @code{externally_visible} become static functions
7488 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.
7489 While this option is equivalent to proper use of the @code{static} keyword for
7490 programs consisting of a single file, in combination with option
7491 @option{-flto} this flag can be used to
7492 compile many smaller scale programs since the functions and variables become
7493 local for the whole combined compilation unit, not for the single source file
7496 This option implies @option{-fwhole-file} for Fortran programs.
7498 @item -flto[=@var{n}]
7500 This option runs the standard link-time optimizer. When invoked
7501 with source code, it generates GIMPLE (one of GCC's internal
7502 representations) and writes it to special ELF sections in the object
7503 file. When the object files are linked together, all the function
7504 bodies are read from these ELF sections and instantiated as if they
7505 had been part of the same translation unit.
7507 To use the link-timer optimizer, @option{-flto} needs to be specified at
7508 compile time and during the final link. For example,
7511 gcc -c -O2 -flto foo.c
7512 gcc -c -O2 -flto bar.c
7513 gcc -o myprog -flto -O2 foo.o bar.o
7516 The first two invocations to GCC will save a bytecode representation
7517 of GIMPLE into special ELF sections inside @file{foo.o} and
7518 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7519 @file{foo.o} and @file{bar.o}, merge the two files into a single
7520 internal image, and compile the result as usual. Since both
7521 @file{foo.o} and @file{bar.o} are merged into a single image, this
7522 causes all the inter-procedural analyses and optimizations in GCC to
7523 work across the two files as if they were a single one. This means,
7524 for example, that the inliner will be able to inline functions in
7525 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7527 Another (simpler) way to enable link-time optimization is,
7530 gcc -o myprog -flto -O2 foo.c bar.c
7533 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7534 merge them together into a single GIMPLE representation and optimize
7535 them as usual to produce @file{myprog}.
7537 The only important thing to keep in mind is that to enable link-time
7538 optimizations the @option{-flto} flag needs to be passed to both the
7539 compile and the link commands.
7541 To make whole program optimization effective, it is necessary to make
7542 certain whole program assumptions. The compiler needs to know
7543 what functions and variables can be accessed by libraries and runtime
7544 outside of the link time optimized unit. When supported by the linker,
7545 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7546 compiler information about used and externally visible symbols. When
7547 the linker plugin is not available, @option{-fwhole-program} should be
7548 used to allow the compiler to make these assumptions, which will lead
7549 to more aggressive optimization decisions.
7551 Note that when a file is compiled with @option{-flto}, the generated
7552 object file will be larger than a regular object file because it will
7553 contain GIMPLE bytecodes and the usual final code. This means that
7554 object files with LTO information can be linked as a normal object
7555 file. So, in the previous example, if the final link is done with
7558 gcc -o myprog foo.o bar.o
7561 The only difference will be that no inter-procedural optimizations
7562 will be applied to produce @file{myprog}. The two object files
7563 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7566 Additionally, the optimization flags used to compile individual files
7567 are not necessarily related to those used at link-time. For instance,
7570 gcc -c -O0 -flto foo.c
7571 gcc -c -O0 -flto bar.c
7572 gcc -o myprog -flto -O3 foo.o bar.o
7575 This will produce individual object files with unoptimized assembler
7576 code, but the resulting binary @file{myprog} will be optimized at
7577 @option{-O3}. Now, if the final binary is generated without
7578 @option{-flto}, then @file{myprog} will not be optimized.
7580 When producing the final binary with @option{-flto}, GCC will only
7581 apply link-time optimizations to those files that contain bytecode.
7582 Therefore, you can mix and match object files and libraries with
7583 GIMPLE bytecodes and final object code. GCC will automatically select
7584 which files to optimize in LTO mode and which files to link without
7587 There are some code generation flags that GCC will preserve when
7588 generating bytecodes, as they need to be used during the final link
7589 stage. Currently, the following options are saved into the GIMPLE
7590 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7591 @option{-m} target flags.
7593 At link time, these options are read-in and reapplied. Note that the
7594 current implementation makes no attempt at recognizing conflicting
7595 values for these options. If two or more files have a conflicting
7596 value (e.g., one file is compiled with @option{-fPIC} and another
7597 isn't), the compiler will simply use the last value read from the
7598 bytecode files. It is recommended, then, that all the files
7599 participating in the same link be compiled with the same options.
7601 Another feature of LTO is that it is possible to apply interprocedural
7602 optimizations on files written in different languages. This requires
7603 some support in the language front end. Currently, the C, C++ and
7604 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7605 something like this should work
7610 gfortran -c -flto baz.f90
7611 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7614 Notice that the final link is done with @command{g++} to get the C++
7615 runtime libraries and @option{-lgfortran} is added to get the Fortran
7616 runtime libraries. In general, when mixing languages in LTO mode, you
7617 should use the same link command used when mixing languages in a
7618 regular (non-LTO) compilation. This means that if your build process
7619 was mixing languages before, all you need to add is @option{-flto} to
7620 all the compile and link commands.
7622 If LTO encounters objects with C linkage declared with incompatible
7623 types in separate translation units to be linked together (undefined
7624 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7625 issued. The behavior is still undefined at runtime.
7627 If object files containing GIMPLE bytecode are stored in a library archive, say
7628 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7629 are using a linker with linker plugin support. To enable this feature, use
7630 the flag @option{-fuse-linker-plugin} at link-time:
7633 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7636 With the linker plugin enabled, the linker will extract the needed
7637 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7638 to make them part of the aggregated GIMPLE image to be optimized.
7640 If you are not using a linker with linker plugin support and/or do not
7641 enable linker plugin then the objects inside @file{libfoo.a}
7642 will be extracted and linked as usual, but they will not participate
7643 in the LTO optimization process.
7645 Link time optimizations do not require the presence of the whole program to
7646 operate. If the program does not require any symbols to be exported, it is
7647 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7648 the interprocedural optimizers to use more aggressive assumptions which may
7649 lead to improved optimization opportunities.
7650 Use of @option{-fwhole-program} is not needed when linker plugin is
7651 active (see @option{-fuse-linker-plugin}).
7653 Regarding portability: the current implementation of LTO makes no
7654 attempt at generating bytecode that can be ported between different
7655 types of hosts. The bytecode files are versioned and there is a
7656 strict version check, so bytecode files generated in one version of
7657 GCC will not work with an older/newer version of GCC.
7659 Link time optimization does not play well with generating debugging
7660 information. Combining @option{-flto} with
7661 @option{-g} is currently experimental and expected to produce wrong
7664 If you specify the optional @var{n}, the optimization and code
7665 generation done at link time is executed in parallel using @var{n}
7666 parallel jobs by utilizing an installed @command{make} program. The
7667 environment variable @env{MAKE} may be used to override the program
7668 used. The default value for @var{n} is 1.
7670 You can also specify @option{-flto=jobserver} to use GNU make's
7671 job server mode to determine the number of parallel jobs. This
7672 is useful when the Makefile calling GCC is already executing in parallel.
7673 The parent Makefile will need a @samp{+} prepended to the command recipe
7674 for this to work. This will likely only work if @env{MAKE} is
7677 This option is disabled by default.
7679 @item -flto-partition=@var{alg}
7680 @opindex flto-partition
7681 Specify the partitioning algorithm used by the link time optimizer.
7682 The value is either @code{1to1} to specify a partitioning mirroring
7683 the original source files or @code{balanced} to specify partitioning
7684 into equally sized chunks (whenever possible). Specifying @code{none}
7685 as an algorithm disables partitioning and streaming completely. The
7686 default value is @code{balanced}.
7688 @item -flto-compression-level=@var{n}
7689 This option specifies the level of compression used for intermediate
7690 language written to LTO object files, and is only meaningful in
7691 conjunction with LTO mode (@option{-flto}). Valid
7692 values are 0 (no compression) to 9 (maximum compression). Values
7693 outside this range are clamped to either 0 or 9. If the option is not
7694 given, a default balanced compression setting is used.
7697 Prints a report with internal details on the workings of the link-time
7698 optimizer. The contents of this report vary from version to version,
7699 it is meant to be useful to GCC developers when processing object
7700 files in LTO mode (via @option{-flto}).
7702 Disabled by default.
7704 @item -fuse-linker-plugin
7705 Enables the use of linker plugin during link time optimization. This option
7706 relies on the linker plugin support in linker that is available in gold
7707 or in GNU ld 2.21 or newer.
7709 This option enables the extraction of object files with GIMPLE bytecode out of
7710 library archives. This improves the quality of optimization by exposing more
7711 code the the link time optimizer. This information specify what symbols
7712 can be accessed externally (by non-LTO object or during dynamic linking).
7713 Resulting code quality improvements on binaries (and shared libraries that do
7714 use hidden visibility) is similar to @code{-fwhole-program}. See
7715 @option{-flto} for a description on the effect of this flag and how to use it.
7717 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7718 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7720 @item -fcompare-elim
7721 @opindex fcompare-elim
7722 After register allocation and post-register allocation instruction splitting,
7723 identify arithmetic instructions that compute processor flags similar to a
7724 comparison operation based on that arithmetic. If possible, eliminate the
7725 explicit comparison operation.
7727 This pass only applies to certain targets that cannot explicitly represent
7728 the comparison operation before register allocation is complete.
7730 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7732 @item -fcprop-registers
7733 @opindex fcprop-registers
7734 After register allocation and post-register allocation instruction splitting,
7735 we perform a copy-propagation pass to try to reduce scheduling dependencies
7736 and occasionally eliminate the copy.
7738 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7740 @item -fprofile-correction
7741 @opindex fprofile-correction
7742 Profiles collected using an instrumented binary for multi-threaded programs may
7743 be inconsistent due to missed counter updates. When this option is specified,
7744 GCC will use heuristics to correct or smooth out such inconsistencies. By
7745 default, GCC will emit an error message when an inconsistent profile is detected.
7747 @item -fprofile-dir=@var{path}
7748 @opindex fprofile-dir
7750 Set the directory to search for the profile data files in to @var{path}.
7751 This option affects only the profile data generated by
7752 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7753 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7754 and its related options.
7755 By default, GCC will use the current directory as @var{path}, thus the
7756 profile data file will appear in the same directory as the object file.
7758 @item -fprofile-generate
7759 @itemx -fprofile-generate=@var{path}
7760 @opindex fprofile-generate
7762 Enable options usually used for instrumenting application to produce
7763 profile useful for later recompilation with profile feedback based
7764 optimization. You must use @option{-fprofile-generate} both when
7765 compiling and when linking your program.
7767 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7769 If @var{path} is specified, GCC will look at the @var{path} to find
7770 the profile feedback data files. See @option{-fprofile-dir}.
7773 @itemx -fprofile-use=@var{path}
7774 @opindex fprofile-use
7775 Enable profile feedback directed optimizations, and optimizations
7776 generally profitable only with profile feedback available.
7778 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7779 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7781 By default, GCC emits an error message if the feedback profiles do not
7782 match the source code. This error can be turned into a warning by using
7783 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7786 If @var{path} is specified, GCC will look at the @var{path} to find
7787 the profile feedback data files. See @option{-fprofile-dir}.
7790 The following options control compiler behavior regarding floating
7791 point arithmetic. These options trade off between speed and
7792 correctness. All must be specifically enabled.
7796 @opindex ffloat-store
7797 Do not store floating point variables in registers, and inhibit other
7798 options that might change whether a floating point value is taken from a
7801 @cindex floating point precision
7802 This option prevents undesirable excess precision on machines such as
7803 the 68000 where the floating registers (of the 68881) keep more
7804 precision than a @code{double} is supposed to have. Similarly for the
7805 x86 architecture. For most programs, the excess precision does only
7806 good, but a few programs rely on the precise definition of IEEE floating
7807 point. Use @option{-ffloat-store} for such programs, after modifying
7808 them to store all pertinent intermediate computations into variables.
7810 @item -fexcess-precision=@var{style}
7811 @opindex fexcess-precision
7812 This option allows further control over excess precision on machines
7813 where floating-point registers have more precision than the IEEE
7814 @code{float} and @code{double} types and the processor does not
7815 support operations rounding to those types. By default,
7816 @option{-fexcess-precision=fast} is in effect; this means that
7817 operations are carried out in the precision of the registers and that
7818 it is unpredictable when rounding to the types specified in the source
7819 code takes place. When compiling C, if
7820 @option{-fexcess-precision=standard} is specified then excess
7821 precision will follow the rules specified in ISO C99; in particular,
7822 both casts and assignments cause values to be rounded to their
7823 semantic types (whereas @option{-ffloat-store} only affects
7824 assignments). This option is enabled by default for C if a strict
7825 conformance option such as @option{-std=c99} is used.
7828 @option{-fexcess-precision=standard} is not implemented for languages
7829 other than C, and has no effect if
7830 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7831 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7832 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7833 semantics apply without excess precision, and in the latter, rounding
7838 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7839 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7840 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7842 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7844 This option is not turned on by any @option{-O} option besides
7845 @option{-Ofast} since it can result in incorrect output for programs
7846 which depend on an exact implementation of IEEE or ISO rules/specifications
7847 for math functions. It may, however, yield faster code for programs
7848 that do not require the guarantees of these specifications.
7850 @item -fno-math-errno
7851 @opindex fno-math-errno
7852 Do not set ERRNO after calling math functions that are executed
7853 with a single instruction, e.g., sqrt. A program that relies on
7854 IEEE exceptions for math error handling may want to use this flag
7855 for speed while maintaining IEEE arithmetic compatibility.
7857 This option is not turned on by any @option{-O} option since
7858 it can result in incorrect output for programs which depend on
7859 an exact implementation of IEEE or ISO rules/specifications for
7860 math functions. It may, however, yield faster code for programs
7861 that do not require the guarantees of these specifications.
7863 The default is @option{-fmath-errno}.
7865 On Darwin systems, the math library never sets @code{errno}. There is
7866 therefore no reason for the compiler to consider the possibility that
7867 it might, and @option{-fno-math-errno} is the default.
7869 @item -funsafe-math-optimizations
7870 @opindex funsafe-math-optimizations
7872 Allow optimizations for floating-point arithmetic that (a) assume
7873 that arguments and results are valid and (b) may violate IEEE or
7874 ANSI standards. When used at link-time, it may include libraries
7875 or startup files that change the default FPU control word or other
7876 similar optimizations.
7878 This option is not turned on by any @option{-O} option since
7879 it can result in incorrect output for programs which depend on
7880 an exact implementation of IEEE or ISO rules/specifications for
7881 math functions. It may, however, yield faster code for programs
7882 that do not require the guarantees of these specifications.
7883 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7884 @option{-fassociative-math} and @option{-freciprocal-math}.
7886 The default is @option{-fno-unsafe-math-optimizations}.
7888 @item -fassociative-math
7889 @opindex fassociative-math
7891 Allow re-association of operands in series of floating-point operations.
7892 This violates the ISO C and C++ language standard by possibly changing
7893 computation result. NOTE: re-ordering may change the sign of zero as
7894 well as ignore NaNs and inhibit or create underflow or overflow (and
7895 thus cannot be used on a code which relies on rounding behavior like
7896 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7897 and thus may not be used when ordered comparisons are required.
7898 This option requires that both @option{-fno-signed-zeros} and
7899 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7900 much sense with @option{-frounding-math}. For Fortran the option
7901 is automatically enabled when both @option{-fno-signed-zeros} and
7902 @option{-fno-trapping-math} are in effect.
7904 The default is @option{-fno-associative-math}.
7906 @item -freciprocal-math
7907 @opindex freciprocal-math
7909 Allow the reciprocal of a value to be used instead of dividing by
7910 the value if this enables optimizations. For example @code{x / y}
7911 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7912 is subject to common subexpression elimination. Note that this loses
7913 precision and increases the number of flops operating on the value.
7915 The default is @option{-fno-reciprocal-math}.
7917 @item -ffinite-math-only
7918 @opindex ffinite-math-only
7919 Allow optimizations for floating-point arithmetic that assume
7920 that arguments and results are not NaNs or +-Infs.
7922 This option is not turned on by any @option{-O} option since
7923 it can result in incorrect output for programs which depend on
7924 an exact implementation of IEEE or ISO rules/specifications for
7925 math functions. It may, however, yield faster code for programs
7926 that do not require the guarantees of these specifications.
7928 The default is @option{-fno-finite-math-only}.
7930 @item -fno-signed-zeros
7931 @opindex fno-signed-zeros
7932 Allow optimizations for floating point arithmetic that ignore the
7933 signedness of zero. IEEE arithmetic specifies the behavior of
7934 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7935 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7936 This option implies that the sign of a zero result isn't significant.
7938 The default is @option{-fsigned-zeros}.
7940 @item -fno-trapping-math
7941 @opindex fno-trapping-math
7942 Compile code assuming that floating-point operations cannot generate
7943 user-visible traps. These traps include division by zero, overflow,
7944 underflow, inexact result and invalid operation. This option requires
7945 that @option{-fno-signaling-nans} be in effect. Setting this option may
7946 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7948 This option should never be turned on by any @option{-O} option since
7949 it can result in incorrect output for programs which depend on
7950 an exact implementation of IEEE or ISO rules/specifications for
7953 The default is @option{-ftrapping-math}.
7955 @item -frounding-math
7956 @opindex frounding-math
7957 Disable transformations and optimizations that assume default floating
7958 point rounding behavior. This is round-to-zero for all floating point
7959 to integer conversions, and round-to-nearest for all other arithmetic
7960 truncations. This option should be specified for programs that change
7961 the FP rounding mode dynamically, or that may be executed with a
7962 non-default rounding mode. This option disables constant folding of
7963 floating point expressions at compile-time (which may be affected by
7964 rounding mode) and arithmetic transformations that are unsafe in the
7965 presence of sign-dependent rounding modes.
7967 The default is @option{-fno-rounding-math}.
7969 This option is experimental and does not currently guarantee to
7970 disable all GCC optimizations that are affected by rounding mode.
7971 Future versions of GCC may provide finer control of this setting
7972 using C99's @code{FENV_ACCESS} pragma. This command line option
7973 will be used to specify the default state for @code{FENV_ACCESS}.
7975 @item -fsignaling-nans
7976 @opindex fsignaling-nans
7977 Compile code assuming that IEEE signaling NaNs may generate user-visible
7978 traps during floating-point operations. Setting this option disables
7979 optimizations that may change the number of exceptions visible with
7980 signaling NaNs. This option implies @option{-ftrapping-math}.
7982 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7985 The default is @option{-fno-signaling-nans}.
7987 This option is experimental and does not currently guarantee to
7988 disable all GCC optimizations that affect signaling NaN behavior.
7990 @item -fsingle-precision-constant
7991 @opindex fsingle-precision-constant
7992 Treat floating point constant as single precision constant instead of
7993 implicitly converting it to double precision constant.
7995 @item -fcx-limited-range
7996 @opindex fcx-limited-range
7997 When enabled, this option states that a range reduction step is not
7998 needed when performing complex division. Also, there is no checking
7999 whether the result of a complex multiplication or division is @code{NaN
8000 + I*NaN}, with an attempt to rescue the situation in that case. The
8001 default is @option{-fno-cx-limited-range}, but is enabled by
8002 @option{-ffast-math}.
8004 This option controls the default setting of the ISO C99
8005 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8008 @item -fcx-fortran-rules
8009 @opindex fcx-fortran-rules
8010 Complex multiplication and division follow Fortran rules. Range
8011 reduction is done as part of complex division, but there is no checking
8012 whether the result of a complex multiplication or division is @code{NaN
8013 + I*NaN}, with an attempt to rescue the situation in that case.
8015 The default is @option{-fno-cx-fortran-rules}.
8019 The following options control optimizations that may improve
8020 performance, but are not enabled by any @option{-O} options. This
8021 section includes experimental options that may produce broken code.
8024 @item -fbranch-probabilities
8025 @opindex fbranch-probabilities
8026 After running a program compiled with @option{-fprofile-arcs}
8027 (@pxref{Debugging Options,, Options for Debugging Your Program or
8028 @command{gcc}}), you can compile it a second time using
8029 @option{-fbranch-probabilities}, to improve optimizations based on
8030 the number of times each branch was taken. When the program
8031 compiled with @option{-fprofile-arcs} exits it saves arc execution
8032 counts to a file called @file{@var{sourcename}.gcda} for each source
8033 file. The information in this data file is very dependent on the
8034 structure of the generated code, so you must use the same source code
8035 and the same optimization options for both compilations.
8037 With @option{-fbranch-probabilities}, GCC puts a
8038 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8039 These can be used to improve optimization. Currently, they are only
8040 used in one place: in @file{reorg.c}, instead of guessing which path a
8041 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8042 exactly determine which path is taken more often.
8044 @item -fprofile-values
8045 @opindex fprofile-values
8046 If combined with @option{-fprofile-arcs}, it adds code so that some
8047 data about values of expressions in the program is gathered.
8049 With @option{-fbranch-probabilities}, it reads back the data gathered
8050 from profiling values of expressions for usage in optimizations.
8052 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8056 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8057 a code to gather information about values of expressions.
8059 With @option{-fbranch-probabilities}, it reads back the data gathered
8060 and actually performs the optimizations based on them.
8061 Currently the optimizations include specialization of division operation
8062 using the knowledge about the value of the denominator.
8064 @item -frename-registers
8065 @opindex frename-registers
8066 Attempt to avoid false dependencies in scheduled code by making use
8067 of registers left over after register allocation. This optimization
8068 will most benefit processors with lots of registers. Depending on the
8069 debug information format adopted by the target, however, it can
8070 make debugging impossible, since variables will no longer stay in
8071 a ``home register''.
8073 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8077 Perform tail duplication to enlarge superblock size. This transformation
8078 simplifies the control flow of the function allowing other optimizations to do
8081 Enabled with @option{-fprofile-use}.
8083 @item -funroll-loops
8084 @opindex funroll-loops
8085 Unroll loops whose number of iterations can be determined at compile time or
8086 upon entry to the loop. @option{-funroll-loops} implies
8087 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8088 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8089 small constant number of iterations). This option makes code larger, and may
8090 or may not make it run faster.
8092 Enabled with @option{-fprofile-use}.
8094 @item -funroll-all-loops
8095 @opindex funroll-all-loops
8096 Unroll all loops, even if their number of iterations is uncertain when
8097 the loop is entered. This usually makes programs run more slowly.
8098 @option{-funroll-all-loops} implies the same options as
8099 @option{-funroll-loops}.
8102 @opindex fpeel-loops
8103 Peels the loops for that there is enough information that they do not
8104 roll much (from profile feedback). It also turns on complete loop peeling
8105 (i.e.@: complete removal of loops with small constant number of iterations).
8107 Enabled with @option{-fprofile-use}.
8109 @item -fmove-loop-invariants
8110 @opindex fmove-loop-invariants
8111 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8112 at level @option{-O1}
8114 @item -funswitch-loops
8115 @opindex funswitch-loops
8116 Move branches with loop invariant conditions out of the loop, with duplicates
8117 of the loop on both branches (modified according to result of the condition).
8119 @item -ffunction-sections
8120 @itemx -fdata-sections
8121 @opindex ffunction-sections
8122 @opindex fdata-sections
8123 Place each function or data item into its own section in the output
8124 file if the target supports arbitrary sections. The name of the
8125 function or the name of the data item determines the section's name
8128 Use these options on systems where the linker can perform optimizations
8129 to improve locality of reference in the instruction space. Most systems
8130 using the ELF object format and SPARC processors running Solaris 2 have
8131 linkers with such optimizations. AIX may have these optimizations in
8134 Only use these options when there are significant benefits from doing
8135 so. When you specify these options, the assembler and linker will
8136 create larger object and executable files and will also be slower.
8137 You will not be able to use @code{gprof} on all systems if you
8138 specify this option and you may have problems with debugging if
8139 you specify both this option and @option{-g}.
8141 @item -fbranch-target-load-optimize
8142 @opindex fbranch-target-load-optimize
8143 Perform branch target register load optimization before prologue / epilogue
8145 The use of target registers can typically be exposed only during reload,
8146 thus hoisting loads out of loops and doing inter-block scheduling needs
8147 a separate optimization pass.
8149 @item -fbranch-target-load-optimize2
8150 @opindex fbranch-target-load-optimize2
8151 Perform branch target register load optimization after prologue / epilogue
8154 @item -fbtr-bb-exclusive
8155 @opindex fbtr-bb-exclusive
8156 When performing branch target register load optimization, don't reuse
8157 branch target registers in within any basic block.
8159 @item -fstack-protector
8160 @opindex fstack-protector
8161 Emit extra code to check for buffer overflows, such as stack smashing
8162 attacks. This is done by adding a guard variable to functions with
8163 vulnerable objects. This includes functions that call alloca, and
8164 functions with buffers larger than 8 bytes. The guards are initialized
8165 when a function is entered and then checked when the function exits.
8166 If a guard check fails, an error message is printed and the program exits.
8168 @item -fstack-protector-all
8169 @opindex fstack-protector-all
8170 Like @option{-fstack-protector} except that all functions are protected.
8172 @item -fsection-anchors
8173 @opindex fsection-anchors
8174 Try to reduce the number of symbolic address calculations by using
8175 shared ``anchor'' symbols to address nearby objects. This transformation
8176 can help to reduce the number of GOT entries and GOT accesses on some
8179 For example, the implementation of the following function @code{foo}:
8183 int foo (void) @{ return a + b + c; @}
8186 would usually calculate the addresses of all three variables, but if you
8187 compile it with @option{-fsection-anchors}, it will access the variables
8188 from a common anchor point instead. The effect is similar to the
8189 following pseudocode (which isn't valid C):
8194 register int *xr = &x;
8195 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8199 Not all targets support this option.
8201 @item --param @var{name}=@var{value}
8203 In some places, GCC uses various constants to control the amount of
8204 optimization that is done. For example, GCC will not inline functions
8205 that contain more that a certain number of instructions. You can
8206 control some of these constants on the command-line using the
8207 @option{--param} option.
8209 The names of specific parameters, and the meaning of the values, are
8210 tied to the internals of the compiler, and are subject to change
8211 without notice in future releases.
8213 In each case, the @var{value} is an integer. The allowable choices for
8214 @var{name} are given in the following table:
8217 @item predictable-branch-outcome
8218 When branch is predicted to be taken with probability lower than this threshold
8219 (in percent), then it is considered well predictable. The default is 10.
8221 @item max-crossjump-edges
8222 The maximum number of incoming edges to consider for crossjumping.
8223 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8224 the number of edges incoming to each block. Increasing values mean
8225 more aggressive optimization, making the compile time increase with
8226 probably small improvement in executable size.
8228 @item min-crossjump-insns
8229 The minimum number of instructions which must be matched at the end
8230 of two blocks before crossjumping will be performed on them. This
8231 value is ignored in the case where all instructions in the block being
8232 crossjumped from are matched. The default value is 5.
8234 @item max-grow-copy-bb-insns
8235 The maximum code size expansion factor when copying basic blocks
8236 instead of jumping. The expansion is relative to a jump instruction.
8237 The default value is 8.
8239 @item max-goto-duplication-insns
8240 The maximum number of instructions to duplicate to a block that jumps
8241 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8242 passes, GCC factors computed gotos early in the compilation process,
8243 and unfactors them as late as possible. Only computed jumps at the
8244 end of a basic blocks with no more than max-goto-duplication-insns are
8245 unfactored. The default value is 8.
8247 @item max-delay-slot-insn-search
8248 The maximum number of instructions to consider when looking for an
8249 instruction to fill a delay slot. If more than this arbitrary number of
8250 instructions is searched, the time savings from filling the delay slot
8251 will be minimal so stop searching. Increasing values mean more
8252 aggressive optimization, making the compile time increase with probably
8253 small improvement in executable run time.
8255 @item max-delay-slot-live-search
8256 When trying to fill delay slots, the maximum number of instructions to
8257 consider when searching for a block with valid live register
8258 information. Increasing this arbitrarily chosen value means more
8259 aggressive optimization, increasing the compile time. This parameter
8260 should be removed when the delay slot code is rewritten to maintain the
8263 @item max-gcse-memory
8264 The approximate maximum amount of memory that will be allocated in
8265 order to perform the global common subexpression elimination
8266 optimization. If more memory than specified is required, the
8267 optimization will not be done.
8269 @item max-gcse-insertion-ratio
8270 If the ratio of expression insertions to deletions is larger than this value
8271 for any expression, then RTL PRE will insert or remove the expression and thus
8272 leave partially redundant computations in the instruction stream. The default value is 20.
8274 @item max-pending-list-length
8275 The maximum number of pending dependencies scheduling will allow
8276 before flushing the current state and starting over. Large functions
8277 with few branches or calls can create excessively large lists which
8278 needlessly consume memory and resources.
8280 @item max-inline-insns-single
8281 Several parameters control the tree inliner used in gcc.
8282 This number sets the maximum number of instructions (counted in GCC's
8283 internal representation) in a single function that the tree inliner
8284 will consider for inlining. This only affects functions declared
8285 inline and methods implemented in a class declaration (C++).
8286 The default value is 400.
8288 @item max-inline-insns-auto
8289 When you use @option{-finline-functions} (included in @option{-O3}),
8290 a lot of functions that would otherwise not be considered for inlining
8291 by the compiler will be investigated. To those functions, a different
8292 (more restrictive) limit compared to functions declared inline can
8294 The default value is 40.
8296 @item large-function-insns
8297 The limit specifying really large functions. For functions larger than this
8298 limit after inlining, inlining is constrained by
8299 @option{--param large-function-growth}. This parameter is useful primarily
8300 to avoid extreme compilation time caused by non-linear algorithms used by the
8302 The default value is 2700.
8304 @item large-function-growth
8305 Specifies maximal growth of large function caused by inlining in percents.
8306 The default value is 100 which limits large function growth to 2.0 times
8309 @item large-unit-insns
8310 The limit specifying large translation unit. Growth caused by inlining of
8311 units larger than this limit is limited by @option{--param inline-unit-growth}.
8312 For small units this might be too tight (consider unit consisting of function A
8313 that is inline and B that just calls A three time. If B is small relative to
8314 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8315 large units consisting of small inlineable functions however the overall unit
8316 growth limit is needed to avoid exponential explosion of code size. Thus for
8317 smaller units, the size is increased to @option{--param large-unit-insns}
8318 before applying @option{--param inline-unit-growth}. The default is 10000
8320 @item inline-unit-growth
8321 Specifies maximal overall growth of the compilation unit caused by inlining.
8322 The default value is 30 which limits unit growth to 1.3 times the original
8325 @item ipcp-unit-growth
8326 Specifies maximal overall growth of the compilation unit caused by
8327 interprocedural constant propagation. The default value is 10 which limits
8328 unit growth to 1.1 times the original size.
8330 @item large-stack-frame
8331 The limit specifying large stack frames. While inlining the algorithm is trying
8332 to not grow past this limit too much. Default value is 256 bytes.
8334 @item large-stack-frame-growth
8335 Specifies maximal growth of large stack frames caused by inlining in percents.
8336 The default value is 1000 which limits large stack frame growth to 11 times
8339 @item max-inline-insns-recursive
8340 @itemx max-inline-insns-recursive-auto
8341 Specifies maximum number of instructions out-of-line copy of self recursive inline
8342 function can grow into by performing recursive inlining.
8344 For functions declared inline @option{--param max-inline-insns-recursive} is
8345 taken into account. For function not declared inline, recursive inlining
8346 happens only when @option{-finline-functions} (included in @option{-O3}) is
8347 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8348 default value is 450.
8350 @item max-inline-recursive-depth
8351 @itemx max-inline-recursive-depth-auto
8352 Specifies maximum recursion depth used by the recursive inlining.
8354 For functions declared inline @option{--param max-inline-recursive-depth} is
8355 taken into account. For function not declared inline, recursive inlining
8356 happens only when @option{-finline-functions} (included in @option{-O3}) is
8357 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8360 @item min-inline-recursive-probability
8361 Recursive inlining is profitable only for function having deep recursion
8362 in average and can hurt for function having little recursion depth by
8363 increasing the prologue size or complexity of function body to other
8366 When profile feedback is available (see @option{-fprofile-generate}) the actual
8367 recursion depth can be guessed from probability that function will recurse via
8368 given call expression. This parameter limits inlining only to call expression
8369 whose probability exceeds given threshold (in percents). The default value is
8372 @item early-inlining-insns
8373 Specify growth that early inliner can make. In effect it increases amount of
8374 inlining for code having large abstraction penalty. The default value is 10.
8376 @item max-early-inliner-iterations
8377 @itemx max-early-inliner-iterations
8378 Limit of iterations of early inliner. This basically bounds number of nested
8379 indirect calls early inliner can resolve. Deeper chains are still handled by
8382 @item comdat-sharing-probability
8383 @itemx comdat-sharing-probability
8384 Probability (in percent) that C++ inline function with comdat visibility
8385 will be shared across multiple compilation units. The default value is 20.
8387 @item min-vect-loop-bound
8388 The minimum number of iterations under which a loop will not get vectorized
8389 when @option{-ftree-vectorize} is used. The number of iterations after
8390 vectorization needs to be greater than the value specified by this option
8391 to allow vectorization. The default value is 0.
8393 @item gcse-cost-distance-ratio
8394 Scaling factor in calculation of maximum distance an expression
8395 can be moved by GCSE optimizations. This is currently supported only in the
8396 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8397 will be with simple expressions, i.e., the expressions which have cost
8398 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8399 hoisting of simple expressions. The default value is 10.
8401 @item gcse-unrestricted-cost
8402 Cost, roughly measured as the cost of a single typical machine
8403 instruction, at which GCSE optimizations will not constrain
8404 the distance an expression can travel. This is currently
8405 supported only in the code hoisting pass. The lesser the cost,
8406 the more aggressive code hoisting will be. Specifying 0 will
8407 allow all expressions to travel unrestricted distances.
8408 The default value is 3.
8410 @item max-hoist-depth
8411 The depth of search in the dominator tree for expressions to hoist.
8412 This is used to avoid quadratic behavior in hoisting algorithm.
8413 The value of 0 will avoid limiting the search, but may slow down compilation
8414 of huge functions. The default value is 30.
8416 @item max-unrolled-insns
8417 The maximum number of instructions that a loop should have if that loop
8418 is unrolled, and if the loop is unrolled, it determines how many times
8419 the loop code is unrolled.
8421 @item max-average-unrolled-insns
8422 The maximum number of instructions biased by probabilities of their execution
8423 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8424 it determines how many times the loop code is unrolled.
8426 @item max-unroll-times
8427 The maximum number of unrollings of a single loop.
8429 @item max-peeled-insns
8430 The maximum number of instructions that a loop should have if that loop
8431 is peeled, and if the loop is peeled, it determines how many times
8432 the loop code is peeled.
8434 @item max-peel-times
8435 The maximum number of peelings of a single loop.
8437 @item max-completely-peeled-insns
8438 The maximum number of insns of a completely peeled loop.
8440 @item max-completely-peel-times
8441 The maximum number of iterations of a loop to be suitable for complete peeling.
8443 @item max-completely-peel-loop-nest-depth
8444 The maximum depth of a loop nest suitable for complete peeling.
8446 @item max-unswitch-insns
8447 The maximum number of insns of an unswitched loop.
8449 @item max-unswitch-level
8450 The maximum number of branches unswitched in a single loop.
8453 The minimum cost of an expensive expression in the loop invariant motion.
8455 @item iv-consider-all-candidates-bound
8456 Bound on number of candidates for induction variables below that
8457 all candidates are considered for each use in induction variable
8458 optimizations. Only the most relevant candidates are considered
8459 if there are more candidates, to avoid quadratic time complexity.
8461 @item iv-max-considered-uses
8462 The induction variable optimizations give up on loops that contain more
8463 induction variable uses.
8465 @item iv-always-prune-cand-set-bound
8466 If number of candidates in the set is smaller than this value,
8467 we always try to remove unnecessary ivs from the set during its
8468 optimization when a new iv is added to the set.
8470 @item scev-max-expr-size
8471 Bound on size of expressions used in the scalar evolutions analyzer.
8472 Large expressions slow the analyzer.
8474 @item scev-max-expr-complexity
8475 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8476 Complex expressions slow the analyzer.
8478 @item omega-max-vars
8479 The maximum number of variables in an Omega constraint system.
8480 The default value is 128.
8482 @item omega-max-geqs
8483 The maximum number of inequalities in an Omega constraint system.
8484 The default value is 256.
8487 The maximum number of equalities in an Omega constraint system.
8488 The default value is 128.
8490 @item omega-max-wild-cards
8491 The maximum number of wildcard variables that the Omega solver will
8492 be able to insert. The default value is 18.
8494 @item omega-hash-table-size
8495 The size of the hash table in the Omega solver. The default value is
8498 @item omega-max-keys
8499 The maximal number of keys used by the Omega solver. The default
8502 @item omega-eliminate-redundant-constraints
8503 When set to 1, use expensive methods to eliminate all redundant
8504 constraints. The default value is 0.
8506 @item vect-max-version-for-alignment-checks
8507 The maximum number of runtime checks that can be performed when
8508 doing loop versioning for alignment in the vectorizer. See option
8509 ftree-vect-loop-version for more information.
8511 @item vect-max-version-for-alias-checks
8512 The maximum number of runtime checks that can be performed when
8513 doing loop versioning for alias in the vectorizer. See option
8514 ftree-vect-loop-version for more information.
8516 @item max-iterations-to-track
8518 The maximum number of iterations of a loop the brute force algorithm
8519 for analysis of # of iterations of the loop tries to evaluate.
8521 @item hot-bb-count-fraction
8522 Select fraction of the maximal count of repetitions of basic block in program
8523 given basic block needs to have to be considered hot.
8525 @item hot-bb-frequency-fraction
8526 Select fraction of the entry block frequency of executions of basic block in
8527 function given basic block needs to have to be considered hot.
8529 @item max-predicted-iterations
8530 The maximum number of loop iterations we predict statically. This is useful
8531 in cases where function contain single loop with known bound and other loop
8532 with unknown. We predict the known number of iterations correctly, while
8533 the unknown number of iterations average to roughly 10. This means that the
8534 loop without bounds would appear artificially cold relative to the other one.
8536 @item align-threshold
8538 Select fraction of the maximal frequency of executions of basic block in
8539 function given basic block will get aligned.
8541 @item align-loop-iterations
8543 A loop expected to iterate at lest the selected number of iterations will get
8546 @item tracer-dynamic-coverage
8547 @itemx tracer-dynamic-coverage-feedback
8549 This value is used to limit superblock formation once the given percentage of
8550 executed instructions is covered. This limits unnecessary code size
8553 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8554 feedback is available. The real profiles (as opposed to statically estimated
8555 ones) are much less balanced allowing the threshold to be larger value.
8557 @item tracer-max-code-growth
8558 Stop tail duplication once code growth has reached given percentage. This is
8559 rather hokey argument, as most of the duplicates will be eliminated later in
8560 cross jumping, so it may be set to much higher values than is the desired code
8563 @item tracer-min-branch-ratio
8565 Stop reverse growth when the reverse probability of best edge is less than this
8566 threshold (in percent).
8568 @item tracer-min-branch-ratio
8569 @itemx tracer-min-branch-ratio-feedback
8571 Stop forward growth if the best edge do have probability lower than this
8574 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8575 compilation for profile feedback and one for compilation without. The value
8576 for compilation with profile feedback needs to be more conservative (higher) in
8577 order to make tracer effective.
8579 @item max-cse-path-length
8581 Maximum number of basic blocks on path that cse considers. The default is 10.
8584 The maximum instructions CSE process before flushing. The default is 1000.
8586 @item ggc-min-expand
8588 GCC uses a garbage collector to manage its own memory allocation. This
8589 parameter specifies the minimum percentage by which the garbage
8590 collector's heap should be allowed to expand between collections.
8591 Tuning this may improve compilation speed; it has no effect on code
8594 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8595 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8596 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8597 GCC is not able to calculate RAM on a particular platform, the lower
8598 bound of 30% is used. Setting this parameter and
8599 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8600 every opportunity. This is extremely slow, but can be useful for
8603 @item ggc-min-heapsize
8605 Minimum size of the garbage collector's heap before it begins bothering
8606 to collect garbage. The first collection occurs after the heap expands
8607 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8608 tuning this may improve compilation speed, and has no effect on code
8611 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8612 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8613 with a lower bound of 4096 (four megabytes) and an upper bound of
8614 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8615 particular platform, the lower bound is used. Setting this parameter
8616 very large effectively disables garbage collection. Setting this
8617 parameter and @option{ggc-min-expand} to zero causes a full collection
8618 to occur at every opportunity.
8620 @item max-reload-search-insns
8621 The maximum number of instruction reload should look backward for equivalent
8622 register. Increasing values mean more aggressive optimization, making the
8623 compile time increase with probably slightly better performance. The default
8626 @item max-cselib-memory-locations
8627 The maximum number of memory locations cselib should take into account.
8628 Increasing values mean more aggressive optimization, making the compile time
8629 increase with probably slightly better performance. The default value is 500.
8631 @item reorder-blocks-duplicate
8632 @itemx reorder-blocks-duplicate-feedback
8634 Used by basic block reordering pass to decide whether to use unconditional
8635 branch or duplicate the code on its destination. Code is duplicated when its
8636 estimated size is smaller than this value multiplied by the estimated size of
8637 unconditional jump in the hot spots of the program.
8639 The @option{reorder-block-duplicate-feedback} is used only when profile
8640 feedback is available and may be set to higher values than
8641 @option{reorder-block-duplicate} since information about the hot spots is more
8644 @item max-sched-ready-insns
8645 The maximum number of instructions ready to be issued the scheduler should
8646 consider at any given time during the first scheduling pass. Increasing
8647 values mean more thorough searches, making the compilation time increase
8648 with probably little benefit. The default value is 100.
