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
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
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
165 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
166 --version -wrapper@@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg}}
168 @item C Language Options
169 @xref{C Dialect Options,,Options Controlling C Dialect}.
170 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
171 -aux-info @var{filename} @gol
172 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
173 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
174 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
175 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
176 -fsigned-bitfields -fsigned-char @gol
177 -funsigned-bitfields -funsigned-char}
179 @item C++ Language Options
180 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
181 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
182 -fconserve-space -ffriend-injection @gol
183 -fno-elide-constructors @gol
184 -fno-enforce-eh-specs @gol
185 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
186 -fno-implicit-templates @gol
187 -fno-implicit-inline-templates @gol
188 -fno-implement-inlines -fms-extensions @gol
189 -fno-nonansi-builtins -fno-operator-names @gol
190 -fno-optional-diags -fpermissive @gol
191 -fno-pretty-templates @gol
192 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
194 -fno-default-inline -fvisibility-inlines-hidden @gol
195 -fvisibility-ms-compat @gol
196 -Wabi -Wctor-dtor-privacy @gol
197 -Wnon-virtual-dtor -Wreorder @gol
198 -Weffc++ -Wstrict-null-sentinel @gol
199 -Wno-non-template-friend -Wold-style-cast @gol
200 -Woverloaded-virtual -Wno-pmf-conversions @gol
203 @item Objective-C and Objective-C++ Language Options
204 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
205 Objective-C and Objective-C++ Dialects}.
206 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
207 -fgnu-runtime -fnext-runtime @gol
208 -fno-nil-receivers @gol
209 -fobjc-call-cxx-cdtors @gol
210 -fobjc-direct-dispatch @gol
211 -fobjc-exceptions @gol
213 -freplace-objc-classes @gol
216 -Wassign-intercept @gol
217 -Wno-protocol -Wselector @gol
218 -Wstrict-selector-match @gol
219 -Wundeclared-selector}
221 @item Language Independent Options
222 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
223 @gccoptlist{-fmessage-length=@var{n} @gol
224 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
225 -fdiagnostics-show-option}
227 @item Warning Options
228 @xref{Warning Options,,Options to Request or Suppress Warnings}.
229 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
230 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
231 -Wno-attributes -Wno-builtin-macro-redefined @gol
232 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
233 -Wchar-subscripts -Wclobbered -Wcomment @gol
234 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
235 -Wno-deprecated-declarations -Wdisabled-optimization @gol
236 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
237 -Werror -Werror=* @gol
238 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
239 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
240 -Wformat-security -Wformat-y2k @gol
241 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
242 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
243 -Winit-self -Winline @gol
244 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
245 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
246 -Wlogical-op -Wlong-long @gol
247 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
248 -Wmissing-format-attribute -Wmissing-include-dirs @gol
249 -Wmissing-noreturn -Wno-mudflap @gol
250 -Wno-multichar -Wnonnull -Wno-overflow @gol
251 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
252 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
253 -Wpointer-arith -Wno-pointer-to-int-cast @gol
254 -Wredundant-decls @gol
255 -Wreturn-type -Wsequence-point -Wshadow @gol
256 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
257 -Wstrict-aliasing -Wstrict-aliasing=n @gol
258 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
259 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
260 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
261 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
262 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
263 -Wunused-label -Wunused-parameter -Wunused-value -Wunused-variable @gol
264 -Wvariadic-macros -Wvla @gol
265 -Wvolatile-register-var -Wwrite-strings}
267 @item C and Objective-C-only Warning Options
268 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
269 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
270 -Wold-style-declaration -Wold-style-definition @gol
271 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
272 -Wdeclaration-after-statement -Wpointer-sign}
274 @item Debugging Options
275 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
276 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
277 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
278 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
279 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
280 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
281 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
282 -fdump-statistics @gol
284 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
288 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-nrv -fdump-tree-vect @gol
298 -fdump-tree-sink @gol
299 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
303 -ftree-vectorizer-verbose=@var{n} @gol
304 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
305 -fdump-final-insns=@var{file} @gol
306 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
307 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
308 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
309 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
310 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
311 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
312 -ftest-coverage -ftime-report -fvar-tracking @gol
313 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
314 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
315 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
316 -fdebug-prefix-map=@var{old}=@var{new} @gol
317 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
318 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
319 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
320 -print-multi-directory -print-multi-lib @gol
321 -print-prog-name=@var{program} -print-search-dirs -Q @gol
322 -print-sysroot -print-sysroot-headers-suffix @gol
323 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
325 @item Optimization Options
326 @xref{Optimize Options,,Options that Control Optimization}.
328 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
329 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
330 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
331 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
332 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
333 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
334 -fdata-sections -fdce -fdce @gol
335 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
336 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
337 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
338 -fforward-propagate -ffunction-sections @gol
339 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
340 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
341 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
342 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
343 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
344 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
345 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
346 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
347 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
348 -floop-block -floop-interchange -floop-strip-mine @gol
349 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
350 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
351 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
352 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
353 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
354 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
355 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
356 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
357 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
358 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
359 -fprofile-generate=@var{path} @gol
360 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
361 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
362 -freorder-blocks-and-partition -freorder-functions @gol
363 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
364 -frounding-math -fsched2-use-superblocks @gol
365 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
366 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
367 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
368 -fselective-scheduling -fselective-scheduling2 @gol
369 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
370 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
371 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
372 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
373 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
374 -ftree-copyrename -ftree-dce @gol
375 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
376 -ftree-phiprop -ftree-loop-distribution @gol
377 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
378 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
379 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
380 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
381 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
382 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
383 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
385 --param @var{name}=@var{value}
386 -O -O0 -O1 -O2 -O3 -Os}
388 @item Preprocessor Options
389 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
390 @gccoptlist{-A@var{question}=@var{answer} @gol
391 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
392 -C -dD -dI -dM -dN @gol
393 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
394 -idirafter @var{dir} @gol
395 -include @var{file} -imacros @var{file} @gol
396 -iprefix @var{file} -iwithprefix @var{dir} @gol
397 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
398 -imultilib @var{dir} -isysroot @var{dir} @gol
399 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
400 -P -fworking-directory -remap @gol
401 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
402 -Xpreprocessor @var{option}}
404 @item Assembler Option
405 @xref{Assembler Options,,Passing Options to the Assembler}.
406 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
409 @xref{Link Options,,Options for Linking}.
410 @gccoptlist{@var{object-file-name} -l@var{library} @gol
411 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
412 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
413 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
416 @item Directory Options
417 @xref{Directory Options,,Options for Directory Search}.
418 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
419 -specs=@var{file} -I- --sysroot=@var{dir}}
422 @c I wrote this xref this way to avoid overfull hbox. -- rms
423 @xref{Target Options}.
424 @gccoptlist{-V @var{version} -b @var{machine}}
426 @item Machine Dependent Options
427 @xref{Submodel Options,,Hardware Models and Configurations}.
428 @c This list is ordered alphanumerically by subsection name.
429 @c Try and put the significant identifier (CPU or system) first,
430 @c so users have a clue at guessing where the ones they want will be.
433 @gccoptlist{-EB -EL @gol
434 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
435 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
438 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
439 -mabi=@var{name} @gol
440 -mapcs-stack-check -mno-apcs-stack-check @gol
441 -mapcs-float -mno-apcs-float @gol
442 -mapcs-reentrant -mno-apcs-reentrant @gol
443 -msched-prolog -mno-sched-prolog @gol
444 -mlittle-endian -mbig-endian -mwords-little-endian @gol
445 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
446 -mfp16-format=@var{name}
447 -mthumb-interwork -mno-thumb-interwork @gol
448 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
449 -mstructure-size-boundary=@var{n} @gol
450 -mabort-on-noreturn @gol
451 -mlong-calls -mno-long-calls @gol
452 -msingle-pic-base -mno-single-pic-base @gol
453 -mpic-register=@var{reg} @gol
454 -mnop-fun-dllimport @gol
455 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
456 -mpoke-function-name @gol
458 -mtpcs-frame -mtpcs-leaf-frame @gol
459 -mcaller-super-interworking -mcallee-super-interworking @gol
461 -mword-relocations @gol
462 -mfix-cortex-m3-ldrd}
465 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
466 -mcall-prologues -mtiny-stack -mint8}
468 @emph{Blackfin Options}
469 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
470 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
471 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
472 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
473 -mno-id-shared-library -mshared-library-id=@var{n} @gol
474 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
475 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
476 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
480 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
481 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
482 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
483 -mstack-align -mdata-align -mconst-align @gol
484 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
485 -melf -maout -melinux -mlinux -sim -sim2 @gol
486 -mmul-bug-workaround -mno-mul-bug-workaround}
489 @gccoptlist{-mmac -mpush-args}
491 @emph{Darwin Options}
492 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
493 -arch_only -bind_at_load -bundle -bundle_loader @gol
494 -client_name -compatibility_version -current_version @gol
496 -dependency-file -dylib_file -dylinker_install_name @gol
497 -dynamic -dynamiclib -exported_symbols_list @gol
498 -filelist -flat_namespace -force_cpusubtype_ALL @gol
499 -force_flat_namespace -headerpad_max_install_names @gol
501 -image_base -init -install_name -keep_private_externs @gol
502 -multi_module -multiply_defined -multiply_defined_unused @gol
503 -noall_load -no_dead_strip_inits_and_terms @gol
504 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
505 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
506 -private_bundle -read_only_relocs -sectalign @gol
507 -sectobjectsymbols -whyload -seg1addr @gol
508 -sectcreate -sectobjectsymbols -sectorder @gol
509 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
510 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
511 -segprot -segs_read_only_addr -segs_read_write_addr @gol
512 -single_module -static -sub_library -sub_umbrella @gol
513 -twolevel_namespace -umbrella -undefined @gol
514 -unexported_symbols_list -weak_reference_mismatches @gol
515 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
516 -mkernel -mone-byte-bool}
518 @emph{DEC Alpha Options}
519 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
520 -mieee -mieee-with-inexact -mieee-conformant @gol
521 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
522 -mtrap-precision=@var{mode} -mbuild-constants @gol
523 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
524 -mbwx -mmax -mfix -mcix @gol
525 -mfloat-vax -mfloat-ieee @gol
526 -mexplicit-relocs -msmall-data -mlarge-data @gol
527 -msmall-text -mlarge-text @gol
528 -mmemory-latency=@var{time}}
530 @emph{DEC Alpha/VMS Options}
531 @gccoptlist{-mvms-return-codes}
534 @gccoptlist{-msmall-model -mno-lsim}
537 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
538 -mhard-float -msoft-float @gol
539 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
540 -mdouble -mno-double @gol
541 -mmedia -mno-media -mmuladd -mno-muladd @gol
542 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
543 -mlinked-fp -mlong-calls -malign-labels @gol
544 -mlibrary-pic -macc-4 -macc-8 @gol
545 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
546 -moptimize-membar -mno-optimize-membar @gol
547 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
548 -mvliw-branch -mno-vliw-branch @gol
549 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
550 -mno-nested-cond-exec -mtomcat-stats @gol
554 @emph{GNU/Linux Options}
555 @gccoptlist{-muclibc}
557 @emph{H8/300 Options}
558 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
561 @gccoptlist{-march=@var{architecture-type} @gol
562 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
563 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
564 -mfixed-range=@var{register-range} @gol
565 -mjump-in-delay -mlinker-opt -mlong-calls @gol
566 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
567 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
568 -mno-jump-in-delay -mno-long-load-store @gol
569 -mno-portable-runtime -mno-soft-float @gol
570 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
571 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
572 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
573 -munix=@var{unix-std} -nolibdld -static -threads}
575 @emph{i386 and x86-64 Options}
576 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
577 -mfpmath=@var{unit} @gol
578 -masm=@var{dialect} -mno-fancy-math-387 @gol
579 -mno-fp-ret-in-387 -msoft-float @gol
580 -mno-wide-multiply -mrtd -malign-double @gol
581 -mpreferred-stack-boundary=@var{num}
582 -mincoming-stack-boundary=@var{num}
583 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
584 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
586 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
587 -mthreads -mno-align-stringops -minline-all-stringops @gol
588 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
589 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
590 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
591 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
592 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
593 -mcmodel=@var{code-model} -mabi=@var{name} @gol
594 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
595 -mfused-madd -mno-fused-madd -msse2avx}
598 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
599 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
600 -mconstant-gp -mauto-pic -mfused-madd @gol
601 -minline-float-divide-min-latency @gol
602 -minline-float-divide-max-throughput @gol
603 -mno-inline-float-divide @gol
604 -minline-int-divide-min-latency @gol
605 -minline-int-divide-max-throughput @gol
606 -mno-inline-int-divide @gol
607 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
608 -mno-inline-sqrt @gol
609 -mdwarf2-asm -mearly-stop-bits @gol
610 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
611 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
612 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
613 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
614 -msched-spec-ldc -msched-spec-control-ldc @gol
615 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
616 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
617 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
618 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
620 @emph{M32R/D Options}
621 @gccoptlist{-m32r2 -m32rx -m32r @gol
623 -malign-loops -mno-align-loops @gol
624 -missue-rate=@var{number} @gol
625 -mbranch-cost=@var{number} @gol
626 -mmodel=@var{code-size-model-type} @gol
627 -msdata=@var{sdata-type} @gol
628 -mno-flush-func -mflush-func=@var{name} @gol
629 -mno-flush-trap -mflush-trap=@var{number} @gol
633 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
635 @emph{M680x0 Options}
636 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
637 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
638 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
639 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
640 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
641 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
642 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
643 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
646 @emph{M68hc1x Options}
647 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
648 -mauto-incdec -minmax -mlong-calls -mshort @gol
649 -msoft-reg-count=@var{count}}
652 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
653 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
654 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
655 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
656 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
659 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
660 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
661 -mips64 -mips64r2 @gol
662 -mips16 -mno-mips16 -mflip-mips16 @gol
663 -minterlink-mips16 -mno-interlink-mips16 @gol
664 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
665 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
666 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
667 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
668 -mfpu=@var{fpu-type} @gol
669 -msmartmips -mno-smartmips @gol
670 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
671 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
672 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
673 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
674 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
675 -membedded-data -mno-embedded-data @gol
676 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
677 -mcode-readable=@var{setting} @gol
678 -msplit-addresses -mno-split-addresses @gol
679 -mexplicit-relocs -mno-explicit-relocs @gol
680 -mcheck-zero-division -mno-check-zero-division @gol
681 -mdivide-traps -mdivide-breaks @gol
682 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
683 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
684 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
685 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
686 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
687 -mflush-func=@var{func} -mno-flush-func @gol
688 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
689 -mfp-exceptions -mno-fp-exceptions @gol
690 -mvr4130-align -mno-vr4130-align}
693 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
694 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
695 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
696 -mno-base-addresses -msingle-exit -mno-single-exit}
698 @emph{MN10300 Options}
699 @gccoptlist{-mmult-bug -mno-mult-bug @gol
700 -mam33 -mno-am33 @gol
701 -mam33-2 -mno-am33-2 @gol
702 -mreturn-pointer-on-d0 @gol
705 @emph{PDP-11 Options}
706 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
707 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
708 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
709 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
710 -mbranch-expensive -mbranch-cheap @gol
711 -msplit -mno-split -munix-asm -mdec-asm}
713 @emph{picoChip Options}
714 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
715 -msymbol-as-address -mno-inefficient-warnings}
717 @emph{PowerPC Options}
718 See RS/6000 and PowerPC Options.
720 @emph{RS/6000 and PowerPC Options}
721 @gccoptlist{-mcpu=@var{cpu-type} @gol
722 -mtune=@var{cpu-type} @gol
723 -mpower -mno-power -mpower2 -mno-power2 @gol
724 -mpowerpc -mpowerpc64 -mno-powerpc @gol
725 -maltivec -mno-altivec @gol
726 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
727 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
728 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
729 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
730 -mnew-mnemonics -mold-mnemonics @gol
731 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
732 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
733 -malign-power -malign-natural @gol
734 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
735 -msingle-float -mdouble-float -msimple-fpu @gol
736 -mstring -mno-string -mupdate -mno-update @gol
737 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
738 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
739 -mstrict-align -mno-strict-align -mrelocatable @gol
740 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
741 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
742 -mdynamic-no-pic -maltivec -mswdiv @gol
743 -mprioritize-restricted-insns=@var{priority} @gol
744 -msched-costly-dep=@var{dependence_type} @gol
745 -minsert-sched-nops=@var{scheme} @gol
746 -mcall-sysv -mcall-netbsd @gol
747 -maix-struct-return -msvr4-struct-return @gol
748 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
749 -misel -mno-isel @gol
750 -misel=yes -misel=no @gol
752 -mspe=yes -mspe=no @gol
754 -mgen-cell-microcode -mwarn-cell-microcode @gol
755 -mvrsave -mno-vrsave @gol
756 -mmulhw -mno-mulhw @gol
757 -mdlmzb -mno-dlmzb @gol
758 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
759 -mprototype -mno-prototype @gol
760 -msim -mmvme -mads -myellowknife -memb -msdata @gol
761 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
763 @emph{S/390 and zSeries Options}
764 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
765 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
766 -mlong-double-64 -mlong-double-128 @gol
767 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
768 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
769 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
770 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
771 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
774 @gccoptlist{-meb -mel @gol
778 -mscore5 -mscore5u -mscore7 -mscore7d}
781 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
782 -m4-nofpu -m4-single-only -m4-single -m4 @gol
783 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
784 -m5-64media -m5-64media-nofpu @gol
785 -m5-32media -m5-32media-nofpu @gol
786 -m5-compact -m5-compact-nofpu @gol
787 -mb -ml -mdalign -mrelax @gol
788 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
789 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
790 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
791 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
792 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
796 @gccoptlist{-mcpu=@var{cpu-type} @gol
797 -mtune=@var{cpu-type} @gol
798 -mcmodel=@var{code-model} @gol
799 -m32 -m64 -mapp-regs -mno-app-regs @gol
800 -mfaster-structs -mno-faster-structs @gol
801 -mfpu -mno-fpu -mhard-float -msoft-float @gol
802 -mhard-quad-float -msoft-quad-float @gol
803 -mimpure-text -mno-impure-text -mlittle-endian @gol
804 -mstack-bias -mno-stack-bias @gol
805 -munaligned-doubles -mno-unaligned-doubles @gol
806 -mv8plus -mno-v8plus -mvis -mno-vis
807 -threads -pthreads -pthread}
810 @gccoptlist{-mwarn-reloc -merror-reloc @gol
811 -msafe-dma -munsafe-dma @gol
813 -msmall-mem -mlarge-mem -mstdmain @gol
814 -mfixed-range=@var{register-range}}
816 @emph{System V Options}
817 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
820 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
821 -mprolog-function -mno-prolog-function -mspace @gol
822 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
823 -mapp-regs -mno-app-regs @gol
824 -mdisable-callt -mno-disable-callt @gol
830 @gccoptlist{-mg -mgnu -munix}
832 @emph{VxWorks Options}
833 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
834 -Xbind-lazy -Xbind-now}
836 @emph{x86-64 Options}
837 See i386 and x86-64 Options.
839 @emph{i386 and x86-64 Windows Options}
840 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
841 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
843 @emph{Xstormy16 Options}
846 @emph{Xtensa Options}
847 @gccoptlist{-mconst16 -mno-const16 @gol
848 -mfused-madd -mno-fused-madd @gol
849 -mserialize-volatile -mno-serialize-volatile @gol
850 -mtext-section-literals -mno-text-section-literals @gol
851 -mtarget-align -mno-target-align @gol
852 -mlongcalls -mno-longcalls}
854 @emph{zSeries Options}
855 See S/390 and zSeries Options.
857 @item Code Generation Options
858 @xref{Code Gen Options,,Options for Code Generation Conventions}.
859 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
860 -ffixed-@var{reg} -fexceptions @gol
861 -fnon-call-exceptions -funwind-tables @gol
862 -fasynchronous-unwind-tables @gol
863 -finhibit-size-directive -finstrument-functions @gol
864 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
865 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
866 -fno-common -fno-ident @gol
867 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
868 -fno-jump-tables @gol
869 -frecord-gcc-switches @gol
870 -freg-struct-return -fshort-enums @gol
871 -fshort-double -fshort-wchar @gol
872 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
873 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
874 -fno-stack-limit -fargument-alias -fargument-noalias @gol
875 -fargument-noalias-global -fargument-noalias-anything @gol
876 -fleading-underscore -ftls-model=@var{model} @gol
877 -ftrapv -fwrapv -fbounds-check @gol
882 * Overall Options:: Controlling the kind of output:
883 an executable, object files, assembler files,
884 or preprocessed source.
885 * C Dialect Options:: Controlling the variant of C language compiled.
886 * C++ Dialect Options:: Variations on C++.
887 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
889 * Language Independent Options:: Controlling how diagnostics should be
891 * Warning Options:: How picky should the compiler be?
892 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
893 * Optimize Options:: How much optimization?
894 * Preprocessor Options:: Controlling header files and macro definitions.
895 Also, getting dependency information for Make.
896 * Assembler Options:: Passing options to the assembler.
897 * Link Options:: Specifying libraries and so on.
898 * Directory Options:: Where to find header files and libraries.
899 Where to find the compiler executable files.
900 * Spec Files:: How to pass switches to sub-processes.
901 * Target Options:: Running a cross-compiler, or an old version of GCC.
904 @node Overall Options
905 @section Options Controlling the Kind of Output
907 Compilation can involve up to four stages: preprocessing, compilation
908 proper, assembly and linking, always in that order. GCC is capable of
909 preprocessing and compiling several files either into several
910 assembler input files, or into one assembler input file; then each
911 assembler input file produces an object file, and linking combines all
912 the object files (those newly compiled, and those specified as input)
913 into an executable file.
915 @cindex file name suffix
916 For any given input file, the file name suffix determines what kind of
921 C source code which must be preprocessed.
924 C source code which should not be preprocessed.
927 C++ source code which should not be preprocessed.
930 Objective-C source code. Note that you must link with the @file{libobjc}
931 library to make an Objective-C program work.
934 Objective-C source code which should not be preprocessed.
938 Objective-C++ source code. Note that you must link with the @file{libobjc}
939 library to make an Objective-C++ program work. Note that @samp{.M} refers
940 to a literal capital M@.
943 Objective-C++ source code which should not be preprocessed.
946 C, C++, Objective-C or Objective-C++ header file to be turned into a
951 @itemx @var{file}.cxx
952 @itemx @var{file}.cpp
953 @itemx @var{file}.CPP
954 @itemx @var{file}.c++
956 C++ source code which must be preprocessed. Note that in @samp{.cxx},
957 the last two letters must both be literally @samp{x}. Likewise,
958 @samp{.C} refers to a literal capital C@.
962 Objective-C++ source code which must be preprocessed.
965 Objective-C++ source code which should not be preprocessed.
970 @itemx @var{file}.hxx
971 @itemx @var{file}.hpp
972 @itemx @var{file}.HPP
973 @itemx @var{file}.h++
974 @itemx @var{file}.tcc
975 C++ header file to be turned into a precompiled header.
978 @itemx @var{file}.for
979 @itemx @var{file}.ftn
980 Fixed form Fortran source code which should not be preprocessed.
983 @itemx @var{file}.FOR
984 @itemx @var{file}.fpp
985 @itemx @var{file}.FPP
986 @itemx @var{file}.FTN
987 Fixed form Fortran source code which must be preprocessed (with the traditional
991 @itemx @var{file}.f95
992 @itemx @var{file}.f03
993 @itemx @var{file}.f08
994 Free form Fortran source code which should not be preprocessed.
997 @itemx @var{file}.F95
998 @itemx @var{file}.F03
999 @itemx @var{file}.F08
1000 Free form Fortran source code which must be preprocessed (with the
1001 traditional preprocessor).
1003 @c FIXME: Descriptions of Java file types.
1009 @item @var{file}.ads
1010 Ada source code file which contains a library unit declaration (a
1011 declaration of a package, subprogram, or generic, or a generic
1012 instantiation), or a library unit renaming declaration (a package,
1013 generic, or subprogram renaming declaration). Such files are also
1016 @item @var{file}.adb
1017 Ada source code file containing a library unit body (a subprogram or
1018 package body). Such files are also called @dfn{bodies}.
1020 @c GCC also knows about some suffixes for languages not yet included:
1031 @itemx @var{file}.sx
1032 Assembler code which must be preprocessed.
1035 An object file to be fed straight into linking.
1036 Any file name with no recognized suffix is treated this way.
1040 You can specify the input language explicitly with the @option{-x} option:
1043 @item -x @var{language}
1044 Specify explicitly the @var{language} for the following input files
1045 (rather than letting the compiler choose a default based on the file
1046 name suffix). This option applies to all following input files until
1047 the next @option{-x} option. Possible values for @var{language} are:
1049 c c-header c-cpp-output
1050 c++ c++-header c++-cpp-output
1051 objective-c objective-c-header objective-c-cpp-output
1052 objective-c++ objective-c++-header objective-c++-cpp-output
1053 assembler assembler-with-cpp
1055 f77 f77-cpp-input f95 f95-cpp-input
1060 Turn off any specification of a language, so that subsequent files are
1061 handled according to their file name suffixes (as they are if @option{-x}
1062 has not been used at all).
1064 @item -pass-exit-codes
1065 @opindex pass-exit-codes
1066 Normally the @command{gcc} program will exit with the code of 1 if any
1067 phase of the compiler returns a non-success return code. If you specify
1068 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1069 numerically highest error produced by any phase that returned an error
1070 indication. The C, C++, and Fortran frontends return 4, if an internal
1071 compiler error is encountered.
1074 If you only want some of the stages of compilation, you can use
1075 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1076 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1077 @command{gcc} is to stop. Note that some combinations (for example,
1078 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1083 Compile or assemble the source files, but do not link. The linking
1084 stage simply is not done. The ultimate output is in the form of an
1085 object file for each source file.
1087 By default, the object file name for a source file is made by replacing
1088 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1090 Unrecognized input files, not requiring compilation or assembly, are
1095 Stop after the stage of compilation proper; do not assemble. The output
1096 is in the form of an assembler code file for each non-assembler input
1099 By default, the assembler file name for a source file is made by
1100 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1102 Input files that don't require compilation are ignored.
1106 Stop after the preprocessing stage; do not run the compiler proper. The
1107 output is in the form of preprocessed source code, which is sent to the
1110 Input files which don't require preprocessing are ignored.
1112 @cindex output file option
1115 Place output in file @var{file}. This applies regardless to whatever
1116 sort of output is being produced, whether it be an executable file,
1117 an object file, an assembler file or preprocessed C code.
1119 If @option{-o} is not specified, the default is to put an executable
1120 file in @file{a.out}, the object file for
1121 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1122 assembler file in @file{@var{source}.s}, a precompiled header file in
1123 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1128 Print (on standard error output) the commands executed to run the stages
1129 of compilation. Also print the version number of the compiler driver
1130 program and of the preprocessor and the compiler proper.
1134 Like @option{-v} except the commands are not executed and all command
1135 arguments are quoted. This is useful for shell scripts to capture the
1136 driver-generated command lines.
1140 Use pipes rather than temporary files for communication between the
1141 various stages of compilation. This fails to work on some systems where
1142 the assembler is unable to read from a pipe; but the GNU assembler has
1147 If you are compiling multiple source files, this option tells the driver
1148 to pass all the source files to the compiler at once (for those
1149 languages for which the compiler can handle this). This will allow
1150 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1151 language for which this is supported is C@. If you pass source files for
1152 multiple languages to the driver, using this option, the driver will invoke
1153 the compiler(s) that support IMA once each, passing each compiler all the
1154 source files appropriate for it. For those languages that do not support
1155 IMA this option will be ignored, and the compiler will be invoked once for
1156 each source file in that language. If you use this option in conjunction
1157 with @option{-save-temps}, the compiler will generate multiple
1159 (one for each source file), but only one (combined) @file{.o} or
1164 Print (on the standard output) a description of the command line options
1165 understood by @command{gcc}. If the @option{-v} option is also specified
1166 then @option{--help} will also be passed on to the various processes
1167 invoked by @command{gcc}, so that they can display the command line options
1168 they accept. If the @option{-Wextra} option has also been specified
1169 (prior to the @option{--help} option), then command line options which
1170 have no documentation associated with them will also be displayed.
1173 @opindex target-help
1174 Print (on the standard output) a description of target-specific command
1175 line options for each tool. For some targets extra target-specific
1176 information may also be printed.
1178 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1179 Print (on the standard output) a description of the command line
1180 options understood by the compiler that fit into all specified classes
1181 and qualifiers. These are the supported classes:
1184 @item @samp{optimizers}
1185 This will display all of the optimization options supported by the
1188 @item @samp{warnings}
1189 This will display all of the options controlling warning messages
1190 produced by the compiler.
1193 This will display target-specific options. Unlike the
1194 @option{--target-help} option however, target-specific options of the
1195 linker and assembler will not be displayed. This is because those
1196 tools do not currently support the extended @option{--help=} syntax.
1199 This will display the values recognized by the @option{--param}
1202 @item @var{language}
1203 This will display the options supported for @var{language}, where
1204 @var{language} is the name of one of the languages supported in this
1208 This will display the options that are common to all languages.
1211 These are the supported qualifiers:
1214 @item @samp{undocumented}
1215 Display only those options which are undocumented.
1218 Display options which take an argument that appears after an equal
1219 sign in the same continuous piece of text, such as:
1220 @samp{--help=target}.
1222 @item @samp{separate}
1223 Display options which take an argument that appears as a separate word
1224 following the original option, such as: @samp{-o output-file}.
1227 Thus for example to display all the undocumented target-specific
1228 switches supported by the compiler the following can be used:
1231 --help=target,undocumented
1234 The sense of a qualifier can be inverted by prefixing it with the
1235 @samp{^} character, so for example to display all binary warning
1236 options (i.e., ones that are either on or off and that do not take an
1237 argument), which have a description the following can be used:
1240 --help=warnings,^joined,^undocumented
1243 The argument to @option{--help=} should not consist solely of inverted
1246 Combining several classes is possible, although this usually
1247 restricts the output by so much that there is nothing to display. One
1248 case where it does work however is when one of the classes is
1249 @var{target}. So for example to display all the target-specific
1250 optimization options the following can be used:
1253 --help=target,optimizers
1256 The @option{--help=} option can be repeated on the command line. Each
1257 successive use will display its requested class of options, skipping
1258 those that have already been displayed.
1260 If the @option{-Q} option appears on the command line before the
1261 @option{--help=} option, then the descriptive text displayed by
1262 @option{--help=} is changed. Instead of describing the displayed
1263 options, an indication is given as to whether the option is enabled,
1264 disabled or set to a specific value (assuming that the compiler
1265 knows this at the point where the @option{--help=} option is used).
1267 Here is a truncated example from the ARM port of @command{gcc}:
1270 % gcc -Q -mabi=2 --help=target -c
1271 The following options are target specific:
1273 -mabort-on-noreturn [disabled]
1277 The output is sensitive to the effects of previous command line
1278 options, so for example it is possible to find out which optimizations
1279 are enabled at @option{-O2} by using:
1282 -Q -O2 --help=optimizers
1285 Alternatively you can discover which binary optimizations are enabled
1286 by @option{-O3} by using:
1289 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1290 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1291 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1296 Display the version number and copyrights of the invoked GCC@.
1300 Invoke all subcommands under a wrapper program. It takes a single
1301 comma separated list as an argument, which will be used to invoke
1305 gcc -c t.c -wrapper gdb,--args
1308 This will invoke all subprograms of gcc under "gdb --args",
1309 thus cc1 invocation will be "gdb --args cc1 ...".
1311 @item -fplugin=@var{name}.so
1312 Load the plugin code in file @var{name}.so, assumed to be a
1313 shared object to be dlopen'd by the compiler. The base name of
1314 the shared object file is used to identify the plugin for the
1315 purposes of argument parsing (See
1316 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1317 Each plugin should define the callback functions specified in the
1320 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1321 Define an argument called @var{key} with a value of @var{value}
1322 for the plugin called @var{name}.
1324 @include @value{srcdir}/../libiberty/at-file.texi
1328 @section Compiling C++ Programs
1330 @cindex suffixes for C++ source
1331 @cindex C++ source file suffixes
1332 C++ source files conventionally use one of the suffixes @samp{.C},
1333 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1334 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1335 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1336 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1337 files with these names and compiles them as C++ programs even if you
1338 call the compiler the same way as for compiling C programs (usually
1339 with the name @command{gcc}).
1343 However, the use of @command{gcc} does not add the C++ library.
1344 @command{g++} is a program that calls GCC and treats @samp{.c},
1345 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1346 files unless @option{-x} is used, and automatically specifies linking
1347 against the C++ library. This program is also useful when
1348 precompiling a C header file with a @samp{.h} extension for use in C++
1349 compilations. On many systems, @command{g++} is also installed with
1350 the name @command{c++}.
1352 @cindex invoking @command{g++}
1353 When you compile C++ programs, you may specify many of the same
1354 command-line options that you use for compiling programs in any
1355 language; or command-line options meaningful for C and related
1356 languages; or options that are meaningful only for C++ programs.
1357 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1358 explanations of options for languages related to C@.
1359 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1360 explanations of options that are meaningful only for C++ programs.
1362 @node C Dialect Options
1363 @section Options Controlling C Dialect
1364 @cindex dialect options
1365 @cindex language dialect options
1366 @cindex options, dialect
1368 The following options control the dialect of C (or languages derived
1369 from C, such as C++, Objective-C and Objective-C++) that the compiler
1373 @cindex ANSI support
1377 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1378 equivalent to @samp{-std=c++98}.
1380 This turns off certain features of GCC that are incompatible with ISO
1381 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1382 such as the @code{asm} and @code{typeof} keywords, and
1383 predefined macros such as @code{unix} and @code{vax} that identify the
1384 type of system you are using. It also enables the undesirable and
1385 rarely used ISO trigraph feature. For the C compiler,
1386 it disables recognition of C++ style @samp{//} comments as well as
1387 the @code{inline} keyword.
1389 The alternate keywords @code{__asm__}, @code{__extension__},
1390 @code{__inline__} and @code{__typeof__} continue to work despite
1391 @option{-ansi}. You would not want to use them in an ISO C program, of
1392 course, but it is useful to put them in header files that might be included
1393 in compilations done with @option{-ansi}. Alternate predefined macros
1394 such as @code{__unix__} and @code{__vax__} are also available, with or
1395 without @option{-ansi}.
1397 The @option{-ansi} option does not cause non-ISO programs to be
1398 rejected gratuitously. For that, @option{-pedantic} is required in
1399 addition to @option{-ansi}. @xref{Warning Options}.
1401 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1402 option is used. Some header files may notice this macro and refrain
1403 from declaring certain functions or defining certain macros that the
1404 ISO standard doesn't call for; this is to avoid interfering with any
1405 programs that might use these names for other things.
1407 Functions that would normally be built in but do not have semantics
1408 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1409 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1410 built-in functions provided by GCC}, for details of the functions
1415 Determine the language standard. @xref{Standards,,Language Standards
1416 Supported by GCC}, for details of these standard versions. This option
1417 is currently only supported when compiling C or C++.
1419 The compiler can accept several base standards, such as @samp{c89} or
1420 @samp{c++98}, and GNU dialects of those standards, such as
1421 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1422 compiler will accept all programs following that standard and those
1423 using GNU extensions that do not contradict it. For example,
1424 @samp{-std=c89} turns off certain features of GCC that are
1425 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1426 keywords, but not other GNU extensions that do not have a meaning in
1427 ISO C90, such as omitting the middle term of a @code{?:}
1428 expression. On the other hand, by specifying a GNU dialect of a
1429 standard, all features the compiler support are enabled, even when
1430 those features change the meaning of the base standard and some
1431 strict-conforming programs may be rejected. The particular standard
1432 is used by @option{-pedantic} to identify which features are GNU
1433 extensions given that version of the standard. For example
1434 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1435 comments, while @samp{-std=gnu99 -pedantic} would not.
1437 A value for this option must be provided; possible values are
1442 Support all ISO C90 programs (certain GNU extensions that conflict
1443 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1445 @item iso9899:199409
1446 ISO C90 as modified in amendment 1.
1452 ISO C99. Note that this standard is not yet fully supported; see
1453 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1454 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1457 GNU dialect of ISO C90 (including some C99 features). This
1458 is the default for C code.
1462 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1463 this will become the default. The name @samp{gnu9x} is deprecated.
1466 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1470 GNU dialect of @option{-std=c++98}. This is the default for
1474 The working draft of the upcoming ISO C++0x standard. This option
1475 enables experimental features that are likely to be included in
1476 C++0x. The working draft is constantly changing, and any feature that is
1477 enabled by this flag may be removed from future versions of GCC if it is
1478 not part of the C++0x standard.
1481 GNU dialect of @option{-std=c++0x}. This option enables
1482 experimental features that may be removed in future versions of GCC.
1485 @item -fgnu89-inline
1486 @opindex fgnu89-inline
1487 The option @option{-fgnu89-inline} tells GCC to use the traditional
1488 GNU semantics for @code{inline} functions when in C99 mode.
1489 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1490 is accepted and ignored by GCC versions 4.1.3 up to but not including
1491 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1492 C99 mode. Using this option is roughly equivalent to adding the
1493 @code{gnu_inline} function attribute to all inline functions
1494 (@pxref{Function Attributes}).
1496 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1497 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1498 specifies the default behavior). This option was first supported in
1499 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1501 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1502 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1503 in effect for @code{inline} functions. @xref{Common Predefined
1504 Macros,,,cpp,The C Preprocessor}.
1506 @item -aux-info @var{filename}
1508 Output to the given filename prototyped declarations for all functions
1509 declared and/or defined in a translation unit, including those in header
1510 files. This option is silently ignored in any language other than C@.
1512 Besides declarations, the file indicates, in comments, the origin of
1513 each declaration (source file and line), whether the declaration was
1514 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1515 @samp{O} for old, respectively, in the first character after the line
1516 number and the colon), and whether it came from a declaration or a
1517 definition (@samp{C} or @samp{F}, respectively, in the following
1518 character). In the case of function definitions, a K&R-style list of
1519 arguments followed by their declarations is also provided, inside
1520 comments, after the declaration.
1524 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1525 keyword, so that code can use these words as identifiers. You can use
1526 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1527 instead. @option{-ansi} implies @option{-fno-asm}.
1529 In C++, this switch only affects the @code{typeof} keyword, since
1530 @code{asm} and @code{inline} are standard keywords. You may want to
1531 use the @option{-fno-gnu-keywords} flag instead, which has the same
1532 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1533 switch only affects the @code{asm} and @code{typeof} keywords, since
1534 @code{inline} is a standard keyword in ISO C99.
1537 @itemx -fno-builtin-@var{function}
1538 @opindex fno-builtin
1539 @cindex built-in functions
1540 Don't recognize built-in functions that do not begin with
1541 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1542 functions provided by GCC}, for details of the functions affected,
1543 including those which are not built-in functions when @option{-ansi} or
1544 @option{-std} options for strict ISO C conformance are used because they
1545 do not have an ISO standard meaning.
1547 GCC normally generates special code to handle certain built-in functions
1548 more efficiently; for instance, calls to @code{alloca} may become single
1549 instructions that adjust the stack directly, and calls to @code{memcpy}
1550 may become inline copy loops. The resulting code is often both smaller
1551 and faster, but since the function calls no longer appear as such, you
1552 cannot set a breakpoint on those calls, nor can you change the behavior
1553 of the functions by linking with a different library. In addition,
1554 when a function is recognized as a built-in function, GCC may use
1555 information about that function to warn about problems with calls to
1556 that function, or to generate more efficient code, even if the
1557 resulting code still contains calls to that function. For example,
1558 warnings are given with @option{-Wformat} for bad calls to
1559 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1560 known not to modify global memory.
1562 With the @option{-fno-builtin-@var{function}} option
1563 only the built-in function @var{function} is
1564 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1565 function is named that is not built-in in this version of GCC, this
1566 option is ignored. There is no corresponding
1567 @option{-fbuiltin-@var{function}} option; if you wish to enable
1568 built-in functions selectively when using @option{-fno-builtin} or
1569 @option{-ffreestanding}, you may define macros such as:
1572 #define abs(n) __builtin_abs ((n))
1573 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1578 @cindex hosted environment
1580 Assert that compilation takes place in a hosted environment. This implies
1581 @option{-fbuiltin}. A hosted environment is one in which the
1582 entire standard library is available, and in which @code{main} has a return
1583 type of @code{int}. Examples are nearly everything except a kernel.
1584 This is equivalent to @option{-fno-freestanding}.
1586 @item -ffreestanding
1587 @opindex ffreestanding
1588 @cindex hosted environment
1590 Assert that compilation takes place in a freestanding environment. This
1591 implies @option{-fno-builtin}. A freestanding environment
1592 is one in which the standard library may not exist, and program startup may
1593 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1594 This is equivalent to @option{-fno-hosted}.
1596 @xref{Standards,,Language Standards Supported by GCC}, for details of
1597 freestanding and hosted environments.
1601 @cindex openmp parallel
1602 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1603 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1604 compiler generates parallel code according to the OpenMP Application
1605 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1606 implies @option{-pthread}, and thus is only supported on targets that
1607 have support for @option{-pthread}.
1609 @item -fms-extensions
1610 @opindex fms-extensions
1611 Accept some non-standard constructs used in Microsoft header files.
1613 Some cases of unnamed fields in structures and unions are only
1614 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1615 fields within structs/unions}, for details.
1619 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1620 options for strict ISO C conformance) implies @option{-trigraphs}.
1622 @item -no-integrated-cpp
1623 @opindex no-integrated-cpp
1624 Performs a compilation in two passes: preprocessing and compiling. This
1625 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1626 @option{-B} option. The user supplied compilation step can then add in
1627 an additional preprocessing step after normal preprocessing but before
1628 compiling. The default is to use the integrated cpp (internal cpp)
1630 The semantics of this option will change if "cc1", "cc1plus", and
1631 "cc1obj" are merged.
1633 @cindex traditional C language
1634 @cindex C language, traditional
1636 @itemx -traditional-cpp
1637 @opindex traditional-cpp
1638 @opindex traditional
1639 Formerly, these options caused GCC to attempt to emulate a pre-standard
1640 C compiler. They are now only supported with the @option{-E} switch.
1641 The preprocessor continues to support a pre-standard mode. See the GNU
1642 CPP manual for details.
1644 @item -fcond-mismatch
1645 @opindex fcond-mismatch
1646 Allow conditional expressions with mismatched types in the second and
1647 third arguments. The value of such an expression is void. This option
1648 is not supported for C++.
1650 @item -flax-vector-conversions
1651 @opindex flax-vector-conversions
1652 Allow implicit conversions between vectors with differing numbers of
1653 elements and/or incompatible element types. This option should not be
1656 @item -funsigned-char
1657 @opindex funsigned-char
1658 Let the type @code{char} be unsigned, like @code{unsigned char}.
1660 Each kind of machine has a default for what @code{char} should
1661 be. It is either like @code{unsigned char} by default or like
1662 @code{signed char} by default.
1664 Ideally, a portable program should always use @code{signed char} or
1665 @code{unsigned char} when it depends on the signedness of an object.
1666 But many programs have been written to use plain @code{char} and
1667 expect it to be signed, or expect it to be unsigned, depending on the
1668 machines they were written for. This option, and its inverse, let you
1669 make such a program work with the opposite default.
1671 The type @code{char} is always a distinct type from each of
1672 @code{signed char} or @code{unsigned char}, even though its behavior
1673 is always just like one of those two.
1676 @opindex fsigned-char
1677 Let the type @code{char} be signed, like @code{signed char}.
1679 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1680 the negative form of @option{-funsigned-char}. Likewise, the option
1681 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1683 @item -fsigned-bitfields
1684 @itemx -funsigned-bitfields
1685 @itemx -fno-signed-bitfields
1686 @itemx -fno-unsigned-bitfields
1687 @opindex fsigned-bitfields
1688 @opindex funsigned-bitfields
1689 @opindex fno-signed-bitfields
1690 @opindex fno-unsigned-bitfields
1691 These options control whether a bit-field is signed or unsigned, when the
1692 declaration does not use either @code{signed} or @code{unsigned}. By
1693 default, such a bit-field is signed, because this is consistent: the
1694 basic integer types such as @code{int} are signed types.
1697 @node C++ Dialect Options
1698 @section Options Controlling C++ Dialect
1700 @cindex compiler options, C++
1701 @cindex C++ options, command line
1702 @cindex options, C++
1703 This section describes the command-line options that are only meaningful
1704 for C++ programs; but you can also use most of the GNU compiler options
1705 regardless of what language your program is in. For example, you
1706 might compile a file @code{firstClass.C} like this:
1709 g++ -g -frepo -O -c firstClass.C
1713 In this example, only @option{-frepo} is an option meant
1714 only for C++ programs; you can use the other options with any
1715 language supported by GCC@.
1717 Here is a list of options that are @emph{only} for compiling C++ programs:
1721 @item -fabi-version=@var{n}
1722 @opindex fabi-version
1723 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1724 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1725 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1726 the version that conforms most closely to the C++ ABI specification.
1727 Therefore, the ABI obtained using version 0 will change as ABI bugs
1730 The default is version 2.
1732 @item -fno-access-control
1733 @opindex fno-access-control
1734 Turn off all access checking. This switch is mainly useful for working
1735 around bugs in the access control code.
1739 Check that the pointer returned by @code{operator new} is non-null
1740 before attempting to modify the storage allocated. This check is
1741 normally unnecessary because the C++ standard specifies that
1742 @code{operator new} will only return @code{0} if it is declared
1743 @samp{throw()}, in which case the compiler will always check the
1744 return value even without this option. In all other cases, when
1745 @code{operator new} has a non-empty exception specification, memory
1746 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1747 @samp{new (nothrow)}.
1749 @item -fconserve-space
1750 @opindex fconserve-space
1751 Put uninitialized or runtime-initialized global variables into the
1752 common segment, as C does. This saves space in the executable at the
1753 cost of not diagnosing duplicate definitions. If you compile with this
1754 flag and your program mysteriously crashes after @code{main()} has
1755 completed, you may have an object that is being destroyed twice because
1756 two definitions were merged.
1758 This option is no longer useful on most targets, now that support has
1759 been added for putting variables into BSS without making them common.
1761 @item -ffriend-injection
1762 @opindex ffriend-injection
1763 Inject friend functions into the enclosing namespace, so that they are
1764 visible outside the scope of the class in which they are declared.
1765 Friend functions were documented to work this way in the old Annotated
1766 C++ Reference Manual, and versions of G++ before 4.1 always worked
1767 that way. However, in ISO C++ a friend function which is not declared
1768 in an enclosing scope can only be found using argument dependent
1769 lookup. This option causes friends to be injected as they were in
1772 This option is for compatibility, and may be removed in a future
1775 @item -fno-elide-constructors
1776 @opindex fno-elide-constructors
1777 The C++ standard allows an implementation to omit creating a temporary
1778 which is only used to initialize another object of the same type.
