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 -Wcpp -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 @gol
316 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
317 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
318 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
319 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
320 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
321 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
322 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
323 -gvms -gxcoff -gxcoff+ @gol
324 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
325 -fdebug-prefix-map=@var{old}=@var{new} @gol
326 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
327 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
328 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
329 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
330 -print-prog-name=@var{program} -print-search-dirs -Q @gol
331 -print-sysroot -print-sysroot-headers-suffix @gol
332 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
334 @item Optimization Options
335 @xref{Optimize Options,,Options that Control Optimization}.
336 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
337 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
338 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
339 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
340 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
341 -fcompare-elim -fcprop-registers -fcrossjumping @gol
342 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
343 -fcx-limited-range @gol
344 -fdata-sections -fdce -fdce -fdelayed-branch @gol
345 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
346 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
347 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
348 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
349 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
350 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
351 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
352 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
353 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
354 -fipa-struct-reorg -fira-algorithm=@var{algorithm} @gol
355 -fira-region=@var{region} @gol
356 -fira-loop-pressure -fno-ira-share-save-slots @gol
357 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
358 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
359 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
360 -floop-parallelize-all -flto -flto-compression-level @gol
361 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
362 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
363 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
364 -fno-default-inline @gol
365 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
366 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
367 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
368 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
369 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
370 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
371 -fprefetch-loop-arrays @gol
372 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
373 -fprofile-generate=@var{path} @gol
374 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
375 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
376 -freorder-blocks-and-partition -freorder-functions @gol
377 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
378 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
379 -fsched-spec-load -fsched-spec-load-dangerous @gol
380 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
381 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
382 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
383 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
384 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
385 -fselective-scheduling -fselective-scheduling2 @gol
386 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
387 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
388 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
389 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
391 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
392 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
393 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
394 -ftree-loop-if-convert-stores -ftree-loop-im @gol
395 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
396 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
397 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
398 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
399 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
400 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
401 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
402 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
403 -fwhole-program -fwpa -fuse-linker-plugin @gol
404 --param @var{name}=@var{value}
405 -O -O0 -O1 -O2 -O3 -Os -Ofast}
407 @item Preprocessor Options
408 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
409 @gccoptlist{-A@var{question}=@var{answer} @gol
410 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
411 -C -dD -dI -dM -dN @gol
412 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
413 -idirafter @var{dir} @gol
414 -include @var{file} -imacros @var{file} @gol
415 -iprefix @var{file} -iwithprefix @var{dir} @gol
416 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
417 -imultilib @var{dir} -isysroot @var{dir} @gol
418 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
419 -P -fworking-directory -remap @gol
420 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
421 -Xpreprocessor @var{option}}
423 @item Assembler Option
424 @xref{Assembler Options,,Passing Options to the Assembler}.
425 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
428 @xref{Link Options,,Options for Linking}.
429 @gccoptlist{@var{object-file-name} -l@var{library} @gol
430 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
431 -s -static -static-libgcc -static-libstdc++ -shared @gol
432 -shared-libgcc -symbolic @gol
433 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
436 @item Directory Options
437 @xref{Directory Options,,Options for Directory Search}.
438 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
439 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
442 @item Machine Dependent Options
443 @xref{Submodel Options,,Hardware Models and Configurations}.
444 @c This list is ordered alphanumerically by subsection name.
445 @c Try and put the significant identifier (CPU or system) first,
446 @c so users have a clue at guessing where the ones they want will be.
449 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
450 -mabi=@var{name} @gol
451 -mapcs-stack-check -mno-apcs-stack-check @gol
452 -mapcs-float -mno-apcs-float @gol
453 -mapcs-reentrant -mno-apcs-reentrant @gol
454 -msched-prolog -mno-sched-prolog @gol
455 -mlittle-endian -mbig-endian -mwords-little-endian @gol
456 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
457 -mfp16-format=@var{name}
458 -mthumb-interwork -mno-thumb-interwork @gol
459 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
460 -mstructure-size-boundary=@var{n} @gol
461 -mabort-on-noreturn @gol
462 -mlong-calls -mno-long-calls @gol
463 -msingle-pic-base -mno-single-pic-base @gol
464 -mpic-register=@var{reg} @gol
465 -mnop-fun-dllimport @gol
466 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
467 -mpoke-function-name @gol
469 -mtpcs-frame -mtpcs-leaf-frame @gol
470 -mcaller-super-interworking -mcallee-super-interworking @gol
472 -mword-relocations @gol
473 -mfix-cortex-m3-ldrd}
476 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
477 -mcall-prologues -mtiny-stack -mint8}
479 @emph{Blackfin Options}
480 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
481 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
482 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
483 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
484 -mno-id-shared-library -mshared-library-id=@var{n} @gol
485 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
486 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
487 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
491 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
492 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
493 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
494 -mstack-align -mdata-align -mconst-align @gol
495 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
496 -melf -maout -melinux -mlinux -sim -sim2 @gol
497 -mmul-bug-workaround -mno-mul-bug-workaround}
499 @emph{Darwin Options}
500 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
501 -arch_only -bind_at_load -bundle -bundle_loader @gol
502 -client_name -compatibility_version -current_version @gol
504 -dependency-file -dylib_file -dylinker_install_name @gol
505 -dynamic -dynamiclib -exported_symbols_list @gol
506 -filelist -flat_namespace -force_cpusubtype_ALL @gol
507 -force_flat_namespace -headerpad_max_install_names @gol
509 -image_base -init -install_name -keep_private_externs @gol
510 -multi_module -multiply_defined -multiply_defined_unused @gol
511 -noall_load -no_dead_strip_inits_and_terms @gol
512 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
513 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
514 -private_bundle -read_only_relocs -sectalign @gol
515 -sectobjectsymbols -whyload -seg1addr @gol
516 -sectcreate -sectobjectsymbols -sectorder @gol
517 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
518 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
519 -segprot -segs_read_only_addr -segs_read_write_addr @gol
520 -single_module -static -sub_library -sub_umbrella @gol
521 -twolevel_namespace -umbrella -undefined @gol
522 -unexported_symbols_list -weak_reference_mismatches @gol
523 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
524 -mkernel -mone-byte-bool}
526 @emph{DEC Alpha Options}
527 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
528 -mieee -mieee-with-inexact -mieee-conformant @gol
529 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
530 -mtrap-precision=@var{mode} -mbuild-constants @gol
531 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
532 -mbwx -mmax -mfix -mcix @gol
533 -mfloat-vax -mfloat-ieee @gol
534 -mexplicit-relocs -msmall-data -mlarge-data @gol
535 -msmall-text -mlarge-text @gol
536 -mmemory-latency=@var{time}}
538 @emph{DEC Alpha/VMS Options}
539 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
542 @gccoptlist{-msmall-model -mno-lsim}
545 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
546 -mhard-float -msoft-float @gol
547 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
548 -mdouble -mno-double @gol
549 -mmedia -mno-media -mmuladd -mno-muladd @gol
550 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
551 -mlinked-fp -mlong-calls -malign-labels @gol
552 -mlibrary-pic -macc-4 -macc-8 @gol
553 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
554 -moptimize-membar -mno-optimize-membar @gol
555 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
556 -mvliw-branch -mno-vliw-branch @gol
557 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
558 -mno-nested-cond-exec -mtomcat-stats @gol
562 @emph{GNU/Linux Options}
563 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
564 -tno-android-cc -tno-android-ld}
566 @emph{H8/300 Options}
567 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
570 @gccoptlist{-march=@var{architecture-type} @gol
571 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
572 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
573 -mfixed-range=@var{register-range} @gol
574 -mjump-in-delay -mlinker-opt -mlong-calls @gol
575 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
576 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
577 -mno-jump-in-delay -mno-long-load-store @gol
578 -mno-portable-runtime -mno-soft-float @gol
579 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
580 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
581 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
582 -munix=@var{unix-std} -nolibdld -static -threads}
584 @emph{i386 and x86-64 Options}
585 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
586 -mfpmath=@var{unit} @gol
587 -masm=@var{dialect} -mno-fancy-math-387 @gol
588 -mno-fp-ret-in-387 -msoft-float @gol
589 -mno-wide-multiply -mrtd -malign-double @gol
590 -mpreferred-stack-boundary=@var{num} @gol
591 -mincoming-stack-boundary=@var{num} @gol
592 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
593 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
594 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
595 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlwp @gol
596 -mthreads -mno-align-stringops -minline-all-stringops @gol
597 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
598 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
599 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
600 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
601 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
602 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
603 -mcmodel=@var{code-model} -mabi=@var{name} @gol
604 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
605 -msse2avx -mfentry -m8bit-idiv}
608 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
609 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
610 -mconstant-gp -mauto-pic -mfused-madd @gol
611 -minline-float-divide-min-latency @gol
612 -minline-float-divide-max-throughput @gol
613 -mno-inline-float-divide @gol
614 -minline-int-divide-min-latency @gol
615 -minline-int-divide-max-throughput @gol
616 -mno-inline-int-divide @gol
617 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
618 -mno-inline-sqrt @gol
619 -mdwarf2-asm -mearly-stop-bits @gol
620 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
621 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
622 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
623 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
624 -msched-spec-ldc -msched-spec-control-ldc @gol
625 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
626 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
627 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
628 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
630 @emph{IA-64/VMS Options}
631 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
634 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
635 -msign-extend-enabled -muser-enabled}
637 @emph{M32R/D Options}
638 @gccoptlist{-m32r2 -m32rx -m32r @gol
640 -malign-loops -mno-align-loops @gol
641 -missue-rate=@var{number} @gol
642 -mbranch-cost=@var{number} @gol
643 -mmodel=@var{code-size-model-type} @gol
644 -msdata=@var{sdata-type} @gol
645 -mno-flush-func -mflush-func=@var{name} @gol
646 -mno-flush-trap -mflush-trap=@var{number} @gol
650 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
652 @emph{M680x0 Options}
653 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
654 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
655 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
656 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
657 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
658 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
659 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
660 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
664 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
665 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
666 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
667 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
668 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
671 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
672 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
673 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
674 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
677 @emph{MicroBlaze Options}
678 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
679 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
680 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
681 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
682 -mxl-mode-@var{app-model}}
685 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
686 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
687 -mips64 -mips64r2 @gol
688 -mips16 -mno-mips16 -mflip-mips16 @gol
689 -minterlink-mips16 -mno-interlink-mips16 @gol
690 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
691 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
692 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
693 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
694 -mfpu=@var{fpu-type} @gol
695 -msmartmips -mno-smartmips @gol
696 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
697 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
698 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
699 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
700 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
701 -membedded-data -mno-embedded-data @gol
702 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
703 -mcode-readable=@var{setting} @gol
704 -msplit-addresses -mno-split-addresses @gol
705 -mexplicit-relocs -mno-explicit-relocs @gol
706 -mcheck-zero-division -mno-check-zero-division @gol
707 -mdivide-traps -mdivide-breaks @gol
708 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
709 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
710 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
711 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
712 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
713 -mflush-func=@var{func} -mno-flush-func @gol
714 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
715 -mfp-exceptions -mno-fp-exceptions @gol
716 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
717 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
720 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
721 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
722 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
723 -mno-base-addresses -msingle-exit -mno-single-exit}
725 @emph{MN10300 Options}
726 @gccoptlist{-mmult-bug -mno-mult-bug @gol
727 -mno-am33 -mam33 -mam33-2 -mam34 @gol
728 -mtune=@var{cpu-type} @gol
729 -mreturn-pointer-on-d0 @gol
730 -mno-crt0 -mrelax -mliw -msetlb}
732 @emph{PDP-11 Options}
733 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
734 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
735 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
736 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
737 -mbranch-expensive -mbranch-cheap @gol
738 -munix-asm -mdec-asm}
740 @emph{picoChip Options}
741 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
742 -msymbol-as-address -mno-inefficient-warnings}
744 @emph{PowerPC Options}
745 See RS/6000 and PowerPC Options.
747 @emph{RS/6000 and PowerPC Options}
748 @gccoptlist{-mcpu=@var{cpu-type} @gol
749 -mtune=@var{cpu-type} @gol
750 -mcmodel=@var{code-model} @gol
751 -mpower -mno-power -mpower2 -mno-power2 @gol
752 -mpowerpc -mpowerpc64 -mno-powerpc @gol
753 -maltivec -mno-altivec @gol
754 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
755 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
756 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
757 -mfprnd -mno-fprnd @gol
758 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
759 -mnew-mnemonics -mold-mnemonics @gol
760 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
761 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
762 -malign-power -malign-natural @gol
763 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
764 -msingle-float -mdouble-float -msimple-fpu @gol
765 -mstring -mno-string -mupdate -mno-update @gol
766 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
767 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
768 -mstrict-align -mno-strict-align -mrelocatable @gol
769 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
770 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
771 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
772 -mprioritize-restricted-insns=@var{priority} @gol
773 -msched-costly-dep=@var{dependence_type} @gol
774 -minsert-sched-nops=@var{scheme} @gol
775 -mcall-sysv -mcall-netbsd @gol
776 -maix-struct-return -msvr4-struct-return @gol
777 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
778 -mblock-move-inline-limit=@var{num} @gol
779 -misel -mno-isel @gol
780 -misel=yes -misel=no @gol
782 -mspe=yes -mspe=no @gol
784 -mgen-cell-microcode -mwarn-cell-microcode @gol
785 -mvrsave -mno-vrsave @gol
786 -mmulhw -mno-mulhw @gol
787 -mdlmzb -mno-dlmzb @gol
788 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
789 -mprototype -mno-prototype @gol
790 -msim -mmvme -mads -myellowknife -memb -msdata @gol
791 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
792 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
793 -mno-recip-precision @gol
794 -mveclibabi=@var{type} -mfriz -mno-friz}
797 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
799 -mbig-endian-data -mlittle-endian-data @gol
802 -mas100-syntax -mno-as100-syntax@gol
804 -mmax-constant-size=@gol
806 -msave-acc-in-interrupts}
808 @emph{S/390 and zSeries Options}
809 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
810 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
811 -mlong-double-64 -mlong-double-128 @gol
812 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
813 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
814 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
815 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
816 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
819 @gccoptlist{-meb -mel @gol
823 -mscore5 -mscore5u -mscore7 -mscore7d}
826 @gccoptlist{-m1 -m2 -m2e @gol
827 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
829 -m4-nofpu -m4-single-only -m4-single -m4 @gol
830 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
831 -m5-64media -m5-64media-nofpu @gol
832 -m5-32media -m5-32media-nofpu @gol
833 -m5-compact -m5-compact-nofpu @gol
834 -mb -ml -mdalign -mrelax @gol
835 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
836 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
837 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
838 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
839 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
840 -maccumulate-outgoing-args -minvalid-symbols}
842 @emph{Solaris 2 Options}
843 @gccoptlist{-mimpure-text -mno-impure-text @gol
847 @gccoptlist{-mcpu=@var{cpu-type} @gol
848 -mtune=@var{cpu-type} @gol
849 -mcmodel=@var{code-model} @gol
850 -m32 -m64 -mapp-regs -mno-app-regs @gol
851 -mfaster-structs -mno-faster-structs @gol
852 -mfpu -mno-fpu -mhard-float -msoft-float @gol
853 -mhard-quad-float -msoft-quad-float @gol
855 -mstack-bias -mno-stack-bias @gol
856 -munaligned-doubles -mno-unaligned-doubles @gol
857 -mv8plus -mno-v8plus -mvis -mno-vis}
860 @gccoptlist{-mwarn-reloc -merror-reloc @gol
861 -msafe-dma -munsafe-dma @gol
863 -msmall-mem -mlarge-mem -mstdmain @gol
864 -mfixed-range=@var{register-range} @gol
866 -maddress-space-conversion -mno-address-space-conversion @gol
867 -mcache-size=@var{cache-size} @gol
868 -matomic-updates -mno-atomic-updates}
870 @emph{System V Options}
871 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
874 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
875 -mprolog-function -mno-prolog-function -mspace @gol
876 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
877 -mapp-regs -mno-app-regs @gol
878 -mdisable-callt -mno-disable-callt @gol
881 -mv850e1 -mv850es @gol
886 @gccoptlist{-mg -mgnu -munix}
888 @emph{VxWorks Options}
889 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
890 -Xbind-lazy -Xbind-now}
892 @emph{x86-64 Options}
893 See i386 and x86-64 Options.
895 @emph{i386 and x86-64 Windows Options}
896 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
897 -mnop-fun-dllimport -mthread @gol
898 -municode -mwin32 -mwindows -fno-set-stack-executable}
900 @emph{Xstormy16 Options}
903 @emph{Xtensa Options}
904 @gccoptlist{-mconst16 -mno-const16 @gol
905 -mfused-madd -mno-fused-madd @gol
907 -mserialize-volatile -mno-serialize-volatile @gol
908 -mtext-section-literals -mno-text-section-literals @gol
909 -mtarget-align -mno-target-align @gol
910 -mlongcalls -mno-longcalls}
912 @emph{zSeries Options}
913 See S/390 and zSeries Options.
915 @item Code Generation Options
916 @xref{Code Gen Options,,Options for Code Generation Conventions}.
917 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
918 -ffixed-@var{reg} -fexceptions @gol
919 -fnon-call-exceptions -funwind-tables @gol
920 -fasynchronous-unwind-tables @gol
921 -finhibit-size-directive -finstrument-functions @gol
922 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
923 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
924 -fno-common -fno-ident @gol
925 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
926 -fno-jump-tables @gol
927 -frecord-gcc-switches @gol
928 -freg-struct-return -fshort-enums @gol
929 -fshort-double -fshort-wchar @gol
930 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
931 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
932 -fno-stack-limit -fsplit-stack @gol
933 -fleading-underscore -ftls-model=@var{model} @gol
934 -ftrapv -fwrapv -fbounds-check @gol
935 -fvisibility -fstrict-volatile-bitfields}
939 * Overall Options:: Controlling the kind of output:
940 an executable, object files, assembler files,
941 or preprocessed source.
942 * C Dialect Options:: Controlling the variant of C language compiled.
943 * C++ Dialect Options:: Variations on C++.
944 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
946 * Language Independent Options:: Controlling how diagnostics should be
948 * Warning Options:: How picky should the compiler be?
949 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
950 * Optimize Options:: How much optimization?
951 * Preprocessor Options:: Controlling header files and macro definitions.
952 Also, getting dependency information for Make.
953 * Assembler Options:: Passing options to the assembler.
954 * Link Options:: Specifying libraries and so on.
955 * Directory Options:: Where to find header files and libraries.
956 Where to find the compiler executable files.
957 * Spec Files:: How to pass switches to sub-processes.
958 * Target Options:: Running a cross-compiler, or an old version of GCC.
961 @node Overall Options
962 @section Options Controlling the Kind of Output
964 Compilation can involve up to four stages: preprocessing, compilation
965 proper, assembly and linking, always in that order. GCC is capable of
966 preprocessing and compiling several files either into several
967 assembler input files, or into one assembler input file; then each
968 assembler input file produces an object file, and linking combines all
969 the object files (those newly compiled, and those specified as input)
970 into an executable file.
972 @cindex file name suffix
973 For any given input file, the file name suffix determines what kind of
978 C source code which must be preprocessed.
981 C source code which should not be preprocessed.
984 C++ source code which should not be preprocessed.
987 Objective-C source code. Note that you must link with the @file{libobjc}
988 library to make an Objective-C program work.
991 Objective-C source code which should not be preprocessed.
995 Objective-C++ source code. Note that you must link with the @file{libobjc}
996 library to make an Objective-C++ program work. Note that @samp{.M} refers
997 to a literal capital M@.
1000 Objective-C++ source code which should not be preprocessed.
1003 C, C++, Objective-C or Objective-C++ header file to be turned into a
1004 precompiled header (default), or C, C++ header file to be turned into an
1005 Ada spec (via the @option{-fdump-ada-spec} switch).
1008 @itemx @var{file}.cp
1009 @itemx @var{file}.cxx
1010 @itemx @var{file}.cpp
1011 @itemx @var{file}.CPP
1012 @itemx @var{file}.c++
1014 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1015 the last two letters must both be literally @samp{x}. Likewise,
1016 @samp{.C} refers to a literal capital C@.
1020 Objective-C++ source code which must be preprocessed.
1022 @item @var{file}.mii
1023 Objective-C++ source code which should not be preprocessed.
1027 @itemx @var{file}.hp
1028 @itemx @var{file}.hxx
1029 @itemx @var{file}.hpp
1030 @itemx @var{file}.HPP
1031 @itemx @var{file}.h++
1032 @itemx @var{file}.tcc
1033 C++ header file to be turned into a precompiled header or Ada spec.
1036 @itemx @var{file}.for
1037 @itemx @var{file}.ftn
1038 Fixed form Fortran source code which should not be preprocessed.
1041 @itemx @var{file}.FOR
1042 @itemx @var{file}.fpp
1043 @itemx @var{file}.FPP
1044 @itemx @var{file}.FTN
1045 Fixed form Fortran source code which must be preprocessed (with the traditional
1048 @item @var{file}.f90
1049 @itemx @var{file}.f95
1050 @itemx @var{file}.f03
1051 @itemx @var{file}.f08
1052 Free form Fortran source code which should not be preprocessed.
1054 @item @var{file}.F90
1055 @itemx @var{file}.F95
1056 @itemx @var{file}.F03
1057 @itemx @var{file}.F08
1058 Free form Fortran source code which must be preprocessed (with the
1059 traditional preprocessor).
1064 @c FIXME: Descriptions of Java file types.
1070 @item @var{file}.ads
1071 Ada source code file which contains a library unit declaration (a
1072 declaration of a package, subprogram, or generic, or a generic
1073 instantiation), or a library unit renaming declaration (a package,
1074 generic, or subprogram renaming declaration). Such files are also
1077 @item @var{file}.adb
1078 Ada source code file containing a library unit body (a subprogram or
1079 package body). Such files are also called @dfn{bodies}.
1081 @c GCC also knows about some suffixes for languages not yet included:
1092 @itemx @var{file}.sx
1093 Assembler code which must be preprocessed.
1096 An object file to be fed straight into linking.
1097 Any file name with no recognized suffix is treated this way.
1101 You can specify the input language explicitly with the @option{-x} option:
1104 @item -x @var{language}
1105 Specify explicitly the @var{language} for the following input files
1106 (rather than letting the compiler choose a default based on the file
1107 name suffix). This option applies to all following input files until
1108 the next @option{-x} option. Possible values for @var{language} are:
1110 c c-header cpp-output
1111 c++ c++-header c++-cpp-output
1112 objective-c objective-c-header objective-c-cpp-output
1113 objective-c++ objective-c++-header objective-c++-cpp-output
1114 assembler assembler-with-cpp
1116 f77 f77-cpp-input f95 f95-cpp-input
1122 Turn off any specification of a language, so that subsequent files are
1123 handled according to their file name suffixes (as they are if @option{-x}
1124 has not been used at all).
1126 @item -pass-exit-codes
1127 @opindex pass-exit-codes
1128 Normally the @command{gcc} program will exit with the code of 1 if any
1129 phase of the compiler returns a non-success return code. If you specify
1130 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1131 numerically highest error produced by any phase that returned an error
1132 indication. The C, C++, and Fortran frontends return 4, if an internal
1133 compiler error is encountered.
1136 If you only want some of the stages of compilation, you can use
1137 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1138 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1139 @command{gcc} is to stop. Note that some combinations (for example,
1140 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1145 Compile or assemble the source files, but do not link. The linking
1146 stage simply is not done. The ultimate output is in the form of an
1147 object file for each source file.
1149 By default, the object file name for a source file is made by replacing
1150 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1152 Unrecognized input files, not requiring compilation or assembly, are
1157 Stop after the stage of compilation proper; do not assemble. The output
1158 is in the form of an assembler code file for each non-assembler input
1161 By default, the assembler file name for a source file is made by
1162 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1164 Input files that don't require compilation are ignored.
1168 Stop after the preprocessing stage; do not run the compiler proper. The
1169 output is in the form of preprocessed source code, which is sent to the
1172 Input files which don't require preprocessing are ignored.
1174 @cindex output file option
1177 Place output in file @var{file}. This applies regardless to whatever
1178 sort of output is being produced, whether it be an executable file,
1179 an object file, an assembler file or preprocessed C code.
1181 If @option{-o} is not specified, the default is to put an executable
1182 file in @file{a.out}, the object file for
1183 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1184 assembler file in @file{@var{source}.s}, a precompiled header file in
1185 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1190 Print (on standard error output) the commands executed to run the stages
1191 of compilation. Also print the version number of the compiler driver
1192 program and of the preprocessor and the compiler proper.
1196 Like @option{-v} except the commands are not executed and arguments
1197 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1198 This is useful for shell scripts to capture the driver-generated command lines.
1202 Use pipes rather than temporary files for communication between the
1203 various stages of compilation. This fails to work on some systems where
1204 the assembler is unable to read from a pipe; but the GNU assembler has
1209 Print (on the standard output) a description of the command line options
1210 understood by @command{gcc}. If the @option{-v} option is also specified
1211 then @option{--help} will also be passed on to the various processes
1212 invoked by @command{gcc}, so that they can display the command line options
1213 they accept. If the @option{-Wextra} option has also been specified
1214 (prior to the @option{--help} option), then command line options which
1215 have no documentation associated with them will also be displayed.
1218 @opindex target-help
1219 Print (on the standard output) a description of target-specific command
1220 line options for each tool. For some targets extra target-specific
1221 information may also be printed.
1223 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1224 Print (on the standard output) a description of the command line
1225 options understood by the compiler that fit into all specified classes
1226 and qualifiers. These are the supported classes:
1229 @item @samp{optimizers}
1230 This will display all of the optimization options supported by the
1233 @item @samp{warnings}
1234 This will display all of the options controlling warning messages
1235 produced by the compiler.
1238 This will display target-specific options. Unlike the
1239 @option{--target-help} option however, target-specific options of the
1240 linker and assembler will not be displayed. This is because those
1241 tools do not currently support the extended @option{--help=} syntax.
1244 This will display the values recognized by the @option{--param}
1247 @item @var{language}
1248 This will display the options supported for @var{language}, where
1249 @var{language} is the name of one of the languages supported in this
1253 This will display the options that are common to all languages.
1256 These are the supported qualifiers:
1259 @item @samp{undocumented}
1260 Display only those options which are undocumented.
1263 Display options which take an argument that appears after an equal
1264 sign in the same continuous piece of text, such as:
1265 @samp{--help=target}.
1267 @item @samp{separate}
1268 Display options which take an argument that appears as a separate word
1269 following the original option, such as: @samp{-o output-file}.
1272 Thus for example to display all the undocumented target-specific
1273 switches supported by the compiler the following can be used:
1276 --help=target,undocumented
1279 The sense of a qualifier can be inverted by prefixing it with the
1280 @samp{^} character, so for example to display all binary warning
1281 options (i.e., ones that are either on or off and that do not take an
1282 argument), which have a description the following can be used:
1285 --help=warnings,^joined,^undocumented
1288 The argument to @option{--help=} should not consist solely of inverted
1291 Combining several classes is possible, although this usually
1292 restricts the output by so much that there is nothing to display. One
1293 case where it does work however is when one of the classes is
1294 @var{target}. So for example to display all the target-specific
1295 optimization options the following can be used:
1298 --help=target,optimizers
1301 The @option{--help=} option can be repeated on the command line. Each
1302 successive use will display its requested class of options, skipping
1303 those that have already been displayed.
1305 If the @option{-Q} option appears on the command line before the
1306 @option{--help=} option, then the descriptive text displayed by
1307 @option{--help=} is changed. Instead of describing the displayed
1308 options, an indication is given as to whether the option is enabled,
1309 disabled or set to a specific value (assuming that the compiler
1310 knows this at the point where the @option{--help=} option is used).
1312 Here is a truncated example from the ARM port of @command{gcc}:
1315 % gcc -Q -mabi=2 --help=target -c
1316 The following options are target specific:
1318 -mabort-on-noreturn [disabled]
1322 The output is sensitive to the effects of previous command line
1323 options, so for example it is possible to find out which optimizations
1324 are enabled at @option{-O2} by using:
1327 -Q -O2 --help=optimizers
1330 Alternatively you can discover which binary optimizations are enabled
1331 by @option{-O3} by using:
1334 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1335 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1336 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1339 @item -no-canonical-prefixes
1340 @opindex no-canonical-prefixes
1341 Do not expand any symbolic links, resolve references to @samp{/../}
1342 or @samp{/./}, or make the path absolute when generating a relative
1347 Display the version number and copyrights of the invoked GCC@.
1351 Invoke all subcommands under a wrapper program. The name of the
1352 wrapper program and its parameters are passed as a comma separated
1356 gcc -c t.c -wrapper gdb,--args
1359 This will invoke all subprograms of @command{gcc} under
1360 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1361 @samp{gdb --args cc1 @dots{}}.
1363 @item -fplugin=@var{name}.so
1364 Load the plugin code in file @var{name}.so, assumed to be a
1365 shared object to be dlopen'd by the compiler. The base name of
1366 the shared object file is used to identify the plugin for the
1367 purposes of argument parsing (See
1368 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1369 Each plugin should define the callback functions specified in the
1372 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1373 Define an argument called @var{key} with a value of @var{value}
1374 for the plugin called @var{name}.
1376 @item -fdump-ada-spec@r{[}-slim@r{]}
1377 For C and C++ source and include files, generate corresponding Ada
1378 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1379 GNAT User's Guide}, which provides detailed documentation on this feature.
1381 @item -fdump-go-spec=@var{file}
1382 For input files in any language, generate corresponding Go
1383 declarations in @var{file}. This generates Go @code{const},
1384 @code{type}, @code{var}, and @code{func} declarations which may be a
1385 useful way to start writing a Go interface to code written in some
1388 @include @value{srcdir}/../libiberty/at-file.texi
1392 @section Compiling C++ Programs
1394 @cindex suffixes for C++ source
1395 @cindex C++ source file suffixes
1396 C++ source files conventionally use one of the suffixes @samp{.C},
1397 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1398 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1399 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1400 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1401 files with these names and compiles them as C++ programs even if you
1402 call the compiler the same way as for compiling C programs (usually
1403 with the name @command{gcc}).
1407 However, the use of @command{gcc} does not add the C++ library.
1408 @command{g++} is a program that calls GCC and treats @samp{.c},
1409 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1410 files unless @option{-x} is used, and automatically specifies linking
1411 against the C++ library. This program is also useful when
1412 precompiling a C header file with a @samp{.h} extension for use in C++
1413 compilations. On many systems, @command{g++} is also installed with
1414 the name @command{c++}.
1416 @cindex invoking @command{g++}
1417 When you compile C++ programs, you may specify many of the same
1418 command-line options that you use for compiling programs in any
1419 language; or command-line options meaningful for C and related
1420 languages; or options that are meaningful only for C++ programs.
1421 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1422 explanations of options for languages related to C@.
1423 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1424 explanations of options that are meaningful only for C++ programs.
1426 @node C Dialect Options
1427 @section Options Controlling C Dialect
1428 @cindex dialect options
1429 @cindex language dialect options
1430 @cindex options, dialect
1432 The following options control the dialect of C (or languages derived
1433 from C, such as C++, Objective-C and Objective-C++) that the compiler
1437 @cindex ANSI support
1441 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1442 equivalent to @samp{-std=c++98}.
1444 This turns off certain features of GCC that are incompatible with ISO
1445 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1446 such as the @code{asm} and @code{typeof} keywords, and
1447 predefined macros such as @code{unix} and @code{vax} that identify the
1448 type of system you are using. It also enables the undesirable and
1449 rarely used ISO trigraph feature. For the C compiler,
1450 it disables recognition of C++ style @samp{//} comments as well as
1451 the @code{inline} keyword.
1453 The alternate keywords @code{__asm__}, @code{__extension__},
1454 @code{__inline__} and @code{__typeof__} continue to work despite
1455 @option{-ansi}. You would not want to use them in an ISO C program, of
1456 course, but it is useful to put them in header files that might be included
1457 in compilations done with @option{-ansi}. Alternate predefined macros
1458 such as @code{__unix__} and @code{__vax__} are also available, with or
1459 without @option{-ansi}.
1461 The @option{-ansi} option does not cause non-ISO programs to be
1462 rejected gratuitously. For that, @option{-pedantic} is required in
1463 addition to @option{-ansi}. @xref{Warning Options}.
1465 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1466 option is used. Some header files may notice this macro and refrain
1467 from declaring certain functions or defining certain macros that the
1468 ISO standard doesn't call for; this is to avoid interfering with any
1469 programs that might use these names for other things.
1471 Functions that would normally be built in but do not have semantics
1472 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1473 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1474 built-in functions provided by GCC}, for details of the functions
1479 Determine the language standard. @xref{Standards,,Language Standards
1480 Supported by GCC}, for details of these standard versions. This option
1481 is currently only supported when compiling C or C++.
1483 The compiler can accept several base standards, such as @samp{c90} or
1484 @samp{c++98}, and GNU dialects of those standards, such as
1485 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1486 compiler will accept all programs following that standard and those
1487 using GNU extensions that do not contradict it. For example,
1488 @samp{-std=c90} turns off certain features of GCC that are
1489 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1490 keywords, but not other GNU extensions that do not have a meaning in
1491 ISO C90, such as omitting the middle term of a @code{?:}
1492 expression. On the other hand, by specifying a GNU dialect of a
1493 standard, all features the compiler support are enabled, even when
1494 those features change the meaning of the base standard and some
1495 strict-conforming programs may be rejected. The particular standard
1496 is used by @option{-pedantic} to identify which features are GNU
1497 extensions given that version of the standard. For example
1498 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1499 comments, while @samp{-std=gnu99 -pedantic} would not.
1501 A value for this option must be provided; possible values are
1507 Support all ISO C90 programs (certain GNU extensions that conflict
1508 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1510 @item iso9899:199409
1511 ISO C90 as modified in amendment 1.
1517 ISO C99. Note that this standard is not yet fully supported; see
1518 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1519 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1522 ISO C1X, the draft of the next revision of the ISO C standard.
1523 Support is limited and experimental and features enabled by this
1524 option may be changed or removed if changed in or removed from the
1529 GNU dialect of ISO C90 (including some C99 features). This
1530 is the default for C code.
1534 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1535 this will become the default. The name @samp{gnu9x} is deprecated.
1538 GNU dialect of ISO C1X. Support is limited and experimental and
1539 features enabled by this option may be changed or removed if changed
1540 in or removed from the standard draft.
1543 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1547 GNU dialect of @option{-std=c++98}. This is the default for
1551 The working draft of the upcoming ISO C++0x standard. This option
1552 enables experimental features that are likely to be included in
1553 C++0x. The working draft is constantly changing, and any feature that is
1554 enabled by this flag may be removed from future versions of GCC if it is
1555 not part of the C++0x standard.
1558 GNU dialect of @option{-std=c++0x}. This option enables
1559 experimental features that may be removed in future versions of GCC.
1562 @item -fgnu89-inline
1563 @opindex fgnu89-inline
1564 The option @option{-fgnu89-inline} tells GCC to use the traditional
1565 GNU semantics for @code{inline} functions when in C99 mode.
1566 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1567 is accepted and ignored by GCC versions 4.1.3 up to but not including
1568 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1569 C99 mode. Using this option is roughly equivalent to adding the
1570 @code{gnu_inline} function attribute to all inline functions
1571 (@pxref{Function Attributes}).
1573 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1574 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1575 specifies the default behavior). This option was first supported in
1576 GCC 4.3. This option is not supported in @option{-std=c90} or
1577 @option{-std=gnu90} mode.
1579 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1580 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1581 in effect for @code{inline} functions. @xref{Common Predefined
1582 Macros,,,cpp,The C Preprocessor}.
1584 @item -aux-info @var{filename}
1586 Output to the given filename prototyped declarations for all functions
1587 declared and/or defined in a translation unit, including those in header
1588 files. This option is silently ignored in any language other than C@.
1590 Besides declarations, the file indicates, in comments, the origin of
1591 each declaration (source file and line), whether the declaration was
1592 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1593 @samp{O} for old, respectively, in the first character after the line
1594 number and the colon), and whether it came from a declaration or a
1595 definition (@samp{C} or @samp{F}, respectively, in the following
1596 character). In the case of function definitions, a K&R-style list of
1597 arguments followed by their declarations is also provided, inside
1598 comments, after the declaration.
1602 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1603 keyword, so that code can use these words as identifiers. You can use
1604 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1605 instead. @option{-ansi} implies @option{-fno-asm}.
1607 In C++, this switch only affects the @code{typeof} keyword, since
1608 @code{asm} and @code{inline} are standard keywords. You may want to
1609 use the @option{-fno-gnu-keywords} flag instead, which has the same
1610 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1611 switch only affects the @code{asm} and @code{typeof} keywords, since
1612 @code{inline} is a standard keyword in ISO C99.
1615 @itemx -fno-builtin-@var{function}
1616 @opindex fno-builtin
1617 @cindex built-in functions
1618 Don't recognize built-in functions that do not begin with
1619 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1620 functions provided by GCC}, for details of the functions affected,
1621 including those which are not built-in functions when @option{-ansi} or
1622 @option{-std} options for strict ISO C conformance are used because they
1623 do not have an ISO standard meaning.
1625 GCC normally generates special code to handle certain built-in functions
1626 more efficiently; for instance, calls to @code{alloca} may become single
1627 instructions that adjust the stack directly, and calls to @code{memcpy}
1628 may become inline copy loops. The resulting code is often both smaller
1629 and faster, but since the function calls no longer appear as such, you
1630 cannot set a breakpoint on those calls, nor can you change the behavior
1631 of the functions by linking with a different library. In addition,
1632 when a function is recognized as a built-in function, GCC may use
1633 information about that function to warn about problems with calls to
1634 that function, or to generate more efficient code, even if the
1635 resulting code still contains calls to that function. For example,
1636 warnings are given with @option{-Wformat} for bad calls to
1637 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1638 known not to modify global memory.
1640 With the @option{-fno-builtin-@var{function}} option
1641 only the built-in function @var{function} is
1642 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1643 function is named that is not built-in in this version of GCC, this
1644 option is ignored. There is no corresponding
1645 @option{-fbuiltin-@var{function}} option; if you wish to enable
1646 built-in functions selectively when using @option{-fno-builtin} or
1647 @option{-ffreestanding}, you may define macros such as:
1650 #define abs(n) __builtin_abs ((n))
1651 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1656 @cindex hosted environment
1658 Assert that compilation takes place in a hosted environment. This implies
1659 @option{-fbuiltin}. A hosted environment is one in which the
1660 entire standard library is available, and in which @code{main} has a return
1661 type of @code{int}. Examples are nearly everything except a kernel.
1662 This is equivalent to @option{-fno-freestanding}.
1664 @item -ffreestanding
1665 @opindex ffreestanding
1666 @cindex hosted environment
1668 Assert that compilation takes place in a freestanding environment. This
1669 implies @option{-fno-builtin}. A freestanding environment
1670 is one in which the standard library may not exist, and program startup may
1671 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1672 This is equivalent to @option{-fno-hosted}.
1674 @xref{Standards,,Language Standards Supported by GCC}, for details of
1675 freestanding and hosted environments.
1679 @cindex OpenMP parallel
1680 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1681 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1682 compiler generates parallel code according to the OpenMP Application
1683 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1684 implies @option{-pthread}, and thus is only supported on targets that
1685 have support for @option{-pthread}.
1687 @item -fms-extensions
1688 @opindex fms-extensions
1689 Accept some non-standard constructs used in Microsoft header files.
1691 In C++ code, this allows member names in structures to be similar
1692 to previous types declarations.
1701 Some cases of unnamed fields in structures and unions are only
1702 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1703 fields within structs/unions}, for details.
1705 @item -fplan9-extensions
1706 Accept some non-standard constructs used in Plan 9 code.
1708 This enables @option{-fms-extensions}, permits passing pointers to
1709 structures with anonymous fields to functions which expect pointers to
1710 elements of the type of the field, and permits referring to anonymous
1711 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1712 struct/union fields within structs/unions}, for details. This is only
1713 supported for C, not C++.
1717 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1718 options for strict ISO C conformance) implies @option{-trigraphs}.
1720 @item -no-integrated-cpp
1721 @opindex no-integrated-cpp
1722 Performs a compilation in two passes: preprocessing and compiling. This
1723 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1724 @option{-B} option. The user supplied compilation step can then add in
1725 an additional preprocessing step after normal preprocessing but before
1726 compiling. The default is to use the integrated cpp (internal cpp)
1728 The semantics of this option will change if "cc1", "cc1plus", and
1729 "cc1obj" are merged.
1731 @cindex traditional C language
1732 @cindex C language, traditional
1734 @itemx -traditional-cpp
1735 @opindex traditional-cpp
1736 @opindex traditional
1737 Formerly, these options caused GCC to attempt to emulate a pre-standard
1738 C compiler. They are now only supported with the @option{-E} switch.
1739 The preprocessor continues to support a pre-standard mode. See the GNU
1740 CPP manual for details.
1742 @item -fcond-mismatch
1743 @opindex fcond-mismatch
1744 Allow conditional expressions with mismatched types in the second and
1745 third arguments. The value of such an expression is void. This option
1746 is not supported for C++.
1748 @item -flax-vector-conversions
1749 @opindex flax-vector-conversions
1750 Allow implicit conversions between vectors with differing numbers of
1751 elements and/or incompatible element types. This option should not be
1754 @item -funsigned-char
1755 @opindex funsigned-char
1756 Let the type @code{char} be unsigned, like @code{unsigned char}.
1758 Each kind of machine has a default for what @code{char} should
1759 be. It is either like @code{unsigned char} by default or like
1760 @code{signed char} by default.
1762 Ideally, a portable program should always use @code{signed char} or
1763 @code{unsigned char} when it depends on the signedness of an object.
1764 But many programs have been written to use plain @code{char} and
1765 expect it to be signed, or expect it to be unsigned, depending on the
1766 machines they were written for. This option, and its inverse, let you
1767 make such a program work with the opposite default.
1769 The type @code{char} is always a distinct type from each of
1770 @code{signed char} or @code{unsigned char}, even though its behavior
1771 is always just like one of those two.
1774 @opindex fsigned-char
1775 Let the type @code{char} be signed, like @code{signed char}.
1777 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1778 the negative form of @option{-funsigned-char}. Likewise, the option
1779 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1781 @item -fsigned-bitfields
1782 @itemx -funsigned-bitfields
1783 @itemx -fno-signed-bitfields
1784 @itemx -fno-unsigned-bitfields
1785 @opindex fsigned-bitfields
1786 @opindex funsigned-bitfields
1787 @opindex fno-signed-bitfields
1788 @opindex fno-unsigned-bitfields
1789 These options control whether a bit-field is signed or unsigned, when the
1790 declaration does not use either @code{signed} or @code{unsigned}. By
1791 default, such a bit-field is signed, because this is consistent: the
1792 basic integer types such as @code{int} are signed types.
1795 @node C++ Dialect Options
1796 @section Options Controlling C++ Dialect
1798 @cindex compiler options, C++
1799 @cindex C++ options, command line
1800 @cindex options, C++
1801 This section describes the command-line options that are only meaningful
1802 for C++ programs; but you can also use most of the GNU compiler options
1803 regardless of what language your program is in. For example, you
1804 might compile a file @code{firstClass.C} like this:
1807 g++ -g -frepo -O -c firstClass.C
1811 In this example, only @option{-frepo} is an option meant
1812 only for C++ programs; you can use the other options with any
1813 language supported by GCC@.
1815 Here is a list of options that are @emph{only} for compiling C++ programs:
1819 @item -fabi-version=@var{n}
1820 @opindex fabi-version
1821 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1822 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1823 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1824 the version that conforms most closely to the C++ ABI specification.
1825 Therefore, the ABI obtained using version 0 will change as ABI bugs
1828 The default is version 2.
1830 Version 3 corrects an error in mangling a constant address as a
1833 Version 4 implements a standard mangling for vector types.
1835 Version 5 corrects the mangling of attribute const/volatile on
1836 function pointer types, decltype of a plain decl, and use of a
1837 function parameter in the declaration of another parameter.
1839 See also @option{-Wabi}.
1841 @item -fno-access-control
1842 @opindex fno-access-control
1843 Turn off all access checking. This switch is mainly useful for working
1844 around bugs in the access control code.
1848 Check that the pointer returned by @code{operator new} is non-null
1849 before attempting to modify the storage allocated. This check is
1850 normally unnecessary because the C++ standard specifies that
1851 @code{operator new} will only return @code{0} if it is declared
1852 @samp{throw()}, in which case the compiler will always check the
1853 return value even without this option. In all other cases, when
1854 @code{operator new} has a non-empty exception specification, memory
1855 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1856 @samp{new (nothrow)}.
1858 @item -fconserve-space
1859 @opindex fconserve-space
1860 Put uninitialized or runtime-initialized global variables into the
1861 common segment, as C does. This saves space in the executable at the
1862 cost of not diagnosing duplicate definitions. If you compile with this
1863 flag and your program mysteriously crashes after @code{main()} has
1864 completed, you may have an object that is being destroyed twice because
1865 two definitions were merged.
1867 This option is no longer useful on most targets, now that support has
1868 been added for putting variables into BSS without making them common.
1870 @item -fconstexpr-depth=@var{n}
1871 @opindex fconstexpr-depth
1872 Set the maximum nested evaluation depth for C++0x constexpr functions
1873 to @var{n}. A limit is needed to detect endless recursion during
1874 constant expression evaluation. The minimum specified by the standard
1877 @item -fno-deduce-init-list
1878 @opindex fno-deduce-init-list
1879 Disable deduction of a template type parameter as
1880 std::initializer_list from a brace-enclosed initializer list, i.e.
1883 template <class T> auto forward(T t) -> decltype (realfn (t))
1890 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1894 This option is present because this deduction is an extension to the
1895 current specification in the C++0x working draft, and there was
1896 some concern about potential overload resolution problems.
1898 @item -ffriend-injection
1899 @opindex ffriend-injection
1900 Inject friend functions into the enclosing namespace, so that they are
1901 visible outside the scope of the class in which they are declared.
1902 Friend functions were documented to work this way in the old Annotated
1903 C++ Reference Manual, and versions of G++ before 4.1 always worked
1904 that way. However, in ISO C++ a friend function which is not declared
1905 in an enclosing scope can only be found using argument dependent
1906 lookup. This option causes friends to be injected as they were in
1909 This option is for compatibility, and may be removed in a future
1912 @item -fno-elide-constructors
1913 @opindex fno-elide-constructors
1914 The C++ standard allows an implementation to omit creating a temporary
1915 which is only used to initialize another object of the same type.
1916 Specifying this option disables that optimization, and forces G++ to
1917 call the copy constructor in all cases.
1919 @item -fno-enforce-eh-specs
1920 @opindex fno-enforce-eh-specs
1921 Don't generate code to check for violation of exception specifications
1922 at runtime. This option violates the C++ standard, but may be useful
1923 for reducing code size in production builds, much like defining
1924 @samp{NDEBUG}. This does not give user code permission to throw
1925 exceptions in violation of the exception specifications; the compiler
1926 will still optimize based on the specifications, so throwing an
1927 unexpected exception will result in undefined behavior.
1930 @itemx -fno-for-scope
1932 @opindex fno-for-scope
1933 If @option{-ffor-scope} is specified, the scope of variables declared in
1934 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1935 as specified by the C++ standard.
1936 If @option{-fno-for-scope} is specified, the scope of variables declared in
1937 a @i{for-init-statement} extends to the end of the enclosing scope,
1938 as was the case in old versions of G++, and other (traditional)
1939 implementations of C++.
1941 The default if neither flag is given to follow the standard,
1942 but to allow and give a warning for old-style code that would
1943 otherwise be invalid, or have different behavior.
1945 @item -fno-gnu-keywords
1946 @opindex fno-gnu-keywords
1947 Do not recognize @code{typeof} as a keyword, so that code can use this
1948 word as an identifier. You can use the keyword @code{__typeof__} instead.
1949 @option{-ansi} implies @option{-fno-gnu-keywords}.
1951 @item -fno-implicit-templates
1952 @opindex fno-implicit-templates
1953 Never emit code for non-inline templates which are instantiated
1954 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1955 @xref{Template Instantiation}, for more information.
1957 @item -fno-implicit-inline-templates
1958 @opindex fno-implicit-inline-templates
1959 Don't emit code for implicit instantiations of inline templates, either.
1960 The default is to handle inlines differently so that compiles with and
1961 without optimization will need the same set of explicit instantiations.
1963 @item -fno-implement-inlines
1964 @opindex fno-implement-inlines
1965 To save space, do not emit out-of-line copies of inline functions
1966 controlled by @samp{#pragma implementation}. This will cause linker
1967 errors if these functions are not inlined everywhere they are called.
1969 @item -fms-extensions
1970 @opindex fms-extensions
1971 Disable pedantic warnings about constructs used in MFC, such as implicit
1972 int and getting a pointer to member function via non-standard syntax.
1974 @item -fno-nonansi-builtins
1975 @opindex fno-nonansi-builtins
1976 Disable built-in declarations of functions that are not mandated by
1977 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1978 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1981 @opindex fnothrow-opt
1982 Treat a @code{throw()} exception specification as though it were a
1983 @code{noexcept} specification to reduce or eliminate the text size
1984 overhead relative to a function with no exception specification. If
1985 the function has local variables of types with non-trivial
1986 destructors, the exception specification will actually make the
1987 function smaller because the EH cleanups for those variables can be
1988 optimized away. The semantic effect is that an exception thrown out of
1989 a function with such an exception specification will result in a call
1990 to @code{terminate} rather than @code{unexpected}.
1992 @item -fno-operator-names
1993 @opindex fno-operator-names
1994 Do not treat the operator name keywords @code{and}, @code{bitand},
1995 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1996 synonyms as keywords.
1998 @item -fno-optional-diags
1999 @opindex fno-optional-diags
2000 Disable diagnostics that the standard says a compiler does not need to
2001 issue. Currently, the only such diagnostic issued by G++ is the one for
2002 a name having multiple meanings within a class.
2005 @opindex fpermissive
2006 Downgrade some diagnostics about nonconformant code from errors to
2007 warnings. Thus, using @option{-fpermissive} will allow some
2008 nonconforming code to compile.
2010 @item -fno-pretty-templates
2011 @opindex fno-pretty-templates
2012 When an error message refers to a specialization of a function
2013 template, the compiler will normally print the signature of the
2014 template followed by the template arguments and any typedefs or
2015 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2016 rather than @code{void f(int)}) so that it's clear which template is
2017 involved. When an error message refers to a specialization of a class
2018 template, the compiler will omit any template arguments which match
2019 the default template arguments for that template. If either of these
2020 behaviors make it harder to understand the error message rather than
2021 easier, using @option{-fno-pretty-templates} will disable them.
2025 Enable automatic template instantiation at link time. This option also
2026 implies @option{-fno-implicit-templates}. @xref{Template
2027 Instantiation}, for more information.
2031 Disable generation of information about every class with virtual
2032 functions for use by the C++ runtime type identification features
2033 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2034 of the language, you can save some space by using this flag. Note that
2035 exception handling uses the same information, but it will generate it as
2036 needed. The @samp{dynamic_cast} operator can still be used for casts that
2037 do not require runtime type information, i.e.@: casts to @code{void *} or to
2038 unambiguous base classes.
2042 Emit statistics about front-end processing at the end of the compilation.
2043 This information is generally only useful to the G++ development team.
2045 @item -fstrict-enums
2046 @opindex fstrict-enums
2047 Allow the compiler to optimize using the assumption that a value of
2048 enumeration type can only be one of the values of the enumeration (as
2049 defined in the C++ standard; basically, a value which can be
2050 represented in the minimum number of bits needed to represent all the
2051 enumerators). This assumption may not be valid if the program uses a
2052 cast to convert an arbitrary integer value to the enumeration type.
2054 @item -ftemplate-depth=@var{n}
2055 @opindex ftemplate-depth
2056 Set the maximum instantiation depth for template classes to @var{n}.
2057 A limit on the template instantiation depth is needed to detect
2058 endless recursions during template class instantiation. ANSI/ISO C++
2059 conforming programs must not rely on a maximum depth greater than 17
2060 (changed to 1024 in C++0x).
2062 @item -fno-threadsafe-statics
2063 @opindex fno-threadsafe-statics
2064 Do not emit the extra code to use the routines specified in the C++
2065 ABI for thread-safe initialization of local statics. You can use this
2066 option to reduce code size slightly in code that doesn't need to be
2069 @item -fuse-cxa-atexit
2070 @opindex fuse-cxa-atexit
2071 Register destructors for objects with static storage duration with the
2072 @code{__cxa_atexit} function rather than the @code{atexit} function.
2073 This option is required for fully standards-compliant handling of static
2074 destructors, but will only work if your C library supports
2075 @code{__cxa_atexit}.
2077 @item -fno-use-cxa-get-exception-ptr
2078 @opindex fno-use-cxa-get-exception-ptr
2079 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2080 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2081 if the runtime routine is not available.
2083 @item -fvisibility-inlines-hidden
2084 @opindex fvisibility-inlines-hidden
2085 This switch declares that the user does not attempt to compare
2086 pointers to inline methods where the addresses of the two functions
2087 were taken in different shared objects.
2089 The effect of this is that GCC may, effectively, mark inline methods with
2090 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2091 appear in the export table of a DSO and do not require a PLT indirection
2092 when used within the DSO@. Enabling this option can have a dramatic effect
2093 on load and link times of a DSO as it massively reduces the size of the
2094 dynamic export table when the library makes heavy use of templates.
2096 The behavior of this switch is not quite the same as marking the
2097 methods as hidden directly, because it does not affect static variables
2098 local to the function or cause the compiler to deduce that
2099 the function is defined in only one shared object.
2101 You may mark a method as having a visibility explicitly to negate the
2102 effect of the switch for that method. For example, if you do want to
2103 compare pointers to a particular inline method, you might mark it as
2104 having default visibility. Marking the enclosing class with explicit
2105 visibility will have no effect.
2107 Explicitly instantiated inline methods are unaffected by this option
2108 as their linkage might otherwise cross a shared library boundary.
2109 @xref{Template Instantiation}.
2111 @item -fvisibility-ms-compat
2112 @opindex fvisibility-ms-compat
2113 This flag attempts to use visibility settings to make GCC's C++
2114 linkage model compatible with that of Microsoft Visual Studio.
2116 The flag makes these changes to GCC's linkage model:
2120 It sets the default visibility to @code{hidden}, like
2121 @option{-fvisibility=hidden}.
2124 Types, but not their members, are not hidden by default.
2127 The One Definition Rule is relaxed for types without explicit
2128 visibility specifications which are defined in more than one different
2129 shared object: those declarations are permitted if they would have
2130 been permitted when this option was not used.
2133 In new code it is better to use @option{-fvisibility=hidden} and
2134 export those classes which are intended to be externally visible.
2135 Unfortunately it is possible for code to rely, perhaps accidentally,
2136 on the Visual Studio behavior.
2138 Among the consequences of these changes are that static data members
2139 of the same type with the same name but defined in different shared
2140 objects will be different, so changing one will not change the other;
2141 and that pointers to function members defined in different shared
2142 objects may not compare equal. When this flag is given, it is a
2143 violation of the ODR to define types with the same name differently.
2147 Do not use weak symbol support, even if it is provided by the linker.
2148 By default, G++ will use weak symbols if they are available. This
2149 option exists only for testing, and should not be used by end-users;
2150 it will result in inferior code and has no benefits. This option may
2151 be removed in a future release of G++.
2155 Do not search for header files in the standard directories specific to
2156 C++, but do still search the other standard directories. (This option
2157 is used when building the C++ library.)
2160 In addition, these optimization, warning, and code generation options
2161 have meanings only for C++ programs:
2164 @item -fno-default-inline
2165 @opindex fno-default-inline
2166 Do not assume @samp{inline} for functions defined inside a class scope.
2167 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2168 functions will have linkage like inline functions; they just won't be
2171 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2174 Warn when G++ generates code that is probably not compatible with the
2175 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2176 all such cases, there are probably some cases that are not warned about,
2177 even though G++ is generating incompatible code. There may also be
2178 cases where warnings are emitted even though the code that is generated
2181 You should rewrite your code to avoid these warnings if you are
2182 concerned about the fact that code generated by G++ may not be binary
2183 compatible with code generated by other compilers.
2185 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2190 A template with a non-type template parameter of reference type is
2191 mangled incorrectly:
2194 template <int &> struct S @{@};
2198 This is fixed in @option{-fabi-version=3}.
2201 SIMD vector types declared using @code{__attribute ((vector_size))} are
2202 mangled in a non-standard way that does not allow for overloading of
2203 functions taking vectors of different sizes.
2205 The mangling is changed in @option{-fabi-version=4}.
2208 The known incompatibilities in @option{-fabi-version=1} include:
2213 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2214 pack data into the same byte as a base class. For example:
2217 struct A @{ virtual void f(); int f1 : 1; @};
2218 struct B : public A @{ int f2 : 1; @};
2222 In this case, G++ will place @code{B::f2} into the same byte
2223 as@code{A::f1}; other compilers will not. You can avoid this problem
2224 by explicitly padding @code{A} so that its size is a multiple of the
2225 byte size on your platform; that will cause G++ and other compilers to
2226 layout @code{B} identically.
2229 Incorrect handling of tail-padding for virtual bases. G++ does not use
2230 tail padding when laying out virtual bases. For example:
2233 struct A @{ virtual void f(); char c1; @};
2234 struct B @{ B(); char c2; @};
2235 struct C : public A, public virtual B @{@};
2239 In this case, G++ will not place @code{B} into the tail-padding for
2240 @code{A}; other compilers will. You can avoid this problem by
2241 explicitly padding @code{A} so that its size is a multiple of its
2242 alignment (ignoring virtual base classes); that will cause G++ and other
2243 compilers to layout @code{C} identically.
2246 Incorrect handling of bit-fields with declared widths greater than that
2247 of their underlying types, when the bit-fields appear in a union. For
2251 union U @{ int i : 4096; @};
2255 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2256 union too small by the number of bits in an @code{int}.
2259 Empty classes can be placed at incorrect offsets. For example:
2269 struct C : public B, public A @{@};
2273 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2274 it should be placed at offset zero. G++ mistakenly believes that the
2275 @code{A} data member of @code{B} is already at offset zero.
2278 Names of template functions whose types involve @code{typename} or
2279 template template parameters can be mangled incorrectly.
2282 template <typename Q>
2283 void f(typename Q::X) @{@}
2285 template <template <typename> class Q>
2286 void f(typename Q<int>::X) @{@}
2290 Instantiations of these templates may be mangled incorrectly.
2294 It also warns psABI related changes. The known psABI changes at this
2300 For SYSV/x86-64, when passing union with long double, it is changed to
2301 pass in memory as specified in psABI. For example:
2311 @code{union U} will always be passed in memory.
2315 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2316 @opindex Wctor-dtor-privacy
2317 @opindex Wno-ctor-dtor-privacy
2318 Warn when a class seems unusable because all the constructors or
2319 destructors in that class are private, and it has neither friends nor
2320 public static member functions.
2322 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2324 @opindex Wno-noexcept
2325 Warn when a noexcept-expression evaluates to false because of a call
2326 to a function that does not have a non-throwing exception
2327 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2328 the compiler to never throw an exception.
2330 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2331 @opindex Wnon-virtual-dtor
2332 @opindex Wno-non-virtual-dtor
2333 Warn when a class has virtual functions and accessible non-virtual
2334 destructor, in which case it would be possible but unsafe to delete
2335 an instance of a derived class through a pointer to the base class.
2336 This warning is also enabled if -Weffc++ is specified.
2338 @item -Wreorder @r{(C++ and Objective-C++ only)}
2340 @opindex Wno-reorder
2341 @cindex reordering, warning
2342 @cindex warning for reordering of member initializers
2343 Warn when the order of member initializers given in the code does not
2344 match the order in which they must be executed. For instance:
2350 A(): j (0), i (1) @{ @}
2354 The compiler will rearrange the member initializers for @samp{i}
2355 and @samp{j} to match the declaration order of the members, emitting
2356 a warning to that effect. This warning is enabled by @option{-Wall}.
2359 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2362 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2365 Warn about violations of the following style guidelines from Scott Meyers'
2366 @cite{Effective C++} book:
2370 Item 11: Define a copy constructor and an assignment operator for classes
2371 with dynamically allocated memory.
2374 Item 12: Prefer initialization to assignment in constructors.
2377 Item 14: Make destructors virtual in base classes.
2380 Item 15: Have @code{operator=} return a reference to @code{*this}.
2383 Item 23: Don't try to return a reference when you must return an object.
2387 Also warn about violations of the following style guidelines from
2388 Scott Meyers' @cite{More Effective C++} book:
2392 Item 6: Distinguish between prefix and postfix forms of increment and
2393 decrement operators.
2396 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2400 When selecting this option, be aware that the standard library
2401 headers do not obey all of these guidelines; use @samp{grep -v}
2402 to filter out those warnings.
2404 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2405 @opindex Wstrict-null-sentinel
2406 @opindex Wno-strict-null-sentinel
2407 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2408 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2409 to @code{__null}. Although it is a null pointer constant not a null pointer,
2410 it is guaranteed to be of the same size as a pointer. But this use is
2411 not portable across different compilers.
2413 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2414 @opindex Wno-non-template-friend
2415 @opindex Wnon-template-friend
2416 Disable warnings when non-templatized friend functions are declared
2417 within a template. Since the advent of explicit template specification
2418 support in G++, if the name of the friend is an unqualified-id (i.e.,
2419 @samp{friend foo(int)}), the C++ language specification demands that the
2420 friend declare or define an ordinary, nontemplate function. (Section
2421 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2422 could be interpreted as a particular specialization of a templatized
2423 function. Because this non-conforming behavior is no longer the default
2424 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2425 check existing code for potential trouble spots and is on by default.
2426 This new compiler behavior can be turned off with
2427 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2428 but disables the helpful warning.
2430 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2431 @opindex Wold-style-cast
2432 @opindex Wno-old-style-cast
2433 Warn if an old-style (C-style) cast to a non-void type is used within
2434 a C++ program. The new-style casts (@samp{dynamic_cast},
2435 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2436 less vulnerable to unintended effects and much easier to search for.
2438 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2439 @opindex Woverloaded-virtual
2440 @opindex Wno-overloaded-virtual
2441 @cindex overloaded virtual function, warning
2442 @cindex warning for overloaded virtual function
2443 Warn when a function declaration hides virtual functions from a
2444 base class. For example, in:
2451 struct B: public A @{
2456 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2464 will fail to compile.
2466 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2467 @opindex Wno-pmf-conversions
2468 @opindex Wpmf-conversions
2469 Disable the diagnostic for converting a bound pointer to member function
2472 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2473 @opindex Wsign-promo
2474 @opindex Wno-sign-promo
2475 Warn when overload resolution chooses a promotion from unsigned or
2476 enumerated type to a signed type, over a conversion to an unsigned type of
2477 the same size. Previous versions of G++ would try to preserve
2478 unsignedness, but the standard mandates the current behavior.
2483 A& operator = (int);
2493 In this example, G++ will synthesize a default @samp{A& operator =
2494 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2497 @node Objective-C and Objective-C++ Dialect Options
2498 @section Options Controlling Objective-C and Objective-C++ Dialects
2500 @cindex compiler options, Objective-C and Objective-C++
2501 @cindex Objective-C and Objective-C++ options, command line
2502 @cindex options, Objective-C and Objective-C++
2503 (NOTE: This manual does not describe the Objective-C and Objective-C++
2504 languages themselves. @xref{Standards,,Language Standards
2505 Supported by GCC}, for references.)
2507 This section describes the command-line options that are only meaningful
2508 for Objective-C and Objective-C++ programs, but you can also use most of
2509 the language-independent GNU compiler options.
2510 For example, you might compile a file @code{some_class.m} like this:
2513 gcc -g -fgnu-runtime -O -c some_class.m
2517 In this example, @option{-fgnu-runtime} is an option meant only for
2518 Objective-C and Objective-C++ programs; you can use the other options with
2519 any language supported by GCC@.
2521 Note that since Objective-C is an extension of the C language, Objective-C
2522 compilations may also use options specific to the C front-end (e.g.,
2523 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2524 C++-specific options (e.g., @option{-Wabi}).
2526 Here is a list of options that are @emph{only} for compiling Objective-C
2527 and Objective-C++ programs:
2530 @item -fconstant-string-class=@var{class-name}
2531 @opindex fconstant-string-class
2532 Use @var{class-name} as the name of the class to instantiate for each
2533 literal string specified with the syntax @code{@@"@dots{}"}. The default
2534 class name is @code{NXConstantString} if the GNU runtime is being used, and
2535 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2536 @option{-fconstant-cfstrings} option, if also present, will override the
2537 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2538 to be laid out as constant CoreFoundation strings.
2541 @opindex fgnu-runtime
2542 Generate object code compatible with the standard GNU Objective-C
2543 runtime. This is the default for most types of systems.
2545 @item -fnext-runtime
2546 @opindex fnext-runtime
2547 Generate output compatible with the NeXT runtime. This is the default
2548 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2549 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2552 @item -fno-nil-receivers
2553 @opindex fno-nil-receivers
2554 Assume that all Objective-C message dispatches (@code{[receiver
2555 message:arg]}) in this translation unit ensure that the receiver is
2556 not @code{nil}. This allows for more efficient entry points in the
2557 runtime to be used. This option is only available in conjunction with
2558 the NeXT runtime and ABI version 0 or 1.
2560 @item -fobjc-abi-version=@var{n}
2561 @opindex fobjc-abi-version
2562 Use version @var{n} of the Objective-C ABI for the selected runtime.
2563 This option is currently supported only for the NeXT runtime. In that
2564 case, Version 0 is the traditional (32-bit) ABI without support for
2565 properties and other Objective-C 2.0 additions. Version 1 is the
2566 traditional (32-bit) ABI with support for properties and other
2567 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2568 nothing is specified, the default is Version 0 on 32-bit target
2569 machines, and Version 2 on 64-bit target machines.
2571 @item -fobjc-call-cxx-cdtors
2572 @opindex fobjc-call-cxx-cdtors
2573 For each Objective-C class, check if any of its instance variables is a
2574 C++ object with a non-trivial default constructor. If so, synthesize a
2575 special @code{- (id) .cxx_construct} instance method that will run
2576 non-trivial default constructors on any such instance variables, in order,
2577 and then return @code{self}. Similarly, check if any instance variable
2578 is a C++ object with a non-trivial destructor, and if so, synthesize a
2579 special @code{- (void) .cxx_destruct} method that will run
2580 all such default destructors, in reverse order.
2582 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2583 methods thusly generated will only operate on instance variables
2584 declared in the current Objective-C class, and not those inherited
2585 from superclasses. It is the responsibility of the Objective-C
2586 runtime to invoke all such methods in an object's inheritance
2587 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2588 by the runtime immediately after a new object instance is allocated;
2589 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2590 before the runtime deallocates an object instance.
2592 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2593 support for invoking the @code{- (id) .cxx_construct} and
2594 @code{- (void) .cxx_destruct} methods.
2596 @item -fobjc-direct-dispatch
2597 @opindex fobjc-direct-dispatch
2598 Allow fast jumps to the message dispatcher. On Darwin this is
2599 accomplished via the comm page.
2601 @item -fobjc-exceptions
2602 @opindex fobjc-exceptions
2603 Enable syntactic support for structured exception handling in
2604 Objective-C, similar to what is offered by C++ and Java. This option
2605 is required to use the Objective-C keywords @code{@@try},
2606 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2607 @code{@@synchronized}. This option is available with both the GNU
2608 runtime and the NeXT runtime (but not available in conjunction with
2609 the NeXT runtime on Mac OS X 10.2 and earlier).
2613 Enable garbage collection (GC) in Objective-C and Objective-C++
2614 programs. This option is only available with the NeXT runtime; the
2615 GNU runtime has a different garbage collection implementation that
2616 does not require special compiler flags.
2618 @item -fobjc-nilcheck
2619 @opindex fobjc-nilcheck
2620 For the NeXT runtime with version 2 of the ABI, check for a nil
2621 receiver in method invocations before doing the actual method call.
2622 This is the default and can be disabled using
2623 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2624 checked for nil in this way no matter what this flag is set to.
2625 Currently this flag does nothing when the GNU runtime, or an older
2626 version of the NeXT runtime ABI, is used.
2628 @item -fobjc-std=objc1
2630 Conform to the language syntax of Objective-C 1.0, the language
2631 recognized by GCC 4.0. This only affects the Objective-C additions to
2632 the C/C++ language; it does not affect conformance to C/C++ standards,
2633 which is controlled by the separate C/C++ dialect option flags. When
2634 this option is used with the Objective-C or Objective-C++ compiler,
2635 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2636 This is useful if you need to make sure that your Objective-C code can
2637 be compiled with older versions of GCC.
2639 @item -freplace-objc-classes
2640 @opindex freplace-objc-classes
2641 Emit a special marker instructing @command{ld(1)} not to statically link in
2642 the resulting object file, and allow @command{dyld(1)} to load it in at
2643 run time instead. This is used in conjunction with the Fix-and-Continue
2644 debugging mode, where the object file in question may be recompiled and
2645 dynamically reloaded in the course of program execution, without the need
2646 to restart the program itself. Currently, Fix-and-Continue functionality
2647 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2652 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2653 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2654 compile time) with static class references that get initialized at load time,
2655 which improves run-time performance. Specifying the @option{-fzero-link} flag
2656 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2657 to be retained. This is useful in Zero-Link debugging mode, since it allows
2658 for individual class implementations to be modified during program execution.
2659 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2660 regardless of command line options.
2664 Dump interface declarations for all classes seen in the source file to a
2665 file named @file{@var{sourcename}.decl}.
2667 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2668 @opindex Wassign-intercept
2669 @opindex Wno-assign-intercept
2670 Warn whenever an Objective-C assignment is being intercepted by the
2673 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2674 @opindex Wno-protocol
2676 If a class is declared to implement a protocol, a warning is issued for
2677 every method in the protocol that is not implemented by the class. The
2678 default behavior is to issue a warning for every method not explicitly
2679 implemented in the class, even if a method implementation is inherited
2680 from the superclass. If you use the @option{-Wno-protocol} option, then
2681 methods inherited from the superclass are considered to be implemented,
2682 and no warning is issued for them.