8650 @item max-sched-region-blocks
8651 The maximum number of blocks in a region to be considered for
8652 interblock scheduling. The default value is 10.
8654 @item max-pipeline-region-blocks
8655 The maximum number of blocks in a region to be considered for
8656 pipelining in the selective scheduler. The default value is 15.
8658 @item max-sched-region-insns
8659 The maximum number of insns in a region to be considered for
8660 interblock scheduling. The default value is 100.
8662 @item max-pipeline-region-insns
8663 The maximum number of insns in a region to be considered for
8664 pipelining in the selective scheduler. The default value is 200.
8667 The minimum probability (in percents) of reaching a source block
8668 for interblock speculative scheduling. The default value is 40.
8670 @item max-sched-extend-regions-iters
8671 The maximum number of iterations through CFG to extend regions.
8672 0 - disable region extension,
8673 N - do at most N iterations.
8674 The default value is 0.
8676 @item max-sched-insn-conflict-delay
8677 The maximum conflict delay for an insn to be considered for speculative motion.
8678 The default value is 3.
8680 @item sched-spec-prob-cutoff
8681 The minimal probability of speculation success (in percents), so that
8682 speculative insn will be scheduled.
8683 The default value is 40.
8685 @item sched-mem-true-dep-cost
8686 Minimal distance (in CPU cycles) between store and load targeting same
8687 memory locations. The default value is 1.
8689 @item selsched-max-lookahead
8690 The maximum size of the lookahead window of selective scheduling. It is a
8691 depth of search for available instructions.
8692 The default value is 50.
8694 @item selsched-max-sched-times
8695 The maximum number of times that an instruction will be scheduled during
8696 selective scheduling. This is the limit on the number of iterations
8697 through which the instruction may be pipelined. The default value is 2.
8699 @item selsched-max-insns-to-rename
8700 The maximum number of best instructions in the ready list that are considered
8701 for renaming in the selective scheduler. The default value is 2.
8703 @item max-last-value-rtl
8704 The maximum size measured as number of RTLs that can be recorded in an expression
8705 in combiner for a pseudo register as last known value of that register. The default
8708 @item integer-share-limit
8709 Small integer constants can use a shared data structure, reducing the
8710 compiler's memory usage and increasing its speed. This sets the maximum
8711 value of a shared integer constant. The default value is 256.
8713 @item min-virtual-mappings
8714 Specifies the minimum number of virtual mappings in the incremental
8715 SSA updater that should be registered to trigger the virtual mappings
8716 heuristic defined by virtual-mappings-ratio. The default value is
8719 @item virtual-mappings-ratio
8720 If the number of virtual mappings is virtual-mappings-ratio bigger
8721 than the number of virtual symbols to be updated, then the incremental
8722 SSA updater switches to a full update for those symbols. The default
8725 @item ssp-buffer-size
8726 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8727 protection when @option{-fstack-protection} is used.
8729 @item max-jump-thread-duplication-stmts
8730 Maximum number of statements allowed in a block that needs to be
8731 duplicated when threading jumps.
8733 @item max-fields-for-field-sensitive
8734 Maximum number of fields in a structure we will treat in
8735 a field sensitive manner during pointer analysis. The default is zero
8736 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8738 @item prefetch-latency
8739 Estimate on average number of instructions that are executed before
8740 prefetch finishes. The distance we prefetch ahead is proportional
8741 to this constant. Increasing this number may also lead to less
8742 streams being prefetched (see @option{simultaneous-prefetches}).
8744 @item simultaneous-prefetches
8745 Maximum number of prefetches that can run at the same time.
8747 @item l1-cache-line-size
8748 The size of cache line in L1 cache, in bytes.
8751 The size of L1 cache, in kilobytes.
8754 The size of L2 cache, in kilobytes.
8756 @item min-insn-to-prefetch-ratio
8757 The minimum ratio between the number of instructions and the
8758 number of prefetches to enable prefetching in a loop.
8760 @item prefetch-min-insn-to-mem-ratio
8761 The minimum ratio between the number of instructions and the
8762 number of memory references to enable prefetching in a loop.
8764 @item use-canonical-types
8765 Whether the compiler should use the ``canonical'' type system. By
8766 default, this should always be 1, which uses a more efficient internal
8767 mechanism for comparing types in C++ and Objective-C++. However, if
8768 bugs in the canonical type system are causing compilation failures,
8769 set this value to 0 to disable canonical types.
8771 @item switch-conversion-max-branch-ratio
8772 Switch initialization conversion will refuse to create arrays that are
8773 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8774 branches in the switch.
8776 @item max-partial-antic-length
8777 Maximum length of the partial antic set computed during the tree
8778 partial redundancy elimination optimization (@option{-ftree-pre}) when
8779 optimizing at @option{-O3} and above. For some sorts of source code
8780 the enhanced partial redundancy elimination optimization can run away,
8781 consuming all of the memory available on the host machine. This
8782 parameter sets a limit on the length of the sets that are computed,
8783 which prevents the runaway behavior. Setting a value of 0 for
8784 this parameter will allow an unlimited set length.
8786 @item sccvn-max-scc-size
8787 Maximum size of a strongly connected component (SCC) during SCCVN
8788 processing. If this limit is hit, SCCVN processing for the whole
8789 function will not be done and optimizations depending on it will
8790 be disabled. The default maximum SCC size is 10000.
8792 @item ira-max-loops-num
8793 IRA uses a regional register allocation by default. If a function
8794 contains loops more than number given by the parameter, only at most
8795 given number of the most frequently executed loops will form regions
8796 for the regional register allocation. The default value of the
8799 @item ira-max-conflict-table-size
8800 Although IRA uses a sophisticated algorithm of compression conflict
8801 table, the table can be still big for huge functions. If the conflict
8802 table for a function could be more than size in MB given by the
8803 parameter, the conflict table is not built and faster, simpler, and
8804 lower quality register allocation algorithm will be used. The
8805 algorithm do not use pseudo-register conflicts. The default value of
8806 the parameter is 2000.
8808 @item ira-loop-reserved-regs
8809 IRA can be used to evaluate more accurate register pressure in loops
8810 for decision to move loop invariants (see @option{-O3}). The number
8811 of available registers reserved for some other purposes is described
8812 by this parameter. The default value of the parameter is 2 which is
8813 minimal number of registers needed for execution of typical
8814 instruction. This value is the best found from numerous experiments.
8816 @item loop-invariant-max-bbs-in-loop
8817 Loop invariant motion can be very expensive, both in compile time and
8818 in amount of needed compile time memory, with very large loops. Loops
8819 with more basic blocks than this parameter won't have loop invariant
8820 motion optimization performed on them. The default value of the
8821 parameter is 1000 for -O1 and 10000 for -O2 and above.
8823 @item max-vartrack-size
8824 Sets a maximum number of hash table slots to use during variable
8825 tracking dataflow analysis of any function. If this limit is exceeded
8826 with variable tracking at assignments enabled, analysis for that
8827 function is retried without it, after removing all debug insns from
8828 the function. If the limit is exceeded even without debug insns, var
8829 tracking analysis is completely disabled for the function. Setting
8830 the parameter to zero makes it unlimited.
8832 @item min-nondebug-insn-uid
8833 Use uids starting at this parameter for nondebug insns. The range below
8834 the parameter is reserved exclusively for debug insns created by
8835 @option{-fvar-tracking-assignments}, but debug insns may get
8836 (non-overlapping) uids above it if the reserved range is exhausted.
8838 @item ipa-sra-ptr-growth-factor
8839 IPA-SRA will replace a pointer to an aggregate with one or more new
8840 parameters only when their cumulative size is less or equal to
8841 @option{ipa-sra-ptr-growth-factor} times the size of the original
8844 @item graphite-max-nb-scop-params
8845 To avoid exponential effects in the Graphite loop transforms, the
8846 number of parameters in a Static Control Part (SCoP) is bounded. The
8847 default value is 10 parameters. A variable whose value is unknown at
8848 compile time and defined outside a SCoP is a parameter of the SCoP.
8850 @item graphite-max-bbs-per-function
8851 To avoid exponential effects in the detection of SCoPs, the size of
8852 the functions analyzed by Graphite is bounded. The default value is
8855 @item loop-block-tile-size
8856 Loop blocking or strip mining transforms, enabled with
8857 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8858 loop in the loop nest by a given number of iterations. The strip
8859 length can be changed using the @option{loop-block-tile-size}
8860 parameter. The default value is 51 iterations.
8862 @item devirt-type-list-size
8863 IPA-CP attempts to track all possible types passed to a function's
8864 parameter in order to perform devirtualization.
8865 @option{devirt-type-list-size} is the maximum number of types it
8866 stores per a single formal parameter of a function.
8868 @item lto-partitions
8869 Specify desired number of partitions produced during WHOPR compilation.
8870 The number of partitions should exceed the number of CPUs used for compilation.
8871 The default value is 32.
8873 @item lto-minpartition
8874 Size of minimal partition for WHOPR (in estimated instructions).
8875 This prevents expenses of splitting very small programs into too many
8878 @item cxx-max-namespaces-for-diagnostic-help
8879 The maximum number of namespaces to consult for suggestions when C++
8880 name lookup fails for an identifier. The default is 1000.
8882 @item max-stores-to-sink
8883 The maximum number of conditional stores paires that can be sunk. Set to 0
8884 if either vectorization (@option{-ftree-vectorize}) or if-conversion
8885 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
8890 @node Preprocessor Options
8891 @section Options Controlling the Preprocessor
8892 @cindex preprocessor options
8893 @cindex options, preprocessor
8895 These options control the C preprocessor, which is run on each C source
8896 file before actual compilation.
8898 If you use the @option{-E} option, nothing is done except preprocessing.
8899 Some of these options make sense only together with @option{-E} because
8900 they cause the preprocessor output to be unsuitable for actual
8904 @item -Wp,@var{option}
8906 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8907 and pass @var{option} directly through to the preprocessor. If
8908 @var{option} contains commas, it is split into multiple options at the
8909 commas. However, many options are modified, translated or interpreted
8910 by the compiler driver before being passed to the preprocessor, and
8911 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8912 interface is undocumented and subject to change, so whenever possible
8913 you should avoid using @option{-Wp} and let the driver handle the
8916 @item -Xpreprocessor @var{option}
8917 @opindex Xpreprocessor
8918 Pass @var{option} as an option to the preprocessor. You can use this to
8919 supply system-specific preprocessor options which GCC does not know how to
8922 If you want to pass an option that takes an argument, you must use
8923 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8926 @include cppopts.texi
8928 @node Assembler Options
8929 @section Passing Options to the Assembler
8931 @c prevent bad page break with this line
8932 You can pass options to the assembler.
8935 @item -Wa,@var{option}
8937 Pass @var{option} as an option to the assembler. If @var{option}
8938 contains commas, it is split into multiple options at the commas.
8940 @item -Xassembler @var{option}
8942 Pass @var{option} as an option to the assembler. You can use this to
8943 supply system-specific assembler options which GCC does not know how to
8946 If you want to pass an option that takes an argument, you must use
8947 @option{-Xassembler} twice, once for the option and once for the argument.
8952 @section Options for Linking
8953 @cindex link options
8954 @cindex options, linking
8956 These options come into play when the compiler links object files into
8957 an executable output file. They are meaningless if the compiler is
8958 not doing a link step.
8962 @item @var{object-file-name}
8963 A file name that does not end in a special recognized suffix is
8964 considered to name an object file or library. (Object files are
8965 distinguished from libraries by the linker according to the file
8966 contents.) If linking is done, these object files are used as input
8975 If any of these options is used, then the linker is not run, and
8976 object file names should not be used as arguments. @xref{Overall
8980 @item -l@var{library}
8981 @itemx -l @var{library}
8983 Search the library named @var{library} when linking. (The second
8984 alternative with the library as a separate argument is only for
8985 POSIX compliance and is not recommended.)
8987 It makes a difference where in the command you write this option; the
8988 linker searches and processes libraries and object files in the order they
8989 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8990 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8991 to functions in @samp{z}, those functions may not be loaded.
8993 The linker searches a standard list of directories for the library,
8994 which is actually a file named @file{lib@var{library}.a}. The linker
8995 then uses this file as if it had been specified precisely by name.
8997 The directories searched include several standard system directories
8998 plus any that you specify with @option{-L}.
9000 Normally the files found this way are library files---archive files
9001 whose members are object files. The linker handles an archive file by
9002 scanning through it for members which define symbols that have so far
9003 been referenced but not defined. But if the file that is found is an
9004 ordinary object file, it is linked in the usual fashion. The only
9005 difference between using an @option{-l} option and specifying a file name
9006 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9007 and searches several directories.
9011 You need this special case of the @option{-l} option in order to
9012 link an Objective-C or Objective-C++ program.
9015 @opindex nostartfiles
9016 Do not use the standard system startup files when linking.
9017 The standard system libraries are used normally, unless @option{-nostdlib}
9018 or @option{-nodefaultlibs} is used.
9020 @item -nodefaultlibs
9021 @opindex nodefaultlibs
9022 Do not use the standard system libraries when linking.
9023 Only the libraries you specify will be passed to the linker, options
9024 specifying linkage of the system libraries, such as @code{-static-libgcc}
9025 or @code{-shared-libgcc}, will be ignored.
9026 The standard startup files are used normally, unless @option{-nostartfiles}
9027 is used. The compiler may generate calls to @code{memcmp},
9028 @code{memset}, @code{memcpy} and @code{memmove}.
9029 These entries are usually resolved by entries in
9030 libc. These entry points should be supplied through some other
9031 mechanism when this option is specified.
9035 Do not use the standard system startup files or libraries when linking.
9036 No startup files and only the libraries you specify will be passed to
9037 the linker, options specifying linkage of the system libraries, such as
9038 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9039 The compiler may generate calls to @code{memcmp}, @code{memset},
9040 @code{memcpy} and @code{memmove}.
9041 These entries are usually resolved by entries in
9042 libc. These entry points should be supplied through some other
9043 mechanism when this option is specified.
9045 @cindex @option{-lgcc}, use with @option{-nostdlib}
9046 @cindex @option{-nostdlib} and unresolved references
9047 @cindex unresolved references and @option{-nostdlib}
9048 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9049 @cindex @option{-nodefaultlibs} and unresolved references
9050 @cindex unresolved references and @option{-nodefaultlibs}
9051 One of the standard libraries bypassed by @option{-nostdlib} and
9052 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9053 that GCC uses to overcome shortcomings of particular machines, or special
9054 needs for some languages.
9055 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9056 Collection (GCC) Internals},
9057 for more discussion of @file{libgcc.a}.)
9058 In most cases, you need @file{libgcc.a} even when you want to avoid
9059 other standard libraries. In other words, when you specify @option{-nostdlib}
9060 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9061 This ensures that you have no unresolved references to internal GCC
9062 library subroutines. (For example, @samp{__main}, used to ensure C++
9063 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9064 GNU Compiler Collection (GCC) Internals}.)
9068 Produce a position independent executable on targets which support it.
9069 For predictable results, you must also specify the same set of options
9070 that were used to generate code (@option{-fpie}, @option{-fPIE},
9071 or model suboptions) when you specify this option.
9075 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9076 that support it. This instructs the linker to add all symbols, not
9077 only used ones, to the dynamic symbol table. This option is needed
9078 for some uses of @code{dlopen} or to allow obtaining backtraces
9079 from within a program.
9083 Remove all symbol table and relocation information from the executable.
9087 On systems that support dynamic linking, this prevents linking with the shared
9088 libraries. On other systems, this option has no effect.
9092 Produce a shared object which can then be linked with other objects to
9093 form an executable. Not all systems support this option. For predictable
9094 results, you must also specify the same set of options that were used to
9095 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9096 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9097 needs to build supplementary stub code for constructors to work. On
9098 multi-libbed systems, @samp{gcc -shared} must select the correct support
9099 libraries to link against. Failing to supply the correct flags may lead
9100 to subtle defects. Supplying them in cases where they are not necessary
9103 @item -shared-libgcc
9104 @itemx -static-libgcc
9105 @opindex shared-libgcc
9106 @opindex static-libgcc
9107 On systems that provide @file{libgcc} as a shared library, these options
9108 force the use of either the shared or static version respectively.
9109 If no shared version of @file{libgcc} was built when the compiler was
9110 configured, these options have no effect.
9112 There are several situations in which an application should use the
9113 shared @file{libgcc} instead of the static version. The most common
9114 of these is when the application wishes to throw and catch exceptions
9115 across different shared libraries. In that case, each of the libraries
9116 as well as the application itself should use the shared @file{libgcc}.
9118 Therefore, the G++ and GCJ drivers automatically add
9119 @option{-shared-libgcc} whenever you build a shared library or a main
9120 executable, because C++ and Java programs typically use exceptions, so
9121 this is the right thing to do.
9123 If, instead, you use the GCC driver to create shared libraries, you may
9124 find that they will not always be linked with the shared @file{libgcc}.
9125 If GCC finds, at its configuration time, that you have a non-GNU linker
9126 or a GNU linker that does not support option @option{--eh-frame-hdr},
9127 it will link the shared version of @file{libgcc} into shared libraries
9128 by default. Otherwise, it will take advantage of the linker and optimize
9129 away the linking with the shared version of @file{libgcc}, linking with
9130 the static version of libgcc by default. This allows exceptions to
9131 propagate through such shared libraries, without incurring relocation
9132 costs at library load time.
9134 However, if a library or main executable is supposed to throw or catch
9135 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9136 for the languages used in the program, or using the option
9137 @option{-shared-libgcc}, such that it is linked with the shared
9140 @item -static-libstdc++
9141 When the @command{g++} program is used to link a C++ program, it will
9142 normally automatically link against @option{libstdc++}. If
9143 @file{libstdc++} is available as a shared library, and the
9144 @option{-static} option is not used, then this will link against the
9145 shared version of @file{libstdc++}. That is normally fine. However, it
9146 is sometimes useful to freeze the version of @file{libstdc++} used by
9147 the program without going all the way to a fully static link. The
9148 @option{-static-libstdc++} option directs the @command{g++} driver to
9149 link @file{libstdc++} statically, without necessarily linking other
9150 libraries statically.
9154 Bind references to global symbols when building a shared object. Warn
9155 about any unresolved references (unless overridden by the link editor
9156 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9159 @item -T @var{script}
9161 @cindex linker script
9162 Use @var{script} as the linker script. This option is supported by most
9163 systems using the GNU linker. On some targets, such as bare-board
9164 targets without an operating system, the @option{-T} option may be required
9165 when linking to avoid references to undefined symbols.
9167 @item -Xlinker @var{option}
9169 Pass @var{option} as an option to the linker. You can use this to
9170 supply system-specific linker options which GCC does not know how to
9173 If you want to pass an option that takes a separate argument, you must use
9174 @option{-Xlinker} twice, once for the option and once for the argument.
9175 For example, to pass @option{-assert definitions}, you must write
9176 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9177 @option{-Xlinker "-assert definitions"}, because this passes the entire
9178 string as a single argument, which is not what the linker expects.
9180 When using the GNU linker, it is usually more convenient to pass
9181 arguments to linker options using the @option{@var{option}=@var{value}}
9182 syntax than as separate arguments. For example, you can specify
9183 @samp{-Xlinker -Map=output.map} rather than
9184 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9185 this syntax for command-line options.
9187 @item -Wl,@var{option}
9189 Pass @var{option} as an option to the linker. If @var{option} contains
9190 commas, it is split into multiple options at the commas. You can use this
9191 syntax to pass an argument to the option.
9192 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9193 linker. When using the GNU linker, you can also get the same effect with
9194 @samp{-Wl,-Map=output.map}.
9196 @item -u @var{symbol}
9198 Pretend the symbol @var{symbol} is undefined, to force linking of
9199 library modules to define it. You can use @option{-u} multiple times with
9200 different symbols to force loading of additional library modules.
9203 @node Directory Options
9204 @section Options for Directory Search
9205 @cindex directory options
9206 @cindex options, directory search
9209 These options specify directories to search for header files, for
9210 libraries and for parts of the compiler:
9215 Add the directory @var{dir} to the head of the list of directories to be
9216 searched for header files. This can be used to override a system header
9217 file, substituting your own version, since these directories are
9218 searched before the system header file directories. However, you should
9219 not use this option to add directories that contain vendor-supplied
9220 system header files (use @option{-isystem} for that). If you use more than
9221 one @option{-I} option, the directories are scanned in left-to-right
9222 order; the standard system directories come after.
9224 If a standard system include directory, or a directory specified with
9225 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9226 option will be ignored. The directory will still be searched but as a
9227 system directory at its normal position in the system include chain.
9228 This is to ensure that GCC's procedure to fix buggy system headers and
9229 the ordering for the include_next directive are not inadvertently changed.
9230 If you really need to change the search order for system directories,
9231 use the @option{-nostdinc} and/or @option{-isystem} options.
9233 @item -iplugindir=@var{dir}
9234 Set the directory to search for plugins which are passed
9235 by @option{-fplugin=@var{name}} instead of
9236 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9237 to be used by the user, but only passed by the driver.
9239 @item -iquote@var{dir}
9241 Add the directory @var{dir} to the head of the list of directories to
9242 be searched for header files only for the case of @samp{#include
9243 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9244 otherwise just like @option{-I}.
9248 Add directory @var{dir} to the list of directories to be searched
9251 @item -B@var{prefix}
9253 This option specifies where to find the executables, libraries,
9254 include files, and data files of the compiler itself.
9256 The compiler driver program runs one or more of the subprograms
9257 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9258 @var{prefix} as a prefix for each program it tries to run, both with and
9259 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9261 For each subprogram to be run, the compiler driver first tries the
9262 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9263 was not specified, the driver tries two standard prefixes, which are
9264 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9265 those results in a file name that is found, the unmodified program
9266 name is searched for using the directories specified in your
9267 @env{PATH} environment variable.
9269 The compiler will check to see if the path provided by the @option{-B}
9270 refers to a directory, and if necessary it will add a directory
9271 separator character at the end of the path.
9273 @option{-B} prefixes that effectively specify directory names also apply
9274 to libraries in the linker, because the compiler translates these
9275 options into @option{-L} options for the linker. They also apply to
9276 includes files in the preprocessor, because the compiler translates these
9277 options into @option{-isystem} options for the preprocessor. In this case,
9278 the compiler appends @samp{include} to the prefix.
9280 The run-time support file @file{libgcc.a} can also be searched for using
9281 the @option{-B} prefix, if needed. If it is not found there, the two
9282 standard prefixes above are tried, and that is all. The file is left
9283 out of the link if it is not found by those means.
9285 Another way to specify a prefix much like the @option{-B} prefix is to use
9286 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9289 As a special kludge, if the path provided by @option{-B} is
9290 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9291 9, then it will be replaced by @file{[dir/]include}. This is to help
9292 with boot-strapping the compiler.
9294 @item -specs=@var{file}
9296 Process @var{file} after the compiler reads in the standard @file{specs}
9297 file, in order to override the defaults that the @file{gcc} driver
9298 program uses when determining what switches to pass to @file{cc1},
9299 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9300 @option{-specs=@var{file}} can be specified on the command line, and they
9301 are processed in order, from left to right.
9303 @item --sysroot=@var{dir}
9305 Use @var{dir} as the logical root directory for headers and libraries.
9306 For example, if the compiler would normally search for headers in
9307 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9308 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9310 If you use both this option and the @option{-isysroot} option, then
9311 the @option{--sysroot} option will apply to libraries, but the
9312 @option{-isysroot} option will apply to header files.
9314 The GNU linker (beginning with version 2.16) has the necessary support
9315 for this option. If your linker does not support this option, the
9316 header file aspect of @option{--sysroot} will still work, but the
9317 library aspect will not.
9321 This option has been deprecated. Please use @option{-iquote} instead for
9322 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9323 Any directories you specify with @option{-I} options before the @option{-I-}
9324 option are searched only for the case of @samp{#include "@var{file}"};
9325 they are not searched for @samp{#include <@var{file}>}.
9327 If additional directories are specified with @option{-I} options after
9328 the @option{-I-}, these directories are searched for all @samp{#include}
9329 directives. (Ordinarily @emph{all} @option{-I} directories are used
9332 In addition, the @option{-I-} option inhibits the use of the current
9333 directory (where the current input file came from) as the first search
9334 directory for @samp{#include "@var{file}"}. There is no way to
9335 override this effect of @option{-I-}. With @option{-I.} you can specify
9336 searching the directory which was current when the compiler was
9337 invoked. That is not exactly the same as what the preprocessor does
9338 by default, but it is often satisfactory.
9340 @option{-I-} does not inhibit the use of the standard system directories
9341 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9348 @section Specifying subprocesses and the switches to pass to them
9351 @command{gcc} is a driver program. It performs its job by invoking a
9352 sequence of other programs to do the work of compiling, assembling and
9353 linking. GCC interprets its command-line parameters and uses these to
9354 deduce which programs it should invoke, and which command-line options
9355 it ought to place on their command lines. This behavior is controlled
9356 by @dfn{spec strings}. In most cases there is one spec string for each
9357 program that GCC can invoke, but a few programs have multiple spec
9358 strings to control their behavior. The spec strings built into GCC can
9359 be overridden by using the @option{-specs=} command-line switch to specify
9362 @dfn{Spec files} are plaintext files that are used to construct spec
9363 strings. They consist of a sequence of directives separated by blank
9364 lines. The type of directive is determined by the first non-whitespace
9365 character on the line and it can be one of the following:
9368 @item %@var{command}
9369 Issues a @var{command} to the spec file processor. The commands that can
9373 @item %include <@var{file}>
9374 @cindex @code{%include}
9375 Search for @var{file} and insert its text at the current point in the
9378 @item %include_noerr <@var{file}>
9379 @cindex @code{%include_noerr}
9380 Just like @samp{%include}, but do not generate an error message if the include
9381 file cannot be found.
9383 @item %rename @var{old_name} @var{new_name}
9384 @cindex @code{%rename}
9385 Rename the spec string @var{old_name} to @var{new_name}.
9389 @item *[@var{spec_name}]:
9390 This tells the compiler to create, override or delete the named spec
9391 string. All lines after this directive up to the next directive or
9392 blank line are considered to be the text for the spec string. If this
9393 results in an empty string then the spec will be deleted. (Or, if the
9394 spec did not exist, then nothing will happened.) Otherwise, if the spec
9395 does not currently exist a new spec will be created. If the spec does
9396 exist then its contents will be overridden by the text of this
9397 directive, unless the first character of that text is the @samp{+}
9398 character, in which case the text will be appended to the spec.
9400 @item [@var{suffix}]:
9401 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9402 and up to the next directive or blank line are considered to make up the
9403 spec string for the indicated suffix. When the compiler encounters an
9404 input file with the named suffix, it will processes the spec string in
9405 order to work out how to compile that file. For example:
9412 This says that any input file whose name ends in @samp{.ZZ} should be
9413 passed to the program @samp{z-compile}, which should be invoked with the
9414 command-line switch @option{-input} and with the result of performing the
9415 @samp{%i} substitution. (See below.)
9417 As an alternative to providing a spec string, the text that follows a
9418 suffix directive can be one of the following:
9421 @item @@@var{language}
9422 This says that the suffix is an alias for a known @var{language}. This is
9423 similar to using the @option{-x} command-line switch to GCC to specify a
9424 language explicitly. For example:
9431 Says that .ZZ files are, in fact, C++ source files.
9434 This causes an error messages saying:
9437 @var{name} compiler not installed on this system.
9441 GCC already has an extensive list of suffixes built into it.
9442 This directive will add an entry to the end of the list of suffixes, but
9443 since the list is searched from the end backwards, it is effectively
9444 possible to override earlier entries using this technique.
9448 GCC has the following spec strings built into it. Spec files can
9449 override these strings or create their own. Note that individual
9450 targets can also add their own spec strings to this list.
9453 asm Options to pass to the assembler
9454 asm_final Options to pass to the assembler post-processor
9455 cpp Options to pass to the C preprocessor
9456 cc1 Options to pass to the C compiler
9457 cc1plus Options to pass to the C++ compiler
9458 endfile Object files to include at the end of the link
9459 link Options to pass to the linker
9460 lib Libraries to include on the command line to the linker
9461 libgcc Decides which GCC support library to pass to the linker
9462 linker Sets the name of the linker
9463 predefines Defines to be passed to the C preprocessor
9464 signed_char Defines to pass to CPP to say whether @code{char} is signed
9466 startfile Object files to include at the start of the link
9469 Here is a small example of a spec file:
9475 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9478 This example renames the spec called @samp{lib} to @samp{old_lib} and
9479 then overrides the previous definition of @samp{lib} with a new one.
9480 The new definition adds in some extra command-line options before
9481 including the text of the old definition.
9483 @dfn{Spec strings} are a list of command-line options to be passed to their
9484 corresponding program. In addition, the spec strings can contain
9485 @samp{%}-prefixed sequences to substitute variable text or to
9486 conditionally insert text into the command line. Using these constructs
9487 it is possible to generate quite complex command lines.
9489 Here is a table of all defined @samp{%}-sequences for spec
9490 strings. Note that spaces are not generated automatically around the
9491 results of expanding these sequences. Therefore you can concatenate them
9492 together or combine them with constant text in a single argument.
9496 Substitute one @samp{%} into the program name or argument.
9499 Substitute the name of the input file being processed.
9502 Substitute the basename of the input file being processed.
9503 This is the substring up to (and not including) the last period
9504 and not including the directory.
9507 This is the same as @samp{%b}, but include the file suffix (text after
9511 Marks the argument containing or following the @samp{%d} as a
9512 temporary file name, so that that file will be deleted if GCC exits
9513 successfully. Unlike @samp{%g}, this contributes no text to the
9516 @item %g@var{suffix}
9517 Substitute a file name that has suffix @var{suffix} and is chosen
9518 once per compilation, and mark the argument in the same way as
9519 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9520 name is now chosen in a way that is hard to predict even when previously
9521 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9522 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9523 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9524 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9525 was simply substituted with a file name chosen once per compilation,
9526 without regard to any appended suffix (which was therefore treated
9527 just like ordinary text), making such attacks more likely to succeed.
9529 @item %u@var{suffix}
9530 Like @samp{%g}, but generates a new temporary file name even if
9531 @samp{%u@var{suffix}} was already seen.
9533 @item %U@var{suffix}
9534 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9535 new one if there is no such last file name. In the absence of any
9536 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9537 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9538 would involve the generation of two distinct file names, one
9539 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9540 simply substituted with a file name chosen for the previous @samp{%u},
9541 without regard to any appended suffix.
9543 @item %j@var{suffix}
9544 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9545 writable, and if save-temps is off; otherwise, substitute the name
9546 of a temporary file, just like @samp{%u}. This temporary file is not
9547 meant for communication between processes, but rather as a junk
9550 @item %|@var{suffix}
9551 @itemx %m@var{suffix}
9552 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9553 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9554 all. These are the two most common ways to instruct a program that it
9555 should read from standard input or write to standard output. If you
9556 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9557 construct: see for example @file{f/lang-specs.h}.
9559 @item %.@var{SUFFIX}
9560 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9561 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9562 terminated by the next space or %.
9565 Marks the argument containing or following the @samp{%w} as the
9566 designated output file of this compilation. This puts the argument
9567 into the sequence of arguments that @samp{%o} will substitute later.
9570 Substitutes the names of all the output files, with spaces
9571 automatically placed around them. You should write spaces
9572 around the @samp{%o} as well or the results are undefined.
9573 @samp{%o} is for use in the specs for running the linker.
9574 Input files whose names have no recognized suffix are not compiled
9575 at all, but they are included among the output files, so they will
9579 Substitutes the suffix for object files. Note that this is
9580 handled specially when it immediately follows @samp{%g, %u, or %U},
9581 because of the need for those to form complete file names. The
9582 handling is such that @samp{%O} is treated exactly as if it had already
9583 been substituted, except that @samp{%g, %u, and %U} do not currently
9584 support additional @var{suffix} characters following @samp{%O} as they would
9585 following, for example, @samp{.o}.
9588 Substitutes the standard macro predefinitions for the
9589 current target machine. Use this when running @code{cpp}.
9592 Like @samp{%p}, but puts @samp{__} before and after the name of each
9593 predefined macro, except for macros that start with @samp{__} or with
9594 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9598 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9599 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9600 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9601 and @option{-imultilib} as necessary.
9604 Current argument is the name of a library or startup file of some sort.
9605 Search for that file in a standard list of directories and substitute
9606 the full name found. The current working directory is included in the
9607 list of directories scanned.
9610 Current argument is the name of a linker script. Search for that file
9611 in the current list of directories to scan for libraries. If the file
9612 is located insert a @option{--script} option into the command line
9613 followed by the full path name found. If the file is not found then
9614 generate an error message. Note: the current working directory is not
9618 Print @var{str} as an error message. @var{str} is terminated by a newline.
9619 Use this when inconsistent options are detected.
9622 Substitute the contents of spec string @var{name} at this point.
9625 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9627 @item %x@{@var{option}@}
9628 Accumulate an option for @samp{%X}.
9631 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9635 Output the accumulated assembler options specified by @option{-Wa}.
9638 Output the accumulated preprocessor options specified by @option{-Wp}.
9641 Process the @code{asm} spec. This is used to compute the
9642 switches to be passed to the assembler.
9645 Process the @code{asm_final} spec. This is a spec string for
9646 passing switches to an assembler post-processor, if such a program is
9650 Process the @code{link} spec. This is the spec for computing the
9651 command line passed to the linker. Typically it will make use of the
9652 @samp{%L %G %S %D and %E} sequences.
9655 Dump out a @option{-L} option for each directory that GCC believes might
9656 contain startup files. If the target supports multilibs then the
9657 current multilib directory will be prepended to each of these paths.
9660 Process the @code{lib} spec. This is a spec string for deciding which
9661 libraries should be included on the command line to the linker.
9664 Process the @code{libgcc} spec. This is a spec string for deciding
9665 which GCC support library should be included on the command line to the linker.
9668 Process the @code{startfile} spec. This is a spec for deciding which
9669 object files should be the first ones passed to the linker. Typically
9670 this might be a file named @file{crt0.o}.
9673 Process the @code{endfile} spec. This is a spec string that specifies
9674 the last object files that will be passed to the linker.
9677 Process the @code{cpp} spec. This is used to construct the arguments
9678 to be passed to the C preprocessor.
9681 Process the @code{cc1} spec. This is used to construct the options to be
9682 passed to the actual C compiler (@samp{cc1}).
9685 Process the @code{cc1plus} spec. This is used to construct the options to be
9686 passed to the actual C++ compiler (@samp{cc1plus}).
9689 Substitute the variable part of a matched option. See below.
9690 Note that each comma in the substituted string is replaced by
9694 Remove all occurrences of @code{-S} from the command line. Note---this
9695 command is position dependent. @samp{%} commands in the spec string
9696 before this one will see @code{-S}, @samp{%} commands in the spec string
9697 after this one will not.
9699 @item %:@var{function}(@var{args})
9700 Call the named function @var{function}, passing it @var{args}.
9701 @var{args} is first processed as a nested spec string, then split
9702 into an argument vector in the usual fashion. The function returns
9703 a string which is processed as if it had appeared literally as part
9704 of the current spec.
9706 The following built-in spec functions are provided:
9710 The @code{getenv} spec function takes two arguments: an environment
9711 variable name and a string. If the environment variable is not
9712 defined, a fatal error is issued. Otherwise, the return value is the
9713 value of the environment variable concatenated with the string. For
9714 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9717 %:getenv(TOPDIR /include)
9720 expands to @file{/path/to/top/include}.
9722 @item @code{if-exists}
9723 The @code{if-exists} spec function takes one argument, an absolute
9724 pathname to a file. If the file exists, @code{if-exists} returns the
9725 pathname. Here is a small example of its usage:
9729 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9732 @item @code{if-exists-else}
9733 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9734 spec function, except that it takes two arguments. The first argument is
9735 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9736 returns the pathname. If it does not exist, it returns the second argument.
9737 This way, @code{if-exists-else} can be used to select one file or another,
9738 based on the existence of the first. Here is a small example of its usage:
9742 crt0%O%s %:if-exists(crti%O%s) \
9743 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9746 @item @code{replace-outfile}
9747 The @code{replace-outfile} spec function takes two arguments. It looks for the
9748 first argument in the outfiles array and replaces it with the second argument. Here
9749 is a small example of its usage:
9752 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9755 @item @code{remove-outfile}
9756 The @code{remove-outfile} spec function takes one argument. It looks for the
9757 first argument in the outfiles array and removes it. Here is a small example
9761 %:remove-outfile(-lm)
9764 @item @code{pass-through-libs}
9765 The @code{pass-through-libs} spec function takes any number of arguments. It
9766 finds any @option{-l} options and any non-options ending in ".a" (which it
9767 assumes are the names of linker input library archive files) and returns a
9768 result containing all the found arguments each prepended by
9769 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9770 intended to be passed to the LTO linker plugin.
9773 %:pass-through-libs(%G %L %G)
9776 @item @code{print-asm-header}
9777 The @code{print-asm-header} function takes no arguments and simply
9778 prints a banner like:
9784 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9787 It is used to separate compiler options from assembler options
9788 in the @option{--target-help} output.
9792 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9793 If that switch was not specified, this substitutes nothing. Note that
9794 the leading dash is omitted when specifying this option, and it is
9795 automatically inserted if the substitution is performed. Thus the spec
9796 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9797 and would output the command line option @option{-foo}.
9799 @item %W@{@code{S}@}
9800 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9803 @item %@{@code{S}*@}
9804 Substitutes all the switches specified to GCC whose names start
9805 with @code{-S}, but which also take an argument. This is used for
9806 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9807 GCC considers @option{-o foo} as being
9808 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9809 text, including the space. Thus two arguments would be generated.
9811 @item %@{@code{S}*&@code{T}*@}
9812 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9813 (the order of @code{S} and @code{T} in the spec is not significant).
9814 There can be any number of ampersand-separated variables; for each the
9815 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9817 @item %@{@code{S}:@code{X}@}
9818 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9820 @item %@{!@code{S}:@code{X}@}
9821 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9823 @item %@{@code{S}*:@code{X}@}
9824 Substitutes @code{X} if one or more switches whose names start with
9825 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9826 once, no matter how many such switches appeared. However, if @code{%*}
9827 appears somewhere in @code{X}, then @code{X} will be substituted once
9828 for each matching switch, with the @code{%*} replaced by the part of
9829 that switch that matched the @code{*}.
9831 @item %@{.@code{S}:@code{X}@}
9832 Substitutes @code{X}, if processing a file with suffix @code{S}.
9834 @item %@{!.@code{S}:@code{X}@}
9835 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9837 @item %@{,@code{S}:@code{X}@}
9838 Substitutes @code{X}, if processing a file for language @code{S}.
9840 @item %@{!,@code{S}:@code{X}@}
9841 Substitutes @code{X}, if not processing a file for language @code{S}.
9843 @item %@{@code{S}|@code{P}:@code{X}@}
9844 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9845 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9846 @code{*} sequences as well, although they have a stronger binding than
9847 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9848 alternatives must be starred, and only the first matching alternative
9851 For example, a spec string like this:
9854 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9857 will output the following command-line options from the following input
9858 command-line options:
9863 -d fred.c -foo -baz -boggle
9864 -d jim.d -bar -baz -boggle
9867 @item %@{S:X; T:Y; :D@}
9869 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9870 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9871 be as many clauses as you need. This may be combined with @code{.},
9872 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9877 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9878 construct may contain other nested @samp{%} constructs or spaces, or
9879 even newlines. They are processed as usual, as described above.
9880 Trailing white space in @code{X} is ignored. White space may also
9881 appear anywhere on the left side of the colon in these constructs,
9882 except between @code{.} or @code{*} and the corresponding word.
9884 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9885 handled specifically in these constructs. If another value of
9886 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9887 @option{-W} switch is found later in the command line, the earlier
9888 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9889 just one letter, which passes all matching options.
9891 The character @samp{|} at the beginning of the predicate text is used to
9892 indicate that a command should be piped to the following command, but
9893 only if @option{-pipe} is specified.
9895 It is built into GCC which switches take arguments and which do not.
9896 (You might think it would be useful to generalize this to allow each
9897 compiler's spec to say which switches take arguments. But this cannot
9898 be done in a consistent fashion. GCC cannot even decide which input
9899 files have been specified without knowing which switches take arguments,
9900 and it must know which input files to compile in order to tell which
9903 GCC also knows implicitly that arguments starting in @option{-l} are to be
9904 treated as compiler output files, and passed to the linker in their
9905 proper position among the other output files.
9907 @c man begin OPTIONS
9909 @node Target Options
9910 @section Specifying Target Machine and Compiler Version
9911 @cindex target options
9912 @cindex cross compiling
9913 @cindex specifying machine version
9914 @cindex specifying compiler version and target machine
9915 @cindex compiler version, specifying
9916 @cindex target machine, specifying
9918 The usual way to run GCC is to run the executable called @command{gcc}, or
9919 @command{@var{machine}-gcc} when cross-compiling, or
9920 @command{@var{machine}-gcc-@var{version}} to run a version other than the
9921 one that was installed last.
9923 @node Submodel Options
9924 @section Hardware Models and Configurations
9925 @cindex submodel options
9926 @cindex specifying hardware config
9927 @cindex hardware models and configurations, specifying
9928 @cindex machine dependent options
9930 Each target machine types can have its own
9931 special options, starting with @samp{-m}, to choose among various
9932 hardware models or configurations---for example, 68010 vs 68020,
9933 floating coprocessor or none. A single installed version of the
9934 compiler can compile for any model or configuration, according to the
9937 Some configurations of the compiler also support additional special
9938 options, usually for compatibility with other compilers on the same
9941 @c This list is ordered alphanumerically by subsection name.