1779 Specifying this option disables that optimization, and forces G++ to
1780 call the copy constructor in all cases.
1782 @item -fno-enforce-eh-specs
1783 @opindex fno-enforce-eh-specs
1784 Don't generate code to check for violation of exception specifications
1785 at runtime. This option violates the C++ standard, but may be useful
1786 for reducing code size in production builds, much like defining
1787 @samp{NDEBUG}. This does not give user code permission to throw
1788 exceptions in violation of the exception specifications; the compiler
1789 will still optimize based on the specifications, so throwing an
1790 unexpected exception will result in undefined behavior.
1793 @itemx -fno-for-scope
1795 @opindex fno-for-scope
1796 If @option{-ffor-scope} is specified, the scope of variables declared in
1797 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1798 as specified by the C++ standard.
1799 If @option{-fno-for-scope} is specified, the scope of variables declared in
1800 a @i{for-init-statement} extends to the end of the enclosing scope,
1801 as was the case in old versions of G++, and other (traditional)
1802 implementations of C++.
1804 The default if neither flag is given to follow the standard,
1805 but to allow and give a warning for old-style code that would
1806 otherwise be invalid, or have different behavior.
1808 @item -fno-gnu-keywords
1809 @opindex fno-gnu-keywords
1810 Do not recognize @code{typeof} as a keyword, so that code can use this
1811 word as an identifier. You can use the keyword @code{__typeof__} instead.
1812 @option{-ansi} implies @option{-fno-gnu-keywords}.
1814 @item -fno-implicit-templates
1815 @opindex fno-implicit-templates
1816 Never emit code for non-inline templates which are instantiated
1817 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1818 @xref{Template Instantiation}, for more information.
1820 @item -fno-implicit-inline-templates
1821 @opindex fno-implicit-inline-templates
1822 Don't emit code for implicit instantiations of inline templates, either.
1823 The default is to handle inlines differently so that compiles with and
1824 without optimization will need the same set of explicit instantiations.
1826 @item -fno-implement-inlines
1827 @opindex fno-implement-inlines
1828 To save space, do not emit out-of-line copies of inline functions
1829 controlled by @samp{#pragma implementation}. This will cause linker
1830 errors if these functions are not inlined everywhere they are called.
1832 @item -fms-extensions
1833 @opindex fms-extensions
1834 Disable pedantic warnings about constructs used in MFC, such as implicit
1835 int and getting a pointer to member function via non-standard syntax.
1837 @item -fno-nonansi-builtins
1838 @opindex fno-nonansi-builtins
1839 Disable built-in declarations of functions that are not mandated by
1840 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1841 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1843 @item -fno-operator-names
1844 @opindex fno-operator-names
1845 Do not treat the operator name keywords @code{and}, @code{bitand},
1846 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1847 synonyms as keywords.
1849 @item -fno-optional-diags
1850 @opindex fno-optional-diags
1851 Disable diagnostics that the standard says a compiler does not need to
1852 issue. Currently, the only such diagnostic issued by G++ is the one for
1853 a name having multiple meanings within a class.
1856 @opindex fpermissive
1857 Downgrade some diagnostics about nonconformant code from errors to
1858 warnings. Thus, using @option{-fpermissive} will allow some
1859 nonconforming code to compile.
1861 @item -fno-pretty-templates
1862 @opindex fno-pretty-templates
1863 When an error message refers to a specialization of a function
1864 template, the compiler will normally print the signature of the
1865 template followed by the template arguments and any typedefs or
1866 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1867 rather than @code{void f(int)}) so that it's clear which template is
1868 involved. When an error message refers to a specialization of a class
1869 template, the compiler will omit any template arguments which match
1870 the default template arguments for that template. If either of these
1871 behaviors make it harder to understand the error message rather than
1872 easier, using @option{-fno-pretty-templates} will disable them.
1876 Enable automatic template instantiation at link time. This option also
1877 implies @option{-fno-implicit-templates}. @xref{Template
1878 Instantiation}, for more information.
1882 Disable generation of information about every class with virtual
1883 functions for use by the C++ runtime type identification features
1884 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1885 of the language, you can save some space by using this flag. Note that
1886 exception handling uses the same information, but it will generate it as
1887 needed. The @samp{dynamic_cast} operator can still be used for casts that
1888 do not require runtime type information, i.e.@: casts to @code{void *} or to
1889 unambiguous base classes.
1893 Emit statistics about front-end processing at the end of the compilation.
1894 This information is generally only useful to the G++ development team.
1896 @item -ftemplate-depth-@var{n}
1897 @opindex ftemplate-depth
1898 Set the maximum instantiation depth for template classes to @var{n}.
1899 A limit on the template instantiation depth is needed to detect
1900 endless recursions during template class instantiation. ANSI/ISO C++
1901 conforming programs must not rely on a maximum depth greater than 17.
1903 @item -fno-threadsafe-statics
1904 @opindex fno-threadsafe-statics
1905 Do not emit the extra code to use the routines specified in the C++
1906 ABI for thread-safe initialization of local statics. You can use this
1907 option to reduce code size slightly in code that doesn't need to be
1910 @item -fuse-cxa-atexit
1911 @opindex fuse-cxa-atexit
1912 Register destructors for objects with static storage duration with the
1913 @code{__cxa_atexit} function rather than the @code{atexit} function.
1914 This option is required for fully standards-compliant handling of static
1915 destructors, but will only work if your C library supports
1916 @code{__cxa_atexit}.
1918 @item -fno-use-cxa-get-exception-ptr
1919 @opindex fno-use-cxa-get-exception-ptr
1920 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1921 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1922 if the runtime routine is not available.
1924 @item -fvisibility-inlines-hidden
1925 @opindex fvisibility-inlines-hidden
1926 This switch declares that the user does not attempt to compare
1927 pointers to inline methods where the addresses of the two functions
1928 were taken in different shared objects.
1930 The effect of this is that GCC may, effectively, mark inline methods with
1931 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1932 appear in the export table of a DSO and do not require a PLT indirection
1933 when used within the DSO@. Enabling this option can have a dramatic effect
1934 on load and link times of a DSO as it massively reduces the size of the
1935 dynamic export table when the library makes heavy use of templates.
1937 The behavior of this switch is not quite the same as marking the
1938 methods as hidden directly, because it does not affect static variables
1939 local to the function or cause the compiler to deduce that
1940 the function is defined in only one shared object.
1942 You may mark a method as having a visibility explicitly to negate the
1943 effect of the switch for that method. For example, if you do want to
1944 compare pointers to a particular inline method, you might mark it as
1945 having default visibility. Marking the enclosing class with explicit
1946 visibility will have no effect.
1948 Explicitly instantiated inline methods are unaffected by this option
1949 as their linkage might otherwise cross a shared library boundary.
1950 @xref{Template Instantiation}.
1952 @item -fvisibility-ms-compat
1953 @opindex fvisibility-ms-compat
1954 This flag attempts to use visibility settings to make GCC's C++
1955 linkage model compatible with that of Microsoft Visual Studio.
1957 The flag makes these changes to GCC's linkage model:
1961 It sets the default visibility to @code{hidden}, like
1962 @option{-fvisibility=hidden}.
1965 Types, but not their members, are not hidden by default.
1968 The One Definition Rule is relaxed for types without explicit
1969 visibility specifications which are defined in more than one different
1970 shared object: those declarations are permitted if they would have
1971 been permitted when this option was not used.
1974 In new code it is better to use @option{-fvisibility=hidden} and
1975 export those classes which are intended to be externally visible.
1976 Unfortunately it is possible for code to rely, perhaps accidentally,
1977 on the Visual Studio behavior.
1979 Among the consequences of these changes are that static data members
1980 of the same type with the same name but defined in different shared
1981 objects will be different, so changing one will not change the other;
1982 and that pointers to function members defined in different shared
1983 objects may not compare equal. When this flag is given, it is a
1984 violation of the ODR to define types with the same name differently.
1988 Do not use weak symbol support, even if it is provided by the linker.
1989 By default, G++ will use weak symbols if they are available. This
1990 option exists only for testing, and should not be used by end-users;
1991 it will result in inferior code and has no benefits. This option may
1992 be removed in a future release of G++.
1996 Do not search for header files in the standard directories specific to
1997 C++, but do still search the other standard directories. (This option
1998 is used when building the C++ library.)
2001 In addition, these optimization, warning, and code generation options
2002 have meanings only for C++ programs:
2005 @item -fno-default-inline
2006 @opindex fno-default-inline
2007 Do not assume @samp{inline} for functions defined inside a class scope.
2008 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2009 functions will have linkage like inline functions; they just won't be
2012 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2015 Warn when G++ generates code that is probably not compatible with the
2016 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2017 all such cases, there are probably some cases that are not warned about,
2018 even though G++ is generating incompatible code. There may also be
2019 cases where warnings are emitted even though the code that is generated
2022 You should rewrite your code to avoid these warnings if you are
2023 concerned about the fact that code generated by G++ may not be binary
2024 compatible with code generated by other compilers.
2026 The known incompatibilities at this point include:
2031 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2032 pack data into the same byte as a base class. For example:
2035 struct A @{ virtual void f(); int f1 : 1; @};
2036 struct B : public A @{ int f2 : 1; @};
2040 In this case, G++ will place @code{B::f2} into the same byte
2041 as@code{A::f1}; other compilers will not. You can avoid this problem
2042 by explicitly padding @code{A} so that its size is a multiple of the
2043 byte size on your platform; that will cause G++ and other compilers to
2044 layout @code{B} identically.
2047 Incorrect handling of tail-padding for virtual bases. G++ does not use
2048 tail padding when laying out virtual bases. For example:
2051 struct A @{ virtual void f(); char c1; @};
2052 struct B @{ B(); char c2; @};
2053 struct C : public A, public virtual B @{@};
2057 In this case, G++ will not place @code{B} into the tail-padding for
2058 @code{A}; other compilers will. You can avoid this problem by
2059 explicitly padding @code{A} so that its size is a multiple of its
2060 alignment (ignoring virtual base classes); that will cause G++ and other
2061 compilers to layout @code{C} identically.
2064 Incorrect handling of bit-fields with declared widths greater than that
2065 of their underlying types, when the bit-fields appear in a union. For
2069 union U @{ int i : 4096; @};
2073 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2074 union too small by the number of bits in an @code{int}.
2077 Empty classes can be placed at incorrect offsets. For example:
2087 struct C : public B, public A @{@};
2091 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2092 it should be placed at offset zero. G++ mistakenly believes that the
2093 @code{A} data member of @code{B} is already at offset zero.
2096 Names of template functions whose types involve @code{typename} or
2097 template template parameters can be mangled incorrectly.
2100 template <typename Q>
2101 void f(typename Q::X) @{@}
2103 template <template <typename> class Q>
2104 void f(typename Q<int>::X) @{@}
2108 Instantiations of these templates may be mangled incorrectly.
2112 It also warns psABI related changes. The known psABI changes at this
2118 For SYSV/x86-64, when passing union with long double, it is changed to
2119 pass in memory as specified in psABI. For example:
2129 @code{union U} will always be passed in memory.
2133 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2134 @opindex Wctor-dtor-privacy
2135 @opindex Wno-ctor-dtor-privacy
2136 Warn when a class seems unusable because all the constructors or
2137 destructors in that class are private, and it has neither friends nor
2138 public static member functions.
2140 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2141 @opindex Wnon-virtual-dtor
2142 @opindex Wno-non-virtual-dtor
2143 Warn when a class has virtual functions and accessible non-virtual
2144 destructor, in which case it would be possible but unsafe to delete
2145 an instance of a derived class through a pointer to the base class.
2146 This warning is also enabled if -Weffc++ is specified.
2148 @item -Wreorder @r{(C++ and Objective-C++ only)}
2150 @opindex Wno-reorder
2151 @cindex reordering, warning
2152 @cindex warning for reordering of member initializers
2153 Warn when the order of member initializers given in the code does not
2154 match the order in which they must be executed. For instance:
2160 A(): j (0), i (1) @{ @}
2164 The compiler will rearrange the member initializers for @samp{i}
2165 and @samp{j} to match the declaration order of the members, emitting
2166 a warning to that effect. This warning is enabled by @option{-Wall}.
2169 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2172 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2175 Warn about violations of the following style guidelines from Scott Meyers'
2176 @cite{Effective C++} book:
2180 Item 11: Define a copy constructor and an assignment operator for classes
2181 with dynamically allocated memory.
2184 Item 12: Prefer initialization to assignment in constructors.
2187 Item 14: Make destructors virtual in base classes.
2190 Item 15: Have @code{operator=} return a reference to @code{*this}.
2193 Item 23: Don't try to return a reference when you must return an object.
2197 Also warn about violations of the following style guidelines from
2198 Scott Meyers' @cite{More Effective C++} book:
2202 Item 6: Distinguish between prefix and postfix forms of increment and
2203 decrement operators.
2206 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2210 When selecting this option, be aware that the standard library
2211 headers do not obey all of these guidelines; use @samp{grep -v}
2212 to filter out those warnings.
2214 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2215 @opindex Wstrict-null-sentinel
2216 @opindex Wno-strict-null-sentinel
2217 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2218 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2219 to @code{__null}. Although it is a null pointer constant not a null pointer,
2220 it is guaranteed to be of the same size as a pointer. But this use is
2221 not portable across different compilers.
2223 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2224 @opindex Wno-non-template-friend
2225 @opindex Wnon-template-friend
2226 Disable warnings when non-templatized friend functions are declared
2227 within a template. Since the advent of explicit template specification
2228 support in G++, if the name of the friend is an unqualified-id (i.e.,
2229 @samp{friend foo(int)}), the C++ language specification demands that the
2230 friend declare or define an ordinary, nontemplate function. (Section
2231 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2232 could be interpreted as a particular specialization of a templatized
2233 function. Because this non-conforming behavior is no longer the default
2234 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2235 check existing code for potential trouble spots and is on by default.
2236 This new compiler behavior can be turned off with
2237 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2238 but disables the helpful warning.
2240 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2241 @opindex Wold-style-cast
2242 @opindex Wno-old-style-cast
2243 Warn if an old-style (C-style) cast to a non-void type is used within
2244 a C++ program. The new-style casts (@samp{dynamic_cast},
2245 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2246 less vulnerable to unintended effects and much easier to search for.
2248 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2249 @opindex Woverloaded-virtual
2250 @opindex Wno-overloaded-virtual
2251 @cindex overloaded virtual fn, warning
2252 @cindex warning for overloaded virtual fn
2253 Warn when a function declaration hides virtual functions from a
2254 base class. For example, in:
2261 struct B: public A @{
2266 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2274 will fail to compile.
2276 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2277 @opindex Wno-pmf-conversions
2278 @opindex Wpmf-conversions
2279 Disable the diagnostic for converting a bound pointer to member function
2282 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2283 @opindex Wsign-promo
2284 @opindex Wno-sign-promo
2285 Warn when overload resolution chooses a promotion from unsigned or
2286 enumerated type to a signed type, over a conversion to an unsigned type of
2287 the same size. Previous versions of G++ would try to preserve
2288 unsignedness, but the standard mandates the current behavior.
2293 A& operator = (int);
2303 In this example, G++ will synthesize a default @samp{A& operator =
2304 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2307 @node Objective-C and Objective-C++ Dialect Options
2308 @section Options Controlling Objective-C and Objective-C++ Dialects
2310 @cindex compiler options, Objective-C and Objective-C++
2311 @cindex Objective-C and Objective-C++ options, command line
2312 @cindex options, Objective-C and Objective-C++
2313 (NOTE: This manual does not describe the Objective-C and Objective-C++
2314 languages themselves. See @xref{Standards,,Language Standards
2315 Supported by GCC}, for references.)
2317 This section describes the command-line options that are only meaningful
2318 for Objective-C and Objective-C++ programs, but you can also use most of
2319 the language-independent GNU compiler options.
2320 For example, you might compile a file @code{some_class.m} like this:
2323 gcc -g -fgnu-runtime -O -c some_class.m
2327 In this example, @option{-fgnu-runtime} is an option meant only for
2328 Objective-C and Objective-C++ programs; you can use the other options with
2329 any language supported by GCC@.
2331 Note that since Objective-C is an extension of the C language, Objective-C
2332 compilations may also use options specific to the C front-end (e.g.,
2333 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2334 C++-specific options (e.g., @option{-Wabi}).
2336 Here is a list of options that are @emph{only} for compiling Objective-C
2337 and Objective-C++ programs:
2340 @item -fconstant-string-class=@var{class-name}
2341 @opindex fconstant-string-class
2342 Use @var{class-name} as the name of the class to instantiate for each
2343 literal string specified with the syntax @code{@@"@dots{}"}. The default
2344 class name is @code{NXConstantString} if the GNU runtime is being used, and
2345 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2346 @option{-fconstant-cfstrings} option, if also present, will override the
2347 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2348 to be laid out as constant CoreFoundation strings.
2351 @opindex fgnu-runtime
2352 Generate object code compatible with the standard GNU Objective-C
2353 runtime. This is the default for most types of systems.
2355 @item -fnext-runtime
2356 @opindex fnext-runtime
2357 Generate output compatible with the NeXT runtime. This is the default
2358 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2359 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2362 @item -fno-nil-receivers
2363 @opindex fno-nil-receivers
2364 Assume that all Objective-C message dispatches (e.g.,
2365 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2366 is not @code{nil}. This allows for more efficient entry points in the runtime
2367 to be used. Currently, this option is only available in conjunction with
2368 the NeXT runtime on Mac OS X 10.3 and later.
2370 @item -fobjc-call-cxx-cdtors
2371 @opindex fobjc-call-cxx-cdtors
2372 For each Objective-C class, check if any of its instance variables is a
2373 C++ object with a non-trivial default constructor. If so, synthesize a
2374 special @code{- (id) .cxx_construct} instance method that will run
2375 non-trivial default constructors on any such instance variables, in order,
2376 and then return @code{self}. Similarly, check if any instance variable
2377 is a C++ object with a non-trivial destructor, and if so, synthesize a
2378 special @code{- (void) .cxx_destruct} method that will run
2379 all such default destructors, in reverse order.
2381 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2382 thusly generated will only operate on instance variables declared in the
2383 current Objective-C class, and not those inherited from superclasses. It
2384 is the responsibility of the Objective-C runtime to invoke all such methods
2385 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2386 will be invoked by the runtime immediately after a new object
2387 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2388 be invoked immediately before the runtime deallocates an object instance.
2390 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2391 support for invoking the @code{- (id) .cxx_construct} and
2392 @code{- (void) .cxx_destruct} methods.
2394 @item -fobjc-direct-dispatch
2395 @opindex fobjc-direct-dispatch
2396 Allow fast jumps to the message dispatcher. On Darwin this is
2397 accomplished via the comm page.
2399 @item -fobjc-exceptions
2400 @opindex fobjc-exceptions
2401 Enable syntactic support for structured exception handling in Objective-C,
2402 similar to what is offered by C++ and Java. This option is
2403 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2412 @@catch (AnObjCClass *exc) @{
2419 @@catch (AnotherClass *exc) @{
2422 @@catch (id allOthers) @{
2432 The @code{@@throw} statement may appear anywhere in an Objective-C or
2433 Objective-C++ program; when used inside of a @code{@@catch} block, the
2434 @code{@@throw} may appear without an argument (as shown above), in which case
2435 the object caught by the @code{@@catch} will be rethrown.
2437 Note that only (pointers to) Objective-C objects may be thrown and
2438 caught using this scheme. When an object is thrown, it will be caught
2439 by the nearest @code{@@catch} clause capable of handling objects of that type,
2440 analogously to how @code{catch} blocks work in C++ and Java. A
2441 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2442 any and all Objective-C exceptions not caught by previous @code{@@catch}
2445 The @code{@@finally} clause, if present, will be executed upon exit from the
2446 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2447 regardless of whether any exceptions are thrown, caught or rethrown
2448 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2449 of the @code{finally} clause in Java.
2451 There are several caveats to using the new exception mechanism:
2455 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2456 idioms provided by the @code{NSException} class, the new
2457 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2458 systems, due to additional functionality needed in the (NeXT) Objective-C
2462 As mentioned above, the new exceptions do not support handling
2463 types other than Objective-C objects. Furthermore, when used from
2464 Objective-C++, the Objective-C exception model does not interoperate with C++
2465 exceptions at this time. This means you cannot @code{@@throw} an exception
2466 from Objective-C and @code{catch} it in C++, or vice versa
2467 (i.e., @code{throw @dots{} @@catch}).
2470 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2471 blocks for thread-safe execution:
2474 @@synchronized (ObjCClass *guard) @{
2479 Upon entering the @code{@@synchronized} block, a thread of execution shall
2480 first check whether a lock has been placed on the corresponding @code{guard}
2481 object by another thread. If it has, the current thread shall wait until
2482 the other thread relinquishes its lock. Once @code{guard} becomes available,
2483 the current thread will place its own lock on it, execute the code contained in
2484 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2485 making @code{guard} available to other threads).
2487 Unlike Java, Objective-C does not allow for entire methods to be marked
2488 @code{@@synchronized}. Note that throwing exceptions out of
2489 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2490 to be unlocked properly.
2494 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2496 @item -freplace-objc-classes
2497 @opindex freplace-objc-classes
2498 Emit a special marker instructing @command{ld(1)} not to statically link in
2499 the resulting object file, and allow @command{dyld(1)} to load it in at
2500 run time instead. This is used in conjunction with the Fix-and-Continue
2501 debugging mode, where the object file in question may be recompiled and
2502 dynamically reloaded in the course of program execution, without the need
2503 to restart the program itself. Currently, Fix-and-Continue functionality
2504 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2509 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2510 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2511 compile time) with static class references that get initialized at load time,
2512 which improves run-time performance. Specifying the @option{-fzero-link} flag
2513 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2514 to be retained. This is useful in Zero-Link debugging mode, since it allows
2515 for individual class implementations to be modified during program execution.
2519 Dump interface declarations for all classes seen in the source file to a
2520 file named @file{@var{sourcename}.decl}.
2522 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2523 @opindex Wassign-intercept
2524 @opindex Wno-assign-intercept
2525 Warn whenever an Objective-C assignment is being intercepted by the
2528 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2529 @opindex Wno-protocol
2531 If a class is declared to implement a protocol, a warning is issued for
2532 every method in the protocol that is not implemented by the class. The
2533 default behavior is to issue a warning for every method not explicitly
2534 implemented in the class, even if a method implementation is inherited
2535 from the superclass. If you use the @option{-Wno-protocol} option, then
2536 methods inherited from the superclass are considered to be implemented,
2537 and no warning is issued for them.
2539 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2541 @opindex Wno-selector
2542 Warn if multiple methods of different types for the same selector are
2543 found during compilation. The check is performed on the list of methods
2544 in the final stage of compilation. Additionally, a check is performed
2545 for each selector appearing in a @code{@@selector(@dots{})}
2546 expression, and a corresponding method for that selector has been found
2547 during compilation. Because these checks scan the method table only at
2548 the end of compilation, these warnings are not produced if the final
2549 stage of compilation is not reached, for example because an error is
2550 found during compilation, or because the @option{-fsyntax-only} option is
2553 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2554 @opindex Wstrict-selector-match
2555 @opindex Wno-strict-selector-match
2556 Warn if multiple methods with differing argument and/or return types are
2557 found for a given selector when attempting to send a message using this
2558 selector to a receiver of type @code{id} or @code{Class}. When this flag
2559 is off (which is the default behavior), the compiler will omit such warnings
2560 if any differences found are confined to types which share the same size
2563 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2564 @opindex Wundeclared-selector
2565 @opindex Wno-undeclared-selector
2566 Warn if a @code{@@selector(@dots{})} expression referring to an
2567 undeclared selector is found. A selector is considered undeclared if no
2568 method with that name has been declared before the
2569 @code{@@selector(@dots{})} expression, either explicitly in an
2570 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2571 an @code{@@implementation} section. This option always performs its
2572 checks as soon as a @code{@@selector(@dots{})} expression is found,
2573 while @option{-Wselector} only performs its checks in the final stage of
2574 compilation. This also enforces the coding style convention
2575 that methods and selectors must be declared before being used.
2577 @item -print-objc-runtime-info
2578 @opindex print-objc-runtime-info
2579 Generate C header describing the largest structure that is passed by
2584 @node Language Independent Options
2585 @section Options to Control Diagnostic Messages Formatting
2586 @cindex options to control diagnostics formatting
2587 @cindex diagnostic messages
2588 @cindex message formatting
2590 Traditionally, diagnostic messages have been formatted irrespective of
2591 the output device's aspect (e.g.@: its width, @dots{}). The options described
2592 below can be used to control the diagnostic messages formatting
2593 algorithm, e.g.@: how many characters per line, how often source location
2594 information should be reported. Right now, only the C++ front end can
2595 honor these options. However it is expected, in the near future, that
2596 the remaining front ends would be able to digest them correctly.
2599 @item -fmessage-length=@var{n}
2600 @opindex fmessage-length
2601 Try to format error messages so that they fit on lines of about @var{n}
2602 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2603 the front ends supported by GCC@. If @var{n} is zero, then no
2604 line-wrapping will be done; each error message will appear on a single
2607 @opindex fdiagnostics-show-location
2608 @item -fdiagnostics-show-location=once
2609 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2610 reporter to emit @emph{once} source location information; that is, in
2611 case the message is too long to fit on a single physical line and has to
2612 be wrapped, the source location won't be emitted (as prefix) again,
2613 over and over, in subsequent continuation lines. This is the default
2616 @item -fdiagnostics-show-location=every-line
2617 Only meaningful in line-wrapping mode. Instructs the diagnostic
2618 messages reporter to emit the same source location information (as
2619 prefix) for physical lines that result from the process of breaking
2620 a message which is too long to fit on a single line.
2622 @item -fdiagnostics-show-option
2623 @opindex fdiagnostics-show-option
2624 This option instructs the diagnostic machinery to add text to each
2625 diagnostic emitted, which indicates which command line option directly
2626 controls that diagnostic, when such an option is known to the
2627 diagnostic machinery.
2629 @item -Wcoverage-mismatch
2630 @opindex Wcoverage-mismatch
2631 Warn if feedback profiles do not match when using the
2632 @option{-fprofile-use} option.
2633 If a source file was changed between @option{-fprofile-gen} and
2634 @option{-fprofile-use}, the files with the profile feedback can fail
2635 to match the source file and GCC can not use the profile feedback
2636 information. By default, GCC emits an error message in this case.
2637 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2638 error. GCC does not use appropriate feedback profiles, so using this
2639 option can result in poorly optimized code. This option is useful
2640 only in the case of very minor changes such as bug fixes to an
2645 @node Warning Options
2646 @section Options to Request or Suppress Warnings
2647 @cindex options to control warnings
2648 @cindex warning messages
2649 @cindex messages, warning
2650 @cindex suppressing warnings
2652 Warnings are diagnostic messages that report constructions which
2653 are not inherently erroneous but which are risky or suggest there
2654 may have been an error.
2656 The following language-independent options do not enable specific
2657 warnings but control the kinds of diagnostics produced by GCC.
2660 @cindex syntax checking
2662 @opindex fsyntax-only
2663 Check the code for syntax errors, but don't do anything beyond that.
2667 Inhibit all warning messages.
2672 Make all warnings into errors.
2677 Make the specified warning into an error. The specifier for a warning
2678 is appended, for example @option{-Werror=switch} turns the warnings
2679 controlled by @option{-Wswitch} into errors. This switch takes a
2680 negative form, to be used to negate @option{-Werror} for specific
2681 warnings, for example @option{-Wno-error=switch} makes
2682 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2683 is in effect. You can use the @option{-fdiagnostics-show-option}
2684 option to have each controllable warning amended with the option which
2685 controls it, to determine what to use with this option.
2687 Note that specifying @option{-Werror=}@var{foo} automatically implies
2688 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2691 @item -Wfatal-errors
2692 @opindex Wfatal-errors
2693 @opindex Wno-fatal-errors
2694 This option causes the compiler to abort compilation on the first error
2695 occurred rather than trying to keep going and printing further error
2700 You can request many specific warnings with options beginning
2701 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2702 implicit declarations. Each of these specific warning options also
2703 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2704 example, @option{-Wno-implicit}. This manual lists only one of the
2705 two forms, whichever is not the default. For further,
2706 language-specific options also refer to @ref{C++ Dialect Options} and
2707 @ref{Objective-C and Objective-C++ Dialect Options}.
2712 Issue all the warnings demanded by strict ISO C and ISO C++;
2713 reject all programs that use forbidden extensions, and some other
2714 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2715 version of the ISO C standard specified by any @option{-std} option used.
2717 Valid ISO C and ISO C++ programs should compile properly with or without
2718 this option (though a rare few will require @option{-ansi} or a
2719 @option{-std} option specifying the required version of ISO C)@. However,
2720 without this option, certain GNU extensions and traditional C and C++
2721 features are supported as well. With this option, they are rejected.
2723 @option{-pedantic} does not cause warning messages for use of the
2724 alternate keywords whose names begin and end with @samp{__}. Pedantic
2725 warnings are also disabled in the expression that follows
2726 @code{__extension__}. However, only system header files should use
2727 these escape routes; application programs should avoid them.
2728 @xref{Alternate Keywords}.
2730 Some users try to use @option{-pedantic} to check programs for strict ISO
2731 C conformance. They soon find that it does not do quite what they want:
2732 it finds some non-ISO practices, but not all---only those for which
2733 ISO C @emph{requires} a diagnostic, and some others for which
2734 diagnostics have been added.
2736 A feature to report any failure to conform to ISO C might be useful in
2737 some instances, but would require considerable additional work and would
2738 be quite different from @option{-pedantic}. We don't have plans to
2739 support such a feature in the near future.
2741 Where the standard specified with @option{-std} represents a GNU
2742 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2743 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2744 extended dialect is based. Warnings from @option{-pedantic} are given
2745 where they are required by the base standard. (It would not make sense
2746 for such warnings to be given only for features not in the specified GNU
2747 C dialect, since by definition the GNU dialects of C include all
2748 features the compiler supports with the given option, and there would be
2749 nothing to warn about.)
2751 @item -pedantic-errors
2752 @opindex pedantic-errors
2753 Like @option{-pedantic}, except that errors are produced rather than
2759 This enables all the warnings about constructions that some users
2760 consider questionable, and that are easy to avoid (or modify to
2761 prevent the warning), even in conjunction with macros. This also
2762 enables some language-specific warnings described in @ref{C++ Dialect
2763 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2765 @option{-Wall} turns on the following warning flags:
2767 @gccoptlist{-Waddress @gol
2768 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2770 -Wchar-subscripts @gol
2771 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2773 -Wimplicit-function-declaration @gol
2776 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2777 -Wmissing-braces @gol
2783 -Wsequence-point @gol
2784 -Wsign-compare @r{(only in C++)} @gol
2785 -Wstrict-aliasing @gol
2786 -Wstrict-overflow=1 @gol
2789 -Wuninitialized @gol
2790 -Wunknown-pragmas @gol
2791 -Wunused-function @gol
2794 -Wunused-variable @gol
2795 -Wvolatile-register-var @gol
2798 Note that some warning flags are not implied by @option{-Wall}. Some of
2799 them warn about constructions that users generally do not consider
2800 questionable, but which occasionally you might wish to check for;
2801 others warn about constructions that are necessary or hard to avoid in
2802 some cases, and there is no simple way to modify the code to suppress
2803 the warning. Some of them are enabled by @option{-Wextra} but many of
2804 them must be enabled individually.
2810 This enables some extra warning flags that are not enabled by
2811 @option{-Wall}. (This option used to be called @option{-W}. The older
2812 name is still supported, but the newer name is more descriptive.)
2814 @gccoptlist{-Wclobbered @gol
2816 -Wignored-qualifiers @gol
2817 -Wmissing-field-initializers @gol
2818 -Wmissing-parameter-type @r{(C only)} @gol
2819 -Wold-style-declaration @r{(C only)} @gol
2820 -Woverride-init @gol
2823 -Wuninitialized @gol
2824 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2827 The option @option{-Wextra} also prints warning messages for the
2833 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2834 @samp{>}, or @samp{>=}.
2837 (C++ only) An enumerator and a non-enumerator both appear in a
2838 conditional expression.
2841 (C++ only) Ambiguous virtual bases.
2844 (C++ only) Subscripting an array which has been declared @samp{register}.
2847 (C++ only) Taking the address of a variable which has been declared
2851 (C++ only) A base class is not initialized in a derived class' copy
2856 @item -Wchar-subscripts
2857 @opindex Wchar-subscripts
2858 @opindex Wno-char-subscripts
2859 Warn if an array subscript has type @code{char}. This is a common cause
2860 of error, as programmers often forget that this type is signed on some
2862 This warning is enabled by @option{-Wall}.
2866 @opindex Wno-comment
2867 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2868 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2869 This warning is enabled by @option{-Wall}.
2874 @opindex ffreestanding
2875 @opindex fno-builtin
2876 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2877 the arguments supplied have types appropriate to the format string
2878 specified, and that the conversions specified in the format string make
2879 sense. This includes standard functions, and others specified by format
2880 attributes (@pxref{Function Attributes}), in the @code{printf},
2881 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2882 not in the C standard) families (or other target-specific families).
2883 Which functions are checked without format attributes having been
2884 specified depends on the standard version selected, and such checks of
2885 functions without the attribute specified are disabled by
2886 @option{-ffreestanding} or @option{-fno-builtin}.
2888 The formats are checked against the format features supported by GNU
2889 libc version 2.2. These include all ISO C90 and C99 features, as well
2890 as features from the Single Unix Specification and some BSD and GNU
2891 extensions. Other library implementations may not support all these
2892 features; GCC does not support warning about features that go beyond a
2893 particular library's limitations. However, if @option{-pedantic} is used
2894 with @option{-Wformat}, warnings will be given about format features not
2895 in the selected standard version (but not for @code{strfmon} formats,
2896 since those are not in any version of the C standard). @xref{C Dialect
2897 Options,,Options Controlling C Dialect}.
2899 Since @option{-Wformat} also checks for null format arguments for
2900 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2902 @option{-Wformat} is included in @option{-Wall}. For more control over some
2903 aspects of format checking, the options @option{-Wformat-y2k},
2904 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2905 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2906 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2909 @opindex Wformat-y2k
2910 @opindex Wno-format-y2k
2911 If @option{-Wformat} is specified, also warn about @code{strftime}
2912 formats which may yield only a two-digit year.
2914 @item -Wno-format-contains-nul
2915 @opindex Wno-format-contains-nul
2916 @opindex Wformat-contains-nul
2917 If @option{-Wformat} is specified, do not warn about format strings that
2920 @item -Wno-format-extra-args
2921 @opindex Wno-format-extra-args
2922 @opindex Wformat-extra-args
2923 If @option{-Wformat} is specified, do not warn about excess arguments to a
2924 @code{printf} or @code{scanf} format function. The C standard specifies
2925 that such arguments are ignored.
2927 Where the unused arguments lie between used arguments that are
2928 specified with @samp{$} operand number specifications, normally
2929 warnings are still given, since the implementation could not know what
2930 type to pass to @code{va_arg} to skip the unused arguments. However,
2931 in the case of @code{scanf} formats, this option will suppress the
2932 warning if the unused arguments are all pointers, since the Single
2933 Unix Specification says that such unused arguments are allowed.
2935 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2936 @opindex Wno-format-zero-length
2937 @opindex Wformat-zero-length
2938 If @option{-Wformat} is specified, do not warn about zero-length formats.
2939 The C standard specifies that zero-length formats are allowed.
2941 @item -Wformat-nonliteral
2942 @opindex Wformat-nonliteral
2943 @opindex Wno-format-nonliteral
2944 If @option{-Wformat} is specified, also warn if the format string is not a
2945 string literal and so cannot be checked, unless the format function
2946 takes its format arguments as a @code{va_list}.
2948 @item -Wformat-security
2949 @opindex Wformat-security
2950 @opindex Wno-format-security
2951 If @option{-Wformat} is specified, also warn about uses of format
2952 functions that represent possible security problems. At present, this
2953 warns about calls to @code{printf} and @code{scanf} functions where the
2954 format string is not a string literal and there are no format arguments,
2955 as in @code{printf (foo);}. This may be a security hole if the format
2956 string came from untrusted input and contains @samp{%n}. (This is
2957 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2958 in future warnings may be added to @option{-Wformat-security} that are not
2959 included in @option{-Wformat-nonliteral}.)
2963 @opindex Wno-format=2
2964 Enable @option{-Wformat} plus format checks not included in
2965 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2966 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2968 @item -Wnonnull @r{(C and Objective-C only)}
2970 @opindex Wno-nonnull
2971 Warn about passing a null pointer for arguments marked as
2972 requiring a non-null value by the @code{nonnull} function attribute.
2974 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2975 can be disabled with the @option{-Wno-nonnull} option.
2977 @item -Wjump-misses-init @r{(C, Objective-C only)}
2978 @opindex Wjump-misses-init
2979 @opindex Wno-jump-misses-init
2980 Warn if a @code{goto} statement or a @code{switch} statement jumps
2981 forward across the initialization of a variable, or jumps backward to a
2982 label after the variable has been initialized. This only warns about
2983 variables which are initialized when they are declared. This warning is
2984 only supported for C and Objective C; in C++ this sort of branch is an
2987 @option{-Wjump-misses-init} is included in @option{-Wall} and
2988 @option{-Wc++-compat}. It can be disabled with the
2989 @option{-Wno-jump-misses-init} option.
2991 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2993 @opindex Wno-init-self
2994 Warn about uninitialized variables which are initialized with themselves.
2995 Note this option can only be used with the @option{-Wuninitialized} option.
2997 For example, GCC will warn about @code{i} being uninitialized in the
2998 following snippet only when @option{-Winit-self} has been specified:
3009 @item -Wimplicit-int @r{(C and Objective-C only)}
3010 @opindex Wimplicit-int
3011 @opindex Wno-implicit-int
3012 Warn when a declaration does not specify a type.
3013 This warning is enabled by @option{-Wall}.
3015 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3016 @opindex Wimplicit-function-declaration
3017 @opindex Wno-implicit-function-declaration
3018 Give a warning whenever a function is used before being declared. In
3019 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3020 enabled by default and it is made into an error by
3021 @option{-pedantic-errors}. This warning is also enabled by
3026 @opindex Wno-implicit
3027 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3028 This warning is enabled by @option{-Wall}.
3030 @item -Wignored-qualifiers @r{(C and C++ only)}
3031 @opindex Wignored-qualifiers
3032 @opindex Wno-ignored-qualifiers
3033 Warn if the return type of a function has a type qualifier
3034 such as @code{const}. For ISO C such a type qualifier has no effect,
3035 since the value returned by a function is not an lvalue.
3036 For C++, the warning is only emitted for scalar types or @code{void}.
3037 ISO C prohibits qualified @code{void} return types on function
3038 definitions, so such return types always receive a warning
3039 even without this option.
3041 This warning is also enabled by @option{-Wextra}.
3046 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3047 a function with external linkage, returning int, taking either zero
3048 arguments, two, or three arguments of appropriate types. This warning
3049 is enabled by default in C++ and is enabled by either @option{-Wall}
3050 or @option{-pedantic}.
3052 @item -Wmissing-braces
3053 @opindex Wmissing-braces
3054 @opindex Wno-missing-braces
3055 Warn if an aggregate or union initializer is not fully bracketed. In
3056 the following example, the initializer for @samp{a} is not fully
3057 bracketed, but that for @samp{b} is fully bracketed.
3060 int a[2][2] = @{ 0, 1, 2, 3 @};
3061 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3064 This warning is enabled by @option{-Wall}.
3066 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3067 @opindex Wmissing-include-dirs
3068 @opindex Wno-missing-include-dirs
3069 Warn if a user-supplied include directory does not exist.
3072 @opindex Wparentheses
3073 @opindex Wno-parentheses
3074 Warn if parentheses are omitted in certain contexts, such
3075 as when there is an assignment in a context where a truth value
3076 is expected, or when operators are nested whose precedence people
3077 often get confused about.
3079 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3080 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3081 interpretation from that of ordinary mathematical notation.
3083 Also warn about constructions where there may be confusion to which
3084 @code{if} statement an @code{else} branch belongs. Here is an example of
3099 In C/C++, every @code{else} branch belongs to the innermost possible
3100 @code{if} statement, which in this example is @code{if (b)}. This is
3101 often not what the programmer expected, as illustrated in the above
3102 example by indentation the programmer chose. When there is the
3103 potential for this confusion, GCC will issue a warning when this flag
3104 is specified. To eliminate the warning, add explicit braces around
3105 the innermost @code{if} statement so there is no way the @code{else}
3106 could belong to the enclosing @code{if}. The resulting code would
3123 This warning is enabled by @option{-Wall}.
3125 @item -Wsequence-point
3126 @opindex Wsequence-point
3127 @opindex Wno-sequence-point
3128 Warn about code that may have undefined semantics because of violations
3129 of sequence point rules in the C and C++ standards.
3131 The C and C++ standards defines the order in which expressions in a C/C++
3132 program are evaluated in terms of @dfn{sequence points}, which represent
3133 a partial ordering between the execution of parts of the program: those
3134 executed before the sequence point, and those executed after it. These
3135 occur after the evaluation of a full expression (one which is not part
3136 of a larger expression), after the evaluation of the first operand of a
3137 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3138 function is called (but after the evaluation of its arguments and the
3139 expression denoting the called function), and in certain other places.
3140 Other than as expressed by the sequence point rules, the order of
3141 evaluation of subexpressions of an expression is not specified. All
3142 these rules describe only a partial order rather than a total order,
3143 since, for example, if two functions are called within one expression
3144 with no sequence point between them, the order in which the functions
3145 are called is not specified. However, the standards committee have
3146 ruled that function calls do not overlap.
3148 It is not specified when between sequence points modifications to the
3149 values of objects take effect. Programs whose behavior depends on this
3150 have undefined behavior; the C and C++ standards specify that ``Between
3151 the previous and next sequence point an object shall have its stored
3152 value modified at most once by the evaluation of an expression.
3153 Furthermore, the prior value shall be read only to determine the value
3154 to be stored.''. If a program breaks these rules, the results on any
3155 particular implementation are entirely unpredictable.
3157 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3158 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3159 diagnosed by this option, and it may give an occasional false positive
3160 result, but in general it has been found fairly effective at detecting
3161 this sort of problem in programs.
3163 The standard is worded confusingly, therefore there is some debate
3164 over the precise meaning of the sequence point rules in subtle cases.
3165 Links to discussions of the problem, including proposed formal
3166 definitions, may be found on the GCC readings page, at
3167 @w{@uref{http://gcc.gnu.org/readings.html}}.
3169 This warning is enabled by @option{-Wall} for C and C++.
3172 @opindex Wreturn-type
3173 @opindex Wno-return-type
3174 Warn whenever a function is defined with a return-type that defaults
3175 to @code{int}. Also warn about any @code{return} statement with no
3176 return-value in a function whose return-type is not @code{void}
3177 (falling off the end of the function body is considered returning
3178 without a value), and about a @code{return} statement with an
3179 expression in a function whose return-type is @code{void}.
3181 For C++, a function without return type always produces a diagnostic
3182 message, even when @option{-Wno-return-type} is specified. The only
3183 exceptions are @samp{main} and functions defined in system headers.
3185 This warning is enabled by @option{-Wall}.
3190 Warn whenever a @code{switch} statement has an index of enumerated type
3191 and lacks a @code{case} for one or more of the named codes of that
3192 enumeration. (The presence of a @code{default} label prevents this
3193 warning.) @code{case} labels outside the enumeration range also
3194 provoke warnings when this option is used (even if there is a
3195 @code{default} label).
3196 This warning is enabled by @option{-Wall}.
3198 @item -Wswitch-default
3199 @opindex Wswitch-default
3200 @opindex Wno-switch-default
3201 Warn whenever a @code{switch} statement does not have a @code{default}
3205 @opindex Wswitch-enum
3206 @opindex Wno-switch-enum
3207 Warn whenever a @code{switch} statement has an index of enumerated type
3208 and lacks a @code{case} for one or more of the named codes of that
3209 enumeration. @code{case} labels outside the enumeration range also
3210 provoke warnings when this option is used. The only difference
3211 between @option{-Wswitch} and this option is that this option gives a
3212 warning about an omitted enumeration code even if there is a
3213 @code{default} label.
3215 @item -Wsync-nand @r{(C and C++ only)}
3217 @opindex Wno-sync-nand
3218 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3219 built-in functions are used. These functions changed semantics in GCC 4.4.
3223 @opindex Wno-trigraphs
3224 Warn if any trigraphs are encountered that might change the meaning of
3225 the program (trigraphs within comments are not warned about).
3226 This warning is enabled by @option{-Wall}.
3228 @item -Wunused-function
3229 @opindex Wunused-function
3230 @opindex Wno-unused-function
3231 Warn whenever a static function is declared but not defined or a
3232 non-inline static function is unused.
3233 This warning is enabled by @option{-Wall}.
3235 @item -Wunused-label
3236 @opindex Wunused-label
3237 @opindex Wno-unused-label
3238 Warn whenever a label is declared but not used.
3239 This warning is enabled by @option{-Wall}.
3241 To suppress this warning use the @samp{unused} attribute
3242 (@pxref{Variable Attributes}).
3244 @item -Wunused-parameter
3245 @opindex Wunused-parameter
3246 @opindex Wno-unused-parameter
3247 Warn whenever a function parameter is unused aside from its declaration.
3249 To suppress this warning use the @samp{unused} attribute
3250 (@pxref{Variable Attributes}).
3252 @item -Wunused-variable
3253 @opindex Wunused-variable
3254 @opindex Wno-unused-variable
3255 Warn whenever a local variable or non-constant static variable is unused
3256 aside from its declaration.
3257 This warning is enabled by @option{-Wall}.
3259 To suppress this warning use the @samp{unused} attribute
3260 (@pxref{Variable Attributes}).
3262 @item -Wunused-value
3263 @opindex Wunused-value
3264 @opindex Wno-unused-value
3265 Warn whenever a statement computes a result that is explicitly not
3266 used. To suppress this warning cast the unused expression to
3267 @samp{void}. This includes an expression-statement or the left-hand
3268 side of a comma expression that contains no side effects. For example,
3269 an expression such as @samp{x[i,j]} will cause a warning, while
3270 @samp{x[(void)i,j]} will not.
3272 This warning is enabled by @option{-Wall}.
3277 All the above @option{-Wunused} options combined.
3279 In order to get a warning about an unused function parameter, you must
3280 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3281 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3283 @item -Wuninitialized
3284 @opindex Wuninitialized
3285 @opindex Wno-uninitialized
3286 Warn if an automatic variable is used without first being initialized
3287 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3288 warn if a non-static reference or non-static @samp{const} member
3289 appears in a class without constructors.