2684 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2686 @opindex Wno-selector
2687 Warn if multiple methods of different types for the same selector are
2688 found during compilation. The check is performed on the list of methods
2689 in the final stage of compilation. Additionally, a check is performed
2690 for each selector appearing in a @code{@@selector(@dots{})}
2691 expression, and a corresponding method for that selector has been found
2692 during compilation. Because these checks scan the method table only at
2693 the end of compilation, these warnings are not produced if the final
2694 stage of compilation is not reached, for example because an error is
2695 found during compilation, or because the @option{-fsyntax-only} option is
2698 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2699 @opindex Wstrict-selector-match
2700 @opindex Wno-strict-selector-match
2701 Warn if multiple methods with differing argument and/or return types are
2702 found for a given selector when attempting to send a message using this
2703 selector to a receiver of type @code{id} or @code{Class}. When this flag
2704 is off (which is the default behavior), the compiler will omit such warnings
2705 if any differences found are confined to types which share the same size
2708 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2709 @opindex Wundeclared-selector
2710 @opindex Wno-undeclared-selector
2711 Warn if a @code{@@selector(@dots{})} expression referring to an
2712 undeclared selector is found. A selector is considered undeclared if no
2713 method with that name has been declared before the
2714 @code{@@selector(@dots{})} expression, either explicitly in an
2715 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2716 an @code{@@implementation} section. This option always performs its
2717 checks as soon as a @code{@@selector(@dots{})} expression is found,
2718 while @option{-Wselector} only performs its checks in the final stage of
2719 compilation. This also enforces the coding style convention
2720 that methods and selectors must be declared before being used.
2722 @item -print-objc-runtime-info
2723 @opindex print-objc-runtime-info
2724 Generate C header describing the largest structure that is passed by
2729 @node Language Independent Options
2730 @section Options to Control Diagnostic Messages Formatting
2731 @cindex options to control diagnostics formatting
2732 @cindex diagnostic messages
2733 @cindex message formatting
2735 Traditionally, diagnostic messages have been formatted irrespective of
2736 the output device's aspect (e.g.@: its width, @dots{}). The options described
2737 below can be used to control the diagnostic messages formatting
2738 algorithm, e.g.@: how many characters per line, how often source location
2739 information should be reported. Right now, only the C++ front end can
2740 honor these options. However it is expected, in the near future, that
2741 the remaining front ends would be able to digest them correctly.
2744 @item -fmessage-length=@var{n}
2745 @opindex fmessage-length
2746 Try to format error messages so that they fit on lines of about @var{n}
2747 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2748 the front ends supported by GCC@. If @var{n} is zero, then no
2749 line-wrapping will be done; each error message will appear on a single
2752 @opindex fdiagnostics-show-location
2753 @item -fdiagnostics-show-location=once
2754 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2755 reporter to emit @emph{once} source location information; that is, in
2756 case the message is too long to fit on a single physical line and has to
2757 be wrapped, the source location won't be emitted (as prefix) again,
2758 over and over, in subsequent continuation lines. This is the default
2761 @item -fdiagnostics-show-location=every-line
2762 Only meaningful in line-wrapping mode. Instructs the diagnostic
2763 messages reporter to emit the same source location information (as
2764 prefix) for physical lines that result from the process of breaking
2765 a message which is too long to fit on a single line.
2767 @item -fno-diagnostics-show-option
2768 @opindex fno-diagnostics-show-option
2769 @opindex fdiagnostics-show-option
2770 By default, each diagnostic emitted includes text which indicates the
2771 command line option that directly controls the diagnostic (if such an
2772 option is known to the diagnostic machinery). Specifying the
2773 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2775 @item -Wcoverage-mismatch
2776 @opindex Wcoverage-mismatch
2777 Warn if feedback profiles do not match when using the
2778 @option{-fprofile-use} option.
2779 If a source file was changed between @option{-fprofile-gen} and
2780 @option{-fprofile-use}, the files with the profile feedback can fail
2781 to match the source file and GCC can not use the profile feedback
2782 information. By default, this warning is enabled and is treated as an
2783 error. @option{-Wno-coverage-mismatch} can be used to disable the
2784 warning or @option{-Wno-error=coverage-mismatch} can be used to
2785 disable the error. Disable the error for this warning can result in
2786 poorly optimized code, so disabling the error is useful only in the
2787 case of very minor changes such as bug fixes to an existing code-base.
2788 Completely disabling the warning is not recommended.
2792 @node Warning Options
2793 @section Options to Request or Suppress Warnings
2794 @cindex options to control warnings
2795 @cindex warning messages
2796 @cindex messages, warning
2797 @cindex suppressing warnings
2799 Warnings are diagnostic messages that report constructions which
2800 are not inherently erroneous but which are risky or suggest there
2801 may have been an error.
2803 The following language-independent options do not enable specific
2804 warnings but control the kinds of diagnostics produced by GCC.
2807 @cindex syntax checking
2809 @opindex fsyntax-only
2810 Check the code for syntax errors, but don't do anything beyond that.
2812 @item -fmax-errors=@var{n}
2813 @opindex fmax-errors
2814 Limits the maximum number of error messages to @var{n}, at which point
2815 GCC bails out rather than attempting to continue processing the source
2816 code. If @var{n} is 0 (the default), there is no limit on the number
2817 of error messages produced. If @option{-Wfatal-errors} is also
2818 specified, then @option{-Wfatal-errors} takes precedence over this
2823 Inhibit all warning messages.
2828 Make all warnings into errors.
2833 Make the specified warning into an error. The specifier for a warning
2834 is appended, for example @option{-Werror=switch} turns the warnings
2835 controlled by @option{-Wswitch} into errors. This switch takes a
2836 negative form, to be used to negate @option{-Werror} for specific
2837 warnings, for example @option{-Wno-error=switch} makes
2838 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2841 The warning message for each controllable warning includes the
2842 option which controls the warning. That option can then be used with
2843 @option{-Werror=} and @option{-Wno-error=} as described above.
2844 (Printing of the option in the warning message can be disabled using the
2845 @option{-fno-diagnostics-show-option} flag.)
2847 Note that specifying @option{-Werror=}@var{foo} automatically implies
2848 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2851 @item -Wfatal-errors
2852 @opindex Wfatal-errors
2853 @opindex Wno-fatal-errors
2854 This option causes the compiler to abort compilation on the first error
2855 occurred rather than trying to keep going and printing further error
2860 You can request many specific warnings with options beginning
2861 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2862 implicit declarations. Each of these specific warning options also
2863 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2864 example, @option{-Wno-implicit}. This manual lists only one of the
2865 two forms, whichever is not the default. For further,
2866 language-specific options also refer to @ref{C++ Dialect Options} and
2867 @ref{Objective-C and Objective-C++ Dialect Options}.
2869 When an unrecognized warning option is requested (e.g.,
2870 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2871 that the option is not recognized. However, if the @option{-Wno-} form
2872 is used, the behavior is slightly different: No diagnostic will be
2873 produced for @option{-Wno-unknown-warning} unless other diagnostics
2874 are being produced. This allows the use of new @option{-Wno-} options
2875 with old compilers, but if something goes wrong, the compiler will
2876 warn that an unrecognized option was used.
2881 Issue all the warnings demanded by strict ISO C and ISO C++;
2882 reject all programs that use forbidden extensions, and some other
2883 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2884 version of the ISO C standard specified by any @option{-std} option used.
2886 Valid ISO C and ISO C++ programs should compile properly with or without
2887 this option (though a rare few will require @option{-ansi} or a
2888 @option{-std} option specifying the required version of ISO C)@. However,
2889 without this option, certain GNU extensions and traditional C and C++
2890 features are supported as well. With this option, they are rejected.
2892 @option{-pedantic} does not cause warning messages for use of the
2893 alternate keywords whose names begin and end with @samp{__}. Pedantic
2894 warnings are also disabled in the expression that follows
2895 @code{__extension__}. However, only system header files should use
2896 these escape routes; application programs should avoid them.
2897 @xref{Alternate Keywords}.
2899 Some users try to use @option{-pedantic} to check programs for strict ISO
2900 C conformance. They soon find that it does not do quite what they want:
2901 it finds some non-ISO practices, but not all---only those for which
2902 ISO C @emph{requires} a diagnostic, and some others for which
2903 diagnostics have been added.
2905 A feature to report any failure to conform to ISO C might be useful in
2906 some instances, but would require considerable additional work and would
2907 be quite different from @option{-pedantic}. We don't have plans to
2908 support such a feature in the near future.
2910 Where the standard specified with @option{-std} represents a GNU
2911 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2912 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2913 extended dialect is based. Warnings from @option{-pedantic} are given
2914 where they are required by the base standard. (It would not make sense
2915 for such warnings to be given only for features not in the specified GNU
2916 C dialect, since by definition the GNU dialects of C include all
2917 features the compiler supports with the given option, and there would be
2918 nothing to warn about.)
2920 @item -pedantic-errors
2921 @opindex pedantic-errors
2922 Like @option{-pedantic}, except that errors are produced rather than
2928 This enables all the warnings about constructions that some users
2929 consider questionable, and that are easy to avoid (or modify to
2930 prevent the warning), even in conjunction with macros. This also
2931 enables some language-specific warnings described in @ref{C++ Dialect
2932 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2934 @option{-Wall} turns on the following warning flags:
2936 @gccoptlist{-Waddress @gol
2937 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2939 -Wchar-subscripts @gol
2940 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2941 -Wimplicit-int @r{(C and Objective-C only)} @gol
2942 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2945 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2946 -Wmissing-braces @gol
2952 -Wsequence-point @gol
2953 -Wsign-compare @r{(only in C++)} @gol
2954 -Wstrict-aliasing @gol
2955 -Wstrict-overflow=1 @gol
2958 -Wuninitialized @gol
2959 -Wunknown-pragmas @gol
2960 -Wunused-function @gol
2963 -Wunused-variable @gol
2964 -Wvolatile-register-var @gol
2967 Note that some warning flags are not implied by @option{-Wall}. Some of
2968 them warn about constructions that users generally do not consider
2969 questionable, but which occasionally you might wish to check for;
2970 others warn about constructions that are necessary or hard to avoid in
2971 some cases, and there is no simple way to modify the code to suppress
2972 the warning. Some of them are enabled by @option{-Wextra} but many of
2973 them must be enabled individually.
2979 This enables some extra warning flags that are not enabled by
2980 @option{-Wall}. (This option used to be called @option{-W}. The older
2981 name is still supported, but the newer name is more descriptive.)
2983 @gccoptlist{-Wclobbered @gol
2985 -Wignored-qualifiers @gol
2986 -Wmissing-field-initializers @gol
2987 -Wmissing-parameter-type @r{(C only)} @gol
2988 -Wold-style-declaration @r{(C only)} @gol
2989 -Woverride-init @gol
2992 -Wuninitialized @gol
2993 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2994 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2997 The option @option{-Wextra} also prints warning messages for the
3003 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3004 @samp{>}, or @samp{>=}.
3007 (C++ only) An enumerator and a non-enumerator both appear in a
3008 conditional expression.
3011 (C++ only) Ambiguous virtual bases.
3014 (C++ only) Subscripting an array which has been declared @samp{register}.
3017 (C++ only) Taking the address of a variable which has been declared
3021 (C++ only) A base class is not initialized in a derived class' copy
3026 @item -Wchar-subscripts
3027 @opindex Wchar-subscripts
3028 @opindex Wno-char-subscripts
3029 Warn if an array subscript has type @code{char}. This is a common cause
3030 of error, as programmers often forget that this type is signed on some
3032 This warning is enabled by @option{-Wall}.
3036 @opindex Wno-comment
3037 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3038 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3039 This warning is enabled by @option{-Wall}.
3042 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3044 Suppress warning messages emitted by @code{#warning} directives.
3046 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3047 @opindex Wdouble-promotion
3048 @opindex Wno-double-promotion
3049 Give a warning when a value of type @code{float} is implicitly
3050 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3051 floating-point unit implement @code{float} in hardware, but emulate
3052 @code{double} in software. On such a machine, doing computations
3053 using @code{double} values is much more expensive because of the
3054 overhead required for software emulation.
3056 It is easy to accidentally do computations with @code{double} because
3057 floating-point literals are implicitly of type @code{double}. For
3061 float area(float radius)
3063 return 3.14159 * radius * radius;
3067 the compiler will perform the entire computation with @code{double}
3068 because the floating-point literal is a @code{double}.
3073 @opindex ffreestanding
3074 @opindex fno-builtin
3075 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3076 the arguments supplied have types appropriate to the format string
3077 specified, and that the conversions specified in the format string make
3078 sense. This includes standard functions, and others specified by format
3079 attributes (@pxref{Function Attributes}), in the @code{printf},
3080 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3081 not in the C standard) families (or other target-specific families).
3082 Which functions are checked without format attributes having been
3083 specified depends on the standard version selected, and such checks of
3084 functions without the attribute specified are disabled by
3085 @option{-ffreestanding} or @option{-fno-builtin}.
3087 The formats are checked against the format features supported by GNU
3088 libc version 2.2. These include all ISO C90 and C99 features, as well
3089 as features from the Single Unix Specification and some BSD and GNU
3090 extensions. Other library implementations may not support all these
3091 features; GCC does not support warning about features that go beyond a
3092 particular library's limitations. However, if @option{-pedantic} is used
3093 with @option{-Wformat}, warnings will be given about format features not
3094 in the selected standard version (but not for @code{strfmon} formats,
3095 since those are not in any version of the C standard). @xref{C Dialect
3096 Options,,Options Controlling C Dialect}.
3098 Since @option{-Wformat} also checks for null format arguments for
3099 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3101 @option{-Wformat} is included in @option{-Wall}. For more control over some
3102 aspects of format checking, the options @option{-Wformat-y2k},
3103 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3104 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3105 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3108 @opindex Wformat-y2k
3109 @opindex Wno-format-y2k
3110 If @option{-Wformat} is specified, also warn about @code{strftime}
3111 formats which may yield only a two-digit year.
3113 @item -Wno-format-contains-nul
3114 @opindex Wno-format-contains-nul
3115 @opindex Wformat-contains-nul
3116 If @option{-Wformat} is specified, do not warn about format strings that
3119 @item -Wno-format-extra-args
3120 @opindex Wno-format-extra-args
3121 @opindex Wformat-extra-args
3122 If @option{-Wformat} is specified, do not warn about excess arguments to a
3123 @code{printf} or @code{scanf} format function. The C standard specifies
3124 that such arguments are ignored.
3126 Where the unused arguments lie between used arguments that are
3127 specified with @samp{$} operand number specifications, normally
3128 warnings are still given, since the implementation could not know what
3129 type to pass to @code{va_arg} to skip the unused arguments. However,
3130 in the case of @code{scanf} formats, this option will suppress the
3131 warning if the unused arguments are all pointers, since the Single
3132 Unix Specification says that such unused arguments are allowed.
3134 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3135 @opindex Wno-format-zero-length
3136 @opindex Wformat-zero-length
3137 If @option{-Wformat} is specified, do not warn about zero-length formats.
3138 The C standard specifies that zero-length formats are allowed.
3140 @item -Wformat-nonliteral
3141 @opindex Wformat-nonliteral
3142 @opindex Wno-format-nonliteral
3143 If @option{-Wformat} is specified, also warn if the format string is not a
3144 string literal and so cannot be checked, unless the format function
3145 takes its format arguments as a @code{va_list}.
3147 @item -Wformat-security
3148 @opindex Wformat-security
3149 @opindex Wno-format-security
3150 If @option{-Wformat} is specified, also warn about uses of format
3151 functions that represent possible security problems. At present, this
3152 warns about calls to @code{printf} and @code{scanf} functions where the
3153 format string is not a string literal and there are no format arguments,
3154 as in @code{printf (foo);}. This may be a security hole if the format
3155 string came from untrusted input and contains @samp{%n}. (This is
3156 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3157 in future warnings may be added to @option{-Wformat-security} that are not
3158 included in @option{-Wformat-nonliteral}.)
3162 @opindex Wno-format=2
3163 Enable @option{-Wformat} plus format checks not included in
3164 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3165 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3167 @item -Wnonnull @r{(C and Objective-C only)}
3169 @opindex Wno-nonnull
3170 Warn about passing a null pointer for arguments marked as
3171 requiring a non-null value by the @code{nonnull} function attribute.
3173 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3174 can be disabled with the @option{-Wno-nonnull} option.
3176 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3178 @opindex Wno-init-self
3179 Warn about uninitialized variables which are initialized with themselves.
3180 Note this option can only be used with the @option{-Wuninitialized} option.
3182 For example, GCC will warn about @code{i} being uninitialized in the
3183 following snippet only when @option{-Winit-self} has been specified:
3194 @item -Wimplicit-int @r{(C and Objective-C only)}
3195 @opindex Wimplicit-int
3196 @opindex Wno-implicit-int
3197 Warn when a declaration does not specify a type.
3198 This warning is enabled by @option{-Wall}.
3200 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3201 @opindex Wimplicit-function-declaration
3202 @opindex Wno-implicit-function-declaration
3203 Give a warning whenever a function is used before being declared. In
3204 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3205 enabled by default and it is made into an error by
3206 @option{-pedantic-errors}. This warning is also enabled by
3209 @item -Wimplicit @r{(C and Objective-C only)}
3211 @opindex Wno-implicit
3212 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3213 This warning is enabled by @option{-Wall}.
3215 @item -Wignored-qualifiers @r{(C and C++ only)}
3216 @opindex Wignored-qualifiers
3217 @opindex Wno-ignored-qualifiers
3218 Warn if the return type of a function has a type qualifier
3219 such as @code{const}. For ISO C such a type qualifier has no effect,
3220 since the value returned by a function is not an lvalue.
3221 For C++, the warning is only emitted for scalar types or @code{void}.
3222 ISO C prohibits qualified @code{void} return types on function
3223 definitions, so such return types always receive a warning
3224 even without this option.
3226 This warning is also enabled by @option{-Wextra}.
3231 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3232 a function with external linkage, returning int, taking either zero
3233 arguments, two, or three arguments of appropriate types. This warning
3234 is enabled by default in C++ and is enabled by either @option{-Wall}
3235 or @option{-pedantic}.
3237 @item -Wmissing-braces
3238 @opindex Wmissing-braces
3239 @opindex Wno-missing-braces
3240 Warn if an aggregate or union initializer is not fully bracketed. In
3241 the following example, the initializer for @samp{a} is not fully
3242 bracketed, but that for @samp{b} is fully bracketed.
3245 int a[2][2] = @{ 0, 1, 2, 3 @};
3246 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3249 This warning is enabled by @option{-Wall}.
3251 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3252 @opindex Wmissing-include-dirs
3253 @opindex Wno-missing-include-dirs
3254 Warn if a user-supplied include directory does not exist.
3257 @opindex Wparentheses
3258 @opindex Wno-parentheses
3259 Warn if parentheses are omitted in certain contexts, such
3260 as when there is an assignment in a context where a truth value
3261 is expected, or when operators are nested whose precedence people
3262 often get confused about.
3264 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3265 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3266 interpretation from that of ordinary mathematical notation.
3268 Also warn about constructions where there may be confusion to which
3269 @code{if} statement an @code{else} branch belongs. Here is an example of
3284 In C/C++, every @code{else} branch belongs to the innermost possible
3285 @code{if} statement, which in this example is @code{if (b)}. This is
3286 often not what the programmer expected, as illustrated in the above
3287 example by indentation the programmer chose. When there is the
3288 potential for this confusion, GCC will issue a warning when this flag
3289 is specified. To eliminate the warning, add explicit braces around
3290 the innermost @code{if} statement so there is no way the @code{else}
3291 could belong to the enclosing @code{if}. The resulting code would
3308 Also warn for dangerous uses of the
3309 ?: with omitted middle operand GNU extension. When the condition
3310 in the ?: operator is a boolean expression the omitted value will
3311 be always 1. Often the user expects it to be a value computed
3312 inside the conditional expression instead.
3314 This warning is enabled by @option{-Wall}.
3316 @item -Wsequence-point
3317 @opindex Wsequence-point
3318 @opindex Wno-sequence-point
3319 Warn about code that may have undefined semantics because of violations
3320 of sequence point rules in the C and C++ standards.
3322 The C and C++ standards defines the order in which expressions in a C/C++
3323 program are evaluated in terms of @dfn{sequence points}, which represent
3324 a partial ordering between the execution of parts of the program: those
3325 executed before the sequence point, and those executed after it. These
3326 occur after the evaluation of a full expression (one which is not part
3327 of a larger expression), after the evaluation of the first operand of a
3328 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3329 function is called (but after the evaluation of its arguments and the
3330 expression denoting the called function), and in certain other places.
3331 Other than as expressed by the sequence point rules, the order of
3332 evaluation of subexpressions of an expression is not specified. All
3333 these rules describe only a partial order rather than a total order,
3334 since, for example, if two functions are called within one expression
3335 with no sequence point between them, the order in which the functions
3336 are called is not specified. However, the standards committee have
3337 ruled that function calls do not overlap.
3339 It is not specified when between sequence points modifications to the
3340 values of objects take effect. Programs whose behavior depends on this
3341 have undefined behavior; the C and C++ standards specify that ``Between
3342 the previous and next sequence point an object shall have its stored
3343 value modified at most once by the evaluation of an expression.
3344 Furthermore, the prior value shall be read only to determine the value
3345 to be stored.''. If a program breaks these rules, the results on any
3346 particular implementation are entirely unpredictable.
3348 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3349 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3350 diagnosed by this option, and it may give an occasional false positive
3351 result, but in general it has been found fairly effective at detecting
3352 this sort of problem in programs.
3354 The standard is worded confusingly, therefore there is some debate
3355 over the precise meaning of the sequence point rules in subtle cases.
3356 Links to discussions of the problem, including proposed formal
3357 definitions, may be found on the GCC readings page, at
3358 @uref{http://gcc.gnu.org/@/readings.html}.
3360 This warning is enabled by @option{-Wall} for C and C++.
3363 @opindex Wreturn-type
3364 @opindex Wno-return-type
3365 Warn whenever a function is defined with a return-type that defaults
3366 to @code{int}. Also warn about any @code{return} statement with no
3367 return-value in a function whose return-type is not @code{void}
3368 (falling off the end of the function body is considered returning
3369 without a value), and about a @code{return} statement with an
3370 expression in a function whose return-type is @code{void}.
3372 For C++, a function without return type always produces a diagnostic
3373 message, even when @option{-Wno-return-type} is specified. The only
3374 exceptions are @samp{main} and functions defined in system headers.
3376 This warning is enabled by @option{-Wall}.
3381 Warn whenever a @code{switch} statement has an index of enumerated type
3382 and lacks a @code{case} for one or more of the named codes of that
3383 enumeration. (The presence of a @code{default} label prevents this
3384 warning.) @code{case} labels outside the enumeration range also
3385 provoke warnings when this option is used (even if there is a
3386 @code{default} label).
3387 This warning is enabled by @option{-Wall}.
3389 @item -Wswitch-default
3390 @opindex Wswitch-default
3391 @opindex Wno-switch-default
3392 Warn whenever a @code{switch} statement does not have a @code{default}
3396 @opindex Wswitch-enum
3397 @opindex Wno-switch-enum
3398 Warn whenever a @code{switch} statement has an index of enumerated type
3399 and lacks a @code{case} for one or more of the named codes of that
3400 enumeration. @code{case} labels outside the enumeration range also
3401 provoke warnings when this option is used. The only difference
3402 between @option{-Wswitch} and this option is that this option gives a
3403 warning about an omitted enumeration code even if there is a
3404 @code{default} label.
3406 @item -Wsync-nand @r{(C and C++ only)}
3408 @opindex Wno-sync-nand
3409 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3410 built-in functions are used. These functions changed semantics in GCC 4.4.
3414 @opindex Wno-trigraphs
3415 Warn if any trigraphs are encountered that might change the meaning of
3416 the program (trigraphs within comments are not warned about).
3417 This warning is enabled by @option{-Wall}.
3419 @item -Wunused-but-set-parameter
3420 @opindex Wunused-but-set-parameter
3421 @opindex Wno-unused-but-set-parameter
3422 Warn whenever a function parameter is assigned to, but otherwise unused
3423 (aside from its declaration).
3425 To suppress this warning use the @samp{unused} attribute
3426 (@pxref{Variable Attributes}).
3428 This warning is also enabled by @option{-Wunused} together with
3431 @item -Wunused-but-set-variable
3432 @opindex Wunused-but-set-variable
3433 @opindex Wno-unused-but-set-variable
3434 Warn whenever a local variable is assigned to, but otherwise unused
3435 (aside from its declaration).
3436 This warning is enabled by @option{-Wall}.
3438 To suppress this warning use the @samp{unused} attribute
3439 (@pxref{Variable Attributes}).
3441 This warning is also enabled by @option{-Wunused}, which is enabled
3444 @item -Wunused-function
3445 @opindex Wunused-function
3446 @opindex Wno-unused-function
3447 Warn whenever a static function is declared but not defined or a
3448 non-inline static function is unused.
3449 This warning is enabled by @option{-Wall}.
3451 @item -Wunused-label
3452 @opindex Wunused-label
3453 @opindex Wno-unused-label
3454 Warn whenever a label is declared but not used.
3455 This warning is enabled by @option{-Wall}.
3457 To suppress this warning use the @samp{unused} attribute
3458 (@pxref{Variable Attributes}).
3460 @item -Wunused-parameter
3461 @opindex Wunused-parameter
3462 @opindex Wno-unused-parameter
3463 Warn whenever a function parameter is unused aside from its declaration.
3465 To suppress this warning use the @samp{unused} attribute
3466 (@pxref{Variable Attributes}).
3468 @item -Wno-unused-result
3469 @opindex Wunused-result
3470 @opindex Wno-unused-result
3471 Do not warn if a caller of a function marked with attribute
3472 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3473 its return value. The default is @option{-Wunused-result}.
3475 @item -Wunused-variable
3476 @opindex Wunused-variable
3477 @opindex Wno-unused-variable
3478 Warn whenever a local variable or non-constant static variable is unused
3479 aside from its declaration.
3480 This warning is enabled by @option{-Wall}.
3482 To suppress this warning use the @samp{unused} attribute
3483 (@pxref{Variable Attributes}).
3485 @item -Wunused-value
3486 @opindex Wunused-value
3487 @opindex Wno-unused-value
3488 Warn whenever a statement computes a result that is explicitly not
3489 used. To suppress this warning cast the unused expression to
3490 @samp{void}. This includes an expression-statement or the left-hand
3491 side of a comma expression that contains no side effects. For example,
3492 an expression such as @samp{x[i,j]} will cause a warning, while
3493 @samp{x[(void)i,j]} will not.
3495 This warning is enabled by @option{-Wall}.
3500 All the above @option{-Wunused} options combined.
3502 In order to get a warning about an unused function parameter, you must
3503 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3504 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3506 @item -Wuninitialized
3507 @opindex Wuninitialized
3508 @opindex Wno-uninitialized
3509 Warn if an automatic variable is used without first being initialized
3510 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3511 warn if a non-static reference or non-static @samp{const} member
3512 appears in a class without constructors.
3514 If you want to warn about code which uses the uninitialized value of the
3515 variable in its own initializer, use the @option{-Winit-self} option.
3517 These warnings occur for individual uninitialized or clobbered
3518 elements of structure, union or array variables as well as for
3519 variables which are uninitialized or clobbered as a whole. They do
3520 not occur for variables or elements declared @code{volatile}. Because
3521 these warnings depend on optimization, the exact variables or elements
3522 for which there are warnings will depend on the precise optimization
3523 options and version of GCC used.
3525 Note that there may be no warning about a variable that is used only
3526 to compute a value that itself is never used, because such
3527 computations may be deleted by data flow analysis before the warnings
3530 These warnings are made optional because GCC is not smart
3531 enough to see all the reasons why the code might be correct
3532 despite appearing to have an error. Here is one example of how
3553 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3554 always initialized, but GCC doesn't know this. Here is
3555 another common case:
3560 if (change_y) save_y = y, y = new_y;
3562 if (change_y) y = save_y;
3567 This has no bug because @code{save_y} is used only if it is set.
3569 @cindex @code{longjmp} warnings
3570 This option also warns when a non-volatile automatic variable might be
3571 changed by a call to @code{longjmp}. These warnings as well are possible
3572 only in optimizing compilation.
3574 The compiler sees only the calls to @code{setjmp}. It cannot know
3575 where @code{longjmp} will be called; in fact, a signal handler could
3576 call it at any point in the code. As a result, you may get a warning
3577 even when there is in fact no problem because @code{longjmp} cannot
3578 in fact be called at the place which would cause a problem.
3580 Some spurious warnings can be avoided if you declare all the functions
3581 you use that never return as @code{noreturn}. @xref{Function
3584 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3586 @item -Wunknown-pragmas
3587 @opindex Wunknown-pragmas
3588 @opindex Wno-unknown-pragmas
3589 @cindex warning for unknown pragmas
3590 @cindex unknown pragmas, warning
3591 @cindex pragmas, warning of unknown
3592 Warn when a #pragma directive is encountered which is not understood by
3593 GCC@. If this command line option is used, warnings will even be issued
3594 for unknown pragmas in system header files. This is not the case if
3595 the warnings were only enabled by the @option{-Wall} command line option.
3598 @opindex Wno-pragmas
3600 Do not warn about misuses of pragmas, such as incorrect parameters,
3601 invalid syntax, or conflicts between pragmas. See also
3602 @samp{-Wunknown-pragmas}.
3604 @item -Wstrict-aliasing
3605 @opindex Wstrict-aliasing
3606 @opindex Wno-strict-aliasing
3607 This option is only active when @option{-fstrict-aliasing} is active.
3608 It warns about code which might break the strict aliasing rules that the
3609 compiler is using for optimization. The warning does not catch all
3610 cases, but does attempt to catch the more common pitfalls. It is
3611 included in @option{-Wall}.
3612 It is equivalent to @option{-Wstrict-aliasing=3}
3614 @item -Wstrict-aliasing=n
3615 @opindex Wstrict-aliasing=n
3616 @opindex Wno-strict-aliasing=n
3617 This option is only active when @option{-fstrict-aliasing} is active.
3618 It warns about code which might break the strict aliasing rules that the
3619 compiler is using for optimization.
3620 Higher levels correspond to higher accuracy (fewer false positives).
3621 Higher levels also correspond to more effort, similar to the way -O works.
3622 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3625 Level 1: Most aggressive, quick, least accurate.
3626 Possibly useful when higher levels
3627 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3628 false negatives. However, it has many false positives.
3629 Warns for all pointer conversions between possibly incompatible types,
3630 even if never dereferenced. Runs in the frontend only.
3632 Level 2: Aggressive, quick, not too precise.
3633 May still have many false positives (not as many as level 1 though),
3634 and few false negatives (but possibly more than level 1).
3635 Unlike level 1, it only warns when an address is taken. Warns about
3636 incomplete types. Runs in the frontend only.
3638 Level 3 (default for @option{-Wstrict-aliasing}):
3639 Should have very few false positives and few false
3640 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3641 Takes care of the common pun+dereference pattern in the frontend:
3642 @code{*(int*)&some_float}.
3643 If optimization is enabled, it also runs in the backend, where it deals
3644 with multiple statement cases using flow-sensitive points-to information.
3645 Only warns when the converted pointer is dereferenced.
3646 Does not warn about incomplete types.
3648 @item -Wstrict-overflow
3649 @itemx -Wstrict-overflow=@var{n}
3650 @opindex Wstrict-overflow
3651 @opindex Wno-strict-overflow
3652 This option is only active when @option{-fstrict-overflow} is active.
3653 It warns about cases where the compiler optimizes based on the
3654 assumption that signed overflow does not occur. Note that it does not
3655 warn about all cases where the code might overflow: it only warns
3656 about cases where the compiler implements some optimization. Thus
3657 this warning depends on the optimization level.
3659 An optimization which assumes that signed overflow does not occur is
3660 perfectly safe if the values of the variables involved are such that
3661 overflow never does, in fact, occur. Therefore this warning can
3662 easily give a false positive: a warning about code which is not
3663 actually a problem. To help focus on important issues, several
3664 warning levels are defined. No warnings are issued for the use of
3665 undefined signed overflow when estimating how many iterations a loop
3666 will require, in particular when determining whether a loop will be
3670 @item -Wstrict-overflow=1
3671 Warn about cases which are both questionable and easy to avoid. For
3672 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3673 compiler will simplify this to @code{1}. This level of
3674 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3675 are not, and must be explicitly requested.
3677 @item -Wstrict-overflow=2
3678 Also warn about other cases where a comparison is simplified to a
3679 constant. For example: @code{abs (x) >= 0}. This can only be
3680 simplified when @option{-fstrict-overflow} is in effect, because
3681 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3682 zero. @option{-Wstrict-overflow} (with no level) is the same as
3683 @option{-Wstrict-overflow=2}.
3685 @item -Wstrict-overflow=3
3686 Also warn about other cases where a comparison is simplified. For
3687 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3689 @item -Wstrict-overflow=4
3690 Also warn about other simplifications not covered by the above cases.
3691 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3693 @item -Wstrict-overflow=5
3694 Also warn about cases where the compiler reduces the magnitude of a
3695 constant involved in a comparison. For example: @code{x + 2 > y} will
3696 be simplified to @code{x + 1 >= y}. This is reported only at the
3697 highest warning level because this simplification applies to many
3698 comparisons, so this warning level will give a very large number of
3702 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3703 @opindex Wsuggest-attribute=
3704 @opindex Wno-suggest-attribute=
3705 Warn for cases where adding an attribute may be beneficial. The
3706 attributes currently supported are listed below.
3709 @item -Wsuggest-attribute=pure
3710 @itemx -Wsuggest-attribute=const
3711 @itemx -Wsuggest-attribute=noreturn
3712 @opindex Wsuggest-attribute=pure
3713 @opindex Wno-suggest-attribute=pure
3714 @opindex Wsuggest-attribute=const
3715 @opindex Wno-suggest-attribute=const
3716 @opindex Wsuggest-attribute=noreturn
3717 @opindex Wno-suggest-attribute=noreturn
3719 Warn about functions which might be candidates for attributes
3720 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3721 functions visible in other compilation units or (in the case of @code{pure} and
3722 @code{const}) if it cannot prove that the function returns normally. A function
3723 returns normally if it doesn't contain an infinite loop nor returns abnormally
3724 by throwing, calling @code{abort()} or trapping. This analysis requires option
3725 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3726 higher. Higher optimization levels improve the accuracy of the analysis.
3729 @item -Warray-bounds
3730 @opindex Wno-array-bounds
3731 @opindex Warray-bounds
3732 This option is only active when @option{-ftree-vrp} is active
3733 (default for @option{-O2} and above). It warns about subscripts to arrays
3734 that are always out of bounds. This warning is enabled by @option{-Wall}.
3736 @item -Wno-div-by-zero
3737 @opindex Wno-div-by-zero
3738 @opindex Wdiv-by-zero
3739 Do not warn about compile-time integer division by zero. Floating point
3740 division by zero is not warned about, as it can be a legitimate way of
3741 obtaining infinities and NaNs.
3743 @item -Wsystem-headers
3744 @opindex Wsystem-headers
3745 @opindex Wno-system-headers
3746 @cindex warnings from system headers
3747 @cindex system headers, warnings from
3748 Print warning messages for constructs found in system header files.
3749 Warnings from system headers are normally suppressed, on the assumption
3750 that they usually do not indicate real problems and would only make the
3751 compiler output harder to read. Using this command line option tells
3752 GCC to emit warnings from system headers as if they occurred in user
3753 code. However, note that using @option{-Wall} in conjunction with this
3754 option will @emph{not} warn about unknown pragmas in system
3755 headers---for that, @option{-Wunknown-pragmas} must also be used.
3758 @opindex Wtrampolines
3759 @opindex Wno-trampolines
3760 Warn about trampolines generated for pointers to nested functions.
3762 A trampoline is a small piece of data or code that is created at run
3763 time on the stack when the address of a nested function is taken, and
3764 is used to call the nested function indirectly. For some targets, it
3765 is made up of data only and thus requires no special treatment. But,
3766 for most targets, it is made up of code and thus requires the stack
3767 to be made executable in order for the program to work properly.
3770 @opindex Wfloat-equal
3771 @opindex Wno-float-equal
3772 Warn if floating point values are used in equality comparisons.
3774 The idea behind this is that sometimes it is convenient (for the
3775 programmer) to consider floating-point values as approximations to
3776 infinitely precise real numbers. If you are doing this, then you need
3777 to compute (by analyzing the code, or in some other way) the maximum or
3778 likely maximum error that the computation introduces, and allow for it
3779 when performing comparisons (and when producing output, but that's a
3780 different problem). In particular, instead of testing for equality, you
3781 would check to see whether the two values have ranges that overlap; and
3782 this is done with the relational operators, so equality comparisons are
3785 @item -Wtraditional @r{(C and Objective-C only)}
3786 @opindex Wtraditional
3787 @opindex Wno-traditional
3788 Warn about certain constructs that behave differently in traditional and
3789 ISO C@. Also warn about ISO C constructs that have no traditional C
3790 equivalent, and/or problematic constructs which should be avoided.
3794 Macro parameters that appear within string literals in the macro body.
3795 In traditional C macro replacement takes place within string literals,
3796 but does not in ISO C@.
3799 In traditional C, some preprocessor directives did not exist.
3800 Traditional preprocessors would only consider a line to be a directive
3801 if the @samp{#} appeared in column 1 on the line. Therefore
3802 @option{-Wtraditional} warns about directives that traditional C
3803 understands but would ignore because the @samp{#} does not appear as the
3804 first character on the line. It also suggests you hide directives like
3805 @samp{#pragma} not understood by traditional C by indenting them. Some
3806 traditional implementations would not recognize @samp{#elif}, so it
3807 suggests avoiding it altogether.
3810 A function-like macro that appears without arguments.
3813 The unary plus operator.
3816 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3817 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3818 constants.) Note, these suffixes appear in macros defined in the system
3819 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3820 Use of these macros in user code might normally lead to spurious
3821 warnings, however GCC's integrated preprocessor has enough context to
3822 avoid warning in these cases.
3825 A function declared external in one block and then used after the end of
3829 A @code{switch} statement has an operand of type @code{long}.
3832 A non-@code{static} function declaration follows a @code{static} one.
3833 This construct is not accepted by some traditional C compilers.
3836 The ISO type of an integer constant has a different width or
3837 signedness from its traditional type. This warning is only issued if
3838 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3839 typically represent bit patterns, are not warned about.
3842 Usage of ISO string concatenation is detected.
3845 Initialization of automatic aggregates.
3848 Identifier conflicts with labels. Traditional C lacks a separate
3849 namespace for labels.
3852 Initialization of unions. If the initializer is zero, the warning is
3853 omitted. This is done under the assumption that the zero initializer in
3854 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3855 initializer warnings and relies on default initialization to zero in the
3859 Conversions by prototypes between fixed/floating point values and vice
3860 versa. The absence of these prototypes when compiling with traditional
3861 C would cause serious problems. This is a subset of the possible
3862 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3865 Use of ISO C style function definitions. This warning intentionally is
3866 @emph{not} issued for prototype declarations or variadic functions
3867 because these ISO C features will appear in your code when using
3868 libiberty's traditional C compatibility macros, @code{PARAMS} and
3869 @code{VPARAMS}. This warning is also bypassed for nested functions
3870 because that feature is already a GCC extension and thus not relevant to
3871 traditional C compatibility.
3874 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3875 @opindex Wtraditional-conversion
3876 @opindex Wno-traditional-conversion
3877 Warn if a prototype causes a type conversion that is different from what
3878 would happen to the same argument in the absence of a prototype. This
3879 includes conversions of fixed point to floating and vice versa, and
3880 conversions changing the width or signedness of a fixed point argument
3881 except when the same as the default promotion.
3883 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3884 @opindex Wdeclaration-after-statement
3885 @opindex Wno-declaration-after-statement
3886 Warn when a declaration is found after a statement in a block. This
3887 construct, known from C++, was introduced with ISO C99 and is by default
3888 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3889 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3894 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3896 @item -Wno-endif-labels
3897 @opindex Wno-endif-labels
3898 @opindex Wendif-labels
3899 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3904 Warn whenever a local variable or type declaration shadows another variable,
3905 parameter, type, or class member (in C++), or whenever a built-in function
3906 is shadowed. Note that in C++, the compiler will not warn if a local variable
3907 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3909 @item -Wlarger-than=@var{len}
3910 @opindex Wlarger-than=@var{len}
3911 @opindex Wlarger-than-@var{len}
3912 Warn whenever an object of larger than @var{len} bytes is defined.
3914 @item -Wframe-larger-than=@var{len}
3915 @opindex Wframe-larger-than
3916 Warn if the size of a function frame is larger than @var{len} bytes.
3917 The computation done to determine the stack frame size is approximate
3918 and not conservative.
3919 The actual requirements may be somewhat greater than @var{len}
3920 even if you do not get a warning. In addition, any space allocated
3921 via @code{alloca}, variable-length arrays, or related constructs
3922 is not included by the compiler when determining
3923 whether or not to issue a warning.
3925 @item -Wunsafe-loop-optimizations
3926 @opindex Wunsafe-loop-optimizations
3927 @opindex Wno-unsafe-loop-optimizations
3928 Warn if the loop cannot be optimized because the compiler could not
3929 assume anything on the bounds of the loop indices. With
3930 @option{-funsafe-loop-optimizations} warn if the compiler made
3933 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3934 @opindex Wno-pedantic-ms-format
3935 @opindex Wpedantic-ms-format
3936 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3937 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3938 depending on the MS runtime, when you are using the options @option{-Wformat}
3939 and @option{-pedantic} without gnu-extensions.
3941 @item -Wpointer-arith
3942 @opindex Wpointer-arith
3943 @opindex Wno-pointer-arith
3944 Warn about anything that depends on the ``size of'' a function type or
3945 of @code{void}. GNU C assigns these types a size of 1, for
3946 convenience in calculations with @code{void *} pointers and pointers
3947 to functions. In C++, warn also when an arithmetic operation involves
3948 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3951 @opindex Wtype-limits
3952 @opindex Wno-type-limits
3953 Warn if a comparison is always true or always false due to the limited
3954 range of the data type, but do not warn for constant expressions. For
3955 example, warn if an unsigned variable is compared against zero with
3956 @samp{<} or @samp{>=}. This warning is also enabled by
3959 @item -Wbad-function-cast @r{(C and Objective-C only)}
3960 @opindex Wbad-function-cast
3961 @opindex Wno-bad-function-cast
3962 Warn whenever a function call is cast to a non-matching type.
3963 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3965 @item -Wc++-compat @r{(C and Objective-C only)}
3966 Warn about ISO C constructs that are outside of the common subset of
3967 ISO C and ISO C++, e.g.@: request for implicit conversion from
3968 @code{void *} to a pointer to non-@code{void} type.
3970 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3971 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3972 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3973 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3977 @opindex Wno-cast-qual
3978 Warn whenever a pointer is cast so as to remove a type qualifier from
3979 the target type. For example, warn if a @code{const char *} is cast
3980 to an ordinary @code{char *}.
3982 Also warn when making a cast which introduces a type qualifier in an
3983 unsafe way. For example, casting @code{char **} to @code{const char **}
3984 is unsafe, as in this example:
3987 /* p is char ** value. */
3988 const char **q = (const char **) p;
3989 /* Assignment of readonly string to const char * is OK. */
3991 /* Now char** pointer points to read-only memory. */
3996 @opindex Wcast-align
3997 @opindex Wno-cast-align
3998 Warn whenever a pointer is cast such that the required alignment of the
3999 target is increased. For example, warn if a @code{char *} is cast to
4000 an @code{int *} on machines where integers can only be accessed at
4001 two- or four-byte boundaries.
4003 @item -Wwrite-strings
4004 @opindex Wwrite-strings
4005 @opindex Wno-write-strings
4006 When compiling C, give string constants the type @code{const
4007 char[@var{length}]} so that copying the address of one into a
4008 non-@code{const} @code{char *} pointer will get a warning. These
4009 warnings will help you find at compile time code that can try to write
4010 into a string constant, but only if you have been very careful about
4011 using @code{const} in declarations and prototypes. Otherwise, it will
4012 just be a nuisance. This is why we did not make @option{-Wall} request
4015 When compiling C++, warn about the deprecated conversion from string
4016 literals to @code{char *}. This warning is enabled by default for C++
4021 @opindex Wno-clobbered
4022 Warn for variables that might be changed by @samp{longjmp} or
4023 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4026 @opindex Wconversion
4027 @opindex Wno-conversion
4028 Warn for implicit conversions that may alter a value. This includes
4029 conversions between real and integer, like @code{abs (x)} when
4030 @code{x} is @code{double}; conversions between signed and unsigned,
4031 like @code{unsigned ui = -1}; and conversions to smaller types, like
4032 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4033 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4034 changed by the conversion like in @code{abs (2.0)}. Warnings about
4035 conversions between signed and unsigned integers can be disabled by
4036 using @option{-Wno-sign-conversion}.
4038 For C++, also warn for confusing overload resolution for user-defined
4039 conversions; and conversions that will never use a type conversion
4040 operator: conversions to @code{void}, the same type, a base class or a
4041 reference to them. Warnings about conversions between signed and
4042 unsigned integers are disabled by default in C++ unless
4043 @option{-Wsign-conversion} is explicitly enabled.
4045 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4046 @opindex Wconversion-null
4047 @opindex Wno-conversion-null
4048 Do not warn for conversions between @code{NULL} and non-pointer
4049 types. @option{-Wconversion-null} is enabled by default.
4052 @opindex Wempty-body
4053 @opindex Wno-empty-body
4054 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4055 while} statement. This warning is also enabled by @option{-Wextra}.
4057 @item -Wenum-compare
4058 @opindex Wenum-compare
4059 @opindex Wno-enum-compare
4060 Warn about a comparison between values of different enum types. In C++
4061 this warning is enabled by default. In C this warning is enabled by
4064 @item -Wjump-misses-init @r{(C, Objective-C only)}
4065 @opindex Wjump-misses-init
4066 @opindex Wno-jump-misses-init
4067 Warn if a @code{goto} statement or a @code{switch} statement jumps
4068 forward across the initialization of a variable, or jumps backward to a
4069 label after the variable has been initialized. This only warns about
4070 variables which are initialized when they are declared. This warning is
4071 only supported for C and Objective C; in C++ this sort of branch is an
4074 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4075 can be disabled with the @option{-Wno-jump-misses-init} option.
4077 @item -Wsign-compare
4078 @opindex Wsign-compare
4079 @opindex Wno-sign-compare
4080 @cindex warning for comparison of signed and unsigned values
4081 @cindex comparison of signed and unsigned values, warning
4082 @cindex signed and unsigned values, comparison warning
4083 Warn when a comparison between signed and unsigned values could produce
4084 an incorrect result when the signed value is converted to unsigned.
4085 This warning is also enabled by @option{-Wextra}; to get the other warnings
4086 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4088 @item -Wsign-conversion
4089 @opindex Wsign-conversion
4090 @opindex Wno-sign-conversion
4091 Warn for implicit conversions that may change the sign of an integer
4092 value, like assigning a signed integer expression to an unsigned
4093 integer variable. An explicit cast silences the warning. In C, this
4094 option is enabled also by @option{-Wconversion}.
4098 @opindex Wno-address
4099 Warn about suspicious uses of memory addresses. These include using
4100 the address of a function in a conditional expression, such as
4101 @code{void func(void); if (func)}, and comparisons against the memory
4102 address of a string literal, such as @code{if (x == "abc")}. Such
4103 uses typically indicate a programmer error: the address of a function
4104 always evaluates to true, so their use in a conditional usually
4105 indicate that the programmer forgot the parentheses in a function
4106 call; and comparisons against string literals result in unspecified
4107 behavior and are not portable in C, so they usually indicate that the
4108 programmer intended to use @code{strcmp}. This warning is enabled by
4112 @opindex Wlogical-op
4113 @opindex Wno-logical-op
4114 Warn about suspicious uses of logical operators in expressions.
4115 This includes using logical operators in contexts where a
4116 bit-wise operator is likely to be expected.
4118 @item -Waggregate-return
4119 @opindex Waggregate-return
4120 @opindex Wno-aggregate-return
4121 Warn if any functions that return structures or unions are defined or
4122 called. (In languages where you can return an array, this also elicits
4125 @item -Wno-attributes
4126 @opindex Wno-attributes
4127 @opindex Wattributes
4128 Do not warn if an unexpected @code{__attribute__} is used, such as
4129 unrecognized attributes, function attributes applied to variables,
4130 etc. This will not stop errors for incorrect use of supported
4133 @item -Wno-builtin-macro-redefined
4134 @opindex Wno-builtin-macro-redefined
4135 @opindex Wbuiltin-macro-redefined
4136 Do not warn if certain built-in macros are redefined. This suppresses
4137 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4138 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4140 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4141 @opindex Wstrict-prototypes
4142 @opindex Wno-strict-prototypes
4143 Warn if a function is declared or defined without specifying the
4144 argument types. (An old-style function definition is permitted without
4145 a warning if preceded by a declaration which specifies the argument
4148 @item -Wold-style-declaration @r{(C and Objective-C only)}
4149 @opindex Wold-style-declaration
4150 @opindex Wno-old-style-declaration
4151 Warn for obsolescent usages, according to the C Standard, in a
4152 declaration. For example, warn if storage-class specifiers like
4153 @code{static} are not the first things in a declaration. This warning
4154 is also enabled by @option{-Wextra}.
4156 @item -Wold-style-definition @r{(C and Objective-C only)}
4157 @opindex Wold-style-definition
4158 @opindex Wno-old-style-definition
4159 Warn if an old-style function definition is used. A warning is given
4160 even if there is a previous prototype.
4162 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4163 @opindex Wmissing-parameter-type
4164 @opindex Wno-missing-parameter-type
4165 A function parameter is declared without a type specifier in K&R-style
4172 This warning is also enabled by @option{-Wextra}.
4174 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4175 @opindex Wmissing-prototypes
4176 @opindex Wno-missing-prototypes
4177 Warn if a global function is defined without a previous prototype
4178 declaration. This warning is issued even if the definition itself
4179 provides a prototype. The aim is to detect global functions that fail
4180 to be declared in header files.
4182 @item -Wmissing-declarations
4183 @opindex Wmissing-declarations
4184 @opindex Wno-missing-declarations
4185 Warn if a global function is defined without a previous declaration.
4186 Do so even if the definition itself provides a prototype.
4187 Use this option to detect global functions that are not declared in
4188 header files. In C++, no warnings are issued for function templates,
4189 or for inline functions, or for functions in anonymous namespaces.
4191 @item -Wmissing-field-initializers
4192 @opindex Wmissing-field-initializers
4193 @opindex Wno-missing-field-initializers
4197 Warn if a structure's initializer has some fields missing. For
4198 example, the following code would cause such a warning, because
4199 @code{x.h} is implicitly zero:
4202 struct s @{ int f, g, h; @};
4203 struct s x = @{ 3, 4 @};
4206 This option does not warn about designated initializers, so the following
4207 modification would not trigger a warning:
4210 struct s @{ int f, g, h; @};
4211 struct s x = @{ .f = 3, .g = 4 @};
4214 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4215 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4217 @item -Wmissing-format-attribute
4218 @opindex Wmissing-format-attribute
4219 @opindex Wno-missing-format-attribute
4222 Warn about function pointers which might be candidates for @code{format}
4223 attributes. Note these are only possible candidates, not absolute ones.
4224 GCC will guess that function pointers with @code{format} attributes that
4225 are used in assignment, initialization, parameter passing or return
4226 statements should have a corresponding @code{format} attribute in the
4227 resulting type. I.e.@: the left-hand side of the assignment or
4228 initialization, the type of the parameter variable, or the return type
4229 of the containing function respectively should also have a @code{format}
4230 attribute to avoid the warning.
4232 GCC will also warn about function definitions which might be
4233 candidates for @code{format} attributes. Again, these are only
4234 possible candidates. GCC will guess that @code{format} attributes
4235 might be appropriate for any function that calls a function like
4236 @code{vprintf} or @code{vscanf}, but this might not always be the
4237 case, and some functions for which @code{format} attributes are
4238 appropriate may not be detected.
4240 @item -Wno-multichar
4241 @opindex Wno-multichar
4243 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4244 Usually they indicate a typo in the user's code, as they have
4245 implementation-defined values, and should not be used in portable code.
4247 @item -Wnormalized=<none|id|nfc|nfkc>
4248 @opindex Wnormalized=
4251 @cindex character set, input normalization
4252 In ISO C and ISO C++, two identifiers are different if they are
4253 different sequences of characters. However, sometimes when characters
4254 outside the basic ASCII character set are used, you can have two
4255 different character sequences that look the same. To avoid confusion,
4256 the ISO 10646 standard sets out some @dfn{normalization rules} which
4257 when applied ensure that two sequences that look the same are turned into
4258 the same sequence. GCC can warn you if you are using identifiers which
4259 have not been normalized; this option controls that warning.
4261 There are four levels of warning that GCC supports. The default is
4262 @option{-Wnormalized=nfc}, which warns about any identifier which is
4263 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4264 recommended form for most uses.
4266 Unfortunately, there are some characters which ISO C and ISO C++ allow
4267 in identifiers that when turned into NFC aren't allowable as
4268 identifiers. That is, there's no way to use these symbols in portable
4269 ISO C or C++ and have all your identifiers in NFC@.
4270 @option{-Wnormalized=id} suppresses the warning for these characters.
4271 It is hoped that future versions of the standards involved will correct
4272 this, which is why this option is not the default.
4274 You can switch the warning off for all characters by writing
4275 @option{-Wnormalized=none}. You would only want to do this if you
4276 were using some other normalization scheme (like ``D''), because
4277 otherwise you can easily create bugs that are literally impossible to see.
4279 Some characters in ISO 10646 have distinct meanings but look identical
4280 in some fonts or display methodologies, especially once formatting has
4281 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4282 LETTER N'', will display just like a regular @code{n} which has been
4283 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4284 normalization scheme to convert all these into a standard form as
4285 well, and GCC will warn if your code is not in NFKC if you use
4286 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4287 about every identifier that contains the letter O because it might be
4288 confused with the digit 0, and so is not the default, but may be
4289 useful as a local coding convention if the programming environment is
4290 unable to be fixed to display these characters distinctly.
4292 @item -Wno-deprecated
4293 @opindex Wno-deprecated
4294 @opindex Wdeprecated
4295 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4297 @item -Wno-deprecated-declarations
4298 @opindex Wno-deprecated-declarations
4299 @opindex Wdeprecated-declarations
4300 Do not warn about uses of functions (@pxref{Function Attributes}),
4301 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4302 Attributes}) marked as deprecated by using the @code{deprecated}
4306 @opindex Wno-overflow
4308 Do not warn about compile-time overflow in constant expressions.
4310 @item -Woverride-init @r{(C and Objective-C only)}
4311 @opindex Woverride-init
4312 @opindex Wno-override-init
4316 Warn if an initialized field without side effects is overridden when
4317 using designated initializers (@pxref{Designated Inits, , Designated
4320 This warning is included in @option{-Wextra}. To get other
4321 @option{-Wextra} warnings without this one, use @samp{-Wextra
4322 -Wno-override-init}.
4327 Warn if a structure is given the packed attribute, but the packed
4328 attribute has no effect on the layout or size of the structure.
4329 Such structures may be mis-aligned for little benefit. For
4330 instance, in this code, the variable @code{f.x} in @code{struct bar}
4331 will be misaligned even though @code{struct bar} does not itself
4332 have the packed attribute:
4339 @} __attribute__((packed));
4347 @item -Wpacked-bitfield-compat
4348 @opindex Wpacked-bitfield-compat
4349 @opindex Wno-packed-bitfield-compat
4350 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4351 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4352 the change can lead to differences in the structure layout. GCC
4353 informs you when the offset of such a field has changed in GCC 4.4.
4354 For example there is no longer a 4-bit padding between field @code{a}
4355 and @code{b} in this structure:
4362 @} __attribute__ ((packed));
4365 This warning is enabled by default. Use
4366 @option{-Wno-packed-bitfield-compat} to disable this warning.
4371 Warn if padding is included in a structure, either to align an element
4372 of the structure or to align the whole structure. Sometimes when this
4373 happens it is possible to rearrange the fields of the structure to
4374 reduce the padding and so make the structure smaller.
4376 @item -Wredundant-decls
4377 @opindex Wredundant-decls
4378 @opindex Wno-redundant-decls
4379 Warn if anything is declared more than once in the same scope, even in
4380 cases where multiple declaration is valid and changes nothing.
4382 @item -Wnested-externs @r{(C and Objective-C only)}
4383 @opindex Wnested-externs
4384 @opindex Wno-nested-externs
4385 Warn if an @code{extern} declaration is encountered within a function.
4390 Warn if a function can not be inlined and it was declared as inline.
4391 Even with this option, the compiler will not warn about failures to
4392 inline functions declared in system headers.
4394 The compiler uses a variety of heuristics to determine whether or not
4395 to inline a function. For example, the compiler takes into account
4396 the size of the function being inlined and the amount of inlining
4397 that has already been done in the current function. Therefore,
4398 seemingly insignificant changes in the source program can cause the
4399 warnings produced by @option{-Winline} to appear or disappear.
4401 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4402 @opindex Wno-invalid-offsetof
4403 @opindex Winvalid-offsetof
4404 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4405 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4406 to a non-POD type is undefined. In existing C++ implementations,
4407 however, @samp{offsetof} typically gives meaningful results even when
4408 applied to certain kinds of non-POD types. (Such as a simple
4409 @samp{struct} that fails to be a POD type only by virtue of having a
4410 constructor.) This flag is for users who are aware that they are
4411 writing nonportable code and who have deliberately chosen to ignore the
4414 The restrictions on @samp{offsetof} may be relaxed in a future version
4415 of the C++ standard.
4417 @item -Wno-int-to-pointer-cast
4418 @opindex Wno-int-to-pointer-cast
4419 @opindex Wint-to-pointer-cast
4420 Suppress warnings from casts to pointer type of an integer of a
4421 different size. In C++, casting to a pointer type of smaller size is
4422 an error. @option{Wint-to-pointer-cast} is enabled by default.
4425 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4426 @opindex Wno-pointer-to-int-cast
4427 @opindex Wpointer-to-int-cast
4428 Suppress warnings from casts from a pointer to an integer type of a
4432 @opindex Winvalid-pch
4433 @opindex Wno-invalid-pch
4434 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4435 the search path but can't be used.
4439 @opindex Wno-long-long
4440 Warn if @samp{long long} type is used. This is enabled by either
4441 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4442 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4444 @item -Wvariadic-macros
4445 @opindex Wvariadic-macros
4446 @opindex Wno-variadic-macros
4447 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4448 alternate syntax when in pedantic ISO C99 mode. This is default.
4449 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4454 Warn if variable length array is used in the code.
4455 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4456 the variable length array.
4458 @item -Wvolatile-register-var
4459 @opindex Wvolatile-register-var
4460 @opindex Wno-volatile-register-var
4461 Warn if a register variable is declared volatile. The volatile
4462 modifier does not inhibit all optimizations that may eliminate reads
4463 and/or writes to register variables. This warning is enabled by
4466 @item -Wdisabled-optimization
4467 @opindex Wdisabled-optimization
4468 @opindex Wno-disabled-optimization
4469 Warn if a requested optimization pass is disabled. This warning does
4470 not generally indicate that there is anything wrong with your code; it
4471 merely indicates that GCC's optimizers were unable to handle the code
4472 effectively. Often, the problem is that your code is too big or too
4473 complex; GCC will refuse to optimize programs when the optimization
4474 itself is likely to take inordinate amounts of time.
4476 @item -Wpointer-sign @r{(C and Objective-C only)}
4477 @opindex Wpointer-sign
4478 @opindex Wno-pointer-sign
4479 Warn for pointer argument passing or assignment with different signedness.
4480 This option is only supported for C and Objective-C@. It is implied by
4481 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4482 @option{-Wno-pointer-sign}.
4484 @item -Wstack-protector
4485 @opindex Wstack-protector
4486 @opindex Wno-stack-protector
4487 This option is only active when @option{-fstack-protector} is active. It
4488 warns about functions that will not be protected against stack smashing.
4491 @opindex Wno-mudflap
4492 Suppress warnings about constructs that cannot be instrumented by
4495 @item -Woverlength-strings
4496 @opindex Woverlength-strings
4497 @opindex Wno-overlength-strings
4498 Warn about string constants which are longer than the ``minimum
4499 maximum'' length specified in the C standard. Modern compilers
4500 generally allow string constants which are much longer than the
4501 standard's minimum limit, but very portable programs should avoid
4502 using longer strings.
4504 The limit applies @emph{after} string constant concatenation, and does
4505 not count the trailing NUL@. In C90, the limit was 509 characters; in
4506 C99, it was raised to 4095. C++98 does not specify a normative
4507 minimum maximum, so we do not diagnose overlength strings in C++@.
4509 This option is implied by @option{-pedantic}, and can be disabled with
4510 @option{-Wno-overlength-strings}.
4512 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4513 @opindex Wunsuffixed-float-constants
4515 GCC will issue a warning for any floating constant that does not have
4516 a suffix. When used together with @option{-Wsystem-headers} it will
4517 warn about such constants in system header files. This can be useful
4518 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4519 from the decimal floating-point extension to C99.
4522 @node Debugging Options
4523 @section Options for Debugging Your Program or GCC
4524 @cindex options, debugging
4525 @cindex debugging information options
4527 GCC has various special options that are used for debugging
4528 either your program or GCC:
4533 Produce debugging information in the operating system's native format
4534 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4537 On most systems that use stabs format, @option{-g} enables use of extra
4538 debugging information that only GDB can use; this extra information
4539 makes debugging work better in GDB but will probably make other debuggers
4541 refuse to read the program. If you want to control for certain whether
4542 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4543 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4545 GCC allows you to use @option{-g} with
4546 @option{-O}. The shortcuts taken by optimized code may occasionally
4547 produce surprising results: some variables you declared may not exist
4548 at all; flow of control may briefly move where you did not expect it;
4549 some statements may not be executed because they compute constant
4550 results or their values were already at hand; some statements may
4551 execute in different places because they were moved out of loops.
4553 Nevertheless it proves possible to debug optimized output. This makes
4554 it reasonable to use the optimizer for programs that might have bugs.
4556 The following options are useful when GCC is generated with the
4557 capability for more than one debugging format.
4561 Produce debugging information for use by GDB@. This means to use the
4562 most expressive format available (DWARF 2, stabs, or the native format
4563 if neither of those are supported), including GDB extensions if at all
4568 Produce debugging information in stabs format (if that is supported),
4569 without GDB extensions. This is the format used by DBX on most BSD
4570 systems. On MIPS, Alpha and System V Release 4 systems this option
4571 produces stabs debugging output which is not understood by DBX or SDB@.
4572 On System V Release 4 systems this option requires the GNU assembler.
4574 @item -feliminate-unused-debug-symbols
4575 @opindex feliminate-unused-debug-symbols
4576 Produce debugging information in stabs format (if that is supported),
4577 for only symbols that are actually used.
4579 @item -femit-class-debug-always
4580 Instead of emitting debugging information for a C++ class in only one
4581 object file, emit it in all object files using the class. This option
4582 should be used only with debuggers that are unable to handle the way GCC
4583 normally emits debugging information for classes because using this
4584 option will increase the size of debugging information by as much as a
4589 Produce debugging information in stabs format (if that is supported),
4590 using GNU extensions understood only by the GNU debugger (GDB)@. The
4591 use of these extensions is likely to make other debuggers crash or
4592 refuse to read the program.
4596 Produce debugging information in COFF format (if that is supported).
4597 This is the format used by SDB on most System V systems prior to
4602 Produce debugging information in XCOFF format (if that is supported).
4603 This is the format used by the DBX debugger on IBM RS/6000 systems.
4607 Produce debugging information in XCOFF format (if that is supported),
4608 using GNU extensions understood only by the GNU debugger (GDB)@. The
4609 use of these extensions is likely to make other debuggers crash or
4610 refuse to read the program, and may cause assemblers other than the GNU
4611 assembler (GAS) to fail with an error.
4613 @item -gdwarf-@var{version}
4614 @opindex gdwarf-@var{version}
4615 Produce debugging information in DWARF format (if that is
4616 supported). This is the format used by DBX on IRIX 6. The value
4617 of @var{version} may be either 2, 3 or 4; the default version is 2.
4619 Note that with DWARF version 2 some ports require, and will always
4620 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4622 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4623 for maximum benefit.
4625 @item -gstrict-dwarf
4626 @opindex gstrict-dwarf
4627 Disallow using extensions of later DWARF standard version than selected
4628 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4629 DWARF extensions from later standard versions is allowed.
4631 @item -gno-strict-dwarf
4632 @opindex gno-strict-dwarf
4633 Allow using extensions of later DWARF standard version than selected with
4634 @option{-gdwarf-@var{version}}.
4638 Produce debugging information in VMS debug format (if that is
4639 supported). This is the format used by DEBUG on VMS systems.
4642 @itemx -ggdb@var{level}
4643 @itemx -gstabs@var{level}
4644 @itemx -gcoff@var{level}
4645 @itemx -gxcoff@var{level}
4646 @itemx -gvms@var{level}
4647 Request debugging information and also use @var{level} to specify how
4648 much information. The default level is 2.
4650 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4653 Level 1 produces minimal information, enough for making backtraces in
4654 parts of the program that you don't plan to debug. This includes
4655 descriptions of functions and external variables, but no information
4656 about local variables and no line numbers.
4658 Level 3 includes extra information, such as all the macro definitions
4659 present in the program. Some debuggers support macro expansion when
4660 you use @option{-g3}.
4662 @option{-gdwarf-2} does not accept a concatenated debug level, because
4663 GCC used to support an option @option{-gdwarf} that meant to generate
4664 debug information in version 1 of the DWARF format (which is very
4665 different from version 2), and it would have been too confusing. That
4666 debug format is long obsolete, but the option cannot be changed now.
4667 Instead use an additional @option{-g@var{level}} option to change the
4668 debug level for DWARF.
4672 Turn off generation of debug info, if leaving out this option would have
4673 generated it, or turn it on at level 2 otherwise. The position of this
4674 argument in the command line does not matter, it takes effect after all
4675 other options are processed, and it does so only once, no matter how
4676 many times it is given. This is mainly intended to be used with
4677 @option{-fcompare-debug}.
4679 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4680 @opindex fdump-final-insns
4681 Dump the final internal representation (RTL) to @var{file}. If the
4682 optional argument is omitted (or if @var{file} is @code{.}), the name
4683 of the dump file will be determined by appending @code{.gkd} to the
4684 compilation output file name.
4686 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4687 @opindex fcompare-debug
4688 @opindex fno-compare-debug
4689 If no error occurs during compilation, run the compiler a second time,
4690 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4691 passed to the second compilation. Dump the final internal
4692 representation in both compilations, and print an error if they differ.
4694 If the equal sign is omitted, the default @option{-gtoggle} is used.
4696 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4697 and nonzero, implicitly enables @option{-fcompare-debug}. If
4698 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4699 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4702 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4703 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4704 of the final representation and the second compilation, preventing even
4705 @env{GCC_COMPARE_DEBUG} from taking effect.
4707 To verify full coverage during @option{-fcompare-debug} testing, set
4708 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4709 which GCC will reject as an invalid option in any actual compilation
4710 (rather than preprocessing, assembly or linking). To get just a
4711 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4712 not overridden} will do.
4714 @item -fcompare-debug-second
4715 @opindex fcompare-debug-second
4716 This option is implicitly passed to the compiler for the second
4717 compilation requested by @option{-fcompare-debug}, along with options to
4718 silence warnings, and omitting other options that would cause
4719 side-effect compiler outputs to files or to the standard output. Dump
4720 files and preserved temporary files are renamed so as to contain the
4721 @code{.gk} additional extension during the second compilation, to avoid
4722 overwriting those generated by the first.
4724 When this option is passed to the compiler driver, it causes the
4725 @emph{first} compilation to be skipped, which makes it useful for little
4726 other than debugging the compiler proper.
4728 @item -feliminate-dwarf2-dups
4729 @opindex feliminate-dwarf2-dups
4730 Compress DWARF2 debugging information by eliminating duplicated
4731 information about each symbol. This option only makes sense when
4732 generating DWARF2 debugging information with @option{-gdwarf-2}.
4734 @item -femit-struct-debug-baseonly
4735 Emit debug information for struct-like types
4736 only when the base name of the compilation source file
4737 matches the base name of file in which the struct was defined.
4739 This option substantially reduces the size of debugging information,
4740 but at significant potential loss in type information to the debugger.
4741 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4742 See @option{-femit-struct-debug-detailed} for more detailed control.
4744 This option works only with DWARF 2.
4746 @item -femit-struct-debug-reduced
4747 Emit debug information for struct-like types
4748 only when the base name of the compilation source file
4749 matches the base name of file in which the type was defined,
4750 unless the struct is a template or defined in a system header.
4752 This option significantly reduces the size of debugging information,
4753 with some potential loss in type information to the debugger.
4754 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4755 See @option{-femit-struct-debug-detailed} for more detailed control.
4757 This option works only with DWARF 2.
4759 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4760 Specify the struct-like types
4761 for which the compiler will generate debug information.
4762 The intent is to reduce duplicate struct debug information
4763 between different object files within the same program.
4765 This option is a detailed version of
4766 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4767 which will serve for most needs.
4769 A specification has the syntax@*
4770 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4772 The optional first word limits the specification to
4773 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4774 A struct type is used directly when it is the type of a variable, member.
4775 Indirect uses arise through pointers to structs.
4776 That is, when use of an incomplete struct would be legal, the use is indirect.
4778 @samp{struct one direct; struct two * indirect;}.
4780 The optional second word limits the specification to
4781 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4782 Generic structs are a bit complicated to explain.
4783 For C++, these are non-explicit specializations of template classes,
4784 or non-template classes within the above.
4785 Other programming languages have generics,
4786 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4788 The third word specifies the source files for those
4789 structs for which the compiler will emit debug information.