9942 @c It should be the same order and spelling as these options are listed
9943 @c in Machine Dependent Options
9948 * Blackfin Options::
9951 * DEC Alpha Options::
9952 * DEC Alpha/VMS Options::
9955 * GNU/Linux Options::
9958 * i386 and x86-64 Options::
9959 * i386 and x86-64 Windows Options::
9961 * IA-64/VMS Options::
9968 * MicroBlaze Options::
9973 * picoChip Options::
9975 * RS/6000 and PowerPC Options::
9977 * S/390 and zSeries Options::
9980 * Solaris 2 Options::
9983 * System V Options::
9988 * Xstormy16 Options::
9994 @subsection ARM Options
9997 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10001 @item -mabi=@var{name}
10003 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10004 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10007 @opindex mapcs-frame
10008 Generate a stack frame that is compliant with the ARM Procedure Call
10009 Standard for all functions, even if this is not strictly necessary for
10010 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10011 with this option will cause the stack frames not to be generated for
10012 leaf functions. The default is @option{-mno-apcs-frame}.
10016 This is a synonym for @option{-mapcs-frame}.
10019 @c not currently implemented
10020 @item -mapcs-stack-check
10021 @opindex mapcs-stack-check
10022 Generate code to check the amount of stack space available upon entry to
10023 every function (that actually uses some stack space). If there is
10024 insufficient space available then either the function
10025 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10026 called, depending upon the amount of stack space required. The run time
10027 system is required to provide these functions. The default is
10028 @option{-mno-apcs-stack-check}, since this produces smaller code.
10030 @c not currently implemented
10032 @opindex mapcs-float
10033 Pass floating point arguments using the float point registers. This is
10034 one of the variants of the APCS@. This option is recommended if the
10035 target hardware has a floating point unit or if a lot of floating point
10036 arithmetic is going to be performed by the code. The default is
10037 @option{-mno-apcs-float}, since integer only code is slightly increased in
10038 size if @option{-mapcs-float} is used.
10040 @c not currently implemented
10041 @item -mapcs-reentrant
10042 @opindex mapcs-reentrant
10043 Generate reentrant, position independent code. The default is
10044 @option{-mno-apcs-reentrant}.
10047 @item -mthumb-interwork
10048 @opindex mthumb-interwork
10049 Generate code which supports calling between the ARM and Thumb
10050 instruction sets. Without this option the two instruction sets cannot
10051 be reliably used inside one program. The default is
10052 @option{-mno-thumb-interwork}, since slightly larger code is generated
10053 when @option{-mthumb-interwork} is specified.
10055 @item -mno-sched-prolog
10056 @opindex mno-sched-prolog
10057 Prevent the reordering of instructions in the function prolog, or the
10058 merging of those instruction with the instructions in the function's
10059 body. This means that all functions will start with a recognizable set
10060 of instructions (or in fact one of a choice from a small set of
10061 different function prologues), and this information can be used to
10062 locate the start if functions inside an executable piece of code. The
10063 default is @option{-msched-prolog}.
10065 @item -mfloat-abi=@var{name}
10066 @opindex mfloat-abi
10067 Specifies which floating-point ABI to use. Permissible values
10068 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10070 Specifying @samp{soft} causes GCC to generate output containing
10071 library calls for floating-point operations.
10072 @samp{softfp} allows the generation of code using hardware floating-point
10073 instructions, but still uses the soft-float calling conventions.
10074 @samp{hard} allows generation of floating-point instructions
10075 and uses FPU-specific calling conventions.
10077 The default depends on the specific target configuration. Note that
10078 the hard-float and soft-float ABIs are not link-compatible; you must
10079 compile your entire program with the same ABI, and link with a
10080 compatible set of libraries.
10082 @item -mlittle-endian
10083 @opindex mlittle-endian
10084 Generate code for a processor running in little-endian mode. This is
10085 the default for all standard configurations.
10088 @opindex mbig-endian
10089 Generate code for a processor running in big-endian mode; the default is
10090 to compile code for a little-endian processor.
10092 @item -mwords-little-endian
10093 @opindex mwords-little-endian
10094 This option only applies when generating code for big-endian processors.
10095 Generate code for a little-endian word order but a big-endian byte
10096 order. That is, a byte order of the form @samp{32107654}. Note: this
10097 option should only be used if you require compatibility with code for
10098 big-endian ARM processors generated by versions of the compiler prior to
10101 @item -mcpu=@var{name}
10103 This specifies the name of the target ARM processor. GCC uses this name
10104 to determine what kind of instructions it can emit when generating
10105 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10106 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10107 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10108 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10109 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10111 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10112 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10113 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10114 @samp{strongarm1110},
10115 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10116 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10117 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10118 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10119 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10120 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10121 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10122 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10123 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
10126 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10128 @item -mtune=@var{name}
10130 This option is very similar to the @option{-mcpu=} option, except that
10131 instead of specifying the actual target processor type, and hence
10132 restricting which instructions can be used, it specifies that GCC should
10133 tune the performance of the code as if the target were of the type
10134 specified in this option, but still choosing the instructions that it
10135 will generate based on the CPU specified by a @option{-mcpu=} option.
10136 For some ARM implementations better performance can be obtained by using
10139 @item -march=@var{name}
10141 This specifies the name of the target ARM architecture. GCC uses this
10142 name to determine what kind of instructions it can emit when generating
10143 assembly code. This option can be used in conjunction with or instead
10144 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10145 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10146 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10147 @samp{armv6}, @samp{armv6j},
10148 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10149 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10150 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10152 @item -mfpu=@var{name}
10153 @itemx -mfpe=@var{number}
10154 @itemx -mfp=@var{number}
10158 This specifies what floating point hardware (or hardware emulation) is
10159 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10160 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10161 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10162 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10163 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10164 @option{-mfp} and @option{-mfpe} are synonyms for
10165 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10168 If @option{-msoft-float} is specified this specifies the format of
10169 floating point values.
10171 If the selected floating-point hardware includes the NEON extension
10172 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10173 operations will not be used by GCC's auto-vectorization pass unless
10174 @option{-funsafe-math-optimizations} is also specified. This is
10175 because NEON hardware does not fully implement the IEEE 754 standard for
10176 floating-point arithmetic (in particular denormal values are treated as
10177 zero), so the use of NEON instructions may lead to a loss of precision.
10179 @item -mfp16-format=@var{name}
10180 @opindex mfp16-format
10181 Specify the format of the @code{__fp16} half-precision floating-point type.
10182 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10183 the default is @samp{none}, in which case the @code{__fp16} type is not
10184 defined. @xref{Half-Precision}, for more information.
10186 @item -mstructure-size-boundary=@var{n}
10187 @opindex mstructure-size-boundary
10188 The size of all structures and unions will be rounded up to a multiple
10189 of the number of bits set by this option. Permissible values are 8, 32
10190 and 64. The default value varies for different toolchains. For the COFF
10191 targeted toolchain the default value is 8. A value of 64 is only allowed
10192 if the underlying ABI supports it.
10194 Specifying the larger number can produce faster, more efficient code, but
10195 can also increase the size of the program. Different values are potentially
10196 incompatible. Code compiled with one value cannot necessarily expect to
10197 work with code or libraries compiled with another value, if they exchange
10198 information using structures or unions.
10200 @item -mabort-on-noreturn
10201 @opindex mabort-on-noreturn
10202 Generate a call to the function @code{abort} at the end of a
10203 @code{noreturn} function. It will be executed if the function tries to
10207 @itemx -mno-long-calls
10208 @opindex mlong-calls
10209 @opindex mno-long-calls
10210 Tells the compiler to perform function calls by first loading the
10211 address of the function into a register and then performing a subroutine
10212 call on this register. This switch is needed if the target function
10213 will lie outside of the 64 megabyte addressing range of the offset based
10214 version of subroutine call instruction.
10216 Even if this switch is enabled, not all function calls will be turned
10217 into long calls. The heuristic is that static functions, functions
10218 which have the @samp{short-call} attribute, functions that are inside
10219 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10220 definitions have already been compiled within the current compilation
10221 unit, will not be turned into long calls. The exception to this rule is
10222 that weak function definitions, functions with the @samp{long-call}
10223 attribute or the @samp{section} attribute, and functions that are within
10224 the scope of a @samp{#pragma long_calls} directive, will always be
10225 turned into long calls.
10227 This feature is not enabled by default. Specifying
10228 @option{-mno-long-calls} will restore the default behavior, as will
10229 placing the function calls within the scope of a @samp{#pragma
10230 long_calls_off} directive. Note these switches have no effect on how
10231 the compiler generates code to handle function calls via function
10234 @item -msingle-pic-base
10235 @opindex msingle-pic-base
10236 Treat the register used for PIC addressing as read-only, rather than
10237 loading it in the prologue for each function. The run-time system is
10238 responsible for initializing this register with an appropriate value
10239 before execution begins.
10241 @item -mpic-register=@var{reg}
10242 @opindex mpic-register
10243 Specify the register to be used for PIC addressing. The default is R10
10244 unless stack-checking is enabled, when R9 is used.
10246 @item -mcirrus-fix-invalid-insns
10247 @opindex mcirrus-fix-invalid-insns
10248 @opindex mno-cirrus-fix-invalid-insns
10249 Insert NOPs into the instruction stream to in order to work around
10250 problems with invalid Maverick instruction combinations. This option
10251 is only valid if the @option{-mcpu=ep9312} option has been used to
10252 enable generation of instructions for the Cirrus Maverick floating
10253 point co-processor. This option is not enabled by default, since the
10254 problem is only present in older Maverick implementations. The default
10255 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10258 @item -mpoke-function-name
10259 @opindex mpoke-function-name
10260 Write the name of each function into the text section, directly
10261 preceding the function prologue. The generated code is similar to this:
10265 .ascii "arm_poke_function_name", 0
10268 .word 0xff000000 + (t1 - t0)
10269 arm_poke_function_name
10271 stmfd sp!, @{fp, ip, lr, pc@}
10275 When performing a stack backtrace, code can inspect the value of
10276 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10277 location @code{pc - 12} and the top 8 bits are set, then we know that
10278 there is a function name embedded immediately preceding this location
10279 and has length @code{((pc[-3]) & 0xff000000)}.
10283 Generate code for the Thumb instruction set. The default is to
10284 use the 32-bit ARM instruction set.
10285 This option automatically enables either 16-bit Thumb-1 or
10286 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10287 and @option{-march=@var{name}} options. This option is not passed to the
10288 assembler. If you want to force assembler files to be interpreted as Thumb code,
10289 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10290 option directly to the assembler by prefixing it with @option{-Wa}.
10293 @opindex mtpcs-frame
10294 Generate a stack frame that is compliant with the Thumb Procedure Call
10295 Standard for all non-leaf functions. (A leaf function is one that does
10296 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10298 @item -mtpcs-leaf-frame
10299 @opindex mtpcs-leaf-frame
10300 Generate a stack frame that is compliant with the Thumb Procedure Call
10301 Standard for all leaf functions. (A leaf function is one that does
10302 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10304 @item -mcallee-super-interworking
10305 @opindex mcallee-super-interworking
10306 Gives all externally visible functions in the file being compiled an ARM
10307 instruction set header which switches to Thumb mode before executing the
10308 rest of the function. This allows these functions to be called from
10309 non-interworking code. This option is not valid in AAPCS configurations
10310 because interworking is enabled by default.
10312 @item -mcaller-super-interworking
10313 @opindex mcaller-super-interworking
10314 Allows calls via function pointers (including virtual functions) to
10315 execute correctly regardless of whether the target code has been
10316 compiled for interworking or not. There is a small overhead in the cost
10317 of executing a function pointer if this option is enabled. This option
10318 is not valid in AAPCS configurations because interworking is enabled
10321 @item -mtp=@var{name}
10323 Specify the access model for the thread local storage pointer. The valid
10324 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10325 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10326 (supported in the arm6k architecture), and @option{auto}, which uses the
10327 best available method for the selected processor. The default setting is
10330 @item -mword-relocations
10331 @opindex mword-relocations
10332 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10333 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10334 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10337 @item -mfix-cortex-m3-ldrd
10338 @opindex mfix-cortex-m3-ldrd
10339 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10340 with overlapping destination and base registers are used. This option avoids
10341 generating these instructions. This option is enabled by default when
10342 @option{-mcpu=cortex-m3} is specified.
10347 @subsection AVR Options
10348 @cindex AVR Options
10350 These options are defined for AVR implementations:
10353 @item -mmcu=@var{mcu}
10355 Specify ATMEL AVR instruction set or MCU type.
10357 Instruction set avr1 is for the minimal AVR core, not supported by the C
10358 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10359 attiny11, attiny12, attiny15, attiny28).
10361 Instruction set avr2 (default) is for the classic AVR core with up to
10362 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10363 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10364 at90c8534, at90s8535).
10366 Instruction set avr3 is for the classic AVR core with up to 128K program
10367 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10369 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10370 memory space (MCU types: atmega8, atmega83, atmega85).
10372 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10373 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10374 atmega64, atmega128, at43usb355, at94k).
10376 @item -mno-interrupts
10377 @opindex mno-interrupts
10378 Generated code is not compatible with hardware interrupts.
10379 Code size will be smaller.
10381 @item -mcall-prologues
10382 @opindex mcall-prologues
10383 Functions prologues/epilogues expanded as call to appropriate
10384 subroutines. Code size will be smaller.
10387 @opindex mtiny-stack
10388 Change only the low 8 bits of the stack pointer.
10392 Assume int to be 8 bit integer. This affects the sizes of all types: A
10393 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10394 and long long will be 4 bytes. Please note that this option does not
10395 comply to the C standards, but it will provide you with smaller code
10399 @node Blackfin Options
10400 @subsection Blackfin Options
10401 @cindex Blackfin Options
10404 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10406 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10407 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10408 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10409 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10410 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10411 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10412 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10414 The optional @var{sirevision} specifies the silicon revision of the target
10415 Blackfin processor. Any workarounds available for the targeted silicon revision
10416 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10417 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10418 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10419 hexadecimal digits representing the major and minor numbers in the silicon
10420 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10421 is not defined. If @var{sirevision} is @samp{any}, the
10422 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10423 If this optional @var{sirevision} is not used, GCC assumes the latest known
10424 silicon revision of the targeted Blackfin processor.
10426 Support for @samp{bf561} is incomplete. For @samp{bf561},
10427 Only the processor macro is defined.
10428 Without this option, @samp{bf532} is used as the processor by default.
10429 The corresponding predefined processor macros for @var{cpu} is to
10430 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10431 provided by libgloss to be linked in if @option{-msim} is not given.
10435 Specifies that the program will be run on the simulator. This causes
10436 the simulator BSP provided by libgloss to be linked in. This option
10437 has effect only for @samp{bfin-elf} toolchain.
10438 Certain other options, such as @option{-mid-shared-library} and
10439 @option{-mfdpic}, imply @option{-msim}.
10441 @item -momit-leaf-frame-pointer
10442 @opindex momit-leaf-frame-pointer
10443 Don't keep the frame pointer in a register for leaf functions. This
10444 avoids the instructions to save, set up and restore frame pointers and
10445 makes an extra register available in leaf functions. The option
10446 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10447 which might make debugging harder.
10449 @item -mspecld-anomaly
10450 @opindex mspecld-anomaly
10451 When enabled, the compiler will ensure that the generated code does not
10452 contain speculative loads after jump instructions. If this option is used,
10453 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10455 @item -mno-specld-anomaly
10456 @opindex mno-specld-anomaly
10457 Don't generate extra code to prevent speculative loads from occurring.
10459 @item -mcsync-anomaly
10460 @opindex mcsync-anomaly
10461 When enabled, the compiler will ensure that the generated code does not
10462 contain CSYNC or SSYNC instructions too soon after conditional branches.
10463 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10465 @item -mno-csync-anomaly
10466 @opindex mno-csync-anomaly
10467 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10468 occurring too soon after a conditional branch.
10472 When enabled, the compiler is free to take advantage of the knowledge that
10473 the entire program fits into the low 64k of memory.
10476 @opindex mno-low-64k
10477 Assume that the program is arbitrarily large. This is the default.
10479 @item -mstack-check-l1
10480 @opindex mstack-check-l1
10481 Do stack checking using information placed into L1 scratchpad memory by the
10484 @item -mid-shared-library
10485 @opindex mid-shared-library
10486 Generate code that supports shared libraries via the library ID method.
10487 This allows for execute in place and shared libraries in an environment
10488 without virtual memory management. This option implies @option{-fPIC}.
10489 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10491 @item -mno-id-shared-library
10492 @opindex mno-id-shared-library
10493 Generate code that doesn't assume ID based shared libraries are being used.
10494 This is the default.
10496 @item -mleaf-id-shared-library
10497 @opindex mleaf-id-shared-library
10498 Generate code that supports shared libraries via the library ID method,
10499 but assumes that this library or executable won't link against any other
10500 ID shared libraries. That allows the compiler to use faster code for jumps
10503 @item -mno-leaf-id-shared-library
10504 @opindex mno-leaf-id-shared-library
10505 Do not assume that the code being compiled won't link against any ID shared
10506 libraries. Slower code will be generated for jump and call insns.
10508 @item -mshared-library-id=n
10509 @opindex mshared-library-id
10510 Specified the identification number of the ID based shared library being
10511 compiled. Specifying a value of 0 will generate more compact code, specifying
10512 other values will force the allocation of that number to the current
10513 library but is no more space or time efficient than omitting this option.
10517 Generate code that allows the data segment to be located in a different
10518 area of memory from the text segment. This allows for execute in place in
10519 an environment without virtual memory management by eliminating relocations
10520 against the text section.
10522 @item -mno-sep-data
10523 @opindex mno-sep-data
10524 Generate code that assumes that the data segment follows the text segment.
10525 This is the default.
10528 @itemx -mno-long-calls
10529 @opindex mlong-calls
10530 @opindex mno-long-calls
10531 Tells the compiler to perform function calls by first loading the
10532 address of the function into a register and then performing a subroutine
10533 call on this register. This switch is needed if the target function
10534 will lie outside of the 24 bit addressing range of the offset based
10535 version of subroutine call instruction.
10537 This feature is not enabled by default. Specifying
10538 @option{-mno-long-calls} will restore the default behavior. Note these
10539 switches have no effect on how the compiler generates code to handle
10540 function calls via function pointers.
10544 Link with the fast floating-point library. This library relaxes some of
10545 the IEEE floating-point standard's rules for checking inputs against
10546 Not-a-Number (NAN), in the interest of performance.
10549 @opindex minline-plt
10550 Enable inlining of PLT entries in function calls to functions that are
10551 not known to bind locally. It has no effect without @option{-mfdpic}.
10554 @opindex mmulticore
10555 Build standalone application for multicore Blackfin processor. Proper
10556 start files and link scripts will be used to support multicore.
10557 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10558 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10559 @option{-mcorea} or @option{-mcoreb}. If it's used without
10560 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10561 programming model is used. In this model, the main function of Core B
10562 should be named as coreb_main. If it's used with @option{-mcorea} or
10563 @option{-mcoreb}, one application per core programming model is used.
10564 If this option is not used, single core application programming
10569 Build standalone application for Core A of BF561 when using
10570 one application per core programming model. Proper start files
10571 and link scripts will be used to support Core A. This option
10572 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10576 Build standalone application for Core B of BF561 when using
10577 one application per core programming model. Proper start files
10578 and link scripts will be used to support Core B. This option
10579 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10580 should be used instead of main. It must be used with
10581 @option{-mmulticore}.
10585 Build standalone application for SDRAM. Proper start files and
10586 link scripts will be used to put the application into SDRAM.
10587 Loader should initialize SDRAM before loading the application
10588 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10592 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10593 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10594 are enabled; for standalone applications the default is off.
10598 @subsection CRIS Options
10599 @cindex CRIS Options
10601 These options are defined specifically for the CRIS ports.
10604 @item -march=@var{architecture-type}
10605 @itemx -mcpu=@var{architecture-type}
10608 Generate code for the specified architecture. The choices for
10609 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10610 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10611 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10614 @item -mtune=@var{architecture-type}
10616 Tune to @var{architecture-type} everything applicable about the generated
10617 code, except for the ABI and the set of available instructions. The
10618 choices for @var{architecture-type} are the same as for
10619 @option{-march=@var{architecture-type}}.
10621 @item -mmax-stack-frame=@var{n}
10622 @opindex mmax-stack-frame
10623 Warn when the stack frame of a function exceeds @var{n} bytes.
10629 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10630 @option{-march=v3} and @option{-march=v8} respectively.
10632 @item -mmul-bug-workaround
10633 @itemx -mno-mul-bug-workaround
10634 @opindex mmul-bug-workaround
10635 @opindex mno-mul-bug-workaround
10636 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10637 models where it applies. This option is active by default.
10641 Enable CRIS-specific verbose debug-related information in the assembly
10642 code. This option also has the effect to turn off the @samp{#NO_APP}
10643 formatted-code indicator to the assembler at the beginning of the
10648 Do not use condition-code results from previous instruction; always emit
10649 compare and test instructions before use of condition codes.
10651 @item -mno-side-effects
10652 @opindex mno-side-effects
10653 Do not emit instructions with side-effects in addressing modes other than
10656 @item -mstack-align
10657 @itemx -mno-stack-align
10658 @itemx -mdata-align
10659 @itemx -mno-data-align
10660 @itemx -mconst-align
10661 @itemx -mno-const-align
10662 @opindex mstack-align
10663 @opindex mno-stack-align
10664 @opindex mdata-align
10665 @opindex mno-data-align
10666 @opindex mconst-align
10667 @opindex mno-const-align
10668 These options (no-options) arranges (eliminate arrangements) for the
10669 stack-frame, individual data and constants to be aligned for the maximum
10670 single data access size for the chosen CPU model. The default is to
10671 arrange for 32-bit alignment. ABI details such as structure layout are
10672 not affected by these options.
10680 Similar to the stack- data- and const-align options above, these options
10681 arrange for stack-frame, writable data and constants to all be 32-bit,
10682 16-bit or 8-bit aligned. The default is 32-bit alignment.
10684 @item -mno-prologue-epilogue
10685 @itemx -mprologue-epilogue
10686 @opindex mno-prologue-epilogue
10687 @opindex mprologue-epilogue
10688 With @option{-mno-prologue-epilogue}, the normal function prologue and
10689 epilogue that sets up the stack-frame are omitted and no return
10690 instructions or return sequences are generated in the code. Use this
10691 option only together with visual inspection of the compiled code: no
10692 warnings or errors are generated when call-saved registers must be saved,
10693 or storage for local variable needs to be allocated.
10697 @opindex mno-gotplt
10699 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10700 instruction sequences that load addresses for functions from the PLT part
10701 of the GOT rather than (traditional on other architectures) calls to the
10702 PLT@. The default is @option{-mgotplt}.
10706 Legacy no-op option only recognized with the cris-axis-elf and
10707 cris-axis-linux-gnu targets.
10711 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10715 This option, recognized for the cris-axis-elf arranges
10716 to link with input-output functions from a simulator library. Code,
10717 initialized data and zero-initialized data are allocated consecutively.
10721 Like @option{-sim}, but pass linker options to locate initialized data at
10722 0x40000000 and zero-initialized data at 0x80000000.
10725 @node Darwin Options
10726 @subsection Darwin Options
10727 @cindex Darwin options
10729 These options are defined for all architectures running the Darwin operating
10732 FSF GCC on Darwin does not create ``fat'' object files; it will create
10733 an object file for the single architecture that it was built to
10734 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10735 @option{-arch} options are used; it does so by running the compiler or
10736 linker multiple times and joining the results together with
10739 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10740 @samp{i686}) is determined by the flags that specify the ISA
10741 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10742 @option{-force_cpusubtype_ALL} option can be used to override this.
10744 The Darwin tools vary in their behavior when presented with an ISA
10745 mismatch. The assembler, @file{as}, will only permit instructions to
10746 be used that are valid for the subtype of the file it is generating,
10747 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10748 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10749 and print an error if asked to create a shared library with a less
10750 restrictive subtype than its input files (for instance, trying to put
10751 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10752 for executables, @file{ld}, will quietly give the executable the most
10753 restrictive subtype of any of its input files.
10758 Add the framework directory @var{dir} to the head of the list of
10759 directories to be searched for header files. These directories are
10760 interleaved with those specified by @option{-I} options and are
10761 scanned in a left-to-right order.
10763 A framework directory is a directory with frameworks in it. A
10764 framework is a directory with a @samp{"Headers"} and/or
10765 @samp{"PrivateHeaders"} directory contained directly in it that ends
10766 in @samp{".framework"}. The name of a framework is the name of this
10767 directory excluding the @samp{".framework"}. Headers associated with
10768 the framework are found in one of those two directories, with
10769 @samp{"Headers"} being searched first. A subframework is a framework
10770 directory that is in a framework's @samp{"Frameworks"} directory.
10771 Includes of subframework headers can only appear in a header of a
10772 framework that contains the subframework, or in a sibling subframework
10773 header. Two subframeworks are siblings if they occur in the same
10774 framework. A subframework should not have the same name as a
10775 framework, a warning will be issued if this is violated. Currently a
10776 subframework cannot have subframeworks, in the future, the mechanism
10777 may be extended to support this. The standard frameworks can be found
10778 in @samp{"/System/Library/Frameworks"} and
10779 @samp{"/Library/Frameworks"}. An example include looks like
10780 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10781 the name of the framework and header.h is found in the
10782 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10784 @item -iframework@var{dir}
10785 @opindex iframework
10786 Like @option{-F} except the directory is a treated as a system
10787 directory. The main difference between this @option{-iframework} and
10788 @option{-F} is that with @option{-iframework} the compiler does not
10789 warn about constructs contained within header files found via
10790 @var{dir}. This option is valid only for the C family of languages.
10794 Emit debugging information for symbols that are used. For STABS
10795 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10796 This is by default ON@.
10800 Emit debugging information for all symbols and types.
10802 @item -mmacosx-version-min=@var{version}
10803 The earliest version of MacOS X that this executable will run on
10804 is @var{version}. Typical values of @var{version} include @code{10.1},
10805 @code{10.2}, and @code{10.3.9}.
10807 If the compiler was built to use the system's headers by default,
10808 then the default for this option is the system version on which the
10809 compiler is running, otherwise the default is to make choices which
10810 are compatible with as many systems and code bases as possible.
10814 Enable kernel development mode. The @option{-mkernel} option sets
10815 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10816 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10817 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10818 applicable. This mode also sets @option{-mno-altivec},
10819 @option{-msoft-float}, @option{-fno-builtin} and
10820 @option{-mlong-branch} for PowerPC targets.
10822 @item -mone-byte-bool
10823 @opindex mone-byte-bool
10824 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10825 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10826 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10827 option has no effect on x86.
10829 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10830 to generate code that is not binary compatible with code generated
10831 without that switch. Using this switch may require recompiling all
10832 other modules in a program, including system libraries. Use this
10833 switch to conform to a non-default data model.
10835 @item -mfix-and-continue
10836 @itemx -ffix-and-continue
10837 @itemx -findirect-data
10838 @opindex mfix-and-continue
10839 @opindex ffix-and-continue
10840 @opindex findirect-data
10841 Generate code suitable for fast turn around development. Needed to
10842 enable gdb to dynamically load @code{.o} files into already running
10843 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10844 are provided for backwards compatibility.
10848 Loads all members of static archive libraries.
10849 See man ld(1) for more information.
10851 @item -arch_errors_fatal
10852 @opindex arch_errors_fatal
10853 Cause the errors having to do with files that have the wrong architecture
10856 @item -bind_at_load
10857 @opindex bind_at_load
10858 Causes the output file to be marked such that the dynamic linker will
10859 bind all undefined references when the file is loaded or launched.
10863 Produce a Mach-o bundle format file.
10864 See man ld(1) for more information.
10866 @item -bundle_loader @var{executable}
10867 @opindex bundle_loader
10868 This option specifies the @var{executable} that will be loading the build
10869 output file being linked. See man ld(1) for more information.
10872 @opindex dynamiclib
10873 When passed this option, GCC will produce a dynamic library instead of
10874 an executable when linking, using the Darwin @file{libtool} command.
10876 @item -force_cpusubtype_ALL
10877 @opindex force_cpusubtype_ALL
10878 This causes GCC's output file to have the @var{ALL} subtype, instead of
10879 one controlled by the @option{-mcpu} or @option{-march} option.
10881 @item -allowable_client @var{client_name}
10882 @itemx -client_name
10883 @itemx -compatibility_version
10884 @itemx -current_version
10886 @itemx -dependency-file
10888 @itemx -dylinker_install_name
10890 @itemx -exported_symbols_list
10893 @itemx -flat_namespace
10894 @itemx -force_flat_namespace
10895 @itemx -headerpad_max_install_names
10898 @itemx -install_name
10899 @itemx -keep_private_externs
10900 @itemx -multi_module
10901 @itemx -multiply_defined
10902 @itemx -multiply_defined_unused
10905 @itemx -no_dead_strip_inits_and_terms
10906 @itemx -nofixprebinding
10907 @itemx -nomultidefs
10909 @itemx -noseglinkedit
10910 @itemx -pagezero_size
10912 @itemx -prebind_all_twolevel_modules
10913 @itemx -private_bundle
10915 @itemx -read_only_relocs
10917 @itemx -sectobjectsymbols
10921 @itemx -sectobjectsymbols
10924 @itemx -segs_read_only_addr
10926 @itemx -segs_read_write_addr
10927 @itemx -seg_addr_table
10928 @itemx -seg_addr_table_filename
10929 @itemx -seglinkedit
10931 @itemx -segs_read_only_addr
10932 @itemx -segs_read_write_addr
10933 @itemx -single_module
10935 @itemx -sub_library
10937 @itemx -sub_umbrella
10938 @itemx -twolevel_namespace
10941 @itemx -unexported_symbols_list
10942 @itemx -weak_reference_mismatches
10943 @itemx -whatsloaded
10944 @opindex allowable_client
10945 @opindex client_name
10946 @opindex compatibility_version
10947 @opindex current_version
10948 @opindex dead_strip
10949 @opindex dependency-file
10950 @opindex dylib_file
10951 @opindex dylinker_install_name
10953 @opindex exported_symbols_list
10955 @opindex flat_namespace
10956 @opindex force_flat_namespace
10957 @opindex headerpad_max_install_names
10958 @opindex image_base
10960 @opindex install_name
10961 @opindex keep_private_externs
10962 @opindex multi_module
10963 @opindex multiply_defined
10964 @opindex multiply_defined_unused
10965 @opindex noall_load
10966 @opindex no_dead_strip_inits_and_terms
10967 @opindex nofixprebinding
10968 @opindex nomultidefs
10970 @opindex noseglinkedit
10971 @opindex pagezero_size
10973 @opindex prebind_all_twolevel_modules
10974 @opindex private_bundle
10975 @opindex read_only_relocs
10977 @opindex sectobjectsymbols
10980 @opindex sectcreate
10981 @opindex sectobjectsymbols
10984 @opindex segs_read_only_addr
10985 @opindex segs_read_write_addr
10986 @opindex seg_addr_table
10987 @opindex seg_addr_table_filename
10988 @opindex seglinkedit
10990 @opindex segs_read_only_addr
10991 @opindex segs_read_write_addr
10992 @opindex single_module
10994 @opindex sub_library
10995 @opindex sub_umbrella
10996 @opindex twolevel_namespace
10999 @opindex unexported_symbols_list
11000 @opindex weak_reference_mismatches
11001 @opindex whatsloaded
11002 These options are passed to the Darwin linker. The Darwin linker man page
11003 describes them in detail.
11006 @node DEC Alpha Options
11007 @subsection DEC Alpha Options
11009 These @samp{-m} options are defined for the DEC Alpha implementations:
11012 @item -mno-soft-float
11013 @itemx -msoft-float
11014 @opindex mno-soft-float
11015 @opindex msoft-float
11016 Use (do not use) the hardware floating-point instructions for
11017 floating-point operations. When @option{-msoft-float} is specified,
11018 functions in @file{libgcc.a} will be used to perform floating-point
11019 operations. Unless they are replaced by routines that emulate the
11020 floating-point operations, or compiled in such a way as to call such
11021 emulations routines, these routines will issue floating-point
11022 operations. If you are compiling for an Alpha without floating-point
11023 operations, you must ensure that the library is built so as not to call
11026 Note that Alpha implementations without floating-point operations are
11027 required to have floating-point registers.
11030 @itemx -mno-fp-regs
11032 @opindex mno-fp-regs
11033 Generate code that uses (does not use) the floating-point register set.
11034 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11035 register set is not used, floating point operands are passed in integer
11036 registers as if they were integers and floating-point results are passed
11037 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11038 so any function with a floating-point argument or return value called by code
11039 compiled with @option{-mno-fp-regs} must also be compiled with that
11042 A typical use of this option is building a kernel that does not use,
11043 and hence need not save and restore, any floating-point registers.
11047 The Alpha architecture implements floating-point hardware optimized for
11048 maximum performance. It is mostly compliant with the IEEE floating
11049 point standard. However, for full compliance, software assistance is
11050 required. This option generates code fully IEEE compliant code
11051 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11052 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11053 defined during compilation. The resulting code is less efficient but is
11054 able to correctly support denormalized numbers and exceptional IEEE
11055 values such as not-a-number and plus/minus infinity. Other Alpha
11056 compilers call this option @option{-ieee_with_no_inexact}.
11058 @item -mieee-with-inexact
11059 @opindex mieee-with-inexact
11060 This is like @option{-mieee} except the generated code also maintains
11061 the IEEE @var{inexact-flag}. Turning on this option causes the
11062 generated code to implement fully-compliant IEEE math. In addition to
11063 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11064 macro. On some Alpha implementations the resulting code may execute
11065 significantly slower than the code generated by default. Since there is
11066 very little code that depends on the @var{inexact-flag}, you should
11067 normally not specify this option. Other Alpha compilers call this
11068 option @option{-ieee_with_inexact}.
11070 @item -mfp-trap-mode=@var{trap-mode}
11071 @opindex mfp-trap-mode
11072 This option controls what floating-point related traps are enabled.
11073 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11074 The trap mode can be set to one of four values:
11078 This is the default (normal) setting. The only traps that are enabled
11079 are the ones that cannot be disabled in software (e.g., division by zero
11083 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11087 Like @samp{u}, but the instructions are marked to be safe for software
11088 completion (see Alpha architecture manual for details).
11091 Like @samp{su}, but inexact traps are enabled as well.
11094 @item -mfp-rounding-mode=@var{rounding-mode}
11095 @opindex mfp-rounding-mode
11096 Selects the IEEE rounding mode. Other Alpha compilers call this option
11097 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11102 Normal IEEE rounding mode. Floating point numbers are rounded towards
11103 the nearest machine number or towards the even machine number in case
11107 Round towards minus infinity.
11110 Chopped rounding mode. Floating point numbers are rounded towards zero.
11113 Dynamic rounding mode. A field in the floating point control register
11114 (@var{fpcr}, see Alpha architecture reference manual) controls the
11115 rounding mode in effect. The C library initializes this register for
11116 rounding towards plus infinity. Thus, unless your program modifies the
11117 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11120 @item -mtrap-precision=@var{trap-precision}
11121 @opindex mtrap-precision
11122 In the Alpha architecture, floating point traps are imprecise. This
11123 means without software assistance it is impossible to recover from a
11124 floating trap and program execution normally needs to be terminated.
11125 GCC can generate code that can assist operating system trap handlers
11126 in determining the exact location that caused a floating point trap.
11127 Depending on the requirements of an application, different levels of
11128 precisions can be selected:
11132 Program precision. This option is the default and means a trap handler
11133 can only identify which program caused a floating point exception.
11136 Function precision. The trap handler can determine the function that
11137 caused a floating point exception.
11140 Instruction precision. The trap handler can determine the exact
11141 instruction that caused a floating point exception.
11144 Other Alpha compilers provide the equivalent options called
11145 @option{-scope_safe} and @option{-resumption_safe}.
11147 @item -mieee-conformant
11148 @opindex mieee-conformant
11149 This option marks the generated code as IEEE conformant. You must not
11150 use this option unless you also specify @option{-mtrap-precision=i} and either
11151 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11152 is to emit the line @samp{.eflag 48} in the function prologue of the
11153 generated assembly file. Under DEC Unix, this has the effect that
11154 IEEE-conformant math library routines will be linked in.
11156 @item -mbuild-constants
11157 @opindex mbuild-constants
11158 Normally GCC examines a 32- or 64-bit integer constant to
11159 see if it can construct it from smaller constants in two or three
11160 instructions. If it cannot, it will output the constant as a literal and
11161 generate code to load it from the data segment at runtime.
11163 Use this option to require GCC to construct @emph{all} integer constants
11164 using code, even if it takes more instructions (the maximum is six).
11166 You would typically use this option to build a shared library dynamic
11167 loader. Itself a shared library, it must relocate itself in memory
11168 before it can find the variables and constants in its own data segment.
11174 Select whether to generate code to be assembled by the vendor-supplied
11175 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11193 Indicate whether GCC should generate code to use the optional BWX,
11194 CIX, FIX and MAX instruction sets. The default is to use the instruction
11195 sets supported by the CPU type specified via @option{-mcpu=} option or that
11196 of the CPU on which GCC was built if none was specified.
11199 @itemx -mfloat-ieee
11200 @opindex mfloat-vax
11201 @opindex mfloat-ieee
11202 Generate code that uses (does not use) VAX F and G floating point
11203 arithmetic instead of IEEE single and double precision.
11205 @item -mexplicit-relocs
11206 @itemx -mno-explicit-relocs
11207 @opindex mexplicit-relocs
11208 @opindex mno-explicit-relocs
11209 Older Alpha assemblers provided no way to generate symbol relocations
11210 except via assembler macros. Use of these macros does not allow
11211 optimal instruction scheduling. GNU binutils as of version 2.12
11212 supports a new syntax that allows the compiler to explicitly mark
11213 which relocations should apply to which instructions. This option
11214 is mostly useful for debugging, as GCC detects the capabilities of
11215 the assembler when it is built and sets the default accordingly.
11218 @itemx -mlarge-data
11219 @opindex msmall-data
11220 @opindex mlarge-data
11221 When @option{-mexplicit-relocs} is in effect, static data is
11222 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11223 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11224 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11225 16-bit relocations off of the @code{$gp} register. This limits the
11226 size of the small data area to 64KB, but allows the variables to be
11227 directly accessed via a single instruction.
11229 The default is @option{-mlarge-data}. With this option the data area
11230 is limited to just below 2GB@. Programs that require more than 2GB of
11231 data must use @code{malloc} or @code{mmap} to allocate the data in the
11232 heap instead of in the program's data segment.
11234 When generating code for shared libraries, @option{-fpic} implies
11235 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11238 @itemx -mlarge-text
11239 @opindex msmall-text
11240 @opindex mlarge-text
11241 When @option{-msmall-text} is used, the compiler assumes that the
11242 code of the entire program (or shared library) fits in 4MB, and is
11243 thus reachable with a branch instruction. When @option{-msmall-data}
11244 is used, the compiler can assume that all local symbols share the
11245 same @code{$gp} value, and thus reduce the number of instructions
11246 required for a function call from 4 to 1.
11248 The default is @option{-mlarge-text}.
11250 @item -mcpu=@var{cpu_type}
11252 Set the instruction set and instruction scheduling parameters for
11253 machine type @var{cpu_type}. You can specify either the @samp{EV}
11254 style name or the corresponding chip number. GCC supports scheduling
11255 parameters for the EV4, EV5 and EV6 family of processors and will
11256 choose the default values for the instruction set from the processor
11257 you specify. If you do not specify a processor type, GCC will default
11258 to the processor on which the compiler was built.
11260 Supported values for @var{cpu_type} are
11266 Schedules as an EV4 and has no instruction set extensions.
11270 Schedules as an EV5 and has no instruction set extensions.
11274 Schedules as an EV5 and supports the BWX extension.
11279 Schedules as an EV5 and supports the BWX and MAX extensions.
11283 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11287 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11290 Native Linux/GNU toolchains also support the value @samp{native},
11291 which selects the best architecture option for the host processor.
11292 @option{-mcpu=native} has no effect if GCC does not recognize
11295 @item -mtune=@var{cpu_type}
11297 Set only the instruction scheduling parameters for machine type
11298 @var{cpu_type}. The instruction set is not changed.
11300 Native Linux/GNU toolchains also support the value @samp{native},
11301 which selects the best architecture option for the host processor.
11302 @option{-mtune=native} has no effect if GCC does not recognize
11305 @item -mmemory-latency=@var{time}
11306 @opindex mmemory-latency
11307 Sets the latency the scheduler should assume for typical memory
11308 references as seen by the application. This number is highly
11309 dependent on the memory access patterns used by the application
11310 and the size of the external cache on the machine.
11312 Valid options for @var{time} are
11316 A decimal number representing clock cycles.
11322 The compiler contains estimates of the number of clock cycles for
11323 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11324 (also called Dcache, Scache, and Bcache), as well as to main memory.
11325 Note that L3 is only valid for EV5.
11330 @node DEC Alpha/VMS Options
11331 @subsection DEC Alpha/VMS Options
11333 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11336 @item -mvms-return-codes
11337 @opindex mvms-return-codes
11338 Return VMS condition codes from main. The default is to return POSIX
11339 style condition (e.g.@: error) codes.
11341 @item -mdebug-main=@var{prefix}
11342 @opindex mdebug-main=@var{prefix}
11343 Flag the first routine whose name starts with @var{prefix} as the main
11344 routine for the debugger.