3291 If you want to warn about code which uses the uninitialized value of the
3292 variable in its own initializer, use the @option{-Winit-self} option.
3294 These warnings occur for individual uninitialized or clobbered
3295 elements of structure, union or array variables as well as for
3296 variables which are uninitialized or clobbered as a whole. They do
3297 not occur for variables or elements declared @code{volatile}. Because
3298 these warnings depend on optimization, the exact variables or elements
3299 for which there are warnings will depend on the precise optimization
3300 options and version of GCC used.
3302 Note that there may be no warning about a variable that is used only
3303 to compute a value that itself is never used, because such
3304 computations may be deleted by data flow analysis before the warnings
3307 These warnings are made optional because GCC is not smart
3308 enough to see all the reasons why the code might be correct
3309 despite appearing to have an error. Here is one example of how
3330 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3331 always initialized, but GCC doesn't know this. Here is
3332 another common case:
3337 if (change_y) save_y = y, y = new_y;
3339 if (change_y) y = save_y;
3344 This has no bug because @code{save_y} is used only if it is set.
3346 @cindex @code{longjmp} warnings
3347 This option also warns when a non-volatile automatic variable might be
3348 changed by a call to @code{longjmp}. These warnings as well are possible
3349 only in optimizing compilation.
3351 The compiler sees only the calls to @code{setjmp}. It cannot know
3352 where @code{longjmp} will be called; in fact, a signal handler could
3353 call it at any point in the code. As a result, you may get a warning
3354 even when there is in fact no problem because @code{longjmp} cannot
3355 in fact be called at the place which would cause a problem.
3357 Some spurious warnings can be avoided if you declare all the functions
3358 you use that never return as @code{noreturn}. @xref{Function
3361 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3363 @item -Wunknown-pragmas
3364 @opindex Wunknown-pragmas
3365 @opindex Wno-unknown-pragmas
3366 @cindex warning for unknown pragmas
3367 @cindex unknown pragmas, warning
3368 @cindex pragmas, warning of unknown
3369 Warn when a #pragma directive is encountered which is not understood by
3370 GCC@. If this command line option is used, warnings will even be issued
3371 for unknown pragmas in system header files. This is not the case if
3372 the warnings were only enabled by the @option{-Wall} command line option.
3375 @opindex Wno-pragmas
3377 Do not warn about misuses of pragmas, such as incorrect parameters,
3378 invalid syntax, or conflicts between pragmas. See also
3379 @samp{-Wunknown-pragmas}.
3381 @item -Wstrict-aliasing
3382 @opindex Wstrict-aliasing
3383 @opindex Wno-strict-aliasing
3384 This option is only active when @option{-fstrict-aliasing} is active.
3385 It warns about code which might break the strict aliasing rules that the
3386 compiler is using for optimization. The warning does not catch all
3387 cases, but does attempt to catch the more common pitfalls. It is
3388 included in @option{-Wall}.
3389 It is equivalent to @option{-Wstrict-aliasing=3}
3391 @item -Wstrict-aliasing=n
3392 @opindex Wstrict-aliasing=n
3393 @opindex Wno-strict-aliasing=n
3394 This option is only active when @option{-fstrict-aliasing} is active.
3395 It warns about code which might break the strict aliasing rules that the
3396 compiler is using for optimization.
3397 Higher levels correspond to higher accuracy (fewer false positives).
3398 Higher levels also correspond to more effort, similar to the way -O works.
3399 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3402 Level 1: Most aggressive, quick, least accurate.
3403 Possibly useful when higher levels
3404 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3405 false negatives. However, it has many false positives.
3406 Warns for all pointer conversions between possibly incompatible types,
3407 even if never dereferenced. Runs in the frontend only.
3409 Level 2: Aggressive, quick, not too precise.
3410 May still have many false positives (not as many as level 1 though),
3411 and few false negatives (but possibly more than level 1).
3412 Unlike level 1, it only warns when an address is taken. Warns about
3413 incomplete types. Runs in the frontend only.
3415 Level 3 (default for @option{-Wstrict-aliasing}):
3416 Should have very few false positives and few false
3417 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3418 Takes care of the common punn+dereference pattern in the frontend:
3419 @code{*(int*)&some_float}.
3420 If optimization is enabled, it also runs in the backend, where it deals
3421 with multiple statement cases using flow-sensitive points-to information.
3422 Only warns when the converted pointer is dereferenced.
3423 Does not warn about incomplete types.
3425 @item -Wstrict-overflow
3426 @itemx -Wstrict-overflow=@var{n}
3427 @opindex Wstrict-overflow
3428 @opindex Wno-strict-overflow
3429 This option is only active when @option{-fstrict-overflow} is active.
3430 It warns about cases where the compiler optimizes based on the
3431 assumption that signed overflow does not occur. Note that it does not
3432 warn about all cases where the code might overflow: it only warns
3433 about cases where the compiler implements some optimization. Thus
3434 this warning depends on the optimization level.
3436 An optimization which assumes that signed overflow does not occur is
3437 perfectly safe if the values of the variables involved are such that
3438 overflow never does, in fact, occur. Therefore this warning can
3439 easily give a false positive: a warning about code which is not
3440 actually a problem. To help focus on important issues, several
3441 warning levels are defined. No warnings are issued for the use of
3442 undefined signed overflow when estimating how many iterations a loop
3443 will require, in particular when determining whether a loop will be
3447 @item -Wstrict-overflow=1
3448 Warn about cases which are both questionable and easy to avoid. For
3449 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3450 compiler will simplify this to @code{1}. This level of
3451 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3452 are not, and must be explicitly requested.
3454 @item -Wstrict-overflow=2
3455 Also warn about other cases where a comparison is simplified to a
3456 constant. For example: @code{abs (x) >= 0}. This can only be
3457 simplified when @option{-fstrict-overflow} is in effect, because
3458 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3459 zero. @option{-Wstrict-overflow} (with no level) is the same as
3460 @option{-Wstrict-overflow=2}.
3462 @item -Wstrict-overflow=3
3463 Also warn about other cases where a comparison is simplified. For
3464 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3466 @item -Wstrict-overflow=4
3467 Also warn about other simplifications not covered by the above cases.
3468 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3470 @item -Wstrict-overflow=5
3471 Also warn about cases where the compiler reduces the magnitude of a
3472 constant involved in a comparison. For example: @code{x + 2 > y} will
3473 be simplified to @code{x + 1 >= y}. This is reported only at the
3474 highest warning level because this simplification applies to many
3475 comparisons, so this warning level will give a very large number of
3479 @item -Warray-bounds
3480 @opindex Wno-array-bounds
3481 @opindex Warray-bounds
3482 This option is only active when @option{-ftree-vrp} is active
3483 (default for -O2 and above). It warns about subscripts to arrays
3484 that are always out of bounds. This warning is enabled by @option{-Wall}.
3486 @item -Wno-div-by-zero
3487 @opindex Wno-div-by-zero
3488 @opindex Wdiv-by-zero
3489 Do not warn about compile-time integer division by zero. Floating point
3490 division by zero is not warned about, as it can be a legitimate way of
3491 obtaining infinities and NaNs.
3493 @item -Wsystem-headers
3494 @opindex Wsystem-headers
3495 @opindex Wno-system-headers
3496 @cindex warnings from system headers
3497 @cindex system headers, warnings from
3498 Print warning messages for constructs found in system header files.
3499 Warnings from system headers are normally suppressed, on the assumption
3500 that they usually do not indicate real problems and would only make the
3501 compiler output harder to read. Using this command line option tells
3502 GCC to emit warnings from system headers as if they occurred in user
3503 code. However, note that using @option{-Wall} in conjunction with this
3504 option will @emph{not} warn about unknown pragmas in system
3505 headers---for that, @option{-Wunknown-pragmas} must also be used.
3508 @opindex Wfloat-equal
3509 @opindex Wno-float-equal
3510 Warn if floating point values are used in equality comparisons.
3512 The idea behind this is that sometimes it is convenient (for the
3513 programmer) to consider floating-point values as approximations to
3514 infinitely precise real numbers. If you are doing this, then you need
3515 to compute (by analyzing the code, or in some other way) the maximum or
3516 likely maximum error that the computation introduces, and allow for it
3517 when performing comparisons (and when producing output, but that's a
3518 different problem). In particular, instead of testing for equality, you
3519 would check to see whether the two values have ranges that overlap; and
3520 this is done with the relational operators, so equality comparisons are
3523 @item -Wtraditional @r{(C and Objective-C only)}
3524 @opindex Wtraditional
3525 @opindex Wno-traditional
3526 Warn about certain constructs that behave differently in traditional and
3527 ISO C@. Also warn about ISO C constructs that have no traditional C
3528 equivalent, and/or problematic constructs which should be avoided.
3532 Macro parameters that appear within string literals in the macro body.
3533 In traditional C macro replacement takes place within string literals,
3534 but does not in ISO C@.
3537 In traditional C, some preprocessor directives did not exist.
3538 Traditional preprocessors would only consider a line to be a directive
3539 if the @samp{#} appeared in column 1 on the line. Therefore
3540 @option{-Wtraditional} warns about directives that traditional C
3541 understands but would ignore because the @samp{#} does not appear as the
3542 first character on the line. It also suggests you hide directives like
3543 @samp{#pragma} not understood by traditional C by indenting them. Some
3544 traditional implementations would not recognize @samp{#elif}, so it
3545 suggests avoiding it altogether.
3548 A function-like macro that appears without arguments.
3551 The unary plus operator.
3554 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3555 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3556 constants.) Note, these suffixes appear in macros defined in the system
3557 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3558 Use of these macros in user code might normally lead to spurious
3559 warnings, however GCC's integrated preprocessor has enough context to
3560 avoid warning in these cases.
3563 A function declared external in one block and then used after the end of
3567 A @code{switch} statement has an operand of type @code{long}.
3570 A non-@code{static} function declaration follows a @code{static} one.
3571 This construct is not accepted by some traditional C compilers.
3574 The ISO type of an integer constant has a different width or
3575 signedness from its traditional type. This warning is only issued if
3576 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3577 typically represent bit patterns, are not warned about.
3580 Usage of ISO string concatenation is detected.
3583 Initialization of automatic aggregates.
3586 Identifier conflicts with labels. Traditional C lacks a separate
3587 namespace for labels.
3590 Initialization of unions. If the initializer is zero, the warning is
3591 omitted. This is done under the assumption that the zero initializer in
3592 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3593 initializer warnings and relies on default initialization to zero in the
3597 Conversions by prototypes between fixed/floating point values and vice
3598 versa. The absence of these prototypes when compiling with traditional
3599 C would cause serious problems. This is a subset of the possible
3600 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3603 Use of ISO C style function definitions. This warning intentionally is
3604 @emph{not} issued for prototype declarations or variadic functions
3605 because these ISO C features will appear in your code when using
3606 libiberty's traditional C compatibility macros, @code{PARAMS} and
3607 @code{VPARAMS}. This warning is also bypassed for nested functions
3608 because that feature is already a GCC extension and thus not relevant to
3609 traditional C compatibility.
3612 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3613 @opindex Wtraditional-conversion
3614 @opindex Wno-traditional-conversion
3615 Warn if a prototype causes a type conversion that is different from what
3616 would happen to the same argument in the absence of a prototype. This
3617 includes conversions of fixed point to floating and vice versa, and
3618 conversions changing the width or signedness of a fixed point argument
3619 except when the same as the default promotion.
3621 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3622 @opindex Wdeclaration-after-statement
3623 @opindex Wno-declaration-after-statement
3624 Warn when a declaration is found after a statement in a block. This
3625 construct, known from C++, was introduced with ISO C99 and is by default
3626 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3627 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3632 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3634 @item -Wno-endif-labels
3635 @opindex Wno-endif-labels
3636 @opindex Wendif-labels
3637 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3642 Warn whenever a local variable shadows another local variable, parameter or
3643 global variable or whenever a built-in function is shadowed.
3645 @item -Wlarger-than=@var{len}
3646 @opindex Wlarger-than=@var{len}
3647 @opindex Wlarger-than-@var{len}
3648 Warn whenever an object of larger than @var{len} bytes is defined.
3650 @item -Wframe-larger-than=@var{len}
3651 @opindex Wframe-larger-than
3652 Warn if the size of a function frame is larger than @var{len} bytes.
3653 The computation done to determine the stack frame size is approximate
3654 and not conservative.
3655 The actual requirements may be somewhat greater than @var{len}
3656 even if you do not get a warning. In addition, any space allocated
3657 via @code{alloca}, variable-length arrays, or related constructs
3658 is not included by the compiler when determining
3659 whether or not to issue a warning.
3661 @item -Wunsafe-loop-optimizations
3662 @opindex Wunsafe-loop-optimizations
3663 @opindex Wno-unsafe-loop-optimizations
3664 Warn if the loop cannot be optimized because the compiler could not
3665 assume anything on the bounds of the loop indices. With
3666 @option{-funsafe-loop-optimizations} warn if the compiler made
3669 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3670 @opindex Wno-pedantic-ms-format
3671 @opindex Wpedantic-ms-format
3672 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3673 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3674 depending on the MS runtime, when you are using the options @option{-Wformat}
3675 and @option{-pedantic} without gnu-extensions.
3677 @item -Wpointer-arith
3678 @opindex Wpointer-arith
3679 @opindex Wno-pointer-arith
3680 Warn about anything that depends on the ``size of'' a function type or
3681 of @code{void}. GNU C assigns these types a size of 1, for
3682 convenience in calculations with @code{void *} pointers and pointers
3683 to functions. In C++, warn also when an arithmetic operation involves
3684 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3687 @opindex Wtype-limits
3688 @opindex Wno-type-limits
3689 Warn if a comparison is always true or always false due to the limited
3690 range of the data type, but do not warn for constant expressions. For
3691 example, warn if an unsigned variable is compared against zero with
3692 @samp{<} or @samp{>=}. This warning is also enabled by
3695 @item -Wbad-function-cast @r{(C and Objective-C only)}
3696 @opindex Wbad-function-cast
3697 @opindex Wno-bad-function-cast
3698 Warn whenever a function call is cast to a non-matching type.
3699 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3701 @item -Wc++-compat @r{(C and Objective-C only)}
3702 Warn about ISO C constructs that are outside of the common subset of
3703 ISO C and ISO C++, e.g.@: request for implicit conversion from
3704 @code{void *} to a pointer to non-@code{void} type.
3706 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3707 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3708 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3709 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3713 @opindex Wno-cast-qual
3714 Warn whenever a pointer is cast so as to remove a type qualifier from
3715 the target type. For example, warn if a @code{const char *} is cast
3716 to an ordinary @code{char *}.
3718 Also warn when making a cast which introduces a type qualifier in an
3719 unsafe way. For example, casting @code{char **} to @code{const char **}
3720 is unsafe, as in this example:
3723 /* p is char ** value. */
3724 const char **q = (const char **) p;
3725 /* Assignment of readonly string to const char * is OK. */
3727 /* Now char** pointer points to read-only memory. */
3732 @opindex Wcast-align
3733 @opindex Wno-cast-align
3734 Warn whenever a pointer is cast such that the required alignment of the
3735 target is increased. For example, warn if a @code{char *} is cast to
3736 an @code{int *} on machines where integers can only be accessed at
3737 two- or four-byte boundaries.
3739 @item -Wwrite-strings
3740 @opindex Wwrite-strings
3741 @opindex Wno-write-strings
3742 When compiling C, give string constants the type @code{const
3743 char[@var{length}]} so that copying the address of one into a
3744 non-@code{const} @code{char *} pointer will get a warning. These
3745 warnings will help you find at compile time code that can try to write
3746 into a string constant, but only if you have been very careful about
3747 using @code{const} in declarations and prototypes. Otherwise, it will
3748 just be a nuisance. This is why we did not make @option{-Wall} request
3751 When compiling C++, warn about the deprecated conversion from string
3752 literals to @code{char *}. This warning is enabled by default for C++
3757 @opindex Wno-clobbered
3758 Warn for variables that might be changed by @samp{longjmp} or
3759 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3762 @opindex Wconversion
3763 @opindex Wno-conversion
3764 Warn for implicit conversions that may alter a value. This includes
3765 conversions between real and integer, like @code{abs (x)} when
3766 @code{x} is @code{double}; conversions between signed and unsigned,
3767 like @code{unsigned ui = -1}; and conversions to smaller types, like
3768 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3769 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3770 changed by the conversion like in @code{abs (2.0)}. Warnings about
3771 conversions between signed and unsigned integers can be disabled by
3772 using @option{-Wno-sign-conversion}.
3774 For C++, also warn for conversions between @code{NULL} and non-pointer
3775 types; confusing overload resolution for user-defined conversions; and
3776 conversions that will never use a type conversion operator:
3777 conversions to @code{void}, the same type, a base class or a reference
3778 to them. Warnings about conversions between signed and unsigned
3779 integers are disabled by default in C++ unless
3780 @option{-Wsign-conversion} is explicitly enabled.
3783 @opindex Wempty-body
3784 @opindex Wno-empty-body
3785 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3786 while} statement. This warning is also enabled by @option{-Wextra}.
3788 @item -Wenum-compare
3789 @opindex Wenum-compare
3790 @opindex Wno-enum-compare
3791 Warn about a comparison between values of different enum types. In C++
3792 this warning is enabled by default. In C this warning is enabled by
3795 @item -Wsign-compare
3796 @opindex Wsign-compare
3797 @opindex Wno-sign-compare
3798 @cindex warning for comparison of signed and unsigned values
3799 @cindex comparison of signed and unsigned values, warning
3800 @cindex signed and unsigned values, comparison warning
3801 Warn when a comparison between signed and unsigned values could produce
3802 an incorrect result when the signed value is converted to unsigned.
3803 This warning is also enabled by @option{-Wextra}; to get the other warnings
3804 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3806 @item -Wsign-conversion
3807 @opindex Wsign-conversion
3808 @opindex Wno-sign-conversion
3809 Warn for implicit conversions that may change the sign of an integer
3810 value, like assigning a signed integer expression to an unsigned
3811 integer variable. An explicit cast silences the warning. In C, this
3812 option is enabled also by @option{-Wconversion}.
3816 @opindex Wno-address
3817 Warn about suspicious uses of memory addresses. These include using
3818 the address of a function in a conditional expression, such as
3819 @code{void func(void); if (func)}, and comparisons against the memory
3820 address of a string literal, such as @code{if (x == "abc")}. Such
3821 uses typically indicate a programmer error: the address of a function
3822 always evaluates to true, so their use in a conditional usually
3823 indicate that the programmer forgot the parentheses in a function
3824 call; and comparisons against string literals result in unspecified
3825 behavior and are not portable in C, so they usually indicate that the
3826 programmer intended to use @code{strcmp}. This warning is enabled by
3830 @opindex Wlogical-op
3831 @opindex Wno-logical-op
3832 Warn about suspicious uses of logical operators in expressions.
3833 This includes using logical operators in contexts where a
3834 bit-wise operator is likely to be expected.
3836 @item -Waggregate-return
3837 @opindex Waggregate-return
3838 @opindex Wno-aggregate-return
3839 Warn if any functions that return structures or unions are defined or
3840 called. (In languages where you can return an array, this also elicits
3843 @item -Wno-attributes
3844 @opindex Wno-attributes
3845 @opindex Wattributes
3846 Do not warn if an unexpected @code{__attribute__} is used, such as
3847 unrecognized attributes, function attributes applied to variables,
3848 etc. This will not stop errors for incorrect use of supported
3851 @item -Wno-builtin-macro-redefined
3852 @opindex Wno-builtin-macro-redefined
3853 @opindex Wbuiltin-macro-redefined
3854 Do not warn if certain built-in macros are redefined. This suppresses
3855 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3856 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3858 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3859 @opindex Wstrict-prototypes
3860 @opindex Wno-strict-prototypes
3861 Warn if a function is declared or defined without specifying the
3862 argument types. (An old-style function definition is permitted without
3863 a warning if preceded by a declaration which specifies the argument
3866 @item -Wold-style-declaration @r{(C and Objective-C only)}
3867 @opindex Wold-style-declaration
3868 @opindex Wno-old-style-declaration
3869 Warn for obsolescent usages, according to the C Standard, in a
3870 declaration. For example, warn if storage-class specifiers like
3871 @code{static} are not the first things in a declaration. This warning
3872 is also enabled by @option{-Wextra}.
3874 @item -Wold-style-definition @r{(C and Objective-C only)}
3875 @opindex Wold-style-definition
3876 @opindex Wno-old-style-definition
3877 Warn if an old-style function definition is used. A warning is given
3878 even if there is a previous prototype.
3880 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3881 @opindex Wmissing-parameter-type
3882 @opindex Wno-missing-parameter-type
3883 A function parameter is declared without a type specifier in K&R-style
3890 This warning is also enabled by @option{-Wextra}.
3892 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3893 @opindex Wmissing-prototypes
3894 @opindex Wno-missing-prototypes
3895 Warn if a global function is defined without a previous prototype
3896 declaration. This warning is issued even if the definition itself
3897 provides a prototype. The aim is to detect global functions that fail
3898 to be declared in header files.
3900 @item -Wmissing-declarations
3901 @opindex Wmissing-declarations
3902 @opindex Wno-missing-declarations
3903 Warn if a global function is defined without a previous declaration.
3904 Do so even if the definition itself provides a prototype.
3905 Use this option to detect global functions that are not declared in
3906 header files. In C++, no warnings are issued for function templates,
3907 or for inline functions, or for functions in anonymous namespaces.
3909 @item -Wmissing-field-initializers
3910 @opindex Wmissing-field-initializers
3911 @opindex Wno-missing-field-initializers
3915 Warn if a structure's initializer has some fields missing. For
3916 example, the following code would cause such a warning, because
3917 @code{x.h} is implicitly zero:
3920 struct s @{ int f, g, h; @};
3921 struct s x = @{ 3, 4 @};
3924 This option does not warn about designated initializers, so the following
3925 modification would not trigger a warning:
3928 struct s @{ int f, g, h; @};
3929 struct s x = @{ .f = 3, .g = 4 @};
3932 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3933 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3935 @item -Wmissing-noreturn
3936 @opindex Wmissing-noreturn
3937 @opindex Wno-missing-noreturn
3938 Warn about functions which might be candidates for attribute @code{noreturn}.
3939 Note these are only possible candidates, not absolute ones. Care should
3940 be taken to manually verify functions actually do not ever return before
3941 adding the @code{noreturn} attribute, otherwise subtle code generation
3942 bugs could be introduced. You will not get a warning for @code{main} in
3943 hosted C environments.
3945 @item -Wmissing-format-attribute
3946 @opindex Wmissing-format-attribute
3947 @opindex Wno-missing-format-attribute
3950 Warn about function pointers which might be candidates for @code{format}
3951 attributes. Note these are only possible candidates, not absolute ones.
3952 GCC will guess that function pointers with @code{format} attributes that
3953 are used in assignment, initialization, parameter passing or return
3954 statements should have a corresponding @code{format} attribute in the
3955 resulting type. I.e.@: the left-hand side of the assignment or
3956 initialization, the type of the parameter variable, or the return type
3957 of the containing function respectively should also have a @code{format}
3958 attribute to avoid the warning.
3960 GCC will also warn about function definitions which might be
3961 candidates for @code{format} attributes. Again, these are only
3962 possible candidates. GCC will guess that @code{format} attributes
3963 might be appropriate for any function that calls a function like
3964 @code{vprintf} or @code{vscanf}, but this might not always be the
3965 case, and some functions for which @code{format} attributes are
3966 appropriate may not be detected.
3968 @item -Wno-multichar
3969 @opindex Wno-multichar
3971 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3972 Usually they indicate a typo in the user's code, as they have
3973 implementation-defined values, and should not be used in portable code.
3975 @item -Wnormalized=<none|id|nfc|nfkc>
3976 @opindex Wnormalized=
3979 @cindex character set, input normalization
3980 In ISO C and ISO C++, two identifiers are different if they are
3981 different sequences of characters. However, sometimes when characters
3982 outside the basic ASCII character set are used, you can have two
3983 different character sequences that look the same. To avoid confusion,
3984 the ISO 10646 standard sets out some @dfn{normalization rules} which
3985 when applied ensure that two sequences that look the same are turned into
3986 the same sequence. GCC can warn you if you are using identifiers which
3987 have not been normalized; this option controls that warning.
3989 There are four levels of warning that GCC supports. The default is
3990 @option{-Wnormalized=nfc}, which warns about any identifier which is
3991 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3992 recommended form for most uses.
3994 Unfortunately, there are some characters which ISO C and ISO C++ allow
3995 in identifiers that when turned into NFC aren't allowable as
3996 identifiers. That is, there's no way to use these symbols in portable
3997 ISO C or C++ and have all your identifiers in NFC@.
3998 @option{-Wnormalized=id} suppresses the warning for these characters.
3999 It is hoped that future versions of the standards involved will correct
4000 this, which is why this option is not the default.
4002 You can switch the warning off for all characters by writing
4003 @option{-Wnormalized=none}. You would only want to do this if you
4004 were using some other normalization scheme (like ``D''), because
4005 otherwise you can easily create bugs that are literally impossible to see.
4007 Some characters in ISO 10646 have distinct meanings but look identical
4008 in some fonts or display methodologies, especially once formatting has
4009 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4010 LETTER N'', will display just like a regular @code{n} which has been
4011 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4012 normalization scheme to convert all these into a standard form as
4013 well, and GCC will warn if your code is not in NFKC if you use
4014 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4015 about every identifier that contains the letter O because it might be
4016 confused with the digit 0, and so is not the default, but may be
4017 useful as a local coding convention if the programming environment is
4018 unable to be fixed to display these characters distinctly.
4020 @item -Wno-deprecated
4021 @opindex Wno-deprecated
4022 @opindex Wdeprecated
4023 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4025 @item -Wno-deprecated-declarations
4026 @opindex Wno-deprecated-declarations
4027 @opindex Wdeprecated-declarations
4028 Do not warn about uses of functions (@pxref{Function Attributes}),
4029 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4030 Attributes}) marked as deprecated by using the @code{deprecated}
4034 @opindex Wno-overflow
4036 Do not warn about compile-time overflow in constant expressions.
4038 @item -Woverride-init @r{(C and Objective-C only)}
4039 @opindex Woverride-init
4040 @opindex Wno-override-init
4044 Warn if an initialized field without side effects is overridden when
4045 using designated initializers (@pxref{Designated Inits, , Designated
4048 This warning is included in @option{-Wextra}. To get other
4049 @option{-Wextra} warnings without this one, use @samp{-Wextra
4050 -Wno-override-init}.
4055 Warn if a structure is given the packed attribute, but the packed
4056 attribute has no effect on the layout or size of the structure.
4057 Such structures may be mis-aligned for little benefit. For
4058 instance, in this code, the variable @code{f.x} in @code{struct bar}
4059 will be misaligned even though @code{struct bar} does not itself
4060 have the packed attribute:
4067 @} __attribute__((packed));
4075 @item -Wpacked-bitfield-compat
4076 @opindex Wpacked-bitfield-compat
4077 @opindex Wno-packed-bitfield-compat
4078 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4079 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4080 the change can lead to differences in the structure layout. GCC
4081 informs you when the offset of such a field has changed in GCC 4.4.
4082 For example there is no longer a 4-bit padding between field @code{a}
4083 and @code{b} in this structure:
4090 @} __attribute__ ((packed));
4093 This warning is enabled by default. Use
4094 @option{-Wno-packed-bitfield-compat} to disable this warning.
4099 Warn if padding is included in a structure, either to align an element
4100 of the structure or to align the whole structure. Sometimes when this
4101 happens it is possible to rearrange the fields of the structure to
4102 reduce the padding and so make the structure smaller.
4104 @item -Wredundant-decls
4105 @opindex Wredundant-decls
4106 @opindex Wno-redundant-decls
4107 Warn if anything is declared more than once in the same scope, even in
4108 cases where multiple declaration is valid and changes nothing.
4110 @item -Wnested-externs @r{(C and Objective-C only)}
4111 @opindex Wnested-externs
4112 @opindex Wno-nested-externs
4113 Warn if an @code{extern} declaration is encountered within a function.
4115 @item -Wunreachable-code
4116 @opindex Wunreachable-code
4117 @opindex Wno-unreachable-code
4118 Warn if the compiler detects that code will never be executed.
4120 This option is intended to warn when the compiler detects that at
4121 least a whole line of source code will never be executed, because
4122 some condition is never satisfied or because it is after a
4123 procedure that never returns.
4125 It is possible for this option to produce a warning even though there
4126 are circumstances under which part of the affected line can be executed,
4127 so care should be taken when removing apparently-unreachable code.
4129 For instance, when a function is inlined, a warning may mean that the
4130 line is unreachable in only one inlined copy of the function.
4132 This option is not made part of @option{-Wall} because in a debugging
4133 version of a program there is often substantial code which checks
4134 correct functioning of the program and is, hopefully, unreachable
4135 because the program does work. Another common use of unreachable
4136 code is to provide behavior which is selectable at compile-time.
4141 Warn if a function can not be inlined and it was declared as inline.
4142 Even with this option, the compiler will not warn about failures to
4143 inline functions declared in system headers.
4145 The compiler uses a variety of heuristics to determine whether or not
4146 to inline a function. For example, the compiler takes into account
4147 the size of the function being inlined and the amount of inlining
4148 that has already been done in the current function. Therefore,
4149 seemingly insignificant changes in the source program can cause the
4150 warnings produced by @option{-Winline} to appear or disappear.
4152 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4153 @opindex Wno-invalid-offsetof
4154 @opindex Winvalid-offsetof
4155 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4156 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4157 to a non-POD type is undefined. In existing C++ implementations,
4158 however, @samp{offsetof} typically gives meaningful results even when
4159 applied to certain kinds of non-POD types. (Such as a simple
4160 @samp{struct} that fails to be a POD type only by virtue of having a
4161 constructor.) This flag is for users who are aware that they are
4162 writing nonportable code and who have deliberately chosen to ignore the
4165 The restrictions on @samp{offsetof} may be relaxed in a future version
4166 of the C++ standard.
4168 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4169 @opindex Wno-int-to-pointer-cast
4170 @opindex Wint-to-pointer-cast
4171 Suppress warnings from casts to pointer type of an integer of a
4174 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4175 @opindex Wno-pointer-to-int-cast
4176 @opindex Wpointer-to-int-cast
4177 Suppress warnings from casts from a pointer to an integer type of a
4181 @opindex Winvalid-pch
4182 @opindex Wno-invalid-pch
4183 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4184 the search path but can't be used.
4188 @opindex Wno-long-long
4189 Warn if @samp{long long} type is used. This is enabled by either
4190 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4191 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4193 @item -Wvariadic-macros
4194 @opindex Wvariadic-macros
4195 @opindex Wno-variadic-macros
4196 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4197 alternate syntax when in pedantic ISO C99 mode. This is default.
4198 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4203 Warn if variable length array is used in the code.
4204 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4205 the variable length array.
4207 @item -Wvolatile-register-var
4208 @opindex Wvolatile-register-var
4209 @opindex Wno-volatile-register-var
4210 Warn if a register variable is declared volatile. The volatile
4211 modifier does not inhibit all optimizations that may eliminate reads
4212 and/or writes to register variables. This warning is enabled by
4215 @item -Wdisabled-optimization
4216 @opindex Wdisabled-optimization
4217 @opindex Wno-disabled-optimization
4218 Warn if a requested optimization pass is disabled. This warning does
4219 not generally indicate that there is anything wrong with your code; it
4220 merely indicates that GCC's optimizers were unable to handle the code
4221 effectively. Often, the problem is that your code is too big or too
4222 complex; GCC will refuse to optimize programs when the optimization
4223 itself is likely to take inordinate amounts of time.
4225 @item -Wpointer-sign @r{(C and Objective-C only)}
4226 @opindex Wpointer-sign
4227 @opindex Wno-pointer-sign
4228 Warn for pointer argument passing or assignment with different signedness.
4229 This option is only supported for C and Objective-C@. It is implied by
4230 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4231 @option{-Wno-pointer-sign}.
4233 @item -Wstack-protector
4234 @opindex Wstack-protector
4235 @opindex Wno-stack-protector
4236 This option is only active when @option{-fstack-protector} is active. It
4237 warns about functions that will not be protected against stack smashing.
4240 @opindex Wno-mudflap
4241 Suppress warnings about constructs that cannot be instrumented by
4244 @item -Woverlength-strings
4245 @opindex Woverlength-strings
4246 @opindex Wno-overlength-strings
4247 Warn about string constants which are longer than the ``minimum
4248 maximum'' length specified in the C standard. Modern compilers
4249 generally allow string constants which are much longer than the
4250 standard's minimum limit, but very portable programs should avoid
4251 using longer strings.
4253 The limit applies @emph{after} string constant concatenation, and does
4254 not count the trailing NUL@. In C89, the limit was 509 characters; in
4255 C99, it was raised to 4095. C++98 does not specify a normative
4256 minimum maximum, so we do not diagnose overlength strings in C++@.
4258 This option is implied by @option{-pedantic}, and can be disabled with
4259 @option{-Wno-overlength-strings}.
4261 @item -Wunsuffixed-float-constants
4262 @opindex Wunsuffixed-float-constants
4264 GCC will issue a warning for any floating constant that does not have
4265 a suffix. When used together with @option{-Wsystem-headers} it will
4266 warn about such constants in system header files. This can be useful
4267 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4268 from the decimal floating-point extension to C99.
4271 @node Debugging Options
4272 @section Options for Debugging Your Program or GCC
4273 @cindex options, debugging
4274 @cindex debugging information options
4276 GCC has various special options that are used for debugging
4277 either your program or GCC:
4282 Produce debugging information in the operating system's native format
4283 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4286 On most systems that use stabs format, @option{-g} enables use of extra
4287 debugging information that only GDB can use; this extra information
4288 makes debugging work better in GDB but will probably make other debuggers
4290 refuse to read the program. If you want to control for certain whether
4291 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4292 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4294 GCC allows you to use @option{-g} with
4295 @option{-O}. The shortcuts taken by optimized code may occasionally
4296 produce surprising results: some variables you declared may not exist
4297 at all; flow of control may briefly move where you did not expect it;
4298 some statements may not be executed because they compute constant
4299 results or their values were already at hand; some statements may
4300 execute in different places because they were moved out of loops.
4302 Nevertheless it proves possible to debug optimized output. This makes
4303 it reasonable to use the optimizer for programs that might have bugs.
4305 The following options are useful when GCC is generated with the
4306 capability for more than one debugging format.
4310 Produce debugging information for use by GDB@. This means to use the
4311 most expressive format available (DWARF 2, stabs, or the native format
4312 if neither of those are supported), including GDB extensions if at all
4317 Produce debugging information in stabs format (if that is supported),
4318 without GDB extensions. This is the format used by DBX on most BSD
4319 systems. On MIPS, Alpha and System V Release 4 systems this option
4320 produces stabs debugging output which is not understood by DBX or SDB@.
4321 On System V Release 4 systems this option requires the GNU assembler.
4323 @item -feliminate-unused-debug-symbols
4324 @opindex feliminate-unused-debug-symbols
4325 Produce debugging information in stabs format (if that is supported),
4326 for only symbols that are actually used.
4328 @item -femit-class-debug-always
4329 Instead of emitting debugging information for a C++ class in only one
4330 object file, emit it in all object files using the class. This option
4331 should be used only with debuggers that are unable to handle the way GCC
4332 normally emits debugging information for classes because using this
4333 option will increase the size of debugging information by as much as a
4338 Produce debugging information in stabs format (if that is supported),
4339 using GNU extensions understood only by the GNU debugger (GDB)@. The
4340 use of these extensions is likely to make other debuggers crash or
4341 refuse to read the program.
4345 Produce debugging information in COFF format (if that is supported).
4346 This is the format used by SDB on most System V systems prior to
4351 Produce debugging information in XCOFF format (if that is supported).
4352 This is the format used by the DBX debugger on IBM RS/6000 systems.
4356 Produce debugging information in XCOFF format (if that is supported),
4357 using GNU extensions understood only by the GNU debugger (GDB)@. The
4358 use of these extensions is likely to make other debuggers crash or
4359 refuse to read the program, and may cause assemblers other than the GNU
4360 assembler (GAS) to fail with an error.
4362 @item -gdwarf-@var{version}
4363 @opindex gdwarf-@var{version}
4364 Produce debugging information in DWARF format (if that is
4365 supported). This is the format used by DBX on IRIX 6. The value
4366 of @var{version} may be either 2 or 3; the default version is 2.
4368 Note that with DWARF version 2 some ports require, and will always
4369 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4373 Produce debugging information in VMS debug format (if that is
4374 supported). This is the format used by DEBUG on VMS systems.
4377 @itemx -ggdb@var{level}
4378 @itemx -gstabs@var{level}
4379 @itemx -gcoff@var{level}
4380 @itemx -gxcoff@var{level}
4381 @itemx -gvms@var{level}
4382 Request debugging information and also use @var{level} to specify how
4383 much information. The default level is 2.
4385 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4388 Level 1 produces minimal information, enough for making backtraces in
4389 parts of the program that you don't plan to debug. This includes
4390 descriptions of functions and external variables, but no information
4391 about local variables and no line numbers.
4393 Level 3 includes extra information, such as all the macro definitions
4394 present in the program. Some debuggers support macro expansion when
4395 you use @option{-g3}.
4397 @option{-gdwarf-2} does not accept a concatenated debug level, because
4398 GCC used to support an option @option{-gdwarf} that meant to generate
4399 debug information in version 1 of the DWARF format (which is very
4400 different from version 2), and it would have been too confusing. That
4401 debug format is long obsolete, but the option cannot be changed now.
4402 Instead use an additional @option{-g@var{level}} option to change the
4403 debug level for DWARF.
4407 Turn off generation of debug info, if leaving out this option would have
4408 generated it, or turn it on at level 2 otherwise. The position of this
4409 argument in the command line does not matter, it takes effect after all
4410 other options are processed, and it does so only once, no matter how
4411 many times it is given. This is mainly intended to be used with
4412 @option{-fcompare-debug}.
4414 @item -fdump-final-insns=@var{file}
4415 @opindex fdump-final-insns=
4416 Dump the final internal representation (RTL) to @var{file}.
4418 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4419 @opindex fcompare-debug
4420 @opindex fno-compare-debug
4421 If no error occurs during compilation, run the compiler a second time,
4422 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4423 passed to the second compilation. Dump the final internal
4424 representation in both compilations, and print an error if they differ.
4426 If the equal sign is omitted, the default @option{-gtoggle} is used.
4428 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4429 and nonzero, implicitly enables @option{-fcompare-debug}. If
4430 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4431 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4434 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4435 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4436 of the final representation and the second compilation, preventing even
4437 @env{GCC_COMPARE_DEBUG} from taking effect.
4439 To verify full coverage during @option{-fcompare-debug} testing, set
4440 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4441 which GCC will reject as an invalid option in any actual compilation
4442 (rather than preprocessing, assembly or linking). To get just a
4443 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4444 not overridden} will do.
4446 @item -fcompare-debug-second
4447 @opindex fcompare-debug-second
4448 This option is implicitly passed to the compiler for the second
4449 compilation requested by @option{-fcompare-debug}, along with options to
4450 silence warnings, and omitting other options that would cause
4451 side-effect compiler outputs to files or to the standard output. Dump
4452 files and preserved temporary files are renamed so as to contain the
4453 @code{.gk} additional extension during the second compilation, to avoid
4454 overwriting those generated by the first.
4456 When this option is passed to the compiler driver, it causes the
4457 @emph{first} compilation to be skipped, which makes it useful for little
4458 other than debugging the compiler proper.
4460 @item -feliminate-dwarf2-dups
4461 @opindex feliminate-dwarf2-dups
4462 Compress DWARF2 debugging information by eliminating duplicated
4463 information about each symbol. This option only makes sense when
4464 generating DWARF2 debugging information with @option{-gdwarf-2}.
4466 @item -femit-struct-debug-baseonly
4467 Emit debug information for struct-like types
4468 only when the base name of the compilation source file
4469 matches the base name of file in which the struct was defined.
4471 This option substantially reduces the size of debugging information,
4472 but at significant potential loss in type information to the debugger.
4473 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4474 See @option{-femit-struct-debug-detailed} for more detailed control.
4476 This option works only with DWARF 2.
4478 @item -femit-struct-debug-reduced
4479 Emit debug information for struct-like types
4480 only when the base name of the compilation source file
4481 matches the base name of file in which the type was defined,
4482 unless the struct is a template or defined in a system header.
4484 This option significantly reduces the size of debugging information,
4485 with some potential loss in type information to the debugger.
4486 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4487 See @option{-femit-struct-debug-detailed} for more detailed control.
4489 This option works only with DWARF 2.
4491 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4492 Specify the struct-like types
4493 for which the compiler will generate debug information.
4494 The intent is to reduce duplicate struct debug information
4495 between different object files within the same program.
4497 This option is a detailed version of
4498 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4499 which will serve for most needs.
4501 A specification has the syntax
4502 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4504 The optional first word limits the specification to
4505 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4506 A struct type is used directly when it is the type of a variable, member.
4507 Indirect uses arise through pointers to structs.
4508 That is, when use of an incomplete struct would be legal, the use is indirect.
4510 @samp{struct one direct; struct two * indirect;}.
4512 The optional second word limits the specification to
4513 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4514 Generic structs are a bit complicated to explain.
4515 For C++, these are non-explicit specializations of template classes,
4516 or non-template classes within the above.
4517 Other programming languages have generics,
4518 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4520 The third word specifies the source files for those
4521 structs for which the compiler will emit debug information.
4522 The values @samp{none} and @samp{any} have the normal meaning.
4523 The value @samp{base} means that
4524 the base of name of the file in which the type declaration appears
4525 must match the base of the name of the main compilation file.
4526 In practice, this means that
4527 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4528 but types declared in other header will not.
4529 The value @samp{sys} means those types satisfying @samp{base}
4530 or declared in system or compiler headers.
4532 You may need to experiment to determine the best settings for your application.
4534 The default is @samp{-femit-struct-debug-detailed=all}.
4536 This option works only with DWARF 2.
4538 @item -fno-merge-debug-strings
4539 @opindex fmerge-debug-strings
4540 @opindex fno-merge-debug-strings
4541 Direct the linker to not merge together strings in the debugging
4542 information which are identical in different object files. Merging is
4543 not supported by all assemblers or linkers. Merging decreases the size
4544 of the debug information in the output file at the cost of increasing
4545 link processing time. Merging is enabled by default.
4547 @item -fdebug-prefix-map=@var{old}=@var{new}
4548 @opindex fdebug-prefix-map
4549 When compiling files in directory @file{@var{old}}, record debugging
4550 information describing them as in @file{@var{new}} instead.
4552 @item -fno-dwarf2-cfi-asm
4553 @opindex fdwarf2-cfi-asm
4554 @opindex fno-dwarf2-cfi-asm
4555 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4556 instead of using GAS @code{.cfi_*} directives.
4558 @cindex @command{prof}
4561 Generate extra code to write profile information suitable for the
4562 analysis program @command{prof}. You must use this option when compiling
4563 the source files you want data about, and you must also use it when
4566 @cindex @command{gprof}
4569 Generate extra code to write profile information suitable for the
4570 analysis program @command{gprof}. You must use this option when compiling
4571 the source files you want data about, and you must also use it when
4576 Makes the compiler print out each function name as it is compiled, and
4577 print some statistics about each pass when it finishes.
4580 @opindex ftime-report
4581 Makes the compiler print some statistics about the time consumed by each
4582 pass when it finishes.
4585 @opindex fmem-report
4586 Makes the compiler print some statistics about permanent memory
4587 allocation when it finishes.
4589 @item -fpre-ipa-mem-report
4590 @opindex fpre-ipa-mem-report
4591 @item -fpost-ipa-mem-report
4592 @opindex fpost-ipa-mem-report
4593 Makes the compiler print some statistics about permanent memory
4594 allocation before or after interprocedural optimization.
4596 @item -fprofile-arcs
4597 @opindex fprofile-arcs
4598 Add code so that program flow @dfn{arcs} are instrumented. During
4599 execution the program records how many times each branch and call is
4600 executed and how many times it is taken or returns. When the compiled
4601 program exits it saves this data to a file called
4602 @file{@var{auxname}.gcda} for each source file. The data may be used for
4603 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4604 test coverage analysis (@option{-ftest-coverage}). Each object file's
4605 @var{auxname} is generated from the name of the output file, if
4606 explicitly specified and it is not the final executable, otherwise it is
4607 the basename of the source file. In both cases any suffix is removed
4608 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4609 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4610 @xref{Cross-profiling}.
4612 @cindex @command{gcov}
4616 This option is used to compile and link code instrumented for coverage
4617 analysis. The option is a synonym for @option{-fprofile-arcs}
4618 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4619 linking). See the documentation for those options for more details.
4624 Compile the source files with @option{-fprofile-arcs} plus optimization
4625 and code generation options. For test coverage analysis, use the
4626 additional @option{-ftest-coverage} option. You do not need to profile
4627 every source file in a program.
4630 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4631 (the latter implies the former).
4634 Run the program on a representative workload to generate the arc profile
4635 information. This may be repeated any number of times. You can run
4636 concurrent instances of your program, and provided that the file system
4637 supports locking, the data files will be correctly updated. Also
4638 @code{fork} calls are detected and correctly handled (double counting
4642 For profile-directed optimizations, compile the source files again with
4643 the same optimization and code generation options plus
4644 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4645 Control Optimization}).
4648 For test coverage analysis, use @command{gcov} to produce human readable
4649 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4650 @command{gcov} documentation for further information.
4654 With @option{-fprofile-arcs}, for each function of your program GCC
4655 creates a program flow graph, then finds a spanning tree for the graph.
4656 Only arcs that are not on the spanning tree have to be instrumented: the
4657 compiler adds code to count the number of times that these arcs are
4658 executed. When an arc is the only exit or only entrance to a block, the
4659 instrumentation code can be added to the block; otherwise, a new basic
4660 block must be created to hold the instrumentation code.
4663 @item -ftest-coverage
4664 @opindex ftest-coverage
4665 Produce a notes file that the @command{gcov} code-coverage utility
4666 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4667 show program coverage. Each source file's note file is called
4668 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4669 above for a description of @var{auxname} and instructions on how to
4670 generate test coverage data. Coverage data will match the source files
4671 more closely, if you do not optimize.
4673 @item -fdbg-cnt-list
4674 @opindex fdbg-cnt-list
4675 Print the name and the counter upperbound for all debug counters.
4677 @item -fdbg-cnt=@var{counter-value-list}
4679 Set the internal debug counter upperbound. @var{counter-value-list}
4680 is a comma-separated list of @var{name}:@var{value} pairs
4681 which sets the upperbound of each debug counter @var{name} to @var{value}.