4790 The values @samp{none} and @samp{any} have the normal meaning.
4791 The value @samp{base} means that
4792 the base of name of the file in which the type declaration appears
4793 must match the base of the name of the main compilation file.
4794 In practice, this means that
4795 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4796 but types declared in other header will not.
4797 The value @samp{sys} means those types satisfying @samp{base}
4798 or declared in system or compiler headers.
4800 You may need to experiment to determine the best settings for your application.
4802 The default is @samp{-femit-struct-debug-detailed=all}.
4804 This option works only with DWARF 2.
4806 @item -fno-merge-debug-strings
4807 @opindex fmerge-debug-strings
4808 @opindex fno-merge-debug-strings
4809 Direct the linker to not merge together strings in the debugging
4810 information which are identical in different object files. Merging is
4811 not supported by all assemblers or linkers. Merging decreases the size
4812 of the debug information in the output file at the cost of increasing
4813 link processing time. Merging is enabled by default.
4815 @item -fdebug-prefix-map=@var{old}=@var{new}
4816 @opindex fdebug-prefix-map
4817 When compiling files in directory @file{@var{old}}, record debugging
4818 information describing them as in @file{@var{new}} instead.
4820 @item -fno-dwarf2-cfi-asm
4821 @opindex fdwarf2-cfi-asm
4822 @opindex fno-dwarf2-cfi-asm
4823 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4824 instead of using GAS @code{.cfi_*} directives.
4826 @cindex @command{prof}
4829 Generate extra code to write profile information suitable for the
4830 analysis program @command{prof}. You must use this option when compiling
4831 the source files you want data about, and you must also use it when
4834 @cindex @command{gprof}
4837 Generate extra code to write profile information suitable for the
4838 analysis program @command{gprof}. You must use this option when compiling
4839 the source files you want data about, and you must also use it when
4844 Makes the compiler print out each function name as it is compiled, and
4845 print some statistics about each pass when it finishes.
4848 @opindex ftime-report
4849 Makes the compiler print some statistics about the time consumed by each
4850 pass when it finishes.
4853 @opindex fmem-report
4854 Makes the compiler print some statistics about permanent memory
4855 allocation when it finishes.
4857 @item -fpre-ipa-mem-report
4858 @opindex fpre-ipa-mem-report
4859 @item -fpost-ipa-mem-report
4860 @opindex fpost-ipa-mem-report
4861 Makes the compiler print some statistics about permanent memory
4862 allocation before or after interprocedural optimization.
4865 @opindex fstack-usage
4866 Makes the compiler output stack usage information for the program, on a
4867 per-function basis. The filename for the dump is made by appending
4868 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4869 the output file, if explicitly specified and it is not an executable,
4870 otherwise it is the basename of the source file. An entry is made up
4875 The name of the function.
4879 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4882 The qualifier @code{static} means that the function manipulates the stack
4883 statically: a fixed number of bytes are allocated for the frame on function
4884 entry and released on function exit; no stack adjustments are otherwise made
4885 in the function. The second field is this fixed number of bytes.
4887 The qualifier @code{dynamic} means that the function manipulates the stack
4888 dynamically: in addition to the static allocation described above, stack
4889 adjustments are made in the body of the function, for example to push/pop
4890 arguments around function calls. If the qualifier @code{bounded} is also
4891 present, the amount of these adjustments is bounded at compile-time and
4892 the second field is an upper bound of the total amount of stack used by
4893 the function. If it is not present, the amount of these adjustments is
4894 not bounded at compile-time and the second field only represents the
4897 @item -fprofile-arcs
4898 @opindex fprofile-arcs
4899 Add code so that program flow @dfn{arcs} are instrumented. During
4900 execution the program records how many times each branch and call is
4901 executed and how many times it is taken or returns. When the compiled
4902 program exits it saves this data to a file called
4903 @file{@var{auxname}.gcda} for each source file. The data may be used for
4904 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4905 test coverage analysis (@option{-ftest-coverage}). Each object file's
4906 @var{auxname} is generated from the name of the output file, if
4907 explicitly specified and it is not the final executable, otherwise it is
4908 the basename of the source file. In both cases any suffix is removed
4909 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4910 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4911 @xref{Cross-profiling}.
4913 @cindex @command{gcov}
4917 This option is used to compile and link code instrumented for coverage
4918 analysis. The option is a synonym for @option{-fprofile-arcs}
4919 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4920 linking). See the documentation for those options for more details.
4925 Compile the source files with @option{-fprofile-arcs} plus optimization
4926 and code generation options. For test coverage analysis, use the
4927 additional @option{-ftest-coverage} option. You do not need to profile
4928 every source file in a program.
4931 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4932 (the latter implies the former).
4935 Run the program on a representative workload to generate the arc profile
4936 information. This may be repeated any number of times. You can run
4937 concurrent instances of your program, and provided that the file system
4938 supports locking, the data files will be correctly updated. Also
4939 @code{fork} calls are detected and correctly handled (double counting
4943 For profile-directed optimizations, compile the source files again with
4944 the same optimization and code generation options plus
4945 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4946 Control Optimization}).
4949 For test coverage analysis, use @command{gcov} to produce human readable
4950 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4951 @command{gcov} documentation for further information.
4955 With @option{-fprofile-arcs}, for each function of your program GCC
4956 creates a program flow graph, then finds a spanning tree for the graph.
4957 Only arcs that are not on the spanning tree have to be instrumented: the
4958 compiler adds code to count the number of times that these arcs are
4959 executed. When an arc is the only exit or only entrance to a block, the
4960 instrumentation code can be added to the block; otherwise, a new basic
4961 block must be created to hold the instrumentation code.
4964 @item -ftest-coverage
4965 @opindex ftest-coverage
4966 Produce a notes file that the @command{gcov} code-coverage utility
4967 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4968 show program coverage. Each source file's note file is called
4969 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4970 above for a description of @var{auxname} and instructions on how to
4971 generate test coverage data. Coverage data will match the source files
4972 more closely, if you do not optimize.
4974 @item -fdbg-cnt-list
4975 @opindex fdbg-cnt-list
4976 Print the name and the counter upper bound for all debug counters.
4978 @item -fdbg-cnt=@var{counter-value-list}
4980 Set the internal debug counter upper bound. @var{counter-value-list}
4981 is a comma-separated list of @var{name}:@var{value} pairs
4982 which sets the upper bound of each debug counter @var{name} to @var{value}.
4983 All debug counters have the initial upper bound of @var{UINT_MAX},
4984 thus dbg_cnt() returns true always unless the upper bound is set by this option.
4985 e.g. With -fdbg-cnt=dce:10,tail_call:0
4986 dbg_cnt(dce) will return true only for first 10 invocations
4987 and dbg_cnt(tail_call) will return false always.
4989 @item -d@var{letters}
4990 @itemx -fdump-rtl-@var{pass}
4992 Says to make debugging dumps during compilation at times specified by
4993 @var{letters}. This is used for debugging the RTL-based passes of the
4994 compiler. The file names for most of the dumps are made by appending
4995 a pass number and a word to the @var{dumpname}, and the files are
4996 created in the directory of the output file. Note that the pass
4997 number is computed statically as passes get registered into the pass
4998 manager. Thus the numbering is not related to the dynamic order of
4999 execution of passes. In particular, a pass installed by a plugin
5000 could have a number over 200 even if it executed quite early.
5001 @var{dumpname} is generated from the name of the output file, if
5002 explicitly specified and it is not an executable, otherwise it is the
5003 basename of the source file. These switches may have different effects
5004 when @option{-E} is used for preprocessing.
5006 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5007 @option{-d} option @var{letters}. Here are the possible
5008 letters for use in @var{pass} and @var{letters}, and their meanings:
5012 @item -fdump-rtl-alignments
5013 @opindex fdump-rtl-alignments
5014 Dump after branch alignments have been computed.
5016 @item -fdump-rtl-asmcons
5017 @opindex fdump-rtl-asmcons
5018 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5020 @item -fdump-rtl-auto_inc_dec
5021 @opindex fdump-rtl-auto_inc_dec
5022 Dump after auto-inc-dec discovery. This pass is only run on
5023 architectures that have auto inc or auto dec instructions.
5025 @item -fdump-rtl-barriers
5026 @opindex fdump-rtl-barriers
5027 Dump after cleaning up the barrier instructions.
5029 @item -fdump-rtl-bbpart
5030 @opindex fdump-rtl-bbpart
5031 Dump after partitioning hot and cold basic blocks.
5033 @item -fdump-rtl-bbro
5034 @opindex fdump-rtl-bbro
5035 Dump after block reordering.
5037 @item -fdump-rtl-btl1
5038 @itemx -fdump-rtl-btl2
5039 @opindex fdump-rtl-btl2
5040 @opindex fdump-rtl-btl2
5041 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5042 after the two branch
5043 target load optimization passes.
5045 @item -fdump-rtl-bypass
5046 @opindex fdump-rtl-bypass
5047 Dump after jump bypassing and control flow optimizations.
5049 @item -fdump-rtl-combine
5050 @opindex fdump-rtl-combine
5051 Dump after the RTL instruction combination pass.
5053 @item -fdump-rtl-compgotos
5054 @opindex fdump-rtl-compgotos
5055 Dump after duplicating the computed gotos.
5057 @item -fdump-rtl-ce1
5058 @itemx -fdump-rtl-ce2
5059 @itemx -fdump-rtl-ce3
5060 @opindex fdump-rtl-ce1
5061 @opindex fdump-rtl-ce2
5062 @opindex fdump-rtl-ce3
5063 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5064 @option{-fdump-rtl-ce3} enable dumping after the three
5065 if conversion passes.
5067 @itemx -fdump-rtl-cprop_hardreg
5068 @opindex fdump-rtl-cprop_hardreg
5069 Dump after hard register copy propagation.
5071 @itemx -fdump-rtl-csa
5072 @opindex fdump-rtl-csa
5073 Dump after combining stack adjustments.
5075 @item -fdump-rtl-cse1
5076 @itemx -fdump-rtl-cse2
5077 @opindex fdump-rtl-cse1
5078 @opindex fdump-rtl-cse2
5079 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5080 the two common sub-expression elimination passes.
5082 @itemx -fdump-rtl-dce
5083 @opindex fdump-rtl-dce
5084 Dump after the standalone dead code elimination passes.
5086 @itemx -fdump-rtl-dbr
5087 @opindex fdump-rtl-dbr
5088 Dump after delayed branch scheduling.
5090 @item -fdump-rtl-dce1
5091 @itemx -fdump-rtl-dce2
5092 @opindex fdump-rtl-dce1
5093 @opindex fdump-rtl-dce2
5094 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5095 the two dead store elimination passes.
5098 @opindex fdump-rtl-eh
5099 Dump after finalization of EH handling code.
5101 @item -fdump-rtl-eh_ranges
5102 @opindex fdump-rtl-eh_ranges
5103 Dump after conversion of EH handling range regions.
5105 @item -fdump-rtl-expand
5106 @opindex fdump-rtl-expand
5107 Dump after RTL generation.
5109 @item -fdump-rtl-fwprop1
5110 @itemx -fdump-rtl-fwprop2
5111 @opindex fdump-rtl-fwprop1
5112 @opindex fdump-rtl-fwprop2
5113 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5114 dumping after the two forward propagation passes.
5116 @item -fdump-rtl-gcse1
5117 @itemx -fdump-rtl-gcse2
5118 @opindex fdump-rtl-gcse1
5119 @opindex fdump-rtl-gcse2
5120 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5121 after global common subexpression elimination.
5123 @item -fdump-rtl-init-regs
5124 @opindex fdump-rtl-init-regs
5125 Dump after the initialization of the registers.
5127 @item -fdump-rtl-initvals
5128 @opindex fdump-rtl-initvals
5129 Dump after the computation of the initial value sets.
5131 @itemx -fdump-rtl-into_cfglayout
5132 @opindex fdump-rtl-into_cfglayout
5133 Dump after converting to cfglayout mode.
5135 @item -fdump-rtl-ira
5136 @opindex fdump-rtl-ira
5137 Dump after iterated register allocation.
5139 @item -fdump-rtl-jump
5140 @opindex fdump-rtl-jump
5141 Dump after the second jump optimization.
5143 @item -fdump-rtl-loop2
5144 @opindex fdump-rtl-loop2
5145 @option{-fdump-rtl-loop2} enables dumping after the rtl
5146 loop optimization passes.
5148 @item -fdump-rtl-mach
5149 @opindex fdump-rtl-mach
5150 Dump after performing the machine dependent reorganization pass, if that
5153 @item -fdump-rtl-mode_sw
5154 @opindex fdump-rtl-mode_sw
5155 Dump after removing redundant mode switches.
5157 @item -fdump-rtl-rnreg
5158 @opindex fdump-rtl-rnreg
5159 Dump after register renumbering.
5161 @itemx -fdump-rtl-outof_cfglayout
5162 @opindex fdump-rtl-outof_cfglayout
5163 Dump after converting from cfglayout mode.
5165 @item -fdump-rtl-peephole2
5166 @opindex fdump-rtl-peephole2
5167 Dump after the peephole pass.
5169 @item -fdump-rtl-postreload
5170 @opindex fdump-rtl-postreload
5171 Dump after post-reload optimizations.
5173 @itemx -fdump-rtl-pro_and_epilogue
5174 @opindex fdump-rtl-pro_and_epilogue
5175 Dump after generating the function pro and epilogues.
5177 @item -fdump-rtl-regmove
5178 @opindex fdump-rtl-regmove
5179 Dump after the register move pass.
5181 @item -fdump-rtl-sched1
5182 @itemx -fdump-rtl-sched2
5183 @opindex fdump-rtl-sched1
5184 @opindex fdump-rtl-sched2
5185 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5186 after the basic block scheduling passes.
5188 @item -fdump-rtl-see
5189 @opindex fdump-rtl-see
5190 Dump after sign extension elimination.
5192 @item -fdump-rtl-seqabstr
5193 @opindex fdump-rtl-seqabstr
5194 Dump after common sequence discovery.
5196 @item -fdump-rtl-shorten
5197 @opindex fdump-rtl-shorten
5198 Dump after shortening branches.
5200 @item -fdump-rtl-sibling
5201 @opindex fdump-rtl-sibling
5202 Dump after sibling call optimizations.
5204 @item -fdump-rtl-split1
5205 @itemx -fdump-rtl-split2
5206 @itemx -fdump-rtl-split3
5207 @itemx -fdump-rtl-split4
5208 @itemx -fdump-rtl-split5
5209 @opindex fdump-rtl-split1
5210 @opindex fdump-rtl-split2
5211 @opindex fdump-rtl-split3
5212 @opindex fdump-rtl-split4
5213 @opindex fdump-rtl-split5
5214 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5215 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5216 @option{-fdump-rtl-split5} enable dumping after five rounds of
5217 instruction splitting.
5219 @item -fdump-rtl-sms
5220 @opindex fdump-rtl-sms
5221 Dump after modulo scheduling. This pass is only run on some
5224 @item -fdump-rtl-stack
5225 @opindex fdump-rtl-stack
5226 Dump after conversion from GCC's "flat register file" registers to the
5227 x87's stack-like registers. This pass is only run on x86 variants.
5229 @item -fdump-rtl-subreg1
5230 @itemx -fdump-rtl-subreg2
5231 @opindex fdump-rtl-subreg1
5232 @opindex fdump-rtl-subreg2
5233 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5234 the two subreg expansion passes.
5236 @item -fdump-rtl-unshare
5237 @opindex fdump-rtl-unshare
5238 Dump after all rtl has been unshared.
5240 @item -fdump-rtl-vartrack
5241 @opindex fdump-rtl-vartrack
5242 Dump after variable tracking.
5244 @item -fdump-rtl-vregs
5245 @opindex fdump-rtl-vregs
5246 Dump after converting virtual registers to hard registers.
5248 @item -fdump-rtl-web
5249 @opindex fdump-rtl-web
5250 Dump after live range splitting.
5252 @item -fdump-rtl-regclass
5253 @itemx -fdump-rtl-subregs_of_mode_init
5254 @itemx -fdump-rtl-subregs_of_mode_finish
5255 @itemx -fdump-rtl-dfinit
5256 @itemx -fdump-rtl-dfinish
5257 @opindex fdump-rtl-regclass
5258 @opindex fdump-rtl-subregs_of_mode_init
5259 @opindex fdump-rtl-subregs_of_mode_finish
5260 @opindex fdump-rtl-dfinit
5261 @opindex fdump-rtl-dfinish
5262 These dumps are defined but always produce empty files.
5264 @item -fdump-rtl-all
5265 @opindex fdump-rtl-all
5266 Produce all the dumps listed above.
5270 Annotate the assembler output with miscellaneous debugging information.
5274 Dump all macro definitions, at the end of preprocessing, in addition to
5279 Produce a core dump whenever an error occurs.
5283 Print statistics on memory usage, at the end of the run, to
5288 Annotate the assembler output with a comment indicating which
5289 pattern and alternative was used. The length of each instruction is
5294 Dump the RTL in the assembler output as a comment before each instruction.
5295 Also turns on @option{-dp} annotation.
5299 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5300 dump a representation of the control flow graph suitable for viewing with VCG
5301 to @file{@var{file}.@var{pass}.vcg}.
5305 Just generate RTL for a function instead of compiling it. Usually used
5306 with @option{-fdump-rtl-expand}.
5310 @opindex fdump-noaddr
5311 When doing debugging dumps, suppress address output. This makes it more
5312 feasible to use diff on debugging dumps for compiler invocations with
5313 different compiler binaries and/or different
5314 text / bss / data / heap / stack / dso start locations.
5316 @item -fdump-unnumbered
5317 @opindex fdump-unnumbered
5318 When doing debugging dumps, suppress instruction numbers and address output.
5319 This makes it more feasible to use diff on debugging dumps for compiler
5320 invocations with different options, in particular with and without
5323 @item -fdump-unnumbered-links
5324 @opindex fdump-unnumbered-links
5325 When doing debugging dumps (see @option{-d} option above), suppress
5326 instruction numbers for the links to the previous and next instructions
5329 @item -fdump-translation-unit @r{(C++ only)}
5330 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5331 @opindex fdump-translation-unit
5332 Dump a representation of the tree structure for the entire translation
5333 unit to a file. The file name is made by appending @file{.tu} to the
5334 source file name, and the file is created in the same directory as the
5335 output file. If the @samp{-@var{options}} form is used, @var{options}
5336 controls the details of the dump as described for the
5337 @option{-fdump-tree} options.
5339 @item -fdump-class-hierarchy @r{(C++ only)}
5340 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5341 @opindex fdump-class-hierarchy
5342 Dump a representation of each class's hierarchy and virtual function
5343 table layout to a file. The file name is made by appending
5344 @file{.class} to the source file name, and the file is created in the
5345 same directory as the output file. If the @samp{-@var{options}} form
5346 is used, @var{options} controls the details of the dump as described
5347 for the @option{-fdump-tree} options.
5349 @item -fdump-ipa-@var{switch}
5351 Control the dumping at various stages of inter-procedural analysis
5352 language tree to a file. The file name is generated by appending a
5353 switch specific suffix to the source file name, and the file is created
5354 in the same directory as the output file. The following dumps are
5359 Enables all inter-procedural analysis dumps.
5362 Dumps information about call-graph optimization, unused function removal,
5363 and inlining decisions.
5366 Dump after function inlining.
5370 @item -fdump-statistics-@var{option}
5371 @opindex fdump-statistics
5372 Enable and control dumping of pass statistics in a separate file. The
5373 file name is generated by appending a suffix ending in
5374 @samp{.statistics} to the source file name, and the file is created in
5375 the same directory as the output file. If the @samp{-@var{option}}
5376 form is used, @samp{-stats} will cause counters to be summed over the
5377 whole compilation unit while @samp{-details} will dump every event as
5378 the passes generate them. The default with no option is to sum
5379 counters for each function compiled.
5381 @item -fdump-tree-@var{switch}
5382 @itemx -fdump-tree-@var{switch}-@var{options}
5384 Control the dumping at various stages of processing the intermediate
5385 language tree to a file. The file name is generated by appending a
5386 switch specific suffix to the source file name, and the file is
5387 created in the same directory as the output file. If the
5388 @samp{-@var{options}} form is used, @var{options} is a list of
5389 @samp{-} separated options that control the details of the dump. Not
5390 all options are applicable to all dumps, those which are not
5391 meaningful will be ignored. The following options are available
5395 Print the address of each node. Usually this is not meaningful as it
5396 changes according to the environment and source file. Its primary use
5397 is for tying up a dump file with a debug environment.
5399 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5400 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5401 use working backward from mangled names in the assembly file.
5403 Inhibit dumping of members of a scope or body of a function merely
5404 because that scope has been reached. Only dump such items when they
5405 are directly reachable by some other path. When dumping pretty-printed
5406 trees, this option inhibits dumping the bodies of control structures.
5408 Print a raw representation of the tree. By default, trees are
5409 pretty-printed into a C-like representation.
5411 Enable more detailed dumps (not honored by every dump option).
5413 Enable dumping various statistics about the pass (not honored by every dump
5416 Enable showing basic block boundaries (disabled in raw dumps).
5418 Enable showing virtual operands for every statement.
5420 Enable showing line numbers for statements.
5422 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5424 Enable showing the tree dump for each statement.
5426 Enable showing the EH region number holding each statement.
5428 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5429 and @option{lineno}.
5432 The following tree dumps are possible:
5436 @opindex fdump-tree-original
5437 Dump before any tree based optimization, to @file{@var{file}.original}.
5440 @opindex fdump-tree-optimized
5441 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5444 @opindex fdump-tree-gimple
5445 Dump each function before and after the gimplification pass to a file. The
5446 file name is made by appending @file{.gimple} to the source file name.
5449 @opindex fdump-tree-cfg
5450 Dump the control flow graph of each function to a file. The file name is
5451 made by appending @file{.cfg} to the source file name.
5454 @opindex fdump-tree-vcg
5455 Dump the control flow graph of each function to a file in VCG format. The
5456 file name is made by appending @file{.vcg} to the source file name. Note
5457 that if the file contains more than one function, the generated file cannot
5458 be used directly by VCG@. You will need to cut and paste each function's
5459 graph into its own separate file first.
5462 @opindex fdump-tree-ch
5463 Dump each function after copying loop headers. The file name is made by
5464 appending @file{.ch} to the source file name.
5467 @opindex fdump-tree-ssa
5468 Dump SSA related information to a file. The file name is made by appending
5469 @file{.ssa} to the source file name.
5472 @opindex fdump-tree-alias
5473 Dump aliasing information for each function. The file name is made by
5474 appending @file{.alias} to the source file name.
5477 @opindex fdump-tree-ccp
5478 Dump each function after CCP@. The file name is made by appending
5479 @file{.ccp} to the source file name.
5482 @opindex fdump-tree-storeccp
5483 Dump each function after STORE-CCP@. The file name is made by appending
5484 @file{.storeccp} to the source file name.
5487 @opindex fdump-tree-pre
5488 Dump trees after partial redundancy elimination. The file name is made
5489 by appending @file{.pre} to the source file name.
5492 @opindex fdump-tree-fre
5493 Dump trees after full redundancy elimination. The file name is made
5494 by appending @file{.fre} to the source file name.
5497 @opindex fdump-tree-copyprop
5498 Dump trees after copy propagation. The file name is made
5499 by appending @file{.copyprop} to the source file name.
5501 @item store_copyprop
5502 @opindex fdump-tree-store_copyprop
5503 Dump trees after store copy-propagation. The file name is made
5504 by appending @file{.store_copyprop} to the source file name.
5507 @opindex fdump-tree-dce
5508 Dump each function after dead code elimination. The file name is made by
5509 appending @file{.dce} to the source file name.
5512 @opindex fdump-tree-mudflap
5513 Dump each function after adding mudflap instrumentation. The file name is
5514 made by appending @file{.mudflap} to the source file name.
5517 @opindex fdump-tree-sra
5518 Dump each function after performing scalar replacement of aggregates. The
5519 file name is made by appending @file{.sra} to the source file name.
5522 @opindex fdump-tree-sink
5523 Dump each function after performing code sinking. The file name is made
5524 by appending @file{.sink} to the source file name.
5527 @opindex fdump-tree-dom
5528 Dump each function after applying dominator tree optimizations. The file
5529 name is made by appending @file{.dom} to the source file name.
5532 @opindex fdump-tree-dse
5533 Dump each function after applying dead store elimination. The file
5534 name is made by appending @file{.dse} to the source file name.
5537 @opindex fdump-tree-phiopt
5538 Dump each function after optimizing PHI nodes into straightline code. The file
5539 name is made by appending @file{.phiopt} to the source file name.
5542 @opindex fdump-tree-forwprop
5543 Dump each function after forward propagating single use variables. The file
5544 name is made by appending @file{.forwprop} to the source file name.
5547 @opindex fdump-tree-copyrename
5548 Dump each function after applying the copy rename optimization. The file
5549 name is made by appending @file{.copyrename} to the source file name.
5552 @opindex fdump-tree-nrv
5553 Dump each function after applying the named return value optimization on
5554 generic trees. The file name is made by appending @file{.nrv} to the source
5558 @opindex fdump-tree-vect
5559 Dump each function after applying vectorization of loops. The file name is
5560 made by appending @file{.vect} to the source file name.
5563 @opindex fdump-tree-slp
5564 Dump each function after applying vectorization of basic blocks. The file name
5565 is made by appending @file{.slp} to the source file name.
5568 @opindex fdump-tree-vrp
5569 Dump each function after Value Range Propagation (VRP). The file name
5570 is made by appending @file{.vrp} to the source file name.
5573 @opindex fdump-tree-all
5574 Enable all the available tree dumps with the flags provided in this option.
5577 @item -ftree-vectorizer-verbose=@var{n}
5578 @opindex ftree-vectorizer-verbose
5579 This option controls the amount of debugging output the vectorizer prints.
5580 This information is written to standard error, unless
5581 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5582 in which case it is output to the usual dump listing file, @file{.vect}.
5583 For @var{n}=0 no diagnostic information is reported.
5584 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5585 and the total number of loops that got vectorized.
5586 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5587 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5588 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5589 level that @option{-fdump-tree-vect-stats} uses.
5590 Higher verbosity levels mean either more information dumped for each
5591 reported loop, or same amount of information reported for more loops:
5592 if @var{n}=3, vectorizer cost model information is reported.
5593 If @var{n}=4, alignment related information is added to the reports.
5594 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5595 memory access-patterns) is added to the reports.
5596 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5597 that did not pass the first analysis phase (i.e., may not be countable, or
5598 may have complicated control-flow).
5599 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5600 If @var{n}=8, SLP related information is added to the reports.
5601 For @var{n}=9, all the information the vectorizer generates during its
5602 analysis and transformation is reported. This is the same verbosity level
5603 that @option{-fdump-tree-vect-details} uses.
5605 @item -frandom-seed=@var{string}
5606 @opindex frandom-seed
5607 This option provides a seed that GCC uses when it would otherwise use
5608 random numbers. It is used to generate certain symbol names
5609 that have to be different in every compiled file. It is also used to
5610 place unique stamps in coverage data files and the object files that
5611 produce them. You can use the @option{-frandom-seed} option to produce
5612 reproducibly identical object files.
5614 The @var{string} should be different for every file you compile.
5616 @item -fsched-verbose=@var{n}
5617 @opindex fsched-verbose
5618 On targets that use instruction scheduling, this option controls the
5619 amount of debugging output the scheduler prints. This information is
5620 written to standard error, unless @option{-fdump-rtl-sched1} or
5621 @option{-fdump-rtl-sched2} is specified, in which case it is output
5622 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5623 respectively. However for @var{n} greater than nine, the output is
5624 always printed to standard error.
5626 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5627 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5628 For @var{n} greater than one, it also output basic block probabilities,
5629 detailed ready list information and unit/insn info. For @var{n} greater
5630 than two, it includes RTL at abort point, control-flow and regions info.
5631 And for @var{n} over four, @option{-fsched-verbose} also includes
5635 @itemx -save-temps=cwd
5637 Store the usual ``temporary'' intermediate files permanently; place them
5638 in the current directory and name them based on the source file. Thus,
5639 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5640 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5641 preprocessed @file{foo.i} output file even though the compiler now
5642 normally uses an integrated preprocessor.
5644 When used in combination with the @option{-x} command line option,
5645 @option{-save-temps} is sensible enough to avoid over writing an
5646 input source file with the same extension as an intermediate file.
5647 The corresponding intermediate file may be obtained by renaming the
5648 source file before using @option{-save-temps}.
5650 If you invoke GCC in parallel, compiling several different source
5651 files that share a common base name in different subdirectories or the
5652 same source file compiled for multiple output destinations, it is
5653 likely that the different parallel compilers will interfere with each
5654 other, and overwrite the temporary files. For instance:
5657 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5658 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5661 may result in @file{foo.i} and @file{foo.o} being written to
5662 simultaneously by both compilers.
5664 @item -save-temps=obj
5665 @opindex save-temps=obj
5666 Store the usual ``temporary'' intermediate files permanently. If the
5667 @option{-o} option is used, the temporary files are based on the
5668 object file. If the @option{-o} option is not used, the
5669 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5674 gcc -save-temps=obj -c foo.c
5675 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5676 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5679 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5680 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5681 @file{dir2/yfoobar.o}.
5683 @item -time@r{[}=@var{file}@r{]}
5685 Report the CPU time taken by each subprocess in the compilation
5686 sequence. For C source files, this is the compiler proper and assembler
5687 (plus the linker if linking is done).
5689 Without the specification of an output file, the output looks like this:
5696 The first number on each line is the ``user time'', that is time spent
5697 executing the program itself. The second number is ``system time'',
5698 time spent executing operating system routines on behalf of the program.
5699 Both numbers are in seconds.
5701 With the specification of an output file, the output is appended to the
5702 named file, and it looks like this:
5705 0.12 0.01 cc1 @var{options}
5706 0.00 0.01 as @var{options}
5709 The ``user time'' and the ``system time'' are moved before the program
5710 name, and the options passed to the program are displayed, so that one
5711 can later tell what file was being compiled, and with which options.
5713 @item -fvar-tracking
5714 @opindex fvar-tracking
5715 Run variable tracking pass. It computes where variables are stored at each
5716 position in code. Better debugging information is then generated
5717 (if the debugging information format supports this information).
5719 It is enabled by default when compiling with optimization (@option{-Os},
5720 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5721 the debug info format supports it.
5723 @item -fvar-tracking-assignments
5724 @opindex fvar-tracking-assignments
5725 @opindex fno-var-tracking-assignments
5726 Annotate assignments to user variables early in the compilation and
5727 attempt to carry the annotations over throughout the compilation all the
5728 way to the end, in an attempt to improve debug information while
5729 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5731 It can be enabled even if var-tracking is disabled, in which case
5732 annotations will be created and maintained, but discarded at the end.
5734 @item -fvar-tracking-assignments-toggle
5735 @opindex fvar-tracking-assignments-toggle
5736 @opindex fno-var-tracking-assignments-toggle
5737 Toggle @option{-fvar-tracking-assignments}, in the same way that
5738 @option{-gtoggle} toggles @option{-g}.
5740 @item -print-file-name=@var{library}
5741 @opindex print-file-name
5742 Print the full absolute name of the library file @var{library} that
5743 would be used when linking---and don't do anything else. With this
5744 option, GCC does not compile or link anything; it just prints the
5747 @item -print-multi-directory
5748 @opindex print-multi-directory
5749 Print the directory name corresponding to the multilib selected by any
5750 other switches present in the command line. This directory is supposed
5751 to exist in @env{GCC_EXEC_PREFIX}.
5753 @item -print-multi-lib
5754 @opindex print-multi-lib
5755 Print the mapping from multilib directory names to compiler switches
5756 that enable them. The directory name is separated from the switches by
5757 @samp{;}, and each switch starts with an @samp{@@} instead of the
5758 @samp{-}, without spaces between multiple switches. This is supposed to
5759 ease shell-processing.
5761 @item -print-multi-os-directory
5762 @opindex print-multi-os-directory
5763 Print the path to OS libraries for the selected
5764 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5765 present in the @file{lib} subdirectory and no multilibs are used, this is
5766 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5767 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5768 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5769 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5771 @item -print-prog-name=@var{program}
5772 @opindex print-prog-name
5773 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5775 @item -print-libgcc-file-name
5776 @opindex print-libgcc-file-name
5777 Same as @option{-print-file-name=libgcc.a}.
5779 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5780 but you do want to link with @file{libgcc.a}. You can do
5783 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5786 @item -print-search-dirs
5787 @opindex print-search-dirs
5788 Print the name of the configured installation directory and a list of
5789 program and library directories @command{gcc} will search---and don't do anything else.
5791 This is useful when @command{gcc} prints the error message
5792 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5793 To resolve this you either need to put @file{cpp0} and the other compiler
5794 components where @command{gcc} expects to find them, or you can set the environment
5795 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5796 Don't forget the trailing @samp{/}.
5797 @xref{Environment Variables}.
5799 @item -print-sysroot
5800 @opindex print-sysroot
5801 Print the target sysroot directory that will be used during
5802 compilation. This is the target sysroot specified either at configure
5803 time or using the @option{--sysroot} option, possibly with an extra
5804 suffix that depends on compilation options. If no target sysroot is
5805 specified, the option prints nothing.
5807 @item -print-sysroot-headers-suffix
5808 @opindex print-sysroot-headers-suffix
5809 Print the suffix added to the target sysroot when searching for
5810 headers, or give an error if the compiler is not configured with such
5811 a suffix---and don't do anything else.
5814 @opindex dumpmachine
5815 Print the compiler's target machine (for example,
5816 @samp{i686-pc-linux-gnu})---and don't do anything else.
5819 @opindex dumpversion
5820 Print the compiler version (for example, @samp{3.0})---and don't do
5825 Print the compiler's built-in specs---and don't do anything else. (This
5826 is used when GCC itself is being built.) @xref{Spec Files}.
5828 @item -feliminate-unused-debug-types
5829 @opindex feliminate-unused-debug-types
5830 Normally, when producing DWARF2 output, GCC will emit debugging
5831 information for all types declared in a compilation
5832 unit, regardless of whether or not they are actually used
5833 in that compilation unit. Sometimes this is useful, such as
5834 if, in the debugger, you want to cast a value to a type that is
5835 not actually used in your program (but is declared). More often,
5836 however, this results in a significant amount of wasted space.
5837 With this option, GCC will avoid producing debug symbol output
5838 for types that are nowhere used in the source file being compiled.
5841 @node Optimize Options
5842 @section Options That Control Optimization
5843 @cindex optimize options
5844 @cindex options, optimization
5846 These options control various sorts of optimizations.
5848 Without any optimization option, the compiler's goal is to reduce the
5849 cost of compilation and to make debugging produce the expected
5850 results. Statements are independent: if you stop the program with a
5851 breakpoint between statements, you can then assign a new value to any
5852 variable or change the program counter to any other statement in the
5853 function and get exactly the results you would expect from the source
5856 Turning on optimization flags makes the compiler attempt to improve
5857 the performance and/or code size at the expense of compilation time
5858 and possibly the ability to debug the program.
5860 The compiler performs optimization based on the knowledge it has of the
5861 program. Compiling multiple files at once to a single output file mode allows
5862 the compiler to use information gained from all of the files when compiling
5865 Not all optimizations are controlled directly by a flag. Only
5866 optimizations that have a flag are listed in this section.
5868 Most optimizations are only enabled if an @option{-O} level is set on
5869 the command line. Otherwise they are disabled, even if individual
5870 optimization flags are specified.
5872 Depending on the target and how GCC was configured, a slightly different
5873 set of optimizations may be enabled at each @option{-O} level than
5874 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5875 to find out the exact set of optimizations that are enabled at each level.
5876 @xref{Overall Options}, for examples.
5883 Optimize. Optimizing compilation takes somewhat more time, and a lot
5884 more memory for a large function.
5886 With @option{-O}, the compiler tries to reduce code size and execution
5887 time, without performing any optimizations that take a great deal of
5890 @option{-O} turns on the following optimization flags:
5894 -fcprop-registers @gol
5897 -fdelayed-branch @gol
5899 -fguess-branch-probability @gol
5900 -fif-conversion2 @gol
5901 -fif-conversion @gol
5902 -fipa-pure-const @gol
5904 -fipa-reference @gol
5906 -fsplit-wide-types @gol
5908 -ftree-builtin-call-dce @gol
5911 -ftree-copyrename @gol
5913 -ftree-dominator-opts @gol
5915 -ftree-forwprop @gol
5923 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5924 where doing so does not interfere with debugging.
5928 Optimize even more. GCC performs nearly all supported optimizations
5929 that do not involve a space-speed tradeoff.
5930 As compared to @option{-O}, this option increases both compilation time
5931 and the performance of the generated code.
5933 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5934 also turns on the following optimization flags:
5935 @gccoptlist{-fthread-jumps @gol
5936 -falign-functions -falign-jumps @gol
5937 -falign-loops -falign-labels @gol
5940 -fcse-follow-jumps -fcse-skip-blocks @gol
5941 -fdelete-null-pointer-checks @gol
5943 -fexpensive-optimizations @gol
5944 -fgcse -fgcse-lm @gol
5945 -finline-small-functions @gol
5946 -findirect-inlining @gol
5948 -foptimize-sibling-calls @gol
5949 -fpartial-inlining @gol
5952 -freorder-blocks -freorder-functions @gol
5953 -frerun-cse-after-loop @gol
5954 -fsched-interblock -fsched-spec @gol
5955 -fschedule-insns -fschedule-insns2 @gol
5956 -fstrict-aliasing -fstrict-overflow @gol
5957 -ftree-switch-conversion @gol
5961 Please note the warning under @option{-fgcse} about
5962 invoking @option{-O2} on programs that use computed gotos.
5966 Optimize yet more. @option{-O3} turns on all optimizations specified
5967 by @option{-O2} and also turns on the @option{-finline-functions},
5968 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5969 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
5970 @option{-fipa-cp-clone} options.
5974 Reduce compilation time and make debugging produce the expected
5975 results. This is the default.
5979 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5980 do not typically increase code size. It also performs further
5981 optimizations designed to reduce code size.
5983 @option{-Os} disables the following optimization flags:
5984 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5985 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5986 -fprefetch-loop-arrays -ftree-vect-loop-version}
5990 Disregard strict standards compliance. @option{-Ofast} enables all
5991 @option{-O3} optimizations. It also enables optimizations that are not
5992 valid for all standard compliant programs.
5993 It turns on @option{-ffast-math}.
5995 If you use multiple @option{-O} options, with or without level numbers,
5996 the last such option is the one that is effective.
5999 Options of the form @option{-f@var{flag}} specify machine-independent
6000 flags. Most flags have both positive and negative forms; the negative
6001 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6002 below, only one of the forms is listed---the one you typically will
6003 use. You can figure out the other form by either removing @samp{no-}
6006 The following options control specific optimizations. They are either
6007 activated by @option{-O} options or are related to ones that are. You
6008 can use the following flags in the rare cases when ``fine-tuning'' of
6009 optimizations to be performed is desired.
6012 @item -fno-default-inline
6013 @opindex fno-default-inline
6014 Do not make member functions inline by default merely because they are
6015 defined inside the class scope (C++ only). Otherwise, when you specify
6016 @w{@option{-O}}, member functions defined inside class scope are compiled
6017 inline by default; i.e., you don't need to add @samp{inline} in front of
6018 the member function name.
6020 @item -fno-defer-pop
6021 @opindex fno-defer-pop
6022 Always pop the arguments to each function call as soon as that function
6023 returns. For machines which must pop arguments after a function call,
6024 the compiler normally lets arguments accumulate on the stack for several
6025 function calls and pops them all at once.
6027 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6029 @item -fforward-propagate
6030 @opindex fforward-propagate
6031 Perform a forward propagation pass on RTL@. The pass tries to combine two
6032 instructions and checks if the result can be simplified. If loop unrolling
6033 is active, two passes are performed and the second is scheduled after
6036 This option is enabled by default at optimization levels @option{-O},
6037 @option{-O2}, @option{-O3}, @option{-Os}.
6039 @item -ffp-contract=@var{style}
6040 @opindex ffp-contract
6041 @option{-ffp-contract=off} disables floating-point expression contraction.
6042 @option{-ffp-contract=fast} enables floating-point expression contraction
6043 such as forming of fused multiply-add operations if the target has
6044 native support for them.
6045 @option{-ffp-contract=on} enables floating-point expression contraction
6046 if allowed by the language standard. This is currently not implemented
6047 and treated equal to @option{-ffp-contract=off}.
6049 The default is @option{-ffp-contract=fast}.
6051 @item -fomit-frame-pointer
6052 @opindex fomit-frame-pointer
6053 Don't keep the frame pointer in a register for functions that
6054 don't need one. This avoids the instructions to save, set up and
6055 restore frame pointers; it also makes an extra register available
6056 in many functions. @strong{It also makes debugging impossible on
6059 On some machines, such as the VAX, this flag has no effect, because
6060 the standard calling sequence automatically handles the frame pointer
6061 and nothing is saved by pretending it doesn't exist. The
6062 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6063 whether a target machine supports this flag. @xref{Registers,,Register
6064 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6066 Starting with GCC version 4.6, the default setting (when not optimizing for
6067 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6068 @option{-fomit-frame-pointer}. The default can be reverted to
6069 @option{-fno-omit-frame-pointer} by configuring GCC with the
6070 @option{--enable-frame-pointer} configure option.
6072 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6074 @item -foptimize-sibling-calls
6075 @opindex foptimize-sibling-calls
6076 Optimize sibling and tail recursive calls.
6078 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6082 Don't pay attention to the @code{inline} keyword. Normally this option
6083 is used to keep the compiler from expanding any functions inline.
6084 Note that if you are not optimizing, no functions can be expanded inline.
6086 @item -finline-small-functions
6087 @opindex finline-small-functions
6088 Integrate functions into their callers when their body is smaller than expected
6089 function call code (so overall size of program gets smaller). The compiler
6090 heuristically decides which functions are simple enough to be worth integrating
6093 Enabled at level @option{-O2}.
6095 @item -findirect-inlining
6096 @opindex findirect-inlining
6097 Inline also indirect calls that are discovered to be known at compile
6098 time thanks to previous inlining. This option has any effect only
6099 when inlining itself is turned on by the @option{-finline-functions}
6100 or @option{-finline-small-functions} options.
6102 Enabled at level @option{-O2}.
6104 @item -finline-functions
6105 @opindex finline-functions
6106 Integrate all simple functions into their callers. The compiler
6107 heuristically decides which functions are simple enough to be worth
6108 integrating in this way.
6110 If all calls to a given function are integrated, and the function is
6111 declared @code{static}, then the function is normally not output as
6112 assembler code in its own right.
6114 Enabled at level @option{-O3}.
6116 @item -finline-functions-called-once
6117 @opindex finline-functions-called-once
6118 Consider all @code{static} functions called once for inlining into their
6119 caller even if they are not marked @code{inline}. If a call to a given
6120 function is integrated, then the function is not output as assembler code
6123 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6125 @item -fearly-inlining
6126 @opindex fearly-inlining
6127 Inline functions marked by @code{always_inline} and functions whose body seems
6128 smaller than the function call overhead early before doing
6129 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6130 makes profiling significantly cheaper and usually inlining faster on programs
6131 having large chains of nested wrapper functions.
6137 Perform interprocedural scalar replacement of aggregates, removal of
6138 unused parameters and replacement of parameters passed by reference
6139 by parameters passed by value.
6141 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6143 @item -finline-limit=@var{n}
6144 @opindex finline-limit
6145 By default, GCC limits the size of functions that can be inlined. This flag
6146 allows coarse control of this limit. @var{n} is the size of functions that
6147 can be inlined in number of pseudo instructions.
6149 Inlining is actually controlled by a number of parameters, which may be
6150 specified individually by using @option{--param @var{name}=@var{value}}.
6151 The @option{-finline-limit=@var{n}} option sets some of these parameters
6155 @item max-inline-insns-single
6156 is set to @var{n}/2.
6157 @item max-inline-insns-auto
6158 is set to @var{n}/2.
6161 See below for a documentation of the individual
6162 parameters controlling inlining and for the defaults of these parameters.
6164 @emph{Note:} there may be no value to @option{-finline-limit} that results
6165 in default behavior.
6167 @emph{Note:} pseudo instruction represents, in this particular context, an
6168 abstract measurement of function's size. In no way does it represent a count
6169 of assembly instructions and as such its exact meaning might change from one
6170 release to an another.
6172 @item -fno-keep-inline-dllexport
6173 @opindex -fno-keep-inline-dllexport
6174 This is a more fine-grained version of @option{-fkeep-inline-functions},
6175 which applies only to functions that are declared using the @code{dllexport}
6176 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6179 @item -fkeep-inline-functions
6180 @opindex fkeep-inline-functions
6181 In C, emit @code{static} functions that are declared @code{inline}
6182 into the object file, even if the function has been inlined into all
6183 of its callers. This switch does not affect functions using the
6184 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6185 inline functions into the object file.
6187 @item -fkeep-static-consts
6188 @opindex fkeep-static-consts
6189 Emit variables declared @code{static const} when optimization isn't turned
6190 on, even if the variables aren't referenced.
6192 GCC enables this option by default. If you want to force the compiler to
6193 check if the variable was referenced, regardless of whether or not
6194 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6196 @item -fmerge-constants
6197 @opindex fmerge-constants
6198 Attempt to merge identical constants (string constants and floating point
6199 constants) across compilation units.
6201 This option is the default for optimized compilation if the assembler and
6202 linker support it. Use @option{-fno-merge-constants} to inhibit this
6205 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6207 @item -fmerge-all-constants
6208 @opindex fmerge-all-constants
6209 Attempt to merge identical constants and identical variables.
6211 This option implies @option{-fmerge-constants}. In addition to
6212 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6213 arrays or initialized constant variables with integral or floating point
6214 types. Languages like C or C++ require each variable, including multiple
6215 instances of the same variable in recursive calls, to have distinct locations,
6216 so using this option will result in non-conforming
6219 @item -fmodulo-sched
6220 @opindex fmodulo-sched
6221 Perform swing modulo scheduling immediately before the first scheduling
6222 pass. This pass looks at innermost loops and reorders their
6223 instructions by overlapping different iterations.
6225 @item -fmodulo-sched-allow-regmoves
6226 @opindex fmodulo-sched-allow-regmoves
6227 Perform more aggressive SMS based modulo scheduling with register moves
6228 allowed. By setting this flag certain anti-dependences edges will be
6229 deleted which will trigger the generation of reg-moves based on the
6230 life-range analysis. This option is effective only with
6231 @option{-fmodulo-sched} enabled.
6233 @item -fno-branch-count-reg
6234 @opindex fno-branch-count-reg
6235 Do not use ``decrement and branch'' instructions on a count register,
6236 but instead generate a sequence of instructions that decrement a
6237 register, compare it against zero, then branch based upon the result.
6238 This option is only meaningful on architectures that support such
6239 instructions, which include x86, PowerPC, IA-64 and S/390.
6241 The default is @option{-fbranch-count-reg}.
6243 @item -fno-function-cse
6244 @opindex fno-function-cse
6245 Do not put function addresses in registers; make each instruction that
6246 calls a constant function contain the function's address explicitly.
6248 This option results in less efficient code, but some strange hacks
6249 that alter the assembler output may be confused by the optimizations
6250 performed when this option is not used.
6252 The default is @option{-ffunction-cse}
6254 @item -fno-zero-initialized-in-bss
6255 @opindex fno-zero-initialized-in-bss
6256 If the target supports a BSS section, GCC by default puts variables that
6257 are initialized to zero into BSS@. This can save space in the resulting
6260 This option turns off this behavior because some programs explicitly
6261 rely on variables going to the data section. E.g., so that the
6262 resulting executable can find the beginning of that section and/or make
6263 assumptions based on that.
6265 The default is @option{-fzero-initialized-in-bss}.
6267 @item -fmudflap -fmudflapth -fmudflapir
6271 @cindex bounds checking
6273 For front-ends that support it (C and C++), instrument all risky
6274 pointer/array dereferencing operations, some standard library
6275 string/heap functions, and some other associated constructs with
6276 range/validity tests. Modules so instrumented should be immune to
6277 buffer overflows, invalid heap use, and some other classes of C/C++
6278 programming errors. The instrumentation relies on a separate runtime
6279 library (@file{libmudflap}), which will be linked into a program if
6280 @option{-fmudflap} is given at link time. Run-time behavior of the
6281 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6282 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6285 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6286 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6287 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6288 instrumentation should ignore pointer reads. This produces less
6289 instrumentation (and therefore faster execution) and still provides
6290 some protection against outright memory corrupting writes, but allows
6291 erroneously read data to propagate within a program.
6293 @item -fthread-jumps
6294 @opindex fthread-jumps
6295 Perform optimizations where we check to see if a jump branches to a
6296 location where another comparison subsumed by the first is found. If
6297 so, the first branch is redirected to either the destination of the
6298 second branch or a point immediately following it, depending on whether
6299 the condition is known to be true or false.
6301 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6303 @item -fsplit-wide-types
6304 @opindex fsplit-wide-types
6305 When using a type that occupies multiple registers, such as @code{long
6306 long} on a 32-bit system, split the registers apart and allocate them
6307 independently. This normally generates better code for those types,
6308 but may make debugging more difficult.
6310 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6313 @item -fcse-follow-jumps
6314 @opindex fcse-follow-jumps
6315 In common subexpression elimination (CSE), scan through jump instructions
6316 when the target of the jump is not reached by any other path. For
6317 example, when CSE encounters an @code{if} statement with an
6318 @code{else} clause, CSE will follow the jump when the condition
6321 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6323 @item -fcse-skip-blocks
6324 @opindex fcse-skip-blocks
6325 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6326 follow jumps which conditionally skip over blocks. When CSE
6327 encounters a simple @code{if} statement with no else clause,
6328 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6329 body of the @code{if}.
6331 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6333 @item -frerun-cse-after-loop
6334 @opindex frerun-cse-after-loop
6335 Re-run common subexpression elimination after loop optimizations has been
6338 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6342 Perform a global common subexpression elimination pass.
6343 This pass also performs global constant and copy propagation.
6345 @emph{Note:} When compiling a program using computed gotos, a GCC
6346 extension, you may get better runtime performance if you disable
6347 the global common subexpression elimination pass by adding
6348 @option{-fno-gcse} to the command line.
6350 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6354 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6355 attempt to move loads which are only killed by stores into themselves. This
6356 allows a loop containing a load/store sequence to be changed to a load outside
6357 the loop, and a copy/store within the loop.
6359 Enabled by default when gcse is enabled.
6363 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6364 global common subexpression elimination. This pass will attempt to move
6365 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6366 loops containing a load/store sequence can be changed to a load before
6367 the loop and a store after the loop.
6369 Not enabled at any optimization level.
6373 When @option{-fgcse-las} is enabled, the global common subexpression
6374 elimination pass eliminates redundant loads that come after stores to the
6375 same memory location (both partial and full redundancies).
6377 Not enabled at any optimization level.
6379 @item -fgcse-after-reload
6380 @opindex fgcse-after-reload
6381 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6382 pass is performed after reload. The purpose of this pass is to cleanup
6385 @item -funsafe-loop-optimizations
6386 @opindex funsafe-loop-optimizations
6387 If given, the loop optimizer will assume that loop indices do not
6388 overflow, and that the loops with nontrivial exit condition are not
6389 infinite. This enables a wider range of loop optimizations even if
6390 the loop optimizer itself cannot prove that these assumptions are valid.
6391 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6392 if it finds this kind of loop.
6394 @item -fcrossjumping
6395 @opindex fcrossjumping
6396 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6397 resulting code may or may not perform better than without cross-jumping.
6399 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6401 @item -fauto-inc-dec
6402 @opindex fauto-inc-dec
6403 Combine increments or decrements of addresses with memory accesses.
6404 This pass is always skipped on architectures that do not have
6405 instructions to support this. Enabled by default at @option{-O} and
6406 higher on architectures that support this.
6410 Perform dead code elimination (DCE) on RTL@.
6411 Enabled by default at @option{-O} and higher.
6415 Perform dead store elimination (DSE) on RTL@.
6416 Enabled by default at @option{-O} and higher.
6418 @item -fif-conversion
6419 @opindex fif-conversion
6420 Attempt to transform conditional jumps into branch-less equivalents. This
6421 include use of conditional moves, min, max, set flags and abs instructions, and
6422 some tricks doable by standard arithmetics. The use of conditional execution
6423 on chips where it is available is controlled by @code{if-conversion2}.
6425 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6427 @item -fif-conversion2
6428 @opindex fif-conversion2
6429 Use conditional execution (where available) to transform conditional jumps into
6430 branch-less equivalents.
6432 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6434 @item -fdelete-null-pointer-checks
6435 @opindex fdelete-null-pointer-checks
6436 Assume that programs cannot safely dereference null pointers, and that
6437 no code or data element resides there. This enables simple constant
6438 folding optimizations at all optimization levels. In addition, other
6439 optimization passes in GCC use this flag to control global dataflow
6440 analyses that eliminate useless checks for null pointers; these assume
6441 that if a pointer is checked after it has already been dereferenced,
6444 Note however that in some environments this assumption is not true.
6445 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6446 for programs which depend on that behavior.
6448 Some targets, especially embedded ones, disable this option at all levels.
6449 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6450 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6451 are enabled independently at different optimization levels.
6453 @item -fdevirtualize
6454 @opindex fdevirtualize
6455 Attempt to convert calls to virtual functions to direct calls. This
6456 is done both within a procedure and interprocedurally as part of
6457 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6458 propagation (@option{-fipa-cp}).
6459 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6461 @item -fexpensive-optimizations
6462 @opindex fexpensive-optimizations
6463 Perform a number of minor optimizations that are relatively expensive.
6465 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6467 @item -foptimize-register-move
6469 @opindex foptimize-register-move
6471 Attempt to reassign register numbers in move instructions and as
6472 operands of other simple instructions in order to maximize the amount of
6473 register tying. This is especially helpful on machines with two-operand
6476 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6479 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6481 @item -fira-algorithm=@var{algorithm}
6482 Use specified coloring algorithm for the integrated register
6483 allocator. The @var{algorithm} argument should be @code{priority} or
6484 @code{CB}. The first algorithm specifies Chow's priority coloring,
6485 the second one specifies Chaitin-Briggs coloring. The second
6486 algorithm can be unimplemented for some architectures. If it is
6487 implemented, it is the default because Chaitin-Briggs coloring as a
6488 rule generates a better code.
6490 @item -fira-region=@var{region}
6491 Use specified regions for the integrated register allocator. The
6492 @var{region} argument should be one of @code{all}, @code{mixed}, or
6493 @code{one}. The first value means using all loops as register
6494 allocation regions, the second value which is the default means using
6495 all loops except for loops with small register pressure as the
6496 regions, and third one means using all function as a single region.
6497 The first value can give best result for machines with small size and
6498 irregular register set, the third one results in faster and generates
6499 decent code and the smallest size code, and the default value usually
6500 give the best results in most cases and for most architectures.
6502 @item -fira-loop-pressure
6503 @opindex fira-loop-pressure
6504 Use IRA to evaluate register pressure in loops for decision to move
6505 loop invariants. Usage of this option usually results in generation
6506 of faster and smaller code on machines with big register files (>= 32
6507 registers) but it can slow compiler down.
6509 This option is enabled at level @option{-O3} for some targets.
6511 @item -fno-ira-share-save-slots
6512 @opindex fno-ira-share-save-slots
6513 Switch off sharing stack slots used for saving call used hard
6514 registers living through a call. Each hard register will get a
6515 separate stack slot and as a result function stack frame will be
6518 @item -fno-ira-share-spill-slots
6519 @opindex fno-ira-share-spill-slots
6520 Switch off sharing stack slots allocated for pseudo-registers. Each
6521 pseudo-register which did not get a hard register will get a separate
6522 stack slot and as a result function stack frame will be bigger.
6524 @item -fira-verbose=@var{n}
6525 @opindex fira-verbose
6526 Set up how verbose dump file for the integrated register allocator
6527 will be. Default value is 5. If the value is greater or equal to 10,
6528 the dump file will be stderr as if the value were @var{n} minus 10.
6530 @item -fdelayed-branch
6531 @opindex fdelayed-branch
6532 If supported for the target machine, attempt to reorder instructions
6533 to exploit instruction slots available after delayed branch
6536 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6538 @item -fschedule-insns
6539 @opindex fschedule-insns
6540 If supported for the target machine, attempt to reorder instructions to
6541 eliminate execution stalls due to required data being unavailable. This
6542 helps machines that have slow floating point or memory load instructions
6543 by allowing other instructions to be issued until the result of the load
6544 or floating point instruction is required.
6546 Enabled at levels @option{-O2}, @option{-O3}.
6548 @item -fschedule-insns2
6549 @opindex fschedule-insns2
6550 Similar to @option{-fschedule-insns}, but requests an additional pass of
6551 instruction scheduling after register allocation has been done. This is
6552 especially useful on machines with a relatively small number of
6553 registers and where memory load instructions take more than one cycle.
6555 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6557 @item -fno-sched-interblock
6558 @opindex fno-sched-interblock
6559 Don't schedule instructions across basic blocks. This is normally
6560 enabled by default when scheduling before register allocation, i.e.@:
6561 with @option{-fschedule-insns} or at @option{-O2} or higher.
6563 @item -fno-sched-spec
6564 @opindex fno-sched-spec
6565 Don't allow speculative motion of non-load instructions. This is normally
6566 enabled by default when scheduling before register allocation, i.e.@:
6567 with @option{-fschedule-insns} or at @option{-O2} or higher.
6569 @item -fsched-pressure
6570 @opindex fsched-pressure
6571 Enable register pressure sensitive insn scheduling before the register
6572 allocation. This only makes sense when scheduling before register
6573 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6574 @option{-O2} or higher. Usage of this option can improve the
6575 generated code and decrease its size by preventing register pressure
6576 increase above the number of available hard registers and as a
6577 consequence register spills in the register allocation.
6579 @item -fsched-spec-load
6580 @opindex fsched-spec-load
6581 Allow speculative motion of some load instructions. This only makes
6582 sense when scheduling before register allocation, i.e.@: with
6583 @option{-fschedule-insns} or at @option{-O2} or higher.
6585 @item -fsched-spec-load-dangerous
6586 @opindex fsched-spec-load-dangerous
6587 Allow speculative motion of more load instructions. This only makes
6588 sense when scheduling before register allocation, i.e.@: with
6589 @option{-fschedule-insns} or at @option{-O2} or higher.
6591 @item -fsched-stalled-insns
6592 @itemx -fsched-stalled-insns=@var{n}
6593 @opindex fsched-stalled-insns
6594 Define how many insns (if any) can be moved prematurely from the queue
6595 of stalled insns into the ready list, during the second scheduling pass.
6596 @option{-fno-sched-stalled-insns} means that no insns will be moved
6597 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6598 on how many queued insns can be moved prematurely.
6599 @option{-fsched-stalled-insns} without a value is equivalent to
6600 @option{-fsched-stalled-insns=1}.
6602 @item -fsched-stalled-insns-dep
6603 @itemx -fsched-stalled-insns-dep=@var{n}
6604 @opindex fsched-stalled-insns-dep
6605 Define how many insn groups (cycles) will be examined for a dependency
6606 on a stalled insn that is candidate for premature removal from the queue
6607 of stalled insns. This has an effect only during the second scheduling pass,
6608 and only if @option{-fsched-stalled-insns} is used.
6609 @option{-fno-sched-stalled-insns-dep} is equivalent to
6610 @option{-fsched-stalled-insns-dep=0}.
6611 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6612 @option{-fsched-stalled-insns-dep=1}.
6614 @item -fsched2-use-superblocks
6615 @opindex fsched2-use-superblocks
6616 When scheduling after register allocation, do use superblock scheduling
6617 algorithm. Superblock scheduling allows motion across basic block boundaries
6618 resulting on faster schedules. This option is experimental, as not all machine
6619 descriptions used by GCC model the CPU closely enough to avoid unreliable
6620 results from the algorithm.
6622 This only makes sense when scheduling after register allocation, i.e.@: with
6623 @option{-fschedule-insns2} or at @option{-O2} or higher.
6625 @item -fsched-group-heuristic
6626 @opindex fsched-group-heuristic
6627 Enable the group heuristic in the scheduler. This heuristic favors
6628 the instruction that belongs to a schedule group. This is enabled
6629 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6630 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6632 @item -fsched-critical-path-heuristic
6633 @opindex fsched-critical-path-heuristic
6634 Enable the critical-path heuristic in the scheduler. This heuristic favors
6635 instructions on the critical path. This is enabled by default when
6636 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6637 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6639 @item -fsched-spec-insn-heuristic
6640 @opindex fsched-spec-insn-heuristic
6641 Enable the speculative instruction heuristic in the scheduler. This
6642 heuristic favors speculative instructions with greater dependency weakness.
6643 This is enabled by default when scheduling is enabled, i.e.@:
6644 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6645 or at @option{-O2} or higher.
6647 @item -fsched-rank-heuristic
6648 @opindex fsched-rank-heuristic
6649 Enable the rank heuristic in the scheduler. This heuristic favors
6650 the instruction belonging to a basic block with greater size or frequency.
6651 This is enabled by default when scheduling is enabled, i.e.@:
6652 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6653 at @option{-O2} or higher.
6655 @item -fsched-last-insn-heuristic
6656 @opindex fsched-last-insn-heuristic
6657 Enable the last-instruction heuristic in the scheduler. This heuristic
6658 favors the instruction that is less dependent on the last instruction
6659 scheduled. This is enabled by default when scheduling is enabled,
6660 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6661 at @option{-O2} or higher.
6663 @item -fsched-dep-count-heuristic
6664 @opindex fsched-dep-count-heuristic
6665 Enable the dependent-count heuristic in the scheduler. This heuristic
6666 favors the instruction that has more instructions depending on it.
6667 This is enabled by default when scheduling is enabled, i.e.@:
6668 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6669 at @option{-O2} or higher.
6671 @item -freschedule-modulo-scheduled-loops
6672 @opindex freschedule-modulo-scheduled-loops
6673 The modulo scheduling comes before the traditional scheduling, if a loop
6674 was modulo scheduled we may want to prevent the later scheduling passes
6675 from changing its schedule, we use this option to control that.
6677 @item -fselective-scheduling
6678 @opindex fselective-scheduling
6679 Schedule instructions using selective scheduling algorithm. Selective
6680 scheduling runs instead of the first scheduler pass.
6682 @item -fselective-scheduling2
6683 @opindex fselective-scheduling2
6684 Schedule instructions using selective scheduling algorithm. Selective
6685 scheduling runs instead of the second scheduler pass.
6687 @item -fsel-sched-pipelining
6688 @opindex fsel-sched-pipelining
6689 Enable software pipelining of innermost loops during selective scheduling.
6690 This option has no effect until one of @option{-fselective-scheduling} or
6691 @option{-fselective-scheduling2} is turned on.
6693 @item -fsel-sched-pipelining-outer-loops
6694 @opindex fsel-sched-pipelining-outer-loops
6695 When pipelining loops during selective scheduling, also pipeline outer loops.
6696 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6698 @item -fcaller-saves
6699 @opindex fcaller-saves
6700 Enable values to be allocated in registers that will be clobbered by
6701 function calls, by emitting extra instructions to save and restore the
6702 registers around such calls. Such allocation is done only when it
6703 seems to result in better code than would otherwise be produced.
6705 This option is always enabled by default on certain machines, usually
6706 those which have no call-preserved registers to use instead.
6708 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6710 @item -fcombine-stack-adjustments
6711 @opindex fcombine-stack-adjustments
6712 Tracks stack adjustments (pushes and pops) and stack memory references
6713 and then tries to find ways to combine them.
6715 Enabled by default at @option{-O1} and higher.
6717 @item -fconserve-stack
6718 @opindex fconserve-stack
6719 Attempt to minimize stack usage. The compiler will attempt to use less
6720 stack space, even if that makes the program slower. This option
6721 implies setting the @option{large-stack-frame} parameter to 100
6722 and the @option{large-stack-frame-growth} parameter to 400.
6724 @item -ftree-reassoc
6725 @opindex ftree-reassoc
6726 Perform reassociation on trees. This flag is enabled by default
6727 at @option{-O} and higher.
6731 Perform partial redundancy elimination (PRE) on trees. This flag is
6732 enabled by default at @option{-O2} and @option{-O3}.
6734 @item -ftree-forwprop
6735 @opindex ftree-forwprop
6736 Perform forward propagation on trees. This flag is enabled by default
6737 at @option{-O} and higher.
6741 Perform full redundancy elimination (FRE) on trees. The difference
6742 between FRE and PRE is that FRE only considers expressions
6743 that are computed on all paths leading to the redundant computation.
6744 This analysis is faster than PRE, though it exposes fewer redundancies.
6745 This flag is enabled by default at @option{-O} and higher.
6747 @item -ftree-phiprop
6748 @opindex ftree-phiprop
6749 Perform hoisting of loads from conditional pointers on trees. This
6750 pass is enabled by default at @option{-O} and higher.
6752 @item -ftree-copy-prop
6753 @opindex ftree-copy-prop
6754 Perform copy propagation on trees. This pass eliminates unnecessary
6755 copy operations. This flag is enabled by default at @option{-O} and
6758 @item -fipa-pure-const
6759 @opindex fipa-pure-const
6760 Discover which functions are pure or constant.
6761 Enabled by default at @option{-O} and higher.
6763 @item -fipa-reference
6764 @opindex fipa-reference
6765 Discover which static variables do not escape cannot escape the
6767 Enabled by default at @option{-O} and higher.
6769 @item -fipa-struct-reorg
6770 @opindex fipa-struct-reorg
6771 Perform structure reorganization optimization, that change C-like structures
6772 layout in order to better utilize spatial locality. This transformation is
6773 affective for programs containing arrays of structures. Available in two
6774 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6775 or static (which uses built-in heuristics). It works only in whole program
6776 mode, so it requires @option{-fwhole-program} to be
6777 enabled. Structures considered @samp{cold} by this transformation are not
6778 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6780 With this flag, the program debug info reflects a new structure layout.
6784 Perform interprocedural pointer analysis and interprocedural modification
6785 and reference analysis. This option can cause excessive memory and
6786 compile-time usage on large compilation units. It is not enabled by
6787 default at any optimization level.
6790 @opindex fipa-profile
6791 Perform interprocedural profile propagation. The functions called only from
6792 cold functions are marked as cold. Also functions executed once (such as
6793 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6794 functions and loop less parts of functions executed once are then optimized for
6796 Enabled by default at @option{-O} and higher.
6800 Perform interprocedural constant propagation.
6801 This optimization analyzes the program to determine when values passed
6802 to functions are constants and then optimizes accordingly.
6803 This optimization can substantially increase performance
6804 if the application has constants passed to functions.
6805 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6807 @item -fipa-cp-clone
6808 @opindex fipa-cp-clone
6809 Perform function cloning to make interprocedural constant propagation stronger.
6810 When enabled, interprocedural constant propagation will perform function cloning
6811 when externally visible function can be called with constant arguments.
6812 Because this optimization can create multiple copies of functions,
6813 it may significantly increase code size
6814 (see @option{--param ipcp-unit-growth=@var{value}}).
6815 This flag is enabled by default at @option{-O3}.
6817 @item -fipa-matrix-reorg
6818 @opindex fipa-matrix-reorg
6819 Perform matrix flattening and transposing.
6820 Matrix flattening tries to replace an @math{m}-dimensional matrix
6821 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6822 This reduces the level of indirection needed for accessing the elements
6823 of the matrix. The second optimization is matrix transposing that
6824 attempts to change the order of the matrix's dimensions in order to
6825 improve cache locality.
6826 Both optimizations need the @option{-fwhole-program} flag.
6827 Transposing is enabled only if profiling information is available.
6831 Perform forward store motion on trees. This flag is
6832 enabled by default at @option{-O} and higher.
6834 @item -ftree-bit-ccp
6835 @opindex ftree-bit-ccp
6836 Perform sparse conditional bit constant propagation on trees and propagate
6837 pointer alignment information.
6838 This pass only operates on local scalar variables and is enabled by default
6839 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
6843 Perform sparse conditional constant propagation (CCP) on trees. This
6844 pass only operates on local scalar variables and is enabled by default
6845 at @option{-O} and higher.
6847 @item -ftree-switch-conversion
6848 Perform conversion of simple initializations in a switch to
6849 initializations from a scalar array. This flag is enabled by default
6850 at @option{-O2} and higher.
6854 Perform dead code elimination (DCE) on trees. This flag is enabled by
6855 default at @option{-O} and higher.
6857 @item -ftree-builtin-call-dce
6858 @opindex ftree-builtin-call-dce
6859 Perform conditional dead code elimination (DCE) for calls to builtin functions
6860 that may set @code{errno} but are otherwise side-effect free. This flag is
6861 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6864 @item -ftree-dominator-opts
6865 @opindex ftree-dominator-opts
6866 Perform a variety of simple scalar cleanups (constant/copy
6867 propagation, redundancy elimination, range propagation and expression
6868 simplification) based on a dominator tree traversal. This also
6869 performs jump threading (to reduce jumps to jumps). This flag is
6870 enabled by default at @option{-O} and higher.
6874 Perform dead store elimination (DSE) on trees. A dead store is a store into
6875 a memory location which will later be overwritten by another store without
6876 any intervening loads. In this case the earlier store can be deleted. This
6877 flag is enabled by default at @option{-O} and higher.
6881 Perform loop header copying on trees. This is beneficial since it increases
6882 effectiveness of code motion optimizations. It also saves one jump. This flag
6883 is enabled by default at @option{-O} and higher. It is not enabled
6884 for @option{-Os}, since it usually increases code size.
6886 @item -ftree-loop-optimize
6887 @opindex ftree-loop-optimize
6888 Perform loop optimizations on trees. This flag is enabled by default
6889 at @option{-O} and higher.
6891 @item -ftree-loop-linear
6892 @opindex ftree-loop-linear
6893 Perform loop interchange transformations on tree. Same as
6894 @option{-floop-interchange}. To use this code transformation, GCC has
6895 to be configured with @option{--with-ppl} and @option{--with-cloog} to
6896 enable the Graphite loop transformation infrastructure.