11348 Default to 64bit memory allocation routines.
11352 @subsection FR30 Options
11353 @cindex FR30 Options
11355 These options are defined specifically for the FR30 port.
11359 @item -msmall-model
11360 @opindex msmall-model
11361 Use the small address space model. This can produce smaller code, but
11362 it does assume that all symbolic values and addresses will fit into a
11367 Assume that run-time support has been provided and so there is no need
11368 to include the simulator library (@file{libsim.a}) on the linker
11374 @subsection FRV Options
11375 @cindex FRV Options
11381 Only use the first 32 general purpose registers.
11386 Use all 64 general purpose registers.
11391 Use only the first 32 floating point registers.
11396 Use all 64 floating point registers
11399 @opindex mhard-float
11401 Use hardware instructions for floating point operations.
11404 @opindex msoft-float
11406 Use library routines for floating point operations.
11411 Dynamically allocate condition code registers.
11416 Do not try to dynamically allocate condition code registers, only
11417 use @code{icc0} and @code{fcc0}.
11422 Change ABI to use double word insns.
11427 Do not use double word instructions.
11432 Use floating point double instructions.
11435 @opindex mno-double
11437 Do not use floating point double instructions.
11442 Use media instructions.
11447 Do not use media instructions.
11452 Use multiply and add/subtract instructions.
11455 @opindex mno-muladd
11457 Do not use multiply and add/subtract instructions.
11462 Select the FDPIC ABI, that uses function descriptors to represent
11463 pointers to functions. Without any PIC/PIE-related options, it
11464 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11465 assumes GOT entries and small data are within a 12-bit range from the
11466 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11467 are computed with 32 bits.
11468 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11471 @opindex minline-plt
11473 Enable inlining of PLT entries in function calls to functions that are
11474 not known to bind locally. It has no effect without @option{-mfdpic}.
11475 It's enabled by default if optimizing for speed and compiling for
11476 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11477 optimization option such as @option{-O3} or above is present in the
11483 Assume a large TLS segment when generating thread-local code.
11488 Do not assume a large TLS segment when generating thread-local code.
11493 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11494 that is known to be in read-only sections. It's enabled by default,
11495 except for @option{-fpic} or @option{-fpie}: even though it may help
11496 make the global offset table smaller, it trades 1 instruction for 4.
11497 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11498 one of which may be shared by multiple symbols, and it avoids the need
11499 for a GOT entry for the referenced symbol, so it's more likely to be a
11500 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11502 @item -multilib-library-pic
11503 @opindex multilib-library-pic
11505 Link with the (library, not FD) pic libraries. It's implied by
11506 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11507 @option{-fpic} without @option{-mfdpic}. You should never have to use
11511 @opindex mlinked-fp
11513 Follow the EABI requirement of always creating a frame pointer whenever
11514 a stack frame is allocated. This option is enabled by default and can
11515 be disabled with @option{-mno-linked-fp}.
11518 @opindex mlong-calls
11520 Use indirect addressing to call functions outside the current
11521 compilation unit. This allows the functions to be placed anywhere
11522 within the 32-bit address space.
11524 @item -malign-labels
11525 @opindex malign-labels
11527 Try to align labels to an 8-byte boundary by inserting nops into the
11528 previous packet. This option only has an effect when VLIW packing
11529 is enabled. It doesn't create new packets; it merely adds nops to
11532 @item -mlibrary-pic
11533 @opindex mlibrary-pic
11535 Generate position-independent EABI code.
11540 Use only the first four media accumulator registers.
11545 Use all eight media accumulator registers.
11550 Pack VLIW instructions.
11555 Do not pack VLIW instructions.
11558 @opindex mno-eflags
11560 Do not mark ABI switches in e_flags.
11563 @opindex mcond-move
11565 Enable the use of conditional-move instructions (default).
11567 This switch is mainly for debugging the compiler and will likely be removed
11568 in a future version.
11570 @item -mno-cond-move
11571 @opindex mno-cond-move
11573 Disable the use of conditional-move instructions.
11575 This switch is mainly for debugging the compiler and will likely be removed
11576 in a future version.
11581 Enable the use of conditional set instructions (default).
11583 This switch is mainly for debugging the compiler and will likely be removed
11584 in a future version.
11589 Disable the use of conditional set instructions.
11591 This switch is mainly for debugging the compiler and will likely be removed
11592 in a future version.
11595 @opindex mcond-exec
11597 Enable the use of conditional execution (default).
11599 This switch is mainly for debugging the compiler and will likely be removed
11600 in a future version.
11602 @item -mno-cond-exec
11603 @opindex mno-cond-exec
11605 Disable the use of conditional execution.
11607 This switch is mainly for debugging the compiler and will likely be removed
11608 in a future version.
11610 @item -mvliw-branch
11611 @opindex mvliw-branch
11613 Run a pass to pack branches into VLIW instructions (default).
11615 This switch is mainly for debugging the compiler and will likely be removed
11616 in a future version.
11618 @item -mno-vliw-branch
11619 @opindex mno-vliw-branch
11621 Do not run a pass to pack branches into VLIW instructions.
11623 This switch is mainly for debugging the compiler and will likely be removed
11624 in a future version.
11626 @item -mmulti-cond-exec
11627 @opindex mmulti-cond-exec
11629 Enable optimization of @code{&&} and @code{||} in conditional execution
11632 This switch is mainly for debugging the compiler and will likely be removed
11633 in a future version.
11635 @item -mno-multi-cond-exec
11636 @opindex mno-multi-cond-exec
11638 Disable optimization of @code{&&} and @code{||} in conditional execution.
11640 This switch is mainly for debugging the compiler and will likely be removed
11641 in a future version.
11643 @item -mnested-cond-exec
11644 @opindex mnested-cond-exec
11646 Enable nested conditional execution optimizations (default).
11648 This switch is mainly for debugging the compiler and will likely be removed
11649 in a future version.
11651 @item -mno-nested-cond-exec
11652 @opindex mno-nested-cond-exec
11654 Disable nested conditional execution optimizations.
11656 This switch is mainly for debugging the compiler and will likely be removed
11657 in a future version.
11659 @item -moptimize-membar
11660 @opindex moptimize-membar
11662 This switch removes redundant @code{membar} instructions from the
11663 compiler generated code. It is enabled by default.
11665 @item -mno-optimize-membar
11666 @opindex mno-optimize-membar
11668 This switch disables the automatic removal of redundant @code{membar}
11669 instructions from the generated code.
11671 @item -mtomcat-stats
11672 @opindex mtomcat-stats
11674 Cause gas to print out tomcat statistics.
11676 @item -mcpu=@var{cpu}
11679 Select the processor type for which to generate code. Possible values are
11680 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11681 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11685 @node GNU/Linux Options
11686 @subsection GNU/Linux Options
11688 These @samp{-m} options are defined for GNU/Linux targets:
11693 Use the GNU C library. This is the default except
11694 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11698 Use uClibc C library. This is the default on
11699 @samp{*-*-linux-*uclibc*} targets.
11703 Use Bionic C library. This is the default on
11704 @samp{*-*-linux-*android*} targets.
11708 Compile code compatible with Android platform. This is the default on
11709 @samp{*-*-linux-*android*} targets.
11711 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11712 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11713 this option makes the GCC driver pass Android-specific options to the linker.
11714 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11717 @item -tno-android-cc
11718 @opindex tno-android-cc
11719 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11720 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11721 @option{-fno-rtti} by default.
11723 @item -tno-android-ld
11724 @opindex tno-android-ld
11725 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11726 linking options to the linker.
11730 @node H8/300 Options
11731 @subsection H8/300 Options
11733 These @samp{-m} options are defined for the H8/300 implementations:
11738 Shorten some address references at link time, when possible; uses the
11739 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11740 ld, Using ld}, for a fuller description.
11744 Generate code for the H8/300H@.
11748 Generate code for the H8S@.
11752 Generate code for the H8S and H8/300H in the normal mode. This switch
11753 must be used either with @option{-mh} or @option{-ms}.
11757 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11761 Make @code{int} data 32 bits by default.
11764 @opindex malign-300
11765 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11766 The default for the H8/300H and H8S is to align longs and floats on 4
11768 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11769 This option has no effect on the H8/300.
11773 @subsection HPPA Options
11774 @cindex HPPA Options
11776 These @samp{-m} options are defined for the HPPA family of computers:
11779 @item -march=@var{architecture-type}
11781 Generate code for the specified architecture. The choices for
11782 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11783 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11784 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11785 architecture option for your machine. Code compiled for lower numbered
11786 architectures will run on higher numbered architectures, but not the
11789 @item -mpa-risc-1-0
11790 @itemx -mpa-risc-1-1
11791 @itemx -mpa-risc-2-0
11792 @opindex mpa-risc-1-0
11793 @opindex mpa-risc-1-1
11794 @opindex mpa-risc-2-0
11795 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11798 @opindex mbig-switch
11799 Generate code suitable for big switch tables. Use this option only if
11800 the assembler/linker complain about out of range branches within a switch
11803 @item -mjump-in-delay
11804 @opindex mjump-in-delay
11805 Fill delay slots of function calls with unconditional jump instructions
11806 by modifying the return pointer for the function call to be the target
11807 of the conditional jump.
11809 @item -mdisable-fpregs
11810 @opindex mdisable-fpregs
11811 Prevent floating point registers from being used in any manner. This is
11812 necessary for compiling kernels which perform lazy context switching of
11813 floating point registers. If you use this option and attempt to perform
11814 floating point operations, the compiler will abort.
11816 @item -mdisable-indexing
11817 @opindex mdisable-indexing
11818 Prevent the compiler from using indexing address modes. This avoids some
11819 rather obscure problems when compiling MIG generated code under MACH@.
11821 @item -mno-space-regs
11822 @opindex mno-space-regs
11823 Generate code that assumes the target has no space registers. This allows
11824 GCC to generate faster indirect calls and use unscaled index address modes.
11826 Such code is suitable for level 0 PA systems and kernels.
11828 @item -mfast-indirect-calls
11829 @opindex mfast-indirect-calls
11830 Generate code that assumes calls never cross space boundaries. This
11831 allows GCC to emit code which performs faster indirect calls.
11833 This option will not work in the presence of shared libraries or nested
11836 @item -mfixed-range=@var{register-range}
11837 @opindex mfixed-range
11838 Generate code treating the given register range as fixed registers.
11839 A fixed register is one that the register allocator can not use. This is
11840 useful when compiling kernel code. A register range is specified as
11841 two registers separated by a dash. Multiple register ranges can be
11842 specified separated by a comma.
11844 @item -mlong-load-store
11845 @opindex mlong-load-store
11846 Generate 3-instruction load and store sequences as sometimes required by
11847 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11850 @item -mportable-runtime
11851 @opindex mportable-runtime
11852 Use the portable calling conventions proposed by HP for ELF systems.
11856 Enable the use of assembler directives only GAS understands.
11858 @item -mschedule=@var{cpu-type}
11860 Schedule code according to the constraints for the machine type
11861 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11862 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11863 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11864 proper scheduling option for your machine. The default scheduling is
11868 @opindex mlinker-opt
11869 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11870 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11871 linkers in which they give bogus error messages when linking some programs.
11874 @opindex msoft-float
11875 Generate output containing library calls for floating point.
11876 @strong{Warning:} the requisite libraries are not available for all HPPA
11877 targets. Normally the facilities of the machine's usual C compiler are
11878 used, but this cannot be done directly in cross-compilation. You must make
11879 your own arrangements to provide suitable library functions for
11882 @option{-msoft-float} changes the calling convention in the output file;
11883 therefore, it is only useful if you compile @emph{all} of a program with
11884 this option. In particular, you need to compile @file{libgcc.a}, the
11885 library that comes with GCC, with @option{-msoft-float} in order for
11890 Generate the predefine, @code{_SIO}, for server IO@. The default is
11891 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11892 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11893 options are available under HP-UX and HI-UX@.
11897 Use GNU ld specific options. This passes @option{-shared} to ld when
11898 building a shared library. It is the default when GCC is configured,
11899 explicitly or implicitly, with the GNU linker. This option does not
11900 have any affect on which ld is called, it only changes what parameters
11901 are passed to that ld. The ld that is called is determined by the
11902 @option{--with-ld} configure option, GCC's program search path, and
11903 finally by the user's @env{PATH}. The linker used by GCC can be printed
11904 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11905 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11909 Use HP ld specific options. This passes @option{-b} to ld when building
11910 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11911 links. It is the default when GCC is configured, explicitly or
11912 implicitly, with the HP linker. This option does not have any affect on
11913 which ld is called, it only changes what parameters are passed to that
11914 ld. The ld that is called is determined by the @option{--with-ld}
11915 configure option, GCC's program search path, and finally by the user's
11916 @env{PATH}. The linker used by GCC can be printed using @samp{which
11917 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11918 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11921 @opindex mno-long-calls
11922 Generate code that uses long call sequences. This ensures that a call
11923 is always able to reach linker generated stubs. The default is to generate
11924 long calls only when the distance from the call site to the beginning
11925 of the function or translation unit, as the case may be, exceeds a
11926 predefined limit set by the branch type being used. The limits for
11927 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11928 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11931 Distances are measured from the beginning of functions when using the
11932 @option{-ffunction-sections} option, or when using the @option{-mgas}
11933 and @option{-mno-portable-runtime} options together under HP-UX with
11936 It is normally not desirable to use this option as it will degrade
11937 performance. However, it may be useful in large applications,
11938 particularly when partial linking is used to build the application.
11940 The types of long calls used depends on the capabilities of the
11941 assembler and linker, and the type of code being generated. The
11942 impact on systems that support long absolute calls, and long pic
11943 symbol-difference or pc-relative calls should be relatively small.
11944 However, an indirect call is used on 32-bit ELF systems in pic code
11945 and it is quite long.
11947 @item -munix=@var{unix-std}
11949 Generate compiler predefines and select a startfile for the specified
11950 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11951 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11952 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11953 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11954 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11957 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11958 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11959 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11960 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11961 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11962 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11964 It is @emph{important} to note that this option changes the interfaces
11965 for various library routines. It also affects the operational behavior
11966 of the C library. Thus, @emph{extreme} care is needed in using this
11969 Library code that is intended to operate with more than one UNIX
11970 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11971 as appropriate. Most GNU software doesn't provide this capability.
11975 Suppress the generation of link options to search libdld.sl when the
11976 @option{-static} option is specified on HP-UX 10 and later.
11980 The HP-UX implementation of setlocale in libc has a dependency on
11981 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11982 when the @option{-static} option is specified, special link options
11983 are needed to resolve this dependency.
11985 On HP-UX 10 and later, the GCC driver adds the necessary options to
11986 link with libdld.sl when the @option{-static} option is specified.
11987 This causes the resulting binary to be dynamic. On the 64-bit port,
11988 the linkers generate dynamic binaries by default in any case. The
11989 @option{-nolibdld} option can be used to prevent the GCC driver from
11990 adding these link options.
11994 Add support for multithreading with the @dfn{dce thread} library
11995 under HP-UX@. This option sets flags for both the preprocessor and
11999 @node i386 and x86-64 Options
12000 @subsection Intel 386 and AMD x86-64 Options
12001 @cindex i386 Options
12002 @cindex x86-64 Options
12003 @cindex Intel 386 Options
12004 @cindex AMD x86-64 Options
12006 These @samp{-m} options are defined for the i386 and x86-64 family of
12010 @item -mtune=@var{cpu-type}
12012 Tune to @var{cpu-type} everything applicable about the generated code, except
12013 for the ABI and the set of available instructions. The choices for
12014 @var{cpu-type} are:
12017 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12018 If you know the CPU on which your code will run, then you should use
12019 the corresponding @option{-mtune} option instead of
12020 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12021 of your application will have, then you should use this option.
12023 As new processors are deployed in the marketplace, the behavior of this
12024 option will change. Therefore, if you upgrade to a newer version of
12025 GCC, the code generated option will change to reflect the processors
12026 that were most common when that version of GCC was released.
12028 There is no @option{-march=generic} option because @option{-march}
12029 indicates the instruction set the compiler can use, and there is no
12030 generic instruction set applicable to all processors. In contrast,
12031 @option{-mtune} indicates the processor (or, in this case, collection of
12032 processors) for which the code is optimized.
12034 This selects the CPU to tune for at compilation time by determining
12035 the processor type of the compiling machine. Using @option{-mtune=native}
12036 will produce code optimized for the local machine under the constraints
12037 of the selected instruction set. Using @option{-march=native} will
12038 enable all instruction subsets supported by the local machine (hence
12039 the result might not run on different machines).
12041 Original Intel's i386 CPU@.
12043 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12044 @item i586, pentium
12045 Intel Pentium CPU with no MMX support.
12047 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12049 Intel PentiumPro CPU@.
12051 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12052 instruction set will be used, so the code will run on all i686 family chips.
12054 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12055 @item pentium3, pentium3m
12056 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12059 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12060 support. Used by Centrino notebooks.
12061 @item pentium4, pentium4m
12062 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12064 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12067 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12068 SSE2 and SSE3 instruction set support.
12070 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12071 instruction set support.
12073 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12074 and SSE4.2 instruction set support.
12076 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12077 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12079 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12080 instruction set support.
12082 AMD K6 CPU with MMX instruction set support.
12084 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12085 @item athlon, athlon-tbird
12086 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12088 @item athlon-4, athlon-xp, athlon-mp
12089 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12090 instruction set support.
12091 @item k8, opteron, athlon64, athlon-fx
12092 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12093 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12094 @item k8-sse3, opteron-sse3, athlon64-sse3
12095 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12096 @item amdfam10, barcelona
12097 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12098 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12099 instruction set extensions.)
12101 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12104 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12105 instruction set support.
12107 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12108 implemented for this chip.)
12110 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12111 implemented for this chip.)
12113 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12116 While picking a specific @var{cpu-type} will schedule things appropriately
12117 for that particular chip, the compiler will not generate any code that
12118 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12121 @item -march=@var{cpu-type}
12123 Generate instructions for the machine type @var{cpu-type}. The choices
12124 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12125 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12127 @item -mcpu=@var{cpu-type}
12129 A deprecated synonym for @option{-mtune}.
12131 @item -mfpmath=@var{unit}
12133 Generate floating point arithmetics for selected unit @var{unit}. The choices
12134 for @var{unit} are:
12138 Use the standard 387 floating point coprocessor present majority of chips and
12139 emulated otherwise. Code compiled with this option will run almost everywhere.
12140 The temporary results are computed in 80bit precision instead of precision
12141 specified by the type resulting in slightly different results compared to most
12142 of other chips. See @option{-ffloat-store} for more detailed description.
12144 This is the default choice for i386 compiler.
12147 Use scalar floating point instructions present in the SSE instruction set.
12148 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12149 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12150 instruction set supports only single precision arithmetics, thus the double and
12151 extended precision arithmetics is still done using 387. Later version, present
12152 only in Pentium4 and the future AMD x86-64 chips supports double precision
12155 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12156 or @option{-msse2} switches to enable SSE extensions and make this option
12157 effective. For the x86-64 compiler, these extensions are enabled by default.
12159 The resulting code should be considerably faster in the majority of cases and avoid
12160 the numerical instability problems of 387 code, but may break some existing
12161 code that expects temporaries to be 80bit.
12163 This is the default choice for the x86-64 compiler.
12168 Attempt to utilize both instruction sets at once. This effectively double the
12169 amount of available registers and on chips with separate execution units for
12170 387 and SSE the execution resources too. Use this option with care, as it is
12171 still experimental, because the GCC register allocator does not model separate
12172 functional units well resulting in instable performance.
12175 @item -masm=@var{dialect}
12176 @opindex masm=@var{dialect}
12177 Output asm instructions using selected @var{dialect}. Supported
12178 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12179 not support @samp{intel}.
12182 @itemx -mno-ieee-fp
12184 @opindex mno-ieee-fp
12185 Control whether or not the compiler uses IEEE floating point
12186 comparisons. These handle correctly the case where the result of a
12187 comparison is unordered.
12190 @opindex msoft-float
12191 Generate output containing library calls for floating point.
12192 @strong{Warning:} the requisite libraries are not part of GCC@.
12193 Normally the facilities of the machine's usual C compiler are used, but
12194 this can't be done directly in cross-compilation. You must make your
12195 own arrangements to provide suitable library functions for
12198 On machines where a function returns floating point results in the 80387
12199 register stack, some floating point opcodes may be emitted even if
12200 @option{-msoft-float} is used.
12202 @item -mno-fp-ret-in-387
12203 @opindex mno-fp-ret-in-387
12204 Do not use the FPU registers for return values of functions.
12206 The usual calling convention has functions return values of types
12207 @code{float} and @code{double} in an FPU register, even if there
12208 is no FPU@. The idea is that the operating system should emulate
12211 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12212 in ordinary CPU registers instead.
12214 @item -mno-fancy-math-387
12215 @opindex mno-fancy-math-387
12216 Some 387 emulators do not support the @code{sin}, @code{cos} and
12217 @code{sqrt} instructions for the 387. Specify this option to avoid
12218 generating those instructions. This option is the default on FreeBSD,
12219 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12220 indicates that the target CPU will always have an FPU and so the
12221 instruction will not need emulation. As of revision 2.6.1, these
12222 instructions are not generated unless you also use the
12223 @option{-funsafe-math-optimizations} switch.
12225 @item -malign-double
12226 @itemx -mno-align-double
12227 @opindex malign-double
12228 @opindex mno-align-double
12229 Control whether GCC aligns @code{double}, @code{long double}, and
12230 @code{long long} variables on a two word boundary or a one word
12231 boundary. Aligning @code{double} variables on a two word boundary will
12232 produce code that runs somewhat faster on a @samp{Pentium} at the
12233 expense of more memory.
12235 On x86-64, @option{-malign-double} is enabled by default.
12237 @strong{Warning:} if you use the @option{-malign-double} switch,
12238 structures containing the above types will be aligned differently than
12239 the published application binary interface specifications for the 386
12240 and will not be binary compatible with structures in code compiled
12241 without that switch.
12243 @item -m96bit-long-double
12244 @itemx -m128bit-long-double
12245 @opindex m96bit-long-double
12246 @opindex m128bit-long-double
12247 These switches control the size of @code{long double} type. The i386
12248 application binary interface specifies the size to be 96 bits,
12249 so @option{-m96bit-long-double} is the default in 32 bit mode.
12251 Modern architectures (Pentium and newer) would prefer @code{long double}
12252 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12253 conforming to the ABI, this would not be possible. So specifying a
12254 @option{-m128bit-long-double} will align @code{long double}
12255 to a 16 byte boundary by padding the @code{long double} with an additional
12258 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12259 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12261 Notice that neither of these options enable any extra precision over the x87
12262 standard of 80 bits for a @code{long double}.
12264 @strong{Warning:} if you override the default value for your target ABI, the
12265 structures and arrays containing @code{long double} variables will change
12266 their size as well as function calling convention for function taking
12267 @code{long double} will be modified. Hence they will not be binary
12268 compatible with arrays or structures in code compiled without that switch.
12270 @item -mlarge-data-threshold=@var{number}
12271 @opindex mlarge-data-threshold=@var{number}
12272 When @option{-mcmodel=medium} is specified, the data greater than
12273 @var{threshold} are placed in large data section. This value must be the
12274 same across all object linked into the binary and defaults to 65535.
12278 Use a different function-calling convention, in which functions that
12279 take a fixed number of arguments return with the @code{ret} @var{num}
12280 instruction, which pops their arguments while returning. This saves one
12281 instruction in the caller since there is no need to pop the arguments
12284 You can specify that an individual function is called with this calling
12285 sequence with the function attribute @samp{stdcall}. You can also
12286 override the @option{-mrtd} option by using the function attribute
12287 @samp{cdecl}. @xref{Function Attributes}.
12289 @strong{Warning:} this calling convention is incompatible with the one
12290 normally used on Unix, so you cannot use it if you need to call
12291 libraries compiled with the Unix compiler.
12293 Also, you must provide function prototypes for all functions that
12294 take variable numbers of arguments (including @code{printf});
12295 otherwise incorrect code will be generated for calls to those
12298 In addition, seriously incorrect code will result if you call a
12299 function with too many arguments. (Normally, extra arguments are
12300 harmlessly ignored.)
12302 @item -mregparm=@var{num}
12304 Control how many registers are used to pass integer arguments. By
12305 default, no registers are used to pass arguments, and at most 3
12306 registers can be used. You can control this behavior for a specific
12307 function by using the function attribute @samp{regparm}.
12308 @xref{Function Attributes}.
12310 @strong{Warning:} if you use this switch, and
12311 @var{num} is nonzero, then you must build all modules with the same
12312 value, including any libraries. This includes the system libraries and
12316 @opindex msseregparm
12317 Use SSE register passing conventions for float and double arguments
12318 and return values. You can control this behavior for a specific
12319 function by using the function attribute @samp{sseregparm}.
12320 @xref{Function Attributes}.
12322 @strong{Warning:} if you use this switch then you must build all
12323 modules with the same value, including any libraries. This includes
12324 the system libraries and startup modules.
12326 @item -mvect8-ret-in-mem
12327 @opindex mvect8-ret-in-mem
12328 Return 8-byte vectors in memory instead of MMX registers. This is the
12329 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12330 Studio compilers until version 12. Later compiler versions (starting
12331 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12332 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12333 you need to remain compatible with existing code produced by those
12334 previous compiler versions or older versions of GCC.
12343 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12344 is specified, the significands of results of floating-point operations are
12345 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12346 significands of results of floating-point operations to 53 bits (double
12347 precision) and @option{-mpc80} rounds the significands of results of
12348 floating-point operations to 64 bits (extended double precision), which is
12349 the default. When this option is used, floating-point operations in higher
12350 precisions are not available to the programmer without setting the FPU
12351 control word explicitly.
12353 Setting the rounding of floating-point operations to less than the default
12354 80 bits can speed some programs by 2% or more. Note that some mathematical
12355 libraries assume that extended precision (80 bit) floating-point operations
12356 are enabled by default; routines in such libraries could suffer significant
12357 loss of accuracy, typically through so-called "catastrophic cancellation",
12358 when this option is used to set the precision to less than extended precision.
12360 @item -mstackrealign
12361 @opindex mstackrealign
12362 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12363 option will generate an alternate prologue and epilogue that realigns the
12364 runtime stack if necessary. This supports mixing legacy codes that keep
12365 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12366 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12367 applicable to individual functions.
12369 @item -mpreferred-stack-boundary=@var{num}
12370 @opindex mpreferred-stack-boundary
12371 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12372 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12373 the default is 4 (16 bytes or 128 bits).
12375 @item -mincoming-stack-boundary=@var{num}
12376 @opindex mincoming-stack-boundary
12377 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12378 boundary. If @option{-mincoming-stack-boundary} is not specified,
12379 the one specified by @option{-mpreferred-stack-boundary} will be used.
12381 On Pentium and PentiumPro, @code{double} and @code{long double} values
12382 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12383 suffer significant run time performance penalties. On Pentium III, the
12384 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12385 properly if it is not 16 byte aligned.
12387 To ensure proper alignment of this values on the stack, the stack boundary
12388 must be as aligned as that required by any value stored on the stack.
12389 Further, every function must be generated such that it keeps the stack
12390 aligned. Thus calling a function compiled with a higher preferred
12391 stack boundary from a function compiled with a lower preferred stack
12392 boundary will most likely misalign the stack. It is recommended that
12393 libraries that use callbacks always use the default setting.
12395 This extra alignment does consume extra stack space, and generally
12396 increases code size. Code that is sensitive to stack space usage, such
12397 as embedded systems and operating system kernels, may want to reduce the
12398 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12425 @itemx -mno-fsgsbase
12455 These switches enable or disable the use of instructions in the MMX,
12456 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12457 F16C, SSE4A, FMA4, XOP, LWP, ABM, BMI, or 3DNow!@: extended instruction sets.
12458 These extensions are also available as built-in functions: see
12459 @ref{X86 Built-in Functions}, for details of the functions enabled and
12460 disabled by these switches.
12462 To have SSE/SSE2 instructions generated automatically from floating-point
12463 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12465 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12466 generates new AVX instructions or AVX equivalence for all SSEx instructions
12469 These options will enable GCC to use these extended instructions in
12470 generated code, even without @option{-mfpmath=sse}. Applications which
12471 perform runtime CPU detection must compile separate files for each
12472 supported architecture, using the appropriate flags. In particular,
12473 the file containing the CPU detection code should be compiled without
12477 @itemx -mno-fused-madd
12478 @opindex mfused-madd
12479 @opindex mno-fused-madd
12480 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12481 instructions. The default is to use these instructions.
12485 This option instructs GCC to emit a @code{cld} instruction in the prologue
12486 of functions that use string instructions. String instructions depend on
12487 the DF flag to select between autoincrement or autodecrement mode. While the
12488 ABI specifies the DF flag to be cleared on function entry, some operating
12489 systems violate this specification by not clearing the DF flag in their
12490 exception dispatchers. The exception handler can be invoked with the DF flag
12491 set which leads to wrong direction mode, when string instructions are used.
12492 This option can be enabled by default on 32-bit x86 targets by configuring
12493 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12494 instructions can be suppressed with the @option{-mno-cld} compiler option
12498 @opindex mvzeroupper
12499 This option instructs GCC to emit a @code{vzeroupper} instruction
12500 before a transfer of control flow out of the function to minimize
12501 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12506 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12507 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12508 data types. This is useful for high resolution counters that could be updated
12509 by multiple processors (or cores). This instruction is generated as part of
12510 atomic built-in functions: see @ref{Atomic Builtins} for details.
12514 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12515 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12516 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12517 SAHF are load and store instructions, respectively, for certain status flags.
12518 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12519 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12523 This option will enable GCC to use movbe instruction to implement
12524 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12528 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12529 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12530 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12534 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12535 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12536 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12537 variants) for single precision floating point arguments. These instructions
12538 are generated only when @option{-funsafe-math-optimizations} is enabled
12539 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12540 Note that while the throughput of the sequence is higher than the throughput
12541 of the non-reciprocal instruction, the precision of the sequence can be
12542 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12544 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12545 already with @option{-ffast-math} (or the above option combination), and
12546 doesn't need @option{-mrecip}.
12548 @item -mveclibabi=@var{type}
12549 @opindex mveclibabi
12550 Specifies the ABI type to use for vectorizing intrinsics using an
12551 external library. Supported types are @code{svml} for the Intel short
12552 vector math library and @code{acml} for the AMD math core library style
12553 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12554 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12555 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12556 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12557 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12558 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12559 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12560 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12561 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12562 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12563 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12564 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12565 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12566 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12567 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12568 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12569 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12570 compatible library will have to be specified at link time.
12572 @item -mabi=@var{name}
12574 Generate code for the specified calling convention. Permissible values
12575 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12576 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12577 ABI when targeting Windows. On all other systems, the default is the
12578 SYSV ABI. You can control this behavior for a specific function by
12579 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12580 @xref{Function Attributes}.
12583 @itemx -mno-push-args
12584 @opindex mpush-args
12585 @opindex mno-push-args
12586 Use PUSH operations to store outgoing parameters. This method is shorter
12587 and usually equally fast as method using SUB/MOV operations and is enabled
12588 by default. In some cases disabling it may improve performance because of
12589 improved scheduling and reduced dependencies.
12591 @item -maccumulate-outgoing-args
12592 @opindex maccumulate-outgoing-args
12593 If enabled, the maximum amount of space required for outgoing arguments will be
12594 computed in the function prologue. This is faster on most modern CPUs
12595 because of reduced dependencies, improved scheduling and reduced stack usage
12596 when preferred stack boundary is not equal to 2. The drawback is a notable
12597 increase in code size. This switch implies @option{-mno-push-args}.
12601 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12602 on thread-safe exception handling must compile and link all code with the
12603 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12604 @option{-D_MT}; when linking, it links in a special thread helper library
12605 @option{-lmingwthrd} which cleans up per thread exception handling data.
12607 @item -mno-align-stringops
12608 @opindex mno-align-stringops
12609 Do not align destination of inlined string operations. This switch reduces
12610 code size and improves performance in case the destination is already aligned,
12611 but GCC doesn't know about it.
12613 @item -minline-all-stringops
12614 @opindex minline-all-stringops
12615 By default GCC inlines string operations only when destination is known to be
12616 aligned at least to 4 byte boundary. This enables more inlining, increase code
12617 size, but may improve performance of code that depends on fast memcpy, strlen
12618 and memset for short lengths.
12620 @item -minline-stringops-dynamically
12621 @opindex minline-stringops-dynamically
12622 For string operation of unknown size, inline runtime checks so for small
12623 blocks inline code is used, while for large blocks library call is used.
12625 @item -mstringop-strategy=@var{alg}
12626 @opindex mstringop-strategy=@var{alg}
12627 Overwrite internal decision heuristic about particular algorithm to inline
12628 string operation with. The allowed values are @code{rep_byte},
12629 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12630 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12631 expanding inline loop, @code{libcall} for always expanding library call.
12633 @item -momit-leaf-frame-pointer
12634 @opindex momit-leaf-frame-pointer
12635 Don't keep the frame pointer in a register for leaf functions. This
12636 avoids the instructions to save, set up and restore frame pointers and
12637 makes an extra register available in leaf functions. The option
12638 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12639 which might make debugging harder.
12641 @item -mtls-direct-seg-refs
12642 @itemx -mno-tls-direct-seg-refs
12643 @opindex mtls-direct-seg-refs
12644 Controls whether TLS variables may be accessed with offsets from the
12645 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12646 or whether the thread base pointer must be added. Whether or not this
12647 is legal depends on the operating system, and whether it maps the
12648 segment to cover the entire TLS area.
12650 For systems that use GNU libc, the default is on.
12653 @itemx -mno-sse2avx
12655 Specify that the assembler should encode SSE instructions with VEX
12656 prefix. The option @option{-mavx} turns this on by default.
12661 If profiling is active @option{-pg} put the profiling
12662 counter call before prologue.
12663 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12664 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12667 @itemx -mno-8bit-idiv
12669 On some processors, like Intel Atom, 8bit unsigned integer divide is
12670 much faster than 32bit/64bit integer divide. This option will generate a
12671 runt-time check. If both dividend and divisor are within range of 0
12672 to 255, 8bit unsigned integer divide will be used instead of
12673 32bit/64bit integer divide.
12675 @item -mavx256-split-unaligned-load
12676 @item -mavx256-split-unaligned-store
12677 @opindex avx256-split-unaligned-load
12678 @opindex avx256-split-unaligned-store
12679 Split 32-byte AVX unaligned load and store.
12683 These @samp{-m} switches are supported in addition to the above
12684 on AMD x86-64 processors in 64-bit environments.
12691 Generate code for a 32-bit or 64-bit environment.
12692 The 32-bit environment sets int, long and pointer to 32 bits and
12693 generates code that runs on any i386 system.
12694 The 64-bit environment sets int to 32 bits and long and pointer
12695 to 64 bits and generates code for AMD's x86-64 architecture. For
12696 darwin only the -m64 option turns off the @option{-fno-pic} and
12697 @option{-mdynamic-no-pic} options.
12699 @item -mno-red-zone
12700 @opindex mno-red-zone
12701 Do not use a so called red zone for x86-64 code. The red zone is mandated
12702 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12703 stack pointer that will not be modified by signal or interrupt handlers
12704 and therefore can be used for temporary data without adjusting the stack
12705 pointer. The flag @option{-mno-red-zone} disables this red zone.
12707 @item -mcmodel=small
12708 @opindex mcmodel=small
12709 Generate code for the small code model: the program and its symbols must
12710 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12711 Programs can be statically or dynamically linked. This is the default
12714 @item -mcmodel=kernel
12715 @opindex mcmodel=kernel
12716 Generate code for the kernel code model. The kernel runs in the
12717 negative 2 GB of the address space.
12718 This model has to be used for Linux kernel code.
12720 @item -mcmodel=medium
12721 @opindex mcmodel=medium
12722 Generate code for the medium model: The program is linked in the lower 2
12723 GB of the address space. Small symbols are also placed there. Symbols
12724 with sizes larger than @option{-mlarge-data-threshold} are put into
12725 large data or bss sections and can be located above 2GB. Programs can
12726 be statically or dynamically linked.
12728 @item -mcmodel=large
12729 @opindex mcmodel=large
12730 Generate code for the large model: This model makes no assumptions
12731 about addresses and sizes of sections.
12734 @node i386 and x86-64 Windows Options
12735 @subsection i386 and x86-64 Windows Options
12736 @cindex i386 and x86-64 Windows Options
12738 These additional options are available for Windows targets:
12743 This option is available for Cygwin and MinGW targets. It
12744 specifies that a console application is to be generated, by
12745 instructing the linker to set the PE header subsystem type
12746 required for console applications.
12747 This is the default behavior for Cygwin and MinGW targets.
12751 This option is available for Cygwin and MinGW targets. It
12752 specifies that a DLL - a dynamic link library - is to be
12753 generated, enabling the selection of the required runtime
12754 startup object and entry point.
12756 @item -mnop-fun-dllimport
12757 @opindex mnop-fun-dllimport
12758 This option is available for Cygwin and MinGW targets. It
12759 specifies that the dllimport attribute should be ignored.
12763 This option is available for MinGW targets. It specifies
12764 that MinGW-specific thread support is to be used.
12768 This option is available for mingw-w64 targets. It specifies
12769 that the UNICODE macro is getting pre-defined and that the
12770 unicode capable runtime startup code is chosen.
12774 This option is available for Cygwin and MinGW targets. It
12775 specifies that the typical Windows pre-defined macros are to
12776 be set in the pre-processor, but does not influence the choice
12777 of runtime library/startup code.
12781 This option is available for Cygwin and MinGW targets. It
12782 specifies that a GUI application is to be generated by
12783 instructing the linker to set the PE header subsystem type
12786 @item -fno-set-stack-executable
12787 @opindex fno-set-stack-executable
12788 This option is available for MinGW targets. It specifies that
12789 the executable flag for stack used by nested functions isn't
12790 set. This is necessary for binaries running in kernel mode of
12791 Windows, as there the user32 API, which is used to set executable
12792 privileges, isn't available.
12794 @item -mpe-aligned-commons
12795 @opindex mpe-aligned-commons
12796 This option is available for Cygwin and MinGW targets. It
12797 specifies that the GNU extension to the PE file format that
12798 permits the correct alignment of COMMON variables should be
12799 used when generating code. It will be enabled by default if
12800 GCC detects that the target assembler found during configuration
12801 supports the feature.
12804 See also under @ref{i386 and x86-64 Options} for standard options.
12806 @node IA-64 Options
12807 @subsection IA-64 Options
12808 @cindex IA-64 Options
12810 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12814 @opindex mbig-endian
12815 Generate code for a big endian target. This is the default for HP-UX@.
12817 @item -mlittle-endian
12818 @opindex mlittle-endian
12819 Generate code for a little endian target. This is the default for AIX5
12825 @opindex mno-gnu-as
12826 Generate (or don't) code for the GNU assembler. This is the default.
12827 @c Also, this is the default if the configure option @option{--with-gnu-as}
12833 @opindex mno-gnu-ld
12834 Generate (or don't) code for the GNU linker. This is the default.
12835 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12840 Generate code that does not use a global pointer register. The result
12841 is not position independent code, and violates the IA-64 ABI@.
12843 @item -mvolatile-asm-stop
12844 @itemx -mno-volatile-asm-stop
12845 @opindex mvolatile-asm-stop
12846 @opindex mno-volatile-asm-stop
12847 Generate (or don't) a stop bit immediately before and after volatile asm
12850 @item -mregister-names
12851 @itemx -mno-register-names
12852 @opindex mregister-names
12853 @opindex mno-register-names
12854 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12855 the stacked registers. This may make assembler output more readable.
12861 Disable (or enable) optimizations that use the small data section. This may
12862 be useful for working around optimizer bugs.
12864 @item -mconstant-gp
12865 @opindex mconstant-gp
12866 Generate code that uses a single constant global pointer value. This is
12867 useful when compiling kernel code.
12871 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12872 This is useful when compiling firmware code.
12874 @item -minline-float-divide-min-latency
12875 @opindex minline-float-divide-min-latency
12876 Generate code for inline divides of floating point values
12877 using the minimum latency algorithm.
12879 @item -minline-float-divide-max-throughput
12880 @opindex minline-float-divide-max-throughput
12881 Generate code for inline divides of floating point values
12882 using the maximum throughput algorithm.
12884 @item -mno-inline-float-divide
12885 @opindex mno-inline-float-divide
12886 Do not generate inline code for divides of floating point values.
12888 @item -minline-int-divide-min-latency
12889 @opindex minline-int-divide-min-latency
12890 Generate code for inline divides of integer values
12891 using the minimum latency algorithm.
12893 @item -minline-int-divide-max-throughput
12894 @opindex minline-int-divide-max-throughput
12895 Generate code for inline divides of integer values
12896 using the maximum throughput algorithm.
12898 @item -mno-inline-int-divide
12899 @opindex mno-inline-int-divide
12900 Do not generate inline code for divides of integer values.
12902 @item -minline-sqrt-min-latency
12903 @opindex minline-sqrt-min-latency
12904 Generate code for inline square roots
12905 using the minimum latency algorithm.
12907 @item -minline-sqrt-max-throughput
12908 @opindex minline-sqrt-max-throughput
12909 Generate code for inline square roots
12910 using the maximum throughput algorithm.
12912 @item -mno-inline-sqrt
12913 @opindex mno-inline-sqrt
12914 Do not generate inline code for sqrt.
12917 @itemx -mno-fused-madd
12918 @opindex mfused-madd
12919 @opindex mno-fused-madd
12920 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12921 instructions. The default is to use these instructions.