4682 All debug counters have the initial upperbound of @var{UINT_MAX},
4683 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4684 e.g. With -fdbg-cnt=dce:10,tail_call:0
4685 dbg_cnt(dce) will return true only for first 10 invocations
4686 and dbg_cnt(tail_call) will return false always.
4688 @item -d@var{letters}
4689 @itemx -fdump-rtl-@var{pass}
4691 Says to make debugging dumps during compilation at times specified by
4692 @var{letters}. This is used for debugging the RTL-based passes of the
4693 compiler. The file names for most of the dumps are made by appending
4694 a pass number and a word to the @var{dumpname}, and the files are
4695 created in the directory of the output file. @var{dumpname} is
4696 generated from the name of the output file, if explicitly specified
4697 and it is not an executable, otherwise it is the basename of the
4698 source file. These switches may have different effects when
4699 @option{-E} is used for preprocessing.
4701 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4702 @option{-d} option @var{letters}. Here are the possible
4703 letters for use in @var{pass} and @var{letters}, and their meanings:
4707 @item -fdump-rtl-alignments
4708 @opindex fdump-rtl-alignments
4709 Dump after branch alignments have been computed.
4711 @item -fdump-rtl-asmcons
4712 @opindex fdump-rtl-asmcons
4713 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4715 @item -fdump-rtl-auto_inc_dec
4716 @opindex fdump-rtl-auto_inc_dec
4717 Dump after auto-inc-dec discovery. This pass is only run on
4718 architectures that have auto inc or auto dec instructions.
4720 @item -fdump-rtl-barriers
4721 @opindex fdump-rtl-barriers
4722 Dump after cleaning up the barrier instructions.
4724 @item -fdump-rtl-bbpart
4725 @opindex fdump-rtl-bbpart
4726 Dump after partitioning hot and cold basic blocks.
4728 @item -fdump-rtl-bbro
4729 @opindex fdump-rtl-bbro
4730 Dump after block reordering.
4732 @item -fdump-rtl-btl1
4733 @itemx -fdump-rtl-btl2
4734 @opindex fdump-rtl-btl2
4735 @opindex fdump-rtl-btl2
4736 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4737 after the two branch
4738 target load optimization passes.
4740 @item -fdump-rtl-bypass
4741 @opindex fdump-rtl-bypass
4742 Dump after jump bypassing and control flow optimizations.
4744 @item -fdump-rtl-combine
4745 @opindex fdump-rtl-combine
4746 Dump after the RTL instruction combination pass.
4748 @item -fdump-rtl-compgotos
4749 @opindex fdump-rtl-compgotos
4750 Dump after duplicating the computed gotos.
4752 @item -fdump-rtl-ce1
4753 @itemx -fdump-rtl-ce2
4754 @itemx -fdump-rtl-ce3
4755 @opindex fdump-rtl-ce1
4756 @opindex fdump-rtl-ce2
4757 @opindex fdump-rtl-ce3
4758 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4759 @option{-fdump-rtl-ce3} enable dumping after the three
4760 if conversion passes.
4762 @itemx -fdump-rtl-cprop_hardreg
4763 @opindex fdump-rtl-cprop_hardreg
4764 Dump after hard register copy propagation.
4766 @itemx -fdump-rtl-csa
4767 @opindex fdump-rtl-csa
4768 Dump after combining stack adjustments.
4770 @item -fdump-rtl-cse1
4771 @itemx -fdump-rtl-cse2
4772 @opindex fdump-rtl-cse1
4773 @opindex fdump-rtl-cse2
4774 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4775 the two common sub-expression elimination passes.
4777 @itemx -fdump-rtl-dce
4778 @opindex fdump-rtl-dce
4779 Dump after the standalone dead code elimination passes.
4781 @itemx -fdump-rtl-dbr
4782 @opindex fdump-rtl-dbr
4783 Dump after delayed branch scheduling.
4785 @item -fdump-rtl-dce1
4786 @itemx -fdump-rtl-dce2
4787 @opindex fdump-rtl-dce1
4788 @opindex fdump-rtl-dce2
4789 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4790 the two dead store elimination passes.
4793 @opindex fdump-rtl-eh
4794 Dump after finalization of EH handling code.
4796 @item -fdump-rtl-eh_ranges
4797 @opindex fdump-rtl-eh_ranges
4798 Dump after conversion of EH handling range regions.
4800 @item -fdump-rtl-expand
4801 @opindex fdump-rtl-expand
4802 Dump after RTL generation.
4804 @item -fdump-rtl-fwprop1
4805 @itemx -fdump-rtl-fwprop2
4806 @opindex fdump-rtl-fwprop1
4807 @opindex fdump-rtl-fwprop2
4808 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4809 dumping after the two forward propagation passes.
4811 @item -fdump-rtl-gcse1
4812 @itemx -fdump-rtl-gcse2
4813 @opindex fdump-rtl-gcse1
4814 @opindex fdump-rtl-gcse2
4815 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4816 after global common subexpression elimination.
4818 @item -fdump-rtl-init-regs
4819 @opindex fdump-rtl-init-regs
4820 Dump after the initialization of the registers.
4822 @item -fdump-rtl-initvals
4823 @opindex fdump-rtl-initvals
4824 Dump after the computation of the initial value sets.
4826 @itemx -fdump-rtl-into_cfglayout
4827 @opindex fdump-rtl-into_cfglayout
4828 Dump after converting to cfglayout mode.
4830 @item -fdump-rtl-ira
4831 @opindex fdump-rtl-ira
4832 Dump after iterated register allocation.
4834 @item -fdump-rtl-jump
4835 @opindex fdump-rtl-jump
4836 Dump after the second jump optimization.
4838 @item -fdump-rtl-loop2
4839 @opindex fdump-rtl-loop2
4840 @option{-fdump-rtl-loop2} enables dumping after the rtl
4841 loop optimization passes.
4843 @item -fdump-rtl-mach
4844 @opindex fdump-rtl-mach
4845 Dump after performing the machine dependent reorganization pass, if that
4848 @item -fdump-rtl-mode_sw
4849 @opindex fdump-rtl-mode_sw
4850 Dump after removing redundant mode switches.
4852 @item -fdump-rtl-rnreg
4853 @opindex fdump-rtl-rnreg
4854 Dump after register renumbering.
4856 @itemx -fdump-rtl-outof_cfglayout
4857 @opindex fdump-rtl-outof_cfglayout
4858 Dump after converting from cfglayout mode.
4860 @item -fdump-rtl-peephole2
4861 @opindex fdump-rtl-peephole2
4862 Dump after the peephole pass.
4864 @item -fdump-rtl-postreload
4865 @opindex fdump-rtl-postreload
4866 Dump after post-reload optimizations.
4868 @itemx -fdump-rtl-pro_and_epilogue
4869 @opindex fdump-rtl-pro_and_epilogue
4870 Dump after generating the function pro and epilogues.
4872 @item -fdump-rtl-regmove
4873 @opindex fdump-rtl-regmove
4874 Dump after the register move pass.
4876 @item -fdump-rtl-sched1
4877 @itemx -fdump-rtl-sched2
4878 @opindex fdump-rtl-sched1
4879 @opindex fdump-rtl-sched2
4880 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4881 after the basic block scheduling passes.
4883 @item -fdump-rtl-see
4884 @opindex fdump-rtl-see
4885 Dump after sign extension elimination.
4887 @item -fdump-rtl-seqabstr
4888 @opindex fdump-rtl-seqabstr
4889 Dump after common sequence discovery.
4891 @item -fdump-rtl-shorten
4892 @opindex fdump-rtl-shorten
4893 Dump after shortening branches.
4895 @item -fdump-rtl-sibling
4896 @opindex fdump-rtl-sibling
4897 Dump after sibling call optimizations.
4899 @item -fdump-rtl-split1
4900 @itemx -fdump-rtl-split2
4901 @itemx -fdump-rtl-split3
4902 @itemx -fdump-rtl-split4
4903 @itemx -fdump-rtl-split5
4904 @opindex fdump-rtl-split1
4905 @opindex fdump-rtl-split2
4906 @opindex fdump-rtl-split3
4907 @opindex fdump-rtl-split4
4908 @opindex fdump-rtl-split5
4909 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4910 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4911 @option{-fdump-rtl-split5} enable dumping after five rounds of
4912 instruction splitting.
4914 @item -fdump-rtl-sms
4915 @opindex fdump-rtl-sms
4916 Dump after modulo scheduling. This pass is only run on some
4919 @item -fdump-rtl-stack
4920 @opindex fdump-rtl-stack
4921 Dump after conversion from GCC's "flat register file" registers to the
4922 x87's stack-like registers. This pass is only run on x86 variants.
4924 @item -fdump-rtl-subreg1
4925 @itemx -fdump-rtl-subreg2
4926 @opindex fdump-rtl-subreg1
4927 @opindex fdump-rtl-subreg2
4928 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
4929 the two subreg expansion passes.
4931 @item -fdump-rtl-unshare
4932 @opindex fdump-rtl-unshare
4933 Dump after all rtl has been unshared.
4935 @item -fdump-rtl-vartrack
4936 @opindex fdump-rtl-vartrack
4937 Dump after variable tracking.
4939 @item -fdump-rtl-vregs
4940 @opindex fdump-rtl-vregs
4941 Dump after converting virtual registers to hard registers.
4943 @item -fdump-rtl-web
4944 @opindex fdump-rtl-web
4945 Dump after live range splitting.
4947 @item -fdump-rtl-regclass
4948 @itemx -fdump-rtl-subregs_of_mode_init
4949 @itemx -fdump-rtl-subregs_of_mode_finish
4950 @itemx -fdump-rtl-dfinit
4951 @itemx -fdump-rtl-dfinish
4952 @opindex fdump-rtl-regclass
4953 @opindex fdump-rtl-subregs_of_mode_init
4954 @opindex fdump-rtl-subregs_of_mode_finish
4955 @opindex fdump-rtl-dfinit
4956 @opindex fdump-rtl-dfinish
4957 These dumps are defined but always produce empty files.
4959 @item -fdump-rtl-all
4960 @opindex fdump-rtl-all
4961 Produce all the dumps listed above.
4965 Annotate the assembler output with miscellaneous debugging information.
4969 Dump all macro definitions, at the end of preprocessing, in addition to
4974 Produce a core dump whenever an error occurs.
4978 Print statistics on memory usage, at the end of the run, to
4983 Annotate the assembler output with a comment indicating which
4984 pattern and alternative was used. The length of each instruction is
4989 Dump the RTL in the assembler output as a comment before each instruction.
4990 Also turns on @option{-dp} annotation.
4994 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
4995 dump a representation of the control flow graph suitable for viewing with VCG
4996 to @file{@var{file}.@var{pass}.vcg}.
5000 Just generate RTL for a function instead of compiling it. Usually used
5001 with @option{-fdump-rtl-expand}.
5005 Dump debugging information during parsing, to standard error.
5009 @opindex fdump-noaddr
5010 When doing debugging dumps, suppress address output. This makes it more
5011 feasible to use diff on debugging dumps for compiler invocations with
5012 different compiler binaries and/or different
5013 text / bss / data / heap / stack / dso start locations.
5015 @item -fdump-unnumbered
5016 @opindex fdump-unnumbered
5017 When doing debugging dumps, suppress instruction numbers and address output.
5018 This makes it more feasible to use diff on debugging dumps for compiler
5019 invocations with different options, in particular with and without
5022 @item -fdump-unnumbered-links
5023 @opindex fdump-unnumbered-links
5024 When doing debugging dumps (see @option{-d} option above), suppress
5025 instruction numbers for the links to the previous and next instructions
5028 @item -fdump-translation-unit @r{(C++ only)}
5029 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5030 @opindex fdump-translation-unit
5031 Dump a representation of the tree structure for the entire translation
5032 unit to a file. The file name is made by appending @file{.tu} to the
5033 source file name, and the file is created in the same directory as the
5034 output file. If the @samp{-@var{options}} form is used, @var{options}
5035 controls the details of the dump as described for the
5036 @option{-fdump-tree} options.
5038 @item -fdump-class-hierarchy @r{(C++ only)}
5039 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5040 @opindex fdump-class-hierarchy
5041 Dump a representation of each class's hierarchy and virtual function
5042 table layout to a file. The file name is made by appending
5043 @file{.class} to the source file name, and the file is created in the
5044 same directory as the output file. If the @samp{-@var{options}} form
5045 is used, @var{options} controls the details of the dump as described
5046 for the @option{-fdump-tree} options.
5048 @item -fdump-ipa-@var{switch}
5050 Control the dumping at various stages of inter-procedural analysis
5051 language tree to a file. The file name is generated by appending a
5052 switch specific suffix to the source file name, and the file is created
5053 in the same directory as the output file. The following dumps are
5058 Enables all inter-procedural analysis dumps.
5061 Dumps information about call-graph optimization, unused function removal,
5062 and inlining decisions.
5065 Dump after function inlining.
5069 @item -fdump-statistics-@var{option}
5070 @opindex fdump-statistics
5071 Enable and control dumping of pass statistics in a separate file. The
5072 file name is generated by appending a suffix ending in
5073 @samp{.statistics} to the source file name, and the file is created in
5074 the same directory as the output file. If the @samp{-@var{option}}
5075 form is used, @samp{-stats} will cause counters to be summed over the
5076 whole compilation unit while @samp{-details} will dump every event as
5077 the passes generate them. The default with no option is to sum
5078 counters for each function compiled.
5080 @item -fdump-tree-@var{switch}
5081 @itemx -fdump-tree-@var{switch}-@var{options}
5083 Control the dumping at various stages of processing the intermediate
5084 language tree to a file. The file name is generated by appending a
5085 switch specific suffix to the source file name, and the file is
5086 created in the same directory as the output file. If the
5087 @samp{-@var{options}} form is used, @var{options} is a list of
5088 @samp{-} separated options that control the details of the dump. Not
5089 all options are applicable to all dumps, those which are not
5090 meaningful will be ignored. The following options are available
5094 Print the address of each node. Usually this is not meaningful as it
5095 changes according to the environment and source file. Its primary use
5096 is for tying up a dump file with a debug environment.
5098 Inhibit dumping of members of a scope or body of a function merely
5099 because that scope has been reached. Only dump such items when they
5100 are directly reachable by some other path. When dumping pretty-printed
5101 trees, this option inhibits dumping the bodies of control structures.
5103 Print a raw representation of the tree. By default, trees are
5104 pretty-printed into a C-like representation.
5106 Enable more detailed dumps (not honored by every dump option).
5108 Enable dumping various statistics about the pass (not honored by every dump
5111 Enable showing basic block boundaries (disabled in raw dumps).
5113 Enable showing virtual operands for every statement.
5115 Enable showing line numbers for statements.
5117 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5119 Enable showing the tree dump for each statement.
5121 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5122 and @option{lineno}.
5125 The following tree dumps are possible:
5129 @opindex fdump-tree-original
5130 Dump before any tree based optimization, to @file{@var{file}.original}.
5133 @opindex fdump-tree-optimized
5134 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5137 @opindex fdump-tree-gimple
5138 Dump each function before and after the gimplification pass to a file. The
5139 file name is made by appending @file{.gimple} to the source file name.
5142 @opindex fdump-tree-cfg
5143 Dump the control flow graph of each function to a file. The file name is
5144 made by appending @file{.cfg} to the source file name.
5147 @opindex fdump-tree-vcg
5148 Dump the control flow graph of each function to a file in VCG format. The
5149 file name is made by appending @file{.vcg} to the source file name. Note
5150 that if the file contains more than one function, the generated file cannot
5151 be used directly by VCG@. You will need to cut and paste each function's
5152 graph into its own separate file first.
5155 @opindex fdump-tree-ch
5156 Dump each function after copying loop headers. The file name is made by
5157 appending @file{.ch} to the source file name.
5160 @opindex fdump-tree-ssa
5161 Dump SSA related information to a file. The file name is made by appending
5162 @file{.ssa} to the source file name.
5165 @opindex fdump-tree-alias
5166 Dump aliasing information for each function. The file name is made by
5167 appending @file{.alias} to the source file name.
5170 @opindex fdump-tree-ccp
5171 Dump each function after CCP@. The file name is made by appending
5172 @file{.ccp} to the source file name.
5175 @opindex fdump-tree-storeccp
5176 Dump each function after STORE-CCP@. The file name is made by appending
5177 @file{.storeccp} to the source file name.
5180 @opindex fdump-tree-pre
5181 Dump trees after partial redundancy elimination. The file name is made
5182 by appending @file{.pre} to the source file name.
5185 @opindex fdump-tree-fre
5186 Dump trees after full redundancy elimination. The file name is made
5187 by appending @file{.fre} to the source file name.
5190 @opindex fdump-tree-copyprop
5191 Dump trees after copy propagation. The file name is made
5192 by appending @file{.copyprop} to the source file name.
5194 @item store_copyprop
5195 @opindex fdump-tree-store_copyprop
5196 Dump trees after store copy-propagation. The file name is made
5197 by appending @file{.store_copyprop} to the source file name.
5200 @opindex fdump-tree-dce
5201 Dump each function after dead code elimination. The file name is made by
5202 appending @file{.dce} to the source file name.
5205 @opindex fdump-tree-mudflap
5206 Dump each function after adding mudflap instrumentation. The file name is
5207 made by appending @file{.mudflap} to the source file name.
5210 @opindex fdump-tree-sra
5211 Dump each function after performing scalar replacement of aggregates. The
5212 file name is made by appending @file{.sra} to the source file name.
5215 @opindex fdump-tree-sink
5216 Dump each function after performing code sinking. The file name is made
5217 by appending @file{.sink} to the source file name.
5220 @opindex fdump-tree-dom
5221 Dump each function after applying dominator tree optimizations. The file
5222 name is made by appending @file{.dom} to the source file name.
5225 @opindex fdump-tree-dse
5226 Dump each function after applying dead store elimination. The file
5227 name is made by appending @file{.dse} to the source file name.
5230 @opindex fdump-tree-phiopt
5231 Dump each function after optimizing PHI nodes into straightline code. The file
5232 name is made by appending @file{.phiopt} to the source file name.
5235 @opindex fdump-tree-forwprop
5236 Dump each function after forward propagating single use variables. The file
5237 name is made by appending @file{.forwprop} to the source file name.
5240 @opindex fdump-tree-copyrename
5241 Dump each function after applying the copy rename optimization. The file
5242 name is made by appending @file{.copyrename} to the source file name.
5245 @opindex fdump-tree-nrv
5246 Dump each function after applying the named return value optimization on
5247 generic trees. The file name is made by appending @file{.nrv} to the source
5251 @opindex fdump-tree-vect
5252 Dump each function after applying vectorization of loops. The file name is
5253 made by appending @file{.vect} to the source file name.
5256 @opindex fdump-tree-vrp
5257 Dump each function after Value Range Propagation (VRP). The file name
5258 is made by appending @file{.vrp} to the source file name.
5261 @opindex fdump-tree-all
5262 Enable all the available tree dumps with the flags provided in this option.
5265 @item -ftree-vectorizer-verbose=@var{n}
5266 @opindex ftree-vectorizer-verbose
5267 This option controls the amount of debugging output the vectorizer prints.
5268 This information is written to standard error, unless
5269 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5270 in which case it is output to the usual dump listing file, @file{.vect}.
5271 For @var{n}=0 no diagnostic information is reported.
5272 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5273 and the total number of loops that got vectorized.
5274 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5275 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5276 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5277 level that @option{-fdump-tree-vect-stats} uses.
5278 Higher verbosity levels mean either more information dumped for each
5279 reported loop, or same amount of information reported for more loops:
5280 If @var{n}=3, alignment related information is added to the reports.
5281 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5282 memory access-patterns) is added to the reports.
5283 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5284 that did not pass the first analysis phase (i.e., may not be countable, or
5285 may have complicated control-flow).
5286 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5287 For @var{n}=7, all the information the vectorizer generates during its
5288 analysis and transformation is reported. This is the same verbosity level
5289 that @option{-fdump-tree-vect-details} uses.
5291 @item -frandom-seed=@var{string}
5292 @opindex frandom-seed
5293 This option provides a seed that GCC uses when it would otherwise use
5294 random numbers. It is used to generate certain symbol names
5295 that have to be different in every compiled file. It is also used to
5296 place unique stamps in coverage data files and the object files that
5297 produce them. You can use the @option{-frandom-seed} option to produce
5298 reproducibly identical object files.
5300 The @var{string} should be different for every file you compile.
5302 @item -fsched-verbose=@var{n}
5303 @opindex fsched-verbose
5304 On targets that use instruction scheduling, this option controls the
5305 amount of debugging output the scheduler prints. This information is
5306 written to standard error, unless @option{-fdump-rtl-sched1} or
5307 @option{-fdump-rtl-sched2} is specified, in which case it is output
5308 to the usual dump listing file, @file{.sched} or @file{.sched2}
5309 respectively. However for @var{n} greater than nine, the output is
5310 always printed to standard error.
5312 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5313 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5314 For @var{n} greater than one, it also output basic block probabilities,
5315 detailed ready list information and unit/insn info. For @var{n} greater
5316 than two, it includes RTL at abort point, control-flow and regions info.
5317 And for @var{n} over four, @option{-fsched-verbose} also includes
5321 @itemx -save-temps=cwd
5323 Store the usual ``temporary'' intermediate files permanently; place them
5324 in the current directory and name them based on the source file. Thus,
5325 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5326 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5327 preprocessed @file{foo.i} output file even though the compiler now
5328 normally uses an integrated preprocessor.
5330 When used in combination with the @option{-x} command line option,
5331 @option{-save-temps} is sensible enough to avoid over writing an
5332 input source file with the same extension as an intermediate file.
5333 The corresponding intermediate file may be obtained by renaming the
5334 source file before using @option{-save-temps}.
5336 If you invoke GCC in parallel, compiling several different source
5337 files that share a common base name in different subdirectories or the
5338 same source file compiled for multiple output destinations, it is
5339 likely that the different parallel compilers will interfere with each
5340 other, and overwrite the temporary files. For instance:
5343 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5344 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5347 may result in @file{foo.i} and @file{foo.o} being written to
5348 simultaneously by both compilers.
5350 @item -save-temps=obj
5351 @opindex save-temps=obj
5352 Store the usual ``temporary'' intermediate files permanently. If the
5353 @option{-o} option is used, the temporary files are based on the
5354 object file. If the @option{-o} option is not used, the
5355 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5360 gcc -save-temps=obj -c foo.c
5361 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5362 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5365 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5366 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5367 @file{dir2/yfoobar.o}.
5369 @item -time@r{[}=@var{file}@r{]}
5371 Report the CPU time taken by each subprocess in the compilation
5372 sequence. For C source files, this is the compiler proper and assembler
5373 (plus the linker if linking is done).
5375 Without the specification of an output file, the output looks like this:
5382 The first number on each line is the ``user time'', that is time spent
5383 executing the program itself. The second number is ``system time'',
5384 time spent executing operating system routines on behalf of the program.
5385 Both numbers are in seconds.
5387 With the specification of an output file, the output is appended to the
5388 named file, and it looks like this:
5391 0.12 0.01 cc1 @var{options}
5392 0.00 0.01 as @var{options}
5395 The ``user time'' and the ``system time'' are moved before the program
5396 name, and the options passed to the program are displayed, so that one
5397 can later tell what file was being compiled, and with which options.
5399 @item -fvar-tracking
5400 @opindex fvar-tracking
5401 Run variable tracking pass. It computes where variables are stored at each
5402 position in code. Better debugging information is then generated
5403 (if the debugging information format supports this information).
5405 It is enabled by default when compiling with optimization (@option{-Os},
5406 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5407 the debug info format supports it.
5409 @item -print-file-name=@var{library}
5410 @opindex print-file-name
5411 Print the full absolute name of the library file @var{library} that
5412 would be used when linking---and don't do anything else. With this
5413 option, GCC does not compile or link anything; it just prints the
5416 @item -print-multi-directory
5417 @opindex print-multi-directory
5418 Print the directory name corresponding to the multilib selected by any
5419 other switches present in the command line. This directory is supposed
5420 to exist in @env{GCC_EXEC_PREFIX}.
5422 @item -print-multi-lib
5423 @opindex print-multi-lib
5424 Print the mapping from multilib directory names to compiler switches
5425 that enable them. The directory name is separated from the switches by
5426 @samp{;}, and each switch starts with an @samp{@@} instead of the
5427 @samp{-}, without spaces between multiple switches. This is supposed to
5428 ease shell-processing.
5430 @item -print-prog-name=@var{program}
5431 @opindex print-prog-name
5432 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5434 @item -print-libgcc-file-name
5435 @opindex print-libgcc-file-name
5436 Same as @option{-print-file-name=libgcc.a}.
5438 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5439 but you do want to link with @file{libgcc.a}. You can do
5442 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5445 @item -print-search-dirs
5446 @opindex print-search-dirs
5447 Print the name of the configured installation directory and a list of
5448 program and library directories @command{gcc} will search---and don't do anything else.
5450 This is useful when @command{gcc} prints the error message
5451 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5452 To resolve this you either need to put @file{cpp0} and the other compiler
5453 components where @command{gcc} expects to find them, or you can set the environment
5454 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5455 Don't forget the trailing @samp{/}.
5456 @xref{Environment Variables}.
5458 @item -print-sysroot
5459 @opindex print-sysroot
5460 Print the target sysroot directory that will be used during
5461 compilation. This is the target sysroot specified either at configure
5462 time or using the @option{--sysroot} option, possibly with an extra
5463 suffix that depends on compilation options. If no target sysroot is
5464 specified, the option prints nothing.
5466 @item -print-sysroot-headers-suffix
5467 @opindex print-sysroot-headers-suffix
5468 Print the suffix added to the target sysroot when searching for
5469 headers, or give an error if the compiler is not configured with such
5470 a suffix---and don't do anything else.
5473 @opindex dumpmachine
5474 Print the compiler's target machine (for example,
5475 @samp{i686-pc-linux-gnu})---and don't do anything else.
5478 @opindex dumpversion
5479 Print the compiler version (for example, @samp{3.0})---and don't do
5484 Print the compiler's built-in specs---and don't do anything else. (This
5485 is used when GCC itself is being built.) @xref{Spec Files}.
5487 @item -feliminate-unused-debug-types
5488 @opindex feliminate-unused-debug-types
5489 Normally, when producing DWARF2 output, GCC will emit debugging
5490 information for all types declared in a compilation
5491 unit, regardless of whether or not they are actually used
5492 in that compilation unit. Sometimes this is useful, such as
5493 if, in the debugger, you want to cast a value to a type that is
5494 not actually used in your program (but is declared). More often,
5495 however, this results in a significant amount of wasted space.
5496 With this option, GCC will avoid producing debug symbol output
5497 for types that are nowhere used in the source file being compiled.
5500 @node Optimize Options
5501 @section Options That Control Optimization
5502 @cindex optimize options
5503 @cindex options, optimization
5505 These options control various sorts of optimizations.
5507 Without any optimization option, the compiler's goal is to reduce the
5508 cost of compilation and to make debugging produce the expected
5509 results. Statements are independent: if you stop the program with a
5510 breakpoint between statements, you can then assign a new value to any
5511 variable or change the program counter to any other statement in the
5512 function and get exactly the results you would expect from the source
5515 Turning on optimization flags makes the compiler attempt to improve
5516 the performance and/or code size at the expense of compilation time
5517 and possibly the ability to debug the program.
5519 The compiler performs optimization based on the knowledge it has of the
5520 program. Compiling multiple files at once to a single output file mode allows
5521 the compiler to use information gained from all of the files when compiling
5524 Not all optimizations are controlled directly by a flag. Only
5525 optimizations that have a flag are listed in this section.
5527 Depending on the target and how GCC was configured, a slightly different
5528 set of optimizations may be enabled at each @option{-O} level than
5529 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5530 to find out the exact set of optimizations that are enabled at each level.
5531 @xref{Overall Options}, for examples.
5538 Optimize. Optimizing compilation takes somewhat more time, and a lot
5539 more memory for a large function.
5541 With @option{-O}, the compiler tries to reduce code size and execution
5542 time, without performing any optimizations that take a great deal of
5545 @option{-O} turns on the following optimization flags:
5548 -fcprop-registers @gol
5551 -fdelayed-branch @gol
5553 -fguess-branch-probability @gol
5554 -fif-conversion2 @gol
5555 -fif-conversion @gol
5556 -finline-small-functions @gol
5557 -fipa-pure-const @gol
5558 -fipa-reference @gol
5560 -fsplit-wide-types @gol
5561 -ftree-builtin-call-dce @gol
5564 -ftree-copyrename @gol
5566 -ftree-dominator-opts @gol
5568 -ftree-forwprop @gol
5576 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5577 where doing so does not interfere with debugging.
5581 Optimize even more. GCC performs nearly all supported optimizations
5582 that do not involve a space-speed tradeoff.
5583 As compared to @option{-O}, this option increases both compilation time
5584 and the performance of the generated code.
5586 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5587 also turns on the following optimization flags:
5588 @gccoptlist{-fthread-jumps @gol
5589 -falign-functions -falign-jumps @gol
5590 -falign-loops -falign-labels @gol
5593 -fcse-follow-jumps -fcse-skip-blocks @gol
5594 -fdelete-null-pointer-checks @gol
5595 -fexpensive-optimizations @gol
5596 -fgcse -fgcse-lm @gol
5597 -findirect-inlining @gol
5598 -foptimize-sibling-calls @gol
5601 -freorder-blocks -freorder-functions @gol
5602 -frerun-cse-after-loop @gol
5603 -fsched-interblock -fsched-spec @gol
5604 -fschedule-insns -fschedule-insns2 @gol
5605 -fstrict-aliasing -fstrict-overflow @gol
5606 -ftree-switch-conversion @gol
5610 Please note the warning under @option{-fgcse} about
5611 invoking @option{-O2} on programs that use computed gotos.
5615 Optimize yet more. @option{-O3} turns on all optimizations specified
5616 by @option{-O2} and also turns on the @option{-finline-functions},
5617 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5618 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5622 Reduce compilation time and make debugging produce the expected
5623 results. This is the default.
5627 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5628 do not typically increase code size. It also performs further
5629 optimizations designed to reduce code size.
5631 @option{-Os} disables the following optimization flags:
5632 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5633 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5634 -fprefetch-loop-arrays -ftree-vect-loop-version}
5636 If you use multiple @option{-O} options, with or without level numbers,
5637 the last such option is the one that is effective.
5640 Options of the form @option{-f@var{flag}} specify machine-independent
5641 flags. Most flags have both positive and negative forms; the negative
5642 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5643 below, only one of the forms is listed---the one you typically will
5644 use. You can figure out the other form by either removing @samp{no-}
5647 The following options control specific optimizations. They are either
5648 activated by @option{-O} options or are related to ones that are. You
5649 can use the following flags in the rare cases when ``fine-tuning'' of
5650 optimizations to be performed is desired.
5653 @item -fno-default-inline
5654 @opindex fno-default-inline
5655 Do not make member functions inline by default merely because they are
5656 defined inside the class scope (C++ only). Otherwise, when you specify
5657 @w{@option{-O}}, member functions defined inside class scope are compiled
5658 inline by default; i.e., you don't need to add @samp{inline} in front of
5659 the member function name.
5661 @item -fno-defer-pop
5662 @opindex fno-defer-pop
5663 Always pop the arguments to each function call as soon as that function
5664 returns. For machines which must pop arguments after a function call,
5665 the compiler normally lets arguments accumulate on the stack for several
5666 function calls and pops them all at once.
5668 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5670 @item -fforward-propagate
5671 @opindex fforward-propagate
5672 Perform a forward propagation pass on RTL@. The pass tries to combine two
5673 instructions and checks if the result can be simplified. If loop unrolling
5674 is active, two passes are performed and the second is scheduled after
5677 This option is enabled by default at optimization levels @option{-O},
5678 @option{-O2}, @option{-O3}, @option{-Os}.
5680 @item -fomit-frame-pointer
5681 @opindex fomit-frame-pointer
5682 Don't keep the frame pointer in a register for functions that
5683 don't need one. This avoids the instructions to save, set up and
5684 restore frame pointers; it also makes an extra register available
5685 in many functions. @strong{It also makes debugging impossible on
5688 On some machines, such as the VAX, this flag has no effect, because
5689 the standard calling sequence automatically handles the frame pointer
5690 and nothing is saved by pretending it doesn't exist. The
5691 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5692 whether a target machine supports this flag. @xref{Registers,,Register
5693 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5695 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5697 @item -foptimize-sibling-calls
5698 @opindex foptimize-sibling-calls
5699 Optimize sibling and tail recursive calls.
5701 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5705 Don't pay attention to the @code{inline} keyword. Normally this option
5706 is used to keep the compiler from expanding any functions inline.
5707 Note that if you are not optimizing, no functions can be expanded inline.
5709 @item -finline-small-functions
5710 @opindex finline-small-functions
5711 Integrate functions into their callers when their body is smaller than expected
5712 function call code (so overall size of program gets smaller). The compiler
5713 heuristically decides which functions are simple enough to be worth integrating
5716 Enabled at level @option{-O2}.
5718 @item -findirect-inlining
5719 @opindex findirect-inlining
5720 Inline also indirect calls that are discovered to be known at compile
5721 time thanks to previous inlining. This option has any effect only
5722 when inlining itself is turned on by the @option{-finline-functions}
5723 or @option{-finline-small-functions} options.
5725 Enabled at level @option{-O2}.
5727 @item -finline-functions
5728 @opindex finline-functions
5729 Integrate all simple functions into their callers. The compiler
5730 heuristically decides which functions are simple enough to be worth
5731 integrating in this way.
5733 If all calls to a given function are integrated, and the function is
5734 declared @code{static}, then the function is normally not output as
5735 assembler code in its own right.
5737 Enabled at level @option{-O3}.
5739 @item -finline-functions-called-once
5740 @opindex finline-functions-called-once
5741 Consider all @code{static} functions called once for inlining into their
5742 caller even if they are not marked @code{inline}. If a call to a given
5743 function is integrated, then the function is not output as assembler code
5746 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5748 @item -fearly-inlining
5749 @opindex fearly-inlining
5750 Inline functions marked by @code{always_inline} and functions whose body seems
5751 smaller than the function call overhead early before doing
5752 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5753 makes profiling significantly cheaper and usually inlining faster on programs
5754 having large chains of nested wrapper functions.
5758 @item -finline-limit=@var{n}
5759 @opindex finline-limit
5760 By default, GCC limits the size of functions that can be inlined. This flag
5761 allows coarse control of this limit. @var{n} is the size of functions that
5762 can be inlined in number of pseudo instructions.
5764 Inlining is actually controlled by a number of parameters, which may be
5765 specified individually by using @option{--param @var{name}=@var{value}}.
5766 The @option{-finline-limit=@var{n}} option sets some of these parameters
5770 @item max-inline-insns-single
5771 is set to @var{n}/2.
5772 @item max-inline-insns-auto
5773 is set to @var{n}/2.
5776 See below for a documentation of the individual
5777 parameters controlling inlining and for the defaults of these parameters.
5779 @emph{Note:} there may be no value to @option{-finline-limit} that results
5780 in default behavior.
5782 @emph{Note:} pseudo instruction represents, in this particular context, an
5783 abstract measurement of function's size. In no way does it represent a count
5784 of assembly instructions and as such its exact meaning might change from one
5785 release to an another.
5787 @item -fkeep-inline-functions
5788 @opindex fkeep-inline-functions
5789 In C, emit @code{static} functions that are declared @code{inline}
5790 into the object file, even if the function has been inlined into all
5791 of its callers. This switch does not affect functions using the
5792 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5793 inline functions into the object file.
5795 @item -fkeep-static-consts
5796 @opindex fkeep-static-consts
5797 Emit variables declared @code{static const} when optimization isn't turned
5798 on, even if the variables aren't referenced.
5800 GCC enables this option by default. If you want to force the compiler to
5801 check if the variable was referenced, regardless of whether or not
5802 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5804 @item -fmerge-constants
5805 @opindex fmerge-constants
5806 Attempt to merge identical constants (string constants and floating point
5807 constants) across compilation units.
5809 This option is the default for optimized compilation if the assembler and
5810 linker support it. Use @option{-fno-merge-constants} to inhibit this
5813 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5815 @item -fmerge-all-constants
5816 @opindex fmerge-all-constants
5817 Attempt to merge identical constants and identical variables.
5819 This option implies @option{-fmerge-constants}. In addition to
5820 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5821 arrays or initialized constant variables with integral or floating point
5822 types. Languages like C or C++ require each variable, including multiple
5823 instances of the same variable in recursive calls, to have distinct locations,
5824 so using this option will result in non-conforming
5827 @item -fmodulo-sched
5828 @opindex fmodulo-sched
5829 Perform swing modulo scheduling immediately before the first scheduling
5830 pass. This pass looks at innermost loops and reorders their
5831 instructions by overlapping different iterations.
5833 @item -fmodulo-sched-allow-regmoves
5834 @opindex fmodulo-sched-allow-regmoves
5835 Perform more aggressive SMS based modulo scheduling with register moves
5836 allowed. By setting this flag certain anti-dependences edges will be
5837 deleted which will trigger the generation of reg-moves based on the
5838 life-range analysis. This option is effective only with
5839 @option{-fmodulo-sched} enabled.
5841 @item -fno-branch-count-reg
5842 @opindex fno-branch-count-reg
5843 Do not use ``decrement and branch'' instructions on a count register,
5844 but instead generate a sequence of instructions that decrement a
5845 register, compare it against zero, then branch based upon the result.
5846 This option is only meaningful on architectures that support such
5847 instructions, which include x86, PowerPC, IA-64 and S/390.
5849 The default is @option{-fbranch-count-reg}.
5851 @item -fno-function-cse
5852 @opindex fno-function-cse
5853 Do not put function addresses in registers; make each instruction that
5854 calls a constant function contain the function's address explicitly.
5856 This option results in less efficient code, but some strange hacks
5857 that alter the assembler output may be confused by the optimizations
5858 performed when this option is not used.
5860 The default is @option{-ffunction-cse}
5862 @item -fno-zero-initialized-in-bss
5863 @opindex fno-zero-initialized-in-bss
5864 If the target supports a BSS section, GCC by default puts variables that
5865 are initialized to zero into BSS@. This can save space in the resulting
5868 This option turns off this behavior because some programs explicitly
5869 rely on variables going to the data section. E.g., so that the
5870 resulting executable can find the beginning of that section and/or make
5871 assumptions based on that.
5873 The default is @option{-fzero-initialized-in-bss}.
5875 @item -fmudflap -fmudflapth -fmudflapir
5879 @cindex bounds checking
5881 For front-ends that support it (C and C++), instrument all risky
5882 pointer/array dereferencing operations, some standard library
5883 string/heap functions, and some other associated constructs with
5884 range/validity tests. Modules so instrumented should be immune to
5885 buffer overflows, invalid heap use, and some other classes of C/C++
5886 programming errors. The instrumentation relies on a separate runtime
5887 library (@file{libmudflap}), which will be linked into a program if
5888 @option{-fmudflap} is given at link time. Run-time behavior of the
5889 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5890 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5893 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5894 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5895 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5896 instrumentation should ignore pointer reads. This produces less
5897 instrumentation (and therefore faster execution) and still provides
5898 some protection against outright memory corrupting writes, but allows
5899 erroneously read data to propagate within a program.
5901 @item -fthread-jumps
5902 @opindex fthread-jumps
5903 Perform optimizations where we check to see if a jump branches to a
5904 location where another comparison subsumed by the first is found. If
5905 so, the first branch is redirected to either the destination of the
5906 second branch or a point immediately following it, depending on whether
5907 the condition is known to be true or false.
5909 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5911 @item -fsplit-wide-types
5912 @opindex fsplit-wide-types
5913 When using a type that occupies multiple registers, such as @code{long
5914 long} on a 32-bit system, split the registers apart and allocate them
5915 independently. This normally generates better code for those types,
5916 but may make debugging more difficult.
5918 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5921 @item -fcse-follow-jumps
5922 @opindex fcse-follow-jumps
5923 In common subexpression elimination (CSE), scan through jump instructions
5924 when the target of the jump is not reached by any other path. For
5925 example, when CSE encounters an @code{if} statement with an
5926 @code{else} clause, CSE will follow the jump when the condition
5929 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5931 @item -fcse-skip-blocks
5932 @opindex fcse-skip-blocks
5933 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5934 follow jumps which conditionally skip over blocks. When CSE
5935 encounters a simple @code{if} statement with no else clause,
5936 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5937 body of the @code{if}.
5939 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5941 @item -frerun-cse-after-loop
5942 @opindex frerun-cse-after-loop
5943 Re-run common subexpression elimination after loop optimizations has been
5946 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5950 Perform a global common subexpression elimination pass.
5951 This pass also performs global constant and copy propagation.
5953 @emph{Note:} When compiling a program using computed gotos, a GCC
5954 extension, you may get better runtime performance if you disable
5955 the global common subexpression elimination pass by adding
5956 @option{-fno-gcse} to the command line.
5958 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5962 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5963 attempt to move loads which are only killed by stores into themselves. This
5964 allows a loop containing a load/store sequence to be changed to a load outside
5965 the loop, and a copy/store within the loop.
5967 Enabled by default when gcse is enabled.
5971 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5972 global common subexpression elimination. This pass will attempt to move
5973 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5974 loops containing a load/store sequence can be changed to a load before
5975 the loop and a store after the loop.
5977 Not enabled at any optimization level.
5981 When @option{-fgcse-las} is enabled, the global common subexpression
5982 elimination pass eliminates redundant loads that come after stores to the
5983 same memory location (both partial and full redundancies).
5985 Not enabled at any optimization level.
5987 @item -fgcse-after-reload
5988 @opindex fgcse-after-reload
5989 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5990 pass is performed after reload. The purpose of this pass is to cleanup
5993 @item -funsafe-loop-optimizations
5994 @opindex funsafe-loop-optimizations
5995 If given, the loop optimizer will assume that loop indices do not
5996 overflow, and that the loops with nontrivial exit condition are not
5997 infinite. This enables a wider range of loop optimizations even if
5998 the loop optimizer itself cannot prove that these assumptions are valid.
5999 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6000 if it finds this kind of loop.
6002 @item -fcrossjumping
6003 @opindex fcrossjumping
6004 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6005 resulting code may or may not perform better than without cross-jumping.
6007 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6009 @item -fauto-inc-dec
6010 @opindex fauto-inc-dec
6011 Combine increments or decrements of addresses with memory accesses.
6012 This pass is always skipped on architectures that do not have
6013 instructions to support this. Enabled by default at @option{-O} and
6014 higher on architectures that support this.
6018 Perform dead code elimination (DCE) on RTL@.
6019 Enabled by default at @option{-O} and higher.
6023 Perform dead store elimination (DSE) on RTL@.
6024 Enabled by default at @option{-O} and higher.
6026 @item -fif-conversion
6027 @opindex fif-conversion
6028 Attempt to transform conditional jumps into branch-less equivalents. This
6029 include use of conditional moves, min, max, set flags and abs instructions, and
6030 some tricks doable by standard arithmetics. The use of conditional execution
6031 on chips where it is available is controlled by @code{if-conversion2}.
6033 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6035 @item -fif-conversion2
6036 @opindex fif-conversion2
6037 Use conditional execution (where available) to transform conditional jumps into
6038 branch-less equivalents.
6040 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6042 @item -fdelete-null-pointer-checks
6043 @opindex fdelete-null-pointer-checks
6044 Assume that programs cannot safely dereference null pointers, and that
6045 no code or data element resides there. This enables simple constant
6046 folding optimizations at all optimization levels. In addition, other
6047 optimization passes in GCC use this flag to control global dataflow
6048 analyses that eliminate useless checks for null pointers; these assume
6049 that if a pointer is checked after it has already been dereferenced,
6052 Note however that in some environments this assumption is not true.
6053 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6054 for programs which depend on that behavior.
6056 Some targets, especially embedded ones, disable this option at all levels.
6057 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6058 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6059 are enabled independently at different optimization levels.
6061 @item -fexpensive-optimizations
6062 @opindex fexpensive-optimizations
6063 Perform a number of minor optimizations that are relatively expensive.
6065 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6067 @item -foptimize-register-move
6069 @opindex foptimize-register-move
6071 Attempt to reassign register numbers in move instructions and as
6072 operands of other simple instructions in order to maximize the amount of
6073 register tying. This is especially helpful on machines with two-operand
6076 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6079 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6081 @item -fira-algorithm=@var{algorithm}
6082 Use specified coloring algorithm for the integrated register
6083 allocator. The @var{algorithm} argument should be @code{priority} or
6084 @code{CB}. The first algorithm specifies Chow's priority coloring,
6085 the second one specifies Chaitin-Briggs coloring. The second
6086 algorithm can be unimplemented for some architectures. If it is
6087 implemented, it is the default because Chaitin-Briggs coloring as a
6088 rule generates a better code.
6090 @item -fira-region=@var{region}
6091 Use specified regions for the integrated register allocator. The
6092 @var{region} argument should be one of @code{all}, @code{mixed}, or
6093 @code{one}. The first value means using all loops as register
6094 allocation regions, the second value which is the default means using
6095 all loops except for loops with small register pressure as the
6096 regions, and third one means using all function as a single region.
6097 The first value can give best result for machines with small size and
6098 irregular register set, the third one results in faster and generates
6099 decent code and the smallest size code, and the default value usually
6100 give the best results in most cases and for most architectures.
6102 @item -fira-coalesce
6103 @opindex fira-coalesce
6104 Do optimistic register coalescing. This option might be profitable for
6105 architectures with big regular register files.
6107 @item -fno-ira-share-save-slots
6108 @opindex fno-ira-share-save-slots
6109 Switch off sharing stack slots used for saving call used hard
6110 registers living through a call. Each hard register will get a
6111 separate stack slot and as a result function stack frame will be
6114 @item -fno-ira-share-spill-slots
6115 @opindex fno-ira-share-spill-slots
6116 Switch off sharing stack slots allocated for pseudo-registers. Each
6117 pseudo-register which did not get a hard register will get a separate
6118 stack slot and as a result function stack frame will be bigger.
6120 @item -fira-verbose=@var{n}
6121 @opindex fira-verbose
6122 Set up how verbose dump file for the integrated register allocator
6123 will be. Default value is 5. If the value is greater or equal to 10,
6124 the dump file will be stderr as if the value were @var{n} minus 10.
6126 @item -fdelayed-branch
6127 @opindex fdelayed-branch
6128 If supported for the target machine, attempt to reorder instructions
6129 to exploit instruction slots available after delayed branch
6132 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6134 @item -fschedule-insns
6135 @opindex fschedule-insns
6136 If supported for the target machine, attempt to reorder instructions to
6137 eliminate execution stalls due to required data being unavailable. This
6138 helps machines that have slow floating point or memory load instructions
6139 by allowing other instructions to be issued until the result of the load
6140 or floating point instruction is required.