6898 @item -floop-interchange
6899 @opindex floop-interchange
6900 Perform loop interchange transformations on loops. Interchanging two
6901 nested loops switches the inner and outer loops. For example, given a
6906 A(J, I) = A(J, I) * C
6910 loop interchange will transform the loop as if the user had written:
6914 A(J, I) = A(J, I) * C
6918 which can be beneficial when @code{N} is larger than the caches,
6919 because in Fortran, the elements of an array are stored in memory
6920 contiguously by column, and the original loop iterates over rows,
6921 potentially creating at each access a cache miss. This optimization
6922 applies to all the languages supported by GCC and is not limited to
6923 Fortran. To use this code transformation, GCC has to be configured
6924 with @option{--with-ppl} and @option{--with-cloog} to enable the
6925 Graphite loop transformation infrastructure.
6927 @item -floop-strip-mine
6928 @opindex floop-strip-mine
6929 Perform loop strip mining transformations on loops. Strip mining
6930 splits a loop into two nested loops. The outer loop has strides
6931 equal to the strip size and the inner loop has strides of the
6932 original loop within a strip. The strip length can be changed
6933 using the @option{loop-block-tile-size} parameter. For example,
6940 loop strip mining will transform the loop as if the user had written:
6943 DO I = II, min (II + 50, N)
6948 This optimization applies to all the languages supported by GCC and is
6949 not limited to Fortran. To use this code transformation, GCC has to
6950 be configured with @option{--with-ppl} and @option{--with-cloog} to
6951 enable the Graphite loop transformation infrastructure.
6954 @opindex floop-block
6955 Perform loop blocking transformations on loops. Blocking strip mines
6956 each loop in the loop nest such that the memory accesses of the
6957 element loops fit inside caches. The strip length can be changed
6958 using the @option{loop-block-tile-size} parameter. For example, given
6963 A(J, I) = B(I) + C(J)
6967 loop blocking will transform the loop as if the user had written:
6971 DO I = II, min (II + 50, N)
6972 DO J = JJ, min (JJ + 50, M)
6973 A(J, I) = B(I) + C(J)
6979 which can be beneficial when @code{M} is larger than the caches,
6980 because the innermost loop will iterate over a smaller amount of data
6981 that can be kept in the caches. This optimization applies to all the
6982 languages supported by GCC and is not limited to Fortran. To use this
6983 code transformation, GCC has to be configured with @option{--with-ppl}
6984 and @option{--with-cloog} to enable the Graphite loop transformation
6987 @item -fgraphite-identity
6988 @opindex fgraphite-identity
6989 Enable the identity transformation for graphite. For every SCoP we generate
6990 the polyhedral representation and transform it back to gimple. Using
6991 @option{-fgraphite-identity} we can check the costs or benefits of the
6992 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6993 are also performed by the code generator CLooG, like index splitting and
6994 dead code elimination in loops.
6996 @item -floop-flatten
6997 @opindex floop-flatten
6998 Removes the loop nesting structure: transforms the loop nest into a
6999 single loop. This transformation can be useful to vectorize all the
7000 levels of the loop nest.
7002 @item -floop-parallelize-all
7003 @opindex floop-parallelize-all
7004 Use the Graphite data dependence analysis to identify loops that can
7005 be parallelized. Parallelize all the loops that can be analyzed to
7006 not contain loop carried dependences without checking that it is
7007 profitable to parallelize the loops.
7009 @item -fcheck-data-deps
7010 @opindex fcheck-data-deps
7011 Compare the results of several data dependence analyzers. This option
7012 is used for debugging the data dependence analyzers.
7014 @item -ftree-loop-if-convert
7015 Attempt to transform conditional jumps in the innermost loops to
7016 branch-less equivalents. The intent is to remove control-flow from
7017 the innermost loops in order to improve the ability of the
7018 vectorization pass to handle these loops. This is enabled by default
7019 if vectorization is enabled.
7021 @item -ftree-loop-if-convert-stores
7022 Attempt to also if-convert conditional jumps containing memory writes.
7023 This transformation can be unsafe for multi-threaded programs as it
7024 transforms conditional memory writes into unconditional memory writes.
7027 for (i = 0; i < N; i++)
7031 would be transformed to
7033 for (i = 0; i < N; i++)
7034 A[i] = cond ? expr : A[i];
7036 potentially producing data races.
7038 @item -ftree-loop-distribution
7039 Perform loop distribution. This flag can improve cache performance on
7040 big loop bodies and allow further loop optimizations, like
7041 parallelization or vectorization, to take place. For example, the loop
7058 @item -ftree-loop-distribute-patterns
7059 Perform loop distribution of patterns that can be code generated with
7060 calls to a library. This flag is enabled by default at @option{-O3}.
7062 This pass distributes the initialization loops and generates a call to
7063 memset zero. For example, the loop
7079 and the initialization loop is transformed into a call to memset zero.
7081 @item -ftree-loop-im
7082 @opindex ftree-loop-im
7083 Perform loop invariant motion on trees. This pass moves only invariants that
7084 would be hard to handle at RTL level (function calls, operations that expand to
7085 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7086 operands of conditions that are invariant out of the loop, so that we can use
7087 just trivial invariantness analysis in loop unswitching. The pass also includes
7090 @item -ftree-loop-ivcanon
7091 @opindex ftree-loop-ivcanon
7092 Create a canonical counter for number of iterations in the loop for that
7093 determining number of iterations requires complicated analysis. Later
7094 optimizations then may determine the number easily. Useful especially
7095 in connection with unrolling.
7099 Perform induction variable optimizations (strength reduction, induction
7100 variable merging and induction variable elimination) on trees.
7102 @item -ftree-parallelize-loops=n
7103 @opindex ftree-parallelize-loops
7104 Parallelize loops, i.e., split their iteration space to run in n threads.
7105 This is only possible for loops whose iterations are independent
7106 and can be arbitrarily reordered. The optimization is only
7107 profitable on multiprocessor machines, for loops that are CPU-intensive,
7108 rather than constrained e.g.@: by memory bandwidth. This option
7109 implies @option{-pthread}, and thus is only supported on targets
7110 that have support for @option{-pthread}.
7114 Perform function-local points-to analysis on trees. This flag is
7115 enabled by default at @option{-O} and higher.
7119 Perform scalar replacement of aggregates. This pass replaces structure
7120 references with scalars to prevent committing structures to memory too
7121 early. This flag is enabled by default at @option{-O} and higher.
7123 @item -ftree-copyrename
7124 @opindex ftree-copyrename
7125 Perform copy renaming on trees. This pass attempts to rename compiler
7126 temporaries to other variables at copy locations, usually resulting in
7127 variable names which more closely resemble the original variables. This flag
7128 is enabled by default at @option{-O} and higher.
7132 Perform temporary expression replacement during the SSA->normal phase. Single
7133 use/single def temporaries are replaced at their use location with their
7134 defining expression. This results in non-GIMPLE code, but gives the expanders
7135 much more complex trees to work on resulting in better RTL generation. This is
7136 enabled by default at @option{-O} and higher.
7138 @item -ftree-vectorize
7139 @opindex ftree-vectorize
7140 Perform loop vectorization on trees. This flag is enabled by default at
7143 @item -ftree-slp-vectorize
7144 @opindex ftree-slp-vectorize
7145 Perform basic block vectorization on trees. This flag is enabled by default at
7146 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7148 @item -ftree-vect-loop-version
7149 @opindex ftree-vect-loop-version
7150 Perform loop versioning when doing loop vectorization on trees. When a loop
7151 appears to be vectorizable except that data alignment or data dependence cannot
7152 be determined at compile time then vectorized and non-vectorized versions of
7153 the loop are generated along with runtime checks for alignment or dependence
7154 to control which version is executed. This option is enabled by default
7155 except at level @option{-Os} where it is disabled.
7157 @item -fvect-cost-model
7158 @opindex fvect-cost-model
7159 Enable cost model for vectorization.
7163 Perform Value Range Propagation on trees. This is similar to the
7164 constant propagation pass, but instead of values, ranges of values are
7165 propagated. This allows the optimizers to remove unnecessary range
7166 checks like array bound checks and null pointer checks. This is
7167 enabled by default at @option{-O2} and higher. Null pointer check
7168 elimination is only done if @option{-fdelete-null-pointer-checks} is
7173 Perform tail duplication to enlarge superblock size. This transformation
7174 simplifies the control flow of the function allowing other optimizations to do
7177 @item -funroll-loops
7178 @opindex funroll-loops
7179 Unroll loops whose number of iterations can be determined at compile
7180 time or upon entry to the loop. @option{-funroll-loops} implies
7181 @option{-frerun-cse-after-loop}. This option makes code larger,
7182 and may or may not make it run faster.
7184 @item -funroll-all-loops
7185 @opindex funroll-all-loops
7186 Unroll all loops, even if their number of iterations is uncertain when
7187 the loop is entered. This usually makes programs run more slowly.
7188 @option{-funroll-all-loops} implies the same options as
7189 @option{-funroll-loops},
7191 @item -fsplit-ivs-in-unroller
7192 @opindex fsplit-ivs-in-unroller
7193 Enables expressing of values of induction variables in later iterations
7194 of the unrolled loop using the value in the first iteration. This breaks
7195 long dependency chains, thus improving efficiency of the scheduling passes.
7197 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7198 same effect. However in cases the loop body is more complicated than
7199 a single basic block, this is not reliable. It also does not work at all
7200 on some of the architectures due to restrictions in the CSE pass.
7202 This optimization is enabled by default.
7204 @item -fvariable-expansion-in-unroller
7205 @opindex fvariable-expansion-in-unroller
7206 With this option, the compiler will create multiple copies of some
7207 local variables when unrolling a loop which can result in superior code.
7209 @item -fpartial-inlining
7210 @opindex fpartial-inlining
7211 Inline parts of functions. This option has any effect only
7212 when inlining itself is turned on by the @option{-finline-functions}
7213 or @option{-finline-small-functions} options.
7215 Enabled at level @option{-O2}.
7217 @item -fpredictive-commoning
7218 @opindex fpredictive-commoning
7219 Perform predictive commoning optimization, i.e., reusing computations
7220 (especially memory loads and stores) performed in previous
7221 iterations of loops.
7223 This option is enabled at level @option{-O3}.
7225 @item -fprefetch-loop-arrays
7226 @opindex fprefetch-loop-arrays
7227 If supported by the target machine, generate instructions to prefetch
7228 memory to improve the performance of loops that access large arrays.
7230 This option may generate better or worse code; results are highly
7231 dependent on the structure of loops within the source code.
7233 Disabled at level @option{-Os}.
7236 @itemx -fno-peephole2
7237 @opindex fno-peephole
7238 @opindex fno-peephole2
7239 Disable any machine-specific peephole optimizations. The difference
7240 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7241 are implemented in the compiler; some targets use one, some use the
7242 other, a few use both.
7244 @option{-fpeephole} is enabled by default.
7245 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7247 @item -fno-guess-branch-probability
7248 @opindex fno-guess-branch-probability
7249 Do not guess branch probabilities using heuristics.
7251 GCC will use heuristics to guess branch probabilities if they are
7252 not provided by profiling feedback (@option{-fprofile-arcs}). These
7253 heuristics are based on the control flow graph. If some branch probabilities
7254 are specified by @samp{__builtin_expect}, then the heuristics will be
7255 used to guess branch probabilities for the rest of the control flow graph,
7256 taking the @samp{__builtin_expect} info into account. The interactions
7257 between the heuristics and @samp{__builtin_expect} can be complex, and in
7258 some cases, it may be useful to disable the heuristics so that the effects
7259 of @samp{__builtin_expect} are easier to understand.
7261 The default is @option{-fguess-branch-probability} at levels
7262 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7264 @item -freorder-blocks
7265 @opindex freorder-blocks
7266 Reorder basic blocks in the compiled function in order to reduce number of
7267 taken branches and improve code locality.
7269 Enabled at levels @option{-O2}, @option{-O3}.
7271 @item -freorder-blocks-and-partition
7272 @opindex freorder-blocks-and-partition
7273 In addition to reordering basic blocks in the compiled function, in order
7274 to reduce number of taken branches, partitions hot and cold basic blocks
7275 into separate sections of the assembly and .o files, to improve
7276 paging and cache locality performance.
7278 This optimization is automatically turned off in the presence of
7279 exception handling, for linkonce sections, for functions with a user-defined
7280 section attribute and on any architecture that does not support named
7283 @item -freorder-functions
7284 @opindex freorder-functions
7285 Reorder functions in the object file in order to
7286 improve code locality. This is implemented by using special
7287 subsections @code{.text.hot} for most frequently executed functions and
7288 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7289 the linker so object file format must support named sections and linker must
7290 place them in a reasonable way.
7292 Also profile feedback must be available in to make this option effective. See
7293 @option{-fprofile-arcs} for details.
7295 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7297 @item -fstrict-aliasing
7298 @opindex fstrict-aliasing
7299 Allow the compiler to assume the strictest aliasing rules applicable to
7300 the language being compiled. For C (and C++), this activates
7301 optimizations based on the type of expressions. In particular, an
7302 object of one type is assumed never to reside at the same address as an
7303 object of a different type, unless the types are almost the same. For
7304 example, an @code{unsigned int} can alias an @code{int}, but not a
7305 @code{void*} or a @code{double}. A character type may alias any other
7308 @anchor{Type-punning}Pay special attention to code like this:
7321 The practice of reading from a different union member than the one most
7322 recently written to (called ``type-punning'') is common. Even with
7323 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7324 is accessed through the union type. So, the code above will work as
7325 expected. @xref{Structures unions enumerations and bit-fields
7326 implementation}. However, this code might not:
7337 Similarly, access by taking the address, casting the resulting pointer
7338 and dereferencing the result has undefined behavior, even if the cast
7339 uses a union type, e.g.:
7343 return ((union a_union *) &d)->i;
7347 The @option{-fstrict-aliasing} option is enabled at levels
7348 @option{-O2}, @option{-O3}, @option{-Os}.
7350 @item -fstrict-overflow
7351 @opindex fstrict-overflow
7352 Allow the compiler to assume strict signed overflow rules, depending
7353 on the language being compiled. For C (and C++) this means that
7354 overflow when doing arithmetic with signed numbers is undefined, which
7355 means that the compiler may assume that it will not happen. This
7356 permits various optimizations. For example, the compiler will assume
7357 that an expression like @code{i + 10 > i} will always be true for
7358 signed @code{i}. This assumption is only valid if signed overflow is
7359 undefined, as the expression is false if @code{i + 10} overflows when
7360 using twos complement arithmetic. When this option is in effect any
7361 attempt to determine whether an operation on signed numbers will
7362 overflow must be written carefully to not actually involve overflow.
7364 This option also allows the compiler to assume strict pointer
7365 semantics: given a pointer to an object, if adding an offset to that
7366 pointer does not produce a pointer to the same object, the addition is
7367 undefined. This permits the compiler to conclude that @code{p + u >
7368 p} is always true for a pointer @code{p} and unsigned integer
7369 @code{u}. This assumption is only valid because pointer wraparound is
7370 undefined, as the expression is false if @code{p + u} overflows using
7371 twos complement arithmetic.
7373 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7374 that integer signed overflow is fully defined: it wraps. When
7375 @option{-fwrapv} is used, there is no difference between
7376 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7377 integers. With @option{-fwrapv} certain types of overflow are
7378 permitted. For example, if the compiler gets an overflow when doing
7379 arithmetic on constants, the overflowed value can still be used with
7380 @option{-fwrapv}, but not otherwise.
7382 The @option{-fstrict-overflow} option is enabled at levels
7383 @option{-O2}, @option{-O3}, @option{-Os}.
7385 @item -falign-functions
7386 @itemx -falign-functions=@var{n}
7387 @opindex falign-functions
7388 Align the start of functions to the next power-of-two greater than
7389 @var{n}, skipping up to @var{n} bytes. For instance,
7390 @option{-falign-functions=32} aligns functions to the next 32-byte
7391 boundary, but @option{-falign-functions=24} would align to the next
7392 32-byte boundary only if this can be done by skipping 23 bytes or less.
7394 @option{-fno-align-functions} and @option{-falign-functions=1} are
7395 equivalent and mean that functions will not be aligned.
7397 Some assemblers only support this flag when @var{n} is a power of two;
7398 in that case, it is rounded up.
7400 If @var{n} is not specified or is zero, use a machine-dependent default.
7402 Enabled at levels @option{-O2}, @option{-O3}.
7404 @item -falign-labels
7405 @itemx -falign-labels=@var{n}
7406 @opindex falign-labels
7407 Align all branch targets to a power-of-two boundary, skipping up to
7408 @var{n} bytes like @option{-falign-functions}. This option can easily
7409 make code slower, because it must insert dummy operations for when the
7410 branch target is reached in the usual flow of the code.
7412 @option{-fno-align-labels} and @option{-falign-labels=1} are
7413 equivalent and mean that labels will not be aligned.
7415 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7416 are greater than this value, then their values are used instead.
7418 If @var{n} is not specified or is zero, use a machine-dependent default
7419 which is very likely to be @samp{1}, meaning no alignment.
7421 Enabled at levels @option{-O2}, @option{-O3}.
7424 @itemx -falign-loops=@var{n}
7425 @opindex falign-loops
7426 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7427 like @option{-falign-functions}. The hope is that the loop will be
7428 executed many times, which will make up for any execution of the dummy
7431 @option{-fno-align-loops} and @option{-falign-loops=1} are
7432 equivalent and mean that loops will not be aligned.
7434 If @var{n} is not specified or is zero, use a machine-dependent default.
7436 Enabled at levels @option{-O2}, @option{-O3}.
7439 @itemx -falign-jumps=@var{n}
7440 @opindex falign-jumps
7441 Align branch targets to a power-of-two boundary, for branch targets
7442 where the targets can only be reached by jumping, skipping up to @var{n}
7443 bytes like @option{-falign-functions}. In this case, no dummy operations
7446 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7447 equivalent and mean that loops will not be aligned.
7449 If @var{n} is not specified or is zero, use a machine-dependent default.
7451 Enabled at levels @option{-O2}, @option{-O3}.
7453 @item -funit-at-a-time
7454 @opindex funit-at-a-time
7455 This option is left for compatibility reasons. @option{-funit-at-a-time}
7456 has no effect, while @option{-fno-unit-at-a-time} implies
7457 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7461 @item -fno-toplevel-reorder
7462 @opindex fno-toplevel-reorder
7463 Do not reorder top-level functions, variables, and @code{asm}
7464 statements. Output them in the same order that they appear in the
7465 input file. When this option is used, unreferenced static variables
7466 will not be removed. This option is intended to support existing code
7467 which relies on a particular ordering. For new code, it is better to
7470 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7471 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7476 Constructs webs as commonly used for register allocation purposes and assign
7477 each web individual pseudo register. This allows the register allocation pass
7478 to operate on pseudos directly, but also strengthens several other optimization
7479 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7480 however, make debugging impossible, since variables will no longer stay in a
7483 Enabled by default with @option{-funroll-loops}.
7485 @item -fwhole-program
7486 @opindex fwhole-program
7487 Assume that the current compilation unit represents the whole program being
7488 compiled. All public functions and variables with the exception of @code{main}
7489 and those merged by attribute @code{externally_visible} become static functions
7490 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.
7491 While this option is equivalent to proper use of the @code{static} keyword for
7492 programs consisting of a single file, in combination with option
7493 @option{-flto} this flag can be used to
7494 compile many smaller scale programs since the functions and variables become
7495 local for the whole combined compilation unit, not for the single source file
7498 This option implies @option{-fwhole-file} for Fortran programs.
7500 @item -flto[=@var{n}]
7502 This option runs the standard link-time optimizer. When invoked
7503 with source code, it generates GIMPLE (one of GCC's internal
7504 representations) and writes it to special ELF sections in the object
7505 file. When the object files are linked together, all the function
7506 bodies are read from these ELF sections and instantiated as if they
7507 had been part of the same translation unit.
7509 To use the link-timer optimizer, @option{-flto} needs to be specified at
7510 compile time and during the final link. For example,
7513 gcc -c -O2 -flto foo.c
7514 gcc -c -O2 -flto bar.c
7515 gcc -o myprog -flto -O2 foo.o bar.o
7518 The first two invocations to GCC will save a bytecode representation
7519 of GIMPLE into special ELF sections inside @file{foo.o} and
7520 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7521 @file{foo.o} and @file{bar.o}, merge the two files into a single
7522 internal image, and compile the result as usual. Since both
7523 @file{foo.o} and @file{bar.o} are merged into a single image, this
7524 causes all the inter-procedural analyses and optimizations in GCC to
7525 work across the two files as if they were a single one. This means,
7526 for example, that the inliner will be able to inline functions in
7527 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7529 Another (simpler) way to enable link-time optimization is,
7532 gcc -o myprog -flto -O2 foo.c bar.c
7535 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7536 merge them together into a single GIMPLE representation and optimize
7537 them as usual to produce @file{myprog}.
7539 The only important thing to keep in mind is that to enable link-time
7540 optimizations the @option{-flto} flag needs to be passed to both the
7541 compile and the link commands.
7543 To make whole program optimization effective, it is necessary to make
7544 certain whole program assumptions. The compiler needs to know
7545 what functions and variables can be accessed by libraries and runtime
7546 outside of the link time optimized unit. When supported by the linker,
7547 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7548 compiler information about used and externally visible symbols. When
7549 the linker plugin is not available, @option{-fwhole-program} should be
7550 used to allow the compiler to make these assumptions, which will lead
7551 to more aggressive optimization decisions.
7553 Note that when a file is compiled with @option{-flto}, the generated
7554 object file will be larger than a regular object file because it will
7555 contain GIMPLE bytecodes and the usual final code. This means that
7556 object files with LTO information can be linked as a normal object
7557 file. So, in the previous example, if the final link is done with
7560 gcc -o myprog foo.o bar.o
7563 The only difference will be that no inter-procedural optimizations
7564 will be applied to produce @file{myprog}. The two object files
7565 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7568 Additionally, the optimization flags used to compile individual files
7569 are not necessarily related to those used at link-time. For instance,
7572 gcc -c -O0 -flto foo.c
7573 gcc -c -O0 -flto bar.c
7574 gcc -o myprog -flto -O3 foo.o bar.o
7577 This will produce individual object files with unoptimized assembler
7578 code, but the resulting binary @file{myprog} will be optimized at
7579 @option{-O3}. Now, if the final binary is generated without
7580 @option{-flto}, then @file{myprog} will not be optimized.
7582 When producing the final binary with @option{-flto}, GCC will only
7583 apply link-time optimizations to those files that contain bytecode.
7584 Therefore, you can mix and match object files and libraries with
7585 GIMPLE bytecodes and final object code. GCC will automatically select
7586 which files to optimize in LTO mode and which files to link without
7589 There are some code generation flags that GCC will preserve when
7590 generating bytecodes, as they need to be used during the final link
7591 stage. Currently, the following options are saved into the GIMPLE
7592 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7593 @option{-m} target flags.
7595 At link time, these options are read-in and reapplied. Note that the
7596 current implementation makes no attempt at recognizing conflicting
7597 values for these options. If two or more files have a conflicting
7598 value (e.g., one file is compiled with @option{-fPIC} and another
7599 isn't), the compiler will simply use the last value read from the
7600 bytecode files. It is recommended, then, that all the files
7601 participating in the same link be compiled with the same options.
7603 Another feature of LTO is that it is possible to apply interprocedural
7604 optimizations on files written in different languages. This requires
7605 some support in the language front end. Currently, the C, C++ and
7606 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7607 something like this should work
7612 gfortran -c -flto baz.f90
7613 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7616 Notice that the final link is done with @command{g++} to get the C++
7617 runtime libraries and @option{-lgfortran} is added to get the Fortran
7618 runtime libraries. In general, when mixing languages in LTO mode, you
7619 should use the same link command used when mixing languages in a
7620 regular (non-LTO) compilation. This means that if your build process
7621 was mixing languages before, all you need to add is @option{-flto} to
7622 all the compile and link commands.
7624 If LTO encounters objects with C linkage declared with incompatible
7625 types in separate translation units to be linked together (undefined
7626 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7627 issued. The behavior is still undefined at runtime.
7629 If object files containing GIMPLE bytecode are stored in a library archive, say
7630 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7631 are using a linker with linker plugin support. To enable this feature, use
7632 the flag @option{-fuse-linker-plugin} at link-time:
7635 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7638 With the linker plugin enabled, the linker will extract the needed
7639 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7640 to make them part of the aggregated GIMPLE image to be optimized.
7642 If you are not using a linker with linker plugin support and/or do not
7643 enable linker plugin then the objects inside @file{libfoo.a}
7644 will be extracted and linked as usual, but they will not participate
7645 in the LTO optimization process.
7647 Link time optimizations do not require the presence of the whole program to
7648 operate. If the program does not require any symbols to be exported, it is
7649 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7650 the interprocedural optimizers to use more aggressive assumptions which may
7651 lead to improved optimization opportunities.
7652 Use of @option{-fwhole-program} is not needed when linker plugin is
7653 active (see @option{-fuse-linker-plugin}).
7655 Regarding portability: the current implementation of LTO makes no
7656 attempt at generating bytecode that can be ported between different
7657 types of hosts. The bytecode files are versioned and there is a
7658 strict version check, so bytecode files generated in one version of
7659 GCC will not work with an older/newer version of GCC.
7661 Link time optimization does not play well with generating debugging
7662 information. Combining @option{-flto} with
7663 @option{-g} is currently experimental and expected to produce wrong
7666 If you specify the optional @var{n}, the optimization and code
7667 generation done at link time is executed in parallel using @var{n}
7668 parallel jobs by utilizing an installed @command{make} program. The
7669 environment variable @env{MAKE} may be used to override the program
7670 used. The default value for @var{n} is 1.
7672 You can also specify @option{-flto=jobserver} to use GNU make's
7673 job server mode to determine the number of parallel jobs. This
7674 is useful when the Makefile calling GCC is already executing in parallel.
7675 The parent Makefile will need a @samp{+} prepended to the command recipe
7676 for this to work. This will likely only work if @env{MAKE} is
7679 This option is disabled by default.
7681 @item -flto-partition=@var{alg}
7682 @opindex flto-partition
7683 Specify the partitioning algorithm used by the link time optimizer.
7684 The value is either @code{1to1} to specify a partitioning mirroring
7685 the original source files or @code{balanced} to specify partitioning
7686 into equally sized chunks (whenever possible). Specifying @code{none}
7687 as an algorithm disables partitioning and streaming completely. The
7688 default value is @code{balanced}.
7690 @item -flto-compression-level=@var{n}
7691 This option specifies the level of compression used for intermediate
7692 language written to LTO object files, and is only meaningful in
7693 conjunction with LTO mode (@option{-flto}). Valid
7694 values are 0 (no compression) to 9 (maximum compression). Values
7695 outside this range are clamped to either 0 or 9. If the option is not
7696 given, a default balanced compression setting is used.
7699 Prints a report with internal details on the workings of the link-time
7700 optimizer. The contents of this report vary from version to version,
7701 it is meant to be useful to GCC developers when processing object
7702 files in LTO mode (via @option{-flto}).
7704 Disabled by default.
7706 @item -fuse-linker-plugin
7707 Enables the use of linker plugin during link time optimization. This option
7708 relies on the linker plugin support in linker that is available in gold
7709 or in GNU ld 2.21 or newer.
7711 This option enables the extraction of object files with GIMPLE bytecode out of
7712 library archives. This improves the quality of optimization by exposing more
7713 code the the link time optimizer. This information specify what symbols
7714 can be accessed externally (by non-LTO object or during dynamic linking).
7715 Resulting code quality improvements on binaries (and shared libraries that do
7716 use hidden visibility) is similar to @code{-fwhole-program}. See
7717 @option{-flto} for a description on the effect of this flag and how to use it.
7719 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7720 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7722 @item -fcompare-elim
7723 @opindex fcompare-elim
7724 After register allocation and post-register allocation instruction splitting,
7725 identify arithmetic instructions that compute processor flags similar to a
7726 comparison operation based on that arithmetic. If possible, eliminate the
7727 explicit comparison operation.
7729 This pass only applies to certain targets that cannot explicitly represent
7730 the comparison operation before register allocation is complete.
7732 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7734 @item -fcprop-registers
7735 @opindex fcprop-registers
7736 After register allocation and post-register allocation instruction splitting,
7737 we perform a copy-propagation pass to try to reduce scheduling dependencies
7738 and occasionally eliminate the copy.
7740 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7742 @item -fprofile-correction
7743 @opindex fprofile-correction
7744 Profiles collected using an instrumented binary for multi-threaded programs may
7745 be inconsistent due to missed counter updates. When this option is specified,
7746 GCC will use heuristics to correct or smooth out such inconsistencies. By
7747 default, GCC will emit an error message when an inconsistent profile is detected.
7749 @item -fprofile-dir=@var{path}
7750 @opindex fprofile-dir
7752 Set the directory to search for the profile data files in to @var{path}.
7753 This option affects only the profile data generated by
7754 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7755 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7756 and its related options.
7757 By default, GCC will use the current directory as @var{path}, thus the
7758 profile data file will appear in the same directory as the object file.
7760 @item -fprofile-generate
7761 @itemx -fprofile-generate=@var{path}
7762 @opindex fprofile-generate
7764 Enable options usually used for instrumenting application to produce
7765 profile useful for later recompilation with profile feedback based
7766 optimization. You must use @option{-fprofile-generate} both when
7767 compiling and when linking your program.
7769 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7771 If @var{path} is specified, GCC will look at the @var{path} to find
7772 the profile feedback data files. See @option{-fprofile-dir}.
7775 @itemx -fprofile-use=@var{path}
7776 @opindex fprofile-use
7777 Enable profile feedback directed optimizations, and optimizations
7778 generally profitable only with profile feedback available.
7780 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7781 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7783 By default, GCC emits an error message if the feedback profiles do not
7784 match the source code. This error can be turned into a warning by using
7785 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7788 If @var{path} is specified, GCC will look at the @var{path} to find
7789 the profile feedback data files. See @option{-fprofile-dir}.
7792 The following options control compiler behavior regarding floating
7793 point arithmetic. These options trade off between speed and
7794 correctness. All must be specifically enabled.
7798 @opindex ffloat-store
7799 Do not store floating point variables in registers, and inhibit other
7800 options that might change whether a floating point value is taken from a
7803 @cindex floating point precision
7804 This option prevents undesirable excess precision on machines such as
7805 the 68000 where the floating registers (of the 68881) keep more
7806 precision than a @code{double} is supposed to have. Similarly for the
7807 x86 architecture. For most programs, the excess precision does only
7808 good, but a few programs rely on the precise definition of IEEE floating
7809 point. Use @option{-ffloat-store} for such programs, after modifying
7810 them to store all pertinent intermediate computations into variables.
7812 @item -fexcess-precision=@var{style}
7813 @opindex fexcess-precision
7814 This option allows further control over excess precision on machines
7815 where floating-point registers have more precision than the IEEE
7816 @code{float} and @code{double} types and the processor does not
7817 support operations rounding to those types. By default,
7818 @option{-fexcess-precision=fast} is in effect; this means that
7819 operations are carried out in the precision of the registers and that
7820 it is unpredictable when rounding to the types specified in the source
7821 code takes place. When compiling C, if
7822 @option{-fexcess-precision=standard} is specified then excess
7823 precision will follow the rules specified in ISO C99; in particular,
7824 both casts and assignments cause values to be rounded to their
7825 semantic types (whereas @option{-ffloat-store} only affects
7826 assignments). This option is enabled by default for C if a strict
7827 conformance option such as @option{-std=c99} is used.
7830 @option{-fexcess-precision=standard} is not implemented for languages
7831 other than C, and has no effect if
7832 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7833 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7834 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7835 semantics apply without excess precision, and in the latter, rounding
7840 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7841 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7842 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7844 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7846 This option is not turned on by any @option{-O} option besides
7847 @option{-Ofast} since it can result in incorrect output for programs
7848 which depend on an exact implementation of IEEE or ISO rules/specifications
7849 for math functions. It may, however, yield faster code for programs
7850 that do not require the guarantees of these specifications.
7852 @item -fno-math-errno
7853 @opindex fno-math-errno
7854 Do not set ERRNO after calling math functions that are executed
7855 with a single instruction, e.g., sqrt. A program that relies on
7856 IEEE exceptions for math error handling may want to use this flag
7857 for speed while maintaining IEEE arithmetic compatibility.
7859 This option is not turned on by any @option{-O} option since
7860 it can result in incorrect output for programs which depend on
7861 an exact implementation of IEEE or ISO rules/specifications for
7862 math functions. It may, however, yield faster code for programs
7863 that do not require the guarantees of these specifications.
7865 The default is @option{-fmath-errno}.
7867 On Darwin systems, the math library never sets @code{errno}. There is
7868 therefore no reason for the compiler to consider the possibility that
7869 it might, and @option{-fno-math-errno} is the default.
7871 @item -funsafe-math-optimizations
7872 @opindex funsafe-math-optimizations
7874 Allow optimizations for floating-point arithmetic that (a) assume
7875 that arguments and results are valid and (b) may violate IEEE or
7876 ANSI standards. When used at link-time, it may include libraries
7877 or startup files that change the default FPU control word or other
7878 similar optimizations.
7880 This option is not turned on by any @option{-O} option since
7881 it can result in incorrect output for programs which depend on
7882 an exact implementation of IEEE or ISO rules/specifications for
7883 math functions. It may, however, yield faster code for programs
7884 that do not require the guarantees of these specifications.
7885 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7886 @option{-fassociative-math} and @option{-freciprocal-math}.
7888 The default is @option{-fno-unsafe-math-optimizations}.
7890 @item -fassociative-math
7891 @opindex fassociative-math
7893 Allow re-association of operands in series of floating-point operations.
7894 This violates the ISO C and C++ language standard by possibly changing
7895 computation result. NOTE: re-ordering may change the sign of zero as
7896 well as ignore NaNs and inhibit or create underflow or overflow (and
7897 thus cannot be used on a code which relies on rounding behavior like
7898 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7899 and thus may not be used when ordered comparisons are required.
7900 This option requires that both @option{-fno-signed-zeros} and
7901 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7902 much sense with @option{-frounding-math}. For Fortran the option
7903 is automatically enabled when both @option{-fno-signed-zeros} and
7904 @option{-fno-trapping-math} are in effect.
7906 The default is @option{-fno-associative-math}.
7908 @item -freciprocal-math
7909 @opindex freciprocal-math
7911 Allow the reciprocal of a value to be used instead of dividing by
7912 the value if this enables optimizations. For example @code{x / y}
7913 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7914 is subject to common subexpression elimination. Note that this loses
7915 precision and increases the number of flops operating on the value.
7917 The default is @option{-fno-reciprocal-math}.
7919 @item -ffinite-math-only
7920 @opindex ffinite-math-only
7921 Allow optimizations for floating-point arithmetic that assume
7922 that arguments and results are not NaNs or +-Infs.
7924 This option is not turned on by any @option{-O} option since
7925 it can result in incorrect output for programs which depend on
7926 an exact implementation of IEEE or ISO rules/specifications for
7927 math functions. It may, however, yield faster code for programs
7928 that do not require the guarantees of these specifications.
7930 The default is @option{-fno-finite-math-only}.
7932 @item -fno-signed-zeros
7933 @opindex fno-signed-zeros
7934 Allow optimizations for floating point arithmetic that ignore the
7935 signedness of zero. IEEE arithmetic specifies the behavior of
7936 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7937 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7938 This option implies that the sign of a zero result isn't significant.
7940 The default is @option{-fsigned-zeros}.
7942 @item -fno-trapping-math
7943 @opindex fno-trapping-math
7944 Compile code assuming that floating-point operations cannot generate
7945 user-visible traps. These traps include division by zero, overflow,
7946 underflow, inexact result and invalid operation. This option requires
7947 that @option{-fno-signaling-nans} be in effect. Setting this option may
7948 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7950 This option should never be turned on by any @option{-O} option since
7951 it can result in incorrect output for programs which depend on
7952 an exact implementation of IEEE or ISO rules/specifications for
7955 The default is @option{-ftrapping-math}.
7957 @item -frounding-math
7958 @opindex frounding-math
7959 Disable transformations and optimizations that assume default floating
7960 point rounding behavior. This is round-to-zero for all floating point
7961 to integer conversions, and round-to-nearest for all other arithmetic
7962 truncations. This option should be specified for programs that change
7963 the FP rounding mode dynamically, or that may be executed with a
7964 non-default rounding mode. This option disables constant folding of
7965 floating point expressions at compile-time (which may be affected by
7966 rounding mode) and arithmetic transformations that are unsafe in the
7967 presence of sign-dependent rounding modes.
7969 The default is @option{-fno-rounding-math}.
7971 This option is experimental and does not currently guarantee to
7972 disable all GCC optimizations that are affected by rounding mode.
7973 Future versions of GCC may provide finer control of this setting
7974 using C99's @code{FENV_ACCESS} pragma. This command line option
7975 will be used to specify the default state for @code{FENV_ACCESS}.
7977 @item -fsignaling-nans
7978 @opindex fsignaling-nans
7979 Compile code assuming that IEEE signaling NaNs may generate user-visible
7980 traps during floating-point operations. Setting this option disables
7981 optimizations that may change the number of exceptions visible with
7982 signaling NaNs. This option implies @option{-ftrapping-math}.
7984 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7987 The default is @option{-fno-signaling-nans}.
7989 This option is experimental and does not currently guarantee to
7990 disable all GCC optimizations that affect signaling NaN behavior.
7992 @item -fsingle-precision-constant
7993 @opindex fsingle-precision-constant
7994 Treat floating point constant as single precision constant instead of
7995 implicitly converting it to double precision constant.
7997 @item -fcx-limited-range
7998 @opindex fcx-limited-range
7999 When enabled, this option states that a range reduction step is not
8000 needed when performing complex division. Also, there is no checking
8001 whether the result of a complex multiplication or division is @code{NaN
8002 + I*NaN}, with an attempt to rescue the situation in that case. The
8003 default is @option{-fno-cx-limited-range}, but is enabled by
8004 @option{-ffast-math}.
8006 This option controls the default setting of the ISO C99
8007 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8010 @item -fcx-fortran-rules
8011 @opindex fcx-fortran-rules
8012 Complex multiplication and division follow Fortran rules. Range
8013 reduction is done as part of complex division, but there is no checking
8014 whether the result of a complex multiplication or division is @code{NaN
8015 + I*NaN}, with an attempt to rescue the situation in that case.
8017 The default is @option{-fno-cx-fortran-rules}.
8021 The following options control optimizations that may improve
8022 performance, but are not enabled by any @option{-O} options. This
8023 section includes experimental options that may produce broken code.
8026 @item -fbranch-probabilities
8027 @opindex fbranch-probabilities
8028 After running a program compiled with @option{-fprofile-arcs}
8029 (@pxref{Debugging Options,, Options for Debugging Your Program or
8030 @command{gcc}}), you can compile it a second time using
8031 @option{-fbranch-probabilities}, to improve optimizations based on
8032 the number of times each branch was taken. When the program
8033 compiled with @option{-fprofile-arcs} exits it saves arc execution
8034 counts to a file called @file{@var{sourcename}.gcda} for each source
8035 file. The information in this data file is very dependent on the
8036 structure of the generated code, so you must use the same source code
8037 and the same optimization options for both compilations.
8039 With @option{-fbranch-probabilities}, GCC puts a
8040 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8041 These can be used to improve optimization. Currently, they are only
8042 used in one place: in @file{reorg.c}, instead of guessing which path a
8043 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8044 exactly determine which path is taken more often.
8046 @item -fprofile-values
8047 @opindex fprofile-values
8048 If combined with @option{-fprofile-arcs}, it adds code so that some
8049 data about values of expressions in the program is gathered.
8051 With @option{-fbranch-probabilities}, it reads back the data gathered
8052 from profiling values of expressions for usage in optimizations.
8054 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8058 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8059 a code to gather information about values of expressions.
8061 With @option{-fbranch-probabilities}, it reads back the data gathered
8062 and actually performs the optimizations based on them.
8063 Currently the optimizations include specialization of division operation
8064 using the knowledge about the value of the denominator.
8066 @item -frename-registers
8067 @opindex frename-registers
8068 Attempt to avoid false dependencies in scheduled code by making use
8069 of registers left over after register allocation. This optimization
8070 will most benefit processors with lots of registers. Depending on the
8071 debug information format adopted by the target, however, it can
8072 make debugging impossible, since variables will no longer stay in
8073 a ``home register''.
8075 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8079 Perform tail duplication to enlarge superblock size. This transformation
8080 simplifies the control flow of the function allowing other optimizations to do
8083 Enabled with @option{-fprofile-use}.
8085 @item -funroll-loops
8086 @opindex funroll-loops
8087 Unroll loops whose number of iterations can be determined at compile time or
8088 upon entry to the loop. @option{-funroll-loops} implies
8089 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8090 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8091 small constant number of iterations). This option makes code larger, and may
8092 or may not make it run faster.
8094 Enabled with @option{-fprofile-use}.
8096 @item -funroll-all-loops
8097 @opindex funroll-all-loops
8098 Unroll all loops, even if their number of iterations is uncertain when
8099 the loop is entered. This usually makes programs run more slowly.
8100 @option{-funroll-all-loops} implies the same options as
8101 @option{-funroll-loops}.
8104 @opindex fpeel-loops
8105 Peels the loops for that there is enough information that they do not
8106 roll much (from profile feedback). It also turns on complete loop peeling
8107 (i.e.@: complete removal of loops with small constant number of iterations).
8109 Enabled with @option{-fprofile-use}.
8111 @item -fmove-loop-invariants
8112 @opindex fmove-loop-invariants
8113 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8114 at level @option{-O1}
8116 @item -funswitch-loops
8117 @opindex funswitch-loops
8118 Move branches with loop invariant conditions out of the loop, with duplicates
8119 of the loop on both branches (modified according to result of the condition).
8121 @item -ffunction-sections
8122 @itemx -fdata-sections
8123 @opindex ffunction-sections
8124 @opindex fdata-sections
8125 Place each function or data item into its own section in the output
8126 file if the target supports arbitrary sections. The name of the
8127 function or the name of the data item determines the section's name
8130 Use these options on systems where the linker can perform optimizations
8131 to improve locality of reference in the instruction space. Most systems
8132 using the ELF object format and SPARC processors running Solaris 2 have
8133 linkers with such optimizations. AIX may have these optimizations in
8136 Only use these options when there are significant benefits from doing
8137 so. When you specify these options, the assembler and linker will
8138 create larger object and executable files and will also be slower.
8139 You will not be able to use @code{gprof} on all systems if you
8140 specify this option and you may have problems with debugging if
8141 you specify both this option and @option{-g}.
8143 @item -fbranch-target-load-optimize
8144 @opindex fbranch-target-load-optimize
8145 Perform branch target register load optimization before prologue / epilogue
8147 The use of target registers can typically be exposed only during reload,
8148 thus hoisting loads out of loops and doing inter-block scheduling needs
8149 a separate optimization pass.
8151 @item -fbranch-target-load-optimize2
8152 @opindex fbranch-target-load-optimize2
8153 Perform branch target register load optimization after prologue / epilogue
8156 @item -fbtr-bb-exclusive
8157 @opindex fbtr-bb-exclusive
8158 When performing branch target register load optimization, don't reuse
8159 branch target registers in within any basic block.
8161 @item -fstack-protector
8162 @opindex fstack-protector
8163 Emit extra code to check for buffer overflows, such as stack smashing
8164 attacks. This is done by adding a guard variable to functions with
8165 vulnerable objects. This includes functions that call alloca, and
8166 functions with buffers larger than 8 bytes. The guards are initialized
8167 when a function is entered and then checked when the function exits.
8168 If a guard check fails, an error message is printed and the program exits.
8170 @item -fstack-protector-all
8171 @opindex fstack-protector-all
8172 Like @option{-fstack-protector} except that all functions are protected.
8174 @item -fsection-anchors
8175 @opindex fsection-anchors
8176 Try to reduce the number of symbolic address calculations by using
8177 shared ``anchor'' symbols to address nearby objects. This transformation
8178 can help to reduce the number of GOT entries and GOT accesses on some
8181 For example, the implementation of the following function @code{foo}:
8185 int foo (void) @{ return a + b + c; @}
8188 would usually calculate the addresses of all three variables, but if you
8189 compile it with @option{-fsection-anchors}, it will access the variables
8190 from a common anchor point instead. The effect is similar to the
8191 following pseudocode (which isn't valid C):
8196 register int *xr = &x;
8197 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8201 Not all targets support this option.
8203 @item --param @var{name}=@var{value}
8205 In some places, GCC uses various constants to control the amount of
8206 optimization that is done. For example, GCC will not inline functions
8207 that contain more that a certain number of instructions. You can
8208 control some of these constants on the command-line using the
8209 @option{--param} option.
8211 The names of specific parameters, and the meaning of the values, are
8212 tied to the internals of the compiler, and are subject to change
8213 without notice in future releases.
8215 In each case, the @var{value} is an integer. The allowable choices for
8216 @var{name} are given in the following table:
8219 @item struct-reorg-cold-struct-ratio
8220 The threshold ratio (as a percentage) between a structure frequency
8221 and the frequency of the hottest structure in the program. This parameter
8222 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
8223 We say that if the ratio of a structure frequency, calculated by profiling,
8224 to the hottest structure frequency in the program is less than this
8225 parameter, then structure reorganization is not applied to this structure.
8228 @item predictable-branch-outcome
8229 When branch is predicted to be taken with probability lower than this threshold
8230 (in percent), then it is considered well predictable. The default is 10.
8232 @item max-crossjump-edges
8233 The maximum number of incoming edges to consider for crossjumping.
8234 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8235 the number of edges incoming to each block. Increasing values mean
8236 more aggressive optimization, making the compile time increase with
8237 probably small improvement in executable size.
8239 @item min-crossjump-insns
8240 The minimum number of instructions which must be matched at the end
8241 of two blocks before crossjumping will be performed on them. This
8242 value is ignored in the case where all instructions in the block being
8243 crossjumped from are matched. The default value is 5.
8245 @item max-grow-copy-bb-insns
8246 The maximum code size expansion factor when copying basic blocks
8247 instead of jumping. The expansion is relative to a jump instruction.
8248 The default value is 8.
8250 @item max-goto-duplication-insns
8251 The maximum number of instructions to duplicate to a block that jumps
8252 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8253 passes, GCC factors computed gotos early in the compilation process,
8254 and unfactors them as late as possible. Only computed jumps at the
8255 end of a basic blocks with no more than max-goto-duplication-insns are
8256 unfactored. The default value is 8.
8258 @item max-delay-slot-insn-search
8259 The maximum number of instructions to consider when looking for an
8260 instruction to fill a delay slot. If more than this arbitrary number of
8261 instructions is searched, the time savings from filling the delay slot
8262 will be minimal so stop searching. Increasing values mean more
8263 aggressive optimization, making the compile time increase with probably
8264 small improvement in executable run time.
8266 @item max-delay-slot-live-search
8267 When trying to fill delay slots, the maximum number of instructions to
8268 consider when searching for a block with valid live register
8269 information. Increasing this arbitrarily chosen value means more
8270 aggressive optimization, increasing the compile time. This parameter
8271 should be removed when the delay slot code is rewritten to maintain the
8274 @item max-gcse-memory
8275 The approximate maximum amount of memory that will be allocated in
8276 order to perform the global common subexpression elimination
8277 optimization. If more memory than specified is required, the
8278 optimization will not be done.
8280 @item max-gcse-insertion-ratio
8281 If the ratio of expression insertions to deletions is larger than this value
8282 for any expression, then RTL PRE will insert or remove the expression and thus
8283 leave partially redundant computations in the instruction stream. The default value is 20.
8285 @item max-pending-list-length
8286 The maximum number of pending dependencies scheduling will allow
8287 before flushing the current state and starting over. Large functions
8288 with few branches or calls can create excessively large lists which
8289 needlessly consume memory and resources.
8291 @item max-inline-insns-single
8292 Several parameters control the tree inliner used in gcc.
8293 This number sets the maximum number of instructions (counted in GCC's
8294 internal representation) in a single function that the tree inliner
8295 will consider for inlining. This only affects functions declared
8296 inline and methods implemented in a class declaration (C++).
8297 The default value is 400.
8299 @item max-inline-insns-auto
8300 When you use @option{-finline-functions} (included in @option{-O3}),
8301 a lot of functions that would otherwise not be considered for inlining
8302 by the compiler will be investigated. To those functions, a different
8303 (more restrictive) limit compared to functions declared inline can
8305 The default value is 40.
8307 @item large-function-insns
8308 The limit specifying really large functions. For functions larger than this
8309 limit after inlining, inlining is constrained by
8310 @option{--param large-function-growth}. This parameter is useful primarily
8311 to avoid extreme compilation time caused by non-linear algorithms used by the
8313 The default value is 2700.
8315 @item large-function-growth
8316 Specifies maximal growth of large function caused by inlining in percents.
8317 The default value is 100 which limits large function growth to 2.0 times
8320 @item large-unit-insns
8321 The limit specifying large translation unit. Growth caused by inlining of
8322 units larger than this limit is limited by @option{--param inline-unit-growth}.
8323 For small units this might be too tight (consider unit consisting of function A
8324 that is inline and B that just calls A three time. If B is small relative to
8325 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8326 large units consisting of small inlineable functions however the overall unit
8327 growth limit is needed to avoid exponential explosion of code size. Thus for
8328 smaller units, the size is increased to @option{--param large-unit-insns}
8329 before applying @option{--param inline-unit-growth}. The default is 10000
8331 @item inline-unit-growth
8332 Specifies maximal overall growth of the compilation unit caused by inlining.
8333 The default value is 30 which limits unit growth to 1.3 times the original
8336 @item ipcp-unit-growth
8337 Specifies maximal overall growth of the compilation unit caused by
8338 interprocedural constant propagation. The default value is 10 which limits
8339 unit growth to 1.1 times the original size.
8341 @item large-stack-frame
8342 The limit specifying large stack frames. While inlining the algorithm is trying
8343 to not grow past this limit too much. Default value is 256 bytes.
8345 @item large-stack-frame-growth
8346 Specifies maximal growth of large stack frames caused by inlining in percents.
8347 The default value is 1000 which limits large stack frame growth to 11 times
8350 @item max-inline-insns-recursive
8351 @itemx max-inline-insns-recursive-auto
8352 Specifies maximum number of instructions out-of-line copy of self recursive inline
8353 function can grow into by performing recursive inlining.
8355 For functions declared inline @option{--param max-inline-insns-recursive} is
8356 taken into account. For function not declared inline, recursive inlining
8357 happens only when @option{-finline-functions} (included in @option{-O3}) is
8358 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8359 default value is 450.
8361 @item max-inline-recursive-depth
8362 @itemx max-inline-recursive-depth-auto
8363 Specifies maximum recursion depth used by the recursive inlining.
8365 For functions declared inline @option{--param max-inline-recursive-depth} is
8366 taken into account. For function not declared inline, recursive inlining
8367 happens only when @option{-finline-functions} (included in @option{-O3}) is
8368 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8371 @item min-inline-recursive-probability
8372 Recursive inlining is profitable only for function having deep recursion
8373 in average and can hurt for function having little recursion depth by
8374 increasing the prologue size or complexity of function body to other
8377 When profile feedback is available (see @option{-fprofile-generate}) the actual
8378 recursion depth can be guessed from probability that function will recurse via
8379 given call expression. This parameter limits inlining only to call expression
8380 whose probability exceeds given threshold (in percents). The default value is
8383 @item early-inlining-insns
8384 Specify growth that early inliner can make. In effect it increases amount of
8385 inlining for code having large abstraction penalty. The default value is 10.
8387 @item max-early-inliner-iterations
8388 @itemx max-early-inliner-iterations
8389 Limit of iterations of early inliner. This basically bounds number of nested
8390 indirect calls early inliner can resolve. Deeper chains are still handled by
8393 @item comdat-sharing-probability
8394 @itemx comdat-sharing-probability
8395 Probability (in percent) that C++ inline function with comdat visibility
8396 will be shared across multiple compilation units. The default value is 20.
8398 @item min-vect-loop-bound
8399 The minimum number of iterations under which a loop will not get vectorized
8400 when @option{-ftree-vectorize} is used. The number of iterations after
8401 vectorization needs to be greater than the value specified by this option
8402 to allow vectorization. The default value is 0.
8404 @item gcse-cost-distance-ratio
8405 Scaling factor in calculation of maximum distance an expression
8406 can be moved by GCSE optimizations. This is currently supported only in the
8407 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8408 will be with simple expressions, i.e., the expressions which have cost
8409 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8410 hoisting of simple expressions. The default value is 10.
8412 @item gcse-unrestricted-cost
8413 Cost, roughly measured as the cost of a single typical machine
8414 instruction, at which GCSE optimizations will not constrain
8415 the distance an expression can travel. This is currently
8416 supported only in the code hoisting pass. The lesser the cost,
8417 the more aggressive code hoisting will be. Specifying 0 will
8418 allow all expressions to travel unrestricted distances.
8419 The default value is 3.
8421 @item max-hoist-depth
8422 The depth of search in the dominator tree for expressions to hoist.
8423 This is used to avoid quadratic behavior in hoisting algorithm.
8424 The value of 0 will avoid limiting the search, but may slow down compilation
8425 of huge functions. The default value is 30.
8427 @item max-unrolled-insns
8428 The maximum number of instructions that a loop should have if that loop
8429 is unrolled, and if the loop is unrolled, it determines how many times
8430 the loop code is unrolled.
8432 @item max-average-unrolled-insns
8433 The maximum number of instructions biased by probabilities of their execution
8434 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8435 it determines how many times the loop code is unrolled.
8437 @item max-unroll-times
8438 The maximum number of unrollings of a single loop.
8440 @item max-peeled-insns
8441 The maximum number of instructions that a loop should have if that loop
8442 is peeled, and if the loop is peeled, it determines how many times
8443 the loop code is peeled.
8445 @item max-peel-times
8446 The maximum number of peelings of a single loop.
8448 @item max-completely-peeled-insns
8449 The maximum number of insns of a completely peeled loop.
8451 @item max-completely-peel-times
8452 The maximum number of iterations of a loop to be suitable for complete peeling.
8454 @item max-completely-peel-loop-nest-depth
8455 The maximum depth of a loop nest suitable for complete peeling.
8457 @item max-unswitch-insns
8458 The maximum number of insns of an unswitched loop.
8460 @item max-unswitch-level
8461 The maximum number of branches unswitched in a single loop.
8464 The minimum cost of an expensive expression in the loop invariant motion.
8466 @item iv-consider-all-candidates-bound
8467 Bound on number of candidates for induction variables below that
8468 all candidates are considered for each use in induction variable
8469 optimizations. Only the most relevant candidates are considered
8470 if there are more candidates, to avoid quadratic time complexity.
8472 @item iv-max-considered-uses
8473 The induction variable optimizations give up on loops that contain more
8474 induction variable uses.
8476 @item iv-always-prune-cand-set-bound
8477 If number of candidates in the set is smaller than this value,
8478 we always try to remove unnecessary ivs from the set during its
8479 optimization when a new iv is added to the set.
8481 @item scev-max-expr-size
8482 Bound on size of expressions used in the scalar evolutions analyzer.
8483 Large expressions slow the analyzer.
8485 @item scev-max-expr-complexity
8486 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8487 Complex expressions slow the analyzer.
8489 @item omega-max-vars
8490 The maximum number of variables in an Omega constraint system.
8491 The default value is 128.
8493 @item omega-max-geqs
8494 The maximum number of inequalities in an Omega constraint system.
8495 The default value is 256.
8498 The maximum number of equalities in an Omega constraint system.
8499 The default value is 128.
8501 @item omega-max-wild-cards
8502 The maximum number of wildcard variables that the Omega solver will
8503 be able to insert. The default value is 18.
8505 @item omega-hash-table-size
8506 The size of the hash table in the Omega solver. The default value is
8509 @item omega-max-keys
8510 The maximal number of keys used by the Omega solver. The default
8513 @item omega-eliminate-redundant-constraints
8514 When set to 1, use expensive methods to eliminate all redundant
8515 constraints. The default value is 0.
8517 @item vect-max-version-for-alignment-checks
8518 The maximum number of runtime checks that can be performed when
8519 doing loop versioning for alignment in the vectorizer. See option
8520 ftree-vect-loop-version for more information.
8522 @item vect-max-version-for-alias-checks
8523 The maximum number of runtime checks that can be performed when
8524 doing loop versioning for alias in the vectorizer. See option
8525 ftree-vect-loop-version for more information.
8527 @item max-iterations-to-track
8529 The maximum number of iterations of a loop the brute force algorithm
8530 for analysis of # of iterations of the loop tries to evaluate.
8532 @item hot-bb-count-fraction
8533 Select fraction of the maximal count of repetitions of basic block in program
8534 given basic block needs to have to be considered hot.
8536 @item hot-bb-frequency-fraction
8537 Select fraction of the entry block frequency of executions of basic block in
8538 function given basic block needs to have to be considered hot.
8540 @item max-predicted-iterations
8541 The maximum number of loop iterations we predict statically. This is useful
8542 in cases where function contain single loop with known bound and other loop
8543 with unknown. We predict the known number of iterations correctly, while
8544 the unknown number of iterations average to roughly 10. This means that the
8545 loop without bounds would appear artificially cold relative to the other one.
8547 @item align-threshold
8549 Select fraction of the maximal frequency of executions of basic block in
8550 function given basic block will get aligned.
8552 @item align-loop-iterations
8554 A loop expected to iterate at lest the selected number of iterations will get
8557 @item tracer-dynamic-coverage
8558 @itemx tracer-dynamic-coverage-feedback
8560 This value is used to limit superblock formation once the given percentage of
8561 executed instructions is covered. This limits unnecessary code size
8564 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8565 feedback is available. The real profiles (as opposed to statically estimated
8566 ones) are much less balanced allowing the threshold to be larger value.
8568 @item tracer-max-code-growth
8569 Stop tail duplication once code growth has reached given percentage. This is
8570 rather hokey argument, as most of the duplicates will be eliminated later in
8571 cross jumping, so it may be set to much higher values than is the desired code
8574 @item tracer-min-branch-ratio
8576 Stop reverse growth when the reverse probability of best edge is less than this
8577 threshold (in percent).
8579 @item tracer-min-branch-ratio
8580 @itemx tracer-min-branch-ratio-feedback
8582 Stop forward growth if the best edge do have probability lower than this
8585 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8586 compilation for profile feedback and one for compilation without. The value
8587 for compilation with profile feedback needs to be more conservative (higher) in
8588 order to make tracer effective.
8590 @item max-cse-path-length
8592 Maximum number of basic blocks on path that cse considers. The default is 10.
8595 The maximum instructions CSE process before flushing. The default is 1000.
8597 @item ggc-min-expand
8599 GCC uses a garbage collector to manage its own memory allocation. This
8600 parameter specifies the minimum percentage by which the garbage
8601 collector's heap should be allowed to expand between collections.
8602 Tuning this may improve compilation speed; it has no effect on code
8605 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8606 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8607 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8608 GCC is not able to calculate RAM on a particular platform, the lower
8609 bound of 30% is used. Setting this parameter and
8610 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8611 every opportunity. This is extremely slow, but can be useful for
8614 @item ggc-min-heapsize
8616 Minimum size of the garbage collector's heap before it begins bothering
8617 to collect garbage. The first collection occurs after the heap expands
8618 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8619 tuning this may improve compilation speed, and has no effect on code
8622 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8623 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8624 with a lower bound of 4096 (four megabytes) and an upper bound of
8625 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8626 particular platform, the lower bound is used. Setting this parameter
8627 very large effectively disables garbage collection. Setting this
8628 parameter and @option{ggc-min-expand} to zero causes a full collection
8629 to occur at every opportunity.
8631 @item max-reload-search-insns
8632 The maximum number of instruction reload should look backward for equivalent
8633 register. Increasing values mean more aggressive optimization, making the
8634 compile time increase with probably slightly better performance. The default
8637 @item max-cselib-memory-locations
8638 The maximum number of memory locations cselib should take into account.
8639 Increasing values mean more aggressive optimization, making the compile time
8640 increase with probably slightly better performance. The default value is 500.
8642 @item reorder-blocks-duplicate
8643 @itemx reorder-blocks-duplicate-feedback
8645 Used by basic block reordering pass to decide whether to use unconditional
8646 branch or duplicate the code on its destination. Code is duplicated when its
8647 estimated size is smaller than this value multiplied by the estimated size of
8648 unconditional jump in the hot spots of the program.
8650 The @option{reorder-block-duplicate-feedback} is used only when profile
8651 feedback is available and may be set to higher values than
8652 @option{reorder-block-duplicate} since information about the hot spots is more
8655 @item max-sched-ready-insns
8656 The maximum number of instructions ready to be issued the scheduler should
8657 consider at any given time during the first scheduling pass. Increasing
8658 values mean more thorough searches, making the compilation time increase
8659 with probably little benefit. The default value is 100.
8661 @item max-sched-region-blocks
8662 The maximum number of blocks in a region to be considered for
8663 interblock scheduling. The default value is 10.
8665 @item max-pipeline-region-blocks
8666 The maximum number of blocks in a region to be considered for
8667 pipelining in the selective scheduler. The default value is 15.
8669 @item max-sched-region-insns
8670 The maximum number of insns in a region to be considered for
8671 interblock scheduling. The default value is 100.
8673 @item max-pipeline-region-insns
8674 The maximum number of insns in a region to be considered for
8675 pipelining in the selective scheduler. The default value is 200.
8678 The minimum probability (in percents) of reaching a source block
8679 for interblock speculative scheduling. The default value is 40.
8681 @item max-sched-extend-regions-iters
8682 The maximum number of iterations through CFG to extend regions.
8683 0 - disable region extension,
8684 N - do at most N iterations.
8685 The default value is 0.
8687 @item max-sched-insn-conflict-delay
8688 The maximum conflict delay for an insn to be considered for speculative motion.
8689 The default value is 3.
8691 @item sched-spec-prob-cutoff
8692 The minimal probability of speculation success (in percents), so that
8693 speculative insn will be scheduled.
8694 The default value is 40.
8696 @item sched-mem-true-dep-cost
8697 Minimal distance (in CPU cycles) between store and load targeting same
8698 memory locations. The default value is 1.
8700 @item selsched-max-lookahead
8701 The maximum size of the lookahead window of selective scheduling. It is a
8702 depth of search for available instructions.
8703 The default value is 50.
8705 @item selsched-max-sched-times
8706 The maximum number of times that an instruction will be scheduled during
8707 selective scheduling. This is the limit on the number of iterations
8708 through which the instruction may be pipelined. The default value is 2.
8710 @item selsched-max-insns-to-rename
8711 The maximum number of best instructions in the ready list that are considered
8712 for renaming in the selective scheduler. The default value is 2.
8714 @item max-last-value-rtl
8715 The maximum size measured as number of RTLs that can be recorded in an expression
8716 in combiner for a pseudo register as last known value of that register. The default
8719 @item integer-share-limit
8720 Small integer constants can use a shared data structure, reducing the
8721 compiler's memory usage and increasing its speed. This sets the maximum
8722 value of a shared integer constant. The default value is 256.
8724 @item min-virtual-mappings
8725 Specifies the minimum number of virtual mappings in the incremental
8726 SSA updater that should be registered to trigger the virtual mappings
8727 heuristic defined by virtual-mappings-ratio. The default value is
8730 @item virtual-mappings-ratio
8731 If the number of virtual mappings is virtual-mappings-ratio bigger
8732 than the number of virtual symbols to be updated, then the incremental
8733 SSA updater switches to a full update for those symbols. The default
8736 @item ssp-buffer-size
8737 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8738 protection when @option{-fstack-protection} is used.
8740 @item max-jump-thread-duplication-stmts
8741 Maximum number of statements allowed in a block that needs to be
8742 duplicated when threading jumps.
8744 @item max-fields-for-field-sensitive
8745 Maximum number of fields in a structure we will treat in
8746 a field sensitive manner during pointer analysis. The default is zero
8747 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8749 @item prefetch-latency
8750 Estimate on average number of instructions that are executed before
8751 prefetch finishes. The distance we prefetch ahead is proportional
8752 to this constant. Increasing this number may also lead to less
8753 streams being prefetched (see @option{simultaneous-prefetches}).
8755 @item simultaneous-prefetches
8756 Maximum number of prefetches that can run at the same time.
8758 @item l1-cache-line-size
8759 The size of cache line in L1 cache, in bytes.
8762 The size of L1 cache, in kilobytes.
8765 The size of L2 cache, in kilobytes.
8767 @item min-insn-to-prefetch-ratio
8768 The minimum ratio between the number of instructions and the
8769 number of prefetches to enable prefetching in a loop.
8771 @item prefetch-min-insn-to-mem-ratio
8772 The minimum ratio between the number of instructions and the
8773 number of memory references to enable prefetching in a loop.
8775 @item use-canonical-types
8776 Whether the compiler should use the ``canonical'' type system. By
8777 default, this should always be 1, which uses a more efficient internal
8778 mechanism for comparing types in C++ and Objective-C++. However, if
8779 bugs in the canonical type system are causing compilation failures,
8780 set this value to 0 to disable canonical types.
8782 @item switch-conversion-max-branch-ratio
8783 Switch initialization conversion will refuse to create arrays that are
8784 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8785 branches in the switch.
8787 @item max-partial-antic-length
8788 Maximum length of the partial antic set computed during the tree
8789 partial redundancy elimination optimization (@option{-ftree-pre}) when
8790 optimizing at @option{-O3} and above. For some sorts of source code
8791 the enhanced partial redundancy elimination optimization can run away,
8792 consuming all of the memory available on the host machine. This
8793 parameter sets a limit on the length of the sets that are computed,
8794 which prevents the runaway behavior. Setting a value of 0 for
8795 this parameter will allow an unlimited set length.
8797 @item sccvn-max-scc-size
8798 Maximum size of a strongly connected component (SCC) during SCCVN
8799 processing. If this limit is hit, SCCVN processing for the whole
8800 function will not be done and optimizations depending on it will
8801 be disabled. The default maximum SCC size is 10000.
8803 @item ira-max-loops-num
8804 IRA uses a regional register allocation by default. If a function
8805 contains loops more than number given by the parameter, only at most
8806 given number of the most frequently executed loops will form regions
8807 for the regional register allocation. The default value of the
8810 @item ira-max-conflict-table-size
8811 Although IRA uses a sophisticated algorithm of compression conflict
8812 table, the table can be still big for huge functions. If the conflict
8813 table for a function could be more than size in MB given by the
8814 parameter, the conflict table is not built and faster, simpler, and
8815 lower quality register allocation algorithm will be used. The
8816 algorithm do not use pseudo-register conflicts. The default value of
8817 the parameter is 2000.
8819 @item ira-loop-reserved-regs
8820 IRA can be used to evaluate more accurate register pressure in loops
8821 for decision to move loop invariants (see @option{-O3}). The number
8822 of available registers reserved for some other purposes is described
8823 by this parameter. The default value of the parameter is 2 which is
8824 minimal number of registers needed for execution of typical
8825 instruction. This value is the best found from numerous experiments.
8827 @item loop-invariant-max-bbs-in-loop
8828 Loop invariant motion can be very expensive, both in compile time and
8829 in amount of needed compile time memory, with very large loops. Loops
8830 with more basic blocks than this parameter won't have loop invariant
8831 motion optimization performed on them. The default value of the
8832 parameter is 1000 for -O1 and 10000 for -O2 and above.
8834 @item max-vartrack-size
8835 Sets a maximum number of hash table slots to use during variable
8836 tracking dataflow analysis of any function. If this limit is exceeded
8837 with variable tracking at assignments enabled, analysis for that
8838 function is retried without it, after removing all debug insns from
8839 the function. If the limit is exceeded even without debug insns, var
8840 tracking analysis is completely disabled for the function. Setting
8841 the parameter to zero makes it unlimited.
8843 @item min-nondebug-insn-uid
8844 Use uids starting at this parameter for nondebug insns. The range below
8845 the parameter is reserved exclusively for debug insns created by
8846 @option{-fvar-tracking-assignments}, but debug insns may get
8847 (non-overlapping) uids above it if the reserved range is exhausted.
8849 @item ipa-sra-ptr-growth-factor
8850 IPA-SRA will replace a pointer to an aggregate with one or more new
8851 parameters only when their cumulative size is less or equal to
8852 @option{ipa-sra-ptr-growth-factor} times the size of the original
8855 @item graphite-max-nb-scop-params
8856 To avoid exponential effects in the Graphite loop transforms, the
8857 number of parameters in a Static Control Part (SCoP) is bounded. The
8858 default value is 10 parameters. A variable whose value is unknown at
8859 compile time and defined outside a SCoP is a parameter of the SCoP.
8861 @item graphite-max-bbs-per-function
8862 To avoid exponential effects in the detection of SCoPs, the size of
8863 the functions analyzed by Graphite is bounded. The default value is
8866 @item loop-block-tile-size
8867 Loop blocking or strip mining transforms, enabled with
8868 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8869 loop in the loop nest by a given number of iterations. The strip
8870 length can be changed using the @option{loop-block-tile-size}
8871 parameter. The default value is 51 iterations.
8873 @item devirt-type-list-size
8874 IPA-CP attempts to track all possible types passed to a function's
8875 parameter in order to perform devirtualization.
8876 @option{devirt-type-list-size} is the maximum number of types it
8877 stores per a single formal parameter of a function.
8879 @item lto-partitions
8880 Specify desired number of partitions produced during WHOPR compilation.
8881 The number of partitions should exceed the number of CPUs used for compilation.
8882 The default value is 32.
8884 @item lto-minpartition
8885 Size of minimal partition for WHOPR (in estimated instructions).
8886 This prevents expenses of splitting very small programs into too many
8889 @item cxx-max-namespaces-for-diagnostic-help
8890 The maximum number of namespaces to consult for suggestions when C++
8891 name lookup fails for an identifier. The default is 1000.
8896 @node Preprocessor Options
8897 @section Options Controlling the Preprocessor
8898 @cindex preprocessor options
8899 @cindex options, preprocessor
8901 These options control the C preprocessor, which is run on each C source
8902 file before actual compilation.