12923 @item -mno-dwarf2-asm
12924 @itemx -mdwarf2-asm
12925 @opindex mno-dwarf2-asm
12926 @opindex mdwarf2-asm
12927 Don't (or do) generate assembler code for the DWARF2 line number debugging
12928 info. This may be useful when not using the GNU assembler.
12930 @item -mearly-stop-bits
12931 @itemx -mno-early-stop-bits
12932 @opindex mearly-stop-bits
12933 @opindex mno-early-stop-bits
12934 Allow stop bits to be placed earlier than immediately preceding the
12935 instruction that triggered the stop bit. This can improve instruction
12936 scheduling, but does not always do so.
12938 @item -mfixed-range=@var{register-range}
12939 @opindex mfixed-range
12940 Generate code treating the given register range as fixed registers.
12941 A fixed register is one that the register allocator can not use. This is
12942 useful when compiling kernel code. A register range is specified as
12943 two registers separated by a dash. Multiple register ranges can be
12944 specified separated by a comma.
12946 @item -mtls-size=@var{tls-size}
12948 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12951 @item -mtune=@var{cpu-type}
12953 Tune the instruction scheduling for a particular CPU, Valid values are
12954 itanium, itanium1, merced, itanium2, and mckinley.
12960 Generate code for a 32-bit or 64-bit environment.
12961 The 32-bit environment sets int, long and pointer to 32 bits.
12962 The 64-bit environment sets int to 32 bits and long and pointer
12963 to 64 bits. These are HP-UX specific flags.
12965 @item -mno-sched-br-data-spec
12966 @itemx -msched-br-data-spec
12967 @opindex mno-sched-br-data-spec
12968 @opindex msched-br-data-spec
12969 (Dis/En)able data speculative scheduling before reload.
12970 This will result in generation of the ld.a instructions and
12971 the corresponding check instructions (ld.c / chk.a).
12972 The default is 'disable'.
12974 @item -msched-ar-data-spec
12975 @itemx -mno-sched-ar-data-spec
12976 @opindex msched-ar-data-spec
12977 @opindex mno-sched-ar-data-spec
12978 (En/Dis)able data speculative scheduling after reload.
12979 This will result in generation of the ld.a instructions and
12980 the corresponding check instructions (ld.c / chk.a).
12981 The default is 'enable'.
12983 @item -mno-sched-control-spec
12984 @itemx -msched-control-spec
12985 @opindex mno-sched-control-spec
12986 @opindex msched-control-spec
12987 (Dis/En)able control speculative scheduling. This feature is
12988 available only during region scheduling (i.e.@: before reload).
12989 This will result in generation of the ld.s instructions and
12990 the corresponding check instructions chk.s .
12991 The default is 'disable'.
12993 @item -msched-br-in-data-spec
12994 @itemx -mno-sched-br-in-data-spec
12995 @opindex msched-br-in-data-spec
12996 @opindex mno-sched-br-in-data-spec
12997 (En/Dis)able speculative scheduling of the instructions that
12998 are dependent on the data speculative loads before reload.
12999 This is effective only with @option{-msched-br-data-spec} enabled.
13000 The default is 'enable'.
13002 @item -msched-ar-in-data-spec
13003 @itemx -mno-sched-ar-in-data-spec
13004 @opindex msched-ar-in-data-spec
13005 @opindex mno-sched-ar-in-data-spec
13006 (En/Dis)able speculative scheduling of the instructions that
13007 are dependent on the data speculative loads after reload.
13008 This is effective only with @option{-msched-ar-data-spec} enabled.
13009 The default is 'enable'.
13011 @item -msched-in-control-spec
13012 @itemx -mno-sched-in-control-spec
13013 @opindex msched-in-control-spec
13014 @opindex mno-sched-in-control-spec
13015 (En/Dis)able speculative scheduling of the instructions that
13016 are dependent on the control speculative loads.
13017 This is effective only with @option{-msched-control-spec} enabled.
13018 The default is 'enable'.
13020 @item -mno-sched-prefer-non-data-spec-insns
13021 @itemx -msched-prefer-non-data-spec-insns
13022 @opindex mno-sched-prefer-non-data-spec-insns
13023 @opindex msched-prefer-non-data-spec-insns
13024 If enabled, data speculative instructions will be chosen for schedule
13025 only if there are no other choices at the moment. This will make
13026 the use of the data speculation much more conservative.
13027 The default is 'disable'.
13029 @item -mno-sched-prefer-non-control-spec-insns
13030 @itemx -msched-prefer-non-control-spec-insns
13031 @opindex mno-sched-prefer-non-control-spec-insns
13032 @opindex msched-prefer-non-control-spec-insns
13033 If enabled, control speculative instructions will be chosen for schedule
13034 only if there are no other choices at the moment. This will make
13035 the use of the control speculation much more conservative.
13036 The default is 'disable'.
13038 @item -mno-sched-count-spec-in-critical-path
13039 @itemx -msched-count-spec-in-critical-path
13040 @opindex mno-sched-count-spec-in-critical-path
13041 @opindex msched-count-spec-in-critical-path
13042 If enabled, speculative dependencies will be considered during
13043 computation of the instructions priorities. This will make the use of the
13044 speculation a bit more conservative.
13045 The default is 'disable'.
13047 @item -msched-spec-ldc
13048 @opindex msched-spec-ldc
13049 Use a simple data speculation check. This option is on by default.
13051 @item -msched-control-spec-ldc
13052 @opindex msched-spec-ldc
13053 Use a simple check for control speculation. This option is on by default.
13055 @item -msched-stop-bits-after-every-cycle
13056 @opindex msched-stop-bits-after-every-cycle
13057 Place a stop bit after every cycle when scheduling. This option is on
13060 @item -msched-fp-mem-deps-zero-cost
13061 @opindex msched-fp-mem-deps-zero-cost
13062 Assume that floating-point stores and loads are not likely to cause a conflict
13063 when placed into the same instruction group. This option is disabled by
13066 @item -msel-sched-dont-check-control-spec
13067 @opindex msel-sched-dont-check-control-spec
13068 Generate checks for control speculation in selective scheduling.
13069 This flag is disabled by default.
13071 @item -msched-max-memory-insns=@var{max-insns}
13072 @opindex msched-max-memory-insns
13073 Limit on the number of memory insns per instruction group, giving lower
13074 priority to subsequent memory insns attempting to schedule in the same
13075 instruction group. Frequently useful to prevent cache bank conflicts.
13076 The default value is 1.
13078 @item -msched-max-memory-insns-hard-limit
13079 @opindex msched-max-memory-insns-hard-limit
13080 Disallow more than `msched-max-memory-insns' in instruction group.
13081 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13082 when limit is reached but may still schedule memory operations.
13086 @node IA-64/VMS Options
13087 @subsection IA-64/VMS Options
13089 These @samp{-m} options are defined for the IA-64/VMS implementations:
13092 @item -mvms-return-codes
13093 @opindex mvms-return-codes
13094 Return VMS condition codes from main. The default is to return POSIX
13095 style condition (e.g.@ error) codes.
13097 @item -mdebug-main=@var{prefix}
13098 @opindex mdebug-main=@var{prefix}
13099 Flag the first routine whose name starts with @var{prefix} as the main
13100 routine for the debugger.
13104 Default to 64bit memory allocation routines.
13108 @subsection LM32 Options
13109 @cindex LM32 options
13111 These @option{-m} options are defined for the Lattice Mico32 architecture:
13114 @item -mbarrel-shift-enabled
13115 @opindex mbarrel-shift-enabled
13116 Enable barrel-shift instructions.
13118 @item -mdivide-enabled
13119 @opindex mdivide-enabled
13120 Enable divide and modulus instructions.
13122 @item -mmultiply-enabled
13123 @opindex multiply-enabled
13124 Enable multiply instructions.
13126 @item -msign-extend-enabled
13127 @opindex msign-extend-enabled
13128 Enable sign extend instructions.
13130 @item -muser-enabled
13131 @opindex muser-enabled
13132 Enable user-defined instructions.
13137 @subsection M32C Options
13138 @cindex M32C options
13141 @item -mcpu=@var{name}
13143 Select the CPU for which code is generated. @var{name} may be one of
13144 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13145 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13146 the M32C/80 series.
13150 Specifies that the program will be run on the simulator. This causes
13151 an alternate runtime library to be linked in which supports, for
13152 example, file I/O@. You must not use this option when generating
13153 programs that will run on real hardware; you must provide your own
13154 runtime library for whatever I/O functions are needed.
13156 @item -memregs=@var{number}
13158 Specifies the number of memory-based pseudo-registers GCC will use
13159 during code generation. These pseudo-registers will be used like real
13160 registers, so there is a tradeoff between GCC's ability to fit the
13161 code into available registers, and the performance penalty of using
13162 memory instead of registers. Note that all modules in a program must
13163 be compiled with the same value for this option. Because of that, you
13164 must not use this option with the default runtime libraries gcc
13169 @node M32R/D Options
13170 @subsection M32R/D Options
13171 @cindex M32R/D options
13173 These @option{-m} options are defined for Renesas M32R/D architectures:
13178 Generate code for the M32R/2@.
13182 Generate code for the M32R/X@.
13186 Generate code for the M32R@. This is the default.
13188 @item -mmodel=small
13189 @opindex mmodel=small
13190 Assume all objects live in the lower 16MB of memory (so that their addresses
13191 can be loaded with the @code{ld24} instruction), and assume all subroutines
13192 are reachable with the @code{bl} instruction.
13193 This is the default.
13195 The addressability of a particular object can be set with the
13196 @code{model} attribute.
13198 @item -mmodel=medium
13199 @opindex mmodel=medium
13200 Assume objects may be anywhere in the 32-bit address space (the compiler
13201 will generate @code{seth/add3} instructions to load their addresses), and
13202 assume all subroutines are reachable with the @code{bl} instruction.
13204 @item -mmodel=large
13205 @opindex mmodel=large
13206 Assume objects may be anywhere in the 32-bit address space (the compiler
13207 will generate @code{seth/add3} instructions to load their addresses), and
13208 assume subroutines may not be reachable with the @code{bl} instruction
13209 (the compiler will generate the much slower @code{seth/add3/jl}
13210 instruction sequence).
13213 @opindex msdata=none
13214 Disable use of the small data area. Variables will be put into
13215 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13216 @code{section} attribute has been specified).
13217 This is the default.
13219 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13220 Objects may be explicitly put in the small data area with the
13221 @code{section} attribute using one of these sections.
13223 @item -msdata=sdata
13224 @opindex msdata=sdata
13225 Put small global and static data in the small data area, but do not
13226 generate special code to reference them.
13229 @opindex msdata=use
13230 Put small global and static data in the small data area, and generate
13231 special instructions to reference them.
13235 @cindex smaller data references
13236 Put global and static objects less than or equal to @var{num} bytes
13237 into the small data or bss sections instead of the normal data or bss
13238 sections. The default value of @var{num} is 8.
13239 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13240 for this option to have any effect.
13242 All modules should be compiled with the same @option{-G @var{num}} value.
13243 Compiling with different values of @var{num} may or may not work; if it
13244 doesn't the linker will give an error message---incorrect code will not be
13249 Makes the M32R specific code in the compiler display some statistics
13250 that might help in debugging programs.
13252 @item -malign-loops
13253 @opindex malign-loops
13254 Align all loops to a 32-byte boundary.
13256 @item -mno-align-loops
13257 @opindex mno-align-loops
13258 Do not enforce a 32-byte alignment for loops. This is the default.
13260 @item -missue-rate=@var{number}
13261 @opindex missue-rate=@var{number}
13262 Issue @var{number} instructions per cycle. @var{number} can only be 1
13265 @item -mbranch-cost=@var{number}
13266 @opindex mbranch-cost=@var{number}
13267 @var{number} can only be 1 or 2. If it is 1 then branches will be
13268 preferred over conditional code, if it is 2, then the opposite will
13271 @item -mflush-trap=@var{number}
13272 @opindex mflush-trap=@var{number}
13273 Specifies the trap number to use to flush the cache. The default is
13274 12. Valid numbers are between 0 and 15 inclusive.
13276 @item -mno-flush-trap
13277 @opindex mno-flush-trap
13278 Specifies that the cache cannot be flushed by using a trap.
13280 @item -mflush-func=@var{name}
13281 @opindex mflush-func=@var{name}
13282 Specifies the name of the operating system function to call to flush
13283 the cache. The default is @emph{_flush_cache}, but a function call
13284 will only be used if a trap is not available.
13286 @item -mno-flush-func
13287 @opindex mno-flush-func
13288 Indicates that there is no OS function for flushing the cache.
13292 @node M680x0 Options
13293 @subsection M680x0 Options
13294 @cindex M680x0 options
13296 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13297 The default settings depend on which architecture was selected when
13298 the compiler was configured; the defaults for the most common choices
13302 @item -march=@var{arch}
13304 Generate code for a specific M680x0 or ColdFire instruction set
13305 architecture. Permissible values of @var{arch} for M680x0
13306 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13307 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13308 architectures are selected according to Freescale's ISA classification
13309 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13310 @samp{isab} and @samp{isac}.
13312 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13313 code for a ColdFire target. The @var{arch} in this macro is one of the
13314 @option{-march} arguments given above.
13316 When used together, @option{-march} and @option{-mtune} select code
13317 that runs on a family of similar processors but that is optimized
13318 for a particular microarchitecture.
13320 @item -mcpu=@var{cpu}
13322 Generate code for a specific M680x0 or ColdFire processor.
13323 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13324 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13325 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13326 below, which also classifies the CPUs into families:
13328 @multitable @columnfractions 0.20 0.80
13329 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13330 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13331 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13332 @item @samp{5206e} @tab @samp{5206e}
13333 @item @samp{5208} @tab @samp{5207} @samp{5208}
13334 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13335 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13336 @item @samp{5216} @tab @samp{5214} @samp{5216}
13337 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13338 @item @samp{5225} @tab @samp{5224} @samp{5225}
13339 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13340 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13341 @item @samp{5249} @tab @samp{5249}
13342 @item @samp{5250} @tab @samp{5250}
13343 @item @samp{5271} @tab @samp{5270} @samp{5271}
13344 @item @samp{5272} @tab @samp{5272}
13345 @item @samp{5275} @tab @samp{5274} @samp{5275}
13346 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13347 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13348 @item @samp{5307} @tab @samp{5307}
13349 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13350 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13351 @item @samp{5407} @tab @samp{5407}
13352 @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}
13355 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13356 @var{arch} is compatible with @var{cpu}. Other combinations of
13357 @option{-mcpu} and @option{-march} are rejected.
13359 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13360 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13361 where the value of @var{family} is given by the table above.
13363 @item -mtune=@var{tune}
13365 Tune the code for a particular microarchitecture, within the
13366 constraints set by @option{-march} and @option{-mcpu}.
13367 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13368 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13369 and @samp{cpu32}. The ColdFire microarchitectures
13370 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13372 You can also use @option{-mtune=68020-40} for code that needs
13373 to run relatively well on 68020, 68030 and 68040 targets.
13374 @option{-mtune=68020-60} is similar but includes 68060 targets
13375 as well. These two options select the same tuning decisions as
13376 @option{-m68020-40} and @option{-m68020-60} respectively.
13378 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13379 when tuning for 680x0 architecture @var{arch}. It also defines
13380 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13381 option is used. If gcc is tuning for a range of architectures,
13382 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13383 it defines the macros for every architecture in the range.
13385 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13386 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13387 of the arguments given above.
13393 Generate output for a 68000. This is the default
13394 when the compiler is configured for 68000-based systems.
13395 It is equivalent to @option{-march=68000}.
13397 Use this option for microcontrollers with a 68000 or EC000 core,
13398 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13402 Generate output for a 68010. This is the default
13403 when the compiler is configured for 68010-based systems.
13404 It is equivalent to @option{-march=68010}.
13410 Generate output for a 68020. This is the default
13411 when the compiler is configured for 68020-based systems.
13412 It is equivalent to @option{-march=68020}.
13416 Generate output for a 68030. This is the default when the compiler is
13417 configured for 68030-based systems. It is equivalent to
13418 @option{-march=68030}.
13422 Generate output for a 68040. This is the default when the compiler is
13423 configured for 68040-based systems. It is equivalent to
13424 @option{-march=68040}.
13426 This option inhibits the use of 68881/68882 instructions that have to be
13427 emulated by software on the 68040. Use this option if your 68040 does not
13428 have code to emulate those instructions.
13432 Generate output for a 68060. This is the default when the compiler is
13433 configured for 68060-based systems. It is equivalent to
13434 @option{-march=68060}.
13436 This option inhibits the use of 68020 and 68881/68882 instructions that
13437 have to be emulated by software on the 68060. Use this option if your 68060
13438 does not have code to emulate those instructions.
13442 Generate output for a CPU32. This is the default
13443 when the compiler is configured for CPU32-based systems.
13444 It is equivalent to @option{-march=cpu32}.
13446 Use this option for microcontrollers with a
13447 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13448 68336, 68340, 68341, 68349 and 68360.
13452 Generate output for a 520X ColdFire CPU@. This is the default
13453 when the compiler is configured for 520X-based systems.
13454 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13455 in favor of that option.
13457 Use this option for microcontroller with a 5200 core, including
13458 the MCF5202, MCF5203, MCF5204 and MCF5206.
13462 Generate output for a 5206e ColdFire CPU@. The option is now
13463 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13467 Generate output for a member of the ColdFire 528X family.
13468 The option is now deprecated in favor of the equivalent
13469 @option{-mcpu=528x}.
13473 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13474 in favor of the equivalent @option{-mcpu=5307}.
13478 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13479 in favor of the equivalent @option{-mcpu=5407}.
13483 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13484 This includes use of hardware floating point instructions.
13485 The option is equivalent to @option{-mcpu=547x}, and is now
13486 deprecated in favor of that option.
13490 Generate output for a 68040, without using any of the new instructions.
13491 This results in code which can run relatively efficiently on either a
13492 68020/68881 or a 68030 or a 68040. The generated code does use the
13493 68881 instructions that are emulated on the 68040.
13495 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13499 Generate output for a 68060, without using any of the new instructions.
13500 This results in code which can run relatively efficiently on either a
13501 68020/68881 or a 68030 or a 68040. The generated code does use the
13502 68881 instructions that are emulated on the 68060.
13504 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13508 @opindex mhard-float
13510 Generate floating-point instructions. This is the default for 68020
13511 and above, and for ColdFire devices that have an FPU@. It defines the
13512 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13513 on ColdFire targets.
13516 @opindex msoft-float
13517 Do not generate floating-point instructions; use library calls instead.
13518 This is the default for 68000, 68010, and 68832 targets. It is also
13519 the default for ColdFire devices that have no FPU.
13525 Generate (do not generate) ColdFire hardware divide and remainder
13526 instructions. If @option{-march} is used without @option{-mcpu},
13527 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13528 architectures. Otherwise, the default is taken from the target CPU
13529 (either the default CPU, or the one specified by @option{-mcpu}). For
13530 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13531 @option{-mcpu=5206e}.
13533 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13537 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13538 Additionally, parameters passed on the stack are also aligned to a
13539 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13543 Do not consider type @code{int} to be 16 bits wide. This is the default.
13546 @itemx -mno-bitfield
13547 @opindex mnobitfield
13548 @opindex mno-bitfield
13549 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13550 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13554 Do use the bit-field instructions. The @option{-m68020} option implies
13555 @option{-mbitfield}. This is the default if you use a configuration
13556 designed for a 68020.
13560 Use a different function-calling convention, in which functions
13561 that take a fixed number of arguments return with the @code{rtd}
13562 instruction, which pops their arguments while returning. This
13563 saves one instruction in the caller since there is no need to pop
13564 the arguments there.
13566 This calling convention is incompatible with the one normally
13567 used on Unix, so you cannot use it if you need to call libraries
13568 compiled with the Unix compiler.
13570 Also, you must provide function prototypes for all functions that
13571 take variable numbers of arguments (including @code{printf});
13572 otherwise incorrect code will be generated for calls to those
13575 In addition, seriously incorrect code will result if you call a
13576 function with too many arguments. (Normally, extra arguments are
13577 harmlessly ignored.)
13579 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13580 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13584 Do not use the calling conventions selected by @option{-mrtd}.
13585 This is the default.
13588 @itemx -mno-align-int
13589 @opindex malign-int
13590 @opindex mno-align-int
13591 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13592 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13593 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13594 Aligning variables on 32-bit boundaries produces code that runs somewhat
13595 faster on processors with 32-bit busses at the expense of more memory.
13597 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13598 align structures containing the above types differently than
13599 most published application binary interface specifications for the m68k.
13603 Use the pc-relative addressing mode of the 68000 directly, instead of
13604 using a global offset table. At present, this option implies @option{-fpic},
13605 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13606 not presently supported with @option{-mpcrel}, though this could be supported for
13607 68020 and higher processors.
13609 @item -mno-strict-align
13610 @itemx -mstrict-align
13611 @opindex mno-strict-align
13612 @opindex mstrict-align
13613 Do not (do) assume that unaligned memory references will be handled by
13617 Generate code that allows the data segment to be located in a different
13618 area of memory from the text segment. This allows for execute in place in
13619 an environment without virtual memory management. This option implies
13622 @item -mno-sep-data
13623 Generate code that assumes that the data segment follows the text segment.
13624 This is the default.
13626 @item -mid-shared-library
13627 Generate code that supports shared libraries via the library ID method.
13628 This allows for execute in place and shared libraries in an environment
13629 without virtual memory management. This option implies @option{-fPIC}.
13631 @item -mno-id-shared-library
13632 Generate code that doesn't assume ID based shared libraries are being used.
13633 This is the default.
13635 @item -mshared-library-id=n
13636 Specified the identification number of the ID based shared library being
13637 compiled. Specifying a value of 0 will generate more compact code, specifying
13638 other values will force the allocation of that number to the current
13639 library but is no more space or time efficient than omitting this option.
13645 When generating position-independent code for ColdFire, generate code
13646 that works if the GOT has more than 8192 entries. This code is
13647 larger and slower than code generated without this option. On M680x0
13648 processors, this option is not needed; @option{-fPIC} suffices.
13650 GCC normally uses a single instruction to load values from the GOT@.
13651 While this is relatively efficient, it only works if the GOT
13652 is smaller than about 64k. Anything larger causes the linker
13653 to report an error such as:
13655 @cindex relocation truncated to fit (ColdFire)
13657 relocation truncated to fit: R_68K_GOT16O foobar
13660 If this happens, you should recompile your code with @option{-mxgot}.
13661 It should then work with very large GOTs. However, code generated with
13662 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13663 the value of a global symbol.
13665 Note that some linkers, including newer versions of the GNU linker,
13666 can create multiple GOTs and sort GOT entries. If you have such a linker,
13667 you should only need to use @option{-mxgot} when compiling a single
13668 object file that accesses more than 8192 GOT entries. Very few do.
13670 These options have no effect unless GCC is generating
13671 position-independent code.
13675 @node MCore Options
13676 @subsection MCore Options
13677 @cindex MCore options
13679 These are the @samp{-m} options defined for the Motorola M*Core
13685 @itemx -mno-hardlit
13687 @opindex mno-hardlit
13688 Inline constants into the code stream if it can be done in two
13689 instructions or less.
13695 Use the divide instruction. (Enabled by default).
13697 @item -mrelax-immediate
13698 @itemx -mno-relax-immediate
13699 @opindex mrelax-immediate
13700 @opindex mno-relax-immediate
13701 Allow arbitrary sized immediates in bit operations.
13703 @item -mwide-bitfields
13704 @itemx -mno-wide-bitfields
13705 @opindex mwide-bitfields
13706 @opindex mno-wide-bitfields
13707 Always treat bit-fields as int-sized.
13709 @item -m4byte-functions
13710 @itemx -mno-4byte-functions
13711 @opindex m4byte-functions
13712 @opindex mno-4byte-functions
13713 Force all functions to be aligned to a four byte boundary.
13715 @item -mcallgraph-data
13716 @itemx -mno-callgraph-data
13717 @opindex mcallgraph-data
13718 @opindex mno-callgraph-data
13719 Emit callgraph information.
13722 @itemx -mno-slow-bytes
13723 @opindex mslow-bytes
13724 @opindex mno-slow-bytes
13725 Prefer word access when reading byte quantities.
13727 @item -mlittle-endian
13728 @itemx -mbig-endian
13729 @opindex mlittle-endian
13730 @opindex mbig-endian
13731 Generate code for a little endian target.
13737 Generate code for the 210 processor.
13741 Assume that run-time support has been provided and so omit the
13742 simulator library (@file{libsim.a)} from the linker command line.
13744 @item -mstack-increment=@var{size}
13745 @opindex mstack-increment
13746 Set the maximum amount for a single stack increment operation. Large
13747 values can increase the speed of programs which contain functions
13748 that need a large amount of stack space, but they can also trigger a
13749 segmentation fault if the stack is extended too much. The default
13755 @subsection MeP Options
13756 @cindex MeP options
13762 Enables the @code{abs} instruction, which is the absolute difference
13763 between two registers.
13767 Enables all the optional instructions - average, multiply, divide, bit
13768 operations, leading zero, absolute difference, min/max, clip, and
13774 Enables the @code{ave} instruction, which computes the average of two
13777 @item -mbased=@var{n}
13779 Variables of size @var{n} bytes or smaller will be placed in the
13780 @code{.based} section by default. Based variables use the @code{$tp}
13781 register as a base register, and there is a 128 byte limit to the
13782 @code{.based} section.
13786 Enables the bit operation instructions - bit test (@code{btstm}), set
13787 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13788 test-and-set (@code{tas}).
13790 @item -mc=@var{name}
13792 Selects which section constant data will be placed in. @var{name} may
13793 be @code{tiny}, @code{near}, or @code{far}.
13797 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13798 useful unless you also provide @code{-mminmax}.
13800 @item -mconfig=@var{name}
13802 Selects one of the build-in core configurations. Each MeP chip has
13803 one or more modules in it; each module has a core CPU and a variety of
13804 coprocessors, optional instructions, and peripherals. The
13805 @code{MeP-Integrator} tool, not part of GCC, provides these
13806 configurations through this option; using this option is the same as
13807 using all the corresponding command line options. The default
13808 configuration is @code{default}.
13812 Enables the coprocessor instructions. By default, this is a 32-bit
13813 coprocessor. Note that the coprocessor is normally enabled via the
13814 @code{-mconfig=} option.
13818 Enables the 32-bit coprocessor's instructions.
13822 Enables the 64-bit coprocessor's instructions.
13826 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13830 Causes constant variables to be placed in the @code{.near} section.
13834 Enables the @code{div} and @code{divu} instructions.
13838 Generate big-endian code.
13842 Generate little-endian code.
13844 @item -mio-volatile
13845 @opindex mio-volatile
13846 Tells the compiler that any variable marked with the @code{io}
13847 attribute is to be considered volatile.
13851 Causes variables to be assigned to the @code{.far} section by default.
13855 Enables the @code{leadz} (leading zero) instruction.
13859 Causes variables to be assigned to the @code{.near} section by default.
13863 Enables the @code{min} and @code{max} instructions.
13867 Enables the multiplication and multiply-accumulate instructions.
13871 Disables all the optional instructions enabled by @code{-mall-opts}.
13875 Enables the @code{repeat} and @code{erepeat} instructions, used for
13876 low-overhead looping.
13880 Causes all variables to default to the @code{.tiny} section. Note
13881 that there is a 65536 byte limit to this section. Accesses to these
13882 variables use the @code{%gp} base register.
13886 Enables the saturation instructions. Note that the compiler does not
13887 currently generate these itself, but this option is included for
13888 compatibility with other tools, like @code{as}.
13892 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13896 Link the simulator runtime libraries.
13900 Link the simulator runtime libraries, excluding built-in support
13901 for reset and exception vectors and tables.
13905 Causes all functions to default to the @code{.far} section. Without
13906 this option, functions default to the @code{.near} section.
13908 @item -mtiny=@var{n}
13910 Variables that are @var{n} bytes or smaller will be allocated to the
13911 @code{.tiny} section. These variables use the @code{$gp} base
13912 register. The default for this option is 4, but note that there's a
13913 65536 byte limit to the @code{.tiny} section.
13917 @node MicroBlaze Options
13918 @subsection MicroBlaze Options
13919 @cindex MicroBlaze Options
13924 @opindex msoft-float
13925 Use software emulation for floating point (default).
13928 @opindex mhard-float
13929 Use hardware floating point instructions.
13933 Do not optimize block moves, use @code{memcpy}.
13935 @item -mno-clearbss
13936 @opindex mno-clearbss
13937 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
13939 @item -mcpu=@var{cpu-type}
13941 Use features of and schedule code for given CPU.
13942 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
13943 where @var{X} is a major version, @var{YY} is the minor version, and
13944 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
13945 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
13947 @item -mxl-soft-mul
13948 @opindex mxl-soft-mul
13949 Use software multiply emulation (default).
13951 @item -mxl-soft-div
13952 @opindex mxl-soft-div
13953 Use software emulation for divides (default).
13955 @item -mxl-barrel-shift
13956 @opindex mxl-barrel-shift
13957 Use the hardware barrel shifter.
13959 @item -mxl-pattern-compare
13960 @opindex mxl-pattern-compare
13961 Use pattern compare instructions.
13963 @item -msmall-divides
13964 @opindex msmall-divides
13965 Use table lookup optimization for small signed integer divisions.
13967 @item -mxl-stack-check
13968 @opindex mxl-stack-check
13969 This option is deprecated. Use -fstack-check instead.
13972 @opindex mxl-gp-opt
13973 Use GP relative sdata/sbss sections.
13975 @item -mxl-multiply-high
13976 @opindex mxl-multiply-high
13977 Use multiply high instructions for high part of 32x32 multiply.
13979 @item -mxl-float-convert
13980 @opindex mxl-float-convert
13981 Use hardware floating point conversion instructions.
13983 @item -mxl-float-sqrt
13984 @opindex mxl-float-sqrt
13985 Use hardware floating point square root instruction.
13987 @item -mxl-mode-@var{app-model}
13988 Select application model @var{app-model}. Valid models are
13991 normal executable (default), uses startup code @file{crt0.o}.
13994 for use with Xilinx Microprocessor Debugger (XMD) based
13995 software intrusive debug agent called xmdstub. This uses startup file
13996 @file{crt1.o} and sets the start address of the program to be 0x800.
13999 for applications that are loaded using a bootloader.
14000 This model uses startup file @file{crt2.o} which does not contain a processor
14001 reset vector handler. This is suitable for transferring control on a
14002 processor reset to the bootloader rather than the application.
14005 for applications that do not require any of the
14006 MicroBlaze vectors. This option may be useful for applications running
14007 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14010 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14011 @option{-mxl-mode-@var{app-model}}.
14016 @subsection MIPS Options
14017 @cindex MIPS options
14023 Generate big-endian code.
14027 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14030 @item -march=@var{arch}
14032 Generate code that will run on @var{arch}, which can be the name of a
14033 generic MIPS ISA, or the name of a particular processor.
14035 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14036 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14037 The processor names are:
14038 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14039 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14040 @samp{5kc}, @samp{5kf},
14042 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14043 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14044 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14045 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14046 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14047 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14051 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14052 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14053 @samp{rm7000}, @samp{rm9000},
14054 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14057 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14058 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14060 The special value @samp{from-abi} selects the
14061 most compatible architecture for the selected ABI (that is,
14062 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14064 Native Linux/GNU toolchains also support the value @samp{native},
14065 which selects the best architecture option for the host processor.
14066 @option{-march=native} has no effect if GCC does not recognize
14069 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14070 (for example, @samp{-march=r2k}). Prefixes are optional, and
14071 @samp{vr} may be written @samp{r}.
14073 Names of the form @samp{@var{n}f2_1} refer to processors with
14074 FPUs clocked at half the rate of the core, names of the form
14075 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14076 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14077 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14078 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14079 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14080 accepted as synonyms for @samp{@var{n}f1_1}.
14082 GCC defines two macros based on the value of this option. The first
14083 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14084 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14085 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14086 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14087 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14089 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14090 above. In other words, it will have the full prefix and will not
14091 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14092 the macro names the resolved architecture (either @samp{"mips1"} or
14093 @samp{"mips3"}). It names the default architecture when no
14094 @option{-march} option is given.
14096 @item -mtune=@var{arch}
14098 Optimize for @var{arch}. Among other things, this option controls
14099 the way instructions are scheduled, and the perceived cost of arithmetic
14100 operations. The list of @var{arch} values is the same as for
14103 When this option is not used, GCC will optimize for the processor
14104 specified by @option{-march}. By using @option{-march} and
14105 @option{-mtune} together, it is possible to generate code that will
14106 run on a family of processors, but optimize the code for one
14107 particular member of that family.
14109 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14110 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14111 @samp{-march} ones described above.
14115 Equivalent to @samp{-march=mips1}.
14119 Equivalent to @samp{-march=mips2}.
14123 Equivalent to @samp{-march=mips3}.
14127 Equivalent to @samp{-march=mips4}.
14131 Equivalent to @samp{-march=mips32}.
14135 Equivalent to @samp{-march=mips32r2}.
14139 Equivalent to @samp{-march=mips64}.
14143 Equivalent to @samp{-march=mips64r2}.
14148 @opindex mno-mips16
14149 Generate (do not generate) MIPS16 code. If GCC is targetting a
14150 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14152 MIPS16 code generation can also be controlled on a per-function basis
14153 by means of @code{mips16} and @code{nomips16} attributes.
14154 @xref{Function Attributes}, for more information.
14156 @item -mflip-mips16
14157 @opindex mflip-mips16
14158 Generate MIPS16 code on alternating functions. This option is provided
14159 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14160 not intended for ordinary use in compiling user code.
14162 @item -minterlink-mips16
14163 @itemx -mno-interlink-mips16
14164 @opindex minterlink-mips16
14165 @opindex mno-interlink-mips16
14166 Require (do not require) that non-MIPS16 code be link-compatible with
14169 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14170 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14171 therefore disables direct jumps unless GCC knows that the target of the
14172 jump is not MIPS16.
14184 Generate code for the given ABI@.
14186 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14187 generates 64-bit code when you select a 64-bit architecture, but you
14188 can use @option{-mgp32} to get 32-bit code instead.
14190 For information about the O64 ABI, see
14191 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14193 GCC supports a variant of the o32 ABI in which floating-point registers
14194 are 64 rather than 32 bits wide. You can select this combination with
14195 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14196 and @samp{mfhc1} instructions and is therefore only supported for
14197 MIPS32R2 processors.
14199 The register assignments for arguments and return values remain the
14200 same, but each scalar value is passed in a single 64-bit register
14201 rather than a pair of 32-bit registers. For example, scalar
14202 floating-point values are returned in @samp{$f0} only, not a
14203 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14204 remains the same, but all 64 bits are saved.
14207 @itemx -mno-abicalls
14209 @opindex mno-abicalls
14210 Generate (do not generate) code that is suitable for SVR4-style
14211 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14216 Generate (do not generate) code that is fully position-independent,
14217 and that can therefore be linked into shared libraries. This option
14218 only affects @option{-mabicalls}.
14220 All @option{-mabicalls} code has traditionally been position-independent,
14221 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14222 as an extension, the GNU toolchain allows executables to use absolute
14223 accesses for locally-binding symbols. It can also use shorter GP
14224 initialization sequences and generate direct calls to locally-defined
14225 functions. This mode is selected by @option{-mno-shared}.
14227 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14228 objects that can only be linked by the GNU linker. However, the option
14229 does not affect the ABI of the final executable; it only affects the ABI
14230 of relocatable objects. Using @option{-mno-shared} will generally make
14231 executables both smaller and quicker.
14233 @option{-mshared} is the default.
14239 Assume (do not assume) that the static and dynamic linkers
14240 support PLTs and copy relocations. This option only affects
14241 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14242 has no effect without @samp{-msym32}.
14244 You can make @option{-mplt} the default by configuring
14245 GCC with @option{--with-mips-plt}. The default is
14246 @option{-mno-plt} otherwise.
14252 Lift (do not lift) the usual restrictions on the size of the global
14255 GCC normally uses a single instruction to load values from the GOT@.
14256 While this is relatively efficient, it will only work if the GOT
14257 is smaller than about 64k. Anything larger will cause the linker
14258 to report an error such as:
14260 @cindex relocation truncated to fit (MIPS)
14262 relocation truncated to fit: R_MIPS_GOT16 foobar
14265 If this happens, you should recompile your code with @option{-mxgot}.
14266 It should then work with very large GOTs, although it will also be
14267 less efficient, since it will take three instructions to fetch the
14268 value of a global symbol.
14270 Note that some linkers can create multiple GOTs. If you have such a
14271 linker, you should only need to use @option{-mxgot} when a single object
14272 file accesses more than 64k's worth of GOT entries. Very few do.
14274 These options have no effect unless GCC is generating position
14279 Assume that general-purpose registers are 32 bits wide.
14283 Assume that general-purpose registers are 64 bits wide.
14287 Assume that floating-point registers are 32 bits wide.
14291 Assume that floating-point registers are 64 bits wide.
14294 @opindex mhard-float
14295 Use floating-point coprocessor instructions.
14298 @opindex msoft-float
14299 Do not use floating-point coprocessor instructions. Implement
14300 floating-point calculations using library calls instead.
14302 @item -msingle-float
14303 @opindex msingle-float
14304 Assume that the floating-point coprocessor only supports single-precision
14307 @item -mdouble-float
14308 @opindex mdouble-float
14309 Assume that the floating-point coprocessor supports double-precision
14310 operations. This is the default.
14316 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14317 implement atomic memory built-in functions. When neither option is
14318 specified, GCC will use the instructions if the target architecture
14321 @option{-mllsc} is useful if the runtime environment can emulate the
14322 instructions and @option{-mno-llsc} can be useful when compiling for
14323 nonstandard ISAs. You can make either option the default by
14324 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14325 respectively. @option{--with-llsc} is the default for some
14326 configurations; see the installation documentation for details.
14332 Use (do not use) revision 1 of the MIPS DSP ASE@.
14333 @xref{MIPS DSP Built-in Functions}. This option defines the
14334 preprocessor macro @samp{__mips_dsp}. It also defines
14335 @samp{__mips_dsp_rev} to 1.
14341 Use (do not use) revision 2 of the MIPS DSP ASE@.
14342 @xref{MIPS DSP Built-in Functions}. This option defines the
14343 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14344 It also defines @samp{__mips_dsp_rev} to 2.
14347 @itemx -mno-smartmips
14348 @opindex msmartmips
14349 @opindex mno-smartmips
14350 Use (do not use) the MIPS SmartMIPS ASE.
14352 @item -mpaired-single
14353 @itemx -mno-paired-single
14354 @opindex mpaired-single
14355 @opindex mno-paired-single
14356 Use (do not use) paired-single floating-point instructions.
14357 @xref{MIPS Paired-Single Support}. This option requires
14358 hardware floating-point support to be enabled.
14364 Use (do not use) MIPS Digital Media Extension instructions.
14365 This option can only be used when generating 64-bit code and requires
14366 hardware floating-point support to be enabled.
14371 @opindex mno-mips3d
14372 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14373 The option @option{-mips3d} implies @option{-mpaired-single}.
14379 Use (do not use) MT Multithreading instructions.
14383 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14384 an explanation of the default and the way that the pointer size is
14389 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14391 The default size of @code{int}s, @code{long}s and pointers depends on
14392 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14393 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14394 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14395 or the same size as integer registers, whichever is smaller.
14401 Assume (do not assume) that all symbols have 32-bit values, regardless
14402 of the selected ABI@. This option is useful in combination with
14403 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14404 to generate shorter and faster references to symbolic addresses.
14408 Put definitions of externally-visible data in a small data section
14409 if that data is no bigger than @var{num} bytes. GCC can then access
14410 the data more efficiently; see @option{-mgpopt} for details.
14412 The default @option{-G} option depends on the configuration.
14414 @item -mlocal-sdata
14415 @itemx -mno-local-sdata
14416 @opindex mlocal-sdata
14417 @opindex mno-local-sdata
14418 Extend (do not extend) the @option{-G} behavior to local data too,
14419 such as to static variables in C@. @option{-mlocal-sdata} is the
14420 default for all configurations.
14422 If the linker complains that an application is using too much small data,
14423 you might want to try rebuilding the less performance-critical parts with
14424 @option{-mno-local-sdata}. You might also want to build large
14425 libraries with @option{-mno-local-sdata}, so that the libraries leave
14426 more room for the main program.
14428 @item -mextern-sdata
14429 @itemx -mno-extern-sdata
14430 @opindex mextern-sdata
14431 @opindex mno-extern-sdata
14432 Assume (do not assume) that externally-defined data will be in
14433 a small data section if that data is within the @option{-G} limit.
14434 @option{-mextern-sdata} is the default for all configurations.
14436 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14437 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14438 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14439 is placed in a small data section. If @var{Var} is defined by another
14440 module, you must either compile that module with a high-enough
14441 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14442 definition. If @var{Var} is common, you must link the application
14443 with a high-enough @option{-G} setting.
14445 The easiest way of satisfying these restrictions is to compile
14446 and link every module with the same @option{-G} option. However,
14447 you may wish to build a library that supports several different
14448 small data limits. You can do this by compiling the library with
14449 the highest supported @option{-G} setting and additionally using
14450 @option{-mno-extern-sdata} to stop the library from making assumptions
14451 about externally-defined data.