6142 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6144 @item -fschedule-insns2
6145 @opindex fschedule-insns2
6146 Similar to @option{-fschedule-insns}, but requests an additional pass of
6147 instruction scheduling after register allocation has been done. This is
6148 especially useful on machines with a relatively small number of
6149 registers and where memory load instructions take more than one cycle.
6151 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6153 @item -fno-sched-interblock
6154 @opindex fno-sched-interblock
6155 Don't schedule instructions across basic blocks. This is normally
6156 enabled by default when scheduling before register allocation, i.e.@:
6157 with @option{-fschedule-insns} or at @option{-O2} or higher.
6159 @item -fno-sched-spec
6160 @opindex fno-sched-spec
6161 Don't allow speculative motion of non-load instructions. This is normally
6162 enabled by default when scheduling before register allocation, i.e.@:
6163 with @option{-fschedule-insns} or at @option{-O2} or higher.
6165 @item -fsched-spec-load
6166 @opindex fsched-spec-load
6167 Allow speculative motion of some load instructions. This only makes
6168 sense when scheduling before register allocation, i.e.@: with
6169 @option{-fschedule-insns} or at @option{-O2} or higher.
6171 @item -fsched-spec-load-dangerous
6172 @opindex fsched-spec-load-dangerous
6173 Allow speculative motion of more load instructions. This only makes
6174 sense when scheduling before register allocation, i.e.@: with
6175 @option{-fschedule-insns} or at @option{-O2} or higher.
6177 @item -fsched-stalled-insns
6178 @itemx -fsched-stalled-insns=@var{n}
6179 @opindex fsched-stalled-insns
6180 Define how many insns (if any) can be moved prematurely from the queue
6181 of stalled insns into the ready list, during the second scheduling pass.
6182 @option{-fno-sched-stalled-insns} means that no insns will be moved
6183 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6184 on how many queued insns can be moved prematurely.
6185 @option{-fsched-stalled-insns} without a value is equivalent to
6186 @option{-fsched-stalled-insns=1}.
6188 @item -fsched-stalled-insns-dep
6189 @itemx -fsched-stalled-insns-dep=@var{n}
6190 @opindex fsched-stalled-insns-dep
6191 Define how many insn groups (cycles) will be examined for a dependency
6192 on a stalled insn that is candidate for premature removal from the queue
6193 of stalled insns. This has an effect only during the second scheduling pass,
6194 and only if @option{-fsched-stalled-insns} is used.
6195 @option{-fno-sched-stalled-insns-dep} is equivalent to
6196 @option{-fsched-stalled-insns-dep=0}.
6197 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6198 @option{-fsched-stalled-insns-dep=1}.
6200 @item -fsched2-use-superblocks
6201 @opindex fsched2-use-superblocks
6202 When scheduling after register allocation, do use superblock scheduling
6203 algorithm. Superblock scheduling allows motion across basic block boundaries
6204 resulting on faster schedules. This option is experimental, as not all machine
6205 descriptions used by GCC model the CPU closely enough to avoid unreliable
6206 results from the algorithm.
6208 This only makes sense when scheduling after register allocation, i.e.@: with
6209 @option{-fschedule-insns2} or at @option{-O2} or higher.
6211 @item -fsched2-use-traces
6212 @opindex fsched2-use-traces
6213 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6214 allocation and additionally perform code duplication in order to increase the
6215 size of superblocks using tracer pass. See @option{-ftracer} for details on
6218 This mode should produce faster but significantly longer programs. Also
6219 without @option{-fbranch-probabilities} the traces constructed may not
6220 match the reality and hurt the performance. This only makes
6221 sense when scheduling after register allocation, i.e.@: with
6222 @option{-fschedule-insns2} or at @option{-O2} or higher.
6224 @item -freschedule-modulo-scheduled-loops
6225 @opindex freschedule-modulo-scheduled-loops
6226 The modulo scheduling comes before the traditional scheduling, if a loop
6227 was modulo scheduled we may want to prevent the later scheduling passes
6228 from changing its schedule, we use this option to control that.
6230 @item -fselective-scheduling
6231 @opindex fselective-scheduling
6232 Schedule instructions using selective scheduling algorithm. Selective
6233 scheduling runs instead of the first scheduler pass.
6235 @item -fselective-scheduling2
6236 @opindex fselective-scheduling2
6237 Schedule instructions using selective scheduling algorithm. Selective
6238 scheduling runs instead of the second scheduler pass.
6240 @item -fsel-sched-pipelining
6241 @opindex fsel-sched-pipelining
6242 Enable software pipelining of innermost loops during selective scheduling.
6243 This option has no effect until one of @option{-fselective-scheduling} or
6244 @option{-fselective-scheduling2} is turned on.
6246 @item -fsel-sched-pipelining-outer-loops
6247 @opindex fsel-sched-pipelining-outer-loops
6248 When pipelining loops during selective scheduling, also pipeline outer loops.
6249 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6251 @item -fcaller-saves
6252 @opindex fcaller-saves
6253 Enable values to be allocated in registers that will be clobbered by
6254 function calls, by emitting extra instructions to save and restore the
6255 registers around such calls. Such allocation is done only when it
6256 seems to result in better code than would otherwise be produced.
6258 This option is always enabled by default on certain machines, usually
6259 those which have no call-preserved registers to use instead.
6261 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6263 @item -fconserve-stack
6264 @opindex fconserve-stack
6265 Attempt to minimize stack usage. The compiler will attempt to use less
6266 stack space, even if that makes the program slower. This option
6267 implies setting the @option{large-stack-frame} parameter to 100
6268 and the @option{large-stack-frame-growth} parameter to 400.
6270 @item -ftree-reassoc
6271 @opindex ftree-reassoc
6272 Perform reassociation on trees. This flag is enabled by default
6273 at @option{-O} and higher.
6277 Perform partial redundancy elimination (PRE) on trees. This flag is
6278 enabled by default at @option{-O2} and @option{-O3}.
6280 @item -ftree-forwprop
6281 @opindex ftree-forwprop
6282 Perform forward propagation on trees. This flag is enabled by default
6283 at @option{-O} and higher.
6287 Perform full redundancy elimination (FRE) on trees. The difference
6288 between FRE and PRE is that FRE only considers expressions
6289 that are computed on all paths leading to the redundant computation.
6290 This analysis is faster than PRE, though it exposes fewer redundancies.
6291 This flag is enabled by default at @option{-O} and higher.
6293 @item -ftree-phiprop
6294 @opindex ftree-phiprop
6295 Perform hoisting of loads from conditional pointers on trees. This
6296 pass is enabled by default at @option{-O} and higher.
6298 @item -ftree-copy-prop
6299 @opindex ftree-copy-prop
6300 Perform copy propagation on trees. This pass eliminates unnecessary
6301 copy operations. This flag is enabled by default at @option{-O} and
6304 @item -fipa-pure-const
6305 @opindex fipa-pure-const
6306 Discover which functions are pure or constant.
6307 Enabled by default at @option{-O} and higher.
6309 @item -fipa-reference
6310 @opindex fipa-reference
6311 Discover which static variables do not escape cannot escape the
6313 Enabled by default at @option{-O} and higher.
6315 @item -fipa-struct-reorg
6316 @opindex fipa-struct-reorg
6317 Perform structure reorganization optimization, that change C-like structures
6318 layout in order to better utilize spatial locality. This transformation is
6319 affective for programs containing arrays of structures. Available in two
6320 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6321 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6322 to provide the safety of this transformation. It works only in whole program
6323 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6324 enabled. Structures considered @samp{cold} by this transformation are not
6325 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6327 With this flag, the program debug info reflects a new structure layout.
6331 Perform interprocedural pointer analysis. This option is experimental
6332 and does not affect generated code.
6336 Perform interprocedural constant propagation.
6337 This optimization analyzes the program to determine when values passed
6338 to functions are constants and then optimizes accordingly.
6339 This optimization can substantially increase performance
6340 if the application has constants passed to functions.
6341 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6343 @item -fipa-cp-clone
6344 @opindex fipa-cp-clone
6345 Perform function cloning to make interprocedural constant propagation stronger.
6346 When enabled, interprocedural constant propagation will perform function cloning
6347 when externally visible function can be called with constant arguments.
6348 Because this optimization can create multiple copies of functions,
6349 it may significantly increase code size
6350 (see @option{--param ipcp-unit-growth=@var{value}}).
6351 This flag is enabled by default at @option{-O3}.
6353 @item -fipa-matrix-reorg
6354 @opindex fipa-matrix-reorg
6355 Perform matrix flattening and transposing.
6356 Matrix flattening tries to replace an @math{m}-dimensional matrix
6357 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6358 This reduces the level of indirection needed for accessing the elements
6359 of the matrix. The second optimization is matrix transposing that
6360 attempts to change the order of the matrix's dimensions in order to
6361 improve cache locality.
6362 Both optimizations need the @option{-fwhole-program} flag.
6363 Transposing is enabled only if profiling information is available.
6368 Perform forward store motion on trees. This flag is
6369 enabled by default at @option{-O} and higher.
6373 Perform sparse conditional constant propagation (CCP) on trees. This
6374 pass only operates on local scalar variables and is enabled by default
6375 at @option{-O} and higher.
6377 @item -ftree-switch-conversion
6378 Perform conversion of simple initializations in a switch to
6379 initializations from a scalar array. This flag is enabled by default
6380 at @option{-O2} and higher.
6384 Perform dead code elimination (DCE) on trees. This flag is enabled by
6385 default at @option{-O} and higher.
6387 @item -ftree-builtin-call-dce
6388 @opindex ftree-builtin-call-dce
6389 Perform conditional dead code elimination (DCE) for calls to builtin functions
6390 that may set @code{errno} but are otherwise side-effect free. This flag is
6391 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6394 @item -ftree-dominator-opts
6395 @opindex ftree-dominator-opts
6396 Perform a variety of simple scalar cleanups (constant/copy
6397 propagation, redundancy elimination, range propagation and expression
6398 simplification) based on a dominator tree traversal. This also
6399 performs jump threading (to reduce jumps to jumps). This flag is
6400 enabled by default at @option{-O} and higher.
6404 Perform dead store elimination (DSE) on trees. A dead store is a store into
6405 a memory location which will later be overwritten by another store without
6406 any intervening loads. In this case the earlier store can be deleted. This
6407 flag is enabled by default at @option{-O} and higher.
6411 Perform loop header copying on trees. This is beneficial since it increases
6412 effectiveness of code motion optimizations. It also saves one jump. This flag
6413 is enabled by default at @option{-O} and higher. It is not enabled
6414 for @option{-Os}, since it usually increases code size.
6416 @item -ftree-loop-optimize
6417 @opindex ftree-loop-optimize
6418 Perform loop optimizations on trees. This flag is enabled by default
6419 at @option{-O} and higher.
6421 @item -ftree-loop-linear
6422 @opindex ftree-loop-linear
6423 Perform linear loop transformations on tree. This flag can improve cache
6424 performance and allow further loop optimizations to take place.
6426 @item -floop-interchange
6427 Perform loop interchange transformations on loops. Interchanging two
6428 nested loops switches the inner and outer loops. For example, given a
6433 A(J, I) = A(J, I) * C
6437 loop interchange will transform the loop as if the user had written:
6441 A(J, I) = A(J, I) * C
6445 which can be beneficial when @code{N} is larger than the caches,
6446 because in Fortran, the elements of an array are stored in memory
6447 contiguously by column, and the original loop iterates over rows,
6448 potentially creating at each access a cache miss. This optimization
6449 applies to all the languages supported by GCC and is not limited to
6450 Fortran. To use this code transformation, GCC has to be configured
6451 with @option{--with-ppl} and @option{--with-cloog} to enable the
6452 Graphite loop transformation infrastructure.
6454 @item -floop-strip-mine
6455 Perform loop strip mining transformations on loops. Strip mining
6456 splits a loop into two nested loops. The outer loop has strides
6457 equal to the strip size and the inner loop has strides of the
6458 original loop within a strip. For example, given a loop like:
6464 loop strip mining will transform the loop as if the user had written:
6467 DO I = II, min (II + 3, N)
6472 This optimization applies to all the languages supported by GCC and is
6473 not limited to Fortran. To use this code transformation, GCC has to
6474 be configured with @option{--with-ppl} and @option{--with-cloog} to
6475 enable the Graphite loop transformation infrastructure.
6478 Perform loop blocking transformations on loops. Blocking strip mines
6479 each loop in the loop nest such that the memory accesses of the
6480 element loops fit inside caches. For example, given a loop like:
6484 A(J, I) = B(I) + C(J)
6488 loop blocking will transform the loop as if the user had written:
6492 DO I = II, min (II + 63, N)
6493 DO J = JJ, min (JJ + 63, M)
6494 A(J, I) = B(I) + C(J)
6500 which can be beneficial when @code{M} is larger than the caches,
6501 because the innermost loop will iterate over a smaller amount of data
6502 that can be kept in the caches. This optimization applies to all the
6503 languages supported by GCC and is not limited to Fortran. To use this
6504 code transformation, GCC has to be configured with @option{--with-ppl}
6505 and @option{--with-cloog} to enable the Graphite loop transformation
6508 @item -fcheck-data-deps
6509 @opindex fcheck-data-deps
6510 Compare the results of several data dependence analyzers. This option
6511 is used for debugging the data dependence analyzers.
6513 @item -ftree-loop-distribution
6514 Perform loop distribution. This flag can improve cache performance on
6515 big loop bodies and allow further loop optimizations, like
6516 parallelization or vectorization, to take place. For example, the loop
6533 @item -ftree-loop-im
6534 @opindex ftree-loop-im
6535 Perform loop invariant motion on trees. This pass moves only invariants that
6536 would be hard to handle at RTL level (function calls, operations that expand to
6537 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6538 operands of conditions that are invariant out of the loop, so that we can use
6539 just trivial invariantness analysis in loop unswitching. The pass also includes
6542 @item -ftree-loop-ivcanon
6543 @opindex ftree-loop-ivcanon
6544 Create a canonical counter for number of iterations in the loop for that
6545 determining number of iterations requires complicated analysis. Later
6546 optimizations then may determine the number easily. Useful especially
6547 in connection with unrolling.
6551 Perform induction variable optimizations (strength reduction, induction
6552 variable merging and induction variable elimination) on trees.
6554 @item -ftree-parallelize-loops=n
6555 @opindex ftree-parallelize-loops
6556 Parallelize loops, i.e., split their iteration space to run in n threads.
6557 This is only possible for loops whose iterations are independent
6558 and can be arbitrarily reordered. The optimization is only
6559 profitable on multiprocessor machines, for loops that are CPU-intensive,
6560 rather than constrained e.g.@: by memory bandwidth. This option
6561 implies @option{-pthread}, and thus is only supported on targets
6562 that have support for @option{-pthread}.
6566 Perform function-local points-to analysis on trees. This flag is
6567 enabled by default at @option{-O} and higher.
6571 Perform scalar replacement of aggregates. This pass replaces structure
6572 references with scalars to prevent committing structures to memory too
6573 early. This flag is enabled by default at @option{-O} and higher.
6575 @item -ftree-copyrename
6576 @opindex ftree-copyrename
6577 Perform copy renaming on trees. This pass attempts to rename compiler
6578 temporaries to other variables at copy locations, usually resulting in
6579 variable names which more closely resemble the original variables. This flag
6580 is enabled by default at @option{-O} and higher.
6584 Perform temporary expression replacement during the SSA->normal phase. Single
6585 use/single def temporaries are replaced at their use location with their
6586 defining expression. This results in non-GIMPLE code, but gives the expanders
6587 much more complex trees to work on resulting in better RTL generation. This is
6588 enabled by default at @option{-O} and higher.
6590 @item -ftree-vectorize
6591 @opindex ftree-vectorize
6592 Perform loop vectorization on trees. This flag is enabled by default at
6595 @item -ftree-vect-loop-version
6596 @opindex ftree-vect-loop-version
6597 Perform loop versioning when doing loop vectorization on trees. When a loop
6598 appears to be vectorizable except that data alignment or data dependence cannot
6599 be determined at compile time then vectorized and non-vectorized versions of
6600 the loop are generated along with runtime checks for alignment or dependence
6601 to control which version is executed. This option is enabled by default
6602 except at level @option{-Os} where it is disabled.
6604 @item -fvect-cost-model
6605 @opindex fvect-cost-model
6606 Enable cost model for vectorization.
6610 Perform Value Range Propagation on trees. This is similar to the
6611 constant propagation pass, but instead of values, ranges of values are
6612 propagated. This allows the optimizers to remove unnecessary range
6613 checks like array bound checks and null pointer checks. This is
6614 enabled by default at @option{-O2} and higher. Null pointer check
6615 elimination is only done if @option{-fdelete-null-pointer-checks} is
6620 Perform tail duplication to enlarge superblock size. This transformation
6621 simplifies the control flow of the function allowing other optimizations to do
6624 @item -funroll-loops
6625 @opindex funroll-loops
6626 Unroll loops whose number of iterations can be determined at compile
6627 time or upon entry to the loop. @option{-funroll-loops} implies
6628 @option{-frerun-cse-after-loop}. This option makes code larger,
6629 and may or may not make it run faster.
6631 @item -funroll-all-loops
6632 @opindex funroll-all-loops
6633 Unroll all loops, even if their number of iterations is uncertain when
6634 the loop is entered. This usually makes programs run more slowly.
6635 @option{-funroll-all-loops} implies the same options as
6636 @option{-funroll-loops},
6638 @item -fsplit-ivs-in-unroller
6639 @opindex fsplit-ivs-in-unroller
6640 Enables expressing of values of induction variables in later iterations
6641 of the unrolled loop using the value in the first iteration. This breaks
6642 long dependency chains, thus improving efficiency of the scheduling passes.
6644 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6645 same effect. However in cases the loop body is more complicated than
6646 a single basic block, this is not reliable. It also does not work at all
6647 on some of the architectures due to restrictions in the CSE pass.
6649 This optimization is enabled by default.
6651 @item -fvariable-expansion-in-unroller
6652 @opindex fvariable-expansion-in-unroller
6653 With this option, the compiler will create multiple copies of some
6654 local variables when unrolling a loop which can result in superior code.
6656 @item -fpredictive-commoning
6657 @opindex fpredictive-commoning
6658 Perform predictive commoning optimization, i.e., reusing computations
6659 (especially memory loads and stores) performed in previous
6660 iterations of loops.
6662 This option is enabled at level @option{-O3}.
6664 @item -fprefetch-loop-arrays
6665 @opindex fprefetch-loop-arrays
6666 If supported by the target machine, generate instructions to prefetch
6667 memory to improve the performance of loops that access large arrays.
6669 This option may generate better or worse code; results are highly
6670 dependent on the structure of loops within the source code.
6672 Disabled at level @option{-Os}.
6675 @itemx -fno-peephole2
6676 @opindex fno-peephole
6677 @opindex fno-peephole2
6678 Disable any machine-specific peephole optimizations. The difference
6679 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6680 are implemented in the compiler; some targets use one, some use the
6681 other, a few use both.
6683 @option{-fpeephole} is enabled by default.
6684 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6686 @item -fno-guess-branch-probability
6687 @opindex fno-guess-branch-probability
6688 Do not guess branch probabilities using heuristics.
6690 GCC will use heuristics to guess branch probabilities if they are
6691 not provided by profiling feedback (@option{-fprofile-arcs}). These
6692 heuristics are based on the control flow graph. If some branch probabilities
6693 are specified by @samp{__builtin_expect}, then the heuristics will be
6694 used to guess branch probabilities for the rest of the control flow graph,
6695 taking the @samp{__builtin_expect} info into account. The interactions
6696 between the heuristics and @samp{__builtin_expect} can be complex, and in
6697 some cases, it may be useful to disable the heuristics so that the effects
6698 of @samp{__builtin_expect} are easier to understand.
6700 The default is @option{-fguess-branch-probability} at levels
6701 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6703 @item -freorder-blocks
6704 @opindex freorder-blocks
6705 Reorder basic blocks in the compiled function in order to reduce number of
6706 taken branches and improve code locality.
6708 Enabled at levels @option{-O2}, @option{-O3}.
6710 @item -freorder-blocks-and-partition
6711 @opindex freorder-blocks-and-partition
6712 In addition to reordering basic blocks in the compiled function, in order
6713 to reduce number of taken branches, partitions hot and cold basic blocks
6714 into separate sections of the assembly and .o files, to improve
6715 paging and cache locality performance.
6717 This optimization is automatically turned off in the presence of
6718 exception handling, for linkonce sections, for functions with a user-defined
6719 section attribute and on any architecture that does not support named
6722 @item -freorder-functions
6723 @opindex freorder-functions
6724 Reorder functions in the object file in order to
6725 improve code locality. This is implemented by using special
6726 subsections @code{.text.hot} for most frequently executed functions and
6727 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6728 the linker so object file format must support named sections and linker must
6729 place them in a reasonable way.
6731 Also profile feedback must be available in to make this option effective. See
6732 @option{-fprofile-arcs} for details.
6734 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6736 @item -fstrict-aliasing
6737 @opindex fstrict-aliasing
6738 Allows the compiler to assume the strictest aliasing rules applicable to
6739 the language being compiled. For C (and C++), this activates
6740 optimizations based on the type of expressions. In particular, an
6741 object of one type is assumed never to reside at the same address as an
6742 object of a different type, unless the types are almost the same. For
6743 example, an @code{unsigned int} can alias an @code{int}, but not a
6744 @code{void*} or a @code{double}. A character type may alias any other
6747 @anchor{Type-punning}Pay special attention to code like this:
6760 The practice of reading from a different union member than the one most
6761 recently written to (called ``type-punning'') is common. Even with
6762 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6763 is accessed through the union type. So, the code above will work as
6764 expected. @xref{Structures unions enumerations and bit-fields
6765 implementation}. However, this code might not:
6776 Similarly, access by taking the address, casting the resulting pointer
6777 and dereferencing the result has undefined behavior, even if the cast
6778 uses a union type, e.g.:
6782 return ((union a_union *) &d)->i;
6786 The @option{-fstrict-aliasing} option is enabled at levels
6787 @option{-O2}, @option{-O3}, @option{-Os}.
6789 @item -fstrict-overflow
6790 @opindex fstrict-overflow
6791 Allow the compiler to assume strict signed overflow rules, depending
6792 on the language being compiled. For C (and C++) this means that
6793 overflow when doing arithmetic with signed numbers is undefined, which
6794 means that the compiler may assume that it will not happen. This
6795 permits various optimizations. For example, the compiler will assume
6796 that an expression like @code{i + 10 > i} will always be true for
6797 signed @code{i}. This assumption is only valid if signed overflow is
6798 undefined, as the expression is false if @code{i + 10} overflows when
6799 using twos complement arithmetic. When this option is in effect any
6800 attempt to determine whether an operation on signed numbers will
6801 overflow must be written carefully to not actually involve overflow.
6803 This option also allows the compiler to assume strict pointer
6804 semantics: given a pointer to an object, if adding an offset to that
6805 pointer does not produce a pointer to the same object, the addition is
6806 undefined. This permits the compiler to conclude that @code{p + u >
6807 p} is always true for a pointer @code{p} and unsigned integer
6808 @code{u}. This assumption is only valid because pointer wraparound is
6809 undefined, as the expression is false if @code{p + u} overflows using
6810 twos complement arithmetic.
6812 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6813 that integer signed overflow is fully defined: it wraps. When
6814 @option{-fwrapv} is used, there is no difference between
6815 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6816 integers. With @option{-fwrapv} certain types of overflow are
6817 permitted. For example, if the compiler gets an overflow when doing
6818 arithmetic on constants, the overflowed value can still be used with
6819 @option{-fwrapv}, but not otherwise.
6821 The @option{-fstrict-overflow} option is enabled at levels
6822 @option{-O2}, @option{-O3}, @option{-Os}.
6824 @item -falign-functions
6825 @itemx -falign-functions=@var{n}
6826 @opindex falign-functions
6827 Align the start of functions to the next power-of-two greater than
6828 @var{n}, skipping up to @var{n} bytes. For instance,
6829 @option{-falign-functions=32} aligns functions to the next 32-byte
6830 boundary, but @option{-falign-functions=24} would align to the next
6831 32-byte boundary only if this can be done by skipping 23 bytes or less.
6833 @option{-fno-align-functions} and @option{-falign-functions=1} are
6834 equivalent and mean that functions will not be aligned.
6836 Some assemblers only support this flag when @var{n} is a power of two;
6837 in that case, it is rounded up.
6839 If @var{n} is not specified or is zero, use a machine-dependent default.
6841 Enabled at levels @option{-O2}, @option{-O3}.
6843 @item -falign-labels
6844 @itemx -falign-labels=@var{n}
6845 @opindex falign-labels
6846 Align all branch targets to a power-of-two boundary, skipping up to
6847 @var{n} bytes like @option{-falign-functions}. This option can easily
6848 make code slower, because it must insert dummy operations for when the
6849 branch target is reached in the usual flow of the code.
6851 @option{-fno-align-labels} and @option{-falign-labels=1} are
6852 equivalent and mean that labels will not be aligned.
6854 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6855 are greater than this value, then their values are used instead.
6857 If @var{n} is not specified or is zero, use a machine-dependent default
6858 which is very likely to be @samp{1}, meaning no alignment.
6860 Enabled at levels @option{-O2}, @option{-O3}.
6863 @itemx -falign-loops=@var{n}
6864 @opindex falign-loops
6865 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6866 like @option{-falign-functions}. The hope is that the loop will be
6867 executed many times, which will make up for any execution of the dummy
6870 @option{-fno-align-loops} and @option{-falign-loops=1} are
6871 equivalent and mean that loops will not be aligned.
6873 If @var{n} is not specified or is zero, use a machine-dependent default.
6875 Enabled at levels @option{-O2}, @option{-O3}.
6878 @itemx -falign-jumps=@var{n}
6879 @opindex falign-jumps
6880 Align branch targets to a power-of-two boundary, for branch targets
6881 where the targets can only be reached by jumping, skipping up to @var{n}
6882 bytes like @option{-falign-functions}. In this case, no dummy operations
6885 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6886 equivalent and mean that loops will not be aligned.
6888 If @var{n} is not specified or is zero, use a machine-dependent default.
6890 Enabled at levels @option{-O2}, @option{-O3}.
6892 @item -funit-at-a-time
6893 @opindex funit-at-a-time
6894 This option is left for compatibility reasons. @option{-funit-at-a-time}
6895 has no effect, while @option{-fno-unit-at-a-time} implies
6896 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6900 @item -fno-toplevel-reorder
6901 @opindex fno-toplevel-reorder
6902 Do not reorder top-level functions, variables, and @code{asm}
6903 statements. Output them in the same order that they appear in the
6904 input file. When this option is used, unreferenced static variables
6905 will not be removed. This option is intended to support existing code
6906 which relies on a particular ordering. For new code, it is better to
6909 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6910 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
6915 Constructs webs as commonly used for register allocation purposes and assign
6916 each web individual pseudo register. This allows the register allocation pass
6917 to operate on pseudos directly, but also strengthens several other optimization
6918 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6919 however, make debugging impossible, since variables will no longer stay in a
6922 Enabled by default with @option{-funroll-loops}.
6924 @item -fwhole-program
6925 @opindex fwhole-program
6926 Assume that the current compilation unit represents the whole program being
6927 compiled. All public functions and variables with the exception of @code{main}
6928 and those merged by attribute @code{externally_visible} become static functions
6929 and in effect are optimized more aggressively by interprocedural optimizers.
6930 While this option is equivalent to proper use of the @code{static} keyword for
6931 programs consisting of a single file, in combination with option
6932 @option{--combine} this flag can be used to compile many smaller scale C
6933 programs since the functions and variables become local for the whole combined
6934 compilation unit, not for the single source file itself.
6936 This option implies @option{-fwhole-file} for Fortran programs.
6938 @item -fcprop-registers
6939 @opindex fcprop-registers
6940 After register allocation and post-register allocation instruction splitting,
6941 we perform a copy-propagation pass to try to reduce scheduling dependencies
6942 and occasionally eliminate the copy.
6944 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6946 @item -fprofile-correction
6947 @opindex fprofile-correction
6948 Profiles collected using an instrumented binary for multi-threaded programs may
6949 be inconsistent due to missed counter updates. When this option is specified,
6950 GCC will use heuristics to correct or smooth out such inconsistencies. By
6951 default, GCC will emit an error message when an inconsistent profile is detected.
6953 @item -fprofile-dir=@var{path}
6954 @opindex fprofile-dir
6956 Set the directory to search the profile data files in to @var{path}.
6957 This option affects only the profile data generated by
6958 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6959 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6960 and its related options.
6961 By default, GCC will use the current directory as @var{path}
6962 thus the profile data file will appear in the same directory as the object file.
6964 @item -fprofile-generate
6965 @itemx -fprofile-generate=@var{path}
6966 @opindex fprofile-generate
6968 Enable options usually used for instrumenting application to produce
6969 profile useful for later recompilation with profile feedback based
6970 optimization. You must use @option{-fprofile-generate} both when
6971 compiling and when linking your program.
6973 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6975 If @var{path} is specified, GCC will look at the @var{path} to find
6976 the profile feedback data files. See @option{-fprofile-dir}.
6979 @itemx -fprofile-use=@var{path}
6980 @opindex fprofile-use
6981 Enable profile feedback directed optimizations, and optimizations
6982 generally profitable only with profile feedback available.
6984 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6985 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6987 By default, GCC emits an error message if the feedback profiles do not
6988 match the source code. This error can be turned into a warning by using
6989 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6992 If @var{path} is specified, GCC will look at the @var{path} to find
6993 the profile feedback data files. See @option{-fprofile-dir}.
6996 The following options control compiler behavior regarding floating
6997 point arithmetic. These options trade off between speed and
6998 correctness. All must be specifically enabled.
7002 @opindex ffloat-store
7003 Do not store floating point variables in registers, and inhibit other
7004 options that might change whether a floating point value is taken from a
7007 @cindex floating point precision
7008 This option prevents undesirable excess precision on machines such as
7009 the 68000 where the floating registers (of the 68881) keep more
7010 precision than a @code{double} is supposed to have. Similarly for the
7011 x86 architecture. For most programs, the excess precision does only
7012 good, but a few programs rely on the precise definition of IEEE floating
7013 point. Use @option{-ffloat-store} for such programs, after modifying
7014 them to store all pertinent intermediate computations into variables.
7016 @item -fexcess-precision=@var{style}
7017 @opindex fexcess-precision
7018 This option allows further control over excess precision on machines
7019 where floating-point registers have more precision than the IEEE
7020 @code{float} and @code{double} types and the processor does not
7021 support operations rounding to those types. By default,
7022 @option{-fexcess-precision=fast} is in effect; this means that
7023 operations are carried out in the precision of the registers and that
7024 it is unpredictable when rounding to the types specified in the source
7025 code takes place. When compiling C, if
7026 @option{-fexcess-precision=standard} is specified then excess
7027 precision will follow the rules specified in ISO C99; in particular,
7028 both casts and assignments cause values to be rounded to their
7029 semantic types (whereas @option{-ffloat-store} only affects
7030 assignments). This option is enabled by default for C if a strict
7031 conformance option such as @option{-std=c99} is used.
7034 @option{-fexcess-precision=standard} is not implemented for languages
7035 other than C, and has no effect if
7036 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7037 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7038 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7039 semantics apply without excess precision, and in the latter, rounding
7044 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7045 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7046 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7048 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7050 This option is not turned on by any @option{-O} option since
7051 it can result in incorrect output for programs which depend on
7052 an exact implementation of IEEE or ISO rules/specifications for
7053 math functions. It may, however, yield faster code for programs
7054 that do not require the guarantees of these specifications.
7056 @item -fno-math-errno
7057 @opindex fno-math-errno
7058 Do not set ERRNO after calling math functions that are executed
7059 with a single instruction, e.g., sqrt. A program that relies on
7060 IEEE exceptions for math error handling may want to use this flag
7061 for speed while maintaining IEEE arithmetic compatibility.
7063 This option is not turned on by any @option{-O} option since
7064 it can result in incorrect output for programs which depend on
7065 an exact implementation of IEEE or ISO rules/specifications for
7066 math functions. It may, however, yield faster code for programs
7067 that do not require the guarantees of these specifications.
7069 The default is @option{-fmath-errno}.
7071 On Darwin systems, the math library never sets @code{errno}. There is
7072 therefore no reason for the compiler to consider the possibility that
7073 it might, and @option{-fno-math-errno} is the default.
7075 @item -funsafe-math-optimizations
7076 @opindex funsafe-math-optimizations
7078 Allow optimizations for floating-point arithmetic that (a) assume
7079 that arguments and results are valid and (b) may violate IEEE or
7080 ANSI standards. When used at link-time, it may include libraries
7081 or startup files that change the default FPU control word or other
7082 similar optimizations.
7084 This option is not turned on by any @option{-O} option since
7085 it can result in incorrect output for programs which depend on
7086 an exact implementation of IEEE or ISO rules/specifications for
7087 math functions. It may, however, yield faster code for programs
7088 that do not require the guarantees of these specifications.
7089 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7090 @option{-fassociative-math} and @option{-freciprocal-math}.
7092 The default is @option{-fno-unsafe-math-optimizations}.
7094 @item -fassociative-math
7095 @opindex fassociative-math
7097 Allow re-association of operands in series of floating-point operations.
7098 This violates the ISO C and C++ language standard by possibly changing
7099 computation result. NOTE: re-ordering may change the sign of zero as
7100 well as ignore NaNs and inhibit or create underflow or overflow (and
7101 thus cannot be used on a code which relies on rounding behavior like
7102 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7103 and thus may not be used when ordered comparisons are required.
7104 This option requires that both @option{-fno-signed-zeros} and
7105 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7106 much sense with @option{-frounding-math}.
7108 The default is @option{-fno-associative-math}.
7110 @item -freciprocal-math
7111 @opindex freciprocal-math
7113 Allow the reciprocal of a value to be used instead of dividing by
7114 the value if this enables optimizations. For example @code{x / y}
7115 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7116 is subject to common subexpression elimination. Note that this loses
7117 precision and increases the number of flops operating on the value.
7119 The default is @option{-fno-reciprocal-math}.
7121 @item -ffinite-math-only
7122 @opindex ffinite-math-only
7123 Allow optimizations for floating-point arithmetic that assume
7124 that arguments and results are not NaNs or +-Infs.
7126 This option is not turned on by any @option{-O} option since
7127 it can result in incorrect output for programs which depend on
7128 an exact implementation of IEEE or ISO rules/specifications for
7129 math functions. It may, however, yield faster code for programs
7130 that do not require the guarantees of these specifications.
7132 The default is @option{-fno-finite-math-only}.
7134 @item -fno-signed-zeros
7135 @opindex fno-signed-zeros
7136 Allow optimizations for floating point arithmetic that ignore the
7137 signedness of zero. IEEE arithmetic specifies the behavior of
7138 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7139 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7140 This option implies that the sign of a zero result isn't significant.
7142 The default is @option{-fsigned-zeros}.
7144 @item -fno-trapping-math
7145 @opindex fno-trapping-math
7146 Compile code assuming that floating-point operations cannot generate
7147 user-visible traps. These traps include division by zero, overflow,
7148 underflow, inexact result and invalid operation. This option requires
7149 that @option{-fno-signaling-nans} be in effect. Setting this option may
7150 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7152 This option should never be turned on by any @option{-O} option since
7153 it can result in incorrect output for programs which depend on
7154 an exact implementation of IEEE or ISO rules/specifications for
7157 The default is @option{-ftrapping-math}.
7159 @item -frounding-math
7160 @opindex frounding-math
7161 Disable transformations and optimizations that assume default floating
7162 point rounding behavior. This is round-to-zero for all floating point
7163 to integer conversions, and round-to-nearest for all other arithmetic
7164 truncations. This option should be specified for programs that change
7165 the FP rounding mode dynamically, or that may be executed with a
7166 non-default rounding mode. This option disables constant folding of
7167 floating point expressions at compile-time (which may be affected by
7168 rounding mode) and arithmetic transformations that are unsafe in the
7169 presence of sign-dependent rounding modes.
7171 The default is @option{-fno-rounding-math}.
7173 This option is experimental and does not currently guarantee to
7174 disable all GCC optimizations that are affected by rounding mode.
7175 Future versions of GCC may provide finer control of this setting
7176 using C99's @code{FENV_ACCESS} pragma. This command line option
7177 will be used to specify the default state for @code{FENV_ACCESS}.
7179 @item -fsignaling-nans
7180 @opindex fsignaling-nans
7181 Compile code assuming that IEEE signaling NaNs may generate user-visible
7182 traps during floating-point operations. Setting this option disables
7183 optimizations that may change the number of exceptions visible with
7184 signaling NaNs. This option implies @option{-ftrapping-math}.
7186 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7189 The default is @option{-fno-signaling-nans}.
7191 This option is experimental and does not currently guarantee to
7192 disable all GCC optimizations that affect signaling NaN behavior.
7194 @item -fsingle-precision-constant
7195 @opindex fsingle-precision-constant
7196 Treat floating point constant as single precision constant instead of
7197 implicitly converting it to double precision constant.
7199 @item -fcx-limited-range
7200 @opindex fcx-limited-range
7201 When enabled, this option states that a range reduction step is not
7202 needed when performing complex division. Also, there is no checking
7203 whether the result of a complex multiplication or division is @code{NaN
7204 + I*NaN}, with an attempt to rescue the situation in that case. The
7205 default is @option{-fno-cx-limited-range}, but is enabled by
7206 @option{-ffast-math}.
7208 This option controls the default setting of the ISO C99
7209 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7212 @item -fcx-fortran-rules
7213 @opindex fcx-fortran-rules
7214 Complex multiplication and division follow Fortran rules. Range
7215 reduction is done as part of complex division, but there is no checking
7216 whether the result of a complex multiplication or division is @code{NaN
7217 + I*NaN}, with an attempt to rescue the situation in that case.
7219 The default is @option{-fno-cx-fortran-rules}.
7223 The following options control optimizations that may improve
7224 performance, but are not enabled by any @option{-O} options. This
7225 section includes experimental options that may produce broken code.
7228 @item -fbranch-probabilities
7229 @opindex fbranch-probabilities
7230 After running a program compiled with @option{-fprofile-arcs}
7231 (@pxref{Debugging Options,, Options for Debugging Your Program or
7232 @command{gcc}}), you can compile it a second time using
7233 @option{-fbranch-probabilities}, to improve optimizations based on
7234 the number of times each branch was taken. When the program
7235 compiled with @option{-fprofile-arcs} exits it saves arc execution
7236 counts to a file called @file{@var{sourcename}.gcda} for each source
7237 file. The information in this data file is very dependent on the
7238 structure of the generated code, so you must use the same source code
7239 and the same optimization options for both compilations.
7241 With @option{-fbranch-probabilities}, GCC puts a
7242 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7243 These can be used to improve optimization. Currently, they are only
7244 used in one place: in @file{reorg.c}, instead of guessing which path a
7245 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7246 exactly determine which path is taken more often.
7248 @item -fprofile-values
7249 @opindex fprofile-values
7250 If combined with @option{-fprofile-arcs}, it adds code so that some
7251 data about values of expressions in the program is gathered.
7253 With @option{-fbranch-probabilities}, it reads back the data gathered
7254 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7255 notes to instructions for their later usage in optimizations.
7257 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7261 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7262 a code to gather information about values of expressions.
7264 With @option{-fbranch-probabilities}, it reads back the data gathered
7265 and actually performs the optimizations based on them.
7266 Currently the optimizations include specialization of division operation
7267 using the knowledge about the value of the denominator.
7269 @item -frename-registers
7270 @opindex frename-registers
7271 Attempt to avoid false dependencies in scheduled code by making use
7272 of registers left over after register allocation. This optimization
7273 will most benefit processors with lots of registers. Depending on the
7274 debug information format adopted by the target, however, it can
7275 make debugging impossible, since variables will no longer stay in
7276 a ``home register''.
7278 Enabled by default with @option{-funroll-loops}.
7282 Perform tail duplication to enlarge superblock size. This transformation
7283 simplifies the control flow of the function allowing other optimizations to do
7286 Enabled with @option{-fprofile-use}.
7288 @item -funroll-loops
7289 @opindex funroll-loops
7290 Unroll loops whose number of iterations can be determined at compile time or
7291 upon entry to the loop. @option{-funroll-loops} implies
7292 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7293 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7294 small constant number of iterations). This option makes code larger, and may
7295 or may not make it run faster.
7297 Enabled with @option{-fprofile-use}.
7299 @item -funroll-all-loops
7300 @opindex funroll-all-loops
7301 Unroll all loops, even if their number of iterations is uncertain when
7302 the loop is entered. This usually makes programs run more slowly.
7303 @option{-funroll-all-loops} implies the same options as
7304 @option{-funroll-loops}.
7307 @opindex fpeel-loops
7308 Peels the loops for that there is enough information that they do not
7309 roll much (from profile feedback). It also turns on complete loop peeling
7310 (i.e.@: complete removal of loops with small constant number of iterations).
7312 Enabled with @option{-fprofile-use}.
7314 @item -fmove-loop-invariants
7315 @opindex fmove-loop-invariants
7316 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7317 at level @option{-O1}
7319 @item -funswitch-loops
7320 @opindex funswitch-loops
7321 Move branches with loop invariant conditions out of the loop, with duplicates
7322 of the loop on both branches (modified according to result of the condition).
7324 @item -ffunction-sections
7325 @itemx -fdata-sections
7326 @opindex ffunction-sections
7327 @opindex fdata-sections
7328 Place each function or data item into its own section in the output
7329 file if the target supports arbitrary sections. The name of the
7330 function or the name of the data item determines the section's name
7333 Use these options on systems where the linker can perform optimizations
7334 to improve locality of reference in the instruction space. Most systems
7335 using the ELF object format and SPARC processors running Solaris 2 have
7336 linkers with such optimizations. AIX may have these optimizations in
7339 Only use these options when there are significant benefits from doing
7340 so. When you specify these options, the assembler and linker will
7341 create larger object and executable files and will also be slower.
7342 You will not be able to use @code{gprof} on all systems if you
7343 specify this option and you may have problems with debugging if
7344 you specify both this option and @option{-g}.
7346 @item -fbranch-target-load-optimize
7347 @opindex fbranch-target-load-optimize
7348 Perform branch target register load optimization before prologue / epilogue
7350 The use of target registers can typically be exposed only during reload,
7351 thus hoisting loads out of loops and doing inter-block scheduling needs
7352 a separate optimization pass.
7354 @item -fbranch-target-load-optimize2
7355 @opindex fbranch-target-load-optimize2
7356 Perform branch target register load optimization after prologue / epilogue
7359 @item -fbtr-bb-exclusive
7360 @opindex fbtr-bb-exclusive
7361 When performing branch target register load optimization, don't reuse
7362 branch target registers in within any basic block.
7364 @item -fstack-protector
7365 @opindex fstack-protector
7366 Emit extra code to check for buffer overflows, such as stack smashing
7367 attacks. This is done by adding a guard variable to functions with
7368 vulnerable objects. This includes functions that call alloca, and
7369 functions with buffers larger than 8 bytes. The guards are initialized
7370 when a function is entered and then checked when the function exits.
7371 If a guard check fails, an error message is printed and the program exits.
7373 @item -fstack-protector-all
7374 @opindex fstack-protector-all
7375 Like @option{-fstack-protector} except that all functions are protected.
7377 @item -fsection-anchors
7378 @opindex fsection-anchors
7379 Try to reduce the number of symbolic address calculations by using
7380 shared ``anchor'' symbols to address nearby objects. This transformation
7381 can help to reduce the number of GOT entries and GOT accesses on some
7384 For example, the implementation of the following function @code{foo}:
7388 int foo (void) @{ return a + b + c; @}
7391 would usually calculate the addresses of all three variables, but if you
7392 compile it with @option{-fsection-anchors}, it will access the variables
7393 from a common anchor point instead. The effect is similar to the
7394 following pseudocode (which isn't valid C):
7399 register int *xr = &x;
7400 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7404 Not all targets support this option.
7406 @item --param @var{name}=@var{value}
7408 In some places, GCC uses various constants to control the amount of
7409 optimization that is done. For example, GCC will not inline functions
7410 that contain more that a certain number of instructions. You can
7411 control some of these constants on the command-line using the
7412 @option{--param} option.
7414 The names of specific parameters, and the meaning of the values, are
7415 tied to the internals of the compiler, and are subject to change
7416 without notice in future releases.
7418 In each case, the @var{value} is an integer. The allowable choices for
7419 @var{name} are given in the following table:
7422 @item struct-reorg-cold-struct-ratio
7423 The threshold ratio (as a percentage) between a structure frequency
7424 and the frequency of the hottest structure in the program. This parameter
7425 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7426 We say that if the ratio of a structure frequency, calculated by profiling,
7427 to the hottest structure frequency in the program is less than this
7428 parameter, then structure reorganization is not applied to this structure.
7431 @item predictable-branch-cost-outcome
7432 When branch is predicted to be taken with probability lower than this threshold
7433 (in percent), then it is considered well predictable. The default is 10.
7435 @item max-crossjump-edges
7436 The maximum number of incoming edges to consider for crossjumping.
7437 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7438 the number of edges incoming to each block. Increasing values mean
7439 more aggressive optimization, making the compile time increase with
7440 probably small improvement in executable size.
7442 @item min-crossjump-insns
7443 The minimum number of instructions which must be matched at the end
7444 of two blocks before crossjumping will be performed on them. This
7445 value is ignored in the case where all instructions in the block being
7446 crossjumped from are matched. The default value is 5.
7448 @item max-grow-copy-bb-insns
7449 The maximum code size expansion factor when copying basic blocks
7450 instead of jumping. The expansion is relative to a jump instruction.
7451 The default value is 8.
7453 @item max-goto-duplication-insns
7454 The maximum number of instructions to duplicate to a block that jumps
7455 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7456 passes, GCC factors computed gotos early in the compilation process,
7457 and unfactors them as late as possible. Only computed jumps at the
7458 end of a basic blocks with no more than max-goto-duplication-insns are
7459 unfactored. The default value is 8.
7461 @item max-delay-slot-insn-search
7462 The maximum number of instructions to consider when looking for an
7463 instruction to fill a delay slot. If more than this arbitrary number of
7464 instructions is searched, the time savings from filling the delay slot
7465 will be minimal so stop searching. Increasing values mean more
7466 aggressive optimization, making the compile time increase with probably
7467 small improvement in executable run time.