8904 If you use the @option{-E} option, nothing is done except preprocessing.
8905 Some of these options make sense only together with @option{-E} because
8906 they cause the preprocessor output to be unsuitable for actual
8910 @item -Wp,@var{option}
8912 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8913 and pass @var{option} directly through to the preprocessor. If
8914 @var{option} contains commas, it is split into multiple options at the
8915 commas. However, many options are modified, translated or interpreted
8916 by the compiler driver before being passed to the preprocessor, and
8917 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8918 interface is undocumented and subject to change, so whenever possible
8919 you should avoid using @option{-Wp} and let the driver handle the
8922 @item -Xpreprocessor @var{option}
8923 @opindex Xpreprocessor
8924 Pass @var{option} as an option to the preprocessor. You can use this to
8925 supply system-specific preprocessor options which GCC does not know how to
8928 If you want to pass an option that takes an argument, you must use
8929 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8932 @include cppopts.texi
8934 @node Assembler Options
8935 @section Passing Options to the Assembler
8937 @c prevent bad page break with this line
8938 You can pass options to the assembler.
8941 @item -Wa,@var{option}
8943 Pass @var{option} as an option to the assembler. If @var{option}
8944 contains commas, it is split into multiple options at the commas.
8946 @item -Xassembler @var{option}
8948 Pass @var{option} as an option to the assembler. You can use this to
8949 supply system-specific assembler options which GCC does not know how to
8952 If you want to pass an option that takes an argument, you must use
8953 @option{-Xassembler} twice, once for the option and once for the argument.
8958 @section Options for Linking
8959 @cindex link options
8960 @cindex options, linking
8962 These options come into play when the compiler links object files into
8963 an executable output file. They are meaningless if the compiler is
8964 not doing a link step.
8968 @item @var{object-file-name}
8969 A file name that does not end in a special recognized suffix is
8970 considered to name an object file or library. (Object files are
8971 distinguished from libraries by the linker according to the file
8972 contents.) If linking is done, these object files are used as input
8981 If any of these options is used, then the linker is not run, and
8982 object file names should not be used as arguments. @xref{Overall
8986 @item -l@var{library}
8987 @itemx -l @var{library}
8989 Search the library named @var{library} when linking. (The second
8990 alternative with the library as a separate argument is only for
8991 POSIX compliance and is not recommended.)
8993 It makes a difference where in the command you write this option; the
8994 linker searches and processes libraries and object files in the order they
8995 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8996 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8997 to functions in @samp{z}, those functions may not be loaded.
8999 The linker searches a standard list of directories for the library,
9000 which is actually a file named @file{lib@var{library}.a}. The linker
9001 then uses this file as if it had been specified precisely by name.
9003 The directories searched include several standard system directories
9004 plus any that you specify with @option{-L}.
9006 Normally the files found this way are library files---archive files
9007 whose members are object files. The linker handles an archive file by
9008 scanning through it for members which define symbols that have so far
9009 been referenced but not defined. But if the file that is found is an
9010 ordinary object file, it is linked in the usual fashion. The only
9011 difference between using an @option{-l} option and specifying a file name
9012 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9013 and searches several directories.
9017 You need this special case of the @option{-l} option in order to
9018 link an Objective-C or Objective-C++ program.
9021 @opindex nostartfiles
9022 Do not use the standard system startup files when linking.
9023 The standard system libraries are used normally, unless @option{-nostdlib}
9024 or @option{-nodefaultlibs} is used.
9026 @item -nodefaultlibs
9027 @opindex nodefaultlibs
9028 Do not use the standard system libraries when linking.
9029 Only the libraries you specify will be passed to the linker, options
9030 specifying linkage of the system libraries, such as @code{-static-libgcc}
9031 or @code{-shared-libgcc}, will be ignored.
9032 The standard startup files are used normally, unless @option{-nostartfiles}
9033 is used. The compiler may generate calls to @code{memcmp},
9034 @code{memset}, @code{memcpy} and @code{memmove}.
9035 These entries are usually resolved by entries in
9036 libc. These entry points should be supplied through some other
9037 mechanism when this option is specified.
9041 Do not use the standard system startup files or libraries when linking.
9042 No startup files and only the libraries you specify will be passed to
9043 the linker, options specifying linkage of the system libraries, such as
9044 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9045 The compiler may generate calls to @code{memcmp}, @code{memset},
9046 @code{memcpy} and @code{memmove}.
9047 These entries are usually resolved by entries in
9048 libc. These entry points should be supplied through some other
9049 mechanism when this option is specified.
9051 @cindex @option{-lgcc}, use with @option{-nostdlib}
9052 @cindex @option{-nostdlib} and unresolved references
9053 @cindex unresolved references and @option{-nostdlib}
9054 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9055 @cindex @option{-nodefaultlibs} and unresolved references
9056 @cindex unresolved references and @option{-nodefaultlibs}
9057 One of the standard libraries bypassed by @option{-nostdlib} and
9058 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9059 that GCC uses to overcome shortcomings of particular machines, or special
9060 needs for some languages.
9061 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9062 Collection (GCC) Internals},
9063 for more discussion of @file{libgcc.a}.)
9064 In most cases, you need @file{libgcc.a} even when you want to avoid
9065 other standard libraries. In other words, when you specify @option{-nostdlib}
9066 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9067 This ensures that you have no unresolved references to internal GCC
9068 library subroutines. (For example, @samp{__main}, used to ensure C++
9069 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9070 GNU Compiler Collection (GCC) Internals}.)
9074 Produce a position independent executable on targets which support it.
9075 For predictable results, you must also specify the same set of options
9076 that were used to generate code (@option{-fpie}, @option{-fPIE},
9077 or model suboptions) when you specify this option.
9081 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9082 that support it. This instructs the linker to add all symbols, not
9083 only used ones, to the dynamic symbol table. This option is needed
9084 for some uses of @code{dlopen} or to allow obtaining backtraces
9085 from within a program.
9089 Remove all symbol table and relocation information from the executable.
9093 On systems that support dynamic linking, this prevents linking with the shared
9094 libraries. On other systems, this option has no effect.
9098 Produce a shared object which can then be linked with other objects to
9099 form an executable. Not all systems support this option. For predictable
9100 results, you must also specify the same set of options that were used to
9101 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9102 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9103 needs to build supplementary stub code for constructors to work. On
9104 multi-libbed systems, @samp{gcc -shared} must select the correct support
9105 libraries to link against. Failing to supply the correct flags may lead
9106 to subtle defects. Supplying them in cases where they are not necessary
9109 @item -shared-libgcc
9110 @itemx -static-libgcc
9111 @opindex shared-libgcc
9112 @opindex static-libgcc
9113 On systems that provide @file{libgcc} as a shared library, these options
9114 force the use of either the shared or static version respectively.
9115 If no shared version of @file{libgcc} was built when the compiler was
9116 configured, these options have no effect.
9118 There are several situations in which an application should use the
9119 shared @file{libgcc} instead of the static version. The most common
9120 of these is when the application wishes to throw and catch exceptions
9121 across different shared libraries. In that case, each of the libraries
9122 as well as the application itself should use the shared @file{libgcc}.
9124 Therefore, the G++ and GCJ drivers automatically add
9125 @option{-shared-libgcc} whenever you build a shared library or a main
9126 executable, because C++ and Java programs typically use exceptions, so
9127 this is the right thing to do.
9129 If, instead, you use the GCC driver to create shared libraries, you may
9130 find that they will not always be linked with the shared @file{libgcc}.
9131 If GCC finds, at its configuration time, that you have a non-GNU linker
9132 or a GNU linker that does not support option @option{--eh-frame-hdr},
9133 it will link the shared version of @file{libgcc} into shared libraries
9134 by default. Otherwise, it will take advantage of the linker and optimize
9135 away the linking with the shared version of @file{libgcc}, linking with
9136 the static version of libgcc by default. This allows exceptions to
9137 propagate through such shared libraries, without incurring relocation
9138 costs at library load time.
9140 However, if a library or main executable is supposed to throw or catch
9141 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9142 for the languages used in the program, or using the option
9143 @option{-shared-libgcc}, such that it is linked with the shared
9146 @item -static-libstdc++
9147 When the @command{g++} program is used to link a C++ program, it will
9148 normally automatically link against @option{libstdc++}. If
9149 @file{libstdc++} is available as a shared library, and the
9150 @option{-static} option is not used, then this will link against the
9151 shared version of @file{libstdc++}. That is normally fine. However, it
9152 is sometimes useful to freeze the version of @file{libstdc++} used by
9153 the program without going all the way to a fully static link. The
9154 @option{-static-libstdc++} option directs the @command{g++} driver to
9155 link @file{libstdc++} statically, without necessarily linking other
9156 libraries statically.
9160 Bind references to global symbols when building a shared object. Warn
9161 about any unresolved references (unless overridden by the link editor
9162 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9165 @item -T @var{script}
9167 @cindex linker script
9168 Use @var{script} as the linker script. This option is supported by most
9169 systems using the GNU linker. On some targets, such as bare-board
9170 targets without an operating system, the @option{-T} option may be required
9171 when linking to avoid references to undefined symbols.
9173 @item -Xlinker @var{option}
9175 Pass @var{option} as an option to the linker. You can use this to
9176 supply system-specific linker options which GCC does not know how to
9179 If you want to pass an option that takes a separate argument, you must use
9180 @option{-Xlinker} twice, once for the option and once for the argument.
9181 For example, to pass @option{-assert definitions}, you must write
9182 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9183 @option{-Xlinker "-assert definitions"}, because this passes the entire
9184 string as a single argument, which is not what the linker expects.
9186 When using the GNU linker, it is usually more convenient to pass
9187 arguments to linker options using the @option{@var{option}=@var{value}}
9188 syntax than as separate arguments. For example, you can specify
9189 @samp{-Xlinker -Map=output.map} rather than
9190 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9191 this syntax for command-line options.
9193 @item -Wl,@var{option}
9195 Pass @var{option} as an option to the linker. If @var{option} contains
9196 commas, it is split into multiple options at the commas. You can use this
9197 syntax to pass an argument to the option.
9198 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9199 linker. When using the GNU linker, you can also get the same effect with
9200 @samp{-Wl,-Map=output.map}.
9202 @item -u @var{symbol}
9204 Pretend the symbol @var{symbol} is undefined, to force linking of
9205 library modules to define it. You can use @option{-u} multiple times with
9206 different symbols to force loading of additional library modules.
9209 @node Directory Options
9210 @section Options for Directory Search
9211 @cindex directory options
9212 @cindex options, directory search
9215 These options specify directories to search for header files, for
9216 libraries and for parts of the compiler:
9221 Add the directory @var{dir} to the head of the list of directories to be
9222 searched for header files. This can be used to override a system header
9223 file, substituting your own version, since these directories are
9224 searched before the system header file directories. However, you should
9225 not use this option to add directories that contain vendor-supplied
9226 system header files (use @option{-isystem} for that). If you use more than
9227 one @option{-I} option, the directories are scanned in left-to-right
9228 order; the standard system directories come after.
9230 If a standard system include directory, or a directory specified with
9231 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9232 option will be ignored. The directory will still be searched but as a
9233 system directory at its normal position in the system include chain.
9234 This is to ensure that GCC's procedure to fix buggy system headers and
9235 the ordering for the include_next directive are not inadvertently changed.
9236 If you really need to change the search order for system directories,
9237 use the @option{-nostdinc} and/or @option{-isystem} options.
9239 @item -iplugindir=@var{dir}
9240 Set the directory to search for plugins which are passed
9241 by @option{-fplugin=@var{name}} instead of
9242 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9243 to be used by the user, but only passed by the driver.
9245 @item -iquote@var{dir}
9247 Add the directory @var{dir} to the head of the list of directories to
9248 be searched for header files only for the case of @samp{#include
9249 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9250 otherwise just like @option{-I}.
9254 Add directory @var{dir} to the list of directories to be searched
9257 @item -B@var{prefix}
9259 This option specifies where to find the executables, libraries,
9260 include files, and data files of the compiler itself.
9262 The compiler driver program runs one or more of the subprograms
9263 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9264 @var{prefix} as a prefix for each program it tries to run, both with and
9265 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9267 For each subprogram to be run, the compiler driver first tries the
9268 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9269 was not specified, the driver tries two standard prefixes, which are
9270 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9271 those results in a file name that is found, the unmodified program
9272 name is searched for using the directories specified in your
9273 @env{PATH} environment variable.
9275 The compiler will check to see if the path provided by the @option{-B}
9276 refers to a directory, and if necessary it will add a directory
9277 separator character at the end of the path.
9279 @option{-B} prefixes that effectively specify directory names also apply
9280 to libraries in the linker, because the compiler translates these
9281 options into @option{-L} options for the linker. They also apply to
9282 includes files in the preprocessor, because the compiler translates these
9283 options into @option{-isystem} options for the preprocessor. In this case,
9284 the compiler appends @samp{include} to the prefix.
9286 The run-time support file @file{libgcc.a} can also be searched for using
9287 the @option{-B} prefix, if needed. If it is not found there, the two
9288 standard prefixes above are tried, and that is all. The file is left
9289 out of the link if it is not found by those means.
9291 Another way to specify a prefix much like the @option{-B} prefix is to use
9292 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9295 As a special kludge, if the path provided by @option{-B} is
9296 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9297 9, then it will be replaced by @file{[dir/]include}. This is to help
9298 with boot-strapping the compiler.
9300 @item -specs=@var{file}
9302 Process @var{file} after the compiler reads in the standard @file{specs}
9303 file, in order to override the defaults that the @file{gcc} driver
9304 program uses when determining what switches to pass to @file{cc1},
9305 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9306 @option{-specs=@var{file}} can be specified on the command line, and they
9307 are processed in order, from left to right.
9309 @item --sysroot=@var{dir}
9311 Use @var{dir} as the logical root directory for headers and libraries.
9312 For example, if the compiler would normally search for headers in
9313 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9314 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9316 If you use both this option and the @option{-isysroot} option, then
9317 the @option{--sysroot} option will apply to libraries, but the
9318 @option{-isysroot} option will apply to header files.
9320 The GNU linker (beginning with version 2.16) has the necessary support
9321 for this option. If your linker does not support this option, the
9322 header file aspect of @option{--sysroot} will still work, but the
9323 library aspect will not.
9327 This option has been deprecated. Please use @option{-iquote} instead for
9328 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9329 Any directories you specify with @option{-I} options before the @option{-I-}
9330 option are searched only for the case of @samp{#include "@var{file}"};
9331 they are not searched for @samp{#include <@var{file}>}.
9333 If additional directories are specified with @option{-I} options after
9334 the @option{-I-}, these directories are searched for all @samp{#include}
9335 directives. (Ordinarily @emph{all} @option{-I} directories are used
9338 In addition, the @option{-I-} option inhibits the use of the current
9339 directory (where the current input file came from) as the first search
9340 directory for @samp{#include "@var{file}"}. There is no way to
9341 override this effect of @option{-I-}. With @option{-I.} you can specify
9342 searching the directory which was current when the compiler was
9343 invoked. That is not exactly the same as what the preprocessor does
9344 by default, but it is often satisfactory.
9346 @option{-I-} does not inhibit the use of the standard system directories
9347 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9354 @section Specifying subprocesses and the switches to pass to them
9357 @command{gcc} is a driver program. It performs its job by invoking a
9358 sequence of other programs to do the work of compiling, assembling and
9359 linking. GCC interprets its command-line parameters and uses these to
9360 deduce which programs it should invoke, and which command-line options
9361 it ought to place on their command lines. This behavior is controlled
9362 by @dfn{spec strings}. In most cases there is one spec string for each
9363 program that GCC can invoke, but a few programs have multiple spec
9364 strings to control their behavior. The spec strings built into GCC can
9365 be overridden by using the @option{-specs=} command-line switch to specify
9368 @dfn{Spec files} are plaintext files that are used to construct spec
9369 strings. They consist of a sequence of directives separated by blank
9370 lines. The type of directive is determined by the first non-whitespace
9371 character on the line and it can be one of the following:
9374 @item %@var{command}
9375 Issues a @var{command} to the spec file processor. The commands that can
9379 @item %include <@var{file}>
9380 @cindex @code{%include}
9381 Search for @var{file} and insert its text at the current point in the
9384 @item %include_noerr <@var{file}>
9385 @cindex @code{%include_noerr}
9386 Just like @samp{%include}, but do not generate an error message if the include
9387 file cannot be found.
9389 @item %rename @var{old_name} @var{new_name}
9390 @cindex @code{%rename}
9391 Rename the spec string @var{old_name} to @var{new_name}.
9395 @item *[@var{spec_name}]:
9396 This tells the compiler to create, override or delete the named spec
9397 string. All lines after this directive up to the next directive or
9398 blank line are considered to be the text for the spec string. If this
9399 results in an empty string then the spec will be deleted. (Or, if the
9400 spec did not exist, then nothing will happened.) Otherwise, if the spec
9401 does not currently exist a new spec will be created. If the spec does
9402 exist then its contents will be overridden by the text of this
9403 directive, unless the first character of that text is the @samp{+}
9404 character, in which case the text will be appended to the spec.
9406 @item [@var{suffix}]:
9407 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9408 and up to the next directive or blank line are considered to make up the
9409 spec string for the indicated suffix. When the compiler encounters an
9410 input file with the named suffix, it will processes the spec string in
9411 order to work out how to compile that file. For example:
9418 This says that any input file whose name ends in @samp{.ZZ} should be
9419 passed to the program @samp{z-compile}, which should be invoked with the
9420 command-line switch @option{-input} and with the result of performing the
9421 @samp{%i} substitution. (See below.)
9423 As an alternative to providing a spec string, the text that follows a
9424 suffix directive can be one of the following:
9427 @item @@@var{language}
9428 This says that the suffix is an alias for a known @var{language}. This is
9429 similar to using the @option{-x} command-line switch to GCC to specify a
9430 language explicitly. For example:
9437 Says that .ZZ files are, in fact, C++ source files.
9440 This causes an error messages saying:
9443 @var{name} compiler not installed on this system.
9447 GCC already has an extensive list of suffixes built into it.
9448 This directive will add an entry to the end of the list of suffixes, but
9449 since the list is searched from the end backwards, it is effectively
9450 possible to override earlier entries using this technique.
9454 GCC has the following spec strings built into it. Spec files can
9455 override these strings or create their own. Note that individual
9456 targets can also add their own spec strings to this list.
9459 asm Options to pass to the assembler
9460 asm_final Options to pass to the assembler post-processor
9461 cpp Options to pass to the C preprocessor
9462 cc1 Options to pass to the C compiler
9463 cc1plus Options to pass to the C++ compiler
9464 endfile Object files to include at the end of the link
9465 link Options to pass to the linker
9466 lib Libraries to include on the command line to the linker
9467 libgcc Decides which GCC support library to pass to the linker
9468 linker Sets the name of the linker
9469 predefines Defines to be passed to the C preprocessor
9470 signed_char Defines to pass to CPP to say whether @code{char} is signed
9472 startfile Object files to include at the start of the link
9475 Here is a small example of a spec file:
9481 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9484 This example renames the spec called @samp{lib} to @samp{old_lib} and
9485 then overrides the previous definition of @samp{lib} with a new one.
9486 The new definition adds in some extra command-line options before
9487 including the text of the old definition.
9489 @dfn{Spec strings} are a list of command-line options to be passed to their
9490 corresponding program. In addition, the spec strings can contain
9491 @samp{%}-prefixed sequences to substitute variable text or to
9492 conditionally insert text into the command line. Using these constructs
9493 it is possible to generate quite complex command lines.
9495 Here is a table of all defined @samp{%}-sequences for spec
9496 strings. Note that spaces are not generated automatically around the
9497 results of expanding these sequences. Therefore you can concatenate them
9498 together or combine them with constant text in a single argument.
9502 Substitute one @samp{%} into the program name or argument.
9505 Substitute the name of the input file being processed.
9508 Substitute the basename of the input file being processed.
9509 This is the substring up to (and not including) the last period
9510 and not including the directory.
9513 This is the same as @samp{%b}, but include the file suffix (text after
9517 Marks the argument containing or following the @samp{%d} as a
9518 temporary file name, so that that file will be deleted if GCC exits
9519 successfully. Unlike @samp{%g}, this contributes no text to the
9522 @item %g@var{suffix}
9523 Substitute a file name that has suffix @var{suffix} and is chosen
9524 once per compilation, and mark the argument in the same way as
9525 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9526 name is now chosen in a way that is hard to predict even when previously
9527 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9528 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9529 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9530 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9531 was simply substituted with a file name chosen once per compilation,
9532 without regard to any appended suffix (which was therefore treated
9533 just like ordinary text), making such attacks more likely to succeed.
9535 @item %u@var{suffix}
9536 Like @samp{%g}, but generates a new temporary file name even if
9537 @samp{%u@var{suffix}} was already seen.
9539 @item %U@var{suffix}
9540 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9541 new one if there is no such last file name. In the absence of any
9542 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9543 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9544 would involve the generation of two distinct file names, one
9545 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9546 simply substituted with a file name chosen for the previous @samp{%u},
9547 without regard to any appended suffix.
9549 @item %j@var{suffix}
9550 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9551 writable, and if save-temps is off; otherwise, substitute the name
9552 of a temporary file, just like @samp{%u}. This temporary file is not
9553 meant for communication between processes, but rather as a junk
9556 @item %|@var{suffix}
9557 @itemx %m@var{suffix}
9558 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9559 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9560 all. These are the two most common ways to instruct a program that it
9561 should read from standard input or write to standard output. If you
9562 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9563 construct: see for example @file{f/lang-specs.h}.
9565 @item %.@var{SUFFIX}
9566 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9567 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9568 terminated by the next space or %.
9571 Marks the argument containing or following the @samp{%w} as the
9572 designated output file of this compilation. This puts the argument
9573 into the sequence of arguments that @samp{%o} will substitute later.
9576 Substitutes the names of all the output files, with spaces
9577 automatically placed around them. You should write spaces
9578 around the @samp{%o} as well or the results are undefined.
9579 @samp{%o} is for use in the specs for running the linker.
9580 Input files whose names have no recognized suffix are not compiled
9581 at all, but they are included among the output files, so they will
9585 Substitutes the suffix for object files. Note that this is
9586 handled specially when it immediately follows @samp{%g, %u, or %U},
9587 because of the need for those to form complete file names. The
9588 handling is such that @samp{%O} is treated exactly as if it had already
9589 been substituted, except that @samp{%g, %u, and %U} do not currently
9590 support additional @var{suffix} characters following @samp{%O} as they would
9591 following, for example, @samp{.o}.
9594 Substitutes the standard macro predefinitions for the
9595 current target machine. Use this when running @code{cpp}.
9598 Like @samp{%p}, but puts @samp{__} before and after the name of each
9599 predefined macro, except for macros that start with @samp{__} or with
9600 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9604 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9605 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9606 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9607 and @option{-imultilib} as necessary.
9610 Current argument is the name of a library or startup file of some sort.
9611 Search for that file in a standard list of directories and substitute
9612 the full name found. The current working directory is included in the
9613 list of directories scanned.
9616 Current argument is the name of a linker script. Search for that file
9617 in the current list of directories to scan for libraries. If the file
9618 is located insert a @option{--script} option into the command line
9619 followed by the full path name found. If the file is not found then
9620 generate an error message. Note: the current working directory is not
9624 Print @var{str} as an error message. @var{str} is terminated by a newline.
9625 Use this when inconsistent options are detected.
9628 Substitute the contents of spec string @var{name} at this point.
9631 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9633 @item %x@{@var{option}@}
9634 Accumulate an option for @samp{%X}.
9637 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9641 Output the accumulated assembler options specified by @option{-Wa}.
9644 Output the accumulated preprocessor options specified by @option{-Wp}.
9647 Process the @code{asm} spec. This is used to compute the
9648 switches to be passed to the assembler.
9651 Process the @code{asm_final} spec. This is a spec string for
9652 passing switches to an assembler post-processor, if such a program is
9656 Process the @code{link} spec. This is the spec for computing the
9657 command line passed to the linker. Typically it will make use of the
9658 @samp{%L %G %S %D and %E} sequences.
9661 Dump out a @option{-L} option for each directory that GCC believes might
9662 contain startup files. If the target supports multilibs then the
9663 current multilib directory will be prepended to each of these paths.
9666 Process the @code{lib} spec. This is a spec string for deciding which
9667 libraries should be included on the command line to the linker.
9670 Process the @code{libgcc} spec. This is a spec string for deciding
9671 which GCC support library should be included on the command line to the linker.
9674 Process the @code{startfile} spec. This is a spec for deciding which
9675 object files should be the first ones passed to the linker. Typically
9676 this might be a file named @file{crt0.o}.
9679 Process the @code{endfile} spec. This is a spec string that specifies
9680 the last object files that will be passed to the linker.
9683 Process the @code{cpp} spec. This is used to construct the arguments
9684 to be passed to the C preprocessor.
9687 Process the @code{cc1} spec. This is used to construct the options to be
9688 passed to the actual C compiler (@samp{cc1}).
9691 Process the @code{cc1plus} spec. This is used to construct the options to be
9692 passed to the actual C++ compiler (@samp{cc1plus}).
9695 Substitute the variable part of a matched option. See below.
9696 Note that each comma in the substituted string is replaced by
9700 Remove all occurrences of @code{-S} from the command line. Note---this
9701 command is position dependent. @samp{%} commands in the spec string
9702 before this one will see @code{-S}, @samp{%} commands in the spec string
9703 after this one will not.
9705 @item %:@var{function}(@var{args})
9706 Call the named function @var{function}, passing it @var{args}.
9707 @var{args} is first processed as a nested spec string, then split
9708 into an argument vector in the usual fashion. The function returns
9709 a string which is processed as if it had appeared literally as part
9710 of the current spec.
9712 The following built-in spec functions are provided:
9716 The @code{getenv} spec function takes two arguments: an environment
9717 variable name and a string. If the environment variable is not
9718 defined, a fatal error is issued. Otherwise, the return value is the
9719 value of the environment variable concatenated with the string. For
9720 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9723 %:getenv(TOPDIR /include)
9726 expands to @file{/path/to/top/include}.
9728 @item @code{if-exists}
9729 The @code{if-exists} spec function takes one argument, an absolute
9730 pathname to a file. If the file exists, @code{if-exists} returns the
9731 pathname. Here is a small example of its usage:
9735 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9738 @item @code{if-exists-else}
9739 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9740 spec function, except that it takes two arguments. The first argument is
9741 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9742 returns the pathname. If it does not exist, it returns the second argument.
9743 This way, @code{if-exists-else} can be used to select one file or another,
9744 based on the existence of the first. Here is a small example of its usage:
9748 crt0%O%s %:if-exists(crti%O%s) \
9749 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9752 @item @code{replace-outfile}
9753 The @code{replace-outfile} spec function takes two arguments. It looks for the
9754 first argument in the outfiles array and replaces it with the second argument. Here
9755 is a small example of its usage:
9758 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9761 @item @code{remove-outfile}
9762 The @code{remove-outfile} spec function takes one argument. It looks for the
9763 first argument in the outfiles array and removes it. Here is a small example
9767 %:remove-outfile(-lm)
9770 @item @code{pass-through-libs}
9771 The @code{pass-through-libs} spec function takes any number of arguments. It
9772 finds any @option{-l} options and any non-options ending in ".a" (which it
9773 assumes are the names of linker input library archive files) and returns a
9774 result containing all the found arguments each prepended by
9775 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9776 intended to be passed to the LTO linker plugin.
9779 %:pass-through-libs(%G %L %G)
9782 @item @code{print-asm-header}
9783 The @code{print-asm-header} function takes no arguments and simply
9784 prints a banner like:
9790 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9793 It is used to separate compiler options from assembler options
9794 in the @option{--target-help} output.
9798 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9799 If that switch was not specified, this substitutes nothing. Note that
9800 the leading dash is omitted when specifying this option, and it is
9801 automatically inserted if the substitution is performed. Thus the spec
9802 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9803 and would output the command line option @option{-foo}.
9805 @item %W@{@code{S}@}
9806 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9809 @item %@{@code{S}*@}
9810 Substitutes all the switches specified to GCC whose names start
9811 with @code{-S}, but which also take an argument. This is used for
9812 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9813 GCC considers @option{-o foo} as being
9814 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9815 text, including the space. Thus two arguments would be generated.
9817 @item %@{@code{S}*&@code{T}*@}
9818 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9819 (the order of @code{S} and @code{T} in the spec is not significant).
9820 There can be any number of ampersand-separated variables; for each the
9821 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9823 @item %@{@code{S}:@code{X}@}
9824 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9826 @item %@{!@code{S}:@code{X}@}
9827 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9829 @item %@{@code{S}*:@code{X}@}
9830 Substitutes @code{X} if one or more switches whose names start with
9831 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9832 once, no matter how many such switches appeared. However, if @code{%*}
9833 appears somewhere in @code{X}, then @code{X} will be substituted once
9834 for each matching switch, with the @code{%*} replaced by the part of
9835 that switch that matched the @code{*}.
9837 @item %@{.@code{S}:@code{X}@}
9838 Substitutes @code{X}, if processing a file with suffix @code{S}.
9840 @item %@{!.@code{S}:@code{X}@}
9841 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9843 @item %@{,@code{S}:@code{X}@}
9844 Substitutes @code{X}, if processing a file for language @code{S}.
9846 @item %@{!,@code{S}:@code{X}@}
9847 Substitutes @code{X}, if not processing a file for language @code{S}.
9849 @item %@{@code{S}|@code{P}:@code{X}@}
9850 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9851 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9852 @code{*} sequences as well, although they have a stronger binding than
9853 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9854 alternatives must be starred, and only the first matching alternative
9857 For example, a spec string like this:
9860 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9863 will output the following command-line options from the following input
9864 command-line options:
9869 -d fred.c -foo -baz -boggle
9870 -d jim.d -bar -baz -boggle
9873 @item %@{S:X; T:Y; :D@}
9875 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9876 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9877 be as many clauses as you need. This may be combined with @code{.},
9878 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9883 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9884 construct may contain other nested @samp{%} constructs or spaces, or
9885 even newlines. They are processed as usual, as described above.
9886 Trailing white space in @code{X} is ignored. White space may also
9887 appear anywhere on the left side of the colon in these constructs,
9888 except between @code{.} or @code{*} and the corresponding word.
9890 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9891 handled specifically in these constructs. If another value of
9892 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9893 @option{-W} switch is found later in the command line, the earlier
9894 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9895 just one letter, which passes all matching options.
9897 The character @samp{|} at the beginning of the predicate text is used to
9898 indicate that a command should be piped to the following command, but
9899 only if @option{-pipe} is specified.
9901 It is built into GCC which switches take arguments and which do not.
9902 (You might think it would be useful to generalize this to allow each
9903 compiler's spec to say which switches take arguments. But this cannot
9904 be done in a consistent fashion. GCC cannot even decide which input
9905 files have been specified without knowing which switches take arguments,
9906 and it must know which input files to compile in order to tell which
9909 GCC also knows implicitly that arguments starting in @option{-l} are to be
9910 treated as compiler output files, and passed to the linker in their
9911 proper position among the other output files.
9913 @c man begin OPTIONS
9915 @node Target Options
9916 @section Specifying Target Machine and Compiler Version
9917 @cindex target options
9918 @cindex cross compiling
9919 @cindex specifying machine version
9920 @cindex specifying compiler version and target machine
9921 @cindex compiler version, specifying
9922 @cindex target machine, specifying
9924 The usual way to run GCC is to run the executable called @command{gcc}, or
9925 @command{@var{machine}-gcc} when cross-compiling, or
9926 @command{@var{machine}-gcc-@var{version}} to run a version other than the
9927 one that was installed last.
9929 @node Submodel Options
9930 @section Hardware Models and Configurations
9931 @cindex submodel options
9932 @cindex specifying hardware config
9933 @cindex hardware models and configurations, specifying
9934 @cindex machine dependent options
9936 Each target machine types can have its own
9937 special options, starting with @samp{-m}, to choose among various
9938 hardware models or configurations---for example, 68010 vs 68020,
9939 floating coprocessor or none. A single installed version of the
9940 compiler can compile for any model or configuration, according to the
9943 Some configurations of the compiler also support additional special
9944 options, usually for compatibility with other compilers on the same
9947 @c This list is ordered alphanumerically by subsection name.
9948 @c It should be the same order and spelling as these options are listed
9949 @c in Machine Dependent Options
9954 * Blackfin Options::
9957 * DEC Alpha Options::
9958 * DEC Alpha/VMS Options::
9961 * GNU/Linux Options::
9964 * i386 and x86-64 Options::
9965 * i386 and x86-64 Windows Options::
9967 * IA-64/VMS Options::
9974 * MicroBlaze Options::
9979 * picoChip Options::
9981 * RS/6000 and PowerPC Options::
9983 * S/390 and zSeries Options::
9986 * Solaris 2 Options::
9989 * System V Options::
9994 * Xstormy16 Options::
10000 @subsection ARM Options
10001 @cindex ARM options
10003 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10007 @item -mabi=@var{name}
10009 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10010 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10013 @opindex mapcs-frame
10014 Generate a stack frame that is compliant with the ARM Procedure Call
10015 Standard for all functions, even if this is not strictly necessary for
10016 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10017 with this option will cause the stack frames not to be generated for
10018 leaf functions. The default is @option{-mno-apcs-frame}.
10022 This is a synonym for @option{-mapcs-frame}.
10025 @c not currently implemented
10026 @item -mapcs-stack-check
10027 @opindex mapcs-stack-check
10028 Generate code to check the amount of stack space available upon entry to
10029 every function (that actually uses some stack space). If there is
10030 insufficient space available then either the function
10031 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10032 called, depending upon the amount of stack space required. The run time
10033 system is required to provide these functions. The default is
10034 @option{-mno-apcs-stack-check}, since this produces smaller code.
10036 @c not currently implemented
10038 @opindex mapcs-float
10039 Pass floating point arguments using the float point registers. This is
10040 one of the variants of the APCS@. This option is recommended if the
10041 target hardware has a floating point unit or if a lot of floating point
10042 arithmetic is going to be performed by the code. The default is
10043 @option{-mno-apcs-float}, since integer only code is slightly increased in
10044 size if @option{-mapcs-float} is used.
10046 @c not currently implemented
10047 @item -mapcs-reentrant
10048 @opindex mapcs-reentrant
10049 Generate reentrant, position independent code. The default is
10050 @option{-mno-apcs-reentrant}.
10053 @item -mthumb-interwork
10054 @opindex mthumb-interwork
10055 Generate code which supports calling between the ARM and Thumb
10056 instruction sets. Without this option the two instruction sets cannot
10057 be reliably used inside one program. The default is
10058 @option{-mno-thumb-interwork}, since slightly larger code is generated
10059 when @option{-mthumb-interwork} is specified.
10061 @item -mno-sched-prolog
10062 @opindex mno-sched-prolog
10063 Prevent the reordering of instructions in the function prolog, or the
10064 merging of those instruction with the instructions in the function's
10065 body. This means that all functions will start with a recognizable set
10066 of instructions (or in fact one of a choice from a small set of
10067 different function prologues), and this information can be used to
10068 locate the start if functions inside an executable piece of code. The
10069 default is @option{-msched-prolog}.
10071 @item -mfloat-abi=@var{name}
10072 @opindex mfloat-abi
10073 Specifies which floating-point ABI to use. Permissible values
10074 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10076 Specifying @samp{soft} causes GCC to generate output containing
10077 library calls for floating-point operations.
10078 @samp{softfp} allows the generation of code using hardware floating-point
10079 instructions, but still uses the soft-float calling conventions.
10080 @samp{hard} allows generation of floating-point instructions
10081 and uses FPU-specific calling conventions.
10083 The default depends on the specific target configuration. Note that
10084 the hard-float and soft-float ABIs are not link-compatible; you must
10085 compile your entire program with the same ABI, and link with a
10086 compatible set of libraries.
10089 @opindex mhard-float
10090 Equivalent to @option{-mfloat-abi=hard}.
10093 @opindex msoft-float
10094 Equivalent to @option{-mfloat-abi=soft}.
10096 @item -mlittle-endian
10097 @opindex mlittle-endian
10098 Generate code for a processor running in little-endian mode. This is
10099 the default for all standard configurations.
10102 @opindex mbig-endian
10103 Generate code for a processor running in big-endian mode; the default is
10104 to compile code for a little-endian processor.
10106 @item -mwords-little-endian
10107 @opindex mwords-little-endian
10108 This option only applies when generating code for big-endian processors.
10109 Generate code for a little-endian word order but a big-endian byte
10110 order. That is, a byte order of the form @samp{32107654}. Note: this
10111 option should only be used if you require compatibility with code for
10112 big-endian ARM processors generated by versions of the compiler prior to
10115 @item -mcpu=@var{name}
10117 This specifies the name of the target ARM processor. GCC uses this name
10118 to determine what kind of instructions it can emit when generating
10119 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10120 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10121 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10122 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10123 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10125 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10126 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10127 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10128 @samp{strongarm1110},
10129 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10130 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10131 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10132 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10133 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10134 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10135 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10136 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10137 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
10140 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10142 @item -mtune=@var{name}
10144 This option is very similar to the @option{-mcpu=} option, except that
10145 instead of specifying the actual target processor type, and hence
10146 restricting which instructions can be used, it specifies that GCC should
10147 tune the performance of the code as if the target were of the type
10148 specified in this option, but still choosing the instructions that it
10149 will generate based on the CPU specified by a @option{-mcpu=} option.
10150 For some ARM implementations better performance can be obtained by using
10153 @item -march=@var{name}
10155 This specifies the name of the target ARM architecture. GCC uses this
10156 name to determine what kind of instructions it can emit when generating
10157 assembly code. This option can be used in conjunction with or instead
10158 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10159 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10160 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10161 @samp{armv6}, @samp{armv6j},
10162 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10163 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10164 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10166 @item -mfpu=@var{name}
10167 @itemx -mfpe=@var{number}
10168 @itemx -mfp=@var{number}
10172 This specifies what floating point hardware (or hardware emulation) is
10173 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10174 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10175 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10176 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10177 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10178 @option{-mfp} and @option{-mfpe} are synonyms for
10179 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10182 If @option{-msoft-float} is specified this specifies the format of
10183 floating point values.
10185 If the selected floating-point hardware includes the NEON extension
10186 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10187 operations will not be used by GCC's auto-vectorization pass unless
10188 @option{-funsafe-math-optimizations} is also specified. This is
10189 because NEON hardware does not fully implement the IEEE 754 standard for
10190 floating-point arithmetic (in particular denormal values are treated as
10191 zero), so the use of NEON instructions may lead to a loss of precision.
10193 @item -mfp16-format=@var{name}
10194 @opindex mfp16-format
10195 Specify the format of the @code{__fp16} half-precision floating-point type.
10196 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10197 the default is @samp{none}, in which case the @code{__fp16} type is not
10198 defined. @xref{Half-Precision}, for more information.
10200 @item -mstructure-size-boundary=@var{n}
10201 @opindex mstructure-size-boundary
10202 The size of all structures and unions will be rounded up to a multiple
10203 of the number of bits set by this option. Permissible values are 8, 32
10204 and 64. The default value varies for different toolchains. For the COFF
10205 targeted toolchain the default value is 8. A value of 64 is only allowed
10206 if the underlying ABI supports it.
10208 Specifying the larger number can produce faster, more efficient code, but
10209 can also increase the size of the program. Different values are potentially
10210 incompatible. Code compiled with one value cannot necessarily expect to
10211 work with code or libraries compiled with another value, if they exchange
10212 information using structures or unions.
10214 @item -mabort-on-noreturn
10215 @opindex mabort-on-noreturn
10216 Generate a call to the function @code{abort} at the end of a
10217 @code{noreturn} function. It will be executed if the function tries to
10221 @itemx -mno-long-calls
10222 @opindex mlong-calls
10223 @opindex mno-long-calls
10224 Tells the compiler to perform function calls by first loading the
10225 address of the function into a register and then performing a subroutine
10226 call on this register. This switch is needed if the target function
10227 will lie outside of the 64 megabyte addressing range of the offset based
10228 version of subroutine call instruction.
10230 Even if this switch is enabled, not all function calls will be turned
10231 into long calls. The heuristic is that static functions, functions
10232 which have the @samp{short-call} attribute, functions that are inside
10233 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10234 definitions have already been compiled within the current compilation
10235 unit, will not be turned into long calls. The exception to this rule is
10236 that weak function definitions, functions with the @samp{long-call}
10237 attribute or the @samp{section} attribute, and functions that are within
10238 the scope of a @samp{#pragma long_calls} directive, will always be
10239 turned into long calls.
10241 This feature is not enabled by default. Specifying
10242 @option{-mno-long-calls} will restore the default behavior, as will
10243 placing the function calls within the scope of a @samp{#pragma
10244 long_calls_off} directive. Note these switches have no effect on how
10245 the compiler generates code to handle function calls via function
10248 @item -msingle-pic-base
10249 @opindex msingle-pic-base
10250 Treat the register used for PIC addressing as read-only, rather than
10251 loading it in the prologue for each function. The run-time system is
10252 responsible for initializing this register with an appropriate value
10253 before execution begins.
10255 @item -mpic-register=@var{reg}
10256 @opindex mpic-register
10257 Specify the register to be used for PIC addressing. The default is R10
10258 unless stack-checking is enabled, when R9 is used.
10260 @item -mcirrus-fix-invalid-insns
10261 @opindex mcirrus-fix-invalid-insns
10262 @opindex mno-cirrus-fix-invalid-insns
10263 Insert NOPs into the instruction stream to in order to work around
10264 problems with invalid Maverick instruction combinations. This option
10265 is only valid if the @option{-mcpu=ep9312} option has been used to
10266 enable generation of instructions for the Cirrus Maverick floating
10267 point co-processor. This option is not enabled by default, since the
10268 problem is only present in older Maverick implementations. The default
10269 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10272 @item -mpoke-function-name
10273 @opindex mpoke-function-name
10274 Write the name of each function into the text section, directly
10275 preceding the function prologue. The generated code is similar to this:
10279 .ascii "arm_poke_function_name", 0
10282 .word 0xff000000 + (t1 - t0)
10283 arm_poke_function_name
10285 stmfd sp!, @{fp, ip, lr, pc@}
10289 When performing a stack backtrace, code can inspect the value of
10290 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10291 location @code{pc - 12} and the top 8 bits are set, then we know that
10292 there is a function name embedded immediately preceding this location
10293 and has length @code{((pc[-3]) & 0xff000000)}.
10297 Generate code for the Thumb instruction set. The default is to
10298 use the 32-bit ARM instruction set.
10299 This option automatically enables either 16-bit Thumb-1 or
10300 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10301 and @option{-march=@var{name}} options. This option is not passed to the
10302 assembler. If you want to force assembler files to be interpreted as Thumb code,
10303 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10304 option directly to the assembler by prefixing it with @option{-Wa}.
10307 @opindex mtpcs-frame
10308 Generate a stack frame that is compliant with the Thumb Procedure Call
10309 Standard for all non-leaf functions. (A leaf function is one that does
10310 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10312 @item -mtpcs-leaf-frame
10313 @opindex mtpcs-leaf-frame
10314 Generate a stack frame that is compliant with the Thumb Procedure Call
10315 Standard for all leaf functions. (A leaf function is one that does
10316 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10318 @item -mcallee-super-interworking
10319 @opindex mcallee-super-interworking
10320 Gives all externally visible functions in the file being compiled an ARM
10321 instruction set header which switches to Thumb mode before executing the
10322 rest of the function. This allows these functions to be called from
10323 non-interworking code. This option is not valid in AAPCS configurations
10324 because interworking is enabled by default.
10326 @item -mcaller-super-interworking
10327 @opindex mcaller-super-interworking
10328 Allows calls via function pointers (including virtual functions) to
10329 execute correctly regardless of whether the target code has been
10330 compiled for interworking or not. There is a small overhead in the cost
10331 of executing a function pointer if this option is enabled. This option
10332 is not valid in AAPCS configurations because interworking is enabled
10335 @item -mtp=@var{name}
10337 Specify the access model for the thread local storage pointer. The valid
10338 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10339 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10340 (supported in the arm6k architecture), and @option{auto}, which uses the
10341 best available method for the selected processor. The default setting is
10344 @item -mword-relocations
10345 @opindex mword-relocations
10346 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10347 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10348 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10351 @item -mfix-cortex-m3-ldrd
10352 @opindex mfix-cortex-m3-ldrd
10353 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10354 with overlapping destination and base registers are used. This option avoids
10355 generating these instructions. This option is enabled by default when
10356 @option{-mcpu=cortex-m3} is specified.
10361 @subsection AVR Options
10362 @cindex AVR Options
10364 These options are defined for AVR implementations:
10367 @item -mmcu=@var{mcu}
10369 Specify ATMEL AVR instruction set or MCU type.
10371 Instruction set avr1 is for the minimal AVR core, not supported by the C
10372 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10373 attiny11, attiny12, attiny15, attiny28).
10375 Instruction set avr2 (default) is for the classic AVR core with up to
10376 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10377 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10378 at90c8534, at90s8535).
10380 Instruction set avr3 is for the classic AVR core with up to 128K program
10381 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10383 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10384 memory space (MCU types: atmega8, atmega83, atmega85).
10386 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10387 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10388 atmega64, atmega128, at43usb355, at94k).
10390 @item -mno-interrupts
10391 @opindex mno-interrupts
10392 Generated code is not compatible with hardware interrupts.
10393 Code size will be smaller.
10395 @item -mcall-prologues
10396 @opindex mcall-prologues
10397 Functions prologues/epilogues expanded as call to appropriate
10398 subroutines. Code size will be smaller.
10401 @opindex mtiny-stack
10402 Change only the low 8 bits of the stack pointer.
10406 Assume int to be 8 bit integer. This affects the sizes of all types: A
10407 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10408 and long long will be 4 bytes. Please note that this option does not
10409 comply to the C standards, but it will provide you with smaller code
10413 @node Blackfin Options
10414 @subsection Blackfin Options
10415 @cindex Blackfin Options
10418 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10420 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10421 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10422 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10423 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10424 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10425 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10426 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10428 The optional @var{sirevision} specifies the silicon revision of the target
10429 Blackfin processor. Any workarounds available for the targeted silicon revision
10430 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10431 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10432 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10433 hexadecimal digits representing the major and minor numbers in the silicon
10434 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10435 is not defined. If @var{sirevision} is @samp{any}, the
10436 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10437 If this optional @var{sirevision} is not used, GCC assumes the latest known
10438 silicon revision of the targeted Blackfin processor.
10440 Support for @samp{bf561} is incomplete. For @samp{bf561},
10441 Only the processor macro is defined.
10442 Without this option, @samp{bf532} is used as the processor by default.
10443 The corresponding predefined processor macros for @var{cpu} is to
10444 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10445 provided by libgloss to be linked in if @option{-msim} is not given.
10449 Specifies that the program will be run on the simulator. This causes
10450 the simulator BSP provided by libgloss to be linked in. This option
10451 has effect only for @samp{bfin-elf} toolchain.
10452 Certain other options, such as @option{-mid-shared-library} and
10453 @option{-mfdpic}, imply @option{-msim}.
10455 @item -momit-leaf-frame-pointer
10456 @opindex momit-leaf-frame-pointer
10457 Don't keep the frame pointer in a register for leaf functions. This
10458 avoids the instructions to save, set up and restore frame pointers and
10459 makes an extra register available in leaf functions. The option
10460 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10461 which might make debugging harder.
10463 @item -mspecld-anomaly
10464 @opindex mspecld-anomaly
10465 When enabled, the compiler will ensure that the generated code does not
10466 contain speculative loads after jump instructions. If this option is used,
10467 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10469 @item -mno-specld-anomaly
10470 @opindex mno-specld-anomaly
10471 Don't generate extra code to prevent speculative loads from occurring.
10473 @item -mcsync-anomaly
10474 @opindex mcsync-anomaly
10475 When enabled, the compiler will ensure that the generated code does not
10476 contain CSYNC or SSYNC instructions too soon after conditional branches.
10477 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10479 @item -mno-csync-anomaly
10480 @opindex mno-csync-anomaly
10481 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10482 occurring too soon after a conditional branch.
10486 When enabled, the compiler is free to take advantage of the knowledge that
10487 the entire program fits into the low 64k of memory.
10490 @opindex mno-low-64k
10491 Assume that the program is arbitrarily large. This is the default.
10493 @item -mstack-check-l1
10494 @opindex mstack-check-l1
10495 Do stack checking using information placed into L1 scratchpad memory by the
10498 @item -mid-shared-library
10499 @opindex mid-shared-library
10500 Generate code that supports shared libraries via the library ID method.
10501 This allows for execute in place and shared libraries in an environment
10502 without virtual memory management. This option implies @option{-fPIC}.
10503 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10505 @item -mno-id-shared-library
10506 @opindex mno-id-shared-library
10507 Generate code that doesn't assume ID based shared libraries are being used.
10508 This is the default.
10510 @item -mleaf-id-shared-library
10511 @opindex mleaf-id-shared-library
10512 Generate code that supports shared libraries via the library ID method,
10513 but assumes that this library or executable won't link against any other
10514 ID shared libraries. That allows the compiler to use faster code for jumps
10517 @item -mno-leaf-id-shared-library
10518 @opindex mno-leaf-id-shared-library
10519 Do not assume that the code being compiled won't link against any ID shared
10520 libraries. Slower code will be generated for jump and call insns.
10522 @item -mshared-library-id=n
10523 @opindex mshared-library-id
10524 Specified the identification number of the ID based shared library being
10525 compiled. Specifying a value of 0 will generate more compact code, specifying
10526 other values will force the allocation of that number to the current
10527 library but is no more space or time efficient than omitting this option.
10531 Generate code that allows the data segment to be located in a different
10532 area of memory from the text segment. This allows for execute in place in
10533 an environment without virtual memory management by eliminating relocations
10534 against the text section.
10536 @item -mno-sep-data
10537 @opindex mno-sep-data
10538 Generate code that assumes that the data segment follows the text segment.
10539 This is the default.
10542 @itemx -mno-long-calls
10543 @opindex mlong-calls
10544 @opindex mno-long-calls
10545 Tells the compiler to perform function calls by first loading the
10546 address of the function into a register and then performing a subroutine
10547 call on this register. This switch is needed if the target function
10548 will lie outside of the 24 bit addressing range of the offset based
10549 version of subroutine call instruction.
10551 This feature is not enabled by default. Specifying
10552 @option{-mno-long-calls} will restore the default behavior. Note these
10553 switches have no effect on how the compiler generates code to handle
10554 function calls via function pointers.
10558 Link with the fast floating-point library. This library relaxes some of
10559 the IEEE floating-point standard's rules for checking inputs against
10560 Not-a-Number (NAN), in the interest of performance.
10563 @opindex minline-plt
10564 Enable inlining of PLT entries in function calls to functions that are
10565 not known to bind locally. It has no effect without @option{-mfdpic}.
10568 @opindex mmulticore
10569 Build standalone application for multicore Blackfin processor. Proper
10570 start files and link scripts will be used to support multicore.
10571 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10572 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10573 @option{-mcorea} or @option{-mcoreb}. If it's used without
10574 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10575 programming model is used. In this model, the main function of Core B
10576 should be named as coreb_main. If it's used with @option{-mcorea} or
10577 @option{-mcoreb}, one application per core programming model is used.
10578 If this option is not used, single core application programming
10583 Build standalone application for Core A of BF561 when using
10584 one application per core programming model. Proper start files
10585 and link scripts will be used to support Core A. This option
10586 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10590 Build standalone application for Core B of BF561 when using
10591 one application per core programming model. Proper start files
10592 and link scripts will be used to support Core B. This option
10593 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10594 should be used instead of main. It must be used with
10595 @option{-mmulticore}.
10599 Build standalone application for SDRAM. Proper start files and
10600 link scripts will be used to put the application into SDRAM.
10601 Loader should initialize SDRAM before loading the application
10602 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10606 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10607 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10608 are enabled; for standalone applications the default is off.
10612 @subsection CRIS Options
10613 @cindex CRIS Options
10615 These options are defined specifically for the CRIS ports.
10618 @item -march=@var{architecture-type}
10619 @itemx -mcpu=@var{architecture-type}
10622 Generate code for the specified architecture. The choices for
10623 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10624 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10625 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10628 @item -mtune=@var{architecture-type}
10630 Tune to @var{architecture-type} everything applicable about the generated
10631 code, except for the ABI and the set of available instructions. The
10632 choices for @var{architecture-type} are the same as for
10633 @option{-march=@var{architecture-type}}.
10635 @item -mmax-stack-frame=@var{n}
10636 @opindex mmax-stack-frame
10637 Warn when the stack frame of a function exceeds @var{n} bytes.
10643 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10644 @option{-march=v3} and @option{-march=v8} respectively.
10646 @item -mmul-bug-workaround
10647 @itemx -mno-mul-bug-workaround
10648 @opindex mmul-bug-workaround
10649 @opindex mno-mul-bug-workaround
10650 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10651 models where it applies. This option is active by default.
10655 Enable CRIS-specific verbose debug-related information in the assembly
10656 code. This option also has the effect to turn off the @samp{#NO_APP}
10657 formatted-code indicator to the assembler at the beginning of the
10662 Do not use condition-code results from previous instruction; always emit
10663 compare and test instructions before use of condition codes.
10665 @item -mno-side-effects
10666 @opindex mno-side-effects
10667 Do not emit instructions with side-effects in addressing modes other than
10670 @item -mstack-align
10671 @itemx -mno-stack-align
10672 @itemx -mdata-align
10673 @itemx -mno-data-align
10674 @itemx -mconst-align
10675 @itemx -mno-const-align
10676 @opindex mstack-align
10677 @opindex mno-stack-align
10678 @opindex mdata-align
10679 @opindex mno-data-align
10680 @opindex mconst-align
10681 @opindex mno-const-align
10682 These options (no-options) arranges (eliminate arrangements) for the
10683 stack-frame, individual data and constants to be aligned for the maximum
10684 single data access size for the chosen CPU model. The default is to
10685 arrange for 32-bit alignment. ABI details such as structure layout are
10686 not affected by these options.
10694 Similar to the stack- data- and const-align options above, these options
10695 arrange for stack-frame, writable data and constants to all be 32-bit,
10696 16-bit or 8-bit aligned. The default is 32-bit alignment.
10698 @item -mno-prologue-epilogue
10699 @itemx -mprologue-epilogue
10700 @opindex mno-prologue-epilogue
10701 @opindex mprologue-epilogue
10702 With @option{-mno-prologue-epilogue}, the normal function prologue and
10703 epilogue that sets up the stack-frame are omitted and no return
10704 instructions or return sequences are generated in the code. Use this
10705 option only together with visual inspection of the compiled code: no
10706 warnings or errors are generated when call-saved registers must be saved,
10707 or storage for local variable needs to be allocated.
10711 @opindex mno-gotplt
10713 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10714 instruction sequences that load addresses for functions from the PLT part
10715 of the GOT rather than (traditional on other architectures) calls to the
10716 PLT@. The default is @option{-mgotplt}.
10720 Legacy no-op option only recognized with the cris-axis-elf and
10721 cris-axis-linux-gnu targets.
10725 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10729 This option, recognized for the cris-axis-elf arranges
10730 to link with input-output functions from a simulator library. Code,
10731 initialized data and zero-initialized data are allocated consecutively.
10735 Like @option{-sim}, but pass linker options to locate initialized data at
10736 0x40000000 and zero-initialized data at 0x80000000.
10739 @node Darwin Options
10740 @subsection Darwin Options
10741 @cindex Darwin options
10743 These options are defined for all architectures running the Darwin operating
10746 FSF GCC on Darwin does not create ``fat'' object files; it will create
10747 an object file for the single architecture that it was built to
10748 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10749 @option{-arch} options are used; it does so by running the compiler or
10750 linker multiple times and joining the results together with
10753 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10754 @samp{i686}) is determined by the flags that specify the ISA
10755 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10756 @option{-force_cpusubtype_ALL} option can be used to override this.
10758 The Darwin tools vary in their behavior when presented with an ISA
10759 mismatch. The assembler, @file{as}, will only permit instructions to
10760 be used that are valid for the subtype of the file it is generating,
10761 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10762 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10763 and print an error if asked to create a shared library with a less
10764 restrictive subtype than its input files (for instance, trying to put
10765 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10766 for executables, @file{ld}, will quietly give the executable the most
10767 restrictive subtype of any of its input files.
10772 Add the framework directory @var{dir} to the head of the list of
10773 directories to be searched for header files. These directories are
10774 interleaved with those specified by @option{-I} options and are
10775 scanned in a left-to-right order.
10777 A framework directory is a directory with frameworks in it. A
10778 framework is a directory with a @samp{"Headers"} and/or
10779 @samp{"PrivateHeaders"} directory contained directly in it that ends
10780 in @samp{".framework"}. The name of a framework is the name of this
10781 directory excluding the @samp{".framework"}. Headers associated with
10782 the framework are found in one of those two directories, with
10783 @samp{"Headers"} being searched first. A subframework is a framework
10784 directory that is in a framework's @samp{"Frameworks"} directory.
10785 Includes of subframework headers can only appear in a header of a
10786 framework that contains the subframework, or in a sibling subframework
10787 header. Two subframeworks are siblings if they occur in the same
10788 framework. A subframework should not have the same name as a
10789 framework, a warning will be issued if this is violated. Currently a
10790 subframework cannot have subframeworks, in the future, the mechanism
10791 may be extended to support this. The standard frameworks can be found
10792 in @samp{"/System/Library/Frameworks"} and
10793 @samp{"/Library/Frameworks"}. An example include looks like
10794 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10795 the name of the framework and header.h is found in the
10796 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10798 @item -iframework@var{dir}
10799 @opindex iframework
10800 Like @option{-F} except the directory is a treated as a system
10801 directory. The main difference between this @option{-iframework} and
10802 @option{-F} is that with @option{-iframework} the compiler does not
10803 warn about constructs contained within header files found via
10804 @var{dir}. This option is valid only for the C family of languages.
10808 Emit debugging information for symbols that are used. For STABS
10809 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10810 This is by default ON@.
10814 Emit debugging information for all symbols and types.
10816 @item -mmacosx-version-min=@var{version}
10817 The earliest version of MacOS X that this executable will run on
10818 is @var{version}. Typical values of @var{version} include @code{10.1},
10819 @code{10.2}, and @code{10.3.9}.
10821 If the compiler was built to use the system's headers by default,
10822 then the default for this option is the system version on which the
10823 compiler is running, otherwise the default is to make choices which
10824 are compatible with as many systems and code bases as possible.
10828 Enable kernel development mode. The @option{-mkernel} option sets
10829 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10830 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10831 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10832 applicable. This mode also sets @option{-mno-altivec},
10833 @option{-msoft-float}, @option{-fno-builtin} and
10834 @option{-mlong-branch} for PowerPC targets.
10836 @item -mone-byte-bool
10837 @opindex mone-byte-bool
10838 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10839 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10840 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10841 option has no effect on x86.
10843 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10844 to generate code that is not binary compatible with code generated
10845 without that switch. Using this switch may require recompiling all
10846 other modules in a program, including system libraries. Use this
10847 switch to conform to a non-default data model.
10849 @item -mfix-and-continue
10850 @itemx -ffix-and-continue
10851 @itemx -findirect-data
10852 @opindex mfix-and-continue
10853 @opindex ffix-and-continue
10854 @opindex findirect-data
10855 Generate code suitable for fast turn around development. Needed to
10856 enable gdb to dynamically load @code{.o} files into already running
10857 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10858 are provided for backwards compatibility.
10862 Loads all members of static archive libraries.
10863 See man ld(1) for more information.
10865 @item -arch_errors_fatal
10866 @opindex arch_errors_fatal
10867 Cause the errors having to do with files that have the wrong architecture
10870 @item -bind_at_load
10871 @opindex bind_at_load
10872 Causes the output file to be marked such that the dynamic linker will
10873 bind all undefined references when the file is loaded or launched.
10877 Produce a Mach-o bundle format file.
10878 See man ld(1) for more information.
10880 @item -bundle_loader @var{executable}
10881 @opindex bundle_loader
10882 This option specifies the @var{executable} that will be loading the build
10883 output file being linked. See man ld(1) for more information.
10886 @opindex dynamiclib
10887 When passed this option, GCC will produce a dynamic library instead of
10888 an executable when linking, using the Darwin @file{libtool} command.
10890 @item -force_cpusubtype_ALL
10891 @opindex force_cpusubtype_ALL
10892 This causes GCC's output file to have the @var{ALL} subtype, instead of
10893 one controlled by the @option{-mcpu} or @option{-march} option.
10895 @item -allowable_client @var{client_name}
10896 @itemx -client_name
10897 @itemx -compatibility_version
10898 @itemx -current_version
10900 @itemx -dependency-file
10902 @itemx -dylinker_install_name
10904 @itemx -exported_symbols_list
10907 @itemx -flat_namespace
10908 @itemx -force_flat_namespace
10909 @itemx -headerpad_max_install_names
10912 @itemx -install_name
10913 @itemx -keep_private_externs
10914 @itemx -multi_module
10915 @itemx -multiply_defined
10916 @itemx -multiply_defined_unused
10919 @itemx -no_dead_strip_inits_and_terms
10920 @itemx -nofixprebinding
10921 @itemx -nomultidefs
10923 @itemx -noseglinkedit
10924 @itemx -pagezero_size
10926 @itemx -prebind_all_twolevel_modules
10927 @itemx -private_bundle
10929 @itemx -read_only_relocs
10931 @itemx -sectobjectsymbols
10935 @itemx -sectobjectsymbols
10938 @itemx -segs_read_only_addr
10940 @itemx -segs_read_write_addr
10941 @itemx -seg_addr_table
10942 @itemx -seg_addr_table_filename
10943 @itemx -seglinkedit
10945 @itemx -segs_read_only_addr
10946 @itemx -segs_read_write_addr
10947 @itemx -single_module
10949 @itemx -sub_library
10951 @itemx -sub_umbrella
10952 @itemx -twolevel_namespace
10955 @itemx -unexported_symbols_list
10956 @itemx -weak_reference_mismatches
10957 @itemx -whatsloaded
10958 @opindex allowable_client
10959 @opindex client_name
10960 @opindex compatibility_version
10961 @opindex current_version
10962 @opindex dead_strip
10963 @opindex dependency-file
10964 @opindex dylib_file
10965 @opindex dylinker_install_name
10967 @opindex exported_symbols_list
10969 @opindex flat_namespace
10970 @opindex force_flat_namespace
10971 @opindex headerpad_max_install_names
10972 @opindex image_base
10974 @opindex install_name
10975 @opindex keep_private_externs
10976 @opindex multi_module
10977 @opindex multiply_defined
10978 @opindex multiply_defined_unused
10979 @opindex noall_load
10980 @opindex no_dead_strip_inits_and_terms
10981 @opindex nofixprebinding
10982 @opindex nomultidefs
10984 @opindex noseglinkedit
10985 @opindex pagezero_size
10987 @opindex prebind_all_twolevel_modules
10988 @opindex private_bundle
10989 @opindex read_only_relocs
10991 @opindex sectobjectsymbols
10994 @opindex sectcreate
10995 @opindex sectobjectsymbols
10998 @opindex segs_read_only_addr
10999 @opindex segs_read_write_addr
11000 @opindex seg_addr_table
11001 @opindex seg_addr_table_filename
11002 @opindex seglinkedit
11004 @opindex segs_read_only_addr
11005 @opindex segs_read_write_addr
11006 @opindex single_module
11008 @opindex sub_library
11009 @opindex sub_umbrella
11010 @opindex twolevel_namespace
11013 @opindex unexported_symbols_list
11014 @opindex weak_reference_mismatches
11015 @opindex whatsloaded
11016 These options are passed to the Darwin linker. The Darwin linker man page
11017 describes them in detail.
11020 @node DEC Alpha Options
11021 @subsection DEC Alpha Options
11023 These @samp{-m} options are defined for the DEC Alpha implementations:
11026 @item -mno-soft-float
11027 @itemx -msoft-float
11028 @opindex mno-soft-float
11029 @opindex msoft-float
11030 Use (do not use) the hardware floating-point instructions for
11031 floating-point operations. When @option{-msoft-float} is specified,
11032 functions in @file{libgcc.a} will be used to perform floating-point
11033 operations. Unless they are replaced by routines that emulate the
11034 floating-point operations, or compiled in such a way as to call such
11035 emulations routines, these routines will issue floating-point
11036 operations. If you are compiling for an Alpha without floating-point
11037 operations, you must ensure that the library is built so as not to call
11040 Note that Alpha implementations without floating-point operations are
11041 required to have floating-point registers.
11044 @itemx -mno-fp-regs
11046 @opindex mno-fp-regs
11047 Generate code that uses (does not use) the floating-point register set.
11048 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11049 register set is not used, floating point operands are passed in integer
11050 registers as if they were integers and floating-point results are passed
11051 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11052 so any function with a floating-point argument or return value called by code
11053 compiled with @option{-mno-fp-regs} must also be compiled with that
11056 A typical use of this option is building a kernel that does not use,
11057 and hence need not save and restore, any floating-point registers.
11061 The Alpha architecture implements floating-point hardware optimized for
11062 maximum performance. It is mostly compliant with the IEEE floating
11063 point standard. However, for full compliance, software assistance is
11064 required. This option generates code fully IEEE compliant code
11065 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11066 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11067 defined during compilation. The resulting code is less efficient but is
11068 able to correctly support denormalized numbers and exceptional IEEE
11069 values such as not-a-number and plus/minus infinity. Other Alpha
11070 compilers call this option @option{-ieee_with_no_inexact}.
11072 @item -mieee-with-inexact
11073 @opindex mieee-with-inexact
11074 This is like @option{-mieee} except the generated code also maintains
11075 the IEEE @var{inexact-flag}. Turning on this option causes the
11076 generated code to implement fully-compliant IEEE math. In addition to
11077 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11078 macro. On some Alpha implementations the resulting code may execute
11079 significantly slower than the code generated by default. Since there is
11080 very little code that depends on the @var{inexact-flag}, you should
11081 normally not specify this option. Other Alpha compilers call this
11082 option @option{-ieee_with_inexact}.
11084 @item -mfp-trap-mode=@var{trap-mode}
11085 @opindex mfp-trap-mode
11086 This option controls what floating-point related traps are enabled.
11087 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11088 The trap mode can be set to one of four values:
11092 This is the default (normal) setting. The only traps that are enabled
11093 are the ones that cannot be disabled in software (e.g., division by zero
11097 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11101 Like @samp{u}, but the instructions are marked to be safe for software
11102 completion (see Alpha architecture manual for details).
11105 Like @samp{su}, but inexact traps are enabled as well.
11108 @item -mfp-rounding-mode=@var{rounding-mode}
11109 @opindex mfp-rounding-mode
11110 Selects the IEEE rounding mode. Other Alpha compilers call this option
11111 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11116 Normal IEEE rounding mode. Floating point numbers are rounded towards
11117 the nearest machine number or towards the even machine number in case
11121 Round towards minus infinity.
11124 Chopped rounding mode. Floating point numbers are rounded towards zero.
11127 Dynamic rounding mode. A field in the floating point control register
11128 (@var{fpcr}, see Alpha architecture reference manual) controls the
11129 rounding mode in effect. The C library initializes this register for
11130 rounding towards plus infinity. Thus, unless your program modifies the
11131 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11134 @item -mtrap-precision=@var{trap-precision}
11135 @opindex mtrap-precision
11136 In the Alpha architecture, floating point traps are imprecise. This
11137 means without software assistance it is impossible to recover from a
11138 floating trap and program execution normally needs to be terminated.
11139 GCC can generate code that can assist operating system trap handlers
11140 in determining the exact location that caused a floating point trap.
11141 Depending on the requirements of an application, different levels of
11142 precisions can be selected:
11146 Program precision. This option is the default and means a trap handler
11147 can only identify which program caused a floating point exception.
11150 Function precision. The trap handler can determine the function that
11151 caused a floating point exception.
11154 Instruction precision. The trap handler can determine the exact
11155 instruction that caused a floating point exception.
11158 Other Alpha compilers provide the equivalent options called
11159 @option{-scope_safe} and @option{-resumption_safe}.
11161 @item -mieee-conformant
11162 @opindex mieee-conformant
11163 This option marks the generated code as IEEE conformant. You must not
11164 use this option unless you also specify @option{-mtrap-precision=i} and either
11165 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11166 is to emit the line @samp{.eflag 48} in the function prologue of the
11167 generated assembly file. Under DEC Unix, this has the effect that
11168 IEEE-conformant math library routines will be linked in.
11170 @item -mbuild-constants
11171 @opindex mbuild-constants
11172 Normally GCC examines a 32- or 64-bit integer constant to
11173 see if it can construct it from smaller constants in two or three
11174 instructions. If it cannot, it will output the constant as a literal and
11175 generate code to load it from the data segment at runtime.
11177 Use this option to require GCC to construct @emph{all} integer constants
11178 using code, even if it takes more instructions (the maximum is six).
11180 You would typically use this option to build a shared library dynamic
11181 loader. Itself a shared library, it must relocate itself in memory
11182 before it can find the variables and constants in its own data segment.
11188 Select whether to generate code to be assembled by the vendor-supplied
11189 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11207 Indicate whether GCC should generate code to use the optional BWX,
11208 CIX, FIX and MAX instruction sets. The default is to use the instruction
11209 sets supported by the CPU type specified via @option{-mcpu=} option or that
11210 of the CPU on which GCC was built if none was specified.
11213 @itemx -mfloat-ieee
11214 @opindex mfloat-vax
11215 @opindex mfloat-ieee
11216 Generate code that uses (does not use) VAX F and G floating point
11217 arithmetic instead of IEEE single and double precision.
11219 @item -mexplicit-relocs
11220 @itemx -mno-explicit-relocs
11221 @opindex mexplicit-relocs
11222 @opindex mno-explicit-relocs
11223 Older Alpha assemblers provided no way to generate symbol relocations
11224 except via assembler macros. Use of these macros does not allow
11225 optimal instruction scheduling. GNU binutils as of version 2.12
11226 supports a new syntax that allows the compiler to explicitly mark
11227 which relocations should apply to which instructions. This option
11228 is mostly useful for debugging, as GCC detects the capabilities of
11229 the assembler when it is built and sets the default accordingly.