14457 Use (do not use) GP-relative accesses for symbols that are known to be
14458 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14459 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14462 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14463 might not hold the value of @code{_gp}. For example, if the code is
14464 part of a library that might be used in a boot monitor, programs that
14465 call boot monitor routines will pass an unknown value in @code{$gp}.
14466 (In such situations, the boot monitor itself would usually be compiled
14467 with @option{-G0}.)
14469 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14470 @option{-mno-extern-sdata}.
14472 @item -membedded-data
14473 @itemx -mno-embedded-data
14474 @opindex membedded-data
14475 @opindex mno-embedded-data
14476 Allocate variables to the read-only data section first if possible, then
14477 next in the small data section if possible, otherwise in data. This gives
14478 slightly slower code than the default, but reduces the amount of RAM required
14479 when executing, and thus may be preferred for some embedded systems.
14481 @item -muninit-const-in-rodata
14482 @itemx -mno-uninit-const-in-rodata
14483 @opindex muninit-const-in-rodata
14484 @opindex mno-uninit-const-in-rodata
14485 Put uninitialized @code{const} variables in the read-only data section.
14486 This option is only meaningful in conjunction with @option{-membedded-data}.
14488 @item -mcode-readable=@var{setting}
14489 @opindex mcode-readable
14490 Specify whether GCC may generate code that reads from executable sections.
14491 There are three possible settings:
14494 @item -mcode-readable=yes
14495 Instructions may freely access executable sections. This is the
14498 @item -mcode-readable=pcrel
14499 MIPS16 PC-relative load instructions can access executable sections,
14500 but other instructions must not do so. This option is useful on 4KSc
14501 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14502 It is also useful on processors that can be configured to have a dual
14503 instruction/data SRAM interface and that, like the M4K, automatically
14504 redirect PC-relative loads to the instruction RAM.
14506 @item -mcode-readable=no
14507 Instructions must not access executable sections. This option can be
14508 useful on targets that are configured to have a dual instruction/data
14509 SRAM interface but that (unlike the M4K) do not automatically redirect
14510 PC-relative loads to the instruction RAM.
14513 @item -msplit-addresses
14514 @itemx -mno-split-addresses
14515 @opindex msplit-addresses
14516 @opindex mno-split-addresses
14517 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14518 relocation operators. This option has been superseded by
14519 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14521 @item -mexplicit-relocs
14522 @itemx -mno-explicit-relocs
14523 @opindex mexplicit-relocs
14524 @opindex mno-explicit-relocs
14525 Use (do not use) assembler relocation operators when dealing with symbolic
14526 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14527 is to use assembler macros instead.
14529 @option{-mexplicit-relocs} is the default if GCC was configured
14530 to use an assembler that supports relocation operators.
14532 @item -mcheck-zero-division
14533 @itemx -mno-check-zero-division
14534 @opindex mcheck-zero-division
14535 @opindex mno-check-zero-division
14536 Trap (do not trap) on integer division by zero.
14538 The default is @option{-mcheck-zero-division}.
14540 @item -mdivide-traps
14541 @itemx -mdivide-breaks
14542 @opindex mdivide-traps
14543 @opindex mdivide-breaks
14544 MIPS systems check for division by zero by generating either a
14545 conditional trap or a break instruction. Using traps results in
14546 smaller code, but is only supported on MIPS II and later. Also, some
14547 versions of the Linux kernel have a bug that prevents trap from
14548 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14549 allow conditional traps on architectures that support them and
14550 @option{-mdivide-breaks} to force the use of breaks.
14552 The default is usually @option{-mdivide-traps}, but this can be
14553 overridden at configure time using @option{--with-divide=breaks}.
14554 Divide-by-zero checks can be completely disabled using
14555 @option{-mno-check-zero-division}.
14560 @opindex mno-memcpy
14561 Force (do not force) the use of @code{memcpy()} for non-trivial block
14562 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14563 most constant-sized copies.
14566 @itemx -mno-long-calls
14567 @opindex mlong-calls
14568 @opindex mno-long-calls
14569 Disable (do not disable) use of the @code{jal} instruction. Calling
14570 functions using @code{jal} is more efficient but requires the caller
14571 and callee to be in the same 256 megabyte segment.
14573 This option has no effect on abicalls code. The default is
14574 @option{-mno-long-calls}.
14580 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14581 instructions, as provided by the R4650 ISA@.
14584 @itemx -mno-fused-madd
14585 @opindex mfused-madd
14586 @opindex mno-fused-madd
14587 Enable (disable) use of the floating point multiply-accumulate
14588 instructions, when they are available. The default is
14589 @option{-mfused-madd}.
14591 When multiply-accumulate instructions are used, the intermediate
14592 product is calculated to infinite precision and is not subject to
14593 the FCSR Flush to Zero bit. This may be undesirable in some
14598 Tell the MIPS assembler to not run its preprocessor over user
14599 assembler files (with a @samp{.s} suffix) when assembling them.
14602 @itemx -mno-fix-r4000
14603 @opindex mfix-r4000
14604 @opindex mno-fix-r4000
14605 Work around certain R4000 CPU errata:
14608 A double-word or a variable shift may give an incorrect result if executed
14609 immediately after starting an integer division.
14611 A double-word or a variable shift may give an incorrect result if executed
14612 while an integer multiplication is in progress.
14614 An integer division may give an incorrect result if started in a delay slot
14615 of a taken branch or a jump.
14619 @itemx -mno-fix-r4400
14620 @opindex mfix-r4400
14621 @opindex mno-fix-r4400
14622 Work around certain R4400 CPU errata:
14625 A double-word or a variable shift may give an incorrect result if executed
14626 immediately after starting an integer division.
14630 @itemx -mno-fix-r10000
14631 @opindex mfix-r10000
14632 @opindex mno-fix-r10000
14633 Work around certain R10000 errata:
14636 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14637 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14640 This option can only be used if the target architecture supports
14641 branch-likely instructions. @option{-mfix-r10000} is the default when
14642 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14646 @itemx -mno-fix-vr4120
14647 @opindex mfix-vr4120
14648 Work around certain VR4120 errata:
14651 @code{dmultu} does not always produce the correct result.
14653 @code{div} and @code{ddiv} do not always produce the correct result if one
14654 of the operands is negative.
14656 The workarounds for the division errata rely on special functions in
14657 @file{libgcc.a}. At present, these functions are only provided by
14658 the @code{mips64vr*-elf} configurations.
14660 Other VR4120 errata require a nop to be inserted between certain pairs of
14661 instructions. These errata are handled by the assembler, not by GCC itself.
14664 @opindex mfix-vr4130
14665 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14666 workarounds are implemented by the assembler rather than by GCC,
14667 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14668 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14669 instructions are available instead.
14672 @itemx -mno-fix-sb1
14674 Work around certain SB-1 CPU core errata.
14675 (This flag currently works around the SB-1 revision 2
14676 ``F1'' and ``F2'' floating point errata.)
14678 @item -mr10k-cache-barrier=@var{setting}
14679 @opindex mr10k-cache-barrier
14680 Specify whether GCC should insert cache barriers to avoid the
14681 side-effects of speculation on R10K processors.
14683 In common with many processors, the R10K tries to predict the outcome
14684 of a conditional branch and speculatively executes instructions from
14685 the ``taken'' branch. It later aborts these instructions if the
14686 predicted outcome was wrong. However, on the R10K, even aborted
14687 instructions can have side effects.
14689 This problem only affects kernel stores and, depending on the system,
14690 kernel loads. As an example, a speculatively-executed store may load
14691 the target memory into cache and mark the cache line as dirty, even if
14692 the store itself is later aborted. If a DMA operation writes to the
14693 same area of memory before the ``dirty'' line is flushed, the cached
14694 data will overwrite the DMA-ed data. See the R10K processor manual
14695 for a full description, including other potential problems.
14697 One workaround is to insert cache barrier instructions before every memory
14698 access that might be speculatively executed and that might have side
14699 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14700 controls GCC's implementation of this workaround. It assumes that
14701 aborted accesses to any byte in the following regions will not have
14706 the memory occupied by the current function's stack frame;
14709 the memory occupied by an incoming stack argument;
14712 the memory occupied by an object with a link-time-constant address.
14715 It is the kernel's responsibility to ensure that speculative
14716 accesses to these regions are indeed safe.
14718 If the input program contains a function declaration such as:
14724 then the implementation of @code{foo} must allow @code{j foo} and
14725 @code{jal foo} to be executed speculatively. GCC honors this
14726 restriction for functions it compiles itself. It expects non-GCC
14727 functions (such as hand-written assembly code) to do the same.
14729 The option has three forms:
14732 @item -mr10k-cache-barrier=load-store
14733 Insert a cache barrier before a load or store that might be
14734 speculatively executed and that might have side effects even
14737 @item -mr10k-cache-barrier=store
14738 Insert a cache barrier before a store that might be speculatively
14739 executed and that might have side effects even if aborted.
14741 @item -mr10k-cache-barrier=none
14742 Disable the insertion of cache barriers. This is the default setting.
14745 @item -mflush-func=@var{func}
14746 @itemx -mno-flush-func
14747 @opindex mflush-func
14748 Specifies the function to call to flush the I and D caches, or to not
14749 call any such function. If called, the function must take the same
14750 arguments as the common @code{_flush_func()}, that is, the address of the
14751 memory range for which the cache is being flushed, the size of the
14752 memory range, and the number 3 (to flush both caches). The default
14753 depends on the target GCC was configured for, but commonly is either
14754 @samp{_flush_func} or @samp{__cpu_flush}.
14756 @item mbranch-cost=@var{num}
14757 @opindex mbranch-cost
14758 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14759 This cost is only a heuristic and is not guaranteed to produce
14760 consistent results across releases. A zero cost redundantly selects
14761 the default, which is based on the @option{-mtune} setting.
14763 @item -mbranch-likely
14764 @itemx -mno-branch-likely
14765 @opindex mbranch-likely
14766 @opindex mno-branch-likely
14767 Enable or disable use of Branch Likely instructions, regardless of the
14768 default for the selected architecture. By default, Branch Likely
14769 instructions may be generated if they are supported by the selected
14770 architecture. An exception is for the MIPS32 and MIPS64 architectures
14771 and processors which implement those architectures; for those, Branch
14772 Likely instructions will not be generated by default because the MIPS32
14773 and MIPS64 architectures specifically deprecate their use.
14775 @item -mfp-exceptions
14776 @itemx -mno-fp-exceptions
14777 @opindex mfp-exceptions
14778 Specifies whether FP exceptions are enabled. This affects how we schedule
14779 FP instructions for some processors. The default is that FP exceptions are
14782 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14783 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14786 @item -mvr4130-align
14787 @itemx -mno-vr4130-align
14788 @opindex mvr4130-align
14789 The VR4130 pipeline is two-way superscalar, but can only issue two
14790 instructions together if the first one is 8-byte aligned. When this
14791 option is enabled, GCC will align pairs of instructions that it
14792 thinks should execute in parallel.
14794 This option only has an effect when optimizing for the VR4130.
14795 It normally makes code faster, but at the expense of making it bigger.
14796 It is enabled by default at optimization level @option{-O3}.
14801 Enable (disable) generation of @code{synci} instructions on
14802 architectures that support it. The @code{synci} instructions (if
14803 enabled) will be generated when @code{__builtin___clear_cache()} is
14806 This option defaults to @code{-mno-synci}, but the default can be
14807 overridden by configuring with @code{--with-synci}.
14809 When compiling code for single processor systems, it is generally safe
14810 to use @code{synci}. However, on many multi-core (SMP) systems, it
14811 will not invalidate the instruction caches on all cores and may lead
14812 to undefined behavior.
14814 @item -mrelax-pic-calls
14815 @itemx -mno-relax-pic-calls
14816 @opindex mrelax-pic-calls
14817 Try to turn PIC calls that are normally dispatched via register
14818 @code{$25} into direct calls. This is only possible if the linker can
14819 resolve the destination at link-time and if the destination is within
14820 range for a direct call.
14822 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14823 an assembler and a linker that supports the @code{.reloc} assembly
14824 directive and @code{-mexplicit-relocs} is in effect. With
14825 @code{-mno-explicit-relocs}, this optimization can be performed by the
14826 assembler and the linker alone without help from the compiler.
14828 @item -mmcount-ra-address
14829 @itemx -mno-mcount-ra-address
14830 @opindex mmcount-ra-address
14831 @opindex mno-mcount-ra-address
14832 Emit (do not emit) code that allows @code{_mcount} to modify the
14833 calling function's return address. When enabled, this option extends
14834 the usual @code{_mcount} interface with a new @var{ra-address}
14835 parameter, which has type @code{intptr_t *} and is passed in register
14836 @code{$12}. @code{_mcount} can then modify the return address by
14837 doing both of the following:
14840 Returning the new address in register @code{$31}.
14842 Storing the new address in @code{*@var{ra-address}},
14843 if @var{ra-address} is nonnull.
14846 The default is @option{-mno-mcount-ra-address}.
14851 @subsection MMIX Options
14852 @cindex MMIX Options
14854 These options are defined for the MMIX:
14858 @itemx -mno-libfuncs
14860 @opindex mno-libfuncs
14861 Specify that intrinsic library functions are being compiled, passing all
14862 values in registers, no matter the size.
14865 @itemx -mno-epsilon
14867 @opindex mno-epsilon
14868 Generate floating-point comparison instructions that compare with respect
14869 to the @code{rE} epsilon register.
14871 @item -mabi=mmixware
14873 @opindex mabi=mmixware
14875 Generate code that passes function parameters and return values that (in
14876 the called function) are seen as registers @code{$0} and up, as opposed to
14877 the GNU ABI which uses global registers @code{$231} and up.
14879 @item -mzero-extend
14880 @itemx -mno-zero-extend
14881 @opindex mzero-extend
14882 @opindex mno-zero-extend
14883 When reading data from memory in sizes shorter than 64 bits, use (do not
14884 use) zero-extending load instructions by default, rather than
14885 sign-extending ones.
14888 @itemx -mno-knuthdiv
14890 @opindex mno-knuthdiv
14891 Make the result of a division yielding a remainder have the same sign as
14892 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14893 remainder follows the sign of the dividend. Both methods are
14894 arithmetically valid, the latter being almost exclusively used.
14896 @item -mtoplevel-symbols
14897 @itemx -mno-toplevel-symbols
14898 @opindex mtoplevel-symbols
14899 @opindex mno-toplevel-symbols
14900 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14901 code can be used with the @code{PREFIX} assembly directive.
14905 Generate an executable in the ELF format, rather than the default
14906 @samp{mmo} format used by the @command{mmix} simulator.
14908 @item -mbranch-predict
14909 @itemx -mno-branch-predict
14910 @opindex mbranch-predict
14911 @opindex mno-branch-predict
14912 Use (do not use) the probable-branch instructions, when static branch
14913 prediction indicates a probable branch.
14915 @item -mbase-addresses
14916 @itemx -mno-base-addresses
14917 @opindex mbase-addresses
14918 @opindex mno-base-addresses
14919 Generate (do not generate) code that uses @emph{base addresses}. Using a
14920 base address automatically generates a request (handled by the assembler
14921 and the linker) for a constant to be set up in a global register. The
14922 register is used for one or more base address requests within the range 0
14923 to 255 from the value held in the register. The generally leads to short
14924 and fast code, but the number of different data items that can be
14925 addressed is limited. This means that a program that uses lots of static
14926 data may require @option{-mno-base-addresses}.
14928 @item -msingle-exit
14929 @itemx -mno-single-exit
14930 @opindex msingle-exit
14931 @opindex mno-single-exit
14932 Force (do not force) generated code to have a single exit point in each
14936 @node MN10300 Options
14937 @subsection MN10300 Options
14938 @cindex MN10300 options
14940 These @option{-m} options are defined for Matsushita MN10300 architectures:
14945 Generate code to avoid bugs in the multiply instructions for the MN10300
14946 processors. This is the default.
14948 @item -mno-mult-bug
14949 @opindex mno-mult-bug
14950 Do not generate code to avoid bugs in the multiply instructions for the
14951 MN10300 processors.
14955 Generate code which uses features specific to the AM33 processor.
14959 Do not generate code which uses features specific to the AM33 processor. This
14964 Generate code which uses features specific to the AM33/2.0 processor.
14968 Generate code which uses features specific to the AM34 processor.
14970 @item -mtune=@var{cpu-type}
14972 Use the timing characteristics of the indicated CPU type when
14973 scheduling instructions. This does not change the targeted processor
14974 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
14975 @samp{am33-2} or @samp{am34}.
14977 @item -mreturn-pointer-on-d0
14978 @opindex mreturn-pointer-on-d0
14979 When generating a function which returns a pointer, return the pointer
14980 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14981 only in a0, and attempts to call such functions without a prototype
14982 would result in errors. Note that this option is on by default; use
14983 @option{-mno-return-pointer-on-d0} to disable it.
14987 Do not link in the C run-time initialization object file.
14991 Indicate to the linker that it should perform a relaxation optimization pass
14992 to shorten branches, calls and absolute memory addresses. This option only
14993 has an effect when used on the command line for the final link step.
14995 This option makes symbolic debugging impossible.
14999 Allow the compiler to generate @emph{Long Instruction Word}
15000 instructions if the target is the @samp{AM33} or later. This is the
15001 default. This option defines the preprocessor macro @samp{__LIW__}.
15005 Do not allow the compiler to generate @emph{Long Instruction Word}
15006 instructions. This option defines the preprocessor macro
15011 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15012 instructions if the target is the @samp{AM33} or later. This is the
15013 default. This option defines the preprocessor macro @samp{__SETLB__}.
15017 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15018 instructions. This option defines the preprocessor macro
15019 @samp{__NO_SETLB__}.
15023 @node PDP-11 Options
15024 @subsection PDP-11 Options
15025 @cindex PDP-11 Options
15027 These options are defined for the PDP-11:
15032 Use hardware FPP floating point. This is the default. (FIS floating
15033 point on the PDP-11/40 is not supported.)
15036 @opindex msoft-float
15037 Do not use hardware floating point.
15041 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15045 Return floating-point results in memory. This is the default.
15049 Generate code for a PDP-11/40.
15053 Generate code for a PDP-11/45. This is the default.
15057 Generate code for a PDP-11/10.
15059 @item -mbcopy-builtin
15060 @opindex mbcopy-builtin
15061 Use inline @code{movmemhi} patterns for copying memory. This is the
15066 Do not use inline @code{movmemhi} patterns for copying memory.
15072 Use 16-bit @code{int}. This is the default.
15078 Use 32-bit @code{int}.
15081 @itemx -mno-float32
15083 @opindex mno-float32
15084 Use 64-bit @code{float}. This is the default.
15087 @itemx -mno-float64
15089 @opindex mno-float64
15090 Use 32-bit @code{float}.
15094 Use @code{abshi2} pattern. This is the default.
15098 Do not use @code{abshi2} pattern.
15100 @item -mbranch-expensive
15101 @opindex mbranch-expensive
15102 Pretend that branches are expensive. This is for experimenting with
15103 code generation only.
15105 @item -mbranch-cheap
15106 @opindex mbranch-cheap
15107 Do not pretend that branches are expensive. This is the default.
15111 Use Unix assembler syntax. This is the default when configured for
15112 @samp{pdp11-*-bsd}.
15116 Use DEC assembler syntax. This is the default when configured for any
15117 PDP-11 target other than @samp{pdp11-*-bsd}.
15120 @node picoChip Options
15121 @subsection picoChip Options
15122 @cindex picoChip options
15124 These @samp{-m} options are defined for picoChip implementations:
15128 @item -mae=@var{ae_type}
15130 Set the instruction set, register set, and instruction scheduling
15131 parameters for array element type @var{ae_type}. Supported values
15132 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15134 @option{-mae=ANY} selects a completely generic AE type. Code
15135 generated with this option will run on any of the other AE types. The
15136 code will not be as efficient as it would be if compiled for a specific
15137 AE type, and some types of operation (e.g., multiplication) will not
15138 work properly on all types of AE.
15140 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15141 for compiled code, and is the default.
15143 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15144 option may suffer from poor performance of byte (char) manipulation,
15145 since the DSP AE does not provide hardware support for byte load/stores.
15147 @item -msymbol-as-address
15148 Enable the compiler to directly use a symbol name as an address in a
15149 load/store instruction, without first loading it into a
15150 register. Typically, the use of this option will generate larger
15151 programs, which run faster than when the option isn't used. However, the
15152 results vary from program to program, so it is left as a user option,
15153 rather than being permanently enabled.
15155 @item -mno-inefficient-warnings
15156 Disables warnings about the generation of inefficient code. These
15157 warnings can be generated, for example, when compiling code which
15158 performs byte-level memory operations on the MAC AE type. The MAC AE has
15159 no hardware support for byte-level memory operations, so all byte
15160 load/stores must be synthesized from word load/store operations. This is
15161 inefficient and a warning will be generated indicating to the programmer
15162 that they should rewrite the code to avoid byte operations, or to target
15163 an AE type which has the necessary hardware support. This option enables
15164 the warning to be turned off.
15168 @node PowerPC Options
15169 @subsection PowerPC Options
15170 @cindex PowerPC options
15172 These are listed under @xref{RS/6000 and PowerPC Options}.
15174 @node RS/6000 and PowerPC Options
15175 @subsection IBM RS/6000 and PowerPC Options
15176 @cindex RS/6000 and PowerPC Options
15177 @cindex IBM RS/6000 and PowerPC Options
15179 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15186 @itemx -mno-powerpc
15187 @itemx -mpowerpc-gpopt
15188 @itemx -mno-powerpc-gpopt
15189 @itemx -mpowerpc-gfxopt
15190 @itemx -mno-powerpc-gfxopt
15193 @itemx -mno-powerpc64
15197 @itemx -mno-popcntb
15199 @itemx -mno-popcntd
15208 @itemx -mno-hard-dfp
15212 @opindex mno-power2
15214 @opindex mno-powerpc
15215 @opindex mpowerpc-gpopt
15216 @opindex mno-powerpc-gpopt
15217 @opindex mpowerpc-gfxopt
15218 @opindex mno-powerpc-gfxopt
15219 @opindex mpowerpc64
15220 @opindex mno-powerpc64
15224 @opindex mno-popcntb
15226 @opindex mno-popcntd
15232 @opindex mno-mfpgpr
15234 @opindex mno-hard-dfp
15235 GCC supports two related instruction set architectures for the
15236 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15237 instructions supported by the @samp{rios} chip set used in the original
15238 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15239 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15240 the IBM 4xx, 6xx, and follow-on microprocessors.
15242 Neither architecture is a subset of the other. However there is a
15243 large common subset of instructions supported by both. An MQ
15244 register is included in processors supporting the POWER architecture.
15246 You use these options to specify which instructions are available on the
15247 processor you are using. The default value of these options is
15248 determined when configuring GCC@. Specifying the
15249 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15250 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15251 rather than the options listed above.
15253 The @option{-mpower} option allows GCC to generate instructions that
15254 are found only in the POWER architecture and to use the MQ register.
15255 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15256 to generate instructions that are present in the POWER2 architecture but
15257 not the original POWER architecture.
15259 The @option{-mpowerpc} option allows GCC to generate instructions that
15260 are found only in the 32-bit subset of the PowerPC architecture.
15261 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15262 GCC to use the optional PowerPC architecture instructions in the
15263 General Purpose group, including floating-point square root. Specifying
15264 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15265 use the optional PowerPC architecture instructions in the Graphics
15266 group, including floating-point select.
15268 The @option{-mmfcrf} option allows GCC to generate the move from
15269 condition register field instruction implemented on the POWER4
15270 processor and other processors that support the PowerPC V2.01
15272 The @option{-mpopcntb} option allows GCC to generate the popcount and
15273 double precision FP reciprocal estimate instruction implemented on the
15274 POWER5 processor and other processors that support the PowerPC V2.02
15276 The @option{-mpopcntd} option allows GCC to generate the popcount
15277 instruction implemented on the POWER7 processor and other processors
15278 that support the PowerPC V2.06 architecture.
15279 The @option{-mfprnd} option allows GCC to generate the FP round to
15280 integer instructions implemented on the POWER5+ processor and other
15281 processors that support the PowerPC V2.03 architecture.
15282 The @option{-mcmpb} option allows GCC to generate the compare bytes
15283 instruction implemented on the POWER6 processor and other processors
15284 that support the PowerPC V2.05 architecture.
15285 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15286 general purpose register instructions implemented on the POWER6X
15287 processor and other processors that support the extended PowerPC V2.05
15289 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15290 point instructions implemented on some POWER processors.
15292 The @option{-mpowerpc64} option allows GCC to generate the additional
15293 64-bit instructions that are found in the full PowerPC64 architecture
15294 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15295 @option{-mno-powerpc64}.
15297 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15298 will use only the instructions in the common subset of both
15299 architectures plus some special AIX common-mode calls, and will not use
15300 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15301 permits GCC to use any instruction from either architecture and to
15302 allow use of the MQ register; specify this for the Motorola MPC601.
15304 @item -mnew-mnemonics
15305 @itemx -mold-mnemonics
15306 @opindex mnew-mnemonics
15307 @opindex mold-mnemonics
15308 Select which mnemonics to use in the generated assembler code. With
15309 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15310 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15311 assembler mnemonics defined for the POWER architecture. Instructions
15312 defined in only one architecture have only one mnemonic; GCC uses that
15313 mnemonic irrespective of which of these options is specified.
15315 GCC defaults to the mnemonics appropriate for the architecture in
15316 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15317 value of these option. Unless you are building a cross-compiler, you
15318 should normally not specify either @option{-mnew-mnemonics} or
15319 @option{-mold-mnemonics}, but should instead accept the default.
15321 @item -mcpu=@var{cpu_type}
15323 Set architecture type, register usage, choice of mnemonics, and
15324 instruction scheduling parameters for machine type @var{cpu_type}.
15325 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15326 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15327 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15328 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15329 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15330 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15331 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15332 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15333 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15334 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15335 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15337 @option{-mcpu=common} selects a completely generic processor. Code
15338 generated under this option will run on any POWER or PowerPC processor.
15339 GCC will use only the instructions in the common subset of both
15340 architectures, and will not use the MQ register. GCC assumes a generic
15341 processor model for scheduling purposes.
15343 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15344 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15345 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15346 types, with an appropriate, generic processor model assumed for
15347 scheduling purposes.
15349 The other options specify a specific processor. Code generated under
15350 those options will run best on that processor, and may not run at all on
15353 The @option{-mcpu} options automatically enable or disable the
15356 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15357 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15358 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15359 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15361 The particular options set for any particular CPU will vary between
15362 compiler versions, depending on what setting seems to produce optimal
15363 code for that CPU; it doesn't necessarily reflect the actual hardware's
15364 capabilities. If you wish to set an individual option to a particular
15365 value, you may specify it after the @option{-mcpu} option, like
15366 @samp{-mcpu=970 -mno-altivec}.
15368 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15369 not enabled or disabled by the @option{-mcpu} option at present because
15370 AIX does not have full support for these options. You may still
15371 enable or disable them individually if you're sure it'll work in your
15374 @item -mtune=@var{cpu_type}
15376 Set the instruction scheduling parameters for machine type
15377 @var{cpu_type}, but do not set the architecture type, register usage, or
15378 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15379 values for @var{cpu_type} are used for @option{-mtune} as for
15380 @option{-mcpu}. If both are specified, the code generated will use the
15381 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15382 scheduling parameters set by @option{-mtune}.
15384 @item -mcmodel=small
15385 @opindex mcmodel=small
15386 Generate PowerPC64 code for the small model: The TOC is limited to
15389 @item -mcmodel=medium
15390 @opindex mcmodel=medium
15391 Generate PowerPC64 code for the medium model: The TOC and other static
15392 data may be up to a total of 4G in size.
15394 @item -mcmodel=large
15395 @opindex mcmodel=large
15396 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15397 in size. Other data and code is only limited by the 64-bit address
15401 @itemx -mno-altivec
15403 @opindex mno-altivec
15404 Generate code that uses (does not use) AltiVec instructions, and also
15405 enable the use of built-in functions that allow more direct access to
15406 the AltiVec instruction set. You may also need to set
15407 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15413 @opindex mno-vrsave
15414 Generate VRSAVE instructions when generating AltiVec code.
15416 @item -mgen-cell-microcode
15417 @opindex mgen-cell-microcode
15418 Generate Cell microcode instructions
15420 @item -mwarn-cell-microcode
15421 @opindex mwarn-cell-microcode
15422 Warning when a Cell microcode instruction is going to emitted. An example
15423 of a Cell microcode instruction is a variable shift.
15426 @opindex msecure-plt
15427 Generate code that allows ld and ld.so to build executables and shared
15428 libraries with non-exec .plt and .got sections. This is a PowerPC
15429 32-bit SYSV ABI option.
15433 Generate code that uses a BSS .plt section that ld.so fills in, and
15434 requires .plt and .got sections that are both writable and executable.
15435 This is a PowerPC 32-bit SYSV ABI option.
15441 This switch enables or disables the generation of ISEL instructions.
15443 @item -misel=@var{yes/no}
15444 This switch has been deprecated. Use @option{-misel} and
15445 @option{-mno-isel} instead.
15451 This switch enables or disables the generation of SPE simd
15457 @opindex mno-paired
15458 This switch enables or disables the generation of PAIRED simd
15461 @item -mspe=@var{yes/no}
15462 This option has been deprecated. Use @option{-mspe} and
15463 @option{-mno-spe} instead.
15469 Generate code that uses (does not use) vector/scalar (VSX)
15470 instructions, and also enable the use of built-in functions that allow
15471 more direct access to the VSX instruction set.
15473 @item -mfloat-gprs=@var{yes/single/double/no}
15474 @itemx -mfloat-gprs
15475 @opindex mfloat-gprs
15476 This switch enables or disables the generation of floating point
15477 operations on the general purpose registers for architectures that
15480 The argument @var{yes} or @var{single} enables the use of
15481 single-precision floating point operations.
15483 The argument @var{double} enables the use of single and
15484 double-precision floating point operations.
15486 The argument @var{no} disables floating point operations on the
15487 general purpose registers.
15489 This option is currently only available on the MPC854x.
15495 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15496 targets (including GNU/Linux). The 32-bit environment sets int, long
15497 and pointer to 32 bits and generates code that runs on any PowerPC
15498 variant. The 64-bit environment sets int to 32 bits and long and
15499 pointer to 64 bits, and generates code for PowerPC64, as for
15500 @option{-mpowerpc64}.
15503 @itemx -mno-fp-in-toc
15504 @itemx -mno-sum-in-toc
15505 @itemx -mminimal-toc
15507 @opindex mno-fp-in-toc
15508 @opindex mno-sum-in-toc
15509 @opindex mminimal-toc
15510 Modify generation of the TOC (Table Of Contents), which is created for
15511 every executable file. The @option{-mfull-toc} option is selected by
15512 default. In that case, GCC will allocate at least one TOC entry for
15513 each unique non-automatic variable reference in your program. GCC
15514 will also place floating-point constants in the TOC@. However, only
15515 16,384 entries are available in the TOC@.
15517 If you receive a linker error message that saying you have overflowed
15518 the available TOC space, you can reduce the amount of TOC space used
15519 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15520 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15521 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15522 generate code to calculate the sum of an address and a constant at
15523 run-time instead of putting that sum into the TOC@. You may specify one
15524 or both of these options. Each causes GCC to produce very slightly
15525 slower and larger code at the expense of conserving TOC space.
15527 If you still run out of space in the TOC even when you specify both of
15528 these options, specify @option{-mminimal-toc} instead. This option causes
15529 GCC to make only one TOC entry for every file. When you specify this
15530 option, GCC will produce code that is slower and larger but which
15531 uses extremely little TOC space. You may wish to use this option
15532 only on files that contain less frequently executed code.
15538 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15539 @code{long} type, and the infrastructure needed to support them.
15540 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15541 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15542 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15545 @itemx -mno-xl-compat
15546 @opindex mxl-compat
15547 @opindex mno-xl-compat
15548 Produce code that conforms more closely to IBM XL compiler semantics
15549 when using AIX-compatible ABI@. Pass floating-point arguments to
15550 prototyped functions beyond the register save area (RSA) on the stack
15551 in addition to argument FPRs. Do not assume that most significant
15552 double in 128-bit long double value is properly rounded when comparing
15553 values and converting to double. Use XL symbol names for long double
15556 The AIX calling convention was extended but not initially documented to
15557 handle an obscure K&R C case of calling a function that takes the
15558 address of its arguments with fewer arguments than declared. IBM XL
15559 compilers access floating point arguments which do not fit in the
15560 RSA from the stack when a subroutine is compiled without
15561 optimization. Because always storing floating-point arguments on the
15562 stack is inefficient and rarely needed, this option is not enabled by
15563 default and only is necessary when calling subroutines compiled by IBM
15564 XL compilers without optimization.
15568 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15569 application written to use message passing with special startup code to
15570 enable the application to run. The system must have PE installed in the
15571 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15572 must be overridden with the @option{-specs=} option to specify the
15573 appropriate directory location. The Parallel Environment does not
15574 support threads, so the @option{-mpe} option and the @option{-pthread}
15575 option are incompatible.
15577 @item -malign-natural
15578 @itemx -malign-power
15579 @opindex malign-natural
15580 @opindex malign-power
15581 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15582 @option{-malign-natural} overrides the ABI-defined alignment of larger
15583 types, such as floating-point doubles, on their natural size-based boundary.
15584 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15585 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15587 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15591 @itemx -mhard-float
15592 @opindex msoft-float
15593 @opindex mhard-float
15594 Generate code that does not use (uses) the floating-point register set.
15595 Software floating point emulation is provided if you use the
15596 @option{-msoft-float} option, and pass the option to GCC when linking.
15598 @item -msingle-float
15599 @itemx -mdouble-float
15600 @opindex msingle-float
15601 @opindex mdouble-float
15602 Generate code for single or double-precision floating point operations.
15603 @option{-mdouble-float} implies @option{-msingle-float}.
15606 @opindex msimple-fpu
15607 Do not generate sqrt and div instructions for hardware floating point unit.
15611 Specify type of floating point unit. Valid values are @var{sp_lite}
15612 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15613 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15614 and @var{dp_full} (equivalent to -mdouble-float).
15617 @opindex mxilinx-fpu
15618 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15621 @itemx -mno-multiple
15623 @opindex mno-multiple
15624 Generate code that uses (does not use) the load multiple word
15625 instructions and the store multiple word instructions. These
15626 instructions are generated by default on POWER systems, and not
15627 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15628 endian PowerPC systems, since those instructions do not work when the
15629 processor is in little endian mode. The exceptions are PPC740 and
15630 PPC750 which permit the instructions usage in little endian mode.
15635 @opindex mno-string
15636 Generate code that uses (does not use) the load string instructions
15637 and the store string word instructions to save multiple registers and
15638 do small block moves. These instructions are generated by default on
15639 POWER systems, and not generated on PowerPC systems. Do not use
15640 @option{-mstring} on little endian PowerPC systems, since those
15641 instructions do not work when the processor is in little endian mode.
15642 The exceptions are PPC740 and PPC750 which permit the instructions
15643 usage in little endian mode.
15648 @opindex mno-update
15649 Generate code that uses (does not use) the load or store instructions
15650 that update the base register to the address of the calculated memory
15651 location. These instructions are generated by default. If you use
15652 @option{-mno-update}, there is a small window between the time that the
15653 stack pointer is updated and the address of the previous frame is
15654 stored, which means code that walks the stack frame across interrupts or
15655 signals may get corrupted data.
15657 @item -mavoid-indexed-addresses
15658 @itemx -mno-avoid-indexed-addresses
15659 @opindex mavoid-indexed-addresses
15660 @opindex mno-avoid-indexed-addresses
15661 Generate code that tries to avoid (not avoid) the use of indexed load
15662 or store instructions. These instructions can incur a performance
15663 penalty on Power6 processors in certain situations, such as when
15664 stepping through large arrays that cross a 16M boundary. This option
15665 is enabled by default when targetting Power6 and disabled otherwise.
15668 @itemx -mno-fused-madd
15669 @opindex mfused-madd
15670 @opindex mno-fused-madd
15671 Generate code that uses (does not use) the floating point multiply and
15672 accumulate instructions. These instructions are generated by default
15673 if hardware floating point is used. The machine dependent
15674 @option{-mfused-madd} option is now mapped to the machine independent
15675 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15676 mapped to @option{-ffp-contract=off}.
15682 Generate code that uses (does not use) the half-word multiply and
15683 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15684 These instructions are generated by default when targetting those
15691 Generate code that uses (does not use) the string-search @samp{dlmzb}
15692 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15693 generated by default when targetting those processors.
15695 @item -mno-bit-align
15697 @opindex mno-bit-align
15698 @opindex mbit-align
15699 On System V.4 and embedded PowerPC systems do not (do) force structures
15700 and unions that contain bit-fields to be aligned to the base type of the
15703 For example, by default a structure containing nothing but 8
15704 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15705 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15706 the structure would be aligned to a 1 byte boundary and be one byte in
15709 @item -mno-strict-align
15710 @itemx -mstrict-align
15711 @opindex mno-strict-align
15712 @opindex mstrict-align
15713 On System V.4 and embedded PowerPC systems do not (do) assume that
15714 unaligned memory references will be handled by the system.
15716 @item -mrelocatable
15717 @itemx -mno-relocatable
15718 @opindex mrelocatable
15719 @opindex mno-relocatable
15720 Generate code that allows (does not allow) a static executable to be
15721 relocated to a different address at runtime. A simple embedded
15722 PowerPC system loader should relocate the entire contents of
15723 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15724 a table of 32-bit addresses generated by this option. For this to
15725 work, all objects linked together must be compiled with
15726 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15727 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15729 @item -mrelocatable-lib
15730 @itemx -mno-relocatable-lib
15731 @opindex mrelocatable-lib
15732 @opindex mno-relocatable-lib
15733 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15734 @code{.fixup} section to allow static executables to be relocated at
15735 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15736 alignment of @option{-mrelocatable}. Objects compiled with
15737 @option{-mrelocatable-lib} may be linked with objects compiled with
15738 any combination of the @option{-mrelocatable} options.
15744 On System V.4 and embedded PowerPC systems do not (do) assume that
15745 register 2 contains a pointer to a global area pointing to the addresses
15746 used in the program.
15749 @itemx -mlittle-endian
15751 @opindex mlittle-endian
15752 On System V.4 and embedded PowerPC systems compile code for the
15753 processor in little endian mode. The @option{-mlittle-endian} option is
15754 the same as @option{-mlittle}.
15757 @itemx -mbig-endian
15759 @opindex mbig-endian
15760 On System V.4 and embedded PowerPC systems compile code for the
15761 processor in big endian mode. The @option{-mbig-endian} option is
15762 the same as @option{-mbig}.
15764 @item -mdynamic-no-pic
15765 @opindex mdynamic-no-pic
15766 On Darwin and Mac OS X systems, compile code so that it is not
15767 relocatable, but that its external references are relocatable. The
15768 resulting code is suitable for applications, but not shared
15771 @item -msingle-pic-base
15772 @opindex msingle-pic-base
15773 Treat the register used for PIC addressing as read-only, rather than
15774 loading it in the prologue for each function. The run-time system is
15775 responsible for initializing this register with an appropriate value
15776 before execution begins.
15778 @item -mprioritize-restricted-insns=@var{priority}
15779 @opindex mprioritize-restricted-insns
15780 This option controls the priority that is assigned to
15781 dispatch-slot restricted instructions during the second scheduling
15782 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15783 @var{no/highest/second-highest} priority to dispatch slot restricted
15786 @item -msched-costly-dep=@var{dependence_type}
15787 @opindex msched-costly-dep
15788 This option controls which dependences are considered costly
15789 by the target during instruction scheduling. The argument
15790 @var{dependence_type} takes one of the following values:
15791 @var{no}: no dependence is costly,
15792 @var{all}: all dependences are costly,
15793 @var{true_store_to_load}: a true dependence from store to load is costly,
15794 @var{store_to_load}: any dependence from store to load is costly,
15795 @var{number}: any dependence which latency >= @var{number} is costly.
15797 @item -minsert-sched-nops=@var{scheme}
15798 @opindex minsert-sched-nops
15799 This option controls which nop insertion scheme will be used during
15800 the second scheduling pass. The argument @var{scheme} takes one of the
15802 @var{no}: Don't insert nops.
15803 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15804 according to the scheduler's grouping.
15805 @var{regroup_exact}: Insert nops to force costly dependent insns into
15806 separate groups. Insert exactly as many nops as needed to force an insn
15807 to a new group, according to the estimated processor grouping.
15808 @var{number}: Insert nops to force costly dependent insns into
15809 separate groups. Insert @var{number} nops to force an insn to a new group.