7469 @item max-delay-slot-live-search
7470 When trying to fill delay slots, the maximum number of instructions to
7471 consider when searching for a block with valid live register
7472 information. Increasing this arbitrarily chosen value means more
7473 aggressive optimization, increasing the compile time. This parameter
7474 should be removed when the delay slot code is rewritten to maintain the
7477 @item max-gcse-memory
7478 The approximate maximum amount of memory that will be allocated in
7479 order to perform the global common subexpression elimination
7480 optimization. If more memory than specified is required, the
7481 optimization will not be done.
7483 @item max-pending-list-length
7484 The maximum number of pending dependencies scheduling will allow
7485 before flushing the current state and starting over. Large functions
7486 with few branches or calls can create excessively large lists which
7487 needlessly consume memory and resources.
7489 @item max-inline-insns-single
7490 Several parameters control the tree inliner used in gcc.
7491 This number sets the maximum number of instructions (counted in GCC's
7492 internal representation) in a single function that the tree inliner
7493 will consider for inlining. This only affects functions declared
7494 inline and methods implemented in a class declaration (C++).
7495 The default value is 300.
7497 @item max-inline-insns-auto
7498 When you use @option{-finline-functions} (included in @option{-O3}),
7499 a lot of functions that would otherwise not be considered for inlining
7500 by the compiler will be investigated. To those functions, a different
7501 (more restrictive) limit compared to functions declared inline can
7503 The default value is 60.
7505 @item large-function-insns
7506 The limit specifying really large functions. For functions larger than this
7507 limit after inlining, inlining is constrained by
7508 @option{--param large-function-growth}. This parameter is useful primarily
7509 to avoid extreme compilation time caused by non-linear algorithms used by the
7511 The default value is 2700.
7513 @item large-function-growth
7514 Specifies maximal growth of large function caused by inlining in percents.
7515 The default value is 100 which limits large function growth to 2.0 times
7518 @item large-unit-insns
7519 The limit specifying large translation unit. Growth caused by inlining of
7520 units larger than this limit is limited by @option{--param inline-unit-growth}.
7521 For small units this might be too tight (consider unit consisting of function A
7522 that is inline and B that just calls A three time. If B is small relative to
7523 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7524 large units consisting of small inlineable functions however the overall unit
7525 growth limit is needed to avoid exponential explosion of code size. Thus for
7526 smaller units, the size is increased to @option{--param large-unit-insns}
7527 before applying @option{--param inline-unit-growth}. The default is 10000
7529 @item inline-unit-growth
7530 Specifies maximal overall growth of the compilation unit caused by inlining.
7531 The default value is 30 which limits unit growth to 1.3 times the original
7534 @item ipcp-unit-growth
7535 Specifies maximal overall growth of the compilation unit caused by
7536 interprocedural constant propagation. The default value is 10 which limits
7537 unit growth to 1.1 times the original size.
7539 @item large-stack-frame
7540 The limit specifying large stack frames. While inlining the algorithm is trying
7541 to not grow past this limit too much. Default value is 256 bytes.
7543 @item large-stack-frame-growth
7544 Specifies maximal growth of large stack frames caused by inlining in percents.
7545 The default value is 1000 which limits large stack frame growth to 11 times
7548 @item max-inline-insns-recursive
7549 @itemx max-inline-insns-recursive-auto
7550 Specifies maximum number of instructions out-of-line copy of self recursive inline
7551 function can grow into by performing recursive inlining.
7553 For functions declared inline @option{--param max-inline-insns-recursive} is
7554 taken into account. For function not declared inline, recursive inlining
7555 happens only when @option{-finline-functions} (included in @option{-O3}) is
7556 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7557 default value is 450.
7559 @item max-inline-recursive-depth
7560 @itemx max-inline-recursive-depth-auto
7561 Specifies maximum recursion depth used by the recursive inlining.
7563 For functions declared inline @option{--param max-inline-recursive-depth} is
7564 taken into account. For function not declared inline, recursive inlining
7565 happens only when @option{-finline-functions} (included in @option{-O3}) is
7566 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7569 @item min-inline-recursive-probability
7570 Recursive inlining is profitable only for function having deep recursion
7571 in average and can hurt for function having little recursion depth by
7572 increasing the prologue size or complexity of function body to other
7575 When profile feedback is available (see @option{-fprofile-generate}) the actual
7576 recursion depth can be guessed from probability that function will recurse via
7577 given call expression. This parameter limits inlining only to call expression
7578 whose probability exceeds given threshold (in percents). The default value is
7581 @item early-inlining-insns
7582 Specify growth that early inliner can make. In effect it increases amount of
7583 inlining for code having large abstraction penalty. The default value is 12.
7585 @item max-early-inliner-iterations
7586 @itemx max-early-inliner-iterations
7587 Limit of iterations of early inliner. This basically bounds number of nested
7588 indirect calls early inliner can resolve. Deeper chains are still handled by
7591 @item min-vect-loop-bound
7592 The minimum number of iterations under which a loop will not get vectorized
7593 when @option{-ftree-vectorize} is used. The number of iterations after
7594 vectorization needs to be greater than the value specified by this option
7595 to allow vectorization. The default value is 0.
7597 @item max-unrolled-insns
7598 The maximum number of instructions that a loop should have if that loop
7599 is unrolled, and if the loop is unrolled, it determines how many times
7600 the loop code is unrolled.
7602 @item max-average-unrolled-insns
7603 The maximum number of instructions biased by probabilities of their execution
7604 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7605 it determines how many times the loop code is unrolled.
7607 @item max-unroll-times
7608 The maximum number of unrollings of a single loop.
7610 @item max-peeled-insns
7611 The maximum number of instructions that a loop should have if that loop
7612 is peeled, and if the loop is peeled, it determines how many times
7613 the loop code is peeled.
7615 @item max-peel-times
7616 The maximum number of peelings of a single loop.
7618 @item max-completely-peeled-insns
7619 The maximum number of insns of a completely peeled loop.
7621 @item max-completely-peel-times
7622 The maximum number of iterations of a loop to be suitable for complete peeling.
7624 @item max-unswitch-insns
7625 The maximum number of insns of an unswitched loop.
7627 @item max-unswitch-level
7628 The maximum number of branches unswitched in a single loop.
7631 The minimum cost of an expensive expression in the loop invariant motion.
7633 @item iv-consider-all-candidates-bound
7634 Bound on number of candidates for induction variables below that
7635 all candidates are considered for each use in induction variable
7636 optimizations. Only the most relevant candidates are considered
7637 if there are more candidates, to avoid quadratic time complexity.
7639 @item iv-max-considered-uses
7640 The induction variable optimizations give up on loops that contain more
7641 induction variable uses.
7643 @item iv-always-prune-cand-set-bound
7644 If number of candidates in the set is smaller than this value,
7645 we always try to remove unnecessary ivs from the set during its
7646 optimization when a new iv is added to the set.
7648 @item scev-max-expr-size
7649 Bound on size of expressions used in the scalar evolutions analyzer.
7650 Large expressions slow the analyzer.
7652 @item omega-max-vars
7653 The maximum number of variables in an Omega constraint system.
7654 The default value is 128.
7656 @item omega-max-geqs
7657 The maximum number of inequalities in an Omega constraint system.
7658 The default value is 256.
7661 The maximum number of equalities in an Omega constraint system.
7662 The default value is 128.
7664 @item omega-max-wild-cards
7665 The maximum number of wildcard variables that the Omega solver will
7666 be able to insert. The default value is 18.
7668 @item omega-hash-table-size
7669 The size of the hash table in the Omega solver. The default value is
7672 @item omega-max-keys
7673 The maximal number of keys used by the Omega solver. The default
7676 @item omega-eliminate-redundant-constraints
7677 When set to 1, use expensive methods to eliminate all redundant
7678 constraints. The default value is 0.
7680 @item vect-max-version-for-alignment-checks
7681 The maximum number of runtime checks that can be performed when
7682 doing loop versioning for alignment in the vectorizer. See option
7683 ftree-vect-loop-version for more information.
7685 @item vect-max-version-for-alias-checks
7686 The maximum number of runtime checks that can be performed when
7687 doing loop versioning for alias in the vectorizer. See option
7688 ftree-vect-loop-version for more information.
7690 @item max-iterations-to-track
7692 The maximum number of iterations of a loop the brute force algorithm
7693 for analysis of # of iterations of the loop tries to evaluate.
7695 @item hot-bb-count-fraction
7696 Select fraction of the maximal count of repetitions of basic block in program
7697 given basic block needs to have to be considered hot.
7699 @item hot-bb-frequency-fraction
7700 Select fraction of the maximal frequency of executions of basic block in
7701 function given basic block needs to have to be considered hot
7703 @item max-predicted-iterations
7704 The maximum number of loop iterations we predict statically. This is useful
7705 in cases where function contain single loop with known bound and other loop
7706 with unknown. We predict the known number of iterations correctly, while
7707 the unknown number of iterations average to roughly 10. This means that the
7708 loop without bounds would appear artificially cold relative to the other one.
7710 @item align-threshold
7712 Select fraction of the maximal frequency of executions of basic block in
7713 function given basic block will get aligned.
7715 @item align-loop-iterations
7717 A loop expected to iterate at lest the selected number of iterations will get
7720 @item tracer-dynamic-coverage
7721 @itemx tracer-dynamic-coverage-feedback
7723 This value is used to limit superblock formation once the given percentage of
7724 executed instructions is covered. This limits unnecessary code size
7727 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7728 feedback is available. The real profiles (as opposed to statically estimated
7729 ones) are much less balanced allowing the threshold to be larger value.
7731 @item tracer-max-code-growth
7732 Stop tail duplication once code growth has reached given percentage. This is
7733 rather hokey argument, as most of the duplicates will be eliminated later in
7734 cross jumping, so it may be set to much higher values than is the desired code
7737 @item tracer-min-branch-ratio
7739 Stop reverse growth when the reverse probability of best edge is less than this
7740 threshold (in percent).
7742 @item tracer-min-branch-ratio
7743 @itemx tracer-min-branch-ratio-feedback
7745 Stop forward growth if the best edge do have probability lower than this
7748 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7749 compilation for profile feedback and one for compilation without. The value
7750 for compilation with profile feedback needs to be more conservative (higher) in
7751 order to make tracer effective.
7753 @item max-cse-path-length
7755 Maximum number of basic blocks on path that cse considers. The default is 10.
7758 The maximum instructions CSE process before flushing. The default is 1000.
7760 @item ggc-min-expand
7762 GCC uses a garbage collector to manage its own memory allocation. This
7763 parameter specifies the minimum percentage by which the garbage
7764 collector's heap should be allowed to expand between collections.
7765 Tuning this may improve compilation speed; it has no effect on code
7768 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7769 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7770 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7771 GCC is not able to calculate RAM on a particular platform, the lower
7772 bound of 30% is used. Setting this parameter and
7773 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7774 every opportunity. This is extremely slow, but can be useful for
7777 @item ggc-min-heapsize
7779 Minimum size of the garbage collector's heap before it begins bothering
7780 to collect garbage. The first collection occurs after the heap expands
7781 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7782 tuning this may improve compilation speed, and has no effect on code
7785 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7786 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7787 with a lower bound of 4096 (four megabytes) and an upper bound of
7788 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7789 particular platform, the lower bound is used. Setting this parameter
7790 very large effectively disables garbage collection. Setting this
7791 parameter and @option{ggc-min-expand} to zero causes a full collection
7792 to occur at every opportunity.
7794 @item max-reload-search-insns
7795 The maximum number of instruction reload should look backward for equivalent
7796 register. Increasing values mean more aggressive optimization, making the
7797 compile time increase with probably slightly better performance. The default
7800 @item max-cselib-memory-locations
7801 The maximum number of memory locations cselib should take into account.
7802 Increasing values mean more aggressive optimization, making the compile time
7803 increase with probably slightly better performance. The default value is 500.
7805 @item reorder-blocks-duplicate
7806 @itemx reorder-blocks-duplicate-feedback
7808 Used by basic block reordering pass to decide whether to use unconditional
7809 branch or duplicate the code on its destination. Code is duplicated when its
7810 estimated size is smaller than this value multiplied by the estimated size of
7811 unconditional jump in the hot spots of the program.
7813 The @option{reorder-block-duplicate-feedback} is used only when profile
7814 feedback is available and may be set to higher values than
7815 @option{reorder-block-duplicate} since information about the hot spots is more
7818 @item max-sched-ready-insns
7819 The maximum number of instructions ready to be issued the scheduler should
7820 consider at any given time during the first scheduling pass. Increasing
7821 values mean more thorough searches, making the compilation time increase
7822 with probably little benefit. The default value is 100.
7824 @item max-sched-region-blocks
7825 The maximum number of blocks in a region to be considered for
7826 interblock scheduling. The default value is 10.
7828 @item max-pipeline-region-blocks
7829 The maximum number of blocks in a region to be considered for
7830 pipelining in the selective scheduler. The default value is 15.
7832 @item max-sched-region-insns
7833 The maximum number of insns in a region to be considered for
7834 interblock scheduling. The default value is 100.
7836 @item max-pipeline-region-insns
7837 The maximum number of insns in a region to be considered for
7838 pipelining in the selective scheduler. The default value is 200.
7841 The minimum probability (in percents) of reaching a source block
7842 for interblock speculative scheduling. The default value is 40.
7844 @item max-sched-extend-regions-iters
7845 The maximum number of iterations through CFG to extend regions.
7846 0 - disable region extension,
7847 N - do at most N iterations.
7848 The default value is 0.
7850 @item max-sched-insn-conflict-delay
7851 The maximum conflict delay for an insn to be considered for speculative motion.
7852 The default value is 3.
7854 @item sched-spec-prob-cutoff
7855 The minimal probability of speculation success (in percents), so that
7856 speculative insn will be scheduled.
7857 The default value is 40.
7859 @item sched-mem-true-dep-cost
7860 Minimal distance (in CPU cycles) between store and load targeting same
7861 memory locations. The default value is 1.
7863 @item selsched-max-lookahead
7864 The maximum size of the lookahead window of selective scheduling. It is a
7865 depth of search for available instructions.
7866 The default value is 50.
7868 @item selsched-max-sched-times
7869 The maximum number of times that an instruction will be scheduled during
7870 selective scheduling. This is the limit on the number of iterations
7871 through which the instruction may be pipelined. The default value is 2.
7873 @item selsched-max-insns-to-rename
7874 The maximum number of best instructions in the ready list that are considered
7875 for renaming in the selective scheduler. The default value is 2.
7877 @item max-last-value-rtl
7878 The maximum size measured as number of RTLs that can be recorded in an expression
7879 in combiner for a pseudo register as last known value of that register. The default
7882 @item integer-share-limit
7883 Small integer constants can use a shared data structure, reducing the
7884 compiler's memory usage and increasing its speed. This sets the maximum
7885 value of a shared integer constant. The default value is 256.
7887 @item min-virtual-mappings
7888 Specifies the minimum number of virtual mappings in the incremental
7889 SSA updater that should be registered to trigger the virtual mappings
7890 heuristic defined by virtual-mappings-ratio. The default value is
7893 @item virtual-mappings-ratio
7894 If the number of virtual mappings is virtual-mappings-ratio bigger
7895 than the number of virtual symbols to be updated, then the incremental
7896 SSA updater switches to a full update for those symbols. The default
7899 @item ssp-buffer-size
7900 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7901 protection when @option{-fstack-protection} is used.
7903 @item max-jump-thread-duplication-stmts
7904 Maximum number of statements allowed in a block that needs to be
7905 duplicated when threading jumps.
7907 @item max-fields-for-field-sensitive
7908 Maximum number of fields in a structure we will treat in
7909 a field sensitive manner during pointer analysis. The default is zero
7910 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7912 @item prefetch-latency
7913 Estimate on average number of instructions that are executed before
7914 prefetch finishes. The distance we prefetch ahead is proportional
7915 to this constant. Increasing this number may also lead to less
7916 streams being prefetched (see @option{simultaneous-prefetches}).
7918 @item simultaneous-prefetches
7919 Maximum number of prefetches that can run at the same time.
7921 @item l1-cache-line-size
7922 The size of cache line in L1 cache, in bytes.
7925 The size of L1 cache, in kilobytes.
7928 The size of L2 cache, in kilobytes.
7930 @item min-insn-to-prefetch-ratio
7931 The minimum ratio between the number of instructions and the
7932 number of prefetches to enable prefetching in a loop with an
7935 @item prefetch-min-insn-to-mem-ratio
7936 The minimum ratio between the number of instructions and the
7937 number of memory references to enable prefetching in a loop.
7939 @item use-canonical-types
7940 Whether the compiler should use the ``canonical'' type system. By
7941 default, this should always be 1, which uses a more efficient internal
7942 mechanism for comparing types in C++ and Objective-C++. However, if
7943 bugs in the canonical type system are causing compilation failures,
7944 set this value to 0 to disable canonical types.
7946 @item switch-conversion-max-branch-ratio
7947 Switch initialization conversion will refuse to create arrays that are
7948 bigger than @option{switch-conversion-max-branch-ratio} times the number of
7949 branches in the switch.
7951 @item max-partial-antic-length
7952 Maximum length of the partial antic set computed during the tree
7953 partial redundancy elimination optimization (@option{-ftree-pre}) when
7954 optimizing at @option{-O3} and above. For some sorts of source code
7955 the enhanced partial redundancy elimination optimization can run away,
7956 consuming all of the memory available on the host machine. This
7957 parameter sets a limit on the length of the sets that are computed,
7958 which prevents the runaway behavior. Setting a value of 0 for
7959 this parameter will allow an unlimited set length.
7961 @item sccvn-max-scc-size
7962 Maximum size of a strongly connected component (SCC) during SCCVN
7963 processing. If this limit is hit, SCCVN processing for the whole
7964 function will not be done and optimizations depending on it will
7965 be disabled. The default maximum SCC size is 10000.
7967 @item ira-max-loops-num
7968 IRA uses a regional register allocation by default. If a function
7969 contains loops more than number given by the parameter, only at most
7970 given number of the most frequently executed loops will form regions
7971 for the regional register allocation. The default value of the
7974 @item ira-max-conflict-table-size
7975 Although IRA uses a sophisticated algorithm of compression conflict
7976 table, the table can be still big for huge functions. If the conflict
7977 table for a function could be more than size in MB given by the
7978 parameter, the conflict table is not built and faster, simpler, and
7979 lower quality register allocation algorithm will be used. The
7980 algorithm do not use pseudo-register conflicts. The default value of
7981 the parameter is 2000.
7983 @item loop-invariant-max-bbs-in-loop
7984 Loop invariant motion can be very expensive, both in compile time and
7985 in amount of needed compile time memory, with very large loops. Loops
7986 with more basic blocks than this parameter won't have loop invariant
7987 motion optimization performed on them. The default value of the
7988 parameter is 1000 for -O1 and 10000 for -O2 and above.
7993 @node Preprocessor Options
7994 @section Options Controlling the Preprocessor
7995 @cindex preprocessor options
7996 @cindex options, preprocessor
7998 These options control the C preprocessor, which is run on each C source
7999 file before actual compilation.
8001 If you use the @option{-E} option, nothing is done except preprocessing.
8002 Some of these options make sense only together with @option{-E} because
8003 they cause the preprocessor output to be unsuitable for actual
8007 @item -Wp,@var{option}
8009 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8010 and pass @var{option} directly through to the preprocessor. If
8011 @var{option} contains commas, it is split into multiple options at the
8012 commas. However, many options are modified, translated or interpreted
8013 by the compiler driver before being passed to the preprocessor, and
8014 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8015 interface is undocumented and subject to change, so whenever possible
8016 you should avoid using @option{-Wp} and let the driver handle the
8019 @item -Xpreprocessor @var{option}
8020 @opindex Xpreprocessor
8021 Pass @var{option} as an option to the preprocessor. You can use this to
8022 supply system-specific preprocessor options which GCC does not know how to
8025 If you want to pass an option that takes an argument, you must use
8026 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8029 @include cppopts.texi
8031 @node Assembler Options
8032 @section Passing Options to the Assembler
8034 @c prevent bad page break with this line
8035 You can pass options to the assembler.
8038 @item -Wa,@var{option}
8040 Pass @var{option} as an option to the assembler. If @var{option}
8041 contains commas, it is split into multiple options at the commas.
8043 @item -Xassembler @var{option}
8045 Pass @var{option} as an option to the assembler. You can use this to
8046 supply system-specific assembler options which GCC does not know how to
8049 If you want to pass an option that takes an argument, you must use
8050 @option{-Xassembler} twice, once for the option and once for the argument.
8055 @section Options for Linking
8056 @cindex link options
8057 @cindex options, linking
8059 These options come into play when the compiler links object files into
8060 an executable output file. They are meaningless if the compiler is
8061 not doing a link step.
8065 @item @var{object-file-name}
8066 A file name that does not end in a special recognized suffix is
8067 considered to name an object file or library. (Object files are
8068 distinguished from libraries by the linker according to the file
8069 contents.) If linking is done, these object files are used as input
8078 If any of these options is used, then the linker is not run, and
8079 object file names should not be used as arguments. @xref{Overall
8083 @item -l@var{library}
8084 @itemx -l @var{library}
8086 Search the library named @var{library} when linking. (The second
8087 alternative with the library as a separate argument is only for
8088 POSIX compliance and is not recommended.)
8090 It makes a difference where in the command you write this option; the
8091 linker searches and processes libraries and object files in the order they
8092 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8093 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8094 to functions in @samp{z}, those functions may not be loaded.
8096 The linker searches a standard list of directories for the library,
8097 which is actually a file named @file{lib@var{library}.a}. The linker
8098 then uses this file as if it had been specified precisely by name.
8100 The directories searched include several standard system directories
8101 plus any that you specify with @option{-L}.
8103 Normally the files found this way are library files---archive files
8104 whose members are object files. The linker handles an archive file by
8105 scanning through it for members which define symbols that have so far
8106 been referenced but not defined. But if the file that is found is an
8107 ordinary object file, it is linked in the usual fashion. The only
8108 difference between using an @option{-l} option and specifying a file name
8109 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8110 and searches several directories.
8114 You need this special case of the @option{-l} option in order to
8115 link an Objective-C or Objective-C++ program.
8118 @opindex nostartfiles
8119 Do not use the standard system startup files when linking.
8120 The standard system libraries are used normally, unless @option{-nostdlib}
8121 or @option{-nodefaultlibs} is used.
8123 @item -nodefaultlibs
8124 @opindex nodefaultlibs
8125 Do not use the standard system libraries when linking.
8126 Only the libraries you specify will be passed to the linker.
8127 The standard startup files are used normally, unless @option{-nostartfiles}
8128 is used. The compiler may generate calls to @code{memcmp},
8129 @code{memset}, @code{memcpy} and @code{memmove}.
8130 These entries are usually resolved by entries in
8131 libc. These entry points should be supplied through some other
8132 mechanism when this option is specified.
8136 Do not use the standard system startup files or libraries when linking.
8137 No startup files and only the libraries you specify will be passed to
8138 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
8139 @code{memcpy} and @code{memmove}.
8140 These entries are usually resolved by entries in
8141 libc. These entry points should be supplied through some other
8142 mechanism when this option is specified.
8144 @cindex @option{-lgcc}, use with @option{-nostdlib}
8145 @cindex @option{-nostdlib} and unresolved references
8146 @cindex unresolved references and @option{-nostdlib}
8147 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8148 @cindex @option{-nodefaultlibs} and unresolved references
8149 @cindex unresolved references and @option{-nodefaultlibs}
8150 One of the standard libraries bypassed by @option{-nostdlib} and
8151 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8152 that GCC uses to overcome shortcomings of particular machines, or special
8153 needs for some languages.
8154 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8155 Collection (GCC) Internals},
8156 for more discussion of @file{libgcc.a}.)
8157 In most cases, you need @file{libgcc.a} even when you want to avoid
8158 other standard libraries. In other words, when you specify @option{-nostdlib}
8159 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8160 This ensures that you have no unresolved references to internal GCC
8161 library subroutines. (For example, @samp{__main}, used to ensure C++
8162 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8163 GNU Compiler Collection (GCC) Internals}.)
8167 Produce a position independent executable on targets which support it.
8168 For predictable results, you must also specify the same set of options
8169 that were used to generate code (@option{-fpie}, @option{-fPIE},
8170 or model suboptions) when you specify this option.
8174 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8175 that support it. This instructs the linker to add all symbols, not
8176 only used ones, to the dynamic symbol table. This option is needed
8177 for some uses of @code{dlopen} or to allow obtaining backtraces
8178 from within a program.
8182 Remove all symbol table and relocation information from the executable.
8186 On systems that support dynamic linking, this prevents linking with the shared
8187 libraries. On other systems, this option has no effect.
8191 Produce a shared object which can then be linked with other objects to
8192 form an executable. Not all systems support this option. For predictable
8193 results, you must also specify the same set of options that were used to
8194 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8195 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8196 needs to build supplementary stub code for constructors to work. On
8197 multi-libbed systems, @samp{gcc -shared} must select the correct support
8198 libraries to link against. Failing to supply the correct flags may lead
8199 to subtle defects. Supplying them in cases where they are not necessary
8202 @item -shared-libgcc
8203 @itemx -static-libgcc
8204 @opindex shared-libgcc
8205 @opindex static-libgcc
8206 On systems that provide @file{libgcc} as a shared library, these options
8207 force the use of either the shared or static version respectively.
8208 If no shared version of @file{libgcc} was built when the compiler was
8209 configured, these options have no effect.
8211 There are several situations in which an application should use the
8212 shared @file{libgcc} instead of the static version. The most common
8213 of these is when the application wishes to throw and catch exceptions
8214 across different shared libraries. In that case, each of the libraries
8215 as well as the application itself should use the shared @file{libgcc}.
8217 Therefore, the G++ and GCJ drivers automatically add
8218 @option{-shared-libgcc} whenever you build a shared library or a main
8219 executable, because C++ and Java programs typically use exceptions, so
8220 this is the right thing to do.
8222 If, instead, you use the GCC driver to create shared libraries, you may
8223 find that they will not always be linked with the shared @file{libgcc}.
8224 If GCC finds, at its configuration time, that you have a non-GNU linker
8225 or a GNU linker that does not support option @option{--eh-frame-hdr},
8226 it will link the shared version of @file{libgcc} into shared libraries
8227 by default. Otherwise, it will take advantage of the linker and optimize
8228 away the linking with the shared version of @file{libgcc}, linking with
8229 the static version of libgcc by default. This allows exceptions to
8230 propagate through such shared libraries, without incurring relocation
8231 costs at library load time.
8233 However, if a library or main executable is supposed to throw or catch
8234 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8235 for the languages used in the program, or using the option
8236 @option{-shared-libgcc}, such that it is linked with the shared
8241 Bind references to global symbols when building a shared object. Warn
8242 about any unresolved references (unless overridden by the link editor
8243 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8246 @item -T @var{script}
8248 @cindex linker script
8249 Use @var{script} as the linker script. This option is supported by most
8250 systems using the GNU linker. On some targets, such as bare-board
8251 targets without an operating system, the @option{-T} option may be required
8252 when linking to avoid references to undefined symbols.
8254 @item -Xlinker @var{option}
8256 Pass @var{option} as an option to the linker. You can use this to
8257 supply system-specific linker options which GCC does not know how to
8260 If you want to pass an option that takes a separate argument, you must use
8261 @option{-Xlinker} twice, once for the option and once for the argument.
8262 For example, to pass @option{-assert definitions}, you must write
8263 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8264 @option{-Xlinker "-assert definitions"}, because this passes the entire
8265 string as a single argument, which is not what the linker expects.
8267 When using the GNU linker, it is usually more convenient to pass
8268 arguments to linker options using the @option{@var{option}=@var{value}}
8269 syntax than as separate arguments. For example, you can specify
8270 @samp{-Xlinker -Map=output.map} rather than
8271 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8272 this syntax for command-line options.
8274 @item -Wl,@var{option}
8276 Pass @var{option} as an option to the linker. If @var{option} contains
8277 commas, it is split into multiple options at the commas. You can use this
8278 syntax to pass an argument to the option.
8279 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8280 linker. When using the GNU linker, you can also get the same effect with
8281 @samp{-Wl,-Map=output.map}.
8283 @item -u @var{symbol}
8285 Pretend the symbol @var{symbol} is undefined, to force linking of
8286 library modules to define it. You can use @option{-u} multiple times with
8287 different symbols to force loading of additional library modules.
8290 @node Directory Options
8291 @section Options for Directory Search
8292 @cindex directory options
8293 @cindex options, directory search
8296 These options specify directories to search for header files, for
8297 libraries and for parts of the compiler:
8302 Add the directory @var{dir} to the head of the list of directories to be
8303 searched for header files. This can be used to override a system header
8304 file, substituting your own version, since these directories are
8305 searched before the system header file directories. However, you should
8306 not use this option to add directories that contain vendor-supplied
8307 system header files (use @option{-isystem} for that). If you use more than
8308 one @option{-I} option, the directories are scanned in left-to-right
8309 order; the standard system directories come after.
8311 If a standard system include directory, or a directory specified with
8312 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8313 option will be ignored. The directory will still be searched but as a
8314 system directory at its normal position in the system include chain.
8315 This is to ensure that GCC's procedure to fix buggy system headers and
8316 the ordering for the include_next directive are not inadvertently changed.
8317 If you really need to change the search order for system directories,
8318 use the @option{-nostdinc} and/or @option{-isystem} options.
8320 @item -iquote@var{dir}
8322 Add the directory @var{dir} to the head of the list of directories to
8323 be searched for header files only for the case of @samp{#include
8324 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8325 otherwise just like @option{-I}.
8329 Add directory @var{dir} to the list of directories to be searched
8332 @item -B@var{prefix}
8334 This option specifies where to find the executables, libraries,
8335 include files, and data files of the compiler itself.
8337 The compiler driver program runs one or more of the subprograms
8338 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8339 @var{prefix} as a prefix for each program it tries to run, both with and
8340 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8342 For each subprogram to be run, the compiler driver first tries the
8343 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8344 was not specified, the driver tries two standard prefixes, which are
8345 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8346 those results in a file name that is found, the unmodified program
8347 name is searched for using the directories specified in your
8348 @env{PATH} environment variable.
8350 The compiler will check to see if the path provided by the @option{-B}
8351 refers to a directory, and if necessary it will add a directory
8352 separator character at the end of the path.
8354 @option{-B} prefixes that effectively specify directory names also apply
8355 to libraries in the linker, because the compiler translates these
8356 options into @option{-L} options for the linker. They also apply to
8357 includes files in the preprocessor, because the compiler translates these
8358 options into @option{-isystem} options for the preprocessor. In this case,
8359 the compiler appends @samp{include} to the prefix.
8361 The run-time support file @file{libgcc.a} can also be searched for using
8362 the @option{-B} prefix, if needed. If it is not found there, the two
8363 standard prefixes above are tried, and that is all. The file is left
8364 out of the link if it is not found by those means.
8366 Another way to specify a prefix much like the @option{-B} prefix is to use
8367 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8370 As a special kludge, if the path provided by @option{-B} is
8371 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8372 9, then it will be replaced by @file{[dir/]include}. This is to help
8373 with boot-strapping the compiler.
8375 @item -specs=@var{file}
8377 Process @var{file} after the compiler reads in the standard @file{specs}
8378 file, in order to override the defaults that the @file{gcc} driver
8379 program uses when determining what switches to pass to @file{cc1},
8380 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8381 @option{-specs=@var{file}} can be specified on the command line, and they
8382 are processed in order, from left to right.
8384 @item --sysroot=@var{dir}
8386 Use @var{dir} as the logical root directory for headers and libraries.
8387 For example, if the compiler would normally search for headers in
8388 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8389 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8391 If you use both this option and the @option{-isysroot} option, then
8392 the @option{--sysroot} option will apply to libraries, but the
8393 @option{-isysroot} option will apply to header files.
8395 The GNU linker (beginning with version 2.16) has the necessary support
8396 for this option. If your linker does not support this option, the
8397 header file aspect of @option{--sysroot} will still work, but the
8398 library aspect will not.
8402 This option has been deprecated. Please use @option{-iquote} instead for
8403 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8404 Any directories you specify with @option{-I} options before the @option{-I-}
8405 option are searched only for the case of @samp{#include "@var{file}"};
8406 they are not searched for @samp{#include <@var{file}>}.
8408 If additional directories are specified with @option{-I} options after
8409 the @option{-I-}, these directories are searched for all @samp{#include}
8410 directives. (Ordinarily @emph{all} @option{-I} directories are used
8413 In addition, the @option{-I-} option inhibits the use of the current
8414 directory (where the current input file came from) as the first search
8415 directory for @samp{#include "@var{file}"}. There is no way to
8416 override this effect of @option{-I-}. With @option{-I.} you can specify
8417 searching the directory which was current when the compiler was
8418 invoked. That is not exactly the same as what the preprocessor does
8419 by default, but it is often satisfactory.
8421 @option{-I-} does not inhibit the use of the standard system directories
8422 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8429 @section Specifying subprocesses and the switches to pass to them
8432 @command{gcc} is a driver program. It performs its job by invoking a
8433 sequence of other programs to do the work of compiling, assembling and
8434 linking. GCC interprets its command-line parameters and uses these to
8435 deduce which programs it should invoke, and which command-line options
8436 it ought to place on their command lines. This behavior is controlled
8437 by @dfn{spec strings}. In most cases there is one spec string for each
8438 program that GCC can invoke, but a few programs have multiple spec
8439 strings to control their behavior. The spec strings built into GCC can
8440 be overridden by using the @option{-specs=} command-line switch to specify
8443 @dfn{Spec files} are plaintext files that are used to construct spec
8444 strings. They consist of a sequence of directives separated by blank
8445 lines. The type of directive is determined by the first non-whitespace
8446 character on the line and it can be one of the following:
8449 @item %@var{command}
8450 Issues a @var{command} to the spec file processor. The commands that can
8454 @item %include <@var{file}>
8456 Search for @var{file} and insert its text at the current point in the
8459 @item %include_noerr <@var{file}>
8460 @cindex %include_noerr
8461 Just like @samp{%include}, but do not generate an error message if the include
8462 file cannot be found.
8464 @item %rename @var{old_name} @var{new_name}
8466 Rename the spec string @var{old_name} to @var{new_name}.
8470 @item *[@var{spec_name}]:
8471 This tells the compiler to create, override or delete the named spec
8472 string. All lines after this directive up to the next directive or
8473 blank line are considered to be the text for the spec string. If this
8474 results in an empty string then the spec will be deleted. (Or, if the
8475 spec did not exist, then nothing will happened.) Otherwise, if the spec
8476 does not currently exist a new spec will be created. If the spec does
8477 exist then its contents will be overridden by the text of this
8478 directive, unless the first character of that text is the @samp{+}
8479 character, in which case the text will be appended to the spec.
8481 @item [@var{suffix}]:
8482 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8483 and up to the next directive or blank line are considered to make up the
8484 spec string for the indicated suffix. When the compiler encounters an
8485 input file with the named suffix, it will processes the spec string in
8486 order to work out how to compile that file. For example:
8493 This says that any input file whose name ends in @samp{.ZZ} should be
8494 passed to the program @samp{z-compile}, which should be invoked with the
8495 command-line switch @option{-input} and with the result of performing the
8496 @samp{%i} substitution. (See below.)
8498 As an alternative to providing a spec string, the text that follows a
8499 suffix directive can be one of the following:
8502 @item @@@var{language}
8503 This says that the suffix is an alias for a known @var{language}. This is
8504 similar to using the @option{-x} command-line switch to GCC to specify a
8505 language explicitly. For example:
8512 Says that .ZZ files are, in fact, C++ source files.
8515 This causes an error messages saying:
8518 @var{name} compiler not installed on this system.
8522 GCC already has an extensive list of suffixes built into it.
8523 This directive will add an entry to the end of the list of suffixes, but
8524 since the list is searched from the end backwards, it is effectively
8525 possible to override earlier entries using this technique.
8529 GCC has the following spec strings built into it. Spec files can
8530 override these strings or create their own. Note that individual
8531 targets can also add their own spec strings to this list.
8534 asm Options to pass to the assembler
8535 asm_final Options to pass to the assembler post-processor
8536 cpp Options to pass to the C preprocessor
8537 cc1 Options to pass to the C compiler
8538 cc1plus Options to pass to the C++ compiler
8539 endfile Object files to include at the end of the link
8540 link Options to pass to the linker
8541 lib Libraries to include on the command line to the linker
8542 libgcc Decides which GCC support library to pass to the linker
8543 linker Sets the name of the linker
8544 predefines Defines to be passed to the C preprocessor
8545 signed_char Defines to pass to CPP to say whether @code{char} is signed
8547 startfile Object files to include at the start of the link
8550 Here is a small example of a spec file:
8556 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8559 This example renames the spec called @samp{lib} to @samp{old_lib} and
8560 then overrides the previous definition of @samp{lib} with a new one.
8561 The new definition adds in some extra command-line options before
8562 including the text of the old definition.
8564 @dfn{Spec strings} are a list of command-line options to be passed to their
8565 corresponding program. In addition, the spec strings can contain
8566 @samp{%}-prefixed sequences to substitute variable text or to
8567 conditionally insert text into the command line. Using these constructs
8568 it is possible to generate quite complex command lines.
8570 Here is a table of all defined @samp{%}-sequences for spec
8571 strings. Note that spaces are not generated automatically around the
8572 results of expanding these sequences. Therefore you can concatenate them
8573 together or combine them with constant text in a single argument.
8577 Substitute one @samp{%} into the program name or argument.
8580 Substitute the name of the input file being processed.
8583 Substitute the basename of the input file being processed.
8584 This is the substring up to (and not including) the last period
8585 and not including the directory.
8588 This is the same as @samp{%b}, but include the file suffix (text after
8592 Marks the argument containing or following the @samp{%d} as a
8593 temporary file name, so that that file will be deleted if GCC exits
8594 successfully. Unlike @samp{%g}, this contributes no text to the
8597 @item %g@var{suffix}
8598 Substitute a file name that has suffix @var{suffix} and is chosen
8599 once per compilation, and mark the argument in the same way as
8600 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8601 name is now chosen in a way that is hard to predict even when previously
8602 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8603 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8604 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8605 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8606 was simply substituted with a file name chosen once per compilation,
8607 without regard to any appended suffix (which was therefore treated
8608 just like ordinary text), making such attacks more likely to succeed.
8610 @item %u@var{suffix}
8611 Like @samp{%g}, but generates a new temporary file name even if
8612 @samp{%u@var{suffix}} was already seen.
8614 @item %U@var{suffix}
8615 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8616 new one if there is no such last file name. In the absence of any
8617 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8618 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8619 would involve the generation of two distinct file names, one
8620 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8621 simply substituted with a file name chosen for the previous @samp{%u},
8622 without regard to any appended suffix.
8624 @item %j@var{suffix}
8625 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8626 writable, and if save-temps is off; otherwise, substitute the name
8627 of a temporary file, just like @samp{%u}. This temporary file is not
8628 meant for communication between processes, but rather as a junk
8631 @item %|@var{suffix}
8632 @itemx %m@var{suffix}
8633 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8634 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8635 all. These are the two most common ways to instruct a program that it
8636 should read from standard input or write to standard output. If you
8637 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8638 construct: see for example @file{f/lang-specs.h}.
8640 @item %.@var{SUFFIX}
8641 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8642 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8643 terminated by the next space or %.
8646 Marks the argument containing or following the @samp{%w} as the
8647 designated output file of this compilation. This puts the argument
8648 into the sequence of arguments that @samp{%o} will substitute later.
8651 Substitutes the names of all the output files, with spaces
8652 automatically placed around them. You should write spaces
8653 around the @samp{%o} as well or the results are undefined.
8654 @samp{%o} is for use in the specs for running the linker.
8655 Input files whose names have no recognized suffix are not compiled
8656 at all, but they are included among the output files, so they will
8660 Substitutes the suffix for object files. Note that this is
8661 handled specially when it immediately follows @samp{%g, %u, or %U},
8662 because of the need for those to form complete file names. The
8663 handling is such that @samp{%O} is treated exactly as if it had already
8664 been substituted, except that @samp{%g, %u, and %U} do not currently
8665 support additional @var{suffix} characters following @samp{%O} as they would
8666 following, for example, @samp{.o}.
8669 Substitutes the standard macro predefinitions for the
8670 current target machine. Use this when running @code{cpp}.
8673 Like @samp{%p}, but puts @samp{__} before and after the name of each
8674 predefined macro, except for macros that start with @samp{__} or with
8675 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8679 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8680 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8681 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8682 and @option{-imultilib} as necessary.
8685 Current argument is the name of a library or startup file of some sort.
8686 Search for that file in a standard list of directories and substitute
8687 the full name found.
8690 Print @var{str} as an error message. @var{str} is terminated by a newline.
8691 Use this when inconsistent options are detected.
8694 Substitute the contents of spec string @var{name} at this point.
8697 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8699 @item %x@{@var{option}@}
8700 Accumulate an option for @samp{%X}.
8703 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8707 Output the accumulated assembler options specified by @option{-Wa}.
8710 Output the accumulated preprocessor options specified by @option{-Wp}.
8713 Process the @code{asm} spec. This is used to compute the
8714 switches to be passed to the assembler.
8717 Process the @code{asm_final} spec. This is a spec string for
8718 passing switches to an assembler post-processor, if such a program is
8722 Process the @code{link} spec. This is the spec for computing the
8723 command line passed to the linker. Typically it will make use of the
8724 @samp{%L %G %S %D and %E} sequences.
8727 Dump out a @option{-L} option for each directory that GCC believes might
8728 contain startup files. If the target supports multilibs then the
8729 current multilib directory will be prepended to each of these paths.
8732 Process the @code{lib} spec. This is a spec string for deciding which
8733 libraries should be included on the command line to the linker.
8736 Process the @code{libgcc} spec. This is a spec string for deciding
8737 which GCC support library should be included on the command line to the linker.
8740 Process the @code{startfile} spec. This is a spec for deciding which
8741 object files should be the first ones passed to the linker. Typically
8742 this might be a file named @file{crt0.o}.
8745 Process the @code{endfile} spec. This is a spec string that specifies
8746 the last object files that will be passed to the linker.
8749 Process the @code{cpp} spec. This is used to construct the arguments
8750 to be passed to the C preprocessor.
8753 Process the @code{cc1} spec. This is used to construct the options to be
8754 passed to the actual C compiler (@samp{cc1}).
8757 Process the @code{cc1plus} spec. This is used to construct the options to be
8758 passed to the actual C++ compiler (@samp{cc1plus}).
8761 Substitute the variable part of a matched option. See below.
8762 Note that each comma in the substituted string is replaced by
8766 Remove all occurrences of @code{-S} from the command line. Note---this
8767 command is position dependent. @samp{%} commands in the spec string
8768 before this one will see @code{-S}, @samp{%} commands in the spec string
8769 after this one will not.
8771 @item %:@var{function}(@var{args})
8772 Call the named function @var{function}, passing it @var{args}.
8773 @var{args} is first processed as a nested spec string, then split
8774 into an argument vector in the usual fashion. The function returns
8775 a string which is processed as if it had appeared literally as part
8776 of the current spec.
8778 The following built-in spec functions are provided:
8782 The @code{getenv} spec function takes two arguments: an environment
8783 variable name and a string. If the environment variable is not
8784 defined, a fatal error is issued. Otherwise, the return value is the
8785 value of the environment variable concatenated with the string. For
8786 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8789 %:getenv(TOPDIR /include)
8792 expands to @file{/path/to/top/include}.
8794 @item @code{if-exists}
8795 The @code{if-exists} spec function takes one argument, an absolute
8796 pathname to a file. If the file exists, @code{if-exists} returns the
8797 pathname. Here is a small example of its usage:
8801 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8804 @item @code{if-exists-else}
8805 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8806 spec function, except that it takes two arguments. The first argument is
8807 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8808 returns the pathname. If it does not exist, it returns the second argument.
8809 This way, @code{if-exists-else} can be used to select one file or another,
8810 based on the existence of the first. Here is a small example of its usage:
8814 crt0%O%s %:if-exists(crti%O%s) \
8815 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8818 @item @code{replace-outfile}
8819 The @code{replace-outfile} spec function takes two arguments. It looks for the
8820 first argument in the outfiles array and replaces it with the second argument. Here
8821 is a small example of its usage:
8824 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8827 @item @code{print-asm-header}
8828 The @code{print-asm-header} function takes no arguments and simply
8829 prints a banner like:
8835 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8838 It is used to separate compiler options from assembler options
8839 in the @option{--target-help} output.
8843 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8844 If that switch was not specified, this substitutes nothing. Note that
8845 the leading dash is omitted when specifying this option, and it is
8846 automatically inserted if the substitution is performed. Thus the spec
8847 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8848 and would output the command line option @option{-foo}.
8850 @item %W@{@code{S}@}
8851 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8854 @item %@{@code{S}*@}
8855 Substitutes all the switches specified to GCC whose names start
8856 with @code{-S}, but which also take an argument. This is used for
8857 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8858 GCC considers @option{-o foo} as being
8859 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8860 text, including the space. Thus two arguments would be generated.
8862 @item %@{@code{S}*&@code{T}*@}
8863 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8864 (the order of @code{S} and @code{T} in the spec is not significant).
8865 There can be any number of ampersand-separated variables; for each the
8866 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8868 @item %@{@code{S}:@code{X}@}
8869 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8871 @item %@{!@code{S}:@code{X}@}
8872 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8874 @item %@{@code{S}*:@code{X}@}
8875 Substitutes @code{X} if one or more switches whose names start with
8876 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8877 once, no matter how many such switches appeared. However, if @code{%*}
8878 appears somewhere in @code{X}, then @code{X} will be substituted once
8879 for each matching switch, with the @code{%*} replaced by the part of
8880 that switch that matched the @code{*}.
8882 @item %@{.@code{S}:@code{X}@}
8883 Substitutes @code{X}, if processing a file with suffix @code{S}.
8885 @item %@{!.@code{S}:@code{X}@}
8886 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8888 @item %@{,@code{S}:@code{X}@}
8889 Substitutes @code{X}, if processing a file for language @code{S}.
8891 @item %@{!,@code{S}:@code{X}@}
8892 Substitutes @code{X}, if not processing a file for language @code{S}.
8894 @item %@{@code{S}|@code{P}:@code{X}@}
8895 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8896 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8897 @code{*} sequences as well, although they have a stronger binding than
8898 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8899 alternatives must be starred, and only the first matching alternative
8902 For example, a spec string like this:
8905 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8908 will output the following command-line options from the following input
8909 command-line options:
8914 -d fred.c -foo -baz -boggle
8915 -d jim.d -bar -baz -boggle
8918 @item %@{S:X; T:Y; :D@}
8920 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8921 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8922 be as many clauses as you need. This may be combined with @code{.},
8923 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8928 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8929 construct may contain other nested @samp{%} constructs or spaces, or
8930 even newlines. They are processed as usual, as described above.