11232 @itemx -mlarge-data
11233 @opindex msmall-data
11234 @opindex mlarge-data
11235 When @option{-mexplicit-relocs} is in effect, static data is
11236 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11237 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11238 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11239 16-bit relocations off of the @code{$gp} register. This limits the
11240 size of the small data area to 64KB, but allows the variables to be
11241 directly accessed via a single instruction.
11243 The default is @option{-mlarge-data}. With this option the data area
11244 is limited to just below 2GB@. Programs that require more than 2GB of
11245 data must use @code{malloc} or @code{mmap} to allocate the data in the
11246 heap instead of in the program's data segment.
11248 When generating code for shared libraries, @option{-fpic} implies
11249 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11252 @itemx -mlarge-text
11253 @opindex msmall-text
11254 @opindex mlarge-text
11255 When @option{-msmall-text} is used, the compiler assumes that the
11256 code of the entire program (or shared library) fits in 4MB, and is
11257 thus reachable with a branch instruction. When @option{-msmall-data}
11258 is used, the compiler can assume that all local symbols share the
11259 same @code{$gp} value, and thus reduce the number of instructions
11260 required for a function call from 4 to 1.
11262 The default is @option{-mlarge-text}.
11264 @item -mcpu=@var{cpu_type}
11266 Set the instruction set and instruction scheduling parameters for
11267 machine type @var{cpu_type}. You can specify either the @samp{EV}
11268 style name or the corresponding chip number. GCC supports scheduling
11269 parameters for the EV4, EV5 and EV6 family of processors and will
11270 choose the default values for the instruction set from the processor
11271 you specify. If you do not specify a processor type, GCC will default
11272 to the processor on which the compiler was built.
11274 Supported values for @var{cpu_type} are
11280 Schedules as an EV4 and has no instruction set extensions.
11284 Schedules as an EV5 and has no instruction set extensions.
11288 Schedules as an EV5 and supports the BWX extension.
11293 Schedules as an EV5 and supports the BWX and MAX extensions.
11297 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11301 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11304 Native Linux/GNU toolchains also support the value @samp{native},
11305 which selects the best architecture option for the host processor.
11306 @option{-mcpu=native} has no effect if GCC does not recognize
11309 @item -mtune=@var{cpu_type}
11311 Set only the instruction scheduling parameters for machine type
11312 @var{cpu_type}. The instruction set is not changed.
11314 Native Linux/GNU toolchains also support the value @samp{native},
11315 which selects the best architecture option for the host processor.
11316 @option{-mtune=native} has no effect if GCC does not recognize
11319 @item -mmemory-latency=@var{time}
11320 @opindex mmemory-latency
11321 Sets the latency the scheduler should assume for typical memory
11322 references as seen by the application. This number is highly
11323 dependent on the memory access patterns used by the application
11324 and the size of the external cache on the machine.
11326 Valid options for @var{time} are
11330 A decimal number representing clock cycles.
11336 The compiler contains estimates of the number of clock cycles for
11337 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11338 (also called Dcache, Scache, and Bcache), as well as to main memory.
11339 Note that L3 is only valid for EV5.
11344 @node DEC Alpha/VMS Options
11345 @subsection DEC Alpha/VMS Options
11347 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11350 @item -mvms-return-codes
11351 @opindex mvms-return-codes
11352 Return VMS condition codes from main. The default is to return POSIX
11353 style condition (e.g.@: error) codes.
11355 @item -mdebug-main=@var{prefix}
11356 @opindex mdebug-main=@var{prefix}
11357 Flag the first routine whose name starts with @var{prefix} as the main
11358 routine for the debugger.
11362 Default to 64bit memory allocation routines.
11366 @subsection FR30 Options
11367 @cindex FR30 Options
11369 These options are defined specifically for the FR30 port.
11373 @item -msmall-model
11374 @opindex msmall-model
11375 Use the small address space model. This can produce smaller code, but
11376 it does assume that all symbolic values and addresses will fit into a
11381 Assume that run-time support has been provided and so there is no need
11382 to include the simulator library (@file{libsim.a}) on the linker
11388 @subsection FRV Options
11389 @cindex FRV Options
11395 Only use the first 32 general purpose registers.
11400 Use all 64 general purpose registers.
11405 Use only the first 32 floating point registers.
11410 Use all 64 floating point registers
11413 @opindex mhard-float
11415 Use hardware instructions for floating point operations.
11418 @opindex msoft-float
11420 Use library routines for floating point operations.
11425 Dynamically allocate condition code registers.
11430 Do not try to dynamically allocate condition code registers, only
11431 use @code{icc0} and @code{fcc0}.
11436 Change ABI to use double word insns.
11441 Do not use double word instructions.
11446 Use floating point double instructions.
11449 @opindex mno-double
11451 Do not use floating point double instructions.
11456 Use media instructions.
11461 Do not use media instructions.
11466 Use multiply and add/subtract instructions.
11469 @opindex mno-muladd
11471 Do not use multiply and add/subtract instructions.
11476 Select the FDPIC ABI, that uses function descriptors to represent
11477 pointers to functions. Without any PIC/PIE-related options, it
11478 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11479 assumes GOT entries and small data are within a 12-bit range from the
11480 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11481 are computed with 32 bits.
11482 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11485 @opindex minline-plt
11487 Enable inlining of PLT entries in function calls to functions that are
11488 not known to bind locally. It has no effect without @option{-mfdpic}.
11489 It's enabled by default if optimizing for speed and compiling for
11490 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11491 optimization option such as @option{-O3} or above is present in the
11497 Assume a large TLS segment when generating thread-local code.
11502 Do not assume a large TLS segment when generating thread-local code.
11507 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11508 that is known to be in read-only sections. It's enabled by default,
11509 except for @option{-fpic} or @option{-fpie}: even though it may help
11510 make the global offset table smaller, it trades 1 instruction for 4.
11511 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11512 one of which may be shared by multiple symbols, and it avoids the need
11513 for a GOT entry for the referenced symbol, so it's more likely to be a
11514 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11516 @item -multilib-library-pic
11517 @opindex multilib-library-pic
11519 Link with the (library, not FD) pic libraries. It's implied by
11520 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11521 @option{-fpic} without @option{-mfdpic}. You should never have to use
11525 @opindex mlinked-fp
11527 Follow the EABI requirement of always creating a frame pointer whenever
11528 a stack frame is allocated. This option is enabled by default and can
11529 be disabled with @option{-mno-linked-fp}.
11532 @opindex mlong-calls
11534 Use indirect addressing to call functions outside the current
11535 compilation unit. This allows the functions to be placed anywhere
11536 within the 32-bit address space.
11538 @item -malign-labels
11539 @opindex malign-labels
11541 Try to align labels to an 8-byte boundary by inserting nops into the
11542 previous packet. This option only has an effect when VLIW packing
11543 is enabled. It doesn't create new packets; it merely adds nops to
11546 @item -mlibrary-pic
11547 @opindex mlibrary-pic
11549 Generate position-independent EABI code.
11554 Use only the first four media accumulator registers.
11559 Use all eight media accumulator registers.
11564 Pack VLIW instructions.
11569 Do not pack VLIW instructions.
11572 @opindex mno-eflags
11574 Do not mark ABI switches in e_flags.
11577 @opindex mcond-move
11579 Enable the use of conditional-move instructions (default).
11581 This switch is mainly for debugging the compiler and will likely be removed
11582 in a future version.
11584 @item -mno-cond-move
11585 @opindex mno-cond-move
11587 Disable the use of conditional-move instructions.
11589 This switch is mainly for debugging the compiler and will likely be removed
11590 in a future version.
11595 Enable the use of conditional set instructions (default).
11597 This switch is mainly for debugging the compiler and will likely be removed
11598 in a future version.
11603 Disable the use of conditional set instructions.
11605 This switch is mainly for debugging the compiler and will likely be removed
11606 in a future version.
11609 @opindex mcond-exec
11611 Enable the use of conditional execution (default).
11613 This switch is mainly for debugging the compiler and will likely be removed
11614 in a future version.
11616 @item -mno-cond-exec
11617 @opindex mno-cond-exec
11619 Disable the use of conditional execution.
11621 This switch is mainly for debugging the compiler and will likely be removed
11622 in a future version.
11624 @item -mvliw-branch
11625 @opindex mvliw-branch
11627 Run a pass to pack branches into VLIW instructions (default).
11629 This switch is mainly for debugging the compiler and will likely be removed
11630 in a future version.
11632 @item -mno-vliw-branch
11633 @opindex mno-vliw-branch
11635 Do not run a pass to pack branches into VLIW instructions.
11637 This switch is mainly for debugging the compiler and will likely be removed
11638 in a future version.
11640 @item -mmulti-cond-exec
11641 @opindex mmulti-cond-exec
11643 Enable optimization of @code{&&} and @code{||} in conditional execution
11646 This switch is mainly for debugging the compiler and will likely be removed
11647 in a future version.
11649 @item -mno-multi-cond-exec
11650 @opindex mno-multi-cond-exec
11652 Disable optimization of @code{&&} and @code{||} in conditional execution.
11654 This switch is mainly for debugging the compiler and will likely be removed
11655 in a future version.
11657 @item -mnested-cond-exec
11658 @opindex mnested-cond-exec
11660 Enable nested conditional execution optimizations (default).
11662 This switch is mainly for debugging the compiler and will likely be removed
11663 in a future version.
11665 @item -mno-nested-cond-exec
11666 @opindex mno-nested-cond-exec
11668 Disable nested conditional execution optimizations.
11670 This switch is mainly for debugging the compiler and will likely be removed
11671 in a future version.
11673 @item -moptimize-membar
11674 @opindex moptimize-membar
11676 This switch removes redundant @code{membar} instructions from the
11677 compiler generated code. It is enabled by default.
11679 @item -mno-optimize-membar
11680 @opindex mno-optimize-membar
11682 This switch disables the automatic removal of redundant @code{membar}
11683 instructions from the generated code.
11685 @item -mtomcat-stats
11686 @opindex mtomcat-stats
11688 Cause gas to print out tomcat statistics.
11690 @item -mcpu=@var{cpu}
11693 Select the processor type for which to generate code. Possible values are
11694 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11695 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11699 @node GNU/Linux Options
11700 @subsection GNU/Linux Options
11702 These @samp{-m} options are defined for GNU/Linux targets:
11707 Use the GNU C library. This is the default except
11708 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11712 Use uClibc C library. This is the default on
11713 @samp{*-*-linux-*uclibc*} targets.
11717 Use Bionic C library. This is the default on
11718 @samp{*-*-linux-*android*} targets.
11722 Compile code compatible with Android platform. This is the default on
11723 @samp{*-*-linux-*android*} targets.
11725 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11726 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11727 this option makes the GCC driver pass Android-specific options to the linker.
11728 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11731 @item -tno-android-cc
11732 @opindex tno-android-cc
11733 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11734 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11735 @option{-fno-rtti} by default.
11737 @item -tno-android-ld
11738 @opindex tno-android-ld
11739 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11740 linking options to the linker.
11744 @node H8/300 Options
11745 @subsection H8/300 Options
11747 These @samp{-m} options are defined for the H8/300 implementations:
11752 Shorten some address references at link time, when possible; uses the
11753 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11754 ld, Using ld}, for a fuller description.
11758 Generate code for the H8/300H@.
11762 Generate code for the H8S@.
11766 Generate code for the H8S and H8/300H in the normal mode. This switch
11767 must be used either with @option{-mh} or @option{-ms}.
11771 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11775 Make @code{int} data 32 bits by default.
11778 @opindex malign-300
11779 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11780 The default for the H8/300H and H8S is to align longs and floats on 4
11782 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11783 This option has no effect on the H8/300.
11787 @subsection HPPA Options
11788 @cindex HPPA Options
11790 These @samp{-m} options are defined for the HPPA family of computers:
11793 @item -march=@var{architecture-type}
11795 Generate code for the specified architecture. The choices for
11796 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11797 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11798 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11799 architecture option for your machine. Code compiled for lower numbered
11800 architectures will run on higher numbered architectures, but not the
11803 @item -mpa-risc-1-0
11804 @itemx -mpa-risc-1-1
11805 @itemx -mpa-risc-2-0
11806 @opindex mpa-risc-1-0
11807 @opindex mpa-risc-1-1
11808 @opindex mpa-risc-2-0
11809 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11812 @opindex mbig-switch
11813 Generate code suitable for big switch tables. Use this option only if
11814 the assembler/linker complain about out of range branches within a switch
11817 @item -mjump-in-delay
11818 @opindex mjump-in-delay
11819 Fill delay slots of function calls with unconditional jump instructions
11820 by modifying the return pointer for the function call to be the target
11821 of the conditional jump.
11823 @item -mdisable-fpregs
11824 @opindex mdisable-fpregs
11825 Prevent floating point registers from being used in any manner. This is
11826 necessary for compiling kernels which perform lazy context switching of
11827 floating point registers. If you use this option and attempt to perform
11828 floating point operations, the compiler will abort.
11830 @item -mdisable-indexing
11831 @opindex mdisable-indexing
11832 Prevent the compiler from using indexing address modes. This avoids some
11833 rather obscure problems when compiling MIG generated code under MACH@.
11835 @item -mno-space-regs
11836 @opindex mno-space-regs
11837 Generate code that assumes the target has no space registers. This allows
11838 GCC to generate faster indirect calls and use unscaled index address modes.
11840 Such code is suitable for level 0 PA systems and kernels.
11842 @item -mfast-indirect-calls
11843 @opindex mfast-indirect-calls
11844 Generate code that assumes calls never cross space boundaries. This
11845 allows GCC to emit code which performs faster indirect calls.
11847 This option will not work in the presence of shared libraries or nested
11850 @item -mfixed-range=@var{register-range}
11851 @opindex mfixed-range
11852 Generate code treating the given register range as fixed registers.
11853 A fixed register is one that the register allocator can not use. This is
11854 useful when compiling kernel code. A register range is specified as
11855 two registers separated by a dash. Multiple register ranges can be
11856 specified separated by a comma.
11858 @item -mlong-load-store
11859 @opindex mlong-load-store
11860 Generate 3-instruction load and store sequences as sometimes required by
11861 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11864 @item -mportable-runtime
11865 @opindex mportable-runtime
11866 Use the portable calling conventions proposed by HP for ELF systems.
11870 Enable the use of assembler directives only GAS understands.
11872 @item -mschedule=@var{cpu-type}
11874 Schedule code according to the constraints for the machine type
11875 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11876 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11877 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11878 proper scheduling option for your machine. The default scheduling is
11882 @opindex mlinker-opt
11883 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11884 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11885 linkers in which they give bogus error messages when linking some programs.
11888 @opindex msoft-float
11889 Generate output containing library calls for floating point.
11890 @strong{Warning:} the requisite libraries are not available for all HPPA
11891 targets. Normally the facilities of the machine's usual C compiler are
11892 used, but this cannot be done directly in cross-compilation. You must make
11893 your own arrangements to provide suitable library functions for
11896 @option{-msoft-float} changes the calling convention in the output file;
11897 therefore, it is only useful if you compile @emph{all} of a program with
11898 this option. In particular, you need to compile @file{libgcc.a}, the
11899 library that comes with GCC, with @option{-msoft-float} in order for
11904 Generate the predefine, @code{_SIO}, for server IO@. The default is
11905 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11906 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11907 options are available under HP-UX and HI-UX@.
11911 Use GNU ld specific options. This passes @option{-shared} to ld when
11912 building a shared library. It is the default when GCC is configured,
11913 explicitly or implicitly, with the GNU linker. This option does not
11914 have any affect on which ld is called, it only changes what parameters
11915 are passed to that ld. The ld that is called is determined by the
11916 @option{--with-ld} configure option, GCC's program search path, and
11917 finally by the user's @env{PATH}. The linker used by GCC can be printed
11918 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11919 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11923 Use HP ld specific options. This passes @option{-b} to ld when building
11924 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11925 links. It is the default when GCC is configured, explicitly or
11926 implicitly, with the HP linker. This option does not have any affect on
11927 which ld is called, it only changes what parameters are passed to that
11928 ld. The ld that is called is determined by the @option{--with-ld}
11929 configure option, GCC's program search path, and finally by the user's
11930 @env{PATH}. The linker used by GCC can be printed using @samp{which
11931 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11932 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11935 @opindex mno-long-calls
11936 Generate code that uses long call sequences. This ensures that a call
11937 is always able to reach linker generated stubs. The default is to generate
11938 long calls only when the distance from the call site to the beginning
11939 of the function or translation unit, as the case may be, exceeds a
11940 predefined limit set by the branch type being used. The limits for
11941 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11942 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11945 Distances are measured from the beginning of functions when using the
11946 @option{-ffunction-sections} option, or when using the @option{-mgas}
11947 and @option{-mno-portable-runtime} options together under HP-UX with
11950 It is normally not desirable to use this option as it will degrade
11951 performance. However, it may be useful in large applications,
11952 particularly when partial linking is used to build the application.
11954 The types of long calls used depends on the capabilities of the
11955 assembler and linker, and the type of code being generated. The
11956 impact on systems that support long absolute calls, and long pic
11957 symbol-difference or pc-relative calls should be relatively small.
11958 However, an indirect call is used on 32-bit ELF systems in pic code
11959 and it is quite long.
11961 @item -munix=@var{unix-std}
11963 Generate compiler predefines and select a startfile for the specified
11964 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11965 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11966 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11967 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11968 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11971 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11972 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11973 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11974 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11975 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11976 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11978 It is @emph{important} to note that this option changes the interfaces
11979 for various library routines. It also affects the operational behavior
11980 of the C library. Thus, @emph{extreme} care is needed in using this
11983 Library code that is intended to operate with more than one UNIX
11984 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11985 as appropriate. Most GNU software doesn't provide this capability.
11989 Suppress the generation of link options to search libdld.sl when the
11990 @option{-static} option is specified on HP-UX 10 and later.
11994 The HP-UX implementation of setlocale in libc has a dependency on
11995 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11996 when the @option{-static} option is specified, special link options
11997 are needed to resolve this dependency.
11999 On HP-UX 10 and later, the GCC driver adds the necessary options to
12000 link with libdld.sl when the @option{-static} option is specified.
12001 This causes the resulting binary to be dynamic. On the 64-bit port,
12002 the linkers generate dynamic binaries by default in any case. The
12003 @option{-nolibdld} option can be used to prevent the GCC driver from
12004 adding these link options.
12008 Add support for multithreading with the @dfn{dce thread} library
12009 under HP-UX@. This option sets flags for both the preprocessor and
12013 @node i386 and x86-64 Options
12014 @subsection Intel 386 and AMD x86-64 Options
12015 @cindex i386 Options
12016 @cindex x86-64 Options
12017 @cindex Intel 386 Options
12018 @cindex AMD x86-64 Options
12020 These @samp{-m} options are defined for the i386 and x86-64 family of
12024 @item -mtune=@var{cpu-type}
12026 Tune to @var{cpu-type} everything applicable about the generated code, except
12027 for the ABI and the set of available instructions. The choices for
12028 @var{cpu-type} are:
12031 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12032 If you know the CPU on which your code will run, then you should use
12033 the corresponding @option{-mtune} option instead of
12034 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12035 of your application will have, then you should use this option.
12037 As new processors are deployed in the marketplace, the behavior of this
12038 option will change. Therefore, if you upgrade to a newer version of
12039 GCC, the code generated option will change to reflect the processors
12040 that were most common when that version of GCC was released.
12042 There is no @option{-march=generic} option because @option{-march}
12043 indicates the instruction set the compiler can use, and there is no
12044 generic instruction set applicable to all processors. In contrast,
12045 @option{-mtune} indicates the processor (or, in this case, collection of
12046 processors) for which the code is optimized.
12048 This selects the CPU to tune for at compilation time by determining
12049 the processor type of the compiling machine. Using @option{-mtune=native}
12050 will produce code optimized for the local machine under the constraints
12051 of the selected instruction set. Using @option{-march=native} will
12052 enable all instruction subsets supported by the local machine (hence
12053 the result might not run on different machines).
12055 Original Intel's i386 CPU@.
12057 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12058 @item i586, pentium
12059 Intel Pentium CPU with no MMX support.
12061 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12063 Intel PentiumPro CPU@.
12065 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12066 instruction set will be used, so the code will run on all i686 family chips.
12068 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12069 @item pentium3, pentium3m
12070 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12073 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12074 support. Used by Centrino notebooks.
12075 @item pentium4, pentium4m
12076 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12078 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12081 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12082 SSE2 and SSE3 instruction set support.
12084 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12085 instruction set support.
12087 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12088 and SSE4.2 instruction set support.
12090 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12091 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12093 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12094 instruction set support.
12096 AMD K6 CPU with MMX instruction set support.
12098 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12099 @item athlon, athlon-tbird
12100 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12102 @item athlon-4, athlon-xp, athlon-mp
12103 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12104 instruction set support.
12105 @item k8, opteron, athlon64, athlon-fx
12106 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12107 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12108 @item k8-sse3, opteron-sse3, athlon64-sse3
12109 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12110 @item amdfam10, barcelona
12111 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12112 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12113 instruction set extensions.)
12115 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12118 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12119 instruction set support.
12121 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12122 implemented for this chip.)
12124 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12125 implemented for this chip.)
12127 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12130 While picking a specific @var{cpu-type} will schedule things appropriately
12131 for that particular chip, the compiler will not generate any code that
12132 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12135 @item -march=@var{cpu-type}
12137 Generate instructions for the machine type @var{cpu-type}. The choices
12138 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12139 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12141 @item -mcpu=@var{cpu-type}
12143 A deprecated synonym for @option{-mtune}.
12145 @item -mfpmath=@var{unit}
12147 Generate floating point arithmetics for selected unit @var{unit}. The choices
12148 for @var{unit} are:
12152 Use the standard 387 floating point coprocessor present majority of chips and
12153 emulated otherwise. Code compiled with this option will run almost everywhere.
12154 The temporary results are computed in 80bit precision instead of precision
12155 specified by the type resulting in slightly different results compared to most
12156 of other chips. See @option{-ffloat-store} for more detailed description.
12158 This is the default choice for i386 compiler.
12161 Use scalar floating point instructions present in the SSE instruction set.
12162 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12163 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12164 instruction set supports only single precision arithmetics, thus the double and
12165 extended precision arithmetics is still done using 387. Later version, present
12166 only in Pentium4 and the future AMD x86-64 chips supports double precision
12169 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12170 or @option{-msse2} switches to enable SSE extensions and make this option
12171 effective. For the x86-64 compiler, these extensions are enabled by default.
12173 The resulting code should be considerably faster in the majority of cases and avoid
12174 the numerical instability problems of 387 code, but may break some existing
12175 code that expects temporaries to be 80bit.
12177 This is the default choice for the x86-64 compiler.
12182 Attempt to utilize both instruction sets at once. This effectively double the
12183 amount of available registers and on chips with separate execution units for
12184 387 and SSE the execution resources too. Use this option with care, as it is
12185 still experimental, because the GCC register allocator does not model separate
12186 functional units well resulting in instable performance.
12189 @item -masm=@var{dialect}
12190 @opindex masm=@var{dialect}
12191 Output asm instructions using selected @var{dialect}. Supported
12192 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12193 not support @samp{intel}.
12196 @itemx -mno-ieee-fp
12198 @opindex mno-ieee-fp
12199 Control whether or not the compiler uses IEEE floating point
12200 comparisons. These handle correctly the case where the result of a
12201 comparison is unordered.
12204 @opindex msoft-float
12205 Generate output containing library calls for floating point.
12206 @strong{Warning:} the requisite libraries are not part of GCC@.
12207 Normally the facilities of the machine's usual C compiler are used, but
12208 this can't be done directly in cross-compilation. You must make your
12209 own arrangements to provide suitable library functions for
12212 On machines where a function returns floating point results in the 80387
12213 register stack, some floating point opcodes may be emitted even if
12214 @option{-msoft-float} is used.
12216 @item -mno-fp-ret-in-387
12217 @opindex mno-fp-ret-in-387
12218 Do not use the FPU registers for return values of functions.
12220 The usual calling convention has functions return values of types
12221 @code{float} and @code{double} in an FPU register, even if there
12222 is no FPU@. The idea is that the operating system should emulate
12225 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12226 in ordinary CPU registers instead.
12228 @item -mno-fancy-math-387
12229 @opindex mno-fancy-math-387
12230 Some 387 emulators do not support the @code{sin}, @code{cos} and
12231 @code{sqrt} instructions for the 387. Specify this option to avoid
12232 generating those instructions. This option is the default on FreeBSD,
12233 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12234 indicates that the target CPU will always have an FPU and so the
12235 instruction will not need emulation. As of revision 2.6.1, these
12236 instructions are not generated unless you also use the
12237 @option{-funsafe-math-optimizations} switch.
12239 @item -malign-double
12240 @itemx -mno-align-double
12241 @opindex malign-double
12242 @opindex mno-align-double
12243 Control whether GCC aligns @code{double}, @code{long double}, and
12244 @code{long long} variables on a two word boundary or a one word
12245 boundary. Aligning @code{double} variables on a two word boundary will
12246 produce code that runs somewhat faster on a @samp{Pentium} at the
12247 expense of more memory.
12249 On x86-64, @option{-malign-double} is enabled by default.
12251 @strong{Warning:} if you use the @option{-malign-double} switch,
12252 structures containing the above types will be aligned differently than
12253 the published application binary interface specifications for the 386
12254 and will not be binary compatible with structures in code compiled
12255 without that switch.
12257 @item -m96bit-long-double
12258 @itemx -m128bit-long-double
12259 @opindex m96bit-long-double
12260 @opindex m128bit-long-double
12261 These switches control the size of @code{long double} type. The i386
12262 application binary interface specifies the size to be 96 bits,
12263 so @option{-m96bit-long-double} is the default in 32 bit mode.
12265 Modern architectures (Pentium and newer) would prefer @code{long double}
12266 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12267 conforming to the ABI, this would not be possible. So specifying a
12268 @option{-m128bit-long-double} will align @code{long double}
12269 to a 16 byte boundary by padding the @code{long double} with an additional
12272 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12273 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12275 Notice that neither of these options enable any extra precision over the x87
12276 standard of 80 bits for a @code{long double}.
12278 @strong{Warning:} if you override the default value for your target ABI, the
12279 structures and arrays containing @code{long double} variables will change
12280 their size as well as function calling convention for function taking
12281 @code{long double} will be modified. Hence they will not be binary
12282 compatible with arrays or structures in code compiled without that switch.
12284 @item -mlarge-data-threshold=@var{number}
12285 @opindex mlarge-data-threshold=@var{number}
12286 When @option{-mcmodel=medium} is specified, the data greater than
12287 @var{threshold} are placed in large data section. This value must be the
12288 same across all object linked into the binary and defaults to 65535.
12292 Use a different function-calling convention, in which functions that
12293 take a fixed number of arguments return with the @code{ret} @var{num}
12294 instruction, which pops their arguments while returning. This saves one
12295 instruction in the caller since there is no need to pop the arguments
12298 You can specify that an individual function is called with this calling
12299 sequence with the function attribute @samp{stdcall}. You can also
12300 override the @option{-mrtd} option by using the function attribute
12301 @samp{cdecl}. @xref{Function Attributes}.
12303 @strong{Warning:} this calling convention is incompatible with the one
12304 normally used on Unix, so you cannot use it if you need to call
12305 libraries compiled with the Unix compiler.
12307 Also, you must provide function prototypes for all functions that
12308 take variable numbers of arguments (including @code{printf});
12309 otherwise incorrect code will be generated for calls to those
12312 In addition, seriously incorrect code will result if you call a
12313 function with too many arguments. (Normally, extra arguments are
12314 harmlessly ignored.)
12316 @item -mregparm=@var{num}
12318 Control how many registers are used to pass integer arguments. By
12319 default, no registers are used to pass arguments, and at most 3
12320 registers can be used. You can control this behavior for a specific
12321 function by using the function attribute @samp{regparm}.
12322 @xref{Function Attributes}.
12324 @strong{Warning:} if you use this switch, and
12325 @var{num} is nonzero, then you must build all modules with the same
12326 value, including any libraries. This includes the system libraries and
12330 @opindex msseregparm
12331 Use SSE register passing conventions for float and double arguments
12332 and return values. You can control this behavior for a specific
12333 function by using the function attribute @samp{sseregparm}.
12334 @xref{Function Attributes}.
12336 @strong{Warning:} if you use this switch then you must build all
12337 modules with the same value, including any libraries. This includes
12338 the system libraries and startup modules.
12340 @item -mvect8-ret-in-mem
12341 @opindex mvect8-ret-in-mem
12342 Return 8-byte vectors in memory instead of MMX registers. This is the
12343 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12344 Studio compilers until version 12. Later compiler versions (starting
12345 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12346 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12347 you need to remain compatible with existing code produced by those
12348 previous compiler versions or older versions of GCC.
12357 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12358 is specified, the significands of results of floating-point operations are
12359 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12360 significands of results of floating-point operations to 53 bits (double
12361 precision) and @option{-mpc80} rounds the significands of results of
12362 floating-point operations to 64 bits (extended double precision), which is
12363 the default. When this option is used, floating-point operations in higher
12364 precisions are not available to the programmer without setting the FPU
12365 control word explicitly.
12367 Setting the rounding of floating-point operations to less than the default
12368 80 bits can speed some programs by 2% or more. Note that some mathematical
12369 libraries assume that extended precision (80 bit) floating-point operations
12370 are enabled by default; routines in such libraries could suffer significant
12371 loss of accuracy, typically through so-called "catastrophic cancellation",
12372 when this option is used to set the precision to less than extended precision.
12374 @item -mstackrealign
12375 @opindex mstackrealign
12376 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12377 option will generate an alternate prologue and epilogue that realigns the
12378 runtime stack if necessary. This supports mixing legacy codes that keep
12379 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12380 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12381 applicable to individual functions.
12383 @item -mpreferred-stack-boundary=@var{num}
12384 @opindex mpreferred-stack-boundary
12385 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12386 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12387 the default is 4 (16 bytes or 128 bits).
12389 @item -mincoming-stack-boundary=@var{num}
12390 @opindex mincoming-stack-boundary
12391 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12392 boundary. If @option{-mincoming-stack-boundary} is not specified,
12393 the one specified by @option{-mpreferred-stack-boundary} will be used.
12395 On Pentium and PentiumPro, @code{double} and @code{long double} values
12396 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12397 suffer significant run time performance penalties. On Pentium III, the
12398 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12399 properly if it is not 16 byte aligned.
12401 To ensure proper alignment of this values on the stack, the stack boundary
12402 must be as aligned as that required by any value stored on the stack.
12403 Further, every function must be generated such that it keeps the stack
12404 aligned. Thus calling a function compiled with a higher preferred
12405 stack boundary from a function compiled with a lower preferred stack
12406 boundary will most likely misalign the stack. It is recommended that
12407 libraries that use callbacks always use the default setting.
12409 This extra alignment does consume extra stack space, and generally
12410 increases code size. Code that is sensitive to stack space usage, such
12411 as embedded systems and operating system kernels, may want to reduce the
12412 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12439 @itemx -mno-fsgsbase
12469 These switches enable or disable the use of instructions in the MMX,
12470 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12471 F16C, SSE4A, FMA4, XOP, LWP, ABM, BMI, or 3DNow!@: extended instruction sets.
12472 These extensions are also available as built-in functions: see
12473 @ref{X86 Built-in Functions}, for details of the functions enabled and
12474 disabled by these switches.
12476 To have SSE/SSE2 instructions generated automatically from floating-point
12477 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12479 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12480 generates new AVX instructions or AVX equivalence for all SSEx instructions
12483 These options will enable GCC to use these extended instructions in
12484 generated code, even without @option{-mfpmath=sse}. Applications which
12485 perform runtime CPU detection must compile separate files for each
12486 supported architecture, using the appropriate flags. In particular,
12487 the file containing the CPU detection code should be compiled without
12491 @itemx -mno-fused-madd
12492 @opindex mfused-madd
12493 @opindex mno-fused-madd
12494 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12495 instructions. The default is to use these instructions.
12499 This option instructs GCC to emit a @code{cld} instruction in the prologue
12500 of functions that use string instructions. String instructions depend on
12501 the DF flag to select between autoincrement or autodecrement mode. While the
12502 ABI specifies the DF flag to be cleared on function entry, some operating
12503 systems violate this specification by not clearing the DF flag in their
12504 exception dispatchers. The exception handler can be invoked with the DF flag
12505 set which leads to wrong direction mode, when string instructions are used.
12506 This option can be enabled by default on 32-bit x86 targets by configuring
12507 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12508 instructions can be suppressed with the @option{-mno-cld} compiler option
12512 @opindex mvzeroupper
12513 This option instructs GCC to emit a @code{vzeroupper} instruction
12514 before a transfer of control flow out of the function to minimize
12515 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12520 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12521 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12522 data types. This is useful for high resolution counters that could be updated
12523 by multiple processors (or cores). This instruction is generated as part of
12524 atomic built-in functions: see @ref{Atomic Builtins} for details.
12528 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12529 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12530 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12531 SAHF are load and store instructions, respectively, for certain status flags.
12532 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12533 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12537 This option will enable GCC to use movbe instruction to implement
12538 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12542 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12543 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12544 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12548 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12549 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12550 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12551 variants) for single precision floating point arguments. These instructions
12552 are generated only when @option{-funsafe-math-optimizations} is enabled
12553 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12554 Note that while the throughput of the sequence is higher than the throughput
12555 of the non-reciprocal instruction, the precision of the sequence can be
12556 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12558 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12559 already with @option{-ffast-math} (or the above option combination), and
12560 doesn't need @option{-mrecip}.
12562 @item -mveclibabi=@var{type}
12563 @opindex mveclibabi
12564 Specifies the ABI type to use for vectorizing intrinsics using an
12565 external library. Supported types are @code{svml} for the Intel short
12566 vector math library and @code{acml} for the AMD math core library style
12567 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12568 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12569 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12570 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12571 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12572 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12573 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12574 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12575 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12576 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12577 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12578 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12579 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12580 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12581 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12582 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12583 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12584 compatible library will have to be specified at link time.
12586 @item -mabi=@var{name}
12588 Generate code for the specified calling convention. Permissible values
12589 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12590 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12591 ABI when targeting Windows. On all other systems, the default is the
12592 SYSV ABI. You can control this behavior for a specific function by
12593 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12594 @xref{Function Attributes}.
12597 @itemx -mno-push-args
12598 @opindex mpush-args
12599 @opindex mno-push-args
12600 Use PUSH operations to store outgoing parameters. This method is shorter
12601 and usually equally fast as method using SUB/MOV operations and is enabled
12602 by default. In some cases disabling it may improve performance because of
12603 improved scheduling and reduced dependencies.
12605 @item -maccumulate-outgoing-args
12606 @opindex maccumulate-outgoing-args
12607 If enabled, the maximum amount of space required for outgoing arguments will be
12608 computed in the function prologue. This is faster on most modern CPUs
12609 because of reduced dependencies, improved scheduling and reduced stack usage
12610 when preferred stack boundary is not equal to 2. The drawback is a notable
12611 increase in code size. This switch implies @option{-mno-push-args}.
12615 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12616 on thread-safe exception handling must compile and link all code with the
12617 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12618 @option{-D_MT}; when linking, it links in a special thread helper library
12619 @option{-lmingwthrd} which cleans up per thread exception handling data.
12621 @item -mno-align-stringops
12622 @opindex mno-align-stringops
12623 Do not align destination of inlined string operations. This switch reduces
12624 code size and improves performance in case the destination is already aligned,
12625 but GCC doesn't know about it.
12627 @item -minline-all-stringops
12628 @opindex minline-all-stringops
12629 By default GCC inlines string operations only when destination is known to be
12630 aligned at least to 4 byte boundary. This enables more inlining, increase code
12631 size, but may improve performance of code that depends on fast memcpy, strlen
12632 and memset for short lengths.
12634 @item -minline-stringops-dynamically
12635 @opindex minline-stringops-dynamically
12636 For string operation of unknown size, inline runtime checks so for small
12637 blocks inline code is used, while for large blocks library call is used.
12639 @item -mstringop-strategy=@var{alg}
12640 @opindex mstringop-strategy=@var{alg}
12641 Overwrite internal decision heuristic about particular algorithm to inline
12642 string operation with. The allowed values are @code{rep_byte},
12643 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12644 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12645 expanding inline loop, @code{libcall} for always expanding library call.
12647 @item -momit-leaf-frame-pointer
12648 @opindex momit-leaf-frame-pointer
12649 Don't keep the frame pointer in a register for leaf functions. This
12650 avoids the instructions to save, set up and restore frame pointers and
12651 makes an extra register available in leaf functions. The option
12652 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12653 which might make debugging harder.
12655 @item -mtls-direct-seg-refs
12656 @itemx -mno-tls-direct-seg-refs
12657 @opindex mtls-direct-seg-refs
12658 Controls whether TLS variables may be accessed with offsets from the
12659 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12660 or whether the thread base pointer must be added. Whether or not this
12661 is legal depends on the operating system, and whether it maps the
12662 segment to cover the entire TLS area.
12664 For systems that use GNU libc, the default is on.
12667 @itemx -mno-sse2avx
12669 Specify that the assembler should encode SSE instructions with VEX
12670 prefix. The option @option{-mavx} turns this on by default.
12675 If profiling is active @option{-pg} put the profiling
12676 counter call before prologue.
12677 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12678 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12681 @itemx -mno-8bit-idiv
12683 On some processors, like Intel Atom, 8bit unsigned integer divide is
12684 much faster than 32bit/64bit integer divide. This option will generate a
12685 runt-time check. If both dividend and divisor are within range of 0
12686 to 255, 8bit unsigned integer divide will be used instead of
12687 32bit/64bit integer divide.
12691 These @samp{-m} switches are supported in addition to the above
12692 on AMD x86-64 processors in 64-bit environments.
12699 Generate code for a 32-bit or 64-bit environment.
12700 The 32-bit environment sets int, long and pointer to 32 bits and
12701 generates code that runs on any i386 system.
12702 The 64-bit environment sets int to 32 bits and long and pointer
12703 to 64 bits and generates code for AMD's x86-64 architecture. For
12704 darwin only the -m64 option turns off the @option{-fno-pic} and
12705 @option{-mdynamic-no-pic} options.
12707 @item -mno-red-zone
12708 @opindex mno-red-zone
12709 Do not use a so called red zone for x86-64 code. The red zone is mandated
12710 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12711 stack pointer that will not be modified by signal or interrupt handlers
12712 and therefore can be used for temporary data without adjusting the stack
12713 pointer. The flag @option{-mno-red-zone} disables this red zone.
12715 @item -mcmodel=small
12716 @opindex mcmodel=small
12717 Generate code for the small code model: the program and its symbols must
12718 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12719 Programs can be statically or dynamically linked. This is the default
12722 @item -mcmodel=kernel
12723 @opindex mcmodel=kernel
12724 Generate code for the kernel code model. The kernel runs in the
12725 negative 2 GB of the address space.
12726 This model has to be used for Linux kernel code.
12728 @item -mcmodel=medium
12729 @opindex mcmodel=medium
12730 Generate code for the medium model: The program is linked in the lower 2
12731 GB of the address space. Small symbols are also placed there. Symbols
12732 with sizes larger than @option{-mlarge-data-threshold} are put into
12733 large data or bss sections and can be located above 2GB. Programs can
12734 be statically or dynamically linked.
12736 @item -mcmodel=large
12737 @opindex mcmodel=large
12738 Generate code for the large model: This model makes no assumptions
12739 about addresses and sizes of sections.
12742 @node IA-64 Options
12743 @subsection IA-64 Options
12744 @cindex IA-64 Options
12746 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12750 @opindex mbig-endian
12751 Generate code for a big endian target. This is the default for HP-UX@.
12753 @item -mlittle-endian
12754 @opindex mlittle-endian
12755 Generate code for a little endian target. This is the default for AIX5
12761 @opindex mno-gnu-as
12762 Generate (or don't) code for the GNU assembler. This is the default.
12763 @c Also, this is the default if the configure option @option{--with-gnu-as}
12769 @opindex mno-gnu-ld
12770 Generate (or don't) code for the GNU linker. This is the default.
12771 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12776 Generate code that does not use a global pointer register. The result
12777 is not position independent code, and violates the IA-64 ABI@.
12779 @item -mvolatile-asm-stop
12780 @itemx -mno-volatile-asm-stop
12781 @opindex mvolatile-asm-stop
12782 @opindex mno-volatile-asm-stop
12783 Generate (or don't) a stop bit immediately before and after volatile asm
12786 @item -mregister-names
12787 @itemx -mno-register-names
12788 @opindex mregister-names
12789 @opindex mno-register-names
12790 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12791 the stacked registers. This may make assembler output more readable.
12797 Disable (or enable) optimizations that use the small data section. This may
12798 be useful for working around optimizer bugs.
12800 @item -mconstant-gp
12801 @opindex mconstant-gp
12802 Generate code that uses a single constant global pointer value. This is
12803 useful when compiling kernel code.
12807 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12808 This is useful when compiling firmware code.
12810 @item -minline-float-divide-min-latency
12811 @opindex minline-float-divide-min-latency
12812 Generate code for inline divides of floating point values
12813 using the minimum latency algorithm.
12815 @item -minline-float-divide-max-throughput
12816 @opindex minline-float-divide-max-throughput
12817 Generate code for inline divides of floating point values
12818 using the maximum throughput algorithm.
12820 @item -mno-inline-float-divide
12821 @opindex mno-inline-float-divide
12822 Do not generate inline code for divides of floating point values.
12824 @item -minline-int-divide-min-latency
12825 @opindex minline-int-divide-min-latency
12826 Generate code for inline divides of integer values
12827 using the minimum latency algorithm.
12829 @item -minline-int-divide-max-throughput
12830 @opindex minline-int-divide-max-throughput
12831 Generate code for inline divides of integer values
12832 using the maximum throughput algorithm.
12834 @item -mno-inline-int-divide
12835 @opindex mno-inline-int-divide
12836 Do not generate inline code for divides of integer values.
12838 @item -minline-sqrt-min-latency
12839 @opindex minline-sqrt-min-latency
12840 Generate code for inline square roots
12841 using the minimum latency algorithm.
12843 @item -minline-sqrt-max-throughput
12844 @opindex minline-sqrt-max-throughput
12845 Generate code for inline square roots
12846 using the maximum throughput algorithm.
12848 @item -mno-inline-sqrt
12849 @opindex mno-inline-sqrt
12850 Do not generate inline code for sqrt.
12853 @itemx -mno-fused-madd
12854 @opindex mfused-madd
12855 @opindex mno-fused-madd
12856 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12857 instructions. The default is to use these instructions.
12859 @item -mno-dwarf2-asm
12860 @itemx -mdwarf2-asm
12861 @opindex mno-dwarf2-asm
12862 @opindex mdwarf2-asm
12863 Don't (or do) generate assembler code for the DWARF2 line number debugging
12864 info. This may be useful when not using the GNU assembler.
12866 @item -mearly-stop-bits
12867 @itemx -mno-early-stop-bits
12868 @opindex mearly-stop-bits
12869 @opindex mno-early-stop-bits
12870 Allow stop bits to be placed earlier than immediately preceding the
12871 instruction that triggered the stop bit. This can improve instruction
12872 scheduling, but does not always do so.
12874 @item -mfixed-range=@var{register-range}
12875 @opindex mfixed-range
12876 Generate code treating the given register range as fixed registers.
12877 A fixed register is one that the register allocator can not use. This is
12878 useful when compiling kernel code. A register range is specified as
12879 two registers separated by a dash. Multiple register ranges can be
12880 specified separated by a comma.
12882 @item -mtls-size=@var{tls-size}
12884 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12887 @item -mtune=@var{cpu-type}
12889 Tune the instruction scheduling for a particular CPU, Valid values are
12890 itanium, itanium1, merced, itanium2, and mckinley.
12896 Generate code for a 32-bit or 64-bit environment.
12897 The 32-bit environment sets int, long and pointer to 32 bits.
12898 The 64-bit environment sets int to 32 bits and long and pointer
12899 to 64 bits. These are HP-UX specific flags.
12901 @item -mno-sched-br-data-spec
12902 @itemx -msched-br-data-spec
12903 @opindex mno-sched-br-data-spec
12904 @opindex msched-br-data-spec
12905 (Dis/En)able data speculative scheduling before reload.
12906 This will result in generation of the ld.a instructions and
12907 the corresponding check instructions (ld.c / chk.a).
12908 The default is 'disable'.
12910 @item -msched-ar-data-spec
12911 @itemx -mno-sched-ar-data-spec
12912 @opindex msched-ar-data-spec
12913 @opindex mno-sched-ar-data-spec
12914 (En/Dis)able data speculative scheduling after reload.
12915 This will result in generation of the ld.a instructions and
12916 the corresponding check instructions (ld.c / chk.a).
12917 The default is 'enable'.
12919 @item -mno-sched-control-spec
12920 @itemx -msched-control-spec
12921 @opindex mno-sched-control-spec
12922 @opindex msched-control-spec
12923 (Dis/En)able control speculative scheduling. This feature is
12924 available only during region scheduling (i.e.@: before reload).
12925 This will result in generation of the ld.s instructions and
12926 the corresponding check instructions chk.s .
12927 The default is 'disable'.
12929 @item -msched-br-in-data-spec
12930 @itemx -mno-sched-br-in-data-spec
12931 @opindex msched-br-in-data-spec
12932 @opindex mno-sched-br-in-data-spec
12933 (En/Dis)able speculative scheduling of the instructions that
12934 are dependent on the data speculative loads before reload.
12935 This is effective only with @option{-msched-br-data-spec} enabled.
12936 The default is 'enable'.
12938 @item -msched-ar-in-data-spec
12939 @itemx -mno-sched-ar-in-data-spec
12940 @opindex msched-ar-in-data-spec
12941 @opindex mno-sched-ar-in-data-spec
12942 (En/Dis)able speculative scheduling of the instructions that
12943 are dependent on the data speculative loads after reload.
12944 This is effective only with @option{-msched-ar-data-spec} enabled.
12945 The default is 'enable'.
12947 @item -msched-in-control-spec
12948 @itemx -mno-sched-in-control-spec
12949 @opindex msched-in-control-spec
12950 @opindex mno-sched-in-control-spec
12951 (En/Dis)able speculative scheduling of the instructions that
12952 are dependent on the control speculative loads.
12953 This is effective only with @option{-msched-control-spec} enabled.
12954 The default is 'enable'.
12956 @item -mno-sched-prefer-non-data-spec-insns
12957 @itemx -msched-prefer-non-data-spec-insns
12958 @opindex mno-sched-prefer-non-data-spec-insns
12959 @opindex msched-prefer-non-data-spec-insns
12960 If enabled, data speculative instructions will be chosen for schedule
12961 only if there are no other choices at the moment. This will make
12962 the use of the data speculation much more conservative.
12963 The default is 'disable'.
12965 @item -mno-sched-prefer-non-control-spec-insns
12966 @itemx -msched-prefer-non-control-spec-insns
12967 @opindex mno-sched-prefer-non-control-spec-insns
12968 @opindex msched-prefer-non-control-spec-insns
12969 If enabled, control speculative instructions will be chosen for schedule
12970 only if there are no other choices at the moment. This will make
12971 the use of the control speculation much more conservative.
12972 The default is 'disable'.
12974 @item -mno-sched-count-spec-in-critical-path
12975 @itemx -msched-count-spec-in-critical-path
12976 @opindex mno-sched-count-spec-in-critical-path
12977 @opindex msched-count-spec-in-critical-path
12978 If enabled, speculative dependencies will be considered during
12979 computation of the instructions priorities. This will make the use of the
12980 speculation a bit more conservative.
12981 The default is 'disable'.
12983 @item -msched-spec-ldc
12984 @opindex msched-spec-ldc
12985 Use a simple data speculation check. This option is on by default.
12987 @item -msched-control-spec-ldc
12988 @opindex msched-spec-ldc
12989 Use a simple check for control speculation. This option is on by default.
12991 @item -msched-stop-bits-after-every-cycle
12992 @opindex msched-stop-bits-after-every-cycle
12993 Place a stop bit after every cycle when scheduling. This option is on
12996 @item -msched-fp-mem-deps-zero-cost
12997 @opindex msched-fp-mem-deps-zero-cost
12998 Assume that floating-point stores and loads are not likely to cause a conflict
12999 when placed into the same instruction group. This option is disabled by
13002 @item -msel-sched-dont-check-control-spec
13003 @opindex msel-sched-dont-check-control-spec
13004 Generate checks for control speculation in selective scheduling.
13005 This flag is disabled by default.
13007 @item -msched-max-memory-insns=@var{max-insns}
13008 @opindex msched-max-memory-insns
13009 Limit on the number of memory insns per instruction group, giving lower
13010 priority to subsequent memory insns attempting to schedule in the same
13011 instruction group. Frequently useful to prevent cache bank conflicts.
13012 The default value is 1.
13014 @item -msched-max-memory-insns-hard-limit
13015 @opindex msched-max-memory-insns-hard-limit
13016 Disallow more than `msched-max-memory-insns' in instruction group.
13017 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13018 when limit is reached but may still schedule memory operations.
13022 @node IA-64/VMS Options
13023 @subsection IA-64/VMS Options
13025 These @samp{-m} options are defined for the IA-64/VMS implementations:
13028 @item -mvms-return-codes
13029 @opindex mvms-return-codes
13030 Return VMS condition codes from main. The default is to return POSIX
13031 style condition (e.g.@ error) codes.
13033 @item -mdebug-main=@var{prefix}
13034 @opindex mdebug-main=@var{prefix}
13035 Flag the first routine whose name starts with @var{prefix} as the main
13036 routine for the debugger.
13040 Default to 64bit memory allocation routines.
13044 @subsection LM32 Options
13045 @cindex LM32 options
13047 These @option{-m} options are defined for the Lattice Mico32 architecture:
13050 @item -mbarrel-shift-enabled
13051 @opindex mbarrel-shift-enabled
13052 Enable barrel-shift instructions.
13054 @item -mdivide-enabled
13055 @opindex mdivide-enabled
13056 Enable divide and modulus instructions.
13058 @item -mmultiply-enabled
13059 @opindex multiply-enabled
13060 Enable multiply instructions.
13062 @item -msign-extend-enabled
13063 @opindex msign-extend-enabled
13064 Enable sign extend instructions.
13066 @item -muser-enabled
13067 @opindex muser-enabled
13068 Enable user-defined instructions.
13073 @subsection M32C Options
13074 @cindex M32C options
13077 @item -mcpu=@var{name}
13079 Select the CPU for which code is generated. @var{name} may be one of
13080 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13081 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13082 the M32C/80 series.
13086 Specifies that the program will be run on the simulator. This causes
13087 an alternate runtime library to be linked in which supports, for
13088 example, file I/O@. You must not use this option when generating
13089 programs that will run on real hardware; you must provide your own
13090 runtime library for whatever I/O functions are needed.
13092 @item -memregs=@var{number}
13094 Specifies the number of memory-based pseudo-registers GCC will use
13095 during code generation. These pseudo-registers will be used like real
13096 registers, so there is a tradeoff between GCC's ability to fit the
13097 code into available registers, and the performance penalty of using
13098 memory instead of registers. Note that all modules in a program must
13099 be compiled with the same value for this option. Because of that, you
13100 must not use this option with the default runtime libraries gcc
13105 @node M32R/D Options
13106 @subsection M32R/D Options
13107 @cindex M32R/D options
13109 These @option{-m} options are defined for Renesas M32R/D architectures:
13114 Generate code for the M32R/2@.
13118 Generate code for the M32R/X@.
13122 Generate code for the M32R@. This is the default.
13124 @item -mmodel=small
13125 @opindex mmodel=small
13126 Assume all objects live in the lower 16MB of memory (so that their addresses
13127 can be loaded with the @code{ld24} instruction), and assume all subroutines
13128 are reachable with the @code{bl} instruction.
13129 This is the default.
13131 The addressability of a particular object can be set with the
13132 @code{model} attribute.
13134 @item -mmodel=medium
13135 @opindex mmodel=medium
13136 Assume objects may be anywhere in the 32-bit address space (the compiler
13137 will generate @code{seth/add3} instructions to load their addresses), and
13138 assume all subroutines are reachable with the @code{bl} instruction.
13140 @item -mmodel=large
13141 @opindex mmodel=large
13142 Assume objects may be anywhere in the 32-bit address space (the compiler
13143 will generate @code{seth/add3} instructions to load their addresses), and
13144 assume subroutines may not be reachable with the @code{bl} instruction
13145 (the compiler will generate the much slower @code{seth/add3/jl}
13146 instruction sequence).
13149 @opindex msdata=none
13150 Disable use of the small data area. Variables will be put into
13151 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13152 @code{section} attribute has been specified).
13153 This is the default.
13155 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13156 Objects may be explicitly put in the small data area with the
13157 @code{section} attribute using one of these sections.
13159 @item -msdata=sdata
13160 @opindex msdata=sdata
13161 Put small global and static data in the small data area, but do not
13162 generate special code to reference them.
13165 @opindex msdata=use
13166 Put small global and static data in the small data area, and generate
13167 special instructions to reference them.
13171 @cindex smaller data references
13172 Put global and static objects less than or equal to @var{num} bytes
13173 into the small data or bss sections instead of the normal data or bss
13174 sections. The default value of @var{num} is 8.
13175 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13176 for this option to have any effect.
13178 All modules should be compiled with the same @option{-G @var{num}} value.
13179 Compiling with different values of @var{num} may or may not work; if it
13180 doesn't the linker will give an error message---incorrect code will not be
13185 Makes the M32R specific code in the compiler display some statistics
13186 that might help in debugging programs.
13188 @item -malign-loops
13189 @opindex malign-loops
13190 Align all loops to a 32-byte boundary.
13192 @item -mno-align-loops
13193 @opindex mno-align-loops
13194 Do not enforce a 32-byte alignment for loops. This is the default.
13196 @item -missue-rate=@var{number}
13197 @opindex missue-rate=@var{number}
13198 Issue @var{number} instructions per cycle. @var{number} can only be 1
13201 @item -mbranch-cost=@var{number}
13202 @opindex mbranch-cost=@var{number}
13203 @var{number} can only be 1 or 2. If it is 1 then branches will be
13204 preferred over conditional code, if it is 2, then the opposite will
13207 @item -mflush-trap=@var{number}
13208 @opindex mflush-trap=@var{number}
13209 Specifies the trap number to use to flush the cache. The default is
13210 12. Valid numbers are between 0 and 15 inclusive.
13212 @item -mno-flush-trap
13213 @opindex mno-flush-trap
13214 Specifies that the cache cannot be flushed by using a trap.
13216 @item -mflush-func=@var{name}
13217 @opindex mflush-func=@var{name}
13218 Specifies the name of the operating system function to call to flush
13219 the cache. The default is @emph{_flush_cache}, but a function call
13220 will only be used if a trap is not available.
13222 @item -mno-flush-func
13223 @opindex mno-flush-func
13224 Indicates that there is no OS function for flushing the cache.
13228 @node M680x0 Options
13229 @subsection M680x0 Options
13230 @cindex M680x0 options
13232 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13233 The default settings depend on which architecture was selected when
13234 the compiler was configured; the defaults for the most common choices
13238 @item -march=@var{arch}
13240 Generate code for a specific M680x0 or ColdFire instruction set
13241 architecture. Permissible values of @var{arch} for M680x0
13242 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13243 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13244 architectures are selected according to Freescale's ISA classification
13245 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13246 @samp{isab} and @samp{isac}.
13248 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13249 code for a ColdFire target. The @var{arch} in this macro is one of the
13250 @option{-march} arguments given above.
13252 When used together, @option{-march} and @option{-mtune} select code
13253 that runs on a family of similar processors but that is optimized
13254 for a particular microarchitecture.
13256 @item -mcpu=@var{cpu}
13258 Generate code for a specific M680x0 or ColdFire processor.
13259 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13260 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13261 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13262 below, which also classifies the CPUs into families:
13264 @multitable @columnfractions 0.20 0.80
13265 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13266 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13267 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13268 @item @samp{5206e} @tab @samp{5206e}
13269 @item @samp{5208} @tab @samp{5207} @samp{5208}
13270 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13271 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13272 @item @samp{5216} @tab @samp{5214} @samp{5216}
13273 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13274 @item @samp{5225} @tab @samp{5224} @samp{5225}
13275 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13276 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13277 @item @samp{5249} @tab @samp{5249}
13278 @item @samp{5250} @tab @samp{5250}
13279 @item @samp{5271} @tab @samp{5270} @samp{5271}
13280 @item @samp{5272} @tab @samp{5272}
13281 @item @samp{5275} @tab @samp{5274} @samp{5275}
13282 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13283 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13284 @item @samp{5307} @tab @samp{5307}
13285 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13286 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13287 @item @samp{5407} @tab @samp{5407}
13288 @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}
13291 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13292 @var{arch} is compatible with @var{cpu}. Other combinations of
13293 @option{-mcpu} and @option{-march} are rejected.
13295 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13296 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13297 where the value of @var{family} is given by the table above.
13299 @item -mtune=@var{tune}
13301 Tune the code for a particular microarchitecture, within the
13302 constraints set by @option{-march} and @option{-mcpu}.
13303 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13304 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13305 and @samp{cpu32}. The ColdFire microarchitectures
13306 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13308 You can also use @option{-mtune=68020-40} for code that needs
13309 to run relatively well on 68020, 68030 and 68040 targets.
13310 @option{-mtune=68020-60} is similar but includes 68060 targets
13311 as well. These two options select the same tuning decisions as
13312 @option{-m68020-40} and @option{-m68020-60} respectively.
13314 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13315 when tuning for 680x0 architecture @var{arch}. It also defines
13316 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13317 option is used. If gcc is tuning for a range of architectures,
13318 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13319 it defines the macros for every architecture in the range.
13321 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13322 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13323 of the arguments given above.
13329 Generate output for a 68000. This is the default
13330 when the compiler is configured for 68000-based systems.
13331 It is equivalent to @option{-march=68000}.
13333 Use this option for microcontrollers with a 68000 or EC000 core,
13334 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13338 Generate output for a 68010. This is the default
13339 when the compiler is configured for 68010-based systems.
13340 It is equivalent to @option{-march=68010}.
13346 Generate output for a 68020. This is the default
13347 when the compiler is configured for 68020-based systems.
13348 It is equivalent to @option{-march=68020}.
13352 Generate output for a 68030. This is the default when the compiler is
13353 configured for 68030-based systems. It is equivalent to
13354 @option{-march=68030}.
13358 Generate output for a 68040. This is the default when the compiler is
13359 configured for 68040-based systems. It is equivalent to
13360 @option{-march=68040}.
13362 This option inhibits the use of 68881/68882 instructions that have to be
13363 emulated by software on the 68040. Use this option if your 68040 does not
13364 have code to emulate those instructions.
13368 Generate output for a 68060. This is the default when the compiler is
13369 configured for 68060-based systems. It is equivalent to
13370 @option{-march=68060}.
13372 This option inhibits the use of 68020 and 68881/68882 instructions that
13373 have to be emulated by software on the 68060. Use this option if your 68060
13374 does not have code to emulate those instructions.
13378 Generate output for a CPU32. This is the default
13379 when the compiler is configured for CPU32-based systems.
13380 It is equivalent to @option{-march=cpu32}.
13382 Use this option for microcontrollers with a
13383 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13384 68336, 68340, 68341, 68349 and 68360.
13388 Generate output for a 520X ColdFire CPU@. This is the default
13389 when the compiler is configured for 520X-based systems.
13390 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13391 in favor of that option.
13393 Use this option for microcontroller with a 5200 core, including
13394 the MCF5202, MCF5203, MCF5204 and MCF5206.
13398 Generate output for a 5206e ColdFire CPU@. The option is now
13399 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13403 Generate output for a member of the ColdFire 528X family.
13404 The option is now deprecated in favor of the equivalent
13405 @option{-mcpu=528x}.
13409 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13410 in favor of the equivalent @option{-mcpu=5307}.
13414 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13415 in favor of the equivalent @option{-mcpu=5407}.
13419 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13420 This includes use of hardware floating point instructions.
13421 The option is equivalent to @option{-mcpu=547x}, and is now
13422 deprecated in favor of that option.
13426 Generate output for a 68040, without using any of the new instructions.
13427 This results in code which can run relatively efficiently on either a
13428 68020/68881 or a 68030 or a 68040. The generated code does use the
13429 68881 instructions that are emulated on the 68040.
13431 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13435 Generate output for a 68060, without using any of the new instructions.
13436 This results in code which can run relatively efficiently on either a
13437 68020/68881 or a 68030 or a 68040. The generated code does use the
13438 68881 instructions that are emulated on the 68060.
13440 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13444 @opindex mhard-float
13446 Generate floating-point instructions. This is the default for 68020
13447 and above, and for ColdFire devices that have an FPU@. It defines the
13448 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13449 on ColdFire targets.
13452 @opindex msoft-float
13453 Do not generate floating-point instructions; use library calls instead.
13454 This is the default for 68000, 68010, and 68832 targets. It is also
13455 the default for ColdFire devices that have no FPU.
13461 Generate (do not generate) ColdFire hardware divide and remainder
13462 instructions. If @option{-march} is used without @option{-mcpu},
13463 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13464 architectures. Otherwise, the default is taken from the target CPU
13465 (either the default CPU, or the one specified by @option{-mcpu}). For
13466 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13467 @option{-mcpu=5206e}.
13469 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13473 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13474 Additionally, parameters passed on the stack are also aligned to a
13475 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13479 Do not consider type @code{int} to be 16 bits wide. This is the default.
13482 @itemx -mno-bitfield
13483 @opindex mnobitfield
13484 @opindex mno-bitfield
13485 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13486 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13490 Do use the bit-field instructions. The @option{-m68020} option implies
13491 @option{-mbitfield}. This is the default if you use a configuration
13492 designed for a 68020.
13496 Use a different function-calling convention, in which functions
13497 that take a fixed number of arguments return with the @code{rtd}
13498 instruction, which pops their arguments while returning. This
13499 saves one instruction in the caller since there is no need to pop
13500 the arguments there.
13502 This calling convention is incompatible with the one normally
13503 used on Unix, so you cannot use it if you need to call libraries
13504 compiled with the Unix compiler.
13506 Also, you must provide function prototypes for all functions that
13507 take variable numbers of arguments (including @code{printf});
13508 otherwise incorrect code will be generated for calls to those
13511 In addition, seriously incorrect code will result if you call a
13512 function with too many arguments. (Normally, extra arguments are
13513 harmlessly ignored.)
13515 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13516 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13520 Do not use the calling conventions selected by @option{-mrtd}.
13521 This is the default.
13524 @itemx -mno-align-int
13525 @opindex malign-int
13526 @opindex mno-align-int
13527 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13528 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13529 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13530 Aligning variables on 32-bit boundaries produces code that runs somewhat
13531 faster on processors with 32-bit busses at the expense of more memory.
13533 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13534 align structures containing the above types differently than
13535 most published application binary interface specifications for the m68k.
13539 Use the pc-relative addressing mode of the 68000 directly, instead of
13540 using a global offset table. At present, this option implies @option{-fpic},
13541 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13542 not presently supported with @option{-mpcrel}, though this could be supported for
13543 68020 and higher processors.
13545 @item -mno-strict-align
13546 @itemx -mstrict-align
13547 @opindex mno-strict-align
13548 @opindex mstrict-align
13549 Do not (do) assume that unaligned memory references will be handled by
13553 Generate code that allows the data segment to be located in a different
13554 area of memory from the text segment. This allows for execute in place in
13555 an environment without virtual memory management. This option implies
13558 @item -mno-sep-data
13559 Generate code that assumes that the data segment follows the text segment.
13560 This is the default.
13562 @item -mid-shared-library
13563 Generate code that supports shared libraries via the library ID method.
13564 This allows for execute in place and shared libraries in an environment
13565 without virtual memory management. This option implies @option{-fPIC}.
13567 @item -mno-id-shared-library
13568 Generate code that doesn't assume ID based shared libraries are being used.
13569 This is the default.
13571 @item -mshared-library-id=n
13572 Specified the identification number of the ID based shared library being
13573 compiled. Specifying a value of 0 will generate more compact code, specifying
13574 other values will force the allocation of that number to the current
13575 library but is no more space or time efficient than omitting this option.
13581 When generating position-independent code for ColdFire, generate code
13582 that works if the GOT has more than 8192 entries. This code is
13583 larger and slower than code generated without this option. On M680x0
13584 processors, this option is not needed; @option{-fPIC} suffices.
13586 GCC normally uses a single instruction to load values from the GOT@.
13587 While this is relatively efficient, it only works if the GOT
13588 is smaller than about 64k. Anything larger causes the linker
13589 to report an error such as:
13591 @cindex relocation truncated to fit (ColdFire)
13593 relocation truncated to fit: R_68K_GOT16O foobar
13596 If this happens, you should recompile your code with @option{-mxgot}.
13597 It should then work with very large GOTs. However, code generated with
13598 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13599 the value of a global symbol.
13601 Note that some linkers, including newer versions of the GNU linker,
13602 can create multiple GOTs and sort GOT entries. If you have such a linker,
13603 you should only need to use @option{-mxgot} when compiling a single
13604 object file that accesses more than 8192 GOT entries. Very few do.
13606 These options have no effect unless GCC is generating
13607 position-independent code.
13611 @node MCore Options
13612 @subsection MCore Options
13613 @cindex MCore options
13615 These are the @samp{-m} options defined for the Motorola M*Core
13621 @itemx -mno-hardlit
13623 @opindex mno-hardlit
13624 Inline constants into the code stream if it can be done in two
13625 instructions or less.
13631 Use the divide instruction. (Enabled by default).
13633 @item -mrelax-immediate
13634 @itemx -mno-relax-immediate
13635 @opindex mrelax-immediate
13636 @opindex mno-relax-immediate
13637 Allow arbitrary sized immediates in bit operations.
13639 @item -mwide-bitfields
13640 @itemx -mno-wide-bitfields
13641 @opindex mwide-bitfields
13642 @opindex mno-wide-bitfields
13643 Always treat bit-fields as int-sized.
13645 @item -m4byte-functions
13646 @itemx -mno-4byte-functions
13647 @opindex m4byte-functions
13648 @opindex mno-4byte-functions
13649 Force all functions to be aligned to a four byte boundary.
13651 @item -mcallgraph-data
13652 @itemx -mno-callgraph-data
13653 @opindex mcallgraph-data
13654 @opindex mno-callgraph-data
13655 Emit callgraph information.
13658 @itemx -mno-slow-bytes
13659 @opindex mslow-bytes
13660 @opindex mno-slow-bytes
13661 Prefer word access when reading byte quantities.
13663 @item -mlittle-endian
13664 @itemx -mbig-endian
13665 @opindex mlittle-endian
13666 @opindex mbig-endian
13667 Generate code for a little endian target.
13673 Generate code for the 210 processor.
13677 Assume that run-time support has been provided and so omit the
13678 simulator library (@file{libsim.a)} from the linker command line.
13680 @item -mstack-increment=@var{size}
13681 @opindex mstack-increment
13682 Set the maximum amount for a single stack increment operation. Large
13683 values can increase the speed of programs which contain functions
13684 that need a large amount of stack space, but they can also trigger a
13685 segmentation fault if the stack is extended too much. The default
13691 @subsection MeP Options
13692 @cindex MeP options
13698 Enables the @code{abs} instruction, which is the absolute difference
13699 between two registers.
13703 Enables all the optional instructions - average, multiply, divide, bit
13704 operations, leading zero, absolute difference, min/max, clip, and
13710 Enables the @code{ave} instruction, which computes the average of two
13713 @item -mbased=@var{n}
13715 Variables of size @var{n} bytes or smaller will be placed in the
13716 @code{.based} section by default. Based variables use the @code{$tp}
13717 register as a base register, and there is a 128 byte limit to the
13718 @code{.based} section.
13722 Enables the bit operation instructions - bit test (@code{btstm}), set
13723 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13724 test-and-set (@code{tas}).
13726 @item -mc=@var{name}
13728 Selects which section constant data will be placed in. @var{name} may
13729 be @code{tiny}, @code{near}, or @code{far}.
13733 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13734 useful unless you also provide @code{-mminmax}.
13736 @item -mconfig=@var{name}
13738 Selects one of the build-in core configurations. Each MeP chip has
13739 one or more modules in it; each module has a core CPU and a variety of
13740 coprocessors, optional instructions, and peripherals. The
13741 @code{MeP-Integrator} tool, not part of GCC, provides these
13742 configurations through this option; using this option is the same as
13743 using all the corresponding command line options. The default
13744 configuration is @code{default}.
13748 Enables the coprocessor instructions. By default, this is a 32-bit
13749 coprocessor. Note that the coprocessor is normally enabled via the
13750 @code{-mconfig=} option.
13754 Enables the 32-bit coprocessor's instructions.
13758 Enables the 64-bit coprocessor's instructions.
13762 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13766 Causes constant variables to be placed in the @code{.near} section.
13770 Enables the @code{div} and @code{divu} instructions.
13774 Generate big-endian code.
13778 Generate little-endian code.
13780 @item -mio-volatile
13781 @opindex mio-volatile
13782 Tells the compiler that any variable marked with the @code{io}
13783 attribute is to be considered volatile.
13787 Causes variables to be assigned to the @code{.far} section by default.
13791 Enables the @code{leadz} (leading zero) instruction.
13795 Causes variables to be assigned to the @code{.near} section by default.
13799 Enables the @code{min} and @code{max} instructions.
13803 Enables the multiplication and multiply-accumulate instructions.
13807 Disables all the optional instructions enabled by @code{-mall-opts}.
13811 Enables the @code{repeat} and @code{erepeat} instructions, used for
13812 low-overhead looping.
13816 Causes all variables to default to the @code{.tiny} section. Note
13817 that there is a 65536 byte limit to this section. Accesses to these
13818 variables use the @code{%gp} base register.
13822 Enables the saturation instructions. Note that the compiler does not
13823 currently generate these itself, but this option is included for
13824 compatibility with other tools, like @code{as}.
13828 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13832 Link the simulator runtime libraries.
13836 Link the simulator runtime libraries, excluding built-in support
13837 for reset and exception vectors and tables.