15812 @opindex mcall-sysv
15813 On System V.4 and embedded PowerPC systems compile code using calling
15814 conventions that adheres to the March 1995 draft of the System V
15815 Application Binary Interface, PowerPC processor supplement. This is the
15816 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15818 @item -mcall-sysv-eabi
15820 @opindex mcall-sysv-eabi
15821 @opindex mcall-eabi
15822 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15824 @item -mcall-sysv-noeabi
15825 @opindex mcall-sysv-noeabi
15826 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15828 @item -mcall-aixdesc
15830 On System V.4 and embedded PowerPC systems compile code for the AIX
15834 @opindex mcall-linux
15835 On System V.4 and embedded PowerPC systems compile code for the
15836 Linux-based GNU system.
15838 @item -mcall-freebsd
15839 @opindex mcall-freebsd
15840 On System V.4 and embedded PowerPC systems compile code for the
15841 FreeBSD operating system.
15843 @item -mcall-netbsd
15844 @opindex mcall-netbsd
15845 On System V.4 and embedded PowerPC systems compile code for the
15846 NetBSD operating system.
15848 @item -mcall-openbsd
15849 @opindex mcall-netbsd
15850 On System V.4 and embedded PowerPC systems compile code for the
15851 OpenBSD operating system.
15853 @item -maix-struct-return
15854 @opindex maix-struct-return
15855 Return all structures in memory (as specified by the AIX ABI)@.
15857 @item -msvr4-struct-return
15858 @opindex msvr4-struct-return
15859 Return structures smaller than 8 bytes in registers (as specified by the
15862 @item -mabi=@var{abi-type}
15864 Extend the current ABI with a particular extension, or remove such extension.
15865 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15866 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15870 Extend the current ABI with SPE ABI extensions. This does not change
15871 the default ABI, instead it adds the SPE ABI extensions to the current
15875 @opindex mabi=no-spe
15876 Disable Booke SPE ABI extensions for the current ABI@.
15878 @item -mabi=ibmlongdouble
15879 @opindex mabi=ibmlongdouble
15880 Change the current ABI to use IBM extended precision long double.
15881 This is a PowerPC 32-bit SYSV ABI option.
15883 @item -mabi=ieeelongdouble
15884 @opindex mabi=ieeelongdouble
15885 Change the current ABI to use IEEE extended precision long double.
15886 This is a PowerPC 32-bit Linux ABI option.
15889 @itemx -mno-prototype
15890 @opindex mprototype
15891 @opindex mno-prototype
15892 On System V.4 and embedded PowerPC systems assume that all calls to
15893 variable argument functions are properly prototyped. Otherwise, the
15894 compiler must insert an instruction before every non prototyped call to
15895 set or clear bit 6 of the condition code register (@var{CR}) to
15896 indicate whether floating point values were passed in the floating point
15897 registers in case the function takes a variable arguments. With
15898 @option{-mprototype}, only calls to prototyped variable argument functions
15899 will set or clear the bit.
15903 On embedded PowerPC systems, assume that the startup module is called
15904 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15905 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15910 On embedded PowerPC systems, assume that the startup module is called
15911 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15916 On embedded PowerPC systems, assume that the startup module is called
15917 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15920 @item -myellowknife
15921 @opindex myellowknife
15922 On embedded PowerPC systems, assume that the startup module is called
15923 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15928 On System V.4 and embedded PowerPC systems, specify that you are
15929 compiling for a VxWorks system.
15933 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15934 header to indicate that @samp{eabi} extended relocations are used.
15940 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15941 Embedded Applications Binary Interface (eabi) which is a set of
15942 modifications to the System V.4 specifications. Selecting @option{-meabi}
15943 means that the stack is aligned to an 8 byte boundary, a function
15944 @code{__eabi} is called to from @code{main} to set up the eabi
15945 environment, and the @option{-msdata} option can use both @code{r2} and
15946 @code{r13} to point to two separate small data areas. Selecting
15947 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15948 do not call an initialization function from @code{main}, and the
15949 @option{-msdata} option will only use @code{r13} to point to a single
15950 small data area. The @option{-meabi} option is on by default if you
15951 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15954 @opindex msdata=eabi
15955 On System V.4 and embedded PowerPC systems, put small initialized
15956 @code{const} global and static data in the @samp{.sdata2} section, which
15957 is pointed to by register @code{r2}. Put small initialized
15958 non-@code{const} global and static data in the @samp{.sdata} section,
15959 which is pointed to by register @code{r13}. Put small uninitialized
15960 global and static data in the @samp{.sbss} section, which is adjacent to
15961 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15962 incompatible with the @option{-mrelocatable} option. The
15963 @option{-msdata=eabi} option also sets the @option{-memb} option.
15966 @opindex msdata=sysv
15967 On System V.4 and embedded PowerPC systems, put small global and static
15968 data in the @samp{.sdata} section, which is pointed to by register
15969 @code{r13}. Put small uninitialized global and static data in the
15970 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15971 The @option{-msdata=sysv} option is incompatible with the
15972 @option{-mrelocatable} option.
15974 @item -msdata=default
15976 @opindex msdata=default
15978 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15979 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15980 same as @option{-msdata=sysv}.
15983 @opindex msdata=data
15984 On System V.4 and embedded PowerPC systems, put small global
15985 data in the @samp{.sdata} section. Put small uninitialized global
15986 data in the @samp{.sbss} section. Do not use register @code{r13}
15987 to address small data however. This is the default behavior unless
15988 other @option{-msdata} options are used.
15992 @opindex msdata=none
15994 On embedded PowerPC systems, put all initialized global and static data
15995 in the @samp{.data} section, and all uninitialized data in the
15996 @samp{.bss} section.
15998 @item -mblock-move-inline-limit=@var{num}
15999 @opindex mblock-move-inline-limit
16000 Inline all block moves (such as calls to @code{memcpy} or structure
16001 copies) less than or equal to @var{num} bytes. The minimum value for
16002 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16003 targets. The default value is target-specific.
16007 @cindex smaller data references (PowerPC)
16008 @cindex .sdata/.sdata2 references (PowerPC)
16009 On embedded PowerPC systems, put global and static items less than or
16010 equal to @var{num} bytes into the small data or bss sections instead of
16011 the normal data or bss section. By default, @var{num} is 8. The
16012 @option{-G @var{num}} switch is also passed to the linker.
16013 All modules should be compiled with the same @option{-G @var{num}} value.
16016 @itemx -mno-regnames
16018 @opindex mno-regnames
16019 On System V.4 and embedded PowerPC systems do (do not) emit register
16020 names in the assembly language output using symbolic forms.
16023 @itemx -mno-longcall
16025 @opindex mno-longcall
16026 By default assume that all calls are far away so that a longer more
16027 expensive calling sequence is required. This is required for calls
16028 further than 32 megabytes (33,554,432 bytes) from the current location.
16029 A short call will be generated if the compiler knows
16030 the call cannot be that far away. This setting can be overridden by
16031 the @code{shortcall} function attribute, or by @code{#pragma
16034 Some linkers are capable of detecting out-of-range calls and generating
16035 glue code on the fly. On these systems, long calls are unnecessary and
16036 generate slower code. As of this writing, the AIX linker can do this,
16037 as can the GNU linker for PowerPC/64. It is planned to add this feature
16038 to the GNU linker for 32-bit PowerPC systems as well.
16040 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16041 callee, L42'', plus a ``branch island'' (glue code). The two target
16042 addresses represent the callee and the ``branch island''. The
16043 Darwin/PPC linker will prefer the first address and generate a ``bl
16044 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16045 otherwise, the linker will generate ``bl L42'' to call the ``branch
16046 island''. The ``branch island'' is appended to the body of the
16047 calling function; it computes the full 32-bit address of the callee
16050 On Mach-O (Darwin) systems, this option directs the compiler emit to
16051 the glue for every direct call, and the Darwin linker decides whether
16052 to use or discard it.
16054 In the future, we may cause GCC to ignore all longcall specifications
16055 when the linker is known to generate glue.
16057 @item -mtls-markers
16058 @itemx -mno-tls-markers
16059 @opindex mtls-markers
16060 @opindex mno-tls-markers
16061 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16062 specifying the function argument. The relocation allows ld to
16063 reliably associate function call with argument setup instructions for
16064 TLS optimization, which in turn allows gcc to better schedule the
16069 Adds support for multithreading with the @dfn{pthreads} library.
16070 This option sets flags for both the preprocessor and linker.
16075 This option will enable GCC to use the reciprocal estimate and
16076 reciprocal square root estimate instructions with additional
16077 Newton-Raphson steps to increase precision instead of doing a divide or
16078 square root and divide for floating point arguments. You should use
16079 the @option{-ffast-math} option when using @option{-mrecip} (or at
16080 least @option{-funsafe-math-optimizations},
16081 @option{-finite-math-only}, @option{-freciprocal-math} and
16082 @option{-fno-trapping-math}). Note that while the throughput of the
16083 sequence is generally higher than the throughput of the non-reciprocal
16084 instruction, the precision of the sequence can be decreased by up to 2
16085 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16088 @item -mrecip=@var{opt}
16089 @opindex mrecip=opt
16090 This option allows to control which reciprocal estimate instructions
16091 may be used. @var{opt} is a comma separated list of options, that may
16092 be preceded by a @code{!} to invert the option:
16093 @code{all}: enable all estimate instructions,
16094 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16095 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16096 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16097 @code{divf}: enable the single precision reciprocal approximation instructions;
16098 @code{divd}: enable the double precision reciprocal approximation instructions;
16099 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16100 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16101 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16103 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16104 all of the reciprocal estimate instructions, except for the
16105 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16106 which handle the double precision reciprocal square root calculations.
16108 @item -mrecip-precision
16109 @itemx -mno-recip-precision
16110 @opindex mrecip-precision
16111 Assume (do not assume) that the reciprocal estimate instructions
16112 provide higher precision estimates than is mandated by the powerpc
16113 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16114 automatically selects @option{-mrecip-precision}. The double
16115 precision square root estimate instructions are not generated by
16116 default on low precision machines, since they do not provide an
16117 estimate that converges after three steps.
16119 @item -mveclibabi=@var{type}
16120 @opindex mveclibabi
16121 Specifies the ABI type to use for vectorizing intrinsics using an
16122 external library. The only type supported at present is @code{mass},
16123 which specifies to use IBM's Mathematical Acceleration Subsystem
16124 (MASS) libraries for vectorizing intrinsics using external libraries.
16125 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16126 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16127 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16128 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16129 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16130 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16131 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16132 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16133 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16134 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16135 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16136 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16137 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16138 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16139 for power7. Both @option{-ftree-vectorize} and
16140 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16141 libraries will have to be specified at link time.
16146 Generate (do not generate) the @code{friz} instruction when the
16147 @option{-funsafe-math-optimizations} option is used to optimize
16148 rounding a floating point value to 64-bit integer and back to floating
16149 point. The @code{friz} instruction does not return the same value if
16150 the floating point number is too large to fit in an integer.
16154 @subsection RX Options
16157 These command line options are defined for RX targets:
16160 @item -m64bit-doubles
16161 @itemx -m32bit-doubles
16162 @opindex m64bit-doubles
16163 @opindex m32bit-doubles
16164 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16165 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16166 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16167 works on 32-bit values, which is why the default is
16168 @option{-m32bit-doubles}.
16174 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16175 floating point hardware. The default is enabled for the @var{RX600}
16176 series and disabled for the @var{RX200} series.
16178 Floating point instructions will only be generated for 32-bit floating
16179 point values however, so if the @option{-m64bit-doubles} option is in
16180 use then the FPU hardware will not be used for doubles.
16182 @emph{Note} If the @option{-fpu} option is enabled then
16183 @option{-funsafe-math-optimizations} is also enabled automatically.
16184 This is because the RX FPU instructions are themselves unsafe.
16186 @item -mcpu=@var{name}
16188 Selects the type of RX CPU to be targeted. Currently three types are
16189 supported, the generic @var{RX600} and @var{RX200} series hardware and
16190 the specific @var{RX610} CPU. The default is @var{RX600}.
16192 The only difference between @var{RX600} and @var{RX610} is that the
16193 @var{RX610} does not support the @code{MVTIPL} instruction.
16195 The @var{RX200} series does not have a hardware floating point unit
16196 and so @option{-nofpu} is enabled by default when this type is
16199 @item -mbig-endian-data
16200 @itemx -mlittle-endian-data
16201 @opindex mbig-endian-data
16202 @opindex mlittle-endian-data
16203 Store data (but not code) in the big-endian format. The default is
16204 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16207 @item -msmall-data-limit=@var{N}
16208 @opindex msmall-data-limit
16209 Specifies the maximum size in bytes of global and static variables
16210 which can be placed into the small data area. Using the small data
16211 area can lead to smaller and faster code, but the size of area is
16212 limited and it is up to the programmer to ensure that the area does
16213 not overflow. Also when the small data area is used one of the RX's
16214 registers (@code{r13}) is reserved for use pointing to this area, so
16215 it is no longer available for use by the compiler. This could result
16216 in slower and/or larger code if variables which once could have been
16217 held in @code{r13} are now pushed onto the stack.
16219 Note, common variables (variables which have not been initialised) and
16220 constants are not placed into the small data area as they are assigned
16221 to other sections in the output executable.
16223 The default value is zero, which disables this feature. Note, this
16224 feature is not enabled by default with higher optimization levels
16225 (@option{-O2} etc) because of the potentially detrimental effects of
16226 reserving register @code{r13}. It is up to the programmer to
16227 experiment and discover whether this feature is of benefit to their
16234 Use the simulator runtime. The default is to use the libgloss board
16237 @item -mas100-syntax
16238 @itemx -mno-as100-syntax
16239 @opindex mas100-syntax
16240 @opindex mno-as100-syntax
16241 When generating assembler output use a syntax that is compatible with
16242 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16243 assembler but it has some restrictions so generating it is not the
16246 @item -mmax-constant-size=@var{N}
16247 @opindex mmax-constant-size
16248 Specifies the maximum size, in bytes, of a constant that can be used as
16249 an operand in a RX instruction. Although the RX instruction set does
16250 allow constants of up to 4 bytes in length to be used in instructions,
16251 a longer value equates to a longer instruction. Thus in some
16252 circumstances it can be beneficial to restrict the size of constants
16253 that are used in instructions. Constants that are too big are instead
16254 placed into a constant pool and referenced via register indirection.
16256 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16257 or 4 means that constants of any size are allowed.
16261 Enable linker relaxation. Linker relaxation is a process whereby the
16262 linker will attempt to reduce the size of a program by finding shorter
16263 versions of various instructions. Disabled by default.
16265 @item -mint-register=@var{N}
16266 @opindex mint-register
16267 Specify the number of registers to reserve for fast interrupt handler
16268 functions. The value @var{N} can be between 0 and 4. A value of 1
16269 means that register @code{r13} will be reserved for the exclusive use
16270 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16271 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16272 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16273 A value of 0, the default, does not reserve any registers.
16275 @item -msave-acc-in-interrupts
16276 @opindex msave-acc-in-interrupts
16277 Specifies that interrupt handler functions should preserve the
16278 accumulator register. This is only necessary if normal code might use
16279 the accumulator register, for example because it performs 64-bit
16280 multiplications. The default is to ignore the accumulator as this
16281 makes the interrupt handlers faster.
16285 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16286 has special significance to the RX port when used with the
16287 @code{interrupt} function attribute. This attribute indicates a
16288 function intended to process fast interrupts. GCC will will ensure
16289 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16290 and/or @code{r13} and only provided that the normal use of the
16291 corresponding registers have been restricted via the
16292 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16295 @node S/390 and zSeries Options
16296 @subsection S/390 and zSeries Options
16297 @cindex S/390 and zSeries Options
16299 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16303 @itemx -msoft-float
16304 @opindex mhard-float
16305 @opindex msoft-float
16306 Use (do not use) the hardware floating-point instructions and registers
16307 for floating-point operations. When @option{-msoft-float} is specified,
16308 functions in @file{libgcc.a} will be used to perform floating-point
16309 operations. When @option{-mhard-float} is specified, the compiler
16310 generates IEEE floating-point instructions. This is the default.
16313 @itemx -mno-hard-dfp
16315 @opindex mno-hard-dfp
16316 Use (do not use) the hardware decimal-floating-point instructions for
16317 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16318 specified, functions in @file{libgcc.a} will be used to perform
16319 decimal-floating-point operations. When @option{-mhard-dfp} is
16320 specified, the compiler generates decimal-floating-point hardware
16321 instructions. This is the default for @option{-march=z9-ec} or higher.
16323 @item -mlong-double-64
16324 @itemx -mlong-double-128
16325 @opindex mlong-double-64
16326 @opindex mlong-double-128
16327 These switches control the size of @code{long double} type. A size
16328 of 64bit makes the @code{long double} type equivalent to the @code{double}
16329 type. This is the default.
16332 @itemx -mno-backchain
16333 @opindex mbackchain
16334 @opindex mno-backchain
16335 Store (do not store) the address of the caller's frame as backchain pointer
16336 into the callee's stack frame.
16337 A backchain may be needed to allow debugging using tools that do not understand
16338 DWARF-2 call frame information.
16339 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16340 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16341 the backchain is placed into the topmost word of the 96/160 byte register
16344 In general, code compiled with @option{-mbackchain} is call-compatible with
16345 code compiled with @option{-mmo-backchain}; however, use of the backchain
16346 for debugging purposes usually requires that the whole binary is built with
16347 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16348 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16349 to build a linux kernel use @option{-msoft-float}.
16351 The default is to not maintain the backchain.
16353 @item -mpacked-stack
16354 @itemx -mno-packed-stack
16355 @opindex mpacked-stack
16356 @opindex mno-packed-stack
16357 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16358 specified, the compiler uses the all fields of the 96/160 byte register save
16359 area only for their default purpose; unused fields still take up stack space.
16360 When @option{-mpacked-stack} is specified, register save slots are densely
16361 packed at the top of the register save area; unused space is reused for other
16362 purposes, allowing for more efficient use of the available stack space.
16363 However, when @option{-mbackchain} is also in effect, the topmost word of
16364 the save area is always used to store the backchain, and the return address
16365 register is always saved two words below the backchain.
16367 As long as the stack frame backchain is not used, code generated with
16368 @option{-mpacked-stack} is call-compatible with code generated with
16369 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16370 S/390 or zSeries generated code that uses the stack frame backchain at run
16371 time, not just for debugging purposes. Such code is not call-compatible
16372 with code compiled with @option{-mpacked-stack}. Also, note that the
16373 combination of @option{-mbackchain},
16374 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16375 to build a linux kernel use @option{-msoft-float}.
16377 The default is to not use the packed stack layout.
16380 @itemx -mno-small-exec
16381 @opindex msmall-exec
16382 @opindex mno-small-exec
16383 Generate (or do not generate) code using the @code{bras} instruction
16384 to do subroutine calls.
16385 This only works reliably if the total executable size does not
16386 exceed 64k. The default is to use the @code{basr} instruction instead,
16387 which does not have this limitation.
16393 When @option{-m31} is specified, generate code compliant to the
16394 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16395 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16396 particular to generate 64-bit instructions. For the @samp{s390}
16397 targets, the default is @option{-m31}, while the @samp{s390x}
16398 targets default to @option{-m64}.
16404 When @option{-mzarch} is specified, generate code using the
16405 instructions available on z/Architecture.
16406 When @option{-mesa} is specified, generate code using the
16407 instructions available on ESA/390. Note that @option{-mesa} is
16408 not possible with @option{-m64}.
16409 When generating code compliant to the GNU/Linux for S/390 ABI,
16410 the default is @option{-mesa}. When generating code compliant
16411 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16417 Generate (or do not generate) code using the @code{mvcle} instruction
16418 to perform block moves. When @option{-mno-mvcle} is specified,
16419 use a @code{mvc} loop instead. This is the default unless optimizing for
16426 Print (or do not print) additional debug information when compiling.
16427 The default is to not print debug information.
16429 @item -march=@var{cpu-type}
16431 Generate code that will run on @var{cpu-type}, which is the name of a system
16432 representing a certain processor type. Possible values for
16433 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16434 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16435 When generating code using the instructions available on z/Architecture,
16436 the default is @option{-march=z900}. Otherwise, the default is
16437 @option{-march=g5}.
16439 @item -mtune=@var{cpu-type}
16441 Tune to @var{cpu-type} everything applicable about the generated code,
16442 except for the ABI and the set of available instructions.
16443 The list of @var{cpu-type} values is the same as for @option{-march}.
16444 The default is the value used for @option{-march}.
16447 @itemx -mno-tpf-trace
16448 @opindex mtpf-trace
16449 @opindex mno-tpf-trace
16450 Generate code that adds (does not add) in TPF OS specific branches to trace
16451 routines in the operating system. This option is off by default, even
16452 when compiling for the TPF OS@.
16455 @itemx -mno-fused-madd
16456 @opindex mfused-madd
16457 @opindex mno-fused-madd
16458 Generate code that uses (does not use) the floating point multiply and
16459 accumulate instructions. These instructions are generated by default if
16460 hardware floating point is used.
16462 @item -mwarn-framesize=@var{framesize}
16463 @opindex mwarn-framesize
16464 Emit a warning if the current function exceeds the given frame size. Because
16465 this is a compile time check it doesn't need to be a real problem when the program
16466 runs. It is intended to identify functions which most probably cause
16467 a stack overflow. It is useful to be used in an environment with limited stack
16468 size e.g.@: the linux kernel.
16470 @item -mwarn-dynamicstack
16471 @opindex mwarn-dynamicstack
16472 Emit a warning if the function calls alloca or uses dynamically
16473 sized arrays. This is generally a bad idea with a limited stack size.
16475 @item -mstack-guard=@var{stack-guard}
16476 @itemx -mstack-size=@var{stack-size}
16477 @opindex mstack-guard
16478 @opindex mstack-size
16479 If these options are provided the s390 back end emits additional instructions in
16480 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16481 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16482 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16483 the frame size of the compiled function is chosen.
16484 These options are intended to be used to help debugging stack overflow problems.
16485 The additionally emitted code causes only little overhead and hence can also be
16486 used in production like systems without greater performance degradation. The given
16487 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16488 @var{stack-guard} without exceeding 64k.
16489 In order to be efficient the extra code makes the assumption that the stack starts
16490 at an address aligned to the value given by @var{stack-size}.
16491 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16494 @node Score Options
16495 @subsection Score Options
16496 @cindex Score Options
16498 These options are defined for Score implementations:
16503 Compile code for big endian mode. This is the default.
16507 Compile code for little endian mode.
16511 Disable generate bcnz instruction.
16515 Enable generate unaligned load and store instruction.
16519 Enable the use of multiply-accumulate instructions. Disabled by default.
16523 Specify the SCORE5 as the target architecture.
16527 Specify the SCORE5U of the target architecture.
16531 Specify the SCORE7 as the target architecture. This is the default.
16535 Specify the SCORE7D as the target architecture.
16539 @subsection SH Options
16541 These @samp{-m} options are defined for the SH implementations:
16546 Generate code for the SH1.
16550 Generate code for the SH2.
16553 Generate code for the SH2e.
16557 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16558 that the floating-point unit is not used.
16560 @item -m2a-single-only
16561 @opindex m2a-single-only
16562 Generate code for the SH2a-FPU, in such a way that no double-precision
16563 floating point operations are used.
16566 @opindex m2a-single
16567 Generate code for the SH2a-FPU assuming the floating-point unit is in
16568 single-precision mode by default.
16572 Generate code for the SH2a-FPU assuming the floating-point unit is in
16573 double-precision mode by default.
16577 Generate code for the SH3.
16581 Generate code for the SH3e.
16585 Generate code for the SH4 without a floating-point unit.
16587 @item -m4-single-only
16588 @opindex m4-single-only
16589 Generate code for the SH4 with a floating-point unit that only
16590 supports single-precision arithmetic.
16594 Generate code for the SH4 assuming the floating-point unit is in
16595 single-precision mode by default.
16599 Generate code for the SH4.
16603 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16604 floating-point unit is not used.
16606 @item -m4a-single-only
16607 @opindex m4a-single-only
16608 Generate code for the SH4a, in such a way that no double-precision
16609 floating point operations are used.
16612 @opindex m4a-single
16613 Generate code for the SH4a assuming the floating-point unit is in
16614 single-precision mode by default.
16618 Generate code for the SH4a.
16622 Same as @option{-m4a-nofpu}, except that it implicitly passes
16623 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16624 instructions at the moment.
16628 Compile code for the processor in big endian mode.
16632 Compile code for the processor in little endian mode.
16636 Align doubles at 64-bit boundaries. Note that this changes the calling
16637 conventions, and thus some functions from the standard C library will
16638 not work unless you recompile it first with @option{-mdalign}.
16642 Shorten some address references at link time, when possible; uses the
16643 linker option @option{-relax}.
16647 Use 32-bit offsets in @code{switch} tables. The default is to use
16652 Enable the use of bit manipulation instructions on SH2A.
16656 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16657 alignment constraints.
16661 Comply with the calling conventions defined by Renesas.
16665 Comply with the calling conventions defined by Renesas.
16669 Comply with the calling conventions defined for GCC before the Renesas
16670 conventions were available. This option is the default for all
16671 targets of the SH toolchain.
16674 @opindex mnomacsave
16675 Mark the @code{MAC} register as call-clobbered, even if
16676 @option{-mhitachi} is given.
16680 Increase IEEE-compliance of floating-point code.
16681 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16682 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16683 comparisons of NANs / infinities incurs extra overhead in every
16684 floating point comparison, therefore the default is set to
16685 @option{-ffinite-math-only}.
16687 @item -minline-ic_invalidate
16688 @opindex minline-ic_invalidate
16689 Inline code to invalidate instruction cache entries after setting up
16690 nested function trampolines.
16691 This option has no effect if -musermode is in effect and the selected
16692 code generation option (e.g. -m4) does not allow the use of the icbi
16694 If the selected code generation option does not allow the use of the icbi
16695 instruction, and -musermode is not in effect, the inlined code will
16696 manipulate the instruction cache address array directly with an associative
16697 write. This not only requires privileged mode, but it will also
16698 fail if the cache line had been mapped via the TLB and has become unmapped.
16702 Dump instruction size and location in the assembly code.
16705 @opindex mpadstruct
16706 This option is deprecated. It pads structures to multiple of 4 bytes,
16707 which is incompatible with the SH ABI@.
16711 Optimize for space instead of speed. Implied by @option{-Os}.
16714 @opindex mprefergot
16715 When generating position-independent code, emit function calls using
16716 the Global Offset Table instead of the Procedure Linkage Table.
16720 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16721 if the inlined code would not work in user mode.
16722 This is the default when the target is @code{sh-*-linux*}.
16724 @item -multcost=@var{number}
16725 @opindex multcost=@var{number}
16726 Set the cost to assume for a multiply insn.
16728 @item -mdiv=@var{strategy}
16729 @opindex mdiv=@var{strategy}
16730 Set the division strategy to use for SHmedia code. @var{strategy} must be
16731 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16732 inv:call2, inv:fp .
16733 "fp" performs the operation in floating point. This has a very high latency,
16734 but needs only a few instructions, so it might be a good choice if
16735 your code has enough easily exploitable ILP to allow the compiler to
16736 schedule the floating point instructions together with other instructions.
16737 Division by zero causes a floating point exception.
16738 "inv" uses integer operations to calculate the inverse of the divisor,
16739 and then multiplies the dividend with the inverse. This strategy allows
16740 cse and hoisting of the inverse calculation. Division by zero calculates
16741 an unspecified result, but does not trap.
16742 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16743 have been found, or if the entire operation has been hoisted to the same
16744 place, the last stages of the inverse calculation are intertwined with the
16745 final multiply to reduce the overall latency, at the expense of using a few
16746 more instructions, and thus offering fewer scheduling opportunities with
16748 "call" calls a library function that usually implements the inv:minlat
16750 This gives high code density for m5-*media-nofpu compilations.
16751 "call2" uses a different entry point of the same library function, where it
16752 assumes that a pointer to a lookup table has already been set up, which
16753 exposes the pointer load to cse / code hoisting optimizations.
16754 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16755 code generation, but if the code stays unoptimized, revert to the "call",
16756 "call2", or "fp" strategies, respectively. Note that the
16757 potentially-trapping side effect of division by zero is carried by a
16758 separate instruction, so it is possible that all the integer instructions
16759 are hoisted out, but the marker for the side effect stays where it is.
16760 A recombination to fp operations or a call is not possible in that case.
16761 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16762 that the inverse calculation was nor separated from the multiply, they speed
16763 up division where the dividend fits into 20 bits (plus sign where applicable),
16764 by inserting a test to skip a number of operations in this case; this test
16765 slows down the case of larger dividends. inv20u assumes the case of a such
16766 a small dividend to be unlikely, and inv20l assumes it to be likely.
16768 @item -maccumulate-outgoing-args
16769 @opindex maccumulate-outgoing-args
16770 Reserve space once for outgoing arguments in the function prologue rather
16771 than around each call. Generally beneficial for performance and size. Also
16772 needed for unwinding to avoid changing the stack frame around conditional code.
16774 @item -mdivsi3_libfunc=@var{name}
16775 @opindex mdivsi3_libfunc=@var{name}
16776 Set the name of the library function used for 32 bit signed division to
16777 @var{name}. This only affect the name used in the call and inv:call
16778 division strategies, and the compiler will still expect the same
16779 sets of input/output/clobbered registers as if this option was not present.
16781 @item -mfixed-range=@var{register-range}
16782 @opindex mfixed-range
16783 Generate code treating the given register range as fixed registers.
16784 A fixed register is one that the register allocator can not use. This is
16785 useful when compiling kernel code. A register range is specified as
16786 two registers separated by a dash. Multiple register ranges can be
16787 specified separated by a comma.
16789 @item -madjust-unroll
16790 @opindex madjust-unroll
16791 Throttle unrolling to avoid thrashing target registers.
16792 This option only has an effect if the gcc code base supports the
16793 TARGET_ADJUST_UNROLL_MAX target hook.
16795 @item -mindexed-addressing
16796 @opindex mindexed-addressing
16797 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16798 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16799 semantics for the indexed addressing mode. The architecture allows the
16800 implementation of processors with 64 bit MMU, which the OS could use to
16801 get 32 bit addressing, but since no current hardware implementation supports
16802 this or any other way to make the indexed addressing mode safe to use in
16803 the 32 bit ABI, the default is -mno-indexed-addressing.
16805 @item -mgettrcost=@var{number}
16806 @opindex mgettrcost=@var{number}
16807 Set the cost assumed for the gettr instruction to @var{number}.
16808 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16812 Assume pt* instructions won't trap. This will generally generate better
16813 scheduled code, but is unsafe on current hardware. The current architecture
16814 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16815 This has the unintentional effect of making it unsafe to schedule ptabs /
16816 ptrel before a branch, or hoist it out of a loop. For example,
16817 __do_global_ctors, a part of libgcc that runs constructors at program
16818 startup, calls functions in a list which is delimited by @minus{}1. With the
16819 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16820 That means that all the constructors will be run a bit quicker, but when
16821 the loop comes to the end of the list, the program crashes because ptabs
16822 loads @minus{}1 into a target register. Since this option is unsafe for any
16823 hardware implementing the current architecture specification, the default
16824 is -mno-pt-fixed. Unless the user specifies a specific cost with
16825 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16826 this deters register allocation using target registers for storing
16829 @item -minvalid-symbols
16830 @opindex minvalid-symbols
16831 Assume symbols might be invalid. Ordinary function symbols generated by
16832 the compiler will always be valid to load with movi/shori/ptabs or
16833 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16834 to generate symbols that will cause ptabs / ptrel to trap.
16835 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16836 It will then prevent cross-basic-block cse, hoisting and most scheduling
16837 of symbol loads. The default is @option{-mno-invalid-symbols}.
16840 @node Solaris 2 Options
16841 @subsection Solaris 2 Options
16842 @cindex Solaris 2 options
16844 These @samp{-m} options are supported on Solaris 2:
16847 @item -mimpure-text
16848 @opindex mimpure-text
16849 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16850 the compiler to not pass @option{-z text} to the linker when linking a
16851 shared object. Using this option, you can link position-dependent
16852 code into a shared object.
16854 @option{-mimpure-text} suppresses the ``relocations remain against
16855 allocatable but non-writable sections'' linker error message.
16856 However, the necessary relocations will trigger copy-on-write, and the
16857 shared object is not actually shared across processes. Instead of
16858 using @option{-mimpure-text}, you should compile all source code with
16859 @option{-fpic} or @option{-fPIC}.
16863 These switches are supported in addition to the above on Solaris 2:
16868 Add support for multithreading using the POSIX threads library. This
16869 option sets flags for both the preprocessor and linker. This option does
16870 not affect the thread safety of object code produced by the compiler or
16871 that of libraries supplied with it.
16875 This is a synonym for @option{-pthreads}.
16878 @node SPARC Options
16879 @subsection SPARC Options
16880 @cindex SPARC options
16882 These @samp{-m} options are supported on the SPARC:
16885 @item -mno-app-regs
16887 @opindex mno-app-regs
16889 Specify @option{-mapp-regs} to generate output using the global registers
16890 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16893 To be fully SVR4 ABI compliant at the cost of some performance loss,
16894 specify @option{-mno-app-regs}. You should compile libraries and system
16895 software with this option.
16898 @itemx -mhard-float
16900 @opindex mhard-float
16901 Generate output containing floating point instructions. This is the
16905 @itemx -msoft-float
16907 @opindex msoft-float
16908 Generate output containing library calls for floating point.
16909 @strong{Warning:} the requisite libraries are not available for all SPARC
16910 targets. Normally the facilities of the machine's usual C compiler are
16911 used, but this cannot be done directly in cross-compilation. You must make
16912 your own arrangements to provide suitable library functions for
16913 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16914 @samp{sparclite-*-*} do provide software floating point support.
16916 @option{-msoft-float} changes the calling convention in the output file;
16917 therefore, it is only useful if you compile @emph{all} of a program with
16918 this option. In particular, you need to compile @file{libgcc.a}, the
16919 library that comes with GCC, with @option{-msoft-float} in order for
16922 @item -mhard-quad-float
16923 @opindex mhard-quad-float
16924 Generate output containing quad-word (long double) floating point
16927 @item -msoft-quad-float
16928 @opindex msoft-quad-float
16929 Generate output containing library calls for quad-word (long double)
16930 floating point instructions. The functions called are those specified
16931 in the SPARC ABI@. This is the default.
16933 As of this writing, there are no SPARC implementations that have hardware
16934 support for the quad-word floating point instructions. They all invoke
16935 a trap handler for one of these instructions, and then the trap handler
16936 emulates the effect of the instruction. Because of the trap handler overhead,
16937 this is much slower than calling the ABI library routines. Thus the
16938 @option{-msoft-quad-float} option is the default.
16940 @item -mno-unaligned-doubles
16941 @itemx -munaligned-doubles
16942 @opindex mno-unaligned-doubles
16943 @opindex munaligned-doubles
16944 Assume that doubles have 8 byte alignment. This is the default.
16946 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16947 alignment only if they are contained in another type, or if they have an
16948 absolute address. Otherwise, it assumes they have 4 byte alignment.
16949 Specifying this option avoids some rare compatibility problems with code
16950 generated by other compilers. It is not the default because it results
16951 in a performance loss, especially for floating point code.
16953 @item -mno-faster-structs
16954 @itemx -mfaster-structs
16955 @opindex mno-faster-structs
16956 @opindex mfaster-structs
16957 With @option{-mfaster-structs}, the compiler assumes that structures
16958 should have 8 byte alignment. This enables the use of pairs of
16959 @code{ldd} and @code{std} instructions for copies in structure
16960 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16961 However, the use of this changed alignment directly violates the SPARC
16962 ABI@. Thus, it's intended only for use on targets where the developer
16963 acknowledges that their resulting code will not be directly in line with
16964 the rules of the ABI@.
16966 @item -mcpu=@var{cpu_type}
16968 Set the instruction set, register set, and instruction scheduling parameters
16969 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16970 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
16971 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
16972 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16973 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16975 Default instruction scheduling parameters are used for values that select
16976 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16977 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16979 Here is a list of each supported architecture and their supported
16984 v8: supersparc, hypersparc, leon
16985 sparclite: f930, f934, sparclite86x
16987 v9: ultrasparc, ultrasparc3, niagara, niagara2
16990 By default (unless configured otherwise), GCC generates code for the V7
16991 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16992 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16993 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16994 SPARCStation 1, 2, IPX etc.
16996 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16997 architecture. The only difference from V7 code is that the compiler emits
16998 the integer multiply and integer divide instructions which exist in SPARC-V8
16999 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17000 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17003 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17004 the SPARC architecture. This adds the integer multiply, integer divide step
17005 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17006 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17007 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17008 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17009 MB86934 chip, which is the more recent SPARClite with FPU@.
17011 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17012 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17013 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17014 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17015 optimizes it for the TEMIC SPARClet chip.
17017 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17018 architecture. This adds 64-bit integer and floating-point move instructions,
17019 3 additional floating-point condition code registers and conditional move
17020 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17021 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17022 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17023 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17024 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17025 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17026 additionally optimizes it for Sun UltraSPARC T2 chips.
17028 @item -mtune=@var{cpu_type}
17030 Set the instruction scheduling parameters for machine type
17031 @var{cpu_type}, but do not set the instruction set or register set that the
17032 option @option{-mcpu=@var{cpu_type}} would.
17034 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17035 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17036 that select a particular CPU implementation. Those are @samp{cypress},
17037 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17038 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17039 @samp{niagara}, and @samp{niagara2}.
17044 @opindex mno-v8plus
17045 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17046 difference from the V8 ABI is that the global and out registers are
17047 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17048 mode for all SPARC-V9 processors.
17054 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17055 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17058 These @samp{-m} options are supported in addition to the above
17059 on SPARC-V9 processors in 64-bit environments:
17062 @item -mlittle-endian
17063 @opindex mlittle-endian
17064 Generate code for a processor running in little-endian mode. It is only
17065 available for a few configurations and most notably not on Solaris and Linux.
17071 Generate code for a 32-bit or 64-bit environment.
17072 The 32-bit environment sets int, long and pointer to 32 bits.
17073 The 64-bit environment sets int to 32 bits and long and pointer
17076 @item -mcmodel=medlow
17077 @opindex mcmodel=medlow
17078 Generate code for the Medium/Low code model: 64-bit addresses, programs
17079 must be linked in the low 32 bits of memory. Programs can be statically
17080 or dynamically linked.
17082 @item -mcmodel=medmid
17083 @opindex mcmodel=medmid
17084 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17085 must be linked in the low 44 bits of memory, the text and data segments must
17086 be less than 2GB in size and the data segment must be located within 2GB of
17089 @item -mcmodel=medany
17090 @opindex mcmodel=medany
17091 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17092 may be linked anywhere in memory, the text and data segments must be less
17093 than 2GB in size and the data segment must be located within 2GB of the
17096 @item -mcmodel=embmedany
17097 @opindex mcmodel=embmedany
17098 Generate code for the Medium/Anywhere code model for embedded systems:
17099 64-bit addresses, the text and data segments must be less than 2GB in
17100 size, both starting anywhere in memory (determined at link time). The
17101 global register %g4 points to the base of the data segment. Programs
17102 are statically linked and PIC is not supported.
17105 @itemx -mno-stack-bias
17106 @opindex mstack-bias
17107 @opindex mno-stack-bias
17108 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17109 frame pointer if present, are offset by @minus{}2047 which must be added back
17110 when making stack frame references. This is the default in 64-bit mode.
17111 Otherwise, assume no such offset is present.
17115 @subsection SPU Options
17116 @cindex SPU options
17118 These @samp{-m} options are supported on the SPU:
17122 @itemx -merror-reloc
17123 @opindex mwarn-reloc
17124 @opindex merror-reloc
17126 The loader for SPU does not handle dynamic relocations. By default, GCC
17127 will give an error when it generates code that requires a dynamic
17128 relocation. @option{-mno-error-reloc} disables the error,
17129 @option{-mwarn-reloc} will generate a warning instead.
17132 @itemx -munsafe-dma
17134 @opindex munsafe-dma
17136 Instructions which initiate or test completion of DMA must not be
17137 reordered with respect to loads and stores of the memory which is being
17138 accessed. Users typically address this problem using the volatile
17139 keyword, but that can lead to inefficient code in places where the
17140 memory is known to not change. Rather than mark the memory as volatile
17141 we treat the DMA instructions as potentially effecting all memory. With
17142 @option{-munsafe-dma} users must use the volatile keyword to protect
17145 @item -mbranch-hints
17146 @opindex mbranch-hints
17148 By default, GCC will generate a branch hint instruction to avoid
17149 pipeline stalls for always taken or probably taken branches. A hint
17150 will not be generated closer than 8 instructions away from its branch.
17151 There is little reason to disable them, except for debugging purposes,
17152 or to make an object a little bit smaller.
17156 @opindex msmall-mem
17157 @opindex mlarge-mem
17159 By default, GCC generates code assuming that addresses are never larger
17160 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17161 a full 32 bit address.
17166 By default, GCC links against startup code that assumes the SPU-style
17167 main function interface (which has an unconventional parameter list).
17168 With @option{-mstdmain}, GCC will link your program against startup
17169 code that assumes a C99-style interface to @code{main}, including a
17170 local copy of @code{argv} strings.
17172 @item -mfixed-range=@var{register-range}
17173 @opindex mfixed-range
17174 Generate code treating the given register range as fixed registers.
17175 A fixed register is one that the register allocator can not use. This is
17176 useful when compiling kernel code. A register range is specified as
17177 two registers separated by a dash. Multiple register ranges can be
17178 specified separated by a comma.
17184 Compile code assuming that pointers to the PPU address space accessed
17185 via the @code{__ea} named address space qualifier are either 32 or 64
17186 bits wide. The default is 32 bits. As this is an ABI changing option,
17187 all object code in an executable must be compiled with the same setting.
17189 @item -maddress-space-conversion
17190 @itemx -mno-address-space-conversion
17191 @opindex maddress-space-conversion
17192 @opindex mno-address-space-conversion
17193 Allow/disallow treating the @code{__ea} address space as superset
17194 of the generic address space. This enables explicit type casts
17195 between @code{__ea} and generic pointer as well as implicit
17196 conversions of generic pointers to @code{__ea} pointers. The
17197 default is to allow address space pointer conversions.
17199 @item -mcache-size=@var{cache-size}
17200 @opindex mcache-size
17201 This option controls the version of libgcc that the compiler links to an
17202 executable and selects a software-managed cache for accessing variables
17203 in the @code{__ea} address space with a particular cache size. Possible
17204 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17205 and @samp{128}. The default cache size is 64KB.