8931 Trailing white space in @code{X} is ignored. White space may also
8932 appear anywhere on the left side of the colon in these constructs,
8933 except between @code{.} or @code{*} and the corresponding word.
8935 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8936 handled specifically in these constructs. If another value of
8937 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8938 @option{-W} switch is found later in the command line, the earlier
8939 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8940 just one letter, which passes all matching options.
8942 The character @samp{|} at the beginning of the predicate text is used to
8943 indicate that a command should be piped to the following command, but
8944 only if @option{-pipe} is specified.
8946 It is built into GCC which switches take arguments and which do not.
8947 (You might think it would be useful to generalize this to allow each
8948 compiler's spec to say which switches take arguments. But this cannot
8949 be done in a consistent fashion. GCC cannot even decide which input
8950 files have been specified without knowing which switches take arguments,
8951 and it must know which input files to compile in order to tell which
8954 GCC also knows implicitly that arguments starting in @option{-l} are to be
8955 treated as compiler output files, and passed to the linker in their
8956 proper position among the other output files.
8958 @c man begin OPTIONS
8960 @node Target Options
8961 @section Specifying Target Machine and Compiler Version
8962 @cindex target options
8963 @cindex cross compiling
8964 @cindex specifying machine version
8965 @cindex specifying compiler version and target machine
8966 @cindex compiler version, specifying
8967 @cindex target machine, specifying
8969 The usual way to run GCC is to run the executable called @file{gcc}, or
8970 @file{<machine>-gcc} when cross-compiling, or
8971 @file{<machine>-gcc-<version>} to run a version other than the one that
8972 was installed last. Sometimes this is inconvenient, so GCC provides
8973 options that will switch to another cross-compiler or version.
8976 @item -b @var{machine}
8978 The argument @var{machine} specifies the target machine for compilation.
8980 The value to use for @var{machine} is the same as was specified as the
8981 machine type when configuring GCC as a cross-compiler. For
8982 example, if a cross-compiler was configured with @samp{configure
8983 arm-elf}, meaning to compile for an arm processor with elf binaries,
8984 then you would specify @option{-b arm-elf} to run that cross compiler.
8985 Because there are other options beginning with @option{-b}, the
8986 configuration must contain a hyphen, or @option{-b} alone should be one
8987 argument followed by the configuration in the next argument.
8989 @item -V @var{version}
8991 The argument @var{version} specifies which version of GCC to run.
8992 This is useful when multiple versions are installed. For example,
8993 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8996 The @option{-V} and @option{-b} options work by running the
8997 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8998 use them if you can just run that directly.
9000 @node Submodel Options
9001 @section Hardware Models and Configurations
9002 @cindex submodel options
9003 @cindex specifying hardware config
9004 @cindex hardware models and configurations, specifying
9005 @cindex machine dependent options
9007 Earlier we discussed the standard option @option{-b} which chooses among
9008 different installed compilers for completely different target
9009 machines, such as VAX vs.@: 68000 vs.@: 80386.
9011 In addition, each of these target machine types can have its own
9012 special options, starting with @samp{-m}, to choose among various
9013 hardware models or configurations---for example, 68010 vs 68020,
9014 floating coprocessor or none. A single installed version of the
9015 compiler can compile for any model or configuration, according to the
9018 Some configurations of the compiler also support additional special
9019 options, usually for compatibility with other compilers on the same
9022 @c This list is ordered alphanumerically by subsection name.
9023 @c It should be the same order and spelling as these options are listed
9024 @c in Machine Dependent Options
9030 * Blackfin Options::
9034 * DEC Alpha Options::
9035 * DEC Alpha/VMS Options::
9038 * GNU/Linux Options::
9041 * i386 and x86-64 Options::
9042 * i386 and x86-64 Windows Options::
9053 * picoChip Options::
9055 * RS/6000 and PowerPC Options::
9056 * S/390 and zSeries Options::
9061 * System V Options::
9066 * Xstormy16 Options::
9072 @subsection ARC Options
9075 These options are defined for ARC implementations:
9080 Compile code for little endian mode. This is the default.
9084 Compile code for big endian mode.
9087 @opindex mmangle-cpu
9088 Prepend the name of the cpu to all public symbol names.
9089 In multiple-processor systems, there are many ARC variants with different
9090 instruction and register set characteristics. This flag prevents code
9091 compiled for one cpu to be linked with code compiled for another.
9092 No facility exists for handling variants that are ``almost identical''.
9093 This is an all or nothing option.
9095 @item -mcpu=@var{cpu}
9097 Compile code for ARC variant @var{cpu}.
9098 Which variants are supported depend on the configuration.
9099 All variants support @option{-mcpu=base}, this is the default.
9101 @item -mtext=@var{text-section}
9102 @itemx -mdata=@var{data-section}
9103 @itemx -mrodata=@var{readonly-data-section}
9107 Put functions, data, and readonly data in @var{text-section},
9108 @var{data-section}, and @var{readonly-data-section} respectively
9109 by default. This can be overridden with the @code{section} attribute.
9110 @xref{Variable Attributes}.
9112 @item -mfix-cortex-m3-ldrd
9113 @opindex mfix-cortex-m3-ldrd
9114 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9115 with overlapping destination and base registers are used. This option avoids
9116 generating these instructions. This option is enabled by default when
9117 @option{-mcpu=cortex-m3} is specified.
9122 @subsection ARM Options
9125 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9129 @item -mabi=@var{name}
9131 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9132 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9135 @opindex mapcs-frame
9136 Generate a stack frame that is compliant with the ARM Procedure Call
9137 Standard for all functions, even if this is not strictly necessary for
9138 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9139 with this option will cause the stack frames not to be generated for
9140 leaf functions. The default is @option{-mno-apcs-frame}.
9144 This is a synonym for @option{-mapcs-frame}.
9147 @c not currently implemented
9148 @item -mapcs-stack-check
9149 @opindex mapcs-stack-check
9150 Generate code to check the amount of stack space available upon entry to
9151 every function (that actually uses some stack space). If there is
9152 insufficient space available then either the function
9153 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9154 called, depending upon the amount of stack space required. The run time
9155 system is required to provide these functions. The default is
9156 @option{-mno-apcs-stack-check}, since this produces smaller code.
9158 @c not currently implemented
9160 @opindex mapcs-float
9161 Pass floating point arguments using the float point registers. This is
9162 one of the variants of the APCS@. This option is recommended if the
9163 target hardware has a floating point unit or if a lot of floating point
9164 arithmetic is going to be performed by the code. The default is
9165 @option{-mno-apcs-float}, since integer only code is slightly increased in
9166 size if @option{-mapcs-float} is used.
9168 @c not currently implemented
9169 @item -mapcs-reentrant
9170 @opindex mapcs-reentrant
9171 Generate reentrant, position independent code. The default is
9172 @option{-mno-apcs-reentrant}.
9175 @item -mthumb-interwork
9176 @opindex mthumb-interwork
9177 Generate code which supports calling between the ARM and Thumb
9178 instruction sets. Without this option the two instruction sets cannot
9179 be reliably used inside one program. The default is
9180 @option{-mno-thumb-interwork}, since slightly larger code is generated
9181 when @option{-mthumb-interwork} is specified.
9183 @item -mno-sched-prolog
9184 @opindex mno-sched-prolog
9185 Prevent the reordering of instructions in the function prolog, or the
9186 merging of those instruction with the instructions in the function's
9187 body. This means that all functions will start with a recognizable set
9188 of instructions (or in fact one of a choice from a small set of
9189 different function prologues), and this information can be used to
9190 locate the start if functions inside an executable piece of code. The
9191 default is @option{-msched-prolog}.
9193 @item -mfloat-abi=@var{name}
9195 Specifies which floating-point ABI to use. Permissible values
9196 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9198 Specifying @samp{soft} causes GCC to generate output containing
9199 library calls for floating-point operations.
9200 @samp{softfp} allows the generation of code using hardware floating-point
9201 instructions, but still uses the soft-float calling conventions.
9202 @samp{hard} allows generation of floating-point instructions
9203 and uses FPU-specific calling conventions.
9205 Using @option{-mfloat-abi=hard} with VFP coprocessors is not supported.
9206 Use @option{-mfloat-abi=softfp} with the appropriate @option{-mfpu} option
9207 to allow the compiler to generate code that makes use of the hardware
9208 floating-point capabilities for these CPUs.
9210 The default depends on the specific target configuration. Note that
9211 the hard-float and soft-float ABIs are not link-compatible; you must
9212 compile your entire program with the same ABI, and link with a
9213 compatible set of libraries.
9216 @opindex mhard-float
9217 Equivalent to @option{-mfloat-abi=hard}.
9220 @opindex msoft-float
9221 Equivalent to @option{-mfloat-abi=soft}.
9223 @item -mlittle-endian
9224 @opindex mlittle-endian
9225 Generate code for a processor running in little-endian mode. This is
9226 the default for all standard configurations.
9229 @opindex mbig-endian
9230 Generate code for a processor running in big-endian mode; the default is
9231 to compile code for a little-endian processor.
9233 @item -mwords-little-endian
9234 @opindex mwords-little-endian
9235 This option only applies when generating code for big-endian processors.
9236 Generate code for a little-endian word order but a big-endian byte
9237 order. That is, a byte order of the form @samp{32107654}. Note: this
9238 option should only be used if you require compatibility with code for
9239 big-endian ARM processors generated by versions of the compiler prior to
9242 @item -mcpu=@var{name}
9244 This specifies the name of the target ARM processor. GCC uses this name
9245 to determine what kind of instructions it can emit when generating
9246 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9247 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9248 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9249 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9250 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9252 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9253 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9254 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9255 @samp{strongarm1110},
9256 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9257 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9258 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9259 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9260 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9261 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9262 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9263 @samp{cortex-a8}, @samp{cortex-a9},
9264 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9266 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9268 @item -mtune=@var{name}
9270 This option is very similar to the @option{-mcpu=} option, except that
9271 instead of specifying the actual target processor type, and hence
9272 restricting which instructions can be used, it specifies that GCC should
9273 tune the performance of the code as if the target were of the type
9274 specified in this option, but still choosing the instructions that it
9275 will generate based on the cpu specified by a @option{-mcpu=} option.
9276 For some ARM implementations better performance can be obtained by using
9279 @item -march=@var{name}
9281 This specifies the name of the target ARM architecture. GCC uses this
9282 name to determine what kind of instructions it can emit when generating
9283 assembly code. This option can be used in conjunction with or instead
9284 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9285 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9286 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9287 @samp{armv6}, @samp{armv6j},
9288 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9289 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9290 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9292 @item -mfpu=@var{name}
9293 @itemx -mfpe=@var{number}
9294 @itemx -mfp=@var{number}
9298 This specifies what floating point hardware (or hardware emulation) is
9299 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9300 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16},
9301 @samp{neon}, and @samp{neon-fp16}. @option{-mfp} and @option{-mfpe}
9302 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9303 with older versions of GCC@.
9305 If @option{-msoft-float} is specified this specifies the format of
9306 floating point values.
9308 @item -mfp16-format=@var{name}
9309 @opindex mfp16-format
9310 Specify the format of the @code{__fp16} half-precision floating-point type.
9311 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9312 the default is @samp{none}, in which case the @code{__fp16} type is not
9313 defined. @xref{Half-Precision}, for more information.
9315 @item -mstructure-size-boundary=@var{n}
9316 @opindex mstructure-size-boundary
9317 The size of all structures and unions will be rounded up to a multiple
9318 of the number of bits set by this option. Permissible values are 8, 32
9319 and 64. The default value varies for different toolchains. For the COFF
9320 targeted toolchain the default value is 8. A value of 64 is only allowed
9321 if the underlying ABI supports it.
9323 Specifying the larger number can produce faster, more efficient code, but
9324 can also increase the size of the program. Different values are potentially
9325 incompatible. Code compiled with one value cannot necessarily expect to
9326 work with code or libraries compiled with another value, if they exchange
9327 information using structures or unions.
9329 @item -mabort-on-noreturn
9330 @opindex mabort-on-noreturn
9331 Generate a call to the function @code{abort} at the end of a
9332 @code{noreturn} function. It will be executed if the function tries to
9336 @itemx -mno-long-calls
9337 @opindex mlong-calls
9338 @opindex mno-long-calls
9339 Tells the compiler to perform function calls by first loading the
9340 address of the function into a register and then performing a subroutine
9341 call on this register. This switch is needed if the target function
9342 will lie outside of the 64 megabyte addressing range of the offset based
9343 version of subroutine call instruction.
9345 Even if this switch is enabled, not all function calls will be turned
9346 into long calls. The heuristic is that static functions, functions
9347 which have the @samp{short-call} attribute, functions that are inside
9348 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9349 definitions have already been compiled within the current compilation
9350 unit, will not be turned into long calls. The exception to this rule is
9351 that weak function definitions, functions with the @samp{long-call}
9352 attribute or the @samp{section} attribute, and functions that are within
9353 the scope of a @samp{#pragma long_calls} directive, will always be
9354 turned into long calls.
9356 This feature is not enabled by default. Specifying
9357 @option{-mno-long-calls} will restore the default behavior, as will
9358 placing the function calls within the scope of a @samp{#pragma
9359 long_calls_off} directive. Note these switches have no effect on how
9360 the compiler generates code to handle function calls via function
9363 @item -msingle-pic-base
9364 @opindex msingle-pic-base
9365 Treat the register used for PIC addressing as read-only, rather than
9366 loading it in the prologue for each function. The run-time system is
9367 responsible for initializing this register with an appropriate value
9368 before execution begins.
9370 @item -mpic-register=@var{reg}
9371 @opindex mpic-register
9372 Specify the register to be used for PIC addressing. The default is R10
9373 unless stack-checking is enabled, when R9 is used.
9375 @item -mcirrus-fix-invalid-insns
9376 @opindex mcirrus-fix-invalid-insns
9377 @opindex mno-cirrus-fix-invalid-insns
9378 Insert NOPs into the instruction stream to in order to work around
9379 problems with invalid Maverick instruction combinations. This option
9380 is only valid if the @option{-mcpu=ep9312} option has been used to
9381 enable generation of instructions for the Cirrus Maverick floating
9382 point co-processor. This option is not enabled by default, since the
9383 problem is only present in older Maverick implementations. The default
9384 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9387 @item -mpoke-function-name
9388 @opindex mpoke-function-name
9389 Write the name of each function into the text section, directly
9390 preceding the function prologue. The generated code is similar to this:
9394 .ascii "arm_poke_function_name", 0
9397 .word 0xff000000 + (t1 - t0)
9398 arm_poke_function_name
9400 stmfd sp!, @{fp, ip, lr, pc@}
9404 When performing a stack backtrace, code can inspect the value of
9405 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9406 location @code{pc - 12} and the top 8 bits are set, then we know that
9407 there is a function name embedded immediately preceding this location
9408 and has length @code{((pc[-3]) & 0xff000000)}.
9412 Generate code for the Thumb instruction set. The default is to
9413 use the 32-bit ARM instruction set.
9414 This option automatically enables either 16-bit Thumb-1 or
9415 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9416 and @option{-march=@var{name}} options. This option is not passed to the
9417 assembler. If you want to force assembler files to be interpreted as Thumb code,
9418 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9419 option directly to the assembler by prefixing it with @option{-Wa}.
9422 @opindex mtpcs-frame
9423 Generate a stack frame that is compliant with the Thumb Procedure Call
9424 Standard for all non-leaf functions. (A leaf function is one that does
9425 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9427 @item -mtpcs-leaf-frame
9428 @opindex mtpcs-leaf-frame
9429 Generate a stack frame that is compliant with the Thumb Procedure Call
9430 Standard for all leaf functions. (A leaf function is one that does
9431 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9433 @item -mcallee-super-interworking
9434 @opindex mcallee-super-interworking
9435 Gives all externally visible functions in the file being compiled an ARM
9436 instruction set header which switches to Thumb mode before executing the
9437 rest of the function. This allows these functions to be called from
9438 non-interworking code.
9440 @item -mcaller-super-interworking
9441 @opindex mcaller-super-interworking
9442 Allows calls via function pointers (including virtual functions) to
9443 execute correctly regardless of whether the target code has been
9444 compiled for interworking or not. There is a small overhead in the cost
9445 of executing a function pointer if this option is enabled.
9447 @item -mtp=@var{name}
9449 Specify the access model for the thread local storage pointer. The valid
9450 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9451 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9452 (supported in the arm6k architecture), and @option{auto}, which uses the
9453 best available method for the selected processor. The default setting is
9456 @item -mword-relocations
9457 @opindex mword-relocations
9458 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9459 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9460 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9466 @subsection AVR Options
9469 These options are defined for AVR implementations:
9472 @item -mmcu=@var{mcu}
9474 Specify ATMEL AVR instruction set or MCU type.
9476 Instruction set avr1 is for the minimal AVR core, not supported by the C
9477 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9478 attiny11, attiny12, attiny15, attiny28).
9480 Instruction set avr2 (default) is for the classic AVR core with up to
9481 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9482 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9483 at90c8534, at90s8535).
9485 Instruction set avr3 is for the classic AVR core with up to 128K program
9486 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9488 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9489 memory space (MCU types: atmega8, atmega83, atmega85).
9491 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9492 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9493 atmega64, atmega128, at43usb355, at94k).
9497 Output instruction sizes to the asm file.
9499 @item -minit-stack=@var{N}
9500 @opindex minit-stack
9501 Specify the initial stack address, which may be a symbol or numeric value,
9502 @samp{__stack} is the default.
9504 @item -mno-interrupts
9505 @opindex mno-interrupts
9506 Generated code is not compatible with hardware interrupts.
9507 Code size will be smaller.
9509 @item -mcall-prologues
9510 @opindex mcall-prologues
9511 Functions prologues/epilogues expanded as call to appropriate
9512 subroutines. Code size will be smaller.
9515 @opindex mtiny-stack
9516 Change only the low 8 bits of the stack pointer.
9520 Assume int to be 8 bit integer. This affects the sizes of all types: A
9521 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
9522 and long long will be 4 bytes. Please note that this option does not
9523 comply to the C standards, but it will provide you with smaller code
9527 @node Blackfin Options
9528 @subsection Blackfin Options
9529 @cindex Blackfin Options
9532 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9534 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9535 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9536 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9537 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9538 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9539 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9541 The optional @var{sirevision} specifies the silicon revision of the target
9542 Blackfin processor. Any workarounds available for the targeted silicon revision
9543 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9544 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9545 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9546 hexadecimal digits representing the major and minor numbers in the silicon
9547 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9548 is not defined. If @var{sirevision} is @samp{any}, the
9549 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9550 If this optional @var{sirevision} is not used, GCC assumes the latest known
9551 silicon revision of the targeted Blackfin processor.
9553 Support for @samp{bf561} is incomplete. For @samp{bf561},
9554 Only the processor macro is defined.
9555 Without this option, @samp{bf532} is used as the processor by default.
9556 The corresponding predefined processor macros for @var{cpu} is to
9557 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9558 provided by libgloss to be linked in if @option{-msim} is not given.
9562 Specifies that the program will be run on the simulator. This causes
9563 the simulator BSP provided by libgloss to be linked in. This option
9564 has effect only for @samp{bfin-elf} toolchain.
9565 Certain other options, such as @option{-mid-shared-library} and
9566 @option{-mfdpic}, imply @option{-msim}.
9568 @item -momit-leaf-frame-pointer
9569 @opindex momit-leaf-frame-pointer
9570 Don't keep the frame pointer in a register for leaf functions. This
9571 avoids the instructions to save, set up and restore frame pointers and
9572 makes an extra register available in leaf functions. The option
9573 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9574 which might make debugging harder.
9576 @item -mspecld-anomaly
9577 @opindex mspecld-anomaly
9578 When enabled, the compiler will ensure that the generated code does not
9579 contain speculative loads after jump instructions. If this option is used,
9580 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9582 @item -mno-specld-anomaly
9583 @opindex mno-specld-anomaly
9584 Don't generate extra code to prevent speculative loads from occurring.
9586 @item -mcsync-anomaly
9587 @opindex mcsync-anomaly
9588 When enabled, the compiler will ensure that the generated code does not
9589 contain CSYNC or SSYNC instructions too soon after conditional branches.
9590 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9592 @item -mno-csync-anomaly
9593 @opindex mno-csync-anomaly
9594 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9595 occurring too soon after a conditional branch.
9599 When enabled, the compiler is free to take advantage of the knowledge that
9600 the entire program fits into the low 64k of memory.
9603 @opindex mno-low-64k
9604 Assume that the program is arbitrarily large. This is the default.
9606 @item -mstack-check-l1
9607 @opindex mstack-check-l1
9608 Do stack checking using information placed into L1 scratchpad memory by the
9611 @item -mid-shared-library
9612 @opindex mid-shared-library
9613 Generate code that supports shared libraries via the library ID method.
9614 This allows for execute in place and shared libraries in an environment
9615 without virtual memory management. This option implies @option{-fPIC}.
9616 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9618 @item -mno-id-shared-library
9619 @opindex mno-id-shared-library
9620 Generate code that doesn't assume ID based shared libraries are being used.
9621 This is the default.
9623 @item -mleaf-id-shared-library
9624 @opindex mleaf-id-shared-library
9625 Generate code that supports shared libraries via the library ID method,
9626 but assumes that this library or executable won't link against any other
9627 ID shared libraries. That allows the compiler to use faster code for jumps
9630 @item -mno-leaf-id-shared-library
9631 @opindex mno-leaf-id-shared-library
9632 Do not assume that the code being compiled won't link against any ID shared
9633 libraries. Slower code will be generated for jump and call insns.
9635 @item -mshared-library-id=n
9636 @opindex mshared-library-id
9637 Specified the identification number of the ID based shared library being
9638 compiled. Specifying a value of 0 will generate more compact code, specifying
9639 other values will force the allocation of that number to the current
9640 library but is no more space or time efficient than omitting this option.
9644 Generate code that allows the data segment to be located in a different
9645 area of memory from the text segment. This allows for execute in place in
9646 an environment without virtual memory management by eliminating relocations
9647 against the text section.
9650 @opindex mno-sep-data
9651 Generate code that assumes that the data segment follows the text segment.
9652 This is the default.
9655 @itemx -mno-long-calls
9656 @opindex mlong-calls
9657 @opindex mno-long-calls
9658 Tells the compiler to perform function calls by first loading the
9659 address of the function into a register and then performing a subroutine
9660 call on this register. This switch is needed if the target function
9661 will lie outside of the 24 bit addressing range of the offset based
9662 version of subroutine call instruction.
9664 This feature is not enabled by default. Specifying
9665 @option{-mno-long-calls} will restore the default behavior. Note these
9666 switches have no effect on how the compiler generates code to handle
9667 function calls via function pointers.
9671 Link with the fast floating-point library. This library relaxes some of
9672 the IEEE floating-point standard's rules for checking inputs against
9673 Not-a-Number (NAN), in the interest of performance.
9676 @opindex minline-plt
9677 Enable inlining of PLT entries in function calls to functions that are
9678 not known to bind locally. It has no effect without @option{-mfdpic}.
9682 Build standalone application for multicore Blackfin processor. Proper
9683 start files and link scripts will be used to support multicore.
9684 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9685 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9686 @option{-mcorea} or @option{-mcoreb}. If it's used without
9687 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9688 programming model is used. In this model, the main function of Core B
9689 should be named as coreb_main. If it's used with @option{-mcorea} or
9690 @option{-mcoreb}, one application per core programming model is used.
9691 If this option is not used, single core application programming
9696 Build standalone application for Core A of BF561 when using
9697 one application per core programming model. Proper start files
9698 and link scripts will be used to support Core A. This option
9699 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9703 Build standalone application for Core B of BF561 when using
9704 one application per core programming model. Proper start files
9705 and link scripts will be used to support Core B. This option
9706 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9707 should be used instead of main. It must be used with
9708 @option{-mmulticore}.
9712 Build standalone application for SDRAM. Proper start files and
9713 link scripts will be used to put the application into SDRAM.
9714 Loader should initialize SDRAM before loading the application
9715 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9719 Assume that ICPLBs are enabled at runtime. This has an effect on certain
9720 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
9721 are enabled; for standalone applications the default is off.
9725 @subsection CRIS Options
9726 @cindex CRIS Options
9728 These options are defined specifically for the CRIS ports.
9731 @item -march=@var{architecture-type}
9732 @itemx -mcpu=@var{architecture-type}
9735 Generate code for the specified architecture. The choices for
9736 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9737 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9738 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9741 @item -mtune=@var{architecture-type}
9743 Tune to @var{architecture-type} everything applicable about the generated
9744 code, except for the ABI and the set of available instructions. The
9745 choices for @var{architecture-type} are the same as for
9746 @option{-march=@var{architecture-type}}.
9748 @item -mmax-stack-frame=@var{n}
9749 @opindex mmax-stack-frame
9750 Warn when the stack frame of a function exceeds @var{n} bytes.
9756 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9757 @option{-march=v3} and @option{-march=v8} respectively.
9759 @item -mmul-bug-workaround
9760 @itemx -mno-mul-bug-workaround
9761 @opindex mmul-bug-workaround
9762 @opindex mno-mul-bug-workaround
9763 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9764 models where it applies. This option is active by default.
9768 Enable CRIS-specific verbose debug-related information in the assembly
9769 code. This option also has the effect to turn off the @samp{#NO_APP}
9770 formatted-code indicator to the assembler at the beginning of the
9775 Do not use condition-code results from previous instruction; always emit
9776 compare and test instructions before use of condition codes.
9778 @item -mno-side-effects
9779 @opindex mno-side-effects
9780 Do not emit instructions with side-effects in addressing modes other than
9784 @itemx -mno-stack-align
9786 @itemx -mno-data-align
9787 @itemx -mconst-align
9788 @itemx -mno-const-align
9789 @opindex mstack-align
9790 @opindex mno-stack-align
9791 @opindex mdata-align
9792 @opindex mno-data-align
9793 @opindex mconst-align
9794 @opindex mno-const-align
9795 These options (no-options) arranges (eliminate arrangements) for the
9796 stack-frame, individual data and constants to be aligned for the maximum
9797 single data access size for the chosen CPU model. The default is to
9798 arrange for 32-bit alignment. ABI details such as structure layout are
9799 not affected by these options.
9807 Similar to the stack- data- and const-align options above, these options
9808 arrange for stack-frame, writable data and constants to all be 32-bit,
9809 16-bit or 8-bit aligned. The default is 32-bit alignment.
9811 @item -mno-prologue-epilogue
9812 @itemx -mprologue-epilogue
9813 @opindex mno-prologue-epilogue
9814 @opindex mprologue-epilogue
9815 With @option{-mno-prologue-epilogue}, the normal function prologue and
9816 epilogue that sets up the stack-frame are omitted and no return
9817 instructions or return sequences are generated in the code. Use this
9818 option only together with visual inspection of the compiled code: no
9819 warnings or errors are generated when call-saved registers must be saved,
9820 or storage for local variable needs to be allocated.
9826 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9827 instruction sequences that load addresses for functions from the PLT part
9828 of the GOT rather than (traditional on other architectures) calls to the
9829 PLT@. The default is @option{-mgotplt}.
9833 Legacy no-op option only recognized with the cris-axis-elf and
9834 cris-axis-linux-gnu targets.
9838 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9842 This option, recognized for the cris-axis-elf arranges
9843 to link with input-output functions from a simulator library. Code,
9844 initialized data and zero-initialized data are allocated consecutively.
9848 Like @option{-sim}, but pass linker options to locate initialized data at
9849 0x40000000 and zero-initialized data at 0x80000000.
9853 @subsection CRX Options
9856 These options are defined specifically for the CRX ports.
9862 Enable the use of multiply-accumulate instructions. Disabled by default.
9866 Push instructions will be used to pass outgoing arguments when functions
9867 are called. Enabled by default.
9870 @node Darwin Options
9871 @subsection Darwin Options
9872 @cindex Darwin options
9874 These options are defined for all architectures running the Darwin operating
9877 FSF GCC on Darwin does not create ``fat'' object files; it will create
9878 an object file for the single architecture that it was built to
9879 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9880 @option{-arch} options are used; it does so by running the compiler or
9881 linker multiple times and joining the results together with
9884 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9885 @samp{i686}) is determined by the flags that specify the ISA
9886 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9887 @option{-force_cpusubtype_ALL} option can be used to override this.
9889 The Darwin tools vary in their behavior when presented with an ISA
9890 mismatch. The assembler, @file{as}, will only permit instructions to
9891 be used that are valid for the subtype of the file it is generating,
9892 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
9893 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9894 and print an error if asked to create a shared library with a less
9895 restrictive subtype than its input files (for instance, trying to put
9896 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9897 for executables, @file{ld}, will quietly give the executable the most
9898 restrictive subtype of any of its input files.
9903 Add the framework directory @var{dir} to the head of the list of
9904 directories to be searched for header files. These directories are
9905 interleaved with those specified by @option{-I} options and are
9906 scanned in a left-to-right order.
9908 A framework directory is a directory with frameworks in it. A
9909 framework is a directory with a @samp{"Headers"} and/or
9910 @samp{"PrivateHeaders"} directory contained directly in it that ends
9911 in @samp{".framework"}. The name of a framework is the name of this
9912 directory excluding the @samp{".framework"}. Headers associated with
9913 the framework are found in one of those two directories, with
9914 @samp{"Headers"} being searched first. A subframework is a framework
9915 directory that is in a framework's @samp{"Frameworks"} directory.
9916 Includes of subframework headers can only appear in a header of a
9917 framework that contains the subframework, or in a sibling subframework
9918 header. Two subframeworks are siblings if they occur in the same
9919 framework. A subframework should not have the same name as a
9920 framework, a warning will be issued if this is violated. Currently a
9921 subframework cannot have subframeworks, in the future, the mechanism
9922 may be extended to support this. The standard frameworks can be found
9923 in @samp{"/System/Library/Frameworks"} and
9924 @samp{"/Library/Frameworks"}. An example include looks like
9925 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9926 the name of the framework and header.h is found in the
9927 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9929 @item -iframework@var{dir}
9931 Like @option{-F} except the directory is a treated as a system
9932 directory. The main difference between this @option{-iframework} and
9933 @option{-F} is that with @option{-iframework} the compiler does not
9934 warn about constructs contained within header files found via
9935 @var{dir}. This option is valid only for the C family of languages.
9939 Emit debugging information for symbols that are used. For STABS
9940 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9941 This is by default ON@.
9945 Emit debugging information for all symbols and types.
9947 @item -mmacosx-version-min=@var{version}
9948 The earliest version of MacOS X that this executable will run on
9949 is @var{version}. Typical values of @var{version} include @code{10.1},
9950 @code{10.2}, and @code{10.3.9}.
9952 If the compiler was built to use the system's headers by default,
9953 then the default for this option is the system version on which the
9954 compiler is running, otherwise the default is to make choices which
9955 are compatible with as many systems and code bases as possible.
9959 Enable kernel development mode. The @option{-mkernel} option sets
9960 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9961 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9962 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9963 applicable. This mode also sets @option{-mno-altivec},
9964 @option{-msoft-float}, @option{-fno-builtin} and
9965 @option{-mlong-branch} for PowerPC targets.
9967 @item -mone-byte-bool
9968 @opindex mone-byte-bool
9969 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9970 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9971 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9972 option has no effect on x86.
9974 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9975 to generate code that is not binary compatible with code generated
9976 without that switch. Using this switch may require recompiling all
9977 other modules in a program, including system libraries. Use this
9978 switch to conform to a non-default data model.
9980 @item -mfix-and-continue
9981 @itemx -ffix-and-continue
9982 @itemx -findirect-data
9983 @opindex mfix-and-continue
9984 @opindex ffix-and-continue
9985 @opindex findirect-data
9986 Generate code suitable for fast turn around development. Needed to
9987 enable gdb to dynamically load @code{.o} files into already running
9988 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9989 are provided for backwards compatibility.
9993 Loads all members of static archive libraries.
9994 See man ld(1) for more information.
9996 @item -arch_errors_fatal
9997 @opindex arch_errors_fatal
9998 Cause the errors having to do with files that have the wrong architecture
10001 @item -bind_at_load
10002 @opindex bind_at_load
10003 Causes the output file to be marked such that the dynamic linker will
10004 bind all undefined references when the file is loaded or launched.
10008 Produce a Mach-o bundle format file.
10009 See man ld(1) for more information.
10011 @item -bundle_loader @var{executable}
10012 @opindex bundle_loader
10013 This option specifies the @var{executable} that will be loading the build
10014 output file being linked. See man ld(1) for more information.
10017 @opindex dynamiclib
10018 When passed this option, GCC will produce a dynamic library instead of
10019 an executable when linking, using the Darwin @file{libtool} command.
10021 @item -force_cpusubtype_ALL
10022 @opindex force_cpusubtype_ALL
10023 This causes GCC's output file to have the @var{ALL} subtype, instead of
10024 one controlled by the @option{-mcpu} or @option{-march} option.
10026 @item -allowable_client @var{client_name}
10027 @itemx -client_name
10028 @itemx -compatibility_version
10029 @itemx -current_version
10031 @itemx -dependency-file
10033 @itemx -dylinker_install_name
10035 @itemx -exported_symbols_list
10037 @itemx -flat_namespace
10038 @itemx -force_flat_namespace
10039 @itemx -headerpad_max_install_names
10042 @itemx -install_name
10043 @itemx -keep_private_externs
10044 @itemx -multi_module
10045 @itemx -multiply_defined
10046 @itemx -multiply_defined_unused
10048 @itemx -no_dead_strip_inits_and_terms
10049 @itemx -nofixprebinding
10050 @itemx -nomultidefs
10052 @itemx -noseglinkedit
10053 @itemx -pagezero_size
10055 @itemx -prebind_all_twolevel_modules
10056 @itemx -private_bundle
10057 @itemx -read_only_relocs
10059 @itemx -sectobjectsymbols
10063 @itemx -sectobjectsymbols
10066 @itemx -segs_read_only_addr
10067 @itemx -segs_read_write_addr
10068 @itemx -seg_addr_table
10069 @itemx -seg_addr_table_filename
10070 @itemx -seglinkedit
10072 @itemx -segs_read_only_addr
10073 @itemx -segs_read_write_addr
10074 @itemx -single_module
10076 @itemx -sub_library
10077 @itemx -sub_umbrella
10078 @itemx -twolevel_namespace
10081 @itemx -unexported_symbols_list
10082 @itemx -weak_reference_mismatches
10083 @itemx -whatsloaded
10084 @opindex allowable_client
10085 @opindex client_name
10086 @opindex compatibility_version
10087 @opindex current_version
10088 @opindex dead_strip
10089 @opindex dependency-file
10090 @opindex dylib_file
10091 @opindex dylinker_install_name
10093 @opindex exported_symbols_list
10095 @opindex flat_namespace
10096 @opindex force_flat_namespace
10097 @opindex headerpad_max_install_names
10098 @opindex image_base
10100 @opindex install_name
10101 @opindex keep_private_externs
10102 @opindex multi_module
10103 @opindex multiply_defined
10104 @opindex multiply_defined_unused
10105 @opindex noall_load
10106 @opindex no_dead_strip_inits_and_terms
10107 @opindex nofixprebinding
10108 @opindex nomultidefs
10110 @opindex noseglinkedit
10111 @opindex pagezero_size
10113 @opindex prebind_all_twolevel_modules
10114 @opindex private_bundle
10115 @opindex read_only_relocs
10117 @opindex sectobjectsymbols
10120 @opindex sectcreate
10121 @opindex sectobjectsymbols
10124 @opindex segs_read_only_addr
10125 @opindex segs_read_write_addr
10126 @opindex seg_addr_table
10127 @opindex seg_addr_table_filename
10128 @opindex seglinkedit
10130 @opindex segs_read_only_addr
10131 @opindex segs_read_write_addr
10132 @opindex single_module
10134 @opindex sub_library
10135 @opindex sub_umbrella
10136 @opindex twolevel_namespace
10139 @opindex unexported_symbols_list
10140 @opindex weak_reference_mismatches
10141 @opindex whatsloaded
10142 These options are passed to the Darwin linker. The Darwin linker man page
10143 describes them in detail.
10146 @node DEC Alpha Options
10147 @subsection DEC Alpha Options
10149 These @samp{-m} options are defined for the DEC Alpha implementations:
10152 @item -mno-soft-float
10153 @itemx -msoft-float
10154 @opindex mno-soft-float
10155 @opindex msoft-float
10156 Use (do not use) the hardware floating-point instructions for
10157 floating-point operations. When @option{-msoft-float} is specified,
10158 functions in @file{libgcc.a} will be used to perform floating-point
10159 operations. Unless they are replaced by routines that emulate the
10160 floating-point operations, or compiled in such a way as to call such
10161 emulations routines, these routines will issue floating-point
10162 operations. If you are compiling for an Alpha without floating-point
10163 operations, you must ensure that the library is built so as not to call
10166 Note that Alpha implementations without floating-point operations are
10167 required to have floating-point registers.
10170 @itemx -mno-fp-regs
10172 @opindex mno-fp-regs
10173 Generate code that uses (does not use) the floating-point register set.
10174 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10175 register set is not used, floating point operands are passed in integer
10176 registers as if they were integers and floating-point results are passed
10177 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10178 so any function with a floating-point argument or return value called by code
10179 compiled with @option{-mno-fp-regs} must also be compiled with that
10182 A typical use of this option is building a kernel that does not use,
10183 and hence need not save and restore, any floating-point registers.
10187 The Alpha architecture implements floating-point hardware optimized for
10188 maximum performance. It is mostly compliant with the IEEE floating
10189 point standard. However, for full compliance, software assistance is
10190 required. This option generates code fully IEEE compliant code
10191 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10192 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10193 defined during compilation. The resulting code is less efficient but is
10194 able to correctly support denormalized numbers and exceptional IEEE
10195 values such as not-a-number and plus/minus infinity. Other Alpha
10196 compilers call this option @option{-ieee_with_no_inexact}.
10198 @item -mieee-with-inexact
10199 @opindex mieee-with-inexact
10200 This is like @option{-mieee} except the generated code also maintains
10201 the IEEE @var{inexact-flag}. Turning on this option causes the
10202 generated code to implement fully-compliant IEEE math. In addition to
10203 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10204 macro. On some Alpha implementations the resulting code may execute
10205 significantly slower than the code generated by default. Since there is
10206 very little code that depends on the @var{inexact-flag}, you should
10207 normally not specify this option. Other Alpha compilers call this
10208 option @option{-ieee_with_inexact}.
10210 @item -mfp-trap-mode=@var{trap-mode}
10211 @opindex mfp-trap-mode
10212 This option controls what floating-point related traps are enabled.
10213 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10214 The trap mode can be set to one of four values:
10218 This is the default (normal) setting. The only traps that are enabled
10219 are the ones that cannot be disabled in software (e.g., division by zero
10223 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10227 Like @samp{u}, but the instructions are marked to be safe for software
10228 completion (see Alpha architecture manual for details).
10231 Like @samp{su}, but inexact traps are enabled as well.
10234 @item -mfp-rounding-mode=@var{rounding-mode}
10235 @opindex mfp-rounding-mode
10236 Selects the IEEE rounding mode. Other Alpha compilers call this option
10237 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10242 Normal IEEE rounding mode. Floating point numbers are rounded towards
10243 the nearest machine number or towards the even machine number in case
10247 Round towards minus infinity.
10250 Chopped rounding mode. Floating point numbers are rounded towards zero.
10253 Dynamic rounding mode. A field in the floating point control register
10254 (@var{fpcr}, see Alpha architecture reference manual) controls the
10255 rounding mode in effect. The C library initializes this register for
10256 rounding towards plus infinity. Thus, unless your program modifies the
10257 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10260 @item -mtrap-precision=@var{trap-precision}
10261 @opindex mtrap-precision
10262 In the Alpha architecture, floating point traps are imprecise. This
10263 means without software assistance it is impossible to recover from a
10264 floating trap and program execution normally needs to be terminated.
10265 GCC can generate code that can assist operating system trap handlers
10266 in determining the exact location that caused a floating point trap.
10267 Depending on the requirements of an application, different levels of
10268 precisions can be selected:
10272 Program precision. This option is the default and means a trap handler
10273 can only identify which program caused a floating point exception.
10276 Function precision. The trap handler can determine the function that
10277 caused a floating point exception.
10280 Instruction precision. The trap handler can determine the exact
10281 instruction that caused a floating point exception.
10284 Other Alpha compilers provide the equivalent options called
10285 @option{-scope_safe} and @option{-resumption_safe}.
10287 @item -mieee-conformant
10288 @opindex mieee-conformant
10289 This option marks the generated code as IEEE conformant. You must not
10290 use this option unless you also specify @option{-mtrap-precision=i} and either
10291 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10292 is to emit the line @samp{.eflag 48} in the function prologue of the
10293 generated assembly file. Under DEC Unix, this has the effect that
10294 IEEE-conformant math library routines will be linked in.
10296 @item -mbuild-constants
10297 @opindex mbuild-constants
10298 Normally GCC examines a 32- or 64-bit integer constant to
10299 see if it can construct it from smaller constants in two or three
10300 instructions. If it cannot, it will output the constant as a literal and
10301 generate code to load it from the data segment at runtime.
10303 Use this option to require GCC to construct @emph{all} integer constants
10304 using code, even if it takes more instructions (the maximum is six).
10306 You would typically use this option to build a shared library dynamic
10307 loader. Itself a shared library, it must relocate itself in memory
10308 before it can find the variables and constants in its own data segment.
10314 Select whether to generate code to be assembled by the vendor-supplied
10315 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10333 Indicate whether GCC should generate code to use the optional BWX,
10334 CIX, FIX and MAX instruction sets. The default is to use the instruction
10335 sets supported by the CPU type specified via @option{-mcpu=} option or that
10336 of the CPU on which GCC was built if none was specified.
10339 @itemx -mfloat-ieee
10340 @opindex mfloat-vax
10341 @opindex mfloat-ieee
10342 Generate code that uses (does not use) VAX F and G floating point
10343 arithmetic instead of IEEE single and double precision.
10345 @item -mexplicit-relocs
10346 @itemx -mno-explicit-relocs
10347 @opindex mexplicit-relocs
10348 @opindex mno-explicit-relocs
10349 Older Alpha assemblers provided no way to generate symbol relocations
10350 except via assembler macros. Use of these macros does not allow
10351 optimal instruction scheduling. GNU binutils as of version 2.12
10352 supports a new syntax that allows the compiler to explicitly mark
10353 which relocations should apply to which instructions. This option
10354 is mostly useful for debugging, as GCC detects the capabilities of
10355 the assembler when it is built and sets the default accordingly.
10358 @itemx -mlarge-data
10359 @opindex msmall-data
10360 @opindex mlarge-data
10361 When @option{-mexplicit-relocs} is in effect, static data is
10362 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10363 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10364 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10365 16-bit relocations off of the @code{$gp} register. This limits the
10366 size of the small data area to 64KB, but allows the variables to be
10367 directly accessed via a single instruction.
10369 The default is @option{-mlarge-data}. With this option the data area
10370 is limited to just below 2GB@. Programs that require more than 2GB of
10371 data must use @code{malloc} or @code{mmap} to allocate the data in the
10372 heap instead of in the program's data segment.
10374 When generating code for shared libraries, @option{-fpic} implies
10375 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10378 @itemx -mlarge-text
10379 @opindex msmall-text
10380 @opindex mlarge-text
10381 When @option{-msmall-text} is used, the compiler assumes that the
10382 code of the entire program (or shared library) fits in 4MB, and is
10383 thus reachable with a branch instruction. When @option{-msmall-data}
10384 is used, the compiler can assume that all local symbols share the
10385 same @code{$gp} value, and thus reduce the number of instructions
10386 required for a function call from 4 to 1.
10388 The default is @option{-mlarge-text}.
10390 @item -mcpu=@var{cpu_type}
10392 Set the instruction set and instruction scheduling parameters for
10393 machine type @var{cpu_type}. You can specify either the @samp{EV}
10394 style name or the corresponding chip number. GCC supports scheduling
10395 parameters for the EV4, EV5 and EV6 family of processors and will
10396 choose the default values for the instruction set from the processor
10397 you specify. If you do not specify a processor type, GCC will default
10398 to the processor on which the compiler was built.
10400 Supported values for @var{cpu_type} are
10406 Schedules as an EV4 and has no instruction set extensions.
10410 Schedules as an EV5 and has no instruction set extensions.
10414 Schedules as an EV5 and supports the BWX extension.
10419 Schedules as an EV5 and supports the BWX and MAX extensions.
10423 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10427 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10430 Native Linux/GNU toolchains also support the value @samp{native},
10431 which selects the best architecture option for the host processor.
10432 @option{-mcpu=native} has no effect if GCC does not recognize
10435 @item -mtune=@var{cpu_type}
10437 Set only the instruction scheduling parameters for machine type
10438 @var{cpu_type}. The instruction set is not changed.
10440 Native Linux/GNU toolchains also support the value @samp{native},
10441 which selects the best architecture option for the host processor.
10442 @option{-mtune=native} has no effect if GCC does not recognize
10445 @item -mmemory-latency=@var{time}
10446 @opindex mmemory-latency
10447 Sets the latency the scheduler should assume for typical memory
10448 references as seen by the application. This number is highly
10449 dependent on the memory access patterns used by the application
10450 and the size of the external cache on the machine.
10452 Valid options for @var{time} are
10456 A decimal number representing clock cycles.
10462 The compiler contains estimates of the number of clock cycles for
10463 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10464 (also called Dcache, Scache, and Bcache), as well as to main memory.
10465 Note that L3 is only valid for EV5.
10470 @node DEC Alpha/VMS Options
10471 @subsection DEC Alpha/VMS Options
10473 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10476 @item -mvms-return-codes
10477 @opindex mvms-return-codes
10478 Return VMS condition codes from main. The default is to return POSIX
10479 style condition (e.g.@: error) codes.
10483 @subsection FR30 Options
10484 @cindex FR30 Options
10486 These options are defined specifically for the FR30 port.
10490 @item -msmall-model
10491 @opindex msmall-model
10492 Use the small address space model. This can produce smaller code, but
10493 it does assume that all symbolic values and addresses will fit into a
10498 Assume that run-time support has been provided and so there is no need
10499 to include the simulator library (@file{libsim.a}) on the linker
10505 @subsection FRV Options
10506 @cindex FRV Options
10512 Only use the first 32 general purpose registers.
10517 Use all 64 general purpose registers.
10522 Use only the first 32 floating point registers.
10527 Use all 64 floating point registers
10530 @opindex mhard-float
10532 Use hardware instructions for floating point operations.
10535 @opindex msoft-float
10537 Use library routines for floating point operations.
10542 Dynamically allocate condition code registers.
10547 Do not try to dynamically allocate condition code registers, only
10548 use @code{icc0} and @code{fcc0}.