13841 Causes all functions to default to the @code{.far} section. Without
13842 this option, functions default to the @code{.near} section.
13844 @item -mtiny=@var{n}
13846 Variables that are @var{n} bytes or smaller will be allocated to the
13847 @code{.tiny} section. These variables use the @code{$gp} base
13848 register. The default for this option is 4, but note that there's a
13849 65536 byte limit to the @code{.tiny} section.
13853 @node MicroBlaze Options
13854 @subsection MicroBlaze Options
13855 @cindex MicroBlaze Options
13860 @opindex msoft-float
13861 Use software emulation for floating point (default).
13864 @opindex mhard-float
13865 Use hardware floating point instructions.
13869 Do not optimize block moves, use @code{memcpy}.
13871 @item -mno-clearbss
13872 @opindex mno-clearbss
13873 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
13875 @item -mcpu=@var{cpu-type}
13877 Use features of and schedule code for given CPU.
13878 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
13879 where @var{X} is a major version, @var{YY} is the minor version, and
13880 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
13881 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
13883 @item -mxl-soft-mul
13884 @opindex mxl-soft-mul
13885 Use software multiply emulation (default).
13887 @item -mxl-soft-div
13888 @opindex mxl-soft-div
13889 Use software emulation for divides (default).
13891 @item -mxl-barrel-shift
13892 @opindex mxl-barrel-shift
13893 Use the hardware barrel shifter.
13895 @item -mxl-pattern-compare
13896 @opindex mxl-pattern-compare
13897 Use pattern compare instructions.
13899 @item -msmall-divides
13900 @opindex msmall-divides
13901 Use table lookup optimization for small signed integer divisions.
13903 @item -mxl-stack-check
13904 @opindex mxl-stack-check
13905 This option is deprecated. Use -fstack-check instead.
13908 @opindex mxl-gp-opt
13909 Use GP relative sdata/sbss sections.
13911 @item -mxl-multiply-high
13912 @opindex mxl-multiply-high
13913 Use multiply high instructions for high part of 32x32 multiply.
13915 @item -mxl-float-convert
13916 @opindex mxl-float-convert
13917 Use hardware floating point conversion instructions.
13919 @item -mxl-float-sqrt
13920 @opindex mxl-float-sqrt
13921 Use hardware floating point square root instruction.
13923 @item -mxl-mode-@var{app-model}
13924 Select application model @var{app-model}. Valid models are
13927 normal executable (default), uses startup code @file{crt0.o}.
13930 for use with Xilinx Microprocessor Debugger (XMD) based
13931 software intrusive debug agent called xmdstub. This uses startup file
13932 @file{crt1.o} and sets the start address of the program to be 0x800.
13935 for applications that are loaded using a bootloader.
13936 This model uses startup file @file{crt2.o} which does not contain a processor
13937 reset vector handler. This is suitable for transferring control on a
13938 processor reset to the bootloader rather than the application.
13941 for applications that do not require any of the
13942 MicroBlaze vectors. This option may be useful for applications running
13943 within a monitoring application. This model uses @file{crt3.o} as a startup file.
13946 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
13947 @option{-mxl-mode-@var{app-model}}.
13952 @subsection MIPS Options
13953 @cindex MIPS options
13959 Generate big-endian code.
13963 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13966 @item -march=@var{arch}
13968 Generate code that will run on @var{arch}, which can be the name of a
13969 generic MIPS ISA, or the name of a particular processor.
13971 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13972 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13973 The processor names are:
13974 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13975 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13976 @samp{5kc}, @samp{5kf},
13978 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13979 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13980 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13981 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13982 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13983 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
13987 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13988 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13989 @samp{rm7000}, @samp{rm9000},
13990 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13993 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13994 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13996 The special value @samp{from-abi} selects the
13997 most compatible architecture for the selected ABI (that is,
13998 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14000 Native Linux/GNU toolchains also support the value @samp{native},
14001 which selects the best architecture option for the host processor.
14002 @option{-march=native} has no effect if GCC does not recognize
14005 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14006 (for example, @samp{-march=r2k}). Prefixes are optional, and
14007 @samp{vr} may be written @samp{r}.
14009 Names of the form @samp{@var{n}f2_1} refer to processors with
14010 FPUs clocked at half the rate of the core, names of the form
14011 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14012 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14013 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14014 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14015 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14016 accepted as synonyms for @samp{@var{n}f1_1}.
14018 GCC defines two macros based on the value of this option. The first
14019 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14020 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14021 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14022 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14023 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14025 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14026 above. In other words, it will have the full prefix and will not
14027 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14028 the macro names the resolved architecture (either @samp{"mips1"} or
14029 @samp{"mips3"}). It names the default architecture when no
14030 @option{-march} option is given.
14032 @item -mtune=@var{arch}
14034 Optimize for @var{arch}. Among other things, this option controls
14035 the way instructions are scheduled, and the perceived cost of arithmetic
14036 operations. The list of @var{arch} values is the same as for
14039 When this option is not used, GCC will optimize for the processor
14040 specified by @option{-march}. By using @option{-march} and
14041 @option{-mtune} together, it is possible to generate code that will
14042 run on a family of processors, but optimize the code for one
14043 particular member of that family.
14045 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14046 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14047 @samp{-march} ones described above.
14051 Equivalent to @samp{-march=mips1}.
14055 Equivalent to @samp{-march=mips2}.
14059 Equivalent to @samp{-march=mips3}.
14063 Equivalent to @samp{-march=mips4}.
14067 Equivalent to @samp{-march=mips32}.
14071 Equivalent to @samp{-march=mips32r2}.
14075 Equivalent to @samp{-march=mips64}.
14079 Equivalent to @samp{-march=mips64r2}.
14084 @opindex mno-mips16
14085 Generate (do not generate) MIPS16 code. If GCC is targetting a
14086 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14088 MIPS16 code generation can also be controlled on a per-function basis
14089 by means of @code{mips16} and @code{nomips16} attributes.
14090 @xref{Function Attributes}, for more information.
14092 @item -mflip-mips16
14093 @opindex mflip-mips16
14094 Generate MIPS16 code on alternating functions. This option is provided
14095 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14096 not intended for ordinary use in compiling user code.
14098 @item -minterlink-mips16
14099 @itemx -mno-interlink-mips16
14100 @opindex minterlink-mips16
14101 @opindex mno-interlink-mips16
14102 Require (do not require) that non-MIPS16 code be link-compatible with
14105 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14106 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14107 therefore disables direct jumps unless GCC knows that the target of the
14108 jump is not MIPS16.
14120 Generate code for the given ABI@.
14122 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14123 generates 64-bit code when you select a 64-bit architecture, but you
14124 can use @option{-mgp32} to get 32-bit code instead.
14126 For information about the O64 ABI, see
14127 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14129 GCC supports a variant of the o32 ABI in which floating-point registers
14130 are 64 rather than 32 bits wide. You can select this combination with
14131 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14132 and @samp{mfhc1} instructions and is therefore only supported for
14133 MIPS32R2 processors.
14135 The register assignments for arguments and return values remain the
14136 same, but each scalar value is passed in a single 64-bit register
14137 rather than a pair of 32-bit registers. For example, scalar
14138 floating-point values are returned in @samp{$f0} only, not a
14139 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14140 remains the same, but all 64 bits are saved.
14143 @itemx -mno-abicalls
14145 @opindex mno-abicalls
14146 Generate (do not generate) code that is suitable for SVR4-style
14147 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14152 Generate (do not generate) code that is fully position-independent,
14153 and that can therefore be linked into shared libraries. This option
14154 only affects @option{-mabicalls}.
14156 All @option{-mabicalls} code has traditionally been position-independent,
14157 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14158 as an extension, the GNU toolchain allows executables to use absolute
14159 accesses for locally-binding symbols. It can also use shorter GP
14160 initialization sequences and generate direct calls to locally-defined
14161 functions. This mode is selected by @option{-mno-shared}.
14163 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14164 objects that can only be linked by the GNU linker. However, the option
14165 does not affect the ABI of the final executable; it only affects the ABI
14166 of relocatable objects. Using @option{-mno-shared} will generally make
14167 executables both smaller and quicker.
14169 @option{-mshared} is the default.
14175 Assume (do not assume) that the static and dynamic linkers
14176 support PLTs and copy relocations. This option only affects
14177 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14178 has no effect without @samp{-msym32}.
14180 You can make @option{-mplt} the default by configuring
14181 GCC with @option{--with-mips-plt}. The default is
14182 @option{-mno-plt} otherwise.
14188 Lift (do not lift) the usual restrictions on the size of the global
14191 GCC normally uses a single instruction to load values from the GOT@.
14192 While this is relatively efficient, it will only work if the GOT
14193 is smaller than about 64k. Anything larger will cause the linker
14194 to report an error such as:
14196 @cindex relocation truncated to fit (MIPS)
14198 relocation truncated to fit: R_MIPS_GOT16 foobar
14201 If this happens, you should recompile your code with @option{-mxgot}.
14202 It should then work with very large GOTs, although it will also be
14203 less efficient, since it will take three instructions to fetch the
14204 value of a global symbol.
14206 Note that some linkers can create multiple GOTs. If you have such a
14207 linker, you should only need to use @option{-mxgot} when a single object
14208 file accesses more than 64k's worth of GOT entries. Very few do.
14210 These options have no effect unless GCC is generating position
14215 Assume that general-purpose registers are 32 bits wide.
14219 Assume that general-purpose registers are 64 bits wide.
14223 Assume that floating-point registers are 32 bits wide.
14227 Assume that floating-point registers are 64 bits wide.
14230 @opindex mhard-float
14231 Use floating-point coprocessor instructions.
14234 @opindex msoft-float
14235 Do not use floating-point coprocessor instructions. Implement
14236 floating-point calculations using library calls instead.
14238 @item -msingle-float
14239 @opindex msingle-float
14240 Assume that the floating-point coprocessor only supports single-precision
14243 @item -mdouble-float
14244 @opindex mdouble-float
14245 Assume that the floating-point coprocessor supports double-precision
14246 operations. This is the default.
14252 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14253 implement atomic memory built-in functions. When neither option is
14254 specified, GCC will use the instructions if the target architecture
14257 @option{-mllsc} is useful if the runtime environment can emulate the
14258 instructions and @option{-mno-llsc} can be useful when compiling for
14259 nonstandard ISAs. You can make either option the default by
14260 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14261 respectively. @option{--with-llsc} is the default for some
14262 configurations; see the installation documentation for details.
14268 Use (do not use) revision 1 of the MIPS DSP ASE@.
14269 @xref{MIPS DSP Built-in Functions}. This option defines the
14270 preprocessor macro @samp{__mips_dsp}. It also defines
14271 @samp{__mips_dsp_rev} to 1.
14277 Use (do not use) revision 2 of the MIPS DSP ASE@.
14278 @xref{MIPS DSP Built-in Functions}. This option defines the
14279 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14280 It also defines @samp{__mips_dsp_rev} to 2.
14283 @itemx -mno-smartmips
14284 @opindex msmartmips
14285 @opindex mno-smartmips
14286 Use (do not use) the MIPS SmartMIPS ASE.
14288 @item -mpaired-single
14289 @itemx -mno-paired-single
14290 @opindex mpaired-single
14291 @opindex mno-paired-single
14292 Use (do not use) paired-single floating-point instructions.
14293 @xref{MIPS Paired-Single Support}. This option requires
14294 hardware floating-point support to be enabled.
14300 Use (do not use) MIPS Digital Media Extension instructions.
14301 This option can only be used when generating 64-bit code and requires
14302 hardware floating-point support to be enabled.
14307 @opindex mno-mips3d
14308 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14309 The option @option{-mips3d} implies @option{-mpaired-single}.
14315 Use (do not use) MT Multithreading instructions.
14319 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14320 an explanation of the default and the way that the pointer size is
14325 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14327 The default size of @code{int}s, @code{long}s and pointers depends on
14328 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14329 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14330 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14331 or the same size as integer registers, whichever is smaller.
14337 Assume (do not assume) that all symbols have 32-bit values, regardless
14338 of the selected ABI@. This option is useful in combination with
14339 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14340 to generate shorter and faster references to symbolic addresses.
14344 Put definitions of externally-visible data in a small data section
14345 if that data is no bigger than @var{num} bytes. GCC can then access
14346 the data more efficiently; see @option{-mgpopt} for details.
14348 The default @option{-G} option depends on the configuration.
14350 @item -mlocal-sdata
14351 @itemx -mno-local-sdata
14352 @opindex mlocal-sdata
14353 @opindex mno-local-sdata
14354 Extend (do not extend) the @option{-G} behavior to local data too,
14355 such as to static variables in C@. @option{-mlocal-sdata} is the
14356 default for all configurations.
14358 If the linker complains that an application is using too much small data,
14359 you might want to try rebuilding the less performance-critical parts with
14360 @option{-mno-local-sdata}. You might also want to build large
14361 libraries with @option{-mno-local-sdata}, so that the libraries leave
14362 more room for the main program.
14364 @item -mextern-sdata
14365 @itemx -mno-extern-sdata
14366 @opindex mextern-sdata
14367 @opindex mno-extern-sdata
14368 Assume (do not assume) that externally-defined data will be in
14369 a small data section if that data is within the @option{-G} limit.
14370 @option{-mextern-sdata} is the default for all configurations.
14372 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14373 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14374 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14375 is placed in a small data section. If @var{Var} is defined by another
14376 module, you must either compile that module with a high-enough
14377 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14378 definition. If @var{Var} is common, you must link the application
14379 with a high-enough @option{-G} setting.
14381 The easiest way of satisfying these restrictions is to compile
14382 and link every module with the same @option{-G} option. However,
14383 you may wish to build a library that supports several different
14384 small data limits. You can do this by compiling the library with
14385 the highest supported @option{-G} setting and additionally using
14386 @option{-mno-extern-sdata} to stop the library from making assumptions
14387 about externally-defined data.
14393 Use (do not use) GP-relative accesses for symbols that are known to be
14394 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14395 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14398 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14399 might not hold the value of @code{_gp}. For example, if the code is
14400 part of a library that might be used in a boot monitor, programs that
14401 call boot monitor routines will pass an unknown value in @code{$gp}.
14402 (In such situations, the boot monitor itself would usually be compiled
14403 with @option{-G0}.)
14405 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14406 @option{-mno-extern-sdata}.
14408 @item -membedded-data
14409 @itemx -mno-embedded-data
14410 @opindex membedded-data
14411 @opindex mno-embedded-data
14412 Allocate variables to the read-only data section first if possible, then
14413 next in the small data section if possible, otherwise in data. This gives
14414 slightly slower code than the default, but reduces the amount of RAM required
14415 when executing, and thus may be preferred for some embedded systems.
14417 @item -muninit-const-in-rodata
14418 @itemx -mno-uninit-const-in-rodata
14419 @opindex muninit-const-in-rodata
14420 @opindex mno-uninit-const-in-rodata
14421 Put uninitialized @code{const} variables in the read-only data section.
14422 This option is only meaningful in conjunction with @option{-membedded-data}.
14424 @item -mcode-readable=@var{setting}
14425 @opindex mcode-readable
14426 Specify whether GCC may generate code that reads from executable sections.
14427 There are three possible settings:
14430 @item -mcode-readable=yes
14431 Instructions may freely access executable sections. This is the
14434 @item -mcode-readable=pcrel
14435 MIPS16 PC-relative load instructions can access executable sections,
14436 but other instructions must not do so. This option is useful on 4KSc
14437 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14438 It is also useful on processors that can be configured to have a dual
14439 instruction/data SRAM interface and that, like the M4K, automatically
14440 redirect PC-relative loads to the instruction RAM.
14442 @item -mcode-readable=no
14443 Instructions must not access executable sections. This option can be
14444 useful on targets that are configured to have a dual instruction/data
14445 SRAM interface but that (unlike the M4K) do not automatically redirect
14446 PC-relative loads to the instruction RAM.
14449 @item -msplit-addresses
14450 @itemx -mno-split-addresses
14451 @opindex msplit-addresses
14452 @opindex mno-split-addresses
14453 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14454 relocation operators. This option has been superseded by
14455 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14457 @item -mexplicit-relocs
14458 @itemx -mno-explicit-relocs
14459 @opindex mexplicit-relocs
14460 @opindex mno-explicit-relocs
14461 Use (do not use) assembler relocation operators when dealing with symbolic
14462 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14463 is to use assembler macros instead.
14465 @option{-mexplicit-relocs} is the default if GCC was configured
14466 to use an assembler that supports relocation operators.
14468 @item -mcheck-zero-division
14469 @itemx -mno-check-zero-division
14470 @opindex mcheck-zero-division
14471 @opindex mno-check-zero-division
14472 Trap (do not trap) on integer division by zero.
14474 The default is @option{-mcheck-zero-division}.
14476 @item -mdivide-traps
14477 @itemx -mdivide-breaks
14478 @opindex mdivide-traps
14479 @opindex mdivide-breaks
14480 MIPS systems check for division by zero by generating either a
14481 conditional trap or a break instruction. Using traps results in
14482 smaller code, but is only supported on MIPS II and later. Also, some
14483 versions of the Linux kernel have a bug that prevents trap from
14484 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14485 allow conditional traps on architectures that support them and
14486 @option{-mdivide-breaks} to force the use of breaks.
14488 The default is usually @option{-mdivide-traps}, but this can be
14489 overridden at configure time using @option{--with-divide=breaks}.
14490 Divide-by-zero checks can be completely disabled using
14491 @option{-mno-check-zero-division}.
14496 @opindex mno-memcpy
14497 Force (do not force) the use of @code{memcpy()} for non-trivial block
14498 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14499 most constant-sized copies.
14502 @itemx -mno-long-calls
14503 @opindex mlong-calls
14504 @opindex mno-long-calls
14505 Disable (do not disable) use of the @code{jal} instruction. Calling
14506 functions using @code{jal} is more efficient but requires the caller
14507 and callee to be in the same 256 megabyte segment.
14509 This option has no effect on abicalls code. The default is
14510 @option{-mno-long-calls}.
14516 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14517 instructions, as provided by the R4650 ISA@.
14520 @itemx -mno-fused-madd
14521 @opindex mfused-madd
14522 @opindex mno-fused-madd
14523 Enable (disable) use of the floating point multiply-accumulate
14524 instructions, when they are available. The default is
14525 @option{-mfused-madd}.
14527 When multiply-accumulate instructions are used, the intermediate
14528 product is calculated to infinite precision and is not subject to
14529 the FCSR Flush to Zero bit. This may be undesirable in some
14534 Tell the MIPS assembler to not run its preprocessor over user
14535 assembler files (with a @samp{.s} suffix) when assembling them.
14538 @itemx -mno-fix-r4000
14539 @opindex mfix-r4000
14540 @opindex mno-fix-r4000
14541 Work around certain R4000 CPU errata:
14544 A double-word or a variable shift may give an incorrect result if executed
14545 immediately after starting an integer division.
14547 A double-word or a variable shift may give an incorrect result if executed
14548 while an integer multiplication is in progress.
14550 An integer division may give an incorrect result if started in a delay slot
14551 of a taken branch or a jump.
14555 @itemx -mno-fix-r4400
14556 @opindex mfix-r4400
14557 @opindex mno-fix-r4400
14558 Work around certain R4400 CPU errata:
14561 A double-word or a variable shift may give an incorrect result if executed
14562 immediately after starting an integer division.
14566 @itemx -mno-fix-r10000
14567 @opindex mfix-r10000
14568 @opindex mno-fix-r10000
14569 Work around certain R10000 errata:
14572 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14573 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14576 This option can only be used if the target architecture supports
14577 branch-likely instructions. @option{-mfix-r10000} is the default when
14578 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14582 @itemx -mno-fix-vr4120
14583 @opindex mfix-vr4120
14584 Work around certain VR4120 errata:
14587 @code{dmultu} does not always produce the correct result.
14589 @code{div} and @code{ddiv} do not always produce the correct result if one
14590 of the operands is negative.
14592 The workarounds for the division errata rely on special functions in
14593 @file{libgcc.a}. At present, these functions are only provided by
14594 the @code{mips64vr*-elf} configurations.
14596 Other VR4120 errata require a nop to be inserted between certain pairs of
14597 instructions. These errata are handled by the assembler, not by GCC itself.
14600 @opindex mfix-vr4130
14601 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14602 workarounds are implemented by the assembler rather than by GCC,
14603 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14604 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14605 instructions are available instead.
14608 @itemx -mno-fix-sb1
14610 Work around certain SB-1 CPU core errata.
14611 (This flag currently works around the SB-1 revision 2
14612 ``F1'' and ``F2'' floating point errata.)
14614 @item -mr10k-cache-barrier=@var{setting}
14615 @opindex mr10k-cache-barrier
14616 Specify whether GCC should insert cache barriers to avoid the
14617 side-effects of speculation on R10K processors.
14619 In common with many processors, the R10K tries to predict the outcome
14620 of a conditional branch and speculatively executes instructions from
14621 the ``taken'' branch. It later aborts these instructions if the
14622 predicted outcome was wrong. However, on the R10K, even aborted
14623 instructions can have side effects.
14625 This problem only affects kernel stores and, depending on the system,
14626 kernel loads. As an example, a speculatively-executed store may load
14627 the target memory into cache and mark the cache line as dirty, even if
14628 the store itself is later aborted. If a DMA operation writes to the
14629 same area of memory before the ``dirty'' line is flushed, the cached
14630 data will overwrite the DMA-ed data. See the R10K processor manual
14631 for a full description, including other potential problems.
14633 One workaround is to insert cache barrier instructions before every memory
14634 access that might be speculatively executed and that might have side
14635 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14636 controls GCC's implementation of this workaround. It assumes that
14637 aborted accesses to any byte in the following regions will not have
14642 the memory occupied by the current function's stack frame;
14645 the memory occupied by an incoming stack argument;
14648 the memory occupied by an object with a link-time-constant address.
14651 It is the kernel's responsibility to ensure that speculative
14652 accesses to these regions are indeed safe.
14654 If the input program contains a function declaration such as:
14660 then the implementation of @code{foo} must allow @code{j foo} and
14661 @code{jal foo} to be executed speculatively. GCC honors this
14662 restriction for functions it compiles itself. It expects non-GCC
14663 functions (such as hand-written assembly code) to do the same.
14665 The option has three forms:
14668 @item -mr10k-cache-barrier=load-store
14669 Insert a cache barrier before a load or store that might be
14670 speculatively executed and that might have side effects even
14673 @item -mr10k-cache-barrier=store
14674 Insert a cache barrier before a store that might be speculatively
14675 executed and that might have side effects even if aborted.
14677 @item -mr10k-cache-barrier=none
14678 Disable the insertion of cache barriers. This is the default setting.
14681 @item -mflush-func=@var{func}
14682 @itemx -mno-flush-func
14683 @opindex mflush-func
14684 Specifies the function to call to flush the I and D caches, or to not
14685 call any such function. If called, the function must take the same
14686 arguments as the common @code{_flush_func()}, that is, the address of the
14687 memory range for which the cache is being flushed, the size of the
14688 memory range, and the number 3 (to flush both caches). The default
14689 depends on the target GCC was configured for, but commonly is either
14690 @samp{_flush_func} or @samp{__cpu_flush}.
14692 @item mbranch-cost=@var{num}
14693 @opindex mbranch-cost
14694 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14695 This cost is only a heuristic and is not guaranteed to produce
14696 consistent results across releases. A zero cost redundantly selects
14697 the default, which is based on the @option{-mtune} setting.
14699 @item -mbranch-likely
14700 @itemx -mno-branch-likely
14701 @opindex mbranch-likely
14702 @opindex mno-branch-likely
14703 Enable or disable use of Branch Likely instructions, regardless of the
14704 default for the selected architecture. By default, Branch Likely
14705 instructions may be generated if they are supported by the selected
14706 architecture. An exception is for the MIPS32 and MIPS64 architectures
14707 and processors which implement those architectures; for those, Branch
14708 Likely instructions will not be generated by default because the MIPS32
14709 and MIPS64 architectures specifically deprecate their use.
14711 @item -mfp-exceptions
14712 @itemx -mno-fp-exceptions
14713 @opindex mfp-exceptions
14714 Specifies whether FP exceptions are enabled. This affects how we schedule
14715 FP instructions for some processors. The default is that FP exceptions are
14718 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14719 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14722 @item -mvr4130-align
14723 @itemx -mno-vr4130-align
14724 @opindex mvr4130-align
14725 The VR4130 pipeline is two-way superscalar, but can only issue two
14726 instructions together if the first one is 8-byte aligned. When this
14727 option is enabled, GCC will align pairs of instructions that it
14728 thinks should execute in parallel.
14730 This option only has an effect when optimizing for the VR4130.
14731 It normally makes code faster, but at the expense of making it bigger.
14732 It is enabled by default at optimization level @option{-O3}.
14737 Enable (disable) generation of @code{synci} instructions on
14738 architectures that support it. The @code{synci} instructions (if
14739 enabled) will be generated when @code{__builtin___clear_cache()} is
14742 This option defaults to @code{-mno-synci}, but the default can be
14743 overridden by configuring with @code{--with-synci}.
14745 When compiling code for single processor systems, it is generally safe
14746 to use @code{synci}. However, on many multi-core (SMP) systems, it
14747 will not invalidate the instruction caches on all cores and may lead
14748 to undefined behavior.
14750 @item -mrelax-pic-calls
14751 @itemx -mno-relax-pic-calls
14752 @opindex mrelax-pic-calls
14753 Try to turn PIC calls that are normally dispatched via register
14754 @code{$25} into direct calls. This is only possible if the linker can
14755 resolve the destination at link-time and if the destination is within
14756 range for a direct call.
14758 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14759 an assembler and a linker that supports the @code{.reloc} assembly
14760 directive and @code{-mexplicit-relocs} is in effect. With
14761 @code{-mno-explicit-relocs}, this optimization can be performed by the
14762 assembler and the linker alone without help from the compiler.
14764 @item -mmcount-ra-address
14765 @itemx -mno-mcount-ra-address
14766 @opindex mmcount-ra-address
14767 @opindex mno-mcount-ra-address
14768 Emit (do not emit) code that allows @code{_mcount} to modify the
14769 calling function's return address. When enabled, this option extends
14770 the usual @code{_mcount} interface with a new @var{ra-address}
14771 parameter, which has type @code{intptr_t *} and is passed in register
14772 @code{$12}. @code{_mcount} can then modify the return address by
14773 doing both of the following:
14776 Returning the new address in register @code{$31}.
14778 Storing the new address in @code{*@var{ra-address}},
14779 if @var{ra-address} is nonnull.
14782 The default is @option{-mno-mcount-ra-address}.
14787 @subsection MMIX Options
14788 @cindex MMIX Options
14790 These options are defined for the MMIX:
14794 @itemx -mno-libfuncs
14796 @opindex mno-libfuncs
14797 Specify that intrinsic library functions are being compiled, passing all
14798 values in registers, no matter the size.
14801 @itemx -mno-epsilon
14803 @opindex mno-epsilon
14804 Generate floating-point comparison instructions that compare with respect
14805 to the @code{rE} epsilon register.
14807 @item -mabi=mmixware
14809 @opindex mabi=mmixware
14811 Generate code that passes function parameters and return values that (in
14812 the called function) are seen as registers @code{$0} and up, as opposed to
14813 the GNU ABI which uses global registers @code{$231} and up.
14815 @item -mzero-extend
14816 @itemx -mno-zero-extend
14817 @opindex mzero-extend
14818 @opindex mno-zero-extend
14819 When reading data from memory in sizes shorter than 64 bits, use (do not
14820 use) zero-extending load instructions by default, rather than
14821 sign-extending ones.
14824 @itemx -mno-knuthdiv
14826 @opindex mno-knuthdiv
14827 Make the result of a division yielding a remainder have the same sign as
14828 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14829 remainder follows the sign of the dividend. Both methods are
14830 arithmetically valid, the latter being almost exclusively used.
14832 @item -mtoplevel-symbols
14833 @itemx -mno-toplevel-symbols
14834 @opindex mtoplevel-symbols
14835 @opindex mno-toplevel-symbols
14836 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14837 code can be used with the @code{PREFIX} assembly directive.
14841 Generate an executable in the ELF format, rather than the default
14842 @samp{mmo} format used by the @command{mmix} simulator.
14844 @item -mbranch-predict
14845 @itemx -mno-branch-predict
14846 @opindex mbranch-predict
14847 @opindex mno-branch-predict
14848 Use (do not use) the probable-branch instructions, when static branch
14849 prediction indicates a probable branch.
14851 @item -mbase-addresses
14852 @itemx -mno-base-addresses
14853 @opindex mbase-addresses
14854 @opindex mno-base-addresses
14855 Generate (do not generate) code that uses @emph{base addresses}. Using a
14856 base address automatically generates a request (handled by the assembler
14857 and the linker) for a constant to be set up in a global register. The
14858 register is used for one or more base address requests within the range 0
14859 to 255 from the value held in the register. The generally leads to short
14860 and fast code, but the number of different data items that can be
14861 addressed is limited. This means that a program that uses lots of static
14862 data may require @option{-mno-base-addresses}.
14864 @item -msingle-exit
14865 @itemx -mno-single-exit
14866 @opindex msingle-exit
14867 @opindex mno-single-exit
14868 Force (do not force) generated code to have a single exit point in each
14872 @node MN10300 Options
14873 @subsection MN10300 Options
14874 @cindex MN10300 options
14876 These @option{-m} options are defined for Matsushita MN10300 architectures:
14881 Generate code to avoid bugs in the multiply instructions for the MN10300
14882 processors. This is the default.
14884 @item -mno-mult-bug
14885 @opindex mno-mult-bug
14886 Do not generate code to avoid bugs in the multiply instructions for the
14887 MN10300 processors.
14891 Generate code which uses features specific to the AM33 processor.
14895 Do not generate code which uses features specific to the AM33 processor. This
14900 Generate code which uses features specific to the AM33/2.0 processor.
14904 Generate code which uses features specific to the AM34 processor.
14906 @item -mtune=@var{cpu-type}
14908 Use the timing characteristics of the indicated CPU type when
14909 scheduling instructions. This does not change the targeted processor
14910 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
14911 @samp{am33-2} or @samp{am34}.
14913 @item -mreturn-pointer-on-d0
14914 @opindex mreturn-pointer-on-d0
14915 When generating a function which returns a pointer, return the pointer
14916 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14917 only in a0, and attempts to call such functions without a prototype
14918 would result in errors. Note that this option is on by default; use
14919 @option{-mno-return-pointer-on-d0} to disable it.
14923 Do not link in the C run-time initialization object file.
14927 Indicate to the linker that it should perform a relaxation optimization pass
14928 to shorten branches, calls and absolute memory addresses. This option only
14929 has an effect when used on the command line for the final link step.
14931 This option makes symbolic debugging impossible.
14935 Allow the compiler to generate @emph{Long Instruction Word}
14936 instructions if the target is the @samp{AM33} or later. This is the
14937 default. This option defines the preprocessor macro @samp{__LIW__}.
14941 Do not allow the compiler to generate @emph{Long Instruction Word}
14942 instructions. This option defines the preprocessor macro
14947 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
14948 instructions if the target is the @samp{AM33} or later. This is the
14949 default. This option defines the preprocessor macro @samp{__SETLB__}.
14953 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
14954 instructions. This option defines the preprocessor macro
14955 @samp{__NO_SETLB__}.
14959 @node PDP-11 Options
14960 @subsection PDP-11 Options
14961 @cindex PDP-11 Options
14963 These options are defined for the PDP-11:
14968 Use hardware FPP floating point. This is the default. (FIS floating
14969 point on the PDP-11/40 is not supported.)
14972 @opindex msoft-float
14973 Do not use hardware floating point.
14977 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14981 Return floating-point results in memory. This is the default.
14985 Generate code for a PDP-11/40.
14989 Generate code for a PDP-11/45. This is the default.
14993 Generate code for a PDP-11/10.
14995 @item -mbcopy-builtin
14996 @opindex mbcopy-builtin
14997 Use inline @code{movmemhi} patterns for copying memory. This is the
15002 Do not use inline @code{movmemhi} patterns for copying memory.
15008 Use 16-bit @code{int}. This is the default.
15014 Use 32-bit @code{int}.
15017 @itemx -mno-float32
15019 @opindex mno-float32
15020 Use 64-bit @code{float}. This is the default.
15023 @itemx -mno-float64
15025 @opindex mno-float64
15026 Use 32-bit @code{float}.
15030 Use @code{abshi2} pattern. This is the default.
15034 Do not use @code{abshi2} pattern.
15036 @item -mbranch-expensive
15037 @opindex mbranch-expensive
15038 Pretend that branches are expensive. This is for experimenting with
15039 code generation only.
15041 @item -mbranch-cheap
15042 @opindex mbranch-cheap
15043 Do not pretend that branches are expensive. This is the default.
15047 Use Unix assembler syntax. This is the default when configured for
15048 @samp{pdp11-*-bsd}.
15052 Use DEC assembler syntax. This is the default when configured for any
15053 PDP-11 target other than @samp{pdp11-*-bsd}.
15056 @node picoChip Options
15057 @subsection picoChip Options
15058 @cindex picoChip options
15060 These @samp{-m} options are defined for picoChip implementations:
15064 @item -mae=@var{ae_type}
15066 Set the instruction set, register set, and instruction scheduling
15067 parameters for array element type @var{ae_type}. Supported values
15068 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15070 @option{-mae=ANY} selects a completely generic AE type. Code
15071 generated with this option will run on any of the other AE types. The
15072 code will not be as efficient as it would be if compiled for a specific
15073 AE type, and some types of operation (e.g., multiplication) will not
15074 work properly on all types of AE.
15076 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15077 for compiled code, and is the default.
15079 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15080 option may suffer from poor performance of byte (char) manipulation,
15081 since the DSP AE does not provide hardware support for byte load/stores.
15083 @item -msymbol-as-address
15084 Enable the compiler to directly use a symbol name as an address in a
15085 load/store instruction, without first loading it into a
15086 register. Typically, the use of this option will generate larger
15087 programs, which run faster than when the option isn't used. However, the
15088 results vary from program to program, so it is left as a user option,
15089 rather than being permanently enabled.
15091 @item -mno-inefficient-warnings
15092 Disables warnings about the generation of inefficient code. These
15093 warnings can be generated, for example, when compiling code which
15094 performs byte-level memory operations on the MAC AE type. The MAC AE has
15095 no hardware support for byte-level memory operations, so all byte
15096 load/stores must be synthesized from word load/store operations. This is
15097 inefficient and a warning will be generated indicating to the programmer
15098 that they should rewrite the code to avoid byte operations, or to target
15099 an AE type which has the necessary hardware support. This option enables
15100 the warning to be turned off.
15104 @node PowerPC Options
15105 @subsection PowerPC Options
15106 @cindex PowerPC options
15108 These are listed under @xref{RS/6000 and PowerPC Options}.
15110 @node RS/6000 and PowerPC Options
15111 @subsection IBM RS/6000 and PowerPC Options
15112 @cindex RS/6000 and PowerPC Options
15113 @cindex IBM RS/6000 and PowerPC Options
15115 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15122 @itemx -mno-powerpc
15123 @itemx -mpowerpc-gpopt
15124 @itemx -mno-powerpc-gpopt
15125 @itemx -mpowerpc-gfxopt
15126 @itemx -mno-powerpc-gfxopt
15129 @itemx -mno-powerpc64
15133 @itemx -mno-popcntb
15135 @itemx -mno-popcntd
15144 @itemx -mno-hard-dfp
15148 @opindex mno-power2
15150 @opindex mno-powerpc
15151 @opindex mpowerpc-gpopt
15152 @opindex mno-powerpc-gpopt
15153 @opindex mpowerpc-gfxopt
15154 @opindex mno-powerpc-gfxopt
15155 @opindex mpowerpc64
15156 @opindex mno-powerpc64
15160 @opindex mno-popcntb
15162 @opindex mno-popcntd
15168 @opindex mno-mfpgpr
15170 @opindex mno-hard-dfp
15171 GCC supports two related instruction set architectures for the
15172 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15173 instructions supported by the @samp{rios} chip set used in the original
15174 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15175 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15176 the IBM 4xx, 6xx, and follow-on microprocessors.
15178 Neither architecture is a subset of the other. However there is a
15179 large common subset of instructions supported by both. An MQ
15180 register is included in processors supporting the POWER architecture.
15182 You use these options to specify which instructions are available on the
15183 processor you are using. The default value of these options is
15184 determined when configuring GCC@. Specifying the
15185 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15186 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15187 rather than the options listed above.
15189 The @option{-mpower} option allows GCC to generate instructions that
15190 are found only in the POWER architecture and to use the MQ register.
15191 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15192 to generate instructions that are present in the POWER2 architecture but
15193 not the original POWER architecture.
15195 The @option{-mpowerpc} option allows GCC to generate instructions that
15196 are found only in the 32-bit subset of the PowerPC architecture.
15197 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15198 GCC to use the optional PowerPC architecture instructions in the
15199 General Purpose group, including floating-point square root. Specifying
15200 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15201 use the optional PowerPC architecture instructions in the Graphics
15202 group, including floating-point select.
15204 The @option{-mmfcrf} option allows GCC to generate the move from
15205 condition register field instruction implemented on the POWER4
15206 processor and other processors that support the PowerPC V2.01
15208 The @option{-mpopcntb} option allows GCC to generate the popcount and
15209 double precision FP reciprocal estimate instruction implemented on the
15210 POWER5 processor and other processors that support the PowerPC V2.02
15212 The @option{-mpopcntd} option allows GCC to generate the popcount
15213 instruction implemented on the POWER7 processor and other processors
15214 that support the PowerPC V2.06 architecture.
15215 The @option{-mfprnd} option allows GCC to generate the FP round to
15216 integer instructions implemented on the POWER5+ processor and other
15217 processors that support the PowerPC V2.03 architecture.
15218 The @option{-mcmpb} option allows GCC to generate the compare bytes
15219 instruction implemented on the POWER6 processor and other processors
15220 that support the PowerPC V2.05 architecture.
15221 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15222 general purpose register instructions implemented on the POWER6X
15223 processor and other processors that support the extended PowerPC V2.05
15225 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15226 point instructions implemented on some POWER processors.
15228 The @option{-mpowerpc64} option allows GCC to generate the additional
15229 64-bit instructions that are found in the full PowerPC64 architecture
15230 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15231 @option{-mno-powerpc64}.
15233 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15234 will use only the instructions in the common subset of both
15235 architectures plus some special AIX common-mode calls, and will not use
15236 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15237 permits GCC to use any instruction from either architecture and to
15238 allow use of the MQ register; specify this for the Motorola MPC601.
15240 @item -mnew-mnemonics
15241 @itemx -mold-mnemonics
15242 @opindex mnew-mnemonics
15243 @opindex mold-mnemonics
15244 Select which mnemonics to use in the generated assembler code. With
15245 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15246 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15247 assembler mnemonics defined for the POWER architecture. Instructions
15248 defined in only one architecture have only one mnemonic; GCC uses that
15249 mnemonic irrespective of which of these options is specified.
15251 GCC defaults to the mnemonics appropriate for the architecture in
15252 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15253 value of these option. Unless you are building a cross-compiler, you
15254 should normally not specify either @option{-mnew-mnemonics} or
15255 @option{-mold-mnemonics}, but should instead accept the default.
15257 @item -mcpu=@var{cpu_type}
15259 Set architecture type, register usage, choice of mnemonics, and
15260 instruction scheduling parameters for machine type @var{cpu_type}.
15261 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15262 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15263 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15264 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15265 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15266 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15267 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15268 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15269 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15270 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15271 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15273 @option{-mcpu=common} selects a completely generic processor. Code
15274 generated under this option will run on any POWER or PowerPC processor.
15275 GCC will use only the instructions in the common subset of both
15276 architectures, and will not use the MQ register. GCC assumes a generic
15277 processor model for scheduling purposes.
15279 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15280 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15281 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15282 types, with an appropriate, generic processor model assumed for
15283 scheduling purposes.
15285 The other options specify a specific processor. Code generated under
15286 those options will run best on that processor, and may not run at all on
15289 The @option{-mcpu} options automatically enable or disable the
15292 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15293 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15294 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15295 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15297 The particular options set for any particular CPU will vary between
15298 compiler versions, depending on what setting seems to produce optimal
15299 code for that CPU; it doesn't necessarily reflect the actual hardware's
15300 capabilities. If you wish to set an individual option to a particular
15301 value, you may specify it after the @option{-mcpu} option, like
15302 @samp{-mcpu=970 -mno-altivec}.
15304 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15305 not enabled or disabled by the @option{-mcpu} option at present because
15306 AIX does not have full support for these options. You may still
15307 enable or disable them individually if you're sure it'll work in your
15310 @item -mtune=@var{cpu_type}
15312 Set the instruction scheduling parameters for machine type
15313 @var{cpu_type}, but do not set the architecture type, register usage, or
15314 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15315 values for @var{cpu_type} are used for @option{-mtune} as for
15316 @option{-mcpu}. If both are specified, the code generated will use the
15317 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15318 scheduling parameters set by @option{-mtune}.
15320 @item -mcmodel=small
15321 @opindex mcmodel=small
15322 Generate PowerPC64 code for the small model: The TOC is limited to
15325 @item -mcmodel=medium
15326 @opindex mcmodel=medium
15327 Generate PowerPC64 code for the medium model: The TOC and other static
15328 data may be up to a total of 4G in size.
15330 @item -mcmodel=large
15331 @opindex mcmodel=large
15332 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15333 in size. Other data and code is only limited by the 64-bit address
15337 @itemx -mno-altivec
15339 @opindex mno-altivec
15340 Generate code that uses (does not use) AltiVec instructions, and also
15341 enable the use of built-in functions that allow more direct access to
15342 the AltiVec instruction set. You may also need to set
15343 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15349 @opindex mno-vrsave
15350 Generate VRSAVE instructions when generating AltiVec code.
15352 @item -mgen-cell-microcode
15353 @opindex mgen-cell-microcode
15354 Generate Cell microcode instructions
15356 @item -mwarn-cell-microcode
15357 @opindex mwarn-cell-microcode
15358 Warning when a Cell microcode instruction is going to emitted. An example
15359 of a Cell microcode instruction is a variable shift.
15362 @opindex msecure-plt
15363 Generate code that allows ld and ld.so to build executables and shared
15364 libraries with non-exec .plt and .got sections. This is a PowerPC
15365 32-bit SYSV ABI option.
15369 Generate code that uses a BSS .plt section that ld.so fills in, and
15370 requires .plt and .got sections that are both writable and executable.
15371 This is a PowerPC 32-bit SYSV ABI option.
15377 This switch enables or disables the generation of ISEL instructions.
15379 @item -misel=@var{yes/no}
15380 This switch has been deprecated. Use @option{-misel} and
15381 @option{-mno-isel} instead.
15387 This switch enables or disables the generation of SPE simd
15393 @opindex mno-paired
15394 This switch enables or disables the generation of PAIRED simd
15397 @item -mspe=@var{yes/no}
15398 This option has been deprecated. Use @option{-mspe} and
15399 @option{-mno-spe} instead.
15405 Generate code that uses (does not use) vector/scalar (VSX)
15406 instructions, and also enable the use of built-in functions that allow
15407 more direct access to the VSX instruction set.
15409 @item -mfloat-gprs=@var{yes/single/double/no}
15410 @itemx -mfloat-gprs
15411 @opindex mfloat-gprs
15412 This switch enables or disables the generation of floating point
15413 operations on the general purpose registers for architectures that
15416 The argument @var{yes} or @var{single} enables the use of
15417 single-precision floating point operations.
15419 The argument @var{double} enables the use of single and
15420 double-precision floating point operations.
15422 The argument @var{no} disables floating point operations on the
15423 general purpose registers.
15425 This option is currently only available on the MPC854x.
15431 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15432 targets (including GNU/Linux). The 32-bit environment sets int, long
15433 and pointer to 32 bits and generates code that runs on any PowerPC
15434 variant. The 64-bit environment sets int to 32 bits and long and
15435 pointer to 64 bits, and generates code for PowerPC64, as for
15436 @option{-mpowerpc64}.
15439 @itemx -mno-fp-in-toc
15440 @itemx -mno-sum-in-toc
15441 @itemx -mminimal-toc
15443 @opindex mno-fp-in-toc
15444 @opindex mno-sum-in-toc
15445 @opindex mminimal-toc
15446 Modify generation of the TOC (Table Of Contents), which is created for
15447 every executable file. The @option{-mfull-toc} option is selected by
15448 default. In that case, GCC will allocate at least one TOC entry for
15449 each unique non-automatic variable reference in your program. GCC
15450 will also place floating-point constants in the TOC@. However, only
15451 16,384 entries are available in the TOC@.
15453 If you receive a linker error message that saying you have overflowed
15454 the available TOC space, you can reduce the amount of TOC space used
15455 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15456 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15457 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15458 generate code to calculate the sum of an address and a constant at
15459 run-time instead of putting that sum into the TOC@. You may specify one
15460 or both of these options. Each causes GCC to produce very slightly
15461 slower and larger code at the expense of conserving TOC space.
15463 If you still run out of space in the TOC even when you specify both of
15464 these options, specify @option{-mminimal-toc} instead. This option causes
15465 GCC to make only one TOC entry for every file. When you specify this
15466 option, GCC will produce code that is slower and larger but which
15467 uses extremely little TOC space. You may wish to use this option
15468 only on files that contain less frequently executed code.
15474 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15475 @code{long} type, and the infrastructure needed to support them.
15476 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15477 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15478 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15481 @itemx -mno-xl-compat
15482 @opindex mxl-compat
15483 @opindex mno-xl-compat
15484 Produce code that conforms more closely to IBM XL compiler semantics
15485 when using AIX-compatible ABI@. Pass floating-point arguments to
15486 prototyped functions beyond the register save area (RSA) on the stack
15487 in addition to argument FPRs. Do not assume that most significant
15488 double in 128-bit long double value is properly rounded when comparing
15489 values and converting to double. Use XL symbol names for long double
15492 The AIX calling convention was extended but not initially documented to
15493 handle an obscure K&R C case of calling a function that takes the
15494 address of its arguments with fewer arguments than declared. IBM XL
15495 compilers access floating point arguments which do not fit in the
15496 RSA from the stack when a subroutine is compiled without
15497 optimization. Because always storing floating-point arguments on the
15498 stack is inefficient and rarely needed, this option is not enabled by
15499 default and only is necessary when calling subroutines compiled by IBM
15500 XL compilers without optimization.
15504 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15505 application written to use message passing with special startup code to
15506 enable the application to run. The system must have PE installed in the
15507 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15508 must be overridden with the @option{-specs=} option to specify the
15509 appropriate directory location. The Parallel Environment does not
15510 support threads, so the @option{-mpe} option and the @option{-pthread}
15511 option are incompatible.
15513 @item -malign-natural
15514 @itemx -malign-power
15515 @opindex malign-natural
15516 @opindex malign-power
15517 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15518 @option{-malign-natural} overrides the ABI-defined alignment of larger
15519 types, such as floating-point doubles, on their natural size-based boundary.
15520 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15521 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15523 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15527 @itemx -mhard-float
15528 @opindex msoft-float
15529 @opindex mhard-float
15530 Generate code that does not use (uses) the floating-point register set.
15531 Software floating point emulation is provided if you use the
15532 @option{-msoft-float} option, and pass the option to GCC when linking.
15534 @item -msingle-float
15535 @itemx -mdouble-float
15536 @opindex msingle-float
15537 @opindex mdouble-float
15538 Generate code for single or double-precision floating point operations.
15539 @option{-mdouble-float} implies @option{-msingle-float}.
15542 @opindex msimple-fpu
15543 Do not generate sqrt and div instructions for hardware floating point unit.
15547 Specify type of floating point unit. Valid values are @var{sp_lite}
15548 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15549 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15550 and @var{dp_full} (equivalent to -mdouble-float).
15553 @opindex mxilinx-fpu
15554 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15557 @itemx -mno-multiple
15559 @opindex mno-multiple
15560 Generate code that uses (does not use) the load multiple word
15561 instructions and the store multiple word instructions. These
15562 instructions are generated by default on POWER systems, and not
15563 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15564 endian PowerPC systems, since those instructions do not work when the
15565 processor is in little endian mode. The exceptions are PPC740 and
15566 PPC750 which permit the instructions usage in little endian mode.
15571 @opindex mno-string
15572 Generate code that uses (does not use) the load string instructions
15573 and the store string word instructions to save multiple registers and
15574 do small block moves. These instructions are generated by default on
15575 POWER systems, and not generated on PowerPC systems. Do not use
15576 @option{-mstring} on little endian PowerPC systems, since those
15577 instructions do not work when the processor is in little endian mode.
15578 The exceptions are PPC740 and PPC750 which permit the instructions
15579 usage in little endian mode.
15584 @opindex mno-update
15585 Generate code that uses (does not use) the load or store instructions
15586 that update the base register to the address of the calculated memory
15587 location. These instructions are generated by default. If you use
15588 @option{-mno-update}, there is a small window between the time that the
15589 stack pointer is updated and the address of the previous frame is
15590 stored, which means code that walks the stack frame across interrupts or
15591 signals may get corrupted data.
15593 @item -mavoid-indexed-addresses
15594 @itemx -mno-avoid-indexed-addresses
15595 @opindex mavoid-indexed-addresses
15596 @opindex mno-avoid-indexed-addresses
15597 Generate code that tries to avoid (not avoid) the use of indexed load
15598 or store instructions. These instructions can incur a performance
15599 penalty on Power6 processors in certain situations, such as when
15600 stepping through large arrays that cross a 16M boundary. This option
15601 is enabled by default when targetting Power6 and disabled otherwise.
15604 @itemx -mno-fused-madd
15605 @opindex mfused-madd
15606 @opindex mno-fused-madd
15607 Generate code that uses (does not use) the floating point multiply and
15608 accumulate instructions. These instructions are generated by default
15609 if hardware floating point is used. The machine dependent
15610 @option{-mfused-madd} option is now mapped to the machine independent
15611 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15612 mapped to @option{-ffp-contract=off}.
15618 Generate code that uses (does not use) the half-word multiply and
15619 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15620 These instructions are generated by default when targetting those
15627 Generate code that uses (does not use) the string-search @samp{dlmzb}
15628 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15629 generated by default when targetting those processors.
15631 @item -mno-bit-align
15633 @opindex mno-bit-align
15634 @opindex mbit-align
15635 On System V.4 and embedded PowerPC systems do not (do) force structures
15636 and unions that contain bit-fields to be aligned to the base type of the
15639 For example, by default a structure containing nothing but 8
15640 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15641 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15642 the structure would be aligned to a 1 byte boundary and be one byte in
15645 @item -mno-strict-align
15646 @itemx -mstrict-align
15647 @opindex mno-strict-align
15648 @opindex mstrict-align
15649 On System V.4 and embedded PowerPC systems do not (do) assume that
15650 unaligned memory references will be handled by the system.
15652 @item -mrelocatable
15653 @itemx -mno-relocatable
15654 @opindex mrelocatable
15655 @opindex mno-relocatable
15656 Generate code that allows (does not allow) a static executable to be
15657 relocated to a different address at runtime. A simple embedded
15658 PowerPC system loader should relocate the entire contents of
15659 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15660 a table of 32-bit addresses generated by this option. For this to
15661 work, all objects linked together must be compiled with
15662 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15663 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15665 @item -mrelocatable-lib
15666 @itemx -mno-relocatable-lib
15667 @opindex mrelocatable-lib
15668 @opindex mno-relocatable-lib
15669 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15670 @code{.fixup} section to allow static executables to be relocated at
15671 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15672 alignment of @option{-mrelocatable}. Objects compiled with
15673 @option{-mrelocatable-lib} may be linked with objects compiled with
15674 any combination of the @option{-mrelocatable} options.
15680 On System V.4 and embedded PowerPC systems do not (do) assume that
15681 register 2 contains a pointer to a global area pointing to the addresses
15682 used in the program.
15685 @itemx -mlittle-endian
15687 @opindex mlittle-endian
15688 On System V.4 and embedded PowerPC systems compile code for the
15689 processor in little endian mode. The @option{-mlittle-endian} option is
15690 the same as @option{-mlittle}.
15693 @itemx -mbig-endian
15695 @opindex mbig-endian
15696 On System V.4 and embedded PowerPC systems compile code for the
15697 processor in big endian mode. The @option{-mbig-endian} option is
15698 the same as @option{-mbig}.
15700 @item -mdynamic-no-pic
15701 @opindex mdynamic-no-pic
15702 On Darwin and Mac OS X systems, compile code so that it is not
15703 relocatable, but that its external references are relocatable. The
15704 resulting code is suitable for applications, but not shared
15707 @item -msingle-pic-base
15708 @opindex msingle-pic-base
15709 Treat the register used for PIC addressing as read-only, rather than
15710 loading it in the prologue for each function. The run-time system is
15711 responsible for initializing this register with an appropriate value
15712 before execution begins.
15714 @item -mprioritize-restricted-insns=@var{priority}
15715 @opindex mprioritize-restricted-insns
15716 This option controls the priority that is assigned to
15717 dispatch-slot restricted instructions during the second scheduling
15718 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15719 @var{no/highest/second-highest} priority to dispatch slot restricted
15722 @item -msched-costly-dep=@var{dependence_type}
15723 @opindex msched-costly-dep
15724 This option controls which dependences are considered costly
15725 by the target during instruction scheduling. The argument
15726 @var{dependence_type} takes one of the following values:
15727 @var{no}: no dependence is costly,
15728 @var{all}: all dependences are costly,
15729 @var{true_store_to_load}: a true dependence from store to load is costly,
15730 @var{store_to_load}: any dependence from store to load is costly,
15731 @var{number}: any dependence which latency >= @var{number} is costly.
15733 @item -minsert-sched-nops=@var{scheme}
15734 @opindex minsert-sched-nops
15735 This option controls which nop insertion scheme will be used during
15736 the second scheduling pass. The argument @var{scheme} takes one of the
15738 @var{no}: Don't insert nops.
15739 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15740 according to the scheduler's grouping.
15741 @var{regroup_exact}: Insert nops to force costly dependent insns into
15742 separate groups. Insert exactly as many nops as needed to force an insn
15743 to a new group, according to the estimated processor grouping.
15744 @var{number}: Insert nops to force costly dependent insns into
15745 separate groups. Insert @var{number} nops to force an insn to a new group.
15748 @opindex mcall-sysv
15749 On System V.4 and embedded PowerPC systems compile code using calling
15750 conventions that adheres to the March 1995 draft of the System V
15751 Application Binary Interface, PowerPC processor supplement. This is the
15752 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15754 @item -mcall-sysv-eabi
15756 @opindex mcall-sysv-eabi
15757 @opindex mcall-eabi
15758 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15760 @item -mcall-sysv-noeabi
15761 @opindex mcall-sysv-noeabi
15762 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15764 @item -mcall-aixdesc
15766 On System V.4 and embedded PowerPC systems compile code for the AIX
15770 @opindex mcall-linux
15771 On System V.4 and embedded PowerPC systems compile code for the
15772 Linux-based GNU system.
15774 @item -mcall-freebsd
15775 @opindex mcall-freebsd
15776 On System V.4 and embedded PowerPC systems compile code for the
15777 FreeBSD operating system.
15779 @item -mcall-netbsd
15780 @opindex mcall-netbsd
15781 On System V.4 and embedded PowerPC systems compile code for the
15782 NetBSD operating system.
15784 @item -mcall-openbsd
15785 @opindex mcall-netbsd
15786 On System V.4 and embedded PowerPC systems compile code for the
15787 OpenBSD operating system.
15789 @item -maix-struct-return
15790 @opindex maix-struct-return
15791 Return all structures in memory (as specified by the AIX ABI)@.
15793 @item -msvr4-struct-return
15794 @opindex msvr4-struct-return
15795 Return structures smaller than 8 bytes in registers (as specified by the
15798 @item -mabi=@var{abi-type}
15800 Extend the current ABI with a particular extension, or remove such extension.
15801 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15802 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15806 Extend the current ABI with SPE ABI extensions. This does not change
15807 the default ABI, instead it adds the SPE ABI extensions to the current
15811 @opindex mabi=no-spe
15812 Disable Booke SPE ABI extensions for the current ABI@.
15814 @item -mabi=ibmlongdouble
15815 @opindex mabi=ibmlongdouble
15816 Change the current ABI to use IBM extended precision long double.
15817 This is a PowerPC 32-bit SYSV ABI option.
15819 @item -mabi=ieeelongdouble
15820 @opindex mabi=ieeelongdouble
15821 Change the current ABI to use IEEE extended precision long double.
15822 This is a PowerPC 32-bit Linux ABI option.
15825 @itemx -mno-prototype
15826 @opindex mprototype
15827 @opindex mno-prototype
15828 On System V.4 and embedded PowerPC systems assume that all calls to
15829 variable argument functions are properly prototyped. Otherwise, the
15830 compiler must insert an instruction before every non prototyped call to
15831 set or clear bit 6 of the condition code register (@var{CR}) to
15832 indicate whether floating point values were passed in the floating point
15833 registers in case the function takes a variable arguments. With
15834 @option{-mprototype}, only calls to prototyped variable argument functions
15835 will set or clear the bit.
15839 On embedded PowerPC systems, assume that the startup module is called
15840 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15841 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15846 On embedded PowerPC systems, assume that the startup module is called
15847 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15852 On embedded PowerPC systems, assume that the startup module is called
15853 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15856 @item -myellowknife
15857 @opindex myellowknife
15858 On embedded PowerPC systems, assume that the startup module is called
15859 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15864 On System V.4 and embedded PowerPC systems, specify that you are
15865 compiling for a VxWorks system.
15869 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15870 header to indicate that @samp{eabi} extended relocations are used.
15876 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15877 Embedded Applications Binary Interface (eabi) which is a set of
15878 modifications to the System V.4 specifications. Selecting @option{-meabi}
15879 means that the stack is aligned to an 8 byte boundary, a function
15880 @code{__eabi} is called to from @code{main} to set up the eabi
15881 environment, and the @option{-msdata} option can use both @code{r2} and
15882 @code{r13} to point to two separate small data areas. Selecting
15883 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15884 do not call an initialization function from @code{main}, and the
15885 @option{-msdata} option will only use @code{r13} to point to a single
15886 small data area. The @option{-meabi} option is on by default if you
15887 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15890 @opindex msdata=eabi
15891 On System V.4 and embedded PowerPC systems, put small initialized
15892 @code{const} global and static data in the @samp{.sdata2} section, which
15893 is pointed to by register @code{r2}. Put small initialized
15894 non-@code{const} global and static data in the @samp{.sdata} section,
15895 which is pointed to by register @code{r13}. Put small uninitialized
15896 global and static data in the @samp{.sbss} section, which is adjacent to
15897 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15898 incompatible with the @option{-mrelocatable} option. The
15899 @option{-msdata=eabi} option also sets the @option{-memb} option.
15902 @opindex msdata=sysv
15903 On System V.4 and embedded PowerPC systems, put small global and static
15904 data in the @samp{.sdata} section, which is pointed to by register
15905 @code{r13}. Put small uninitialized global and static data in the
15906 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15907 The @option{-msdata=sysv} option is incompatible with the
15908 @option{-mrelocatable} option.
15910 @item -msdata=default
15912 @opindex msdata=default
15914 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15915 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15916 same as @option{-msdata=sysv}.
15919 @opindex msdata=data
15920 On System V.4 and embedded PowerPC systems, put small global
15921 data in the @samp{.sdata} section. Put small uninitialized global
15922 data in the @samp{.sbss} section. Do not use register @code{r13}
15923 to address small data however. This is the default behavior unless
15924 other @option{-msdata} options are used.
15928 @opindex msdata=none
15930 On embedded PowerPC systems, put all initialized global and static data
15931 in the @samp{.data} section, and all uninitialized data in the
15932 @samp{.bss} section.
15934 @item -mblock-move-inline-limit=@var{num}
15935 @opindex mblock-move-inline-limit
15936 Inline all block moves (such as calls to @code{memcpy} or structure
15937 copies) less than or equal to @var{num} bytes. The minimum value for
15938 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
15939 targets. The default value is target-specific.
15943 @cindex smaller data references (PowerPC)
15944 @cindex .sdata/.sdata2 references (PowerPC)
15945 On embedded PowerPC systems, put global and static items less than or
15946 equal to @var{num} bytes into the small data or bss sections instead of
15947 the normal data or bss section. By default, @var{num} is 8. The
15948 @option{-G @var{num}} switch is also passed to the linker.
15949 All modules should be compiled with the same @option{-G @var{num}} value.
15952 @itemx -mno-regnames
15954 @opindex mno-regnames
15955 On System V.4 and embedded PowerPC systems do (do not) emit register
15956 names in the assembly language output using symbolic forms.
15959 @itemx -mno-longcall
15961 @opindex mno-longcall
15962 By default assume that all calls are far away so that a longer more
15963 expensive calling sequence is required. This is required for calls
15964 further than 32 megabytes (33,554,432 bytes) from the current location.
15965 A short call will be generated if the compiler knows
15966 the call cannot be that far away. This setting can be overridden by
15967 the @code{shortcall} function attribute, or by @code{#pragma
15970 Some linkers are capable of detecting out-of-range calls and generating
15971 glue code on the fly. On these systems, long calls are unnecessary and
15972 generate slower code. As of this writing, the AIX linker can do this,
15973 as can the GNU linker for PowerPC/64. It is planned to add this feature
15974 to the GNU linker for 32-bit PowerPC systems as well.
15976 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15977 callee, L42'', plus a ``branch island'' (glue code). The two target
15978 addresses represent the callee and the ``branch island''. The
15979 Darwin/PPC linker will prefer the first address and generate a ``bl
15980 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15981 otherwise, the linker will generate ``bl L42'' to call the ``branch
15982 island''. The ``branch island'' is appended to the body of the
15983 calling function; it computes the full 32-bit address of the callee
15986 On Mach-O (Darwin) systems, this option directs the compiler emit to
15987 the glue for every direct call, and the Darwin linker decides whether
15988 to use or discard it.
15990 In the future, we may cause GCC to ignore all longcall specifications
15991 when the linker is known to generate glue.
15993 @item -mtls-markers
15994 @itemx -mno-tls-markers
15995 @opindex mtls-markers
15996 @opindex mno-tls-markers
15997 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15998 specifying the function argument. The relocation allows ld to
15999 reliably associate function call with argument setup instructions for
16000 TLS optimization, which in turn allows gcc to better schedule the
16005 Adds support for multithreading with the @dfn{pthreads} library.
16006 This option sets flags for both the preprocessor and linker.
16011 This option will enable GCC to use the reciprocal estimate and
16012 reciprocal square root estimate instructions with additional
16013 Newton-Raphson steps to increase precision instead of doing a divide or
16014 square root and divide for floating point arguments. You should use
16015 the @option{-ffast-math} option when using @option{-mrecip} (or at
16016 least @option{-funsafe-math-optimizations},
16017 @option{-finite-math-only}, @option{-freciprocal-math} and
16018 @option{-fno-trapping-math}). Note that while the throughput of the
16019 sequence is generally higher than the throughput of the non-reciprocal
16020 instruction, the precision of the sequence can be decreased by up to 2
16021 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16024 @item -mrecip=@var{opt}
16025 @opindex mrecip=opt
16026 This option allows to control which reciprocal estimate instructions
16027 may be used. @var{opt} is a comma separated list of options, that may
16028 be preceded by a @code{!} to invert the option:
16029 @code{all}: enable all estimate instructions,
16030 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16031 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16032 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16033 @code{divf}: enable the single precision reciprocal approximation instructions;
16034 @code{divd}: enable the double precision reciprocal approximation instructions;
16035 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16036 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16037 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16039 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16040 all of the reciprocal estimate instructions, except for the
16041 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16042 which handle the double precision reciprocal square root calculations.
16044 @item -mrecip-precision
16045 @itemx -mno-recip-precision
16046 @opindex mrecip-precision
16047 Assume (do not assume) that the reciprocal estimate instructions
16048 provide higher precision estimates than is mandated by the powerpc
16049 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16050 automatically selects @option{-mrecip-precision}. The double
16051 precision square root estimate instructions are not generated by
16052 default on low precision machines, since they do not provide an
16053 estimate that converges after three steps.
16055 @item -mveclibabi=@var{type}
16056 @opindex mveclibabi
16057 Specifies the ABI type to use for vectorizing intrinsics using an
16058 external library. The only type supported at present is @code{mass},
16059 which specifies to use IBM's Mathematical Acceleration Subsystem
16060 (MASS) libraries for vectorizing intrinsics using external libraries.
16061 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16062 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16063 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16064 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16065 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16066 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16067 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16068 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16069 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16070 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16071 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16072 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16073 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16074 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16075 for power7. Both @option{-ftree-vectorize} and
16076 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16077 libraries will have to be specified at link time.
16082 Generate (do not generate) the @code{friz} instruction when the
16083 @option{-funsafe-math-optimizations} option is used to optimize
16084 rounding a floating point value to 64-bit integer and back to floating
16085 point. The @code{friz} instruction does not return the same value if
16086 the floating point number is too large to fit in an integer.
16090 @subsection RX Options
16093 These command line options are defined for RX targets:
16096 @item -m64bit-doubles
16097 @itemx -m32bit-doubles
16098 @opindex m64bit-doubles
16099 @opindex m32bit-doubles
16100 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16101 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16102 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16103 works on 32-bit values, which is why the default is
16104 @option{-m32bit-doubles}.
16110 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16111 floating point hardware. The default is enabled for the @var{RX600}
16112 series and disabled for the @var{RX200} series.
16114 Floating point instructions will only be generated for 32-bit floating
16115 point values however, so if the @option{-m64bit-doubles} option is in
16116 use then the FPU hardware will not be used for doubles.
16118 @emph{Note} If the @option{-fpu} option is enabled then
16119 @option{-funsafe-math-optimizations} is also enabled automatically.
16120 This is because the RX FPU instructions are themselves unsafe.
16122 @item -mcpu=@var{name}
16124 Selects the type of RX CPU to be targeted. Currently three types are
16125 supported, the generic @var{RX600} and @var{RX200} series hardware and
16126 the specific @var{RX610} CPU. The default is @var{RX600}.
16128 The only difference between @var{RX600} and @var{RX610} is that the
16129 @var{RX610} does not support the @code{MVTIPL} instruction.
16131 The @var{RX200} series does not have a hardware floating point unit
16132 and so @option{-nofpu} is enabled by default when this type is
16135 @item -mbig-endian-data
16136 @itemx -mlittle-endian-data
16137 @opindex mbig-endian-data
16138 @opindex mlittle-endian-data
16139 Store data (but not code) in the big-endian format. The default is
16140 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16143 @item -msmall-data-limit=@var{N}
16144 @opindex msmall-data-limit
16145 Specifies the maximum size in bytes of global and static variables
16146 which can be placed into the small data area. Using the small data
16147 area can lead to smaller and faster code, but the size of area is
16148 limited and it is up to the programmer to ensure that the area does
16149 not overflow. Also when the small data area is used one of the RX's
16150 registers (@code{r13}) is reserved for use pointing to this area, so
16151 it is no longer available for use by the compiler. This could result
16152 in slower and/or larger code if variables which once could have been
16153 held in @code{r13} are now pushed onto the stack.
16155 Note, common variables (variables which have not been initialised) and
16156 constants are not placed into the small data area as they are assigned
16157 to other sections in the output executable.
16159 The default value is zero, which disables this feature. Note, this
16160 feature is not enabled by default with higher optimization levels
16161 (@option{-O2} etc) because of the potentially detrimental effects of
16162 reserving register @code{r13}. It is up to the programmer to
16163 experiment and discover whether this feature is of benefit to their
16170 Use the simulator runtime. The default is to use the libgloss board
16173 @item -mas100-syntax
16174 @itemx -mno-as100-syntax
16175 @opindex mas100-syntax
16176 @opindex mno-as100-syntax
16177 When generating assembler output use a syntax that is compatible with
16178 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16179 assembler but it has some restrictions so generating it is not the
16182 @item -mmax-constant-size=@var{N}
16183 @opindex mmax-constant-size
16184 Specifies the maximum size, in bytes, of a constant that can be used as
16185 an operand in a RX instruction. Although the RX instruction set does
16186 allow constants of up to 4 bytes in length to be used in instructions,
16187 a longer value equates to a longer instruction. Thus in some
16188 circumstances it can be beneficial to restrict the size of constants
16189 that are used in instructions. Constants that are too big are instead
16190 placed into a constant pool and referenced via register indirection.
16192 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16193 or 4 means that constants of any size are allowed.
16197 Enable linker relaxation. Linker relaxation is a process whereby the
16198 linker will attempt to reduce the size of a program by finding shorter
16199 versions of various instructions. Disabled by default.
16201 @item -mint-register=@var{N}
16202 @opindex mint-register
16203 Specify the number of registers to reserve for fast interrupt handler
16204 functions. The value @var{N} can be between 0 and 4. A value of 1
16205 means that register @code{r13} will be reserved for the exclusive use
16206 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16207 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16208 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16209 A value of 0, the default, does not reserve any registers.
16211 @item -msave-acc-in-interrupts
16212 @opindex msave-acc-in-interrupts
16213 Specifies that interrupt handler functions should preserve the
16214 accumulator register. This is only necessary if normal code might use
16215 the accumulator register, for example because it performs 64-bit
16216 multiplications. The default is to ignore the accumulator as this
16217 makes the interrupt handlers faster.
16221 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16222 has special significance to the RX port when used with the
16223 @code{interrupt} function attribute. This attribute indicates a
16224 function intended to process fast interrupts. GCC will will ensure
16225 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16226 and/or @code{r13} and only provided that the normal use of the
16227 corresponding registers have been restricted via the
16228 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16231 @node S/390 and zSeries Options
16232 @subsection S/390 and zSeries Options
16233 @cindex S/390 and zSeries Options
16235 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16239 @itemx -msoft-float
16240 @opindex mhard-float
16241 @opindex msoft-float
16242 Use (do not use) the hardware floating-point instructions and registers
16243 for floating-point operations. When @option{-msoft-float} is specified,
16244 functions in @file{libgcc.a} will be used to perform floating-point
16245 operations. When @option{-mhard-float} is specified, the compiler
16246 generates IEEE floating-point instructions. This is the default.