17207 @item -matomic-updates
17208 @itemx -mno-atomic-updates
17209 @opindex matomic-updates
17210 @opindex mno-atomic-updates
17211 This option controls the version of libgcc that the compiler links to an
17212 executable and selects whether atomic updates to the software-managed
17213 cache of PPU-side variables are used. If you use atomic updates, changes
17214 to a PPU variable from SPU code using the @code{__ea} named address space
17215 qualifier will not interfere with changes to other PPU variables residing
17216 in the same cache line from PPU code. If you do not use atomic updates,
17217 such interference may occur; however, writing back cache lines will be
17218 more efficient. The default behavior is to use atomic updates.
17221 @itemx -mdual-nops=@var{n}
17222 @opindex mdual-nops
17223 By default, GCC will insert nops to increase dual issue when it expects
17224 it to increase performance. @var{n} can be a value from 0 to 10. A
17225 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17226 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17228 @item -mhint-max-nops=@var{n}
17229 @opindex mhint-max-nops
17230 Maximum number of nops to insert for a branch hint. A branch hint must
17231 be at least 8 instructions away from the branch it is effecting. GCC
17232 will insert up to @var{n} nops to enforce this, otherwise it will not
17233 generate the branch hint.
17235 @item -mhint-max-distance=@var{n}
17236 @opindex mhint-max-distance
17237 The encoding of the branch hint instruction limits the hint to be within
17238 256 instructions of the branch it is effecting. By default, GCC makes
17239 sure it is within 125.
17242 @opindex msafe-hints
17243 Work around a hardware bug which causes the SPU to stall indefinitely.
17244 By default, GCC will insert the @code{hbrp} instruction to make sure
17245 this stall won't happen.
17249 @node System V Options
17250 @subsection Options for System V
17252 These additional options are available on System V Release 4 for
17253 compatibility with other compilers on those systems:
17258 Create a shared object.
17259 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17263 Identify the versions of each tool used by the compiler, in a
17264 @code{.ident} assembler directive in the output.
17268 Refrain from adding @code{.ident} directives to the output file (this is
17271 @item -YP,@var{dirs}
17273 Search the directories @var{dirs}, and no others, for libraries
17274 specified with @option{-l}.
17276 @item -Ym,@var{dir}
17278 Look in the directory @var{dir} to find the M4 preprocessor.
17279 The assembler uses this option.
17280 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17281 @c the generic assembler that comes with Solaris takes just -Ym.
17285 @subsection V850 Options
17286 @cindex V850 Options
17288 These @samp{-m} options are defined for V850 implementations:
17292 @itemx -mno-long-calls
17293 @opindex mlong-calls
17294 @opindex mno-long-calls
17295 Treat all calls as being far away (near). If calls are assumed to be
17296 far away, the compiler will always load the functions address up into a
17297 register, and call indirect through the pointer.
17303 Do not optimize (do optimize) basic blocks that use the same index
17304 pointer 4 or more times to copy pointer into the @code{ep} register, and
17305 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17306 option is on by default if you optimize.
17308 @item -mno-prolog-function
17309 @itemx -mprolog-function
17310 @opindex mno-prolog-function
17311 @opindex mprolog-function
17312 Do not use (do use) external functions to save and restore registers
17313 at the prologue and epilogue of a function. The external functions
17314 are slower, but use less code space if more than one function saves
17315 the same number of registers. The @option{-mprolog-function} option
17316 is on by default if you optimize.
17320 Try to make the code as small as possible. At present, this just turns
17321 on the @option{-mep} and @option{-mprolog-function} options.
17323 @item -mtda=@var{n}
17325 Put static or global variables whose size is @var{n} bytes or less into
17326 the tiny data area that register @code{ep} points to. The tiny data
17327 area can hold up to 256 bytes in total (128 bytes for byte references).
17329 @item -msda=@var{n}
17331 Put static or global variables whose size is @var{n} bytes or less into
17332 the small data area that register @code{gp} points to. The small data
17333 area can hold up to 64 kilobytes.
17335 @item -mzda=@var{n}
17337 Put static or global variables whose size is @var{n} bytes or less into
17338 the first 32 kilobytes of memory.
17342 Specify that the target processor is the V850.
17345 @opindex mbig-switch
17346 Generate code suitable for big switch tables. Use this option only if
17347 the assembler/linker complain about out of range branches within a switch
17352 This option will cause r2 and r5 to be used in the code generated by
17353 the compiler. This setting is the default.
17355 @item -mno-app-regs
17356 @opindex mno-app-regs
17357 This option will cause r2 and r5 to be treated as fixed registers.
17361 Specify that the target processor is the V850E2V3. The preprocessor
17362 constants @samp{__v850e2v3__} will be defined if
17363 this option is used.
17367 Specify that the target processor is the V850E2. The preprocessor
17368 constants @samp{__v850e2__} will be defined if this option is used.
17372 Specify that the target processor is the V850E1. The preprocessor
17373 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17374 this option is used.
17378 Specify that the target processor is the V850ES. This is an alias for
17379 the @option{-mv850e1} option.
17383 Specify that the target processor is the V850E@. The preprocessor
17384 constant @samp{__v850e__} will be defined if this option is used.
17386 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17387 nor @option{-mv850e2} nor @option{-mv850e2v3}
17388 are defined then a default target processor will be chosen and the
17389 relevant @samp{__v850*__} preprocessor constant will be defined.
17391 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17392 defined, regardless of which processor variant is the target.
17394 @item -mdisable-callt
17395 @opindex mdisable-callt
17396 This option will suppress generation of the CALLT instruction for the
17397 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17398 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17403 @subsection VAX Options
17404 @cindex VAX options
17406 These @samp{-m} options are defined for the VAX:
17411 Do not output certain jump instructions (@code{aobleq} and so on)
17412 that the Unix assembler for the VAX cannot handle across long
17417 Do output those jump instructions, on the assumption that you
17418 will assemble with the GNU assembler.
17422 Output code for g-format floating point numbers instead of d-format.
17425 @node VxWorks Options
17426 @subsection VxWorks Options
17427 @cindex VxWorks Options
17429 The options in this section are defined for all VxWorks targets.
17430 Options specific to the target hardware are listed with the other
17431 options for that target.
17436 GCC can generate code for both VxWorks kernels and real time processes
17437 (RTPs). This option switches from the former to the latter. It also
17438 defines the preprocessor macro @code{__RTP__}.
17441 @opindex non-static
17442 Link an RTP executable against shared libraries rather than static
17443 libraries. The options @option{-static} and @option{-shared} can
17444 also be used for RTPs (@pxref{Link Options}); @option{-static}
17451 These options are passed down to the linker. They are defined for
17452 compatibility with Diab.
17455 @opindex Xbind-lazy
17456 Enable lazy binding of function calls. This option is equivalent to
17457 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17461 Disable lazy binding of function calls. This option is the default and
17462 is defined for compatibility with Diab.
17465 @node x86-64 Options
17466 @subsection x86-64 Options
17467 @cindex x86-64 options
17469 These are listed under @xref{i386 and x86-64 Options}.
17471 @node Xstormy16 Options
17472 @subsection Xstormy16 Options
17473 @cindex Xstormy16 Options
17475 These options are defined for Xstormy16:
17480 Choose startup files and linker script suitable for the simulator.
17483 @node Xtensa Options
17484 @subsection Xtensa Options
17485 @cindex Xtensa Options
17487 These options are supported for Xtensa targets:
17491 @itemx -mno-const16
17493 @opindex mno-const16
17494 Enable or disable use of @code{CONST16} instructions for loading
17495 constant values. The @code{CONST16} instruction is currently not a
17496 standard option from Tensilica. When enabled, @code{CONST16}
17497 instructions are always used in place of the standard @code{L32R}
17498 instructions. The use of @code{CONST16} is enabled by default only if
17499 the @code{L32R} instruction is not available.
17502 @itemx -mno-fused-madd
17503 @opindex mfused-madd
17504 @opindex mno-fused-madd
17505 Enable or disable use of fused multiply/add and multiply/subtract
17506 instructions in the floating-point option. This has no effect if the
17507 floating-point option is not also enabled. Disabling fused multiply/add
17508 and multiply/subtract instructions forces the compiler to use separate
17509 instructions for the multiply and add/subtract operations. This may be
17510 desirable in some cases where strict IEEE 754-compliant results are
17511 required: the fused multiply add/subtract instructions do not round the
17512 intermediate result, thereby producing results with @emph{more} bits of
17513 precision than specified by the IEEE standard. Disabling fused multiply
17514 add/subtract instructions also ensures that the program output is not
17515 sensitive to the compiler's ability to combine multiply and add/subtract
17518 @item -mserialize-volatile
17519 @itemx -mno-serialize-volatile
17520 @opindex mserialize-volatile
17521 @opindex mno-serialize-volatile
17522 When this option is enabled, GCC inserts @code{MEMW} instructions before
17523 @code{volatile} memory references to guarantee sequential consistency.
17524 The default is @option{-mserialize-volatile}. Use
17525 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17527 @item -mforce-no-pic
17528 @opindex mforce-no-pic
17529 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17530 position-independent code (PIC), this option disables PIC for compiling
17533 @item -mtext-section-literals
17534 @itemx -mno-text-section-literals
17535 @opindex mtext-section-literals
17536 @opindex mno-text-section-literals
17537 Control the treatment of literal pools. The default is
17538 @option{-mno-text-section-literals}, which places literals in a separate
17539 section in the output file. This allows the literal pool to be placed
17540 in a data RAM/ROM, and it also allows the linker to combine literal
17541 pools from separate object files to remove redundant literals and
17542 improve code size. With @option{-mtext-section-literals}, the literals
17543 are interspersed in the text section in order to keep them as close as
17544 possible to their references. This may be necessary for large assembly
17547 @item -mtarget-align
17548 @itemx -mno-target-align
17549 @opindex mtarget-align
17550 @opindex mno-target-align
17551 When this option is enabled, GCC instructs the assembler to
17552 automatically align instructions to reduce branch penalties at the
17553 expense of some code density. The assembler attempts to widen density
17554 instructions to align branch targets and the instructions following call
17555 instructions. If there are not enough preceding safe density
17556 instructions to align a target, no widening will be performed. The
17557 default is @option{-mtarget-align}. These options do not affect the
17558 treatment of auto-aligned instructions like @code{LOOP}, which the
17559 assembler will always align, either by widening density instructions or
17560 by inserting no-op instructions.
17563 @itemx -mno-longcalls
17564 @opindex mlongcalls
17565 @opindex mno-longcalls
17566 When this option is enabled, GCC instructs the assembler to translate
17567 direct calls to indirect calls unless it can determine that the target
17568 of a direct call is in the range allowed by the call instruction. This
17569 translation typically occurs for calls to functions in other source
17570 files. Specifically, the assembler translates a direct @code{CALL}
17571 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17572 The default is @option{-mno-longcalls}. This option should be used in
17573 programs where the call target can potentially be out of range. This
17574 option is implemented in the assembler, not the compiler, so the
17575 assembly code generated by GCC will still show direct call
17576 instructions---look at the disassembled object code to see the actual
17577 instructions. Note that the assembler will use an indirect call for
17578 every cross-file call, not just those that really will be out of range.
17581 @node zSeries Options
17582 @subsection zSeries Options
17583 @cindex zSeries options
17585 These are listed under @xref{S/390 and zSeries Options}.
17587 @node Code Gen Options
17588 @section Options for Code Generation Conventions
17589 @cindex code generation conventions
17590 @cindex options, code generation
17591 @cindex run-time options
17593 These machine-independent options control the interface conventions
17594 used in code generation.
17596 Most of them have both positive and negative forms; the negative form
17597 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17598 one of the forms is listed---the one which is not the default. You
17599 can figure out the other form by either removing @samp{no-} or adding
17603 @item -fbounds-check
17604 @opindex fbounds-check
17605 For front-ends that support it, generate additional code to check that
17606 indices used to access arrays are within the declared range. This is
17607 currently only supported by the Java and Fortran front-ends, where
17608 this option defaults to true and false respectively.
17612 This option generates traps for signed overflow on addition, subtraction,
17613 multiplication operations.
17617 This option instructs the compiler to assume that signed arithmetic
17618 overflow of addition, subtraction and multiplication wraps around
17619 using twos-complement representation. This flag enables some optimizations
17620 and disables others. This option is enabled by default for the Java
17621 front-end, as required by the Java language specification.
17624 @opindex fexceptions
17625 Enable exception handling. Generates extra code needed to propagate
17626 exceptions. For some targets, this implies GCC will generate frame
17627 unwind information for all functions, which can produce significant data
17628 size overhead, although it does not affect execution. If you do not
17629 specify this option, GCC will enable it by default for languages like
17630 C++ which normally require exception handling, and disable it for
17631 languages like C that do not normally require it. However, you may need
17632 to enable this option when compiling C code that needs to interoperate
17633 properly with exception handlers written in C++. You may also wish to
17634 disable this option if you are compiling older C++ programs that don't
17635 use exception handling.
17637 @item -fnon-call-exceptions
17638 @opindex fnon-call-exceptions
17639 Generate code that allows trapping instructions to throw exceptions.
17640 Note that this requires platform-specific runtime support that does
17641 not exist everywhere. Moreover, it only allows @emph{trapping}
17642 instructions to throw exceptions, i.e.@: memory references or floating
17643 point instructions. It does not allow exceptions to be thrown from
17644 arbitrary signal handlers such as @code{SIGALRM}.
17646 @item -funwind-tables
17647 @opindex funwind-tables
17648 Similar to @option{-fexceptions}, except that it will just generate any needed
17649 static data, but will not affect the generated code in any other way.
17650 You will normally not enable this option; instead, a language processor
17651 that needs this handling would enable it on your behalf.
17653 @item -fasynchronous-unwind-tables
17654 @opindex fasynchronous-unwind-tables
17655 Generate unwind table in dwarf2 format, if supported by target machine. The
17656 table is exact at each instruction boundary, so it can be used for stack
17657 unwinding from asynchronous events (such as debugger or garbage collector).
17659 @item -fpcc-struct-return
17660 @opindex fpcc-struct-return
17661 Return ``short'' @code{struct} and @code{union} values in memory like
17662 longer ones, rather than in registers. This convention is less
17663 efficient, but it has the advantage of allowing intercallability between
17664 GCC-compiled files and files compiled with other compilers, particularly
17665 the Portable C Compiler (pcc).
17667 The precise convention for returning structures in memory depends
17668 on the target configuration macros.
17670 Short structures and unions are those whose size and alignment match
17671 that of some integer type.
17673 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17674 switch is not binary compatible with code compiled with the
17675 @option{-freg-struct-return} switch.
17676 Use it to conform to a non-default application binary interface.
17678 @item -freg-struct-return
17679 @opindex freg-struct-return
17680 Return @code{struct} and @code{union} values in registers when possible.
17681 This is more efficient for small structures than
17682 @option{-fpcc-struct-return}.
17684 If you specify neither @option{-fpcc-struct-return} nor
17685 @option{-freg-struct-return}, GCC defaults to whichever convention is
17686 standard for the target. If there is no standard convention, GCC
17687 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17688 the principal compiler. In those cases, we can choose the standard, and
17689 we chose the more efficient register return alternative.
17691 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17692 switch is not binary compatible with code compiled with the
17693 @option{-fpcc-struct-return} switch.
17694 Use it to conform to a non-default application binary interface.
17696 @item -fshort-enums
17697 @opindex fshort-enums
17698 Allocate to an @code{enum} type only as many bytes as it needs for the
17699 declared range of possible values. Specifically, the @code{enum} type
17700 will be equivalent to the smallest integer type which has enough room.
17702 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17703 code that is not binary compatible with code generated without that switch.
17704 Use it to conform to a non-default application binary interface.
17706 @item -fshort-double
17707 @opindex fshort-double
17708 Use the same size for @code{double} as for @code{float}.
17710 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17711 code that is not binary compatible with code generated without that switch.
17712 Use it to conform to a non-default application binary interface.
17714 @item -fshort-wchar
17715 @opindex fshort-wchar
17716 Override the underlying type for @samp{wchar_t} to be @samp{short
17717 unsigned int} instead of the default for the target. This option is
17718 useful for building programs to run under WINE@.
17720 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17721 code that is not binary compatible with code generated without that switch.
17722 Use it to conform to a non-default application binary interface.
17725 @opindex fno-common
17726 In C code, controls the placement of uninitialized global variables.
17727 Unix C compilers have traditionally permitted multiple definitions of
17728 such variables in different compilation units by placing the variables
17730 This is the behavior specified by @option{-fcommon}, and is the default
17731 for GCC on most targets.
17732 On the other hand, this behavior is not required by ISO C, and on some
17733 targets may carry a speed or code size penalty on variable references.
17734 The @option{-fno-common} option specifies that the compiler should place
17735 uninitialized global variables in the data section of the object file,
17736 rather than generating them as common blocks.
17737 This has the effect that if the same variable is declared
17738 (without @code{extern}) in two different compilations,
17739 you will get a multiple-definition error when you link them.
17740 In this case, you must compile with @option{-fcommon} instead.
17741 Compiling with @option{-fno-common} is useful on targets for which
17742 it provides better performance, or if you wish to verify that the
17743 program will work on other systems which always treat uninitialized
17744 variable declarations this way.
17748 Ignore the @samp{#ident} directive.
17750 @item -finhibit-size-directive
17751 @opindex finhibit-size-directive
17752 Don't output a @code{.size} assembler directive, or anything else that
17753 would cause trouble if the function is split in the middle, and the
17754 two halves are placed at locations far apart in memory. This option is
17755 used when compiling @file{crtstuff.c}; you should not need to use it
17758 @item -fverbose-asm
17759 @opindex fverbose-asm
17760 Put extra commentary information in the generated assembly code to
17761 make it more readable. This option is generally only of use to those
17762 who actually need to read the generated assembly code (perhaps while
17763 debugging the compiler itself).
17765 @option{-fno-verbose-asm}, the default, causes the
17766 extra information to be omitted and is useful when comparing two assembler
17769 @item -frecord-gcc-switches
17770 @opindex frecord-gcc-switches
17771 This switch causes the command line that was used to invoke the
17772 compiler to be recorded into the object file that is being created.
17773 This switch is only implemented on some targets and the exact format
17774 of the recording is target and binary file format dependent, but it
17775 usually takes the form of a section containing ASCII text. This
17776 switch is related to the @option{-fverbose-asm} switch, but that
17777 switch only records information in the assembler output file as
17778 comments, so it never reaches the object file.
17782 @cindex global offset table
17784 Generate position-independent code (PIC) suitable for use in a shared
17785 library, if supported for the target machine. Such code accesses all
17786 constant addresses through a global offset table (GOT)@. The dynamic
17787 loader resolves the GOT entries when the program starts (the dynamic
17788 loader is not part of GCC; it is part of the operating system). If
17789 the GOT size for the linked executable exceeds a machine-specific
17790 maximum size, you get an error message from the linker indicating that
17791 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17792 instead. (These maximums are 8k on the SPARC and 32k
17793 on the m68k and RS/6000. The 386 has no such limit.)
17795 Position-independent code requires special support, and therefore works
17796 only on certain machines. For the 386, GCC supports PIC for System V
17797 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17798 position-independent.
17800 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17805 If supported for the target machine, emit position-independent code,
17806 suitable for dynamic linking and avoiding any limit on the size of the
17807 global offset table. This option makes a difference on the m68k,
17808 PowerPC and SPARC@.
17810 Position-independent code requires special support, and therefore works
17811 only on certain machines.
17813 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17820 These options are similar to @option{-fpic} and @option{-fPIC}, but
17821 generated position independent code can be only linked into executables.
17822 Usually these options are used when @option{-pie} GCC option will be
17823 used during linking.
17825 @option{-fpie} and @option{-fPIE} both define the macros
17826 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17827 for @option{-fpie} and 2 for @option{-fPIE}.
17829 @item -fno-jump-tables
17830 @opindex fno-jump-tables
17831 Do not use jump tables for switch statements even where it would be
17832 more efficient than other code generation strategies. This option is
17833 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17834 building code which forms part of a dynamic linker and cannot
17835 reference the address of a jump table. On some targets, jump tables
17836 do not require a GOT and this option is not needed.
17838 @item -ffixed-@var{reg}
17840 Treat the register named @var{reg} as a fixed register; generated code
17841 should never refer to it (except perhaps as a stack pointer, frame
17842 pointer or in some other fixed role).
17844 @var{reg} must be the name of a register. The register names accepted
17845 are machine-specific and are defined in the @code{REGISTER_NAMES}
17846 macro in the machine description macro file.
17848 This flag does not have a negative form, because it specifies a
17851 @item -fcall-used-@var{reg}
17852 @opindex fcall-used
17853 Treat the register named @var{reg} as an allocable register that is
17854 clobbered by function calls. It may be allocated for temporaries or
17855 variables that do not live across a call. Functions compiled this way
17856 will not save and restore the register @var{reg}.
17858 It is an error to used this flag with the frame pointer or stack pointer.
17859 Use of this flag for other registers that have fixed pervasive roles in
17860 the machine's execution model will produce disastrous results.
17862 This flag does not have a negative form, because it specifies a
17865 @item -fcall-saved-@var{reg}
17866 @opindex fcall-saved
17867 Treat the register named @var{reg} as an allocable register saved by
17868 functions. It may be allocated even for temporaries or variables that
17869 live across a call. Functions compiled this way will save and restore
17870 the register @var{reg} if they use it.
17872 It is an error to used this flag with the frame pointer or stack pointer.
17873 Use of this flag for other registers that have fixed pervasive roles in
17874 the machine's execution model will produce disastrous results.
17876 A different sort of disaster will result from the use of this flag for
17877 a register in which function values may be returned.
17879 This flag does not have a negative form, because it specifies a
17882 @item -fpack-struct[=@var{n}]
17883 @opindex fpack-struct
17884 Without a value specified, pack all structure members together without
17885 holes. When a value is specified (which must be a small power of two), pack
17886 structure members according to this value, representing the maximum
17887 alignment (that is, objects with default alignment requirements larger than
17888 this will be output potentially unaligned at the next fitting location.
17890 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17891 code that is not binary compatible with code generated without that switch.
17892 Additionally, it makes the code suboptimal.
17893 Use it to conform to a non-default application binary interface.
17895 @item -finstrument-functions
17896 @opindex finstrument-functions
17897 Generate instrumentation calls for entry and exit to functions. Just
17898 after function entry and just before function exit, the following
17899 profiling functions will be called with the address of the current
17900 function and its call site. (On some platforms,
17901 @code{__builtin_return_address} does not work beyond the current
17902 function, so the call site information may not be available to the
17903 profiling functions otherwise.)
17906 void __cyg_profile_func_enter (void *this_fn,
17908 void __cyg_profile_func_exit (void *this_fn,
17912 The first argument is the address of the start of the current function,
17913 which may be looked up exactly in the symbol table.
17915 This instrumentation is also done for functions expanded inline in other
17916 functions. The profiling calls will indicate where, conceptually, the
17917 inline function is entered and exited. This means that addressable
17918 versions of such functions must be available. If all your uses of a
17919 function are expanded inline, this may mean an additional expansion of
17920 code size. If you use @samp{extern inline} in your C code, an
17921 addressable version of such functions must be provided. (This is
17922 normally the case anyways, but if you get lucky and the optimizer always
17923 expands the functions inline, you might have gotten away without
17924 providing static copies.)
17926 A function may be given the attribute @code{no_instrument_function}, in
17927 which case this instrumentation will not be done. This can be used, for
17928 example, for the profiling functions listed above, high-priority
17929 interrupt routines, and any functions from which the profiling functions
17930 cannot safely be called (perhaps signal handlers, if the profiling
17931 routines generate output or allocate memory).
17933 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17934 @opindex finstrument-functions-exclude-file-list
17936 Set the list of functions that are excluded from instrumentation (see
17937 the description of @code{-finstrument-functions}). If the file that
17938 contains a function definition matches with one of @var{file}, then
17939 that function is not instrumented. The match is done on substrings:
17940 if the @var{file} parameter is a substring of the file name, it is
17941 considered to be a match.
17946 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
17950 will exclude any inline function defined in files whose pathnames
17951 contain @code{/bits/stl} or @code{include/sys}.
17953 If, for some reason, you want to include letter @code{','} in one of
17954 @var{sym}, write @code{'\,'}. For example,
17955 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17956 (note the single quote surrounding the option).
17958 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17959 @opindex finstrument-functions-exclude-function-list
17961 This is similar to @code{-finstrument-functions-exclude-file-list},
17962 but this option sets the list of function names to be excluded from
17963 instrumentation. The function name to be matched is its user-visible
17964 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17965 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17966 match is done on substrings: if the @var{sym} parameter is a substring
17967 of the function name, it is considered to be a match. For C99 and C++
17968 extended identifiers, the function name must be given in UTF-8, not
17969 using universal character names.
17971 @item -fstack-check
17972 @opindex fstack-check
17973 Generate code to verify that you do not go beyond the boundary of the
17974 stack. You should specify this flag if you are running in an
17975 environment with multiple threads, but only rarely need to specify it in
17976 a single-threaded environment since stack overflow is automatically
17977 detected on nearly all systems if there is only one stack.
17979 Note that this switch does not actually cause checking to be done; the
17980 operating system or the language runtime must do that. The switch causes
17981 generation of code to ensure that they see the stack being extended.
17983 You can additionally specify a string parameter: @code{no} means no
17984 checking, @code{generic} means force the use of old-style checking,
17985 @code{specific} means use the best checking method and is equivalent
17986 to bare @option{-fstack-check}.
17988 Old-style checking is a generic mechanism that requires no specific
17989 target support in the compiler but comes with the following drawbacks:
17993 Modified allocation strategy for large objects: they will always be
17994 allocated dynamically if their size exceeds a fixed threshold.
17997 Fixed limit on the size of the static frame of functions: when it is
17998 topped by a particular function, stack checking is not reliable and
17999 a warning is issued by the compiler.
18002 Inefficiency: because of both the modified allocation strategy and the
18003 generic implementation, the performances of the code are hampered.
18006 Note that old-style stack checking is also the fallback method for
18007 @code{specific} if no target support has been added in the compiler.
18009 @item -fstack-limit-register=@var{reg}
18010 @itemx -fstack-limit-symbol=@var{sym}
18011 @itemx -fno-stack-limit
18012 @opindex fstack-limit-register
18013 @opindex fstack-limit-symbol
18014 @opindex fno-stack-limit
18015 Generate code to ensure that the stack does not grow beyond a certain value,
18016 either the value of a register or the address of a symbol. If the stack
18017 would grow beyond the value, a signal is raised. For most targets,
18018 the signal is raised before the stack overruns the boundary, so
18019 it is possible to catch the signal without taking special precautions.
18021 For instance, if the stack starts at absolute address @samp{0x80000000}
18022 and grows downwards, you can use the flags
18023 @option{-fstack-limit-symbol=__stack_limit} and
18024 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18025 of 128KB@. Note that this may only work with the GNU linker.
18027 @item -fsplit-stack
18028 @opindex fsplit-stack
18029 Generate code to automatically split the stack before it overflows.
18030 The resulting program has a discontiguous stack which can only
18031 overflow if the program is unable to allocate any more memory. This
18032 is most useful when running threaded programs, as it is no longer
18033 necessary to calculate a good stack size to use for each thread. This
18034 is currently only implemented for the i386 and x86_64 backends running
18037 When code compiled with @option{-fsplit-stack} calls code compiled
18038 without @option{-fsplit-stack}, there may not be much stack space
18039 available for the latter code to run. If compiling all code,
18040 including library code, with @option{-fsplit-stack} is not an option,
18041 then the linker can fix up these calls so that the code compiled
18042 without @option{-fsplit-stack} always has a large stack. Support for
18043 this is implemented in the gold linker in GNU binutils release 2.21
18046 @item -fleading-underscore
18047 @opindex fleading-underscore
18048 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18049 change the way C symbols are represented in the object file. One use
18050 is to help link with legacy assembly code.
18052 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18053 generate code that is not binary compatible with code generated without that
18054 switch. Use it to conform to a non-default application binary interface.
18055 Not all targets provide complete support for this switch.
18057 @item -ftls-model=@var{model}
18058 @opindex ftls-model
18059 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18060 The @var{model} argument should be one of @code{global-dynamic},
18061 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18063 The default without @option{-fpic} is @code{initial-exec}; with
18064 @option{-fpic} the default is @code{global-dynamic}.
18066 @item -fvisibility=@var{default|internal|hidden|protected}
18067 @opindex fvisibility
18068 Set the default ELF image symbol visibility to the specified option---all
18069 symbols will be marked with this unless overridden within the code.
18070 Using this feature can very substantially improve linking and
18071 load times of shared object libraries, produce more optimized
18072 code, provide near-perfect API export and prevent symbol clashes.
18073 It is @strong{strongly} recommended that you use this in any shared objects
18076 Despite the nomenclature, @code{default} always means public; i.e.,
18077 available to be linked against from outside the shared object.
18078 @code{protected} and @code{internal} are pretty useless in real-world
18079 usage so the only other commonly used option will be @code{hidden}.
18080 The default if @option{-fvisibility} isn't specified is
18081 @code{default}, i.e., make every
18082 symbol public---this causes the same behavior as previous versions of
18085 A good explanation of the benefits offered by ensuring ELF
18086 symbols have the correct visibility is given by ``How To Write
18087 Shared Libraries'' by Ulrich Drepper (which can be found at
18088 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18089 solution made possible by this option to marking things hidden when
18090 the default is public is to make the default hidden and mark things
18091 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18092 and @code{__attribute__ ((visibility("default")))} instead of
18093 @code{__declspec(dllexport)} you get almost identical semantics with
18094 identical syntax. This is a great boon to those working with
18095 cross-platform projects.
18097 For those adding visibility support to existing code, you may find
18098 @samp{#pragma GCC visibility} of use. This works by you enclosing
18099 the declarations you wish to set visibility for with (for example)
18100 @samp{#pragma GCC visibility push(hidden)} and
18101 @samp{#pragma GCC visibility pop}.
18102 Bear in mind that symbol visibility should be viewed @strong{as
18103 part of the API interface contract} and thus all new code should
18104 always specify visibility when it is not the default; i.e., declarations
18105 only for use within the local DSO should @strong{always} be marked explicitly
18106 as hidden as so to avoid PLT indirection overheads---making this
18107 abundantly clear also aids readability and self-documentation of the code.
18108 Note that due to ISO C++ specification requirements, operator new and
18109 operator delete must always be of default visibility.
18111 Be aware that headers from outside your project, in particular system
18112 headers and headers from any other library you use, may not be
18113 expecting to be compiled with visibility other than the default. You
18114 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18115 before including any such headers.
18117 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18118 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18119 no modifications. However, this means that calls to @samp{extern}
18120 functions with no explicit visibility will use the PLT, so it is more
18121 effective to use @samp{__attribute ((visibility))} and/or
18122 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18123 declarations should be treated as hidden.
18125 Note that @samp{-fvisibility} does affect C++ vague linkage
18126 entities. This means that, for instance, an exception class that will
18127 be thrown between DSOs must be explicitly marked with default
18128 visibility so that the @samp{type_info} nodes will be unified between
18131 An overview of these techniques, their benefits and how to use them
18132 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18134 @item -fstrict-volatile-bitfields
18135 @opindex fstrict-volatile-bitfields
18136 This option should be used if accesses to volatile bitfields (or other
18137 structure fields, although the compiler usually honors those types
18138 anyway) should use a single access of the width of the
18139 field's type, aligned to a natural alignment if possible. For
18140 example, targets with memory-mapped peripheral registers might require
18141 all such accesses to be 16 bits wide; with this flag the user could
18142 declare all peripheral bitfields as ``unsigned short'' (assuming short
18143 is 16 bits on these targets) to force GCC to use 16 bit accesses
18144 instead of, perhaps, a more efficient 32 bit access.
18146 If this option is disabled, the compiler will use the most efficient
18147 instruction. In the previous example, that might be a 32-bit load
18148 instruction, even though that will access bytes that do not contain
18149 any portion of the bitfield, or memory-mapped registers unrelated to
18150 the one being updated.
18152 If the target requires strict alignment, and honoring the field
18153 type would require violating this alignment, a warning is issued.
18154 If the field has @code{packed} attribute, the access is done without
18155 honoring the field type. If the field doesn't have @code{packed}
18156 attribute, the access is done honoring the field type. In both cases,
18157 GCC assumes that the user knows something about the target hardware
18158 that it is unaware of.
18160 The default value of this option is determined by the application binary
18161 interface for the target processor.
18167 @node Environment Variables
18168 @section Environment Variables Affecting GCC
18169 @cindex environment variables
18171 @c man begin ENVIRONMENT
18172 This section describes several environment variables that affect how GCC
18173 operates. Some of them work by specifying directories or prefixes to use
18174 when searching for various kinds of files. Some are used to specify other
18175 aspects of the compilation environment.
18177 Note that you can also specify places to search using options such as
18178 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18179 take precedence over places specified using environment variables, which
18180 in turn take precedence over those specified by the configuration of GCC@.
18181 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18182 GNU Compiler Collection (GCC) Internals}.
18187 @c @itemx LC_COLLATE
18189 @c @itemx LC_MONETARY
18190 @c @itemx LC_NUMERIC
18195 @c @findex LC_COLLATE
18196 @findex LC_MESSAGES
18197 @c @findex LC_MONETARY
18198 @c @findex LC_NUMERIC
18202 These environment variables control the way that GCC uses
18203 localization information that allow GCC to work with different
18204 national conventions. GCC inspects the locale categories
18205 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18206 so. These locale categories can be set to any value supported by your
18207 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18208 Kingdom encoded in UTF-8.
18210 The @env{LC_CTYPE} environment variable specifies character
18211 classification. GCC uses it to determine the character boundaries in
18212 a string; this is needed for some multibyte encodings that contain quote
18213 and escape characters that would otherwise be interpreted as a string
18216 The @env{LC_MESSAGES} environment variable specifies the language to
18217 use in diagnostic messages.
18219 If the @env{LC_ALL} environment variable is set, it overrides the value
18220 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18221 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18222 environment variable. If none of these variables are set, GCC
18223 defaults to traditional C English behavior.
18227 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18228 files. GCC uses temporary files to hold the output of one stage of
18229 compilation which is to be used as input to the next stage: for example,
18230 the output of the preprocessor, which is the input to the compiler
18233 @item GCC_EXEC_PREFIX
18234 @findex GCC_EXEC_PREFIX
18235 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18236 names of the subprograms executed by the compiler. No slash is added
18237 when this prefix is combined with the name of a subprogram, but you can
18238 specify a prefix that ends with a slash if you wish.
18240 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18241 an appropriate prefix to use based on the pathname it was invoked with.
18243 If GCC cannot find the subprogram using the specified prefix, it
18244 tries looking in the usual places for the subprogram.
18246 The default value of @env{GCC_EXEC_PREFIX} is
18247 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18248 the installed compiler. In many cases @var{prefix} is the value
18249 of @code{prefix} when you ran the @file{configure} script.
18251 Other prefixes specified with @option{-B} take precedence over this prefix.
18253 This prefix is also used for finding files such as @file{crt0.o} that are
18256 In addition, the prefix is used in an unusual way in finding the
18257 directories to search for header files. For each of the standard
18258 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18259 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18260 replacing that beginning with the specified prefix to produce an
18261 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18262 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18263 These alternate directories are searched first; the standard directories
18264 come next. If a standard directory begins with the configured
18265 @var{prefix} then the value of @var{prefix} is replaced by
18266 @env{GCC_EXEC_PREFIX} when looking for header files.
18268 @item COMPILER_PATH
18269 @findex COMPILER_PATH
18270 The value of @env{COMPILER_PATH} is a colon-separated list of
18271 directories, much like @env{PATH}. GCC tries the directories thus
18272 specified when searching for subprograms, if it can't find the
18273 subprograms using @env{GCC_EXEC_PREFIX}.
18276 @findex LIBRARY_PATH
18277 The value of @env{LIBRARY_PATH} is a colon-separated list of
18278 directories, much like @env{PATH}. When configured as a native compiler,
18279 GCC tries the directories thus specified when searching for special
18280 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18281 using GCC also uses these directories when searching for ordinary
18282 libraries for the @option{-l} option (but directories specified with
18283 @option{-L} come first).
18287 @cindex locale definition
18288 This variable is used to pass locale information to the compiler. One way in
18289 which this information is used is to determine the character set to be used
18290 when character literals, string literals and comments are parsed in C and C++.
18291 When the compiler is configured to allow multibyte characters,
18292 the following values for @env{LANG} are recognized:
18296 Recognize JIS characters.
18298 Recognize SJIS characters.
18300 Recognize EUCJP characters.
18303 If @env{LANG} is not defined, or if it has some other value, then the
18304 compiler will use mblen and mbtowc as defined by the default locale to
18305 recognize and translate multibyte characters.
18309 Some additional environments variables affect the behavior of the
18312 @include cppenv.texi
18316 @node Precompiled Headers
18317 @section Using Precompiled Headers
18318 @cindex precompiled headers
18319 @cindex speed of compilation
18321 Often large projects have many header files that are included in every
18322 source file. The time the compiler takes to process these header files
18323 over and over again can account for nearly all of the time required to
18324 build the project. To make builds faster, GCC allows users to
18325 `precompile' a header file; then, if builds can use the precompiled
18326 header file they will be much faster.
18328 To create a precompiled header file, simply compile it as you would any
18329 other file, if necessary using the @option{-x} option to make the driver
18330 treat it as a C or C++ header file. You will probably want to use a
18331 tool like @command{make} to keep the precompiled header up-to-date when
18332 the headers it contains change.
18334 A precompiled header file will be searched for when @code{#include} is
18335 seen in the compilation. As it searches for the included file
18336 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18337 compiler looks for a precompiled header in each directory just before it
18338 looks for the include file in that directory. The name searched for is
18339 the name specified in the @code{#include} with @samp{.gch} appended. If
18340 the precompiled header file can't be used, it is ignored.
18342 For instance, if you have @code{#include "all.h"}, and you have
18343 @file{all.h.gch} in the same directory as @file{all.h}, then the
18344 precompiled header file will be used if possible, and the original
18345 header will be used otherwise.
18347 Alternatively, you might decide to put the precompiled header file in a
18348 directory and use @option{-I} to ensure that directory is searched
18349 before (or instead of) the directory containing the original header.
18350 Then, if you want to check that the precompiled header file is always
18351 used, you can put a file of the same name as the original header in this
18352 directory containing an @code{#error} command.
18354 This also works with @option{-include}. So yet another way to use
18355 precompiled headers, good for projects not designed with precompiled
18356 header files in mind, is to simply take most of the header files used by
18357 a project, include them from another header file, precompile that header
18358 file, and @option{-include} the precompiled header. If the header files
18359 have guards against multiple inclusion, they will be skipped because
18360 they've already been included (in the precompiled header).
18362 If you need to precompile the same header file for different
18363 languages, targets, or compiler options, you can instead make a
18364 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18365 header in the directory, perhaps using @option{-o}. It doesn't matter
18366 what you call the files in the directory, every precompiled header in
18367 the directory will be considered. The first precompiled header
18368 encountered in the directory that is valid for this compilation will
18369 be used; they're searched in no particular order.
18371 There are many other possibilities, limited only by your imagination,
18372 good sense, and the constraints of your build system.
18374 A precompiled header file can be used only when these conditions apply:
18378 Only one precompiled header can be used in a particular compilation.
18381 A precompiled header can't be used once the first C token is seen. You
18382 can have preprocessor directives before a precompiled header; you can
18383 even include a precompiled header from inside another header, so long as
18384 there are no C tokens before the @code{#include}.
18387 The precompiled header file must be produced for the same language as
18388 the current compilation. You can't use a C precompiled header for a C++
18392 The precompiled header file must have been produced by the same compiler
18393 binary as the current compilation is using.
18396 Any macros defined before the precompiled header is included must
18397 either be defined in the same way as when the precompiled header was
18398 generated, or must not affect the precompiled header, which usually
18399 means that they don't appear in the precompiled header at all.
18401 The @option{-D} option is one way to define a macro before a
18402 precompiled header is included; using a @code{#define} can also do it.
18403 There are also some options that define macros implicitly, like
18404 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18407 @item If debugging information is output when using the precompiled
18408 header, using @option{-g} or similar, the same kind of debugging information
18409 must have been output when building the precompiled header. However,
18410 a precompiled header built using @option{-g} can be used in a compilation
18411 when no debugging information is being output.
18413 @item The same @option{-m} options must generally be used when building
18414 and using the precompiled header. @xref{Submodel Options},
18415 for any cases where this rule is relaxed.
18417 @item Each of the following options must be the same when building and using
18418 the precompiled header:
18420 @gccoptlist{-fexceptions}
18423 Some other command-line options starting with @option{-f},
18424 @option{-p}, or @option{-O} must be defined in the same way as when
18425 the precompiled header was generated. At present, it's not clear
18426 which options are safe to change and which are not; the safest choice
18427 is to use exactly the same options when generating and using the
18428 precompiled header. The following are known to be safe:
18430 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18431 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18432 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18437 For all of these except the last, the compiler will automatically
18438 ignore the precompiled header if the conditions aren't met. If you
18439 find an option combination that doesn't work and doesn't cause the
18440 precompiled header to be ignored, please consider filing a bug report,
18443 If you do use differing options when generating and using the
18444 precompiled header, the actual behavior will be a mixture of the
18445 behavior for the options. For instance, if you use @option{-g} to
18446 generate the precompiled header but not when using it, you may or may
18447 not get debugging information for routines in the precompiled header.