16249 @itemx -mno-hard-dfp
16251 @opindex mno-hard-dfp
16252 Use (do not use) the hardware decimal-floating-point instructions for
16253 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16254 specified, functions in @file{libgcc.a} will be used to perform
16255 decimal-floating-point operations. When @option{-mhard-dfp} is
16256 specified, the compiler generates decimal-floating-point hardware
16257 instructions. This is the default for @option{-march=z9-ec} or higher.
16259 @item -mlong-double-64
16260 @itemx -mlong-double-128
16261 @opindex mlong-double-64
16262 @opindex mlong-double-128
16263 These switches control the size of @code{long double} type. A size
16264 of 64bit makes the @code{long double} type equivalent to the @code{double}
16265 type. This is the default.
16268 @itemx -mno-backchain
16269 @opindex mbackchain
16270 @opindex mno-backchain
16271 Store (do not store) the address of the caller's frame as backchain pointer
16272 into the callee's stack frame.
16273 A backchain may be needed to allow debugging using tools that do not understand
16274 DWARF-2 call frame information.
16275 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16276 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16277 the backchain is placed into the topmost word of the 96/160 byte register
16280 In general, code compiled with @option{-mbackchain} is call-compatible with
16281 code compiled with @option{-mmo-backchain}; however, use of the backchain
16282 for debugging purposes usually requires that the whole binary is built with
16283 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16284 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16285 to build a linux kernel use @option{-msoft-float}.
16287 The default is to not maintain the backchain.
16289 @item -mpacked-stack
16290 @itemx -mno-packed-stack
16291 @opindex mpacked-stack
16292 @opindex mno-packed-stack
16293 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16294 specified, the compiler uses the all fields of the 96/160 byte register save
16295 area only for their default purpose; unused fields still take up stack space.
16296 When @option{-mpacked-stack} is specified, register save slots are densely
16297 packed at the top of the register save area; unused space is reused for other
16298 purposes, allowing for more efficient use of the available stack space.
16299 However, when @option{-mbackchain} is also in effect, the topmost word of
16300 the save area is always used to store the backchain, and the return address
16301 register is always saved two words below the backchain.
16303 As long as the stack frame backchain is not used, code generated with
16304 @option{-mpacked-stack} is call-compatible with code generated with
16305 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16306 S/390 or zSeries generated code that uses the stack frame backchain at run
16307 time, not just for debugging purposes. Such code is not call-compatible
16308 with code compiled with @option{-mpacked-stack}. Also, note that the
16309 combination of @option{-mbackchain},
16310 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16311 to build a linux kernel use @option{-msoft-float}.
16313 The default is to not use the packed stack layout.
16316 @itemx -mno-small-exec
16317 @opindex msmall-exec
16318 @opindex mno-small-exec
16319 Generate (or do not generate) code using the @code{bras} instruction
16320 to do subroutine calls.
16321 This only works reliably if the total executable size does not
16322 exceed 64k. The default is to use the @code{basr} instruction instead,
16323 which does not have this limitation.
16329 When @option{-m31} is specified, generate code compliant to the
16330 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16331 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16332 particular to generate 64-bit instructions. For the @samp{s390}
16333 targets, the default is @option{-m31}, while the @samp{s390x}
16334 targets default to @option{-m64}.
16340 When @option{-mzarch} is specified, generate code using the
16341 instructions available on z/Architecture.
16342 When @option{-mesa} is specified, generate code using the
16343 instructions available on ESA/390. Note that @option{-mesa} is
16344 not possible with @option{-m64}.
16345 When generating code compliant to the GNU/Linux for S/390 ABI,
16346 the default is @option{-mesa}. When generating code compliant
16347 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16353 Generate (or do not generate) code using the @code{mvcle} instruction
16354 to perform block moves. When @option{-mno-mvcle} is specified,
16355 use a @code{mvc} loop instead. This is the default unless optimizing for
16362 Print (or do not print) additional debug information when compiling.
16363 The default is to not print debug information.
16365 @item -march=@var{cpu-type}
16367 Generate code that will run on @var{cpu-type}, which is the name of a system
16368 representing a certain processor type. Possible values for
16369 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16370 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16371 When generating code using the instructions available on z/Architecture,
16372 the default is @option{-march=z900}. Otherwise, the default is
16373 @option{-march=g5}.
16375 @item -mtune=@var{cpu-type}
16377 Tune to @var{cpu-type} everything applicable about the generated code,
16378 except for the ABI and the set of available instructions.
16379 The list of @var{cpu-type} values is the same as for @option{-march}.
16380 The default is the value used for @option{-march}.
16383 @itemx -mno-tpf-trace
16384 @opindex mtpf-trace
16385 @opindex mno-tpf-trace
16386 Generate code that adds (does not add) in TPF OS specific branches to trace
16387 routines in the operating system. This option is off by default, even
16388 when compiling for the TPF OS@.
16391 @itemx -mno-fused-madd
16392 @opindex mfused-madd
16393 @opindex mno-fused-madd
16394 Generate code that uses (does not use) the floating point multiply and
16395 accumulate instructions. These instructions are generated by default if
16396 hardware floating point is used.
16398 @item -mwarn-framesize=@var{framesize}
16399 @opindex mwarn-framesize
16400 Emit a warning if the current function exceeds the given frame size. Because
16401 this is a compile time check it doesn't need to be a real problem when the program
16402 runs. It is intended to identify functions which most probably cause
16403 a stack overflow. It is useful to be used in an environment with limited stack
16404 size e.g.@: the linux kernel.
16406 @item -mwarn-dynamicstack
16407 @opindex mwarn-dynamicstack
16408 Emit a warning if the function calls alloca or uses dynamically
16409 sized arrays. This is generally a bad idea with a limited stack size.
16411 @item -mstack-guard=@var{stack-guard}
16412 @itemx -mstack-size=@var{stack-size}
16413 @opindex mstack-guard
16414 @opindex mstack-size
16415 If these options are provided the s390 back end emits additional instructions in
16416 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16417 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16418 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16419 the frame size of the compiled function is chosen.
16420 These options are intended to be used to help debugging stack overflow problems.
16421 The additionally emitted code causes only little overhead and hence can also be
16422 used in production like systems without greater performance degradation. The given
16423 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16424 @var{stack-guard} without exceeding 64k.
16425 In order to be efficient the extra code makes the assumption that the stack starts
16426 at an address aligned to the value given by @var{stack-size}.
16427 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16430 @node Score Options
16431 @subsection Score Options
16432 @cindex Score Options
16434 These options are defined for Score implementations:
16439 Compile code for big endian mode. This is the default.
16443 Compile code for little endian mode.
16447 Disable generate bcnz instruction.
16451 Enable generate unaligned load and store instruction.
16455 Enable the use of multiply-accumulate instructions. Disabled by default.
16459 Specify the SCORE5 as the target architecture.
16463 Specify the SCORE5U of the target architecture.
16467 Specify the SCORE7 as the target architecture. This is the default.
16471 Specify the SCORE7D as the target architecture.
16475 @subsection SH Options
16477 These @samp{-m} options are defined for the SH implementations:
16482 Generate code for the SH1.
16486 Generate code for the SH2.
16489 Generate code for the SH2e.
16493 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16494 that the floating-point unit is not used.
16496 @item -m2a-single-only
16497 @opindex m2a-single-only
16498 Generate code for the SH2a-FPU, in such a way that no double-precision
16499 floating point operations are used.
16502 @opindex m2a-single
16503 Generate code for the SH2a-FPU assuming the floating-point unit is in
16504 single-precision mode by default.
16508 Generate code for the SH2a-FPU assuming the floating-point unit is in
16509 double-precision mode by default.
16513 Generate code for the SH3.
16517 Generate code for the SH3e.
16521 Generate code for the SH4 without a floating-point unit.
16523 @item -m4-single-only
16524 @opindex m4-single-only
16525 Generate code for the SH4 with a floating-point unit that only
16526 supports single-precision arithmetic.
16530 Generate code for the SH4 assuming the floating-point unit is in
16531 single-precision mode by default.
16535 Generate code for the SH4.
16539 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16540 floating-point unit is not used.
16542 @item -m4a-single-only
16543 @opindex m4a-single-only
16544 Generate code for the SH4a, in such a way that no double-precision
16545 floating point operations are used.
16548 @opindex m4a-single
16549 Generate code for the SH4a assuming the floating-point unit is in
16550 single-precision mode by default.
16554 Generate code for the SH4a.
16558 Same as @option{-m4a-nofpu}, except that it implicitly passes
16559 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16560 instructions at the moment.
16564 Compile code for the processor in big endian mode.
16568 Compile code for the processor in little endian mode.
16572 Align doubles at 64-bit boundaries. Note that this changes the calling
16573 conventions, and thus some functions from the standard C library will
16574 not work unless you recompile it first with @option{-mdalign}.
16578 Shorten some address references at link time, when possible; uses the
16579 linker option @option{-relax}.
16583 Use 32-bit offsets in @code{switch} tables. The default is to use
16588 Enable the use of bit manipulation instructions on SH2A.
16592 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16593 alignment constraints.
16597 Comply with the calling conventions defined by Renesas.
16601 Comply with the calling conventions defined by Renesas.
16605 Comply with the calling conventions defined for GCC before the Renesas
16606 conventions were available. This option is the default for all
16607 targets of the SH toolchain.
16610 @opindex mnomacsave
16611 Mark the @code{MAC} register as call-clobbered, even if
16612 @option{-mhitachi} is given.
16616 Increase IEEE-compliance of floating-point code.
16617 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16618 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16619 comparisons of NANs / infinities incurs extra overhead in every
16620 floating point comparison, therefore the default is set to
16621 @option{-ffinite-math-only}.
16623 @item -minline-ic_invalidate
16624 @opindex minline-ic_invalidate
16625 Inline code to invalidate instruction cache entries after setting up
16626 nested function trampolines.
16627 This option has no effect if -musermode is in effect and the selected
16628 code generation option (e.g. -m4) does not allow the use of the icbi
16630 If the selected code generation option does not allow the use of the icbi
16631 instruction, and -musermode is not in effect, the inlined code will
16632 manipulate the instruction cache address array directly with an associative
16633 write. This not only requires privileged mode, but it will also
16634 fail if the cache line had been mapped via the TLB and has become unmapped.
16638 Dump instruction size and location in the assembly code.
16641 @opindex mpadstruct
16642 This option is deprecated. It pads structures to multiple of 4 bytes,
16643 which is incompatible with the SH ABI@.
16647 Optimize for space instead of speed. Implied by @option{-Os}.
16650 @opindex mprefergot
16651 When generating position-independent code, emit function calls using
16652 the Global Offset Table instead of the Procedure Linkage Table.
16656 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16657 if the inlined code would not work in user mode.
16658 This is the default when the target is @code{sh-*-linux*}.
16660 @item -multcost=@var{number}
16661 @opindex multcost=@var{number}
16662 Set the cost to assume for a multiply insn.
16664 @item -mdiv=@var{strategy}
16665 @opindex mdiv=@var{strategy}
16666 Set the division strategy to use for SHmedia code. @var{strategy} must be
16667 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16668 inv:call2, inv:fp .
16669 "fp" performs the operation in floating point. This has a very high latency,
16670 but needs only a few instructions, so it might be a good choice if
16671 your code has enough easily exploitable ILP to allow the compiler to
16672 schedule the floating point instructions together with other instructions.
16673 Division by zero causes a floating point exception.
16674 "inv" uses integer operations to calculate the inverse of the divisor,
16675 and then multiplies the dividend with the inverse. This strategy allows
16676 cse and hoisting of the inverse calculation. Division by zero calculates
16677 an unspecified result, but does not trap.
16678 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16679 have been found, or if the entire operation has been hoisted to the same
16680 place, the last stages of the inverse calculation are intertwined with the
16681 final multiply to reduce the overall latency, at the expense of using a few
16682 more instructions, and thus offering fewer scheduling opportunities with
16684 "call" calls a library function that usually implements the inv:minlat
16686 This gives high code density for m5-*media-nofpu compilations.
16687 "call2" uses a different entry point of the same library function, where it
16688 assumes that a pointer to a lookup table has already been set up, which
16689 exposes the pointer load to cse / code hoisting optimizations.
16690 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16691 code generation, but if the code stays unoptimized, revert to the "call",
16692 "call2", or "fp" strategies, respectively. Note that the
16693 potentially-trapping side effect of division by zero is carried by a
16694 separate instruction, so it is possible that all the integer instructions
16695 are hoisted out, but the marker for the side effect stays where it is.
16696 A recombination to fp operations or a call is not possible in that case.
16697 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16698 that the inverse calculation was nor separated from the multiply, they speed
16699 up division where the dividend fits into 20 bits (plus sign where applicable),
16700 by inserting a test to skip a number of operations in this case; this test
16701 slows down the case of larger dividends. inv20u assumes the case of a such
16702 a small dividend to be unlikely, and inv20l assumes it to be likely.
16704 @item -maccumulate-outgoing-args
16705 @opindex maccumulate-outgoing-args
16706 Reserve space once for outgoing arguments in the function prologue rather
16707 than around each call. Generally beneficial for performance and size. Also
16708 needed for unwinding to avoid changing the stack frame around conditional code.
16710 @item -mdivsi3_libfunc=@var{name}
16711 @opindex mdivsi3_libfunc=@var{name}
16712 Set the name of the library function used for 32 bit signed division to
16713 @var{name}. This only affect the name used in the call and inv:call
16714 division strategies, and the compiler will still expect the same
16715 sets of input/output/clobbered registers as if this option was not present.
16717 @item -mfixed-range=@var{register-range}
16718 @opindex mfixed-range
16719 Generate code treating the given register range as fixed registers.
16720 A fixed register is one that the register allocator can not use. This is
16721 useful when compiling kernel code. A register range is specified as
16722 two registers separated by a dash. Multiple register ranges can be
16723 specified separated by a comma.
16725 @item -madjust-unroll
16726 @opindex madjust-unroll
16727 Throttle unrolling to avoid thrashing target registers.
16728 This option only has an effect if the gcc code base supports the
16729 TARGET_ADJUST_UNROLL_MAX target hook.
16731 @item -mindexed-addressing
16732 @opindex mindexed-addressing
16733 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16734 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16735 semantics for the indexed addressing mode. The architecture allows the
16736 implementation of processors with 64 bit MMU, which the OS could use to
16737 get 32 bit addressing, but since no current hardware implementation supports
16738 this or any other way to make the indexed addressing mode safe to use in
16739 the 32 bit ABI, the default is -mno-indexed-addressing.
16741 @item -mgettrcost=@var{number}
16742 @opindex mgettrcost=@var{number}
16743 Set the cost assumed for the gettr instruction to @var{number}.
16744 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16748 Assume pt* instructions won't trap. This will generally generate better
16749 scheduled code, but is unsafe on current hardware. The current architecture
16750 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16751 This has the unintentional effect of making it unsafe to schedule ptabs /
16752 ptrel before a branch, or hoist it out of a loop. For example,
16753 __do_global_ctors, a part of libgcc that runs constructors at program
16754 startup, calls functions in a list which is delimited by @minus{}1. With the
16755 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16756 That means that all the constructors will be run a bit quicker, but when
16757 the loop comes to the end of the list, the program crashes because ptabs
16758 loads @minus{}1 into a target register. Since this option is unsafe for any
16759 hardware implementing the current architecture specification, the default
16760 is -mno-pt-fixed. Unless the user specifies a specific cost with
16761 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16762 this deters register allocation using target registers for storing
16765 @item -minvalid-symbols
16766 @opindex minvalid-symbols
16767 Assume symbols might be invalid. Ordinary function symbols generated by
16768 the compiler will always be valid to load with movi/shori/ptabs or
16769 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16770 to generate symbols that will cause ptabs / ptrel to trap.
16771 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16772 It will then prevent cross-basic-block cse, hoisting and most scheduling
16773 of symbol loads. The default is @option{-mno-invalid-symbols}.
16776 @node Solaris 2 Options
16777 @subsection Solaris 2 Options
16778 @cindex Solaris 2 options
16780 These @samp{-m} options are supported on Solaris 2:
16783 @item -mimpure-text
16784 @opindex mimpure-text
16785 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16786 the compiler to not pass @option{-z text} to the linker when linking a
16787 shared object. Using this option, you can link position-dependent
16788 code into a shared object.
16790 @option{-mimpure-text} suppresses the ``relocations remain against
16791 allocatable but non-writable sections'' linker error message.
16792 However, the necessary relocations will trigger copy-on-write, and the
16793 shared object is not actually shared across processes. Instead of
16794 using @option{-mimpure-text}, you should compile all source code with
16795 @option{-fpic} or @option{-fPIC}.
16799 These switches are supported in addition to the above on Solaris 2:
16804 Add support for multithreading using the POSIX threads library. This
16805 option sets flags for both the preprocessor and linker. This option does
16806 not affect the thread safety of object code produced by the compiler or
16807 that of libraries supplied with it.
16811 This is a synonym for @option{-pthreads}.
16814 @node SPARC Options
16815 @subsection SPARC Options
16816 @cindex SPARC options
16818 These @samp{-m} options are supported on the SPARC:
16821 @item -mno-app-regs
16823 @opindex mno-app-regs
16825 Specify @option{-mapp-regs} to generate output using the global registers
16826 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16829 To be fully SVR4 ABI compliant at the cost of some performance loss,
16830 specify @option{-mno-app-regs}. You should compile libraries and system
16831 software with this option.
16834 @itemx -mhard-float
16836 @opindex mhard-float
16837 Generate output containing floating point instructions. This is the
16841 @itemx -msoft-float
16843 @opindex msoft-float
16844 Generate output containing library calls for floating point.
16845 @strong{Warning:} the requisite libraries are not available for all SPARC
16846 targets. Normally the facilities of the machine's usual C compiler are
16847 used, but this cannot be done directly in cross-compilation. You must make
16848 your own arrangements to provide suitable library functions for
16849 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16850 @samp{sparclite-*-*} do provide software floating point support.
16852 @option{-msoft-float} changes the calling convention in the output file;
16853 therefore, it is only useful if you compile @emph{all} of a program with
16854 this option. In particular, you need to compile @file{libgcc.a}, the
16855 library that comes with GCC, with @option{-msoft-float} in order for
16858 @item -mhard-quad-float
16859 @opindex mhard-quad-float
16860 Generate output containing quad-word (long double) floating point
16863 @item -msoft-quad-float
16864 @opindex msoft-quad-float
16865 Generate output containing library calls for quad-word (long double)
16866 floating point instructions. The functions called are those specified
16867 in the SPARC ABI@. This is the default.
16869 As of this writing, there are no SPARC implementations that have hardware
16870 support for the quad-word floating point instructions. They all invoke
16871 a trap handler for one of these instructions, and then the trap handler
16872 emulates the effect of the instruction. Because of the trap handler overhead,
16873 this is much slower than calling the ABI library routines. Thus the
16874 @option{-msoft-quad-float} option is the default.
16876 @item -mno-unaligned-doubles
16877 @itemx -munaligned-doubles
16878 @opindex mno-unaligned-doubles
16879 @opindex munaligned-doubles
16880 Assume that doubles have 8 byte alignment. This is the default.
16882 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16883 alignment only if they are contained in another type, or if they have an
16884 absolute address. Otherwise, it assumes they have 4 byte alignment.
16885 Specifying this option avoids some rare compatibility problems with code
16886 generated by other compilers. It is not the default because it results
16887 in a performance loss, especially for floating point code.
16889 @item -mno-faster-structs
16890 @itemx -mfaster-structs
16891 @opindex mno-faster-structs
16892 @opindex mfaster-structs
16893 With @option{-mfaster-structs}, the compiler assumes that structures
16894 should have 8 byte alignment. This enables the use of pairs of
16895 @code{ldd} and @code{std} instructions for copies in structure
16896 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16897 However, the use of this changed alignment directly violates the SPARC
16898 ABI@. Thus, it's intended only for use on targets where the developer
16899 acknowledges that their resulting code will not be directly in line with
16900 the rules of the ABI@.
16902 @item -mcpu=@var{cpu_type}
16904 Set the instruction set, register set, and instruction scheduling parameters
16905 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16906 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
16907 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
16908 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16909 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16911 Default instruction scheduling parameters are used for values that select
16912 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16913 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16915 Here is a list of each supported architecture and their supported
16920 v8: supersparc, hypersparc, leon
16921 sparclite: f930, f934, sparclite86x
16923 v9: ultrasparc, ultrasparc3, niagara, niagara2
16926 By default (unless configured otherwise), GCC generates code for the V7
16927 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16928 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16929 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16930 SPARCStation 1, 2, IPX etc.
16932 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16933 architecture. The only difference from V7 code is that the compiler emits
16934 the integer multiply and integer divide instructions which exist in SPARC-V8
16935 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16936 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16939 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16940 the SPARC architecture. This adds the integer multiply, integer divide step
16941 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16942 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16943 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16944 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16945 MB86934 chip, which is the more recent SPARClite with FPU@.
16947 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16948 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16949 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16950 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16951 optimizes it for the TEMIC SPARClet chip.
16953 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16954 architecture. This adds 64-bit integer and floating-point move instructions,
16955 3 additional floating-point condition code registers and conditional move
16956 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16957 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16958 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16959 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16960 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16961 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16962 additionally optimizes it for Sun UltraSPARC T2 chips.
16964 @item -mtune=@var{cpu_type}
16966 Set the instruction scheduling parameters for machine type
16967 @var{cpu_type}, but do not set the instruction set or register set that the
16968 option @option{-mcpu=@var{cpu_type}} would.
16970 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16971 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16972 that select a particular CPU implementation. Those are @samp{cypress},
16973 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
16974 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
16975 @samp{niagara}, and @samp{niagara2}.
16980 @opindex mno-v8plus
16981 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16982 difference from the V8 ABI is that the global and out registers are
16983 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16984 mode for all SPARC-V9 processors.
16990 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16991 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16994 These @samp{-m} options are supported in addition to the above
16995 on SPARC-V9 processors in 64-bit environments:
16998 @item -mlittle-endian
16999 @opindex mlittle-endian
17000 Generate code for a processor running in little-endian mode. It is only
17001 available for a few configurations and most notably not on Solaris and Linux.
17007 Generate code for a 32-bit or 64-bit environment.
17008 The 32-bit environment sets int, long and pointer to 32 bits.
17009 The 64-bit environment sets int to 32 bits and long and pointer
17012 @item -mcmodel=medlow
17013 @opindex mcmodel=medlow
17014 Generate code for the Medium/Low code model: 64-bit addresses, programs
17015 must be linked in the low 32 bits of memory. Programs can be statically
17016 or dynamically linked.
17018 @item -mcmodel=medmid
17019 @opindex mcmodel=medmid
17020 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17021 must be linked in the low 44 bits of memory, the text and data segments must
17022 be less than 2GB in size and the data segment must be located within 2GB of
17025 @item -mcmodel=medany
17026 @opindex mcmodel=medany
17027 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17028 may be linked anywhere in memory, the text and data segments must be less
17029 than 2GB in size and the data segment must be located within 2GB of the
17032 @item -mcmodel=embmedany
17033 @opindex mcmodel=embmedany
17034 Generate code for the Medium/Anywhere code model for embedded systems:
17035 64-bit addresses, the text and data segments must be less than 2GB in
17036 size, both starting anywhere in memory (determined at link time). The
17037 global register %g4 points to the base of the data segment. Programs
17038 are statically linked and PIC is not supported.
17041 @itemx -mno-stack-bias
17042 @opindex mstack-bias
17043 @opindex mno-stack-bias
17044 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17045 frame pointer if present, are offset by @minus{}2047 which must be added back
17046 when making stack frame references. This is the default in 64-bit mode.
17047 Otherwise, assume no such offset is present.
17051 @subsection SPU Options
17052 @cindex SPU options
17054 These @samp{-m} options are supported on the SPU:
17058 @itemx -merror-reloc
17059 @opindex mwarn-reloc
17060 @opindex merror-reloc
17062 The loader for SPU does not handle dynamic relocations. By default, GCC
17063 will give an error when it generates code that requires a dynamic
17064 relocation. @option{-mno-error-reloc} disables the error,
17065 @option{-mwarn-reloc} will generate a warning instead.
17068 @itemx -munsafe-dma
17070 @opindex munsafe-dma
17072 Instructions which initiate or test completion of DMA must not be
17073 reordered with respect to loads and stores of the memory which is being
17074 accessed. Users typically address this problem using the volatile
17075 keyword, but that can lead to inefficient code in places where the
17076 memory is known to not change. Rather than mark the memory as volatile
17077 we treat the DMA instructions as potentially effecting all memory. With
17078 @option{-munsafe-dma} users must use the volatile keyword to protect
17081 @item -mbranch-hints
17082 @opindex mbranch-hints
17084 By default, GCC will generate a branch hint instruction to avoid
17085 pipeline stalls for always taken or probably taken branches. A hint
17086 will not be generated closer than 8 instructions away from its branch.
17087 There is little reason to disable them, except for debugging purposes,
17088 or to make an object a little bit smaller.
17092 @opindex msmall-mem
17093 @opindex mlarge-mem
17095 By default, GCC generates code assuming that addresses are never larger
17096 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17097 a full 32 bit address.
17102 By default, GCC links against startup code that assumes the SPU-style
17103 main function interface (which has an unconventional parameter list).
17104 With @option{-mstdmain}, GCC will link your program against startup
17105 code that assumes a C99-style interface to @code{main}, including a
17106 local copy of @code{argv} strings.
17108 @item -mfixed-range=@var{register-range}
17109 @opindex mfixed-range
17110 Generate code treating the given register range as fixed registers.
17111 A fixed register is one that the register allocator can not use. This is
17112 useful when compiling kernel code. A register range is specified as
17113 two registers separated by a dash. Multiple register ranges can be
17114 specified separated by a comma.
17120 Compile code assuming that pointers to the PPU address space accessed
17121 via the @code{__ea} named address space qualifier are either 32 or 64
17122 bits wide. The default is 32 bits. As this is an ABI changing option,
17123 all object code in an executable must be compiled with the same setting.
17125 @item -maddress-space-conversion
17126 @itemx -mno-address-space-conversion
17127 @opindex maddress-space-conversion
17128 @opindex mno-address-space-conversion
17129 Allow/disallow treating the @code{__ea} address space as superset
17130 of the generic address space. This enables explicit type casts
17131 between @code{__ea} and generic pointer as well as implicit
17132 conversions of generic pointers to @code{__ea} pointers. The
17133 default is to allow address space pointer conversions.
17135 @item -mcache-size=@var{cache-size}
17136 @opindex mcache-size
17137 This option controls the version of libgcc that the compiler links to an
17138 executable and selects a software-managed cache for accessing variables
17139 in the @code{__ea} address space with a particular cache size. Possible
17140 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17141 and @samp{128}. The default cache size is 64KB.
17143 @item -matomic-updates
17144 @itemx -mno-atomic-updates
17145 @opindex matomic-updates
17146 @opindex mno-atomic-updates
17147 This option controls the version of libgcc that the compiler links to an
17148 executable and selects whether atomic updates to the software-managed
17149 cache of PPU-side variables are used. If you use atomic updates, changes
17150 to a PPU variable from SPU code using the @code{__ea} named address space
17151 qualifier will not interfere with changes to other PPU variables residing
17152 in the same cache line from PPU code. If you do not use atomic updates,
17153 such interference may occur; however, writing back cache lines will be
17154 more efficient. The default behavior is to use atomic updates.
17157 @itemx -mdual-nops=@var{n}
17158 @opindex mdual-nops
17159 By default, GCC will insert nops to increase dual issue when it expects
17160 it to increase performance. @var{n} can be a value from 0 to 10. A
17161 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17162 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17164 @item -mhint-max-nops=@var{n}
17165 @opindex mhint-max-nops
17166 Maximum number of nops to insert for a branch hint. A branch hint must
17167 be at least 8 instructions away from the branch it is effecting. GCC
17168 will insert up to @var{n} nops to enforce this, otherwise it will not
17169 generate the branch hint.
17171 @item -mhint-max-distance=@var{n}
17172 @opindex mhint-max-distance
17173 The encoding of the branch hint instruction limits the hint to be within
17174 256 instructions of the branch it is effecting. By default, GCC makes
17175 sure it is within 125.
17178 @opindex msafe-hints
17179 Work around a hardware bug which causes the SPU to stall indefinitely.
17180 By default, GCC will insert the @code{hbrp} instruction to make sure
17181 this stall won't happen.
17185 @node System V Options
17186 @subsection Options for System V
17188 These additional options are available on System V Release 4 for
17189 compatibility with other compilers on those systems:
17194 Create a shared object.
17195 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17199 Identify the versions of each tool used by the compiler, in a
17200 @code{.ident} assembler directive in the output.
17204 Refrain from adding @code{.ident} directives to the output file (this is
17207 @item -YP,@var{dirs}
17209 Search the directories @var{dirs}, and no others, for libraries
17210 specified with @option{-l}.
17212 @item -Ym,@var{dir}
17214 Look in the directory @var{dir} to find the M4 preprocessor.
17215 The assembler uses this option.
17216 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17217 @c the generic assembler that comes with Solaris takes just -Ym.
17221 @subsection V850 Options
17222 @cindex V850 Options
17224 These @samp{-m} options are defined for V850 implementations:
17228 @itemx -mno-long-calls
17229 @opindex mlong-calls
17230 @opindex mno-long-calls
17231 Treat all calls as being far away (near). If calls are assumed to be
17232 far away, the compiler will always load the functions address up into a
17233 register, and call indirect through the pointer.
17239 Do not optimize (do optimize) basic blocks that use the same index
17240 pointer 4 or more times to copy pointer into the @code{ep} register, and
17241 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17242 option is on by default if you optimize.
17244 @item -mno-prolog-function
17245 @itemx -mprolog-function
17246 @opindex mno-prolog-function
17247 @opindex mprolog-function
17248 Do not use (do use) external functions to save and restore registers
17249 at the prologue and epilogue of a function. The external functions
17250 are slower, but use less code space if more than one function saves
17251 the same number of registers. The @option{-mprolog-function} option
17252 is on by default if you optimize.
17256 Try to make the code as small as possible. At present, this just turns
17257 on the @option{-mep} and @option{-mprolog-function} options.
17259 @item -mtda=@var{n}
17261 Put static or global variables whose size is @var{n} bytes or less into
17262 the tiny data area that register @code{ep} points to. The tiny data
17263 area can hold up to 256 bytes in total (128 bytes for byte references).
17265 @item -msda=@var{n}
17267 Put static or global variables whose size is @var{n} bytes or less into
17268 the small data area that register @code{gp} points to. The small data
17269 area can hold up to 64 kilobytes.
17271 @item -mzda=@var{n}
17273 Put static or global variables whose size is @var{n} bytes or less into
17274 the first 32 kilobytes of memory.
17278 Specify that the target processor is the V850.
17281 @opindex mbig-switch
17282 Generate code suitable for big switch tables. Use this option only if
17283 the assembler/linker complain about out of range branches within a switch
17288 This option will cause r2 and r5 to be used in the code generated by
17289 the compiler. This setting is the default.
17291 @item -mno-app-regs
17292 @opindex mno-app-regs
17293 This option will cause r2 and r5 to be treated as fixed registers.
17297 Specify that the target processor is the V850E2V3. The preprocessor
17298 constants @samp{__v850e2v3__} will be defined if
17299 this option is used.
17303 Specify that the target processor is the V850E2. The preprocessor
17304 constants @samp{__v850e2__} will be defined if this option is used.
17308 Specify that the target processor is the V850E1. The preprocessor
17309 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17310 this option is used.
17314 Specify that the target processor is the V850ES. This is an alias for
17315 the @option{-mv850e1} option.
17319 Specify that the target processor is the V850E@. The preprocessor
17320 constant @samp{__v850e__} will be defined if this option is used.
17322 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17323 nor @option{-mv850e2} nor @option{-mv850e2v3}
17324 are defined then a default target processor will be chosen and the
17325 relevant @samp{__v850*__} preprocessor constant will be defined.
17327 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17328 defined, regardless of which processor variant is the target.
17330 @item -mdisable-callt
17331 @opindex mdisable-callt
17332 This option will suppress generation of the CALLT instruction for the
17333 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17334 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17339 @subsection VAX Options
17340 @cindex VAX options
17342 These @samp{-m} options are defined for the VAX:
17347 Do not output certain jump instructions (@code{aobleq} and so on)
17348 that the Unix assembler for the VAX cannot handle across long
17353 Do output those jump instructions, on the assumption that you
17354 will assemble with the GNU assembler.
17358 Output code for g-format floating point numbers instead of d-format.
17361 @node VxWorks Options
17362 @subsection VxWorks Options
17363 @cindex VxWorks Options
17365 The options in this section are defined for all VxWorks targets.
17366 Options specific to the target hardware are listed with the other
17367 options for that target.
17372 GCC can generate code for both VxWorks kernels and real time processes
17373 (RTPs). This option switches from the former to the latter. It also
17374 defines the preprocessor macro @code{__RTP__}.
17377 @opindex non-static
17378 Link an RTP executable against shared libraries rather than static
17379 libraries. The options @option{-static} and @option{-shared} can
17380 also be used for RTPs (@pxref{Link Options}); @option{-static}
17387 These options are passed down to the linker. They are defined for
17388 compatibility with Diab.
17391 @opindex Xbind-lazy
17392 Enable lazy binding of function calls. This option is equivalent to
17393 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17397 Disable lazy binding of function calls. This option is the default and
17398 is defined for compatibility with Diab.
17401 @node x86-64 Options
17402 @subsection x86-64 Options
17403 @cindex x86-64 options
17405 These are listed under @xref{i386 and x86-64 Options}.
17407 @node i386 and x86-64 Windows Options
17408 @subsection i386 and x86-64 Windows Options
17409 @cindex i386 and x86-64 Windows Options
17411 These additional options are available for Windows targets:
17416 This option is available for Cygwin and MinGW targets. It
17417 specifies that a console application is to be generated, by
17418 instructing the linker to set the PE header subsystem type
17419 required for console applications.
17420 This is the default behavior for Cygwin and MinGW targets.
17424 This option is available for Cygwin and MinGW targets. It
17425 specifies that a DLL - a dynamic link library - is to be
17426 generated, enabling the selection of the required runtime
17427 startup object and entry point.
17429 @item -mnop-fun-dllimport
17430 @opindex mnop-fun-dllimport
17431 This option is available for Cygwin and MinGW targets. It
17432 specifies that the dllimport attribute should be ignored.
17436 This option is available for MinGW targets. It specifies
17437 that MinGW-specific thread support is to be used.
17441 This option is available for mingw-w64 targets. It specifies
17442 that the UNICODE macro is getting pre-defined and that the
17443 unicode capable runtime startup code is chosen.
17447 This option is available for Cygwin and MinGW targets. It
17448 specifies that the typical Windows pre-defined macros are to
17449 be set in the pre-processor, but does not influence the choice
17450 of runtime library/startup code.
17454 This option is available for Cygwin and MinGW targets. It
17455 specifies that a GUI application is to be generated by
17456 instructing the linker to set the PE header subsystem type
17459 @item -fno-set-stack-executable
17460 @opindex fno-set-stack-executable
17461 This option is available for MinGW targets. It specifies that
17462 the executable flag for stack used by nested functions isn't
17463 set. This is necessary for binaries running in kernel mode of
17464 Windows, as there the user32 API, which is used to set executable
17465 privileges, isn't available.
17467 @item -mpe-aligned-commons
17468 @opindex mpe-aligned-commons
17469 This option is available for Cygwin and MinGW targets. It
17470 specifies that the GNU extension to the PE file format that
17471 permits the correct alignment of COMMON variables should be
17472 used when generating code. It will be enabled by default if
17473 GCC detects that the target assembler found during configuration
17474 supports the feature.
17477 See also under @ref{i386 and x86-64 Options} for standard options.
17479 @node Xstormy16 Options
17480 @subsection Xstormy16 Options
17481 @cindex Xstormy16 Options
17483 These options are defined for Xstormy16:
17488 Choose startup files and linker script suitable for the simulator.
17491 @node Xtensa Options
17492 @subsection Xtensa Options
17493 @cindex Xtensa Options
17495 These options are supported for Xtensa targets:
17499 @itemx -mno-const16
17501 @opindex mno-const16
17502 Enable or disable use of @code{CONST16} instructions for loading
17503 constant values. The @code{CONST16} instruction is currently not a
17504 standard option from Tensilica. When enabled, @code{CONST16}
17505 instructions are always used in place of the standard @code{L32R}
17506 instructions. The use of @code{CONST16} is enabled by default only if
17507 the @code{L32R} instruction is not available.
17510 @itemx -mno-fused-madd
17511 @opindex mfused-madd
17512 @opindex mno-fused-madd
17513 Enable or disable use of fused multiply/add and multiply/subtract
17514 instructions in the floating-point option. This has no effect if the
17515 floating-point option is not also enabled. Disabling fused multiply/add
17516 and multiply/subtract instructions forces the compiler to use separate
17517 instructions for the multiply and add/subtract operations. This may be
17518 desirable in some cases where strict IEEE 754-compliant results are
17519 required: the fused multiply add/subtract instructions do not round the
17520 intermediate result, thereby producing results with @emph{more} bits of
17521 precision than specified by the IEEE standard. Disabling fused multiply
17522 add/subtract instructions also ensures that the program output is not
17523 sensitive to the compiler's ability to combine multiply and add/subtract
17526 @item -mserialize-volatile
17527 @itemx -mno-serialize-volatile
17528 @opindex mserialize-volatile
17529 @opindex mno-serialize-volatile
17530 When this option is enabled, GCC inserts @code{MEMW} instructions before
17531 @code{volatile} memory references to guarantee sequential consistency.
17532 The default is @option{-mserialize-volatile}. Use
17533 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17535 @item -mforce-no-pic
17536 @opindex mforce-no-pic
17537 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17538 position-independent code (PIC), this option disables PIC for compiling
17541 @item -mtext-section-literals
17542 @itemx -mno-text-section-literals
17543 @opindex mtext-section-literals
17544 @opindex mno-text-section-literals
17545 Control the treatment of literal pools. The default is
17546 @option{-mno-text-section-literals}, which places literals in a separate
17547 section in the output file. This allows the literal pool to be placed
17548 in a data RAM/ROM, and it also allows the linker to combine literal
17549 pools from separate object files to remove redundant literals and
17550 improve code size. With @option{-mtext-section-literals}, the literals
17551 are interspersed in the text section in order to keep them as close as
17552 possible to their references. This may be necessary for large assembly
17555 @item -mtarget-align
17556 @itemx -mno-target-align
17557 @opindex mtarget-align
17558 @opindex mno-target-align
17559 When this option is enabled, GCC instructs the assembler to
17560 automatically align instructions to reduce branch penalties at the
17561 expense of some code density. The assembler attempts to widen density
17562 instructions to align branch targets and the instructions following call
17563 instructions. If there are not enough preceding safe density
17564 instructions to align a target, no widening will be performed. The
17565 default is @option{-mtarget-align}. These options do not affect the
17566 treatment of auto-aligned instructions like @code{LOOP}, which the
17567 assembler will always align, either by widening density instructions or
17568 by inserting no-op instructions.
17571 @itemx -mno-longcalls
17572 @opindex mlongcalls
17573 @opindex mno-longcalls
17574 When this option is enabled, GCC instructs the assembler to translate
17575 direct calls to indirect calls unless it can determine that the target
17576 of a direct call is in the range allowed by the call instruction. This
17577 translation typically occurs for calls to functions in other source
17578 files. Specifically, the assembler translates a direct @code{CALL}
17579 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17580 The default is @option{-mno-longcalls}. This option should be used in
17581 programs where the call target can potentially be out of range. This
17582 option is implemented in the assembler, not the compiler, so the
17583 assembly code generated by GCC will still show direct call
17584 instructions---look at the disassembled object code to see the actual
17585 instructions. Note that the assembler will use an indirect call for
17586 every cross-file call, not just those that really will be out of range.
17589 @node zSeries Options
17590 @subsection zSeries Options
17591 @cindex zSeries options
17593 These are listed under @xref{S/390 and zSeries Options}.
17595 @node Code Gen Options
17596 @section Options for Code Generation Conventions
17597 @cindex code generation conventions
17598 @cindex options, code generation
17599 @cindex run-time options
17601 These machine-independent options control the interface conventions
17602 used in code generation.
17604 Most of them have both positive and negative forms; the negative form
17605 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17606 one of the forms is listed---the one which is not the default. You
17607 can figure out the other form by either removing @samp{no-} or adding
17611 @item -fbounds-check
17612 @opindex fbounds-check
17613 For front-ends that support it, generate additional code to check that
17614 indices used to access arrays are within the declared range. This is
17615 currently only supported by the Java and Fortran front-ends, where
17616 this option defaults to true and false respectively.
17620 This option generates traps for signed overflow on addition, subtraction,
17621 multiplication operations.
17625 This option instructs the compiler to assume that signed arithmetic
17626 overflow of addition, subtraction and multiplication wraps around
17627 using twos-complement representation. This flag enables some optimizations
17628 and disables others. This option is enabled by default for the Java
17629 front-end, as required by the Java language specification.
17632 @opindex fexceptions
17633 Enable exception handling. Generates extra code needed to propagate
17634 exceptions. For some targets, this implies GCC will generate frame
17635 unwind information for all functions, which can produce significant data
17636 size overhead, although it does not affect execution. If you do not
17637 specify this option, GCC will enable it by default for languages like
17638 C++ which normally require exception handling, and disable it for
17639 languages like C that do not normally require it. However, you may need
17640 to enable this option when compiling C code that needs to interoperate
17641 properly with exception handlers written in C++. You may also wish to
17642 disable this option if you are compiling older C++ programs that don't
17643 use exception handling.
17645 @item -fnon-call-exceptions
17646 @opindex fnon-call-exceptions
17647 Generate code that allows trapping instructions to throw exceptions.
17648 Note that this requires platform-specific runtime support that does
17649 not exist everywhere. Moreover, it only allows @emph{trapping}
17650 instructions to throw exceptions, i.e.@: memory references or floating
17651 point instructions. It does not allow exceptions to be thrown from
17652 arbitrary signal handlers such as @code{SIGALRM}.
17654 @item -funwind-tables
17655 @opindex funwind-tables
17656 Similar to @option{-fexceptions}, except that it will just generate any needed
17657 static data, but will not affect the generated code in any other way.
17658 You will normally not enable this option; instead, a language processor
17659 that needs this handling would enable it on your behalf.
17661 @item -fasynchronous-unwind-tables
17662 @opindex fasynchronous-unwind-tables
17663 Generate unwind table in dwarf2 format, if supported by target machine. The
17664 table is exact at each instruction boundary, so it can be used for stack
17665 unwinding from asynchronous events (such as debugger or garbage collector).
17667 @item -fpcc-struct-return
17668 @opindex fpcc-struct-return
17669 Return ``short'' @code{struct} and @code{union} values in memory like
17670 longer ones, rather than in registers. This convention is less
17671 efficient, but it has the advantage of allowing intercallability between
17672 GCC-compiled files and files compiled with other compilers, particularly
17673 the Portable C Compiler (pcc).
17675 The precise convention for returning structures in memory depends
17676 on the target configuration macros.
17678 Short structures and unions are those whose size and alignment match
17679 that of some integer type.
17681 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17682 switch is not binary compatible with code compiled with the
17683 @option{-freg-struct-return} switch.
17684 Use it to conform to a non-default application binary interface.
17686 @item -freg-struct-return
17687 @opindex freg-struct-return
17688 Return @code{struct} and @code{union} values in registers when possible.
17689 This is more efficient for small structures than
17690 @option{-fpcc-struct-return}.
17692 If you specify neither @option{-fpcc-struct-return} nor
17693 @option{-freg-struct-return}, GCC defaults to whichever convention is
17694 standard for the target. If there is no standard convention, GCC
17695 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17696 the principal compiler. In those cases, we can choose the standard, and
17697 we chose the more efficient register return alternative.
17699 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17700 switch is not binary compatible with code compiled with the
17701 @option{-fpcc-struct-return} switch.
17702 Use it to conform to a non-default application binary interface.
17704 @item -fshort-enums
17705 @opindex fshort-enums
17706 Allocate to an @code{enum} type only as many bytes as it needs for the
17707 declared range of possible values. Specifically, the @code{enum} type
17708 will be equivalent to the smallest integer type which has enough room.
17710 @strong{Warning:} the @option{-fshort-enums} 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-double
17715 @opindex fshort-double
17716 Use the same size for @code{double} as for @code{float}.
17718 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17719 code that is not binary compatible with code generated without that switch.
17720 Use it to conform to a non-default application binary interface.
17722 @item -fshort-wchar
17723 @opindex fshort-wchar
17724 Override the underlying type for @samp{wchar_t} to be @samp{short
17725 unsigned int} instead of the default for the target. This option is
17726 useful for building programs to run under WINE@.
17728 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17729 code that is not binary compatible with code generated without that switch.
17730 Use it to conform to a non-default application binary interface.
17733 @opindex fno-common
17734 In C code, controls the placement of uninitialized global variables.
17735 Unix C compilers have traditionally permitted multiple definitions of
17736 such variables in different compilation units by placing the variables
17738 This is the behavior specified by @option{-fcommon}, and is the default
17739 for GCC on most targets.
17740 On the other hand, this behavior is not required by ISO C, and on some
17741 targets may carry a speed or code size penalty on variable references.
17742 The @option{-fno-common} option specifies that the compiler should place
17743 uninitialized global variables in the data section of the object file,
17744 rather than generating them as common blocks.
17745 This has the effect that if the same variable is declared
17746 (without @code{extern}) in two different compilations,
17747 you will get a multiple-definition error when you link them.
17748 In this case, you must compile with @option{-fcommon} instead.
17749 Compiling with @option{-fno-common} is useful on targets for which
17750 it provides better performance, or if you wish to verify that the
17751 program will work on other systems which always treat uninitialized
17752 variable declarations this way.
17756 Ignore the @samp{#ident} directive.
17758 @item -finhibit-size-directive
17759 @opindex finhibit-size-directive
17760 Don't output a @code{.size} assembler directive, or anything else that
17761 would cause trouble if the function is split in the middle, and the
17762 two halves are placed at locations far apart in memory. This option is
17763 used when compiling @file{crtstuff.c}; you should not need to use it
17766 @item -fverbose-asm
17767 @opindex fverbose-asm
17768 Put extra commentary information in the generated assembly code to
17769 make it more readable. This option is generally only of use to those
17770 who actually need to read the generated assembly code (perhaps while
17771 debugging the compiler itself).
17773 @option{-fno-verbose-asm}, the default, causes the
17774 extra information to be omitted and is useful when comparing two assembler
17777 @item -frecord-gcc-switches
17778 @opindex frecord-gcc-switches
17779 This switch causes the command line that was used to invoke the
17780 compiler to be recorded into the object file that is being created.
17781 This switch is only implemented on some targets and the exact format
17782 of the recording is target and binary file format dependent, but it
17783 usually takes the form of a section containing ASCII text. This
17784 switch is related to the @option{-fverbose-asm} switch, but that
17785 switch only records information in the assembler output file as
17786 comments, so it never reaches the object file.
17790 @cindex global offset table
17792 Generate position-independent code (PIC) suitable for use in a shared
17793 library, if supported for the target machine. Such code accesses all
17794 constant addresses through a global offset table (GOT)@. The dynamic
17795 loader resolves the GOT entries when the program starts (the dynamic
17796 loader is not part of GCC; it is part of the operating system). If
17797 the GOT size for the linked executable exceeds a machine-specific
17798 maximum size, you get an error message from the linker indicating that
17799 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17800 instead. (These maximums are 8k on the SPARC and 32k
17801 on the m68k and RS/6000. The 386 has no such limit.)
17803 Position-independent code requires special support, and therefore works
17804 only on certain machines. For the 386, GCC supports PIC for System V
17805 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17806 position-independent.
17808 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17813 If supported for the target machine, emit position-independent code,
17814 suitable for dynamic linking and avoiding any limit on the size of the
17815 global offset table. This option makes a difference on the m68k,
17816 PowerPC and SPARC@.
17818 Position-independent code requires special support, and therefore works
17819 only on certain machines.
17821 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17828 These options are similar to @option{-fpic} and @option{-fPIC}, but
17829 generated position independent code can be only linked into executables.
17830 Usually these options are used when @option{-pie} GCC option will be
17831 used during linking.
17833 @option{-fpie} and @option{-fPIE} both define the macros
17834 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17835 for @option{-fpie} and 2 for @option{-fPIE}.
17837 @item -fno-jump-tables
17838 @opindex fno-jump-tables
17839 Do not use jump tables for switch statements even where it would be
17840 more efficient than other code generation strategies. This option is
17841 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17842 building code which forms part of a dynamic linker and cannot
17843 reference the address of a jump table. On some targets, jump tables
17844 do not require a GOT and this option is not needed.
17846 @item -ffixed-@var{reg}
17848 Treat the register named @var{reg} as a fixed register; generated code
17849 should never refer to it (except perhaps as a stack pointer, frame
17850 pointer or in some other fixed role).
17852 @var{reg} must be the name of a register. The register names accepted
17853 are machine-specific and are defined in the @code{REGISTER_NAMES}
17854 macro in the machine description macro file.
17856 This flag does not have a negative form, because it specifies a
17859 @item -fcall-used-@var{reg}
17860 @opindex fcall-used
17861 Treat the register named @var{reg} as an allocable register that is
17862 clobbered by function calls. It may be allocated for temporaries or
17863 variables that do not live across a call. Functions compiled this way
17864 will not save and restore the register @var{reg}.
17866 It is an error to used this flag with the frame pointer or stack pointer.
17867 Use of this flag for other registers that have fixed pervasive roles in
17868 the machine's execution model will produce disastrous results.
17870 This flag does not have a negative form, because it specifies a
17873 @item -fcall-saved-@var{reg}
17874 @opindex fcall-saved
17875 Treat the register named @var{reg} as an allocable register saved by
17876 functions. It may be allocated even for temporaries or variables that
17877 live across a call. Functions compiled this way will save and restore
17878 the register @var{reg} if they use it.
17880 It is an error to used this flag with the frame pointer or stack pointer.
17881 Use of this flag for other registers that have fixed pervasive roles in
17882 the machine's execution model will produce disastrous results.
17884 A different sort of disaster will result from the use of this flag for
17885 a register in which function values may be returned.
17887 This flag does not have a negative form, because it specifies a
17890 @item -fpack-struct[=@var{n}]
17891 @opindex fpack-struct
17892 Without a value specified, pack all structure members together without
17893 holes. When a value is specified (which must be a small power of two), pack
17894 structure members according to this value, representing the maximum
17895 alignment (that is, objects with default alignment requirements larger than
17896 this will be output potentially unaligned at the next fitting location.
17898 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17899 code that is not binary compatible with code generated without that switch.
17900 Additionally, it makes the code suboptimal.
17901 Use it to conform to a non-default application binary interface.
17903 @item -finstrument-functions
17904 @opindex finstrument-functions
17905 Generate instrumentation calls for entry and exit to functions. Just
17906 after function entry and just before function exit, the following
17907 profiling functions will be called with the address of the current
17908 function and its call site. (On some platforms,
17909 @code{__builtin_return_address} does not work beyond the current
17910 function, so the call site information may not be available to the
17911 profiling functions otherwise.)
17914 void __cyg_profile_func_enter (void *this_fn,
17916 void __cyg_profile_func_exit (void *this_fn,
17920 The first argument is the address of the start of the current function,
17921 which may be looked up exactly in the symbol table.
17923 This instrumentation is also done for functions expanded inline in other
17924 functions. The profiling calls will indicate where, conceptually, the
17925 inline function is entered and exited. This means that addressable
17926 versions of such functions must be available. If all your uses of a
17927 function are expanded inline, this may mean an additional expansion of
17928 code size. If you use @samp{extern inline} in your C code, an
17929 addressable version of such functions must be provided. (This is
17930 normally the case anyways, but if you get lucky and the optimizer always
17931 expands the functions inline, you might have gotten away without
17932 providing static copies.)
17934 A function may be given the attribute @code{no_instrument_function}, in
17935 which case this instrumentation will not be done. This can be used, for
17936 example, for the profiling functions listed above, high-priority
17937 interrupt routines, and any functions from which the profiling functions
17938 cannot safely be called (perhaps signal handlers, if the profiling
17939 routines generate output or allocate memory).
17941 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17942 @opindex finstrument-functions-exclude-file-list
17944 Set the list of functions that are excluded from instrumentation (see
17945 the description of @code{-finstrument-functions}). If the file that
17946 contains a function definition matches with one of @var{file}, then
17947 that function is not instrumented. The match is done on substrings:
17948 if the @var{file} parameter is a substring of the file name, it is
17949 considered to be a match.
17954 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
17958 will exclude any inline function defined in files whose pathnames
17959 contain @code{/bits/stl} or @code{include/sys}.
17961 If, for some reason, you want to include letter @code{','} in one of
17962 @var{sym}, write @code{'\,'}. For example,
17963 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17964 (note the single quote surrounding the option).
17966 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17967 @opindex finstrument-functions-exclude-function-list
17969 This is similar to @code{-finstrument-functions-exclude-file-list},
17970 but this option sets the list of function names to be excluded from
17971 instrumentation. The function name to be matched is its user-visible
17972 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17973 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17974 match is done on substrings: if the @var{sym} parameter is a substring
17975 of the function name, it is considered to be a match. For C99 and C++
17976 extended identifiers, the function name must be given in UTF-8, not
17977 using universal character names.
17979 @item -fstack-check
17980 @opindex fstack-check
17981 Generate code to verify that you do not go beyond the boundary of the
17982 stack. You should specify this flag if you are running in an
17983 environment with multiple threads, but only rarely need to specify it in
17984 a single-threaded environment since stack overflow is automatically
17985 detected on nearly all systems if there is only one stack.
17987 Note that this switch does not actually cause checking to be done; the
17988 operating system or the language runtime must do that. The switch causes
17989 generation of code to ensure that they see the stack being extended.
17991 You can additionally specify a string parameter: @code{no} means no
17992 checking, @code{generic} means force the use of old-style checking,
17993 @code{specific} means use the best checking method and is equivalent
17994 to bare @option{-fstack-check}.
17996 Old-style checking is a generic mechanism that requires no specific
17997 target support in the compiler but comes with the following drawbacks:
18001 Modified allocation strategy for large objects: they will always be
18002 allocated dynamically if their size exceeds a fixed threshold.
18005 Fixed limit on the size of the static frame of functions: when it is
18006 topped by a particular function, stack checking is not reliable and
18007 a warning is issued by the compiler.
18010 Inefficiency: because of both the modified allocation strategy and the
18011 generic implementation, the performances of the code are hampered.
18014 Note that old-style stack checking is also the fallback method for
18015 @code{specific} if no target support has been added in the compiler.
18017 @item -fstack-limit-register=@var{reg}
18018 @itemx -fstack-limit-symbol=@var{sym}
18019 @itemx -fno-stack-limit
18020 @opindex fstack-limit-register
18021 @opindex fstack-limit-symbol
18022 @opindex fno-stack-limit
18023 Generate code to ensure that the stack does not grow beyond a certain value,
18024 either the value of a register or the address of a symbol. If the stack
18025 would grow beyond the value, a signal is raised. For most targets,
18026 the signal is raised before the stack overruns the boundary, so
18027 it is possible to catch the signal without taking special precautions.
18029 For instance, if the stack starts at absolute address @samp{0x80000000}
18030 and grows downwards, you can use the flags
18031 @option{-fstack-limit-symbol=__stack_limit} and
18032 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18033 of 128KB@. Note that this may only work with the GNU linker.
18035 @item -fsplit-stack
18036 @opindex fsplit-stack
18037 Generate code to automatically split the stack before it overflows.
18038 The resulting program has a discontiguous stack which can only
18039 overflow if the program is unable to allocate any more memory. This
18040 is most useful when running threaded programs, as it is no longer
18041 necessary to calculate a good stack size to use for each thread. This
18042 is currently only implemented for the i386 and x86_64 backends running
18045 When code compiled with @option{-fsplit-stack} calls code compiled
18046 without @option{-fsplit-stack}, there may not be much stack space
18047 available for the latter code to run. If compiling all code,
18048 including library code, with @option{-fsplit-stack} is not an option,
18049 then the linker can fix up these calls so that the code compiled
18050 without @option{-fsplit-stack} always has a large stack. Support for
18051 this is implemented in the gold linker in GNU binutils release 2.21
18054 @item -fleading-underscore
18055 @opindex fleading-underscore
18056 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18057 change the way C symbols are represented in the object file. One use
18058 is to help link with legacy assembly code.
18060 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18061 generate code that is not binary compatible with code generated without that
18062 switch. Use it to conform to a non-default application binary interface.
18063 Not all targets provide complete support for this switch.
18065 @item -ftls-model=@var{model}
18066 @opindex ftls-model
18067 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18068 The @var{model} argument should be one of @code{global-dynamic},
18069 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18071 The default without @option{-fpic} is @code{initial-exec}; with
18072 @option{-fpic} the default is @code{global-dynamic}.
18074 @item -fvisibility=@var{default|internal|hidden|protected}
18075 @opindex fvisibility
18076 Set the default ELF image symbol visibility to the specified option---all
18077 symbols will be marked with this unless overridden within the code.
18078 Using this feature can very substantially improve linking and
18079 load times of shared object libraries, produce more optimized
18080 code, provide near-perfect API export and prevent symbol clashes.
18081 It is @strong{strongly} recommended that you use this in any shared objects
18084 Despite the nomenclature, @code{default} always means public; i.e.,
18085 available to be linked against from outside the shared object.
18086 @code{protected} and @code{internal} are pretty useless in real-world
18087 usage so the only other commonly used option will be @code{hidden}.
18088 The default if @option{-fvisibility} isn't specified is
18089 @code{default}, i.e., make every
18090 symbol public---this causes the same behavior as previous versions of
18093 A good explanation of the benefits offered by ensuring ELF
18094 symbols have the correct visibility is given by ``How To Write
18095 Shared Libraries'' by Ulrich Drepper (which can be found at
18096 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18097 solution made possible by this option to marking things hidden when
18098 the default is public is to make the default hidden and mark things
18099 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18100 and @code{__attribute__ ((visibility("default")))} instead of
18101 @code{__declspec(dllexport)} you get almost identical semantics with
18102 identical syntax. This is a great boon to those working with
18103 cross-platform projects.
18105 For those adding visibility support to existing code, you may find
18106 @samp{#pragma GCC visibility} of use. This works by you enclosing
18107 the declarations you wish to set visibility for with (for example)
18108 @samp{#pragma GCC visibility push(hidden)} and
18109 @samp{#pragma GCC visibility pop}.
18110 Bear in mind that symbol visibility should be viewed @strong{as
18111 part of the API interface contract} and thus all new code should
18112 always specify visibility when it is not the default; i.e., declarations
18113 only for use within the local DSO should @strong{always} be marked explicitly
18114 as hidden as so to avoid PLT indirection overheads---making this
18115 abundantly clear also aids readability and self-documentation of the code.
18116 Note that due to ISO C++ specification requirements, operator new and
18117 operator delete must always be of default visibility.
18119 Be aware that headers from outside your project, in particular system
18120 headers and headers from any other library you use, may not be
18121 expecting to be compiled with visibility other than the default. You
18122 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18123 before including any such headers.
18125 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18126 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18127 no modifications. However, this means that calls to @samp{extern}
18128 functions with no explicit visibility will use the PLT, so it is more
18129 effective to use @samp{__attribute ((visibility))} and/or
18130 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18131 declarations should be treated as hidden.
18133 Note that @samp{-fvisibility} does affect C++ vague linkage
18134 entities. This means that, for instance, an exception class that will
18135 be thrown between DSOs must be explicitly marked with default
18136 visibility so that the @samp{type_info} nodes will be unified between
18139 An overview of these techniques, their benefits and how to use them
18140 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18142 @item -fstrict-volatile-bitfields
18143 @opindex fstrict-volatile-bitfields
18144 This option should be used if accesses to volatile bitfields (or other
18145 structure fields, although the compiler usually honors those types
18146 anyway) should use a single access of the width of the
18147 field's type, aligned to a natural alignment if possible. For
18148 example, targets with memory-mapped peripheral registers might require
18149 all such accesses to be 16 bits wide; with this flag the user could
18150 declare all peripheral bitfields as ``unsigned short'' (assuming short
18151 is 16 bits on these targets) to force GCC to use 16 bit accesses
18152 instead of, perhaps, a more efficient 32 bit access.
18154 If this option is disabled, the compiler will use the most efficient
18155 instruction. In the previous example, that might be a 32-bit load
18156 instruction, even though that will access bytes that do not contain
18157 any portion of the bitfield, or memory-mapped registers unrelated to
18158 the one being updated.
18160 If the target requires strict alignment, and honoring the field
18161 type would require violating this alignment, a warning is issued.
18162 If the field has @code{packed} attribute, the access is done without
18163 honoring the field type. If the field doesn't have @code{packed}
18164 attribute, the access is done honoring the field type. In both cases,
18165 GCC assumes that the user knows something about the target hardware
18166 that it is unaware of.
18168 The default value of this option is determined by the application binary
18169 interface for the target processor.
18175 @node Environment Variables
18176 @section Environment Variables Affecting GCC
18177 @cindex environment variables
18179 @c man begin ENVIRONMENT
18180 This section describes several environment variables that affect how GCC
18181 operates. Some of them work by specifying directories or prefixes to use
18182 when searching for various kinds of files. Some are used to specify other
18183 aspects of the compilation environment.
18185 Note that you can also specify places to search using options such as
18186 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18187 take precedence over places specified using environment variables, which
18188 in turn take precedence over those specified by the configuration of GCC@.
18189 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18190 GNU Compiler Collection (GCC) Internals}.
18195 @c @itemx LC_COLLATE
18197 @c @itemx LC_MONETARY
18198 @c @itemx LC_NUMERIC
18203 @c @findex LC_COLLATE
18204 @findex LC_MESSAGES
18205 @c @findex LC_MONETARY
18206 @c @findex LC_NUMERIC
18210 These environment variables control the way that GCC uses
18211 localization information that allow GCC to work with different
18212 national conventions. GCC inspects the locale categories
18213 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18214 so. These locale categories can be set to any value supported by your
18215 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18216 Kingdom encoded in UTF-8.
18218 The @env{LC_CTYPE} environment variable specifies character
18219 classification. GCC uses it to determine the character boundaries in
18220 a string; this is needed for some multibyte encodings that contain quote
18221 and escape characters that would otherwise be interpreted as a string
18224 The @env{LC_MESSAGES} environment variable specifies the language to
18225 use in diagnostic messages.
18227 If the @env{LC_ALL} environment variable is set, it overrides the value
18228 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18229 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18230 environment variable. If none of these variables are set, GCC
18231 defaults to traditional C English behavior.
18235 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18236 files. GCC uses temporary files to hold the output of one stage of
18237 compilation which is to be used as input to the next stage: for example,
18238 the output of the preprocessor, which is the input to the compiler
18241 @item GCC_EXEC_PREFIX
18242 @findex GCC_EXEC_PREFIX
18243 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18244 names of the subprograms executed by the compiler. No slash is added
18245 when this prefix is combined with the name of a subprogram, but you can
18246 specify a prefix that ends with a slash if you wish.
18248 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18249 an appropriate prefix to use based on the pathname it was invoked with.
18251 If GCC cannot find the subprogram using the specified prefix, it
18252 tries looking in the usual places for the subprogram.
18254 The default value of @env{GCC_EXEC_PREFIX} is
18255 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18256 the installed compiler. In many cases @var{prefix} is the value
18257 of @code{prefix} when you ran the @file{configure} script.
18259 Other prefixes specified with @option{-B} take precedence over this prefix.
18261 This prefix is also used for finding files such as @file{crt0.o} that are
18264 In addition, the prefix is used in an unusual way in finding the
18265 directories to search for header files. For each of the standard
18266 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18267 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18268 replacing that beginning with the specified prefix to produce an
18269 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18270 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18271 These alternate directories are searched first; the standard directories
18272 come next. If a standard directory begins with the configured
18273 @var{prefix} then the value of @var{prefix} is replaced by
18274 @env{GCC_EXEC_PREFIX} when looking for header files.
18276 @item COMPILER_PATH
18277 @findex COMPILER_PATH
18278 The value of @env{COMPILER_PATH} is a colon-separated list of
18279 directories, much like @env{PATH}. GCC tries the directories thus
18280 specified when searching for subprograms, if it can't find the
18281 subprograms using @env{GCC_EXEC_PREFIX}.
18284 @findex LIBRARY_PATH
18285 The value of @env{LIBRARY_PATH} is a colon-separated list of
18286 directories, much like @env{PATH}. When configured as a native compiler,
18287 GCC tries the directories thus specified when searching for special
18288 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18289 using GCC also uses these directories when searching for ordinary
18290 libraries for the @option{-l} option (but directories specified with
18291 @option{-L} come first).
18295 @cindex locale definition
18296 This variable is used to pass locale information to the compiler. One way in
18297 which this information is used is to determine the character set to be used
18298 when character literals, string literals and comments are parsed in C and C++.
18299 When the compiler is configured to allow multibyte characters,
18300 the following values for @env{LANG} are recognized:
18304 Recognize JIS characters.
18306 Recognize SJIS characters.
18308 Recognize EUCJP characters.
18311 If @env{LANG} is not defined, or if it has some other value, then the
18312 compiler will use mblen and mbtowc as defined by the default locale to
18313 recognize and translate multibyte characters.
18317 Some additional environments variables affect the behavior of the
18320 @include cppenv.texi
18324 @node Precompiled Headers
18325 @section Using Precompiled Headers
18326 @cindex precompiled headers
18327 @cindex speed of compilation
18329 Often large projects have many header files that are included in every
18330 source file. The time the compiler takes to process these header files
18331 over and over again can account for nearly all of the time required to
18332 build the project. To make builds faster, GCC allows users to
18333 `precompile' a header file; then, if builds can use the precompiled
18334 header file they will be much faster.
18336 To create a precompiled header file, simply compile it as you would any
18337 other file, if necessary using the @option{-x} option to make the driver
18338 treat it as a C or C++ header file. You will probably want to use a
18339 tool like @command{make} to keep the precompiled header up-to-date when
18340 the headers it contains change.
18342 A precompiled header file will be searched for when @code{#include} is
18343 seen in the compilation. As it searches for the included file
18344 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18345 compiler looks for a precompiled header in each directory just before it
18346 looks for the include file in that directory. The name searched for is
18347 the name specified in the @code{#include} with @samp{.gch} appended. If
18348 the precompiled header file can't be used, it is ignored.
18350 For instance, if you have @code{#include "all.h"}, and you have
18351 @file{all.h.gch} in the same directory as @file{all.h}, then the
18352 precompiled header file will be used if possible, and the original
18353 header will be used otherwise.
18355 Alternatively, you might decide to put the precompiled header file in a
18356 directory and use @option{-I} to ensure that directory is searched
18357 before (or instead of) the directory containing the original header.
18358 Then, if you want to check that the precompiled header file is always
18359 used, you can put a file of the same name as the original header in this
18360 directory containing an @code{#error} command.
18362 This also works with @option{-include}. So yet another way to use
18363 precompiled headers, good for projects not designed with precompiled
18364 header files in mind, is to simply take most of the header files used by
18365 a project, include them from another header file, precompile that header
18366 file, and @option{-include} the precompiled header. If the header files
18367 have guards against multiple inclusion, they will be skipped because
18368 they've already been included (in the precompiled header).
18370 If you need to precompile the same header file for different
18371 languages, targets, or compiler options, you can instead make a
18372 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18373 header in the directory, perhaps using @option{-o}. It doesn't matter
18374 what you call the files in the directory, every precompiled header in
18375 the directory will be considered. The first precompiled header
18376 encountered in the directory that is valid for this compilation will
18377 be used; they're searched in no particular order.
18379 There are many other possibilities, limited only by your imagination,
18380 good sense, and the constraints of your build system.
18382 A precompiled header file can be used only when these conditions apply:
18386 Only one precompiled header can be used in a particular compilation.
18389 A precompiled header can't be used once the first C token is seen. You
18390 can have preprocessor directives before a precompiled header; you can
18391 even include a precompiled header from inside another header, so long as
18392 there are no C tokens before the @code{#include}.
18395 The precompiled header file must be produced for the same language as
18396 the current compilation. You can't use a C precompiled header for a C++
18400 The precompiled header file must have been produced by the same compiler
18401 binary as the current compilation is using.
18404 Any macros defined before the precompiled header is included must
18405 either be defined in the same way as when the precompiled header was
18406 generated, or must not affect the precompiled header, which usually
18407 means that they don't appear in the precompiled header at all.
18409 The @option{-D} option is one way to define a macro before a
18410 precompiled header is included; using a @code{#define} can also do it.
18411 There are also some options that define macros implicitly, like
18412 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18415 @item If debugging information is output when using the precompiled
18416 header, using @option{-g} or similar, the same kind of debugging information
18417 must have been output when building the precompiled header. However,
18418 a precompiled header built using @option{-g} can be used in a compilation
18419 when no debugging information is being output.
18421 @item The same @option{-m} options must generally be used when building
18422 and using the precompiled header. @xref{Submodel Options},
18423 for any cases where this rule is relaxed.
18425 @item Each of the following options must be the same when building and using
18426 the precompiled header:
18428 @gccoptlist{-fexceptions}
18431 Some other command-line options starting with @option{-f},
18432 @option{-p}, or @option{-O} must be defined in the same way as when
18433 the precompiled header was generated. At present, it's not clear
18434 which options are safe to change and which are not; the safest choice
18435 is to use exactly the same options when generating and using the
18436 precompiled header. The following are known to be safe:
18438 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18439 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18440 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18445 For all of these except the last, the compiler will automatically
18446 ignore the precompiled header if the conditions aren't met. If you
18447 find an option combination that doesn't work and doesn't cause the
18448 precompiled header to be ignored, please consider filing a bug report,
18451 If you do use differing options when generating and using the
18452 precompiled header, the actual behavior will be a mixture of the
18453 behavior for the options. For instance, if you use @option{-g} to
18454 generate the precompiled header but not when using it, you may or may
18455 not get debugging information for routines in the precompiled header.