10553 Change ABI to use double word insns.
10558 Do not use double word instructions.
10563 Use floating point double instructions.
10566 @opindex mno-double
10568 Do not use floating point double instructions.
10573 Use media instructions.
10578 Do not use media instructions.
10583 Use multiply and add/subtract instructions.
10586 @opindex mno-muladd
10588 Do not use multiply and add/subtract instructions.
10593 Select the FDPIC ABI, that uses function descriptors to represent
10594 pointers to functions. Without any PIC/PIE-related options, it
10595 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10596 assumes GOT entries and small data are within a 12-bit range from the
10597 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10598 are computed with 32 bits.
10599 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10602 @opindex minline-plt
10604 Enable inlining of PLT entries in function calls to functions that are
10605 not known to bind locally. It has no effect without @option{-mfdpic}.
10606 It's enabled by default if optimizing for speed and compiling for
10607 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10608 optimization option such as @option{-O3} or above is present in the
10614 Assume a large TLS segment when generating thread-local code.
10619 Do not assume a large TLS segment when generating thread-local code.
10624 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10625 that is known to be in read-only sections. It's enabled by default,
10626 except for @option{-fpic} or @option{-fpie}: even though it may help
10627 make the global offset table smaller, it trades 1 instruction for 4.
10628 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10629 one of which may be shared by multiple symbols, and it avoids the need
10630 for a GOT entry for the referenced symbol, so it's more likely to be a
10631 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10633 @item -multilib-library-pic
10634 @opindex multilib-library-pic
10636 Link with the (library, not FD) pic libraries. It's implied by
10637 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10638 @option{-fpic} without @option{-mfdpic}. You should never have to use
10642 @opindex mlinked-fp
10644 Follow the EABI requirement of always creating a frame pointer whenever
10645 a stack frame is allocated. This option is enabled by default and can
10646 be disabled with @option{-mno-linked-fp}.
10649 @opindex mlong-calls
10651 Use indirect addressing to call functions outside the current
10652 compilation unit. This allows the functions to be placed anywhere
10653 within the 32-bit address space.
10655 @item -malign-labels
10656 @opindex malign-labels
10658 Try to align labels to an 8-byte boundary by inserting nops into the
10659 previous packet. This option only has an effect when VLIW packing
10660 is enabled. It doesn't create new packets; it merely adds nops to
10663 @item -mlibrary-pic
10664 @opindex mlibrary-pic
10666 Generate position-independent EABI code.
10671 Use only the first four media accumulator registers.
10676 Use all eight media accumulator registers.
10681 Pack VLIW instructions.
10686 Do not pack VLIW instructions.
10689 @opindex mno-eflags
10691 Do not mark ABI switches in e_flags.
10694 @opindex mcond-move
10696 Enable the use of conditional-move instructions (default).
10698 This switch is mainly for debugging the compiler and will likely be removed
10699 in a future version.
10701 @item -mno-cond-move
10702 @opindex mno-cond-move
10704 Disable the use of conditional-move instructions.
10706 This switch is mainly for debugging the compiler and will likely be removed
10707 in a future version.
10712 Enable the use of conditional set instructions (default).
10714 This switch is mainly for debugging the compiler and will likely be removed
10715 in a future version.
10720 Disable the use of conditional set instructions.
10722 This switch is mainly for debugging the compiler and will likely be removed
10723 in a future version.
10726 @opindex mcond-exec
10728 Enable the use of conditional execution (default).
10730 This switch is mainly for debugging the compiler and will likely be removed
10731 in a future version.
10733 @item -mno-cond-exec
10734 @opindex mno-cond-exec
10736 Disable the use of conditional execution.
10738 This switch is mainly for debugging the compiler and will likely be removed
10739 in a future version.
10741 @item -mvliw-branch
10742 @opindex mvliw-branch
10744 Run a pass to pack branches into VLIW instructions (default).
10746 This switch is mainly for debugging the compiler and will likely be removed
10747 in a future version.
10749 @item -mno-vliw-branch
10750 @opindex mno-vliw-branch
10752 Do not run a pass to pack branches into VLIW instructions.
10754 This switch is mainly for debugging the compiler and will likely be removed
10755 in a future version.
10757 @item -mmulti-cond-exec
10758 @opindex mmulti-cond-exec
10760 Enable optimization of @code{&&} and @code{||} in conditional execution
10763 This switch is mainly for debugging the compiler and will likely be removed
10764 in a future version.
10766 @item -mno-multi-cond-exec
10767 @opindex mno-multi-cond-exec
10769 Disable optimization of @code{&&} and @code{||} in conditional execution.
10771 This switch is mainly for debugging the compiler and will likely be removed
10772 in a future version.
10774 @item -mnested-cond-exec
10775 @opindex mnested-cond-exec
10777 Enable nested conditional execution optimizations (default).
10779 This switch is mainly for debugging the compiler and will likely be removed
10780 in a future version.
10782 @item -mno-nested-cond-exec
10783 @opindex mno-nested-cond-exec
10785 Disable nested conditional execution optimizations.
10787 This switch is mainly for debugging the compiler and will likely be removed
10788 in a future version.
10790 @item -moptimize-membar
10791 @opindex moptimize-membar
10793 This switch removes redundant @code{membar} instructions from the
10794 compiler generated code. It is enabled by default.
10796 @item -mno-optimize-membar
10797 @opindex mno-optimize-membar
10799 This switch disables the automatic removal of redundant @code{membar}
10800 instructions from the generated code.
10802 @item -mtomcat-stats
10803 @opindex mtomcat-stats
10805 Cause gas to print out tomcat statistics.
10807 @item -mcpu=@var{cpu}
10810 Select the processor type for which to generate code. Possible values are
10811 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10812 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10816 @node GNU/Linux Options
10817 @subsection GNU/Linux Options
10819 These @samp{-m} options are defined for GNU/Linux targets:
10824 Use the GNU C library instead of uClibc. This is the default except
10825 on @samp{*-*-linux-*uclibc*} targets.
10829 Use uClibc instead of the GNU C library. This is the default on
10830 @samp{*-*-linux-*uclibc*} targets.
10833 @node H8/300 Options
10834 @subsection H8/300 Options
10836 These @samp{-m} options are defined for the H8/300 implementations:
10841 Shorten some address references at link time, when possible; uses the
10842 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10843 ld, Using ld}, for a fuller description.
10847 Generate code for the H8/300H@.
10851 Generate code for the H8S@.
10855 Generate code for the H8S and H8/300H in the normal mode. This switch
10856 must be used either with @option{-mh} or @option{-ms}.
10860 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10864 Make @code{int} data 32 bits by default.
10867 @opindex malign-300
10868 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10869 The default for the H8/300H and H8S is to align longs and floats on 4
10871 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10872 This option has no effect on the H8/300.
10876 @subsection HPPA Options
10877 @cindex HPPA Options
10879 These @samp{-m} options are defined for the HPPA family of computers:
10882 @item -march=@var{architecture-type}
10884 Generate code for the specified architecture. The choices for
10885 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10886 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10887 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10888 architecture option for your machine. Code compiled for lower numbered
10889 architectures will run on higher numbered architectures, but not the
10892 @item -mpa-risc-1-0
10893 @itemx -mpa-risc-1-1
10894 @itemx -mpa-risc-2-0
10895 @opindex mpa-risc-1-0
10896 @opindex mpa-risc-1-1
10897 @opindex mpa-risc-2-0
10898 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10901 @opindex mbig-switch
10902 Generate code suitable for big switch tables. Use this option only if
10903 the assembler/linker complain about out of range branches within a switch
10906 @item -mjump-in-delay
10907 @opindex mjump-in-delay
10908 Fill delay slots of function calls with unconditional jump instructions
10909 by modifying the return pointer for the function call to be the target
10910 of the conditional jump.
10912 @item -mdisable-fpregs
10913 @opindex mdisable-fpregs
10914 Prevent floating point registers from being used in any manner. This is
10915 necessary for compiling kernels which perform lazy context switching of
10916 floating point registers. If you use this option and attempt to perform
10917 floating point operations, the compiler will abort.
10919 @item -mdisable-indexing
10920 @opindex mdisable-indexing
10921 Prevent the compiler from using indexing address modes. This avoids some
10922 rather obscure problems when compiling MIG generated code under MACH@.
10924 @item -mno-space-regs
10925 @opindex mno-space-regs
10926 Generate code that assumes the target has no space registers. This allows
10927 GCC to generate faster indirect calls and use unscaled index address modes.
10929 Such code is suitable for level 0 PA systems and kernels.
10931 @item -mfast-indirect-calls
10932 @opindex mfast-indirect-calls
10933 Generate code that assumes calls never cross space boundaries. This
10934 allows GCC to emit code which performs faster indirect calls.
10936 This option will not work in the presence of shared libraries or nested
10939 @item -mfixed-range=@var{register-range}
10940 @opindex mfixed-range
10941 Generate code treating the given register range as fixed registers.
10942 A fixed register is one that the register allocator can not use. This is
10943 useful when compiling kernel code. A register range is specified as
10944 two registers separated by a dash. Multiple register ranges can be
10945 specified separated by a comma.
10947 @item -mlong-load-store
10948 @opindex mlong-load-store
10949 Generate 3-instruction load and store sequences as sometimes required by
10950 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10953 @item -mportable-runtime
10954 @opindex mportable-runtime
10955 Use the portable calling conventions proposed by HP for ELF systems.
10959 Enable the use of assembler directives only GAS understands.
10961 @item -mschedule=@var{cpu-type}
10963 Schedule code according to the constraints for the machine type
10964 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10965 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10966 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10967 proper scheduling option for your machine. The default scheduling is
10971 @opindex mlinker-opt
10972 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10973 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10974 linkers in which they give bogus error messages when linking some programs.
10977 @opindex msoft-float
10978 Generate output containing library calls for floating point.
10979 @strong{Warning:} the requisite libraries are not available for all HPPA
10980 targets. Normally the facilities of the machine's usual C compiler are
10981 used, but this cannot be done directly in cross-compilation. You must make
10982 your own arrangements to provide suitable library functions for
10985 @option{-msoft-float} changes the calling convention in the output file;
10986 therefore, it is only useful if you compile @emph{all} of a program with
10987 this option. In particular, you need to compile @file{libgcc.a}, the
10988 library that comes with GCC, with @option{-msoft-float} in order for
10993 Generate the predefine, @code{_SIO}, for server IO@. The default is
10994 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10995 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10996 options are available under HP-UX and HI-UX@.
11000 Use GNU ld specific options. This passes @option{-shared} to ld when
11001 building a shared library. It is the default when GCC is configured,
11002 explicitly or implicitly, with the GNU linker. This option does not
11003 have any affect on which ld is called, it only changes what parameters
11004 are passed to that ld. The ld that is called is determined by the
11005 @option{--with-ld} configure option, GCC's program search path, and
11006 finally by the user's @env{PATH}. The linker used by GCC can be printed
11007 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11008 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11012 Use HP ld specific options. This passes @option{-b} to ld when building
11013 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11014 links. It is the default when GCC is configured, explicitly or
11015 implicitly, with the HP linker. This option does not have any affect on
11016 which ld is called, it only changes what parameters are passed to that
11017 ld. The ld that is called is determined by the @option{--with-ld}
11018 configure option, GCC's program search path, and finally by the user's
11019 @env{PATH}. The linker used by GCC can be printed using @samp{which
11020 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11021 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11024 @opindex mno-long-calls
11025 Generate code that uses long call sequences. This ensures that a call
11026 is always able to reach linker generated stubs. The default is to generate
11027 long calls only when the distance from the call site to the beginning
11028 of the function or translation unit, as the case may be, exceeds a
11029 predefined limit set by the branch type being used. The limits for
11030 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11031 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11034 Distances are measured from the beginning of functions when using the
11035 @option{-ffunction-sections} option, or when using the @option{-mgas}
11036 and @option{-mno-portable-runtime} options together under HP-UX with
11039 It is normally not desirable to use this option as it will degrade
11040 performance. However, it may be useful in large applications,
11041 particularly when partial linking is used to build the application.
11043 The types of long calls used depends on the capabilities of the
11044 assembler and linker, and the type of code being generated. The
11045 impact on systems that support long absolute calls, and long pic
11046 symbol-difference or pc-relative calls should be relatively small.
11047 However, an indirect call is used on 32-bit ELF systems in pic code
11048 and it is quite long.
11050 @item -munix=@var{unix-std}
11052 Generate compiler predefines and select a startfile for the specified
11053 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11054 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11055 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11056 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11057 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11060 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11061 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11062 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11063 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11064 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11065 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11067 It is @emph{important} to note that this option changes the interfaces
11068 for various library routines. It also affects the operational behavior
11069 of the C library. Thus, @emph{extreme} care is needed in using this
11072 Library code that is intended to operate with more than one UNIX
11073 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11074 as appropriate. Most GNU software doesn't provide this capability.
11078 Suppress the generation of link options to search libdld.sl when the
11079 @option{-static} option is specified on HP-UX 10 and later.
11083 The HP-UX implementation of setlocale in libc has a dependency on
11084 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11085 when the @option{-static} option is specified, special link options
11086 are needed to resolve this dependency.
11088 On HP-UX 10 and later, the GCC driver adds the necessary options to
11089 link with libdld.sl when the @option{-static} option is specified.
11090 This causes the resulting binary to be dynamic. On the 64-bit port,
11091 the linkers generate dynamic binaries by default in any case. The
11092 @option{-nolibdld} option can be used to prevent the GCC driver from
11093 adding these link options.
11097 Add support for multithreading with the @dfn{dce thread} library
11098 under HP-UX@. This option sets flags for both the preprocessor and
11102 @node i386 and x86-64 Options
11103 @subsection Intel 386 and AMD x86-64 Options
11104 @cindex i386 Options
11105 @cindex x86-64 Options
11106 @cindex Intel 386 Options
11107 @cindex AMD x86-64 Options
11109 These @samp{-m} options are defined for the i386 and x86-64 family of
11113 @item -mtune=@var{cpu-type}
11115 Tune to @var{cpu-type} everything applicable about the generated code, except
11116 for the ABI and the set of available instructions. The choices for
11117 @var{cpu-type} are:
11120 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11121 If you know the CPU on which your code will run, then you should use
11122 the corresponding @option{-mtune} option instead of
11123 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11124 of your application will have, then you should use this option.
11126 As new processors are deployed in the marketplace, the behavior of this
11127 option will change. Therefore, if you upgrade to a newer version of
11128 GCC, the code generated option will change to reflect the processors
11129 that were most common when that version of GCC was released.
11131 There is no @option{-march=generic} option because @option{-march}
11132 indicates the instruction set the compiler can use, and there is no
11133 generic instruction set applicable to all processors. In contrast,
11134 @option{-mtune} indicates the processor (or, in this case, collection of
11135 processors) for which the code is optimized.
11137 This selects the CPU to tune for at compilation time by determining
11138 the processor type of the compiling machine. Using @option{-mtune=native}
11139 will produce code optimized for the local machine under the constraints
11140 of the selected instruction set. Using @option{-march=native} will
11141 enable all instruction subsets supported by the local machine (hence
11142 the result might not run on different machines).
11144 Original Intel's i386 CPU@.
11146 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11147 @item i586, pentium
11148 Intel Pentium CPU with no MMX support.
11150 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11152 Intel PentiumPro CPU@.
11154 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11155 instruction set will be used, so the code will run on all i686 family chips.
11157 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11158 @item pentium3, pentium3m
11159 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11162 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11163 support. Used by Centrino notebooks.
11164 @item pentium4, pentium4m
11165 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11167 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11170 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11171 SSE2 and SSE3 instruction set support.
11173 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11174 instruction set support.
11176 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11177 instruction set support.
11179 AMD K6 CPU with MMX instruction set support.
11181 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11182 @item athlon, athlon-tbird
11183 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11185 @item athlon-4, athlon-xp, athlon-mp
11186 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11187 instruction set support.
11188 @item k8, opteron, athlon64, athlon-fx
11189 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11190 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11191 @item k8-sse3, opteron-sse3, athlon64-sse3
11192 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11193 @item amdfam10, barcelona
11194 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11195 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11196 instruction set extensions.)
11198 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11201 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11202 instruction set support.
11204 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11205 implemented for this chip.)
11207 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11208 implemented for this chip.)
11210 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11213 While picking a specific @var{cpu-type} will schedule things appropriately
11214 for that particular chip, the compiler will not generate any code that
11215 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11218 @item -march=@var{cpu-type}
11220 Generate instructions for the machine type @var{cpu-type}. The choices
11221 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11222 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11224 @item -mcpu=@var{cpu-type}
11226 A deprecated synonym for @option{-mtune}.
11228 @item -mfpmath=@var{unit}
11230 Generate floating point arithmetics for selected unit @var{unit}. The choices
11231 for @var{unit} are:
11235 Use the standard 387 floating point coprocessor present majority of chips and
11236 emulated otherwise. Code compiled with this option will run almost everywhere.
11237 The temporary results are computed in 80bit precision instead of precision
11238 specified by the type resulting in slightly different results compared to most
11239 of other chips. See @option{-ffloat-store} for more detailed description.
11241 This is the default choice for i386 compiler.
11244 Use scalar floating point instructions present in the SSE instruction set.
11245 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11246 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11247 instruction set supports only single precision arithmetics, thus the double and
11248 extended precision arithmetics is still done using 387. Later version, present
11249 only in Pentium4 and the future AMD x86-64 chips supports double precision
11252 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11253 or @option{-msse2} switches to enable SSE extensions and make this option
11254 effective. For the x86-64 compiler, these extensions are enabled by default.
11256 The resulting code should be considerably faster in the majority of cases and avoid
11257 the numerical instability problems of 387 code, but may break some existing
11258 code that expects temporaries to be 80bit.
11260 This is the default choice for the x86-64 compiler.
11265 Attempt to utilize both instruction sets at once. This effectively double the
11266 amount of available registers and on chips with separate execution units for
11267 387 and SSE the execution resources too. Use this option with care, as it is
11268 still experimental, because the GCC register allocator does not model separate
11269 functional units well resulting in instable performance.
11272 @item -masm=@var{dialect}
11273 @opindex masm=@var{dialect}
11274 Output asm instructions using selected @var{dialect}. Supported
11275 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11276 not support @samp{intel}.
11279 @itemx -mno-ieee-fp
11281 @opindex mno-ieee-fp
11282 Control whether or not the compiler uses IEEE floating point
11283 comparisons. These handle correctly the case where the result of a
11284 comparison is unordered.
11287 @opindex msoft-float
11288 Generate output containing library calls for floating point.
11289 @strong{Warning:} the requisite libraries are not part of GCC@.
11290 Normally the facilities of the machine's usual C compiler are used, but
11291 this can't be done directly in cross-compilation. You must make your
11292 own arrangements to provide suitable library functions for
11295 On machines where a function returns floating point results in the 80387
11296 register stack, some floating point opcodes may be emitted even if
11297 @option{-msoft-float} is used.
11299 @item -mno-fp-ret-in-387
11300 @opindex mno-fp-ret-in-387
11301 Do not use the FPU registers for return values of functions.
11303 The usual calling convention has functions return values of types
11304 @code{float} and @code{double} in an FPU register, even if there
11305 is no FPU@. The idea is that the operating system should emulate
11308 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11309 in ordinary CPU registers instead.
11311 @item -mno-fancy-math-387
11312 @opindex mno-fancy-math-387
11313 Some 387 emulators do not support the @code{sin}, @code{cos} and
11314 @code{sqrt} instructions for the 387. Specify this option to avoid
11315 generating those instructions. This option is the default on FreeBSD,
11316 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11317 indicates that the target cpu will always have an FPU and so the
11318 instruction will not need emulation. As of revision 2.6.1, these
11319 instructions are not generated unless you also use the
11320 @option{-funsafe-math-optimizations} switch.
11322 @item -malign-double
11323 @itemx -mno-align-double
11324 @opindex malign-double
11325 @opindex mno-align-double
11326 Control whether GCC aligns @code{double}, @code{long double}, and
11327 @code{long long} variables on a two word boundary or a one word
11328 boundary. Aligning @code{double} variables on a two word boundary will
11329 produce code that runs somewhat faster on a @samp{Pentium} at the
11330 expense of more memory.
11332 On x86-64, @option{-malign-double} is enabled by default.
11334 @strong{Warning:} if you use the @option{-malign-double} switch,
11335 structures containing the above types will be aligned differently than
11336 the published application binary interface specifications for the 386
11337 and will not be binary compatible with structures in code compiled
11338 without that switch.
11340 @item -m96bit-long-double
11341 @itemx -m128bit-long-double
11342 @opindex m96bit-long-double
11343 @opindex m128bit-long-double
11344 These switches control the size of @code{long double} type. The i386
11345 application binary interface specifies the size to be 96 bits,
11346 so @option{-m96bit-long-double} is the default in 32 bit mode.
11348 Modern architectures (Pentium and newer) would prefer @code{long double}
11349 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11350 conforming to the ABI, this would not be possible. So specifying a
11351 @option{-m128bit-long-double} will align @code{long double}
11352 to a 16 byte boundary by padding the @code{long double} with an additional
11355 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11356 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11358 Notice that neither of these options enable any extra precision over the x87
11359 standard of 80 bits for a @code{long double}.
11361 @strong{Warning:} if you override the default value for your target ABI, the
11362 structures and arrays containing @code{long double} variables will change
11363 their size as well as function calling convention for function taking
11364 @code{long double} will be modified. Hence they will not be binary
11365 compatible with arrays or structures in code compiled without that switch.
11367 @item -mlarge-data-threshold=@var{number}
11368 @opindex mlarge-data-threshold=@var{number}
11369 When @option{-mcmodel=medium} is specified, the data greater than
11370 @var{threshold} are placed in large data section. This value must be the
11371 same across all object linked into the binary and defaults to 65535.
11375 Use a different function-calling convention, in which functions that
11376 take a fixed number of arguments return with the @code{ret} @var{num}
11377 instruction, which pops their arguments while returning. This saves one
11378 instruction in the caller since there is no need to pop the arguments
11381 You can specify that an individual function is called with this calling
11382 sequence with the function attribute @samp{stdcall}. You can also
11383 override the @option{-mrtd} option by using the function attribute
11384 @samp{cdecl}. @xref{Function Attributes}.
11386 @strong{Warning:} this calling convention is incompatible with the one
11387 normally used on Unix, so you cannot use it if you need to call
11388 libraries compiled with the Unix compiler.
11390 Also, you must provide function prototypes for all functions that
11391 take variable numbers of arguments (including @code{printf});
11392 otherwise incorrect code will be generated for calls to those
11395 In addition, seriously incorrect code will result if you call a
11396 function with too many arguments. (Normally, extra arguments are
11397 harmlessly ignored.)
11399 @item -mregparm=@var{num}
11401 Control how many registers are used to pass integer arguments. By
11402 default, no registers are used to pass arguments, and at most 3
11403 registers can be used. You can control this behavior for a specific
11404 function by using the function attribute @samp{regparm}.
11405 @xref{Function Attributes}.
11407 @strong{Warning:} if you use this switch, and
11408 @var{num} is nonzero, then you must build all modules with the same
11409 value, including any libraries. This includes the system libraries and
11413 @opindex msseregparm
11414 Use SSE register passing conventions for float and double arguments
11415 and return values. You can control this behavior for a specific
11416 function by using the function attribute @samp{sseregparm}.
11417 @xref{Function Attributes}.
11419 @strong{Warning:} if you use this switch then you must build all
11420 modules with the same value, including any libraries. This includes
11421 the system libraries and startup modules.
11430 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11431 is specified, the significands of results of floating-point operations are
11432 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11433 significands of results of floating-point operations to 53 bits (double
11434 precision) and @option{-mpc80} rounds the significands of results of
11435 floating-point operations to 64 bits (extended double precision), which is
11436 the default. When this option is used, floating-point operations in higher
11437 precisions are not available to the programmer without setting the FPU
11438 control word explicitly.
11440 Setting the rounding of floating-point operations to less than the default
11441 80 bits can speed some programs by 2% or more. Note that some mathematical
11442 libraries assume that extended precision (80 bit) floating-point operations
11443 are enabled by default; routines in such libraries could suffer significant
11444 loss of accuracy, typically through so-called "catastrophic cancellation",
11445 when this option is used to set the precision to less than extended precision.
11447 @item -mstackrealign
11448 @opindex mstackrealign
11449 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11450 option will generate an alternate prologue and epilogue that realigns the
11451 runtime stack if necessary. This supports mixing legacy codes that keep
11452 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11453 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11454 applicable to individual functions.
11456 @item -mpreferred-stack-boundary=@var{num}
11457 @opindex mpreferred-stack-boundary
11458 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11459 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11460 the default is 4 (16 bytes or 128 bits).
11462 @item -mincoming-stack-boundary=@var{num}
11463 @opindex mincoming-stack-boundary
11464 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11465 boundary. If @option{-mincoming-stack-boundary} is not specified,
11466 the one specified by @option{-mpreferred-stack-boundary} will be used.
11468 On Pentium and PentiumPro, @code{double} and @code{long double} values
11469 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11470 suffer significant run time performance penalties. On Pentium III, the
11471 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11472 properly if it is not 16 byte aligned.
11474 To ensure proper alignment of this values on the stack, the stack boundary
11475 must be as aligned as that required by any value stored on the stack.
11476 Further, every function must be generated such that it keeps the stack
11477 aligned. Thus calling a function compiled with a higher preferred
11478 stack boundary from a function compiled with a lower preferred stack
11479 boundary will most likely misalign the stack. It is recommended that
11480 libraries that use callbacks always use the default setting.
11482 This extra alignment does consume extra stack space, and generally
11483 increases code size. Code that is sensitive to stack space usage, such
11484 as embedded systems and operating system kernels, may want to reduce the
11485 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11525 These switches enable or disable the use of instructions in the MMX,
11526 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
11527 3DNow!@: extended instruction sets.
11528 These extensions are also available as built-in functions: see
11529 @ref{X86 Built-in Functions}, for details of the functions enabled and
11530 disabled by these switches.
11532 To have SSE/SSE2 instructions generated automatically from floating-point
11533 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11535 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11536 generates new AVX instructions or AVX equivalence for all SSEx instructions
11539 These options will enable GCC to use these extended instructions in
11540 generated code, even without @option{-mfpmath=sse}. Applications which
11541 perform runtime CPU detection must compile separate files for each
11542 supported architecture, using the appropriate flags. In particular,
11543 the file containing the CPU detection code should be compiled without
11548 This option instructs GCC to emit a @code{cld} instruction in the prologue
11549 of functions that use string instructions. String instructions depend on
11550 the DF flag to select between autoincrement or autodecrement mode. While the
11551 ABI specifies the DF flag to be cleared on function entry, some operating
11552 systems violate this specification by not clearing the DF flag in their
11553 exception dispatchers. The exception handler can be invoked with the DF flag
11554 set which leads to wrong direction mode, when string instructions are used.
11555 This option can be enabled by default on 32-bit x86 targets by configuring
11556 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11557 instructions can be suppressed with the @option{-mno-cld} compiler option
11562 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11563 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11564 data types. This is useful for high resolution counters that could be updated
11565 by multiple processors (or cores). This instruction is generated as part of
11566 atomic built-in functions: see @ref{Atomic Builtins} for details.
11570 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11571 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11572 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11573 SAHF are load and store instructions, respectively, for certain status flags.
11574 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11575 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11579 This option will enable GCC to use movbe instruction to implement
11580 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
11584 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
11585 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
11586 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
11590 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11591 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
11592 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11593 variants) for single precision floating point arguments. These instructions
11594 are generated only when @option{-funsafe-math-optimizations} is enabled
11595 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11596 Note that while the throughput of the sequence is higher than the throughput
11597 of the non-reciprocal instruction, the precision of the sequence can be
11598 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11600 @item -mveclibabi=@var{type}
11601 @opindex mveclibabi
11602 Specifies the ABI type to use for vectorizing intrinsics using an
11603 external library. Supported types are @code{svml} for the Intel short
11604 vector math library and @code{acml} for the AMD math core library style
11605 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11606 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11607 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11608 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11609 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11610 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11611 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11612 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11613 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11614 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11615 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11616 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11617 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11618 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11619 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11620 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11621 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11622 compatible library will have to be specified at link time.
11624 @item -mabi=@var{name}
11626 Generate code for the specified calling convention. Permissible values
11627 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
11628 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
11629 ABI when targeting Windows. On all other systems, the default is the
11630 SYSV ABI. You can control this behavior for a specific function by
11631 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
11632 @xref{Function Attributes}.
11635 @itemx -mno-push-args
11636 @opindex mpush-args
11637 @opindex mno-push-args
11638 Use PUSH operations to store outgoing parameters. This method is shorter
11639 and usually equally fast as method using SUB/MOV operations and is enabled
11640 by default. In some cases disabling it may improve performance because of
11641 improved scheduling and reduced dependencies.
11643 @item -maccumulate-outgoing-args
11644 @opindex maccumulate-outgoing-args
11645 If enabled, the maximum amount of space required for outgoing arguments will be
11646 computed in the function prologue. This is faster on most modern CPUs
11647 because of reduced dependencies, improved scheduling and reduced stack usage
11648 when preferred stack boundary is not equal to 2. The drawback is a notable
11649 increase in code size. This switch implies @option{-mno-push-args}.
11653 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11654 on thread-safe exception handling must compile and link all code with the
11655 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11656 @option{-D_MT}; when linking, it links in a special thread helper library
11657 @option{-lmingwthrd} which cleans up per thread exception handling data.
11659 @item -mno-align-stringops
11660 @opindex mno-align-stringops
11661 Do not align destination of inlined string operations. This switch reduces
11662 code size and improves performance in case the destination is already aligned,
11663 but GCC doesn't know about it.
11665 @item -minline-all-stringops
11666 @opindex minline-all-stringops
11667 By default GCC inlines string operations only when destination is known to be
11668 aligned at least to 4 byte boundary. This enables more inlining, increase code
11669 size, but may improve performance of code that depends on fast memcpy, strlen
11670 and memset for short lengths.
11672 @item -minline-stringops-dynamically
11673 @opindex minline-stringops-dynamically
11674 For string operation of unknown size, inline runtime checks so for small
11675 blocks inline code is used, while for large blocks library call is used.
11677 @item -mstringop-strategy=@var{alg}
11678 @opindex mstringop-strategy=@var{alg}
11679 Overwrite internal decision heuristic about particular algorithm to inline
11680 string operation with. The allowed values are @code{rep_byte},
11681 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11682 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11683 expanding inline loop, @code{libcall} for always expanding library call.
11685 @item -momit-leaf-frame-pointer
11686 @opindex momit-leaf-frame-pointer
11687 Don't keep the frame pointer in a register for leaf functions. This
11688 avoids the instructions to save, set up and restore frame pointers and
11689 makes an extra register available in leaf functions. The option
11690 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11691 which might make debugging harder.
11693 @item -mtls-direct-seg-refs
11694 @itemx -mno-tls-direct-seg-refs
11695 @opindex mtls-direct-seg-refs
11696 Controls whether TLS variables may be accessed with offsets from the
11697 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11698 or whether the thread base pointer must be added. Whether or not this
11699 is legal depends on the operating system, and whether it maps the
11700 segment to cover the entire TLS area.
11702 For systems that use GNU libc, the default is on.
11705 @itemx -mno-fused-madd
11706 @opindex mfused-madd
11707 Enable automatic generation of fused floating point multiply-add instructions
11708 if the ISA supports such instructions. The -mfused-madd option is on by
11709 default. The fused multiply-add instructions have a different
11710 rounding behavior compared to executing a multiply followed by an add.
11713 @itemx -mno-sse2avx
11715 Specify that the assembler should encode SSE instructions with VEX
11716 prefix. The option @option{-mavx} turns this on by default.
11719 These @samp{-m} switches are supported in addition to the above
11720 on AMD x86-64 processors in 64-bit environments.
11727 Generate code for a 32-bit or 64-bit environment.
11728 The 32-bit environment sets int, long and pointer to 32 bits and
11729 generates code that runs on any i386 system.
11730 The 64-bit environment sets int to 32 bits and long and pointer
11731 to 64 bits and generates code for AMD's x86-64 architecture. For
11732 darwin only the -m64 option turns off the @option{-fno-pic} and
11733 @option{-mdynamic-no-pic} options.
11735 @item -mno-red-zone
11736 @opindex mno-red-zone
11737 Do not use a so called red zone for x86-64 code. The red zone is mandated
11738 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11739 stack pointer that will not be modified by signal or interrupt handlers
11740 and therefore can be used for temporary data without adjusting the stack
11741 pointer. The flag @option{-mno-red-zone} disables this red zone.
11743 @item -mcmodel=small
11744 @opindex mcmodel=small
11745 Generate code for the small code model: the program and its symbols must
11746 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11747 Programs can be statically or dynamically linked. This is the default
11750 @item -mcmodel=kernel
11751 @opindex mcmodel=kernel
11752 Generate code for the kernel code model. The kernel runs in the
11753 negative 2 GB of the address space.
11754 This model has to be used for Linux kernel code.
11756 @item -mcmodel=medium
11757 @opindex mcmodel=medium
11758 Generate code for the medium model: The program is linked in the lower 2
11759 GB of the address space. Small symbols are also placed there. Symbols
11760 with sizes larger than @option{-mlarge-data-threshold} are put into
11761 large data or bss sections and can be located above 2GB. Programs can
11762 be statically or dynamically linked.
11764 @item -mcmodel=large
11765 @opindex mcmodel=large
11766 Generate code for the large model: This model makes no assumptions
11767 about addresses and sizes of sections.
11770 @node IA-64 Options
11771 @subsection IA-64 Options
11772 @cindex IA-64 Options
11774 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11778 @opindex mbig-endian
11779 Generate code for a big endian target. This is the default for HP-UX@.
11781 @item -mlittle-endian
11782 @opindex mlittle-endian
11783 Generate code for a little endian target. This is the default for AIX5
11789 @opindex mno-gnu-as
11790 Generate (or don't) code for the GNU assembler. This is the default.
11791 @c Also, this is the default if the configure option @option{--with-gnu-as}
11797 @opindex mno-gnu-ld
11798 Generate (or don't) code for the GNU linker. This is the default.
11799 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11804 Generate code that does not use a global pointer register. The result
11805 is not position independent code, and violates the IA-64 ABI@.
11807 @item -mvolatile-asm-stop
11808 @itemx -mno-volatile-asm-stop
11809 @opindex mvolatile-asm-stop
11810 @opindex mno-volatile-asm-stop
11811 Generate (or don't) a stop bit immediately before and after volatile asm
11814 @item -mregister-names
11815 @itemx -mno-register-names
11816 @opindex mregister-names
11817 @opindex mno-register-names
11818 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11819 the stacked registers. This may make assembler output more readable.
11825 Disable (or enable) optimizations that use the small data section. This may
11826 be useful for working around optimizer bugs.
11828 @item -mconstant-gp
11829 @opindex mconstant-gp
11830 Generate code that uses a single constant global pointer value. This is
11831 useful when compiling kernel code.
11835 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11836 This is useful when compiling firmware code.
11838 @item -minline-float-divide-min-latency
11839 @opindex minline-float-divide-min-latency
11840 Generate code for inline divides of floating point values
11841 using the minimum latency algorithm.
11843 @item -minline-float-divide-max-throughput
11844 @opindex minline-float-divide-max-throughput
11845 Generate code for inline divides of floating point values
11846 using the maximum throughput algorithm.
11848 @item -mno-inline-float-divide
11849 @opindex mno-inline-float-divide
11850 Do not generate inline code for divides of floating point values.
11852 @item -minline-int-divide-min-latency
11853 @opindex minline-int-divide-min-latency
11854 Generate code for inline divides of integer values
11855 using the minimum latency algorithm.
11857 @item -minline-int-divide-max-throughput
11858 @opindex minline-int-divide-max-throughput
11859 Generate code for inline divides of integer values
11860 using the maximum throughput algorithm.
11862 @item -mno-inline-int-divide
11863 @opindex mno-inline-int-divide
11864 Do not generate inline code for divides of integer values.
11866 @item -minline-sqrt-min-latency
11867 @opindex minline-sqrt-min-latency
11868 Generate code for inline square roots
11869 using the minimum latency algorithm.
11871 @item -minline-sqrt-max-throughput
11872 @opindex minline-sqrt-max-throughput
11873 Generate code for inline square roots
11874 using the maximum throughput algorithm.
11876 @item -mno-inline-sqrt
11877 @opindex mno-inline-sqrt
11878 Do not generate inline code for sqrt.
11881 @itemx -mno-fused-madd
11882 @opindex mfused-madd
11883 @opindex mno-fused-madd
11884 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
11885 instructions. The default is to use these instructions.
11887 @item -mno-dwarf2-asm
11888 @itemx -mdwarf2-asm
11889 @opindex mno-dwarf2-asm
11890 @opindex mdwarf2-asm
11891 Don't (or do) generate assembler code for the DWARF2 line number debugging
11892 info. This may be useful when not using the GNU assembler.
11894 @item -mearly-stop-bits
11895 @itemx -mno-early-stop-bits
11896 @opindex mearly-stop-bits
11897 @opindex mno-early-stop-bits
11898 Allow stop bits to be placed earlier than immediately preceding the
11899 instruction that triggered the stop bit. This can improve instruction
11900 scheduling, but does not always do so.
11902 @item -mfixed-range=@var{register-range}
11903 @opindex mfixed-range
11904 Generate code treating the given register range as fixed registers.
11905 A fixed register is one that the register allocator can not use. This is
11906 useful when compiling kernel code. A register range is specified as
11907 two registers separated by a dash. Multiple register ranges can be
11908 specified separated by a comma.
11910 @item -mtls-size=@var{tls-size}
11912 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11915 @item -mtune=@var{cpu-type}
11917 Tune the instruction scheduling for a particular CPU, Valid values are
11918 itanium, itanium1, merced, itanium2, and mckinley.
11924 Generate code for a 32-bit or 64-bit environment.
11925 The 32-bit environment sets int, long and pointer to 32 bits.
11926 The 64-bit environment sets int to 32 bits and long and pointer
11927 to 64 bits. These are HP-UX specific flags.
11929 @item -mno-sched-br-data-spec
11930 @itemx -msched-br-data-spec
11931 @opindex mno-sched-br-data-spec
11932 @opindex msched-br-data-spec
11933 (Dis/En)able data speculative scheduling before reload.
11934 This will result in generation of the ld.a instructions and
11935 the corresponding check instructions (ld.c / chk.a).
11936 The default is 'disable'.
11938 @item -msched-ar-data-spec
11939 @itemx -mno-sched-ar-data-spec
11940 @opindex msched-ar-data-spec
11941 @opindex mno-sched-ar-data-spec
11942 (En/Dis)able data speculative scheduling after reload.
11943 This will result in generation of the ld.a instructions and
11944 the corresponding check instructions (ld.c / chk.a).
11945 The default is 'enable'.
11947 @item -mno-sched-control-spec
11948 @itemx -msched-control-spec
11949 @opindex mno-sched-control-spec
11950 @opindex msched-control-spec
11951 (Dis/En)able control speculative scheduling. This feature is
11952 available only during region scheduling (i.e.@: before reload).
11953 This will result in generation of the ld.s instructions and
11954 the corresponding check instructions chk.s .
11955 The default is 'disable'.
11957 @item -msched-br-in-data-spec
11958 @itemx -mno-sched-br-in-data-spec
11959 @opindex msched-br-in-data-spec
11960 @opindex mno-sched-br-in-data-spec
11961 (En/Dis)able speculative scheduling of the instructions that
11962 are dependent on the data speculative loads before reload.
11963 This is effective only with @option{-msched-br-data-spec} enabled.
11964 The default is 'enable'.
11966 @item -msched-ar-in-data-spec
11967 @itemx -mno-sched-ar-in-data-spec
11968 @opindex msched-ar-in-data-spec
11969 @opindex mno-sched-ar-in-data-spec
11970 (En/Dis)able speculative scheduling of the instructions that
11971 are dependent on the data speculative loads after reload.
11972 This is effective only with @option{-msched-ar-data-spec} enabled.
11973 The default is 'enable'.
11975 @item -msched-in-control-spec
11976 @itemx -mno-sched-in-control-spec
11977 @opindex msched-in-control-spec
11978 @opindex mno-sched-in-control-spec
11979 (En/Dis)able speculative scheduling of the instructions that
11980 are dependent on the control speculative loads.
11981 This is effective only with @option{-msched-control-spec} enabled.
11982 The default is 'enable'.
11984 @item -mno-sched-prefer-non-data-spec-insns
11985 @itemx -msched-prefer-non-data-spec-insns
11986 @opindex mno-sched-prefer-non-data-spec-insns
11987 @opindex msched-prefer-non-data-spec-insns
11988 If enabled, data speculative instructions will be chosen for schedule
11989 only if there are no other choices at the moment. This will make
11990 the use of the data speculation much more conservative.
11991 The default is 'disable'.
11993 @item -mno-sched-prefer-non-control-spec-insns
11994 @itemx -msched-prefer-non-control-spec-insns
11995 @opindex mno-sched-prefer-non-control-spec-insns
11996 @opindex msched-prefer-non-control-spec-insns
11997 If enabled, control speculative instructions will be chosen for schedule
11998 only if there are no other choices at the moment. This will make
11999 the use of the control speculation much more conservative.
12000 The default is 'disable'.
12002 @item -mno-sched-count-spec-in-critical-path
12003 @itemx -msched-count-spec-in-critical-path
12004 @opindex mno-sched-count-spec-in-critical-path
12005 @opindex msched-count-spec-in-critical-path
12006 If enabled, speculative dependencies will be considered during
12007 computation of the instructions priorities. This will make the use of the
12008 speculation a bit more conservative.
12009 The default is 'disable'.
12011 @item -msched-spec-ldc
12012 @opindex msched-spec-ldc
12013 Use a simple data speculation check. This option is on by default.
12015 @item -msched-control-spec-ldc
12016 @opindex msched-spec-ldc
12017 Use a simple check for control speculation. This option is on by default.
12019 @item -msched-stop-bits-after-every-cycle
12020 @opindex msched-stop-bits-after-every-cycle
12021 Place a stop bit after every cycle when scheduling. This option is on
12024 @item -msched-fp-mem-deps-zero-cost
12025 @opindex msched-fp-mem-deps-zero-cost
12026 Assume that floating-point stores and loads are not likely to cause a conflict
12027 when placed into the same instruction group. This option is disabled by
12030 @item -msel-sched-dont-check-control-spec
12031 @opindex msel-sched-dont-check-control-spec
12032 Generate checks for control speculation in selective scheduling.
12033 This flag is disabled by default.
12035 @item -msched-max-memory-insns=@var{max-insns}
12036 @opindex msched-max-memory-insns
12037 Limit on the number of memory insns per instruction group, giving lower
12038 priority to subsequent memory insns attempting to schedule in the same
12039 instruction group. Frequently useful to prevent cache bank conflicts.
12040 The default value is 1.
12042 @item -msched-max-memory-insns-hard-limit
12043 @opindex msched-max-memory-insns-hard-limit
12044 Disallow more than `msched-max-memory-insns' in instruction group.
12045 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12046 when limit is reached but may still schedule memory operations.
12051 @subsection M32C Options
12052 @cindex M32C options
12055 @item -mcpu=@var{name}
12057 Select the CPU for which code is generated. @var{name} may be one of
12058 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12059 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12060 the M32C/80 series.
12064 Specifies that the program will be run on the simulator. This causes
12065 an alternate runtime library to be linked in which supports, for
12066 example, file I/O@. You must not use this option when generating
12067 programs that will run on real hardware; you must provide your own
12068 runtime library for whatever I/O functions are needed.
12070 @item -memregs=@var{number}
12072 Specifies the number of memory-based pseudo-registers GCC will use
12073 during code generation. These pseudo-registers will be used like real
12074 registers, so there is a tradeoff between GCC's ability to fit the
12075 code into available registers, and the performance penalty of using
12076 memory instead of registers. Note that all modules in a program must
12077 be compiled with the same value for this option. Because of that, you
12078 must not use this option with the default runtime libraries gcc
12083 @node M32R/D Options
12084 @subsection M32R/D Options
12085 @cindex M32R/D options
12087 These @option{-m} options are defined for Renesas M32R/D architectures:
12092 Generate code for the M32R/2@.
12096 Generate code for the M32R/X@.
12100 Generate code for the M32R@. This is the default.
12102 @item -mmodel=small
12103 @opindex mmodel=small
12104 Assume all objects live in the lower 16MB of memory (so that their addresses
12105 can be loaded with the @code{ld24} instruction), and assume all subroutines
12106 are reachable with the @code{bl} instruction.
12107 This is the default.
12109 The addressability of a particular object can be set with the
12110 @code{model} attribute.
12112 @item -mmodel=medium
12113 @opindex mmodel=medium
12114 Assume objects may be anywhere in the 32-bit address space (the compiler
12115 will generate @code{seth/add3} instructions to load their addresses), and
12116 assume all subroutines are reachable with the @code{bl} instruction.
12118 @item -mmodel=large
12119 @opindex mmodel=large
12120 Assume objects may be anywhere in the 32-bit address space (the compiler
12121 will generate @code{seth/add3} instructions to load their addresses), and
12122 assume subroutines may not be reachable with the @code{bl} instruction
12123 (the compiler will generate the much slower @code{seth/add3/jl}
12124 instruction sequence).
12127 @opindex msdata=none
12128 Disable use of the small data area. Variables will be put into
12129 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12130 @code{section} attribute has been specified).
12131 This is the default.
12133 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12134 Objects may be explicitly put in the small data area with the
12135 @code{section} attribute using one of these sections.
12137 @item -msdata=sdata
12138 @opindex msdata=sdata
12139 Put small global and static data in the small data area, but do not
12140 generate special code to reference them.
12143 @opindex msdata=use
12144 Put small global and static data in the small data area, and generate
12145 special instructions to reference them.
12149 @cindex smaller data references
12150 Put global and static objects less than or equal to @var{num} bytes
12151 into the small data or bss sections instead of the normal data or bss
12152 sections. The default value of @var{num} is 8.
12153 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12154 for this option to have any effect.
12156 All modules should be compiled with the same @option{-G @var{num}} value.
12157 Compiling with different values of @var{num} may or may not work; if it
12158 doesn't the linker will give an error message---incorrect code will not be
12163 Makes the M32R specific code in the compiler display some statistics
12164 that might help in debugging programs.
12166 @item -malign-loops
12167 @opindex malign-loops
12168 Align all loops to a 32-byte boundary.
12170 @item -mno-align-loops
12171 @opindex mno-align-loops
12172 Do not enforce a 32-byte alignment for loops. This is the default.
12174 @item -missue-rate=@var{number}
12175 @opindex missue-rate=@var{number}
12176 Issue @var{number} instructions per cycle. @var{number} can only be 1
12179 @item -mbranch-cost=@var{number}
12180 @opindex mbranch-cost=@var{number}
12181 @var{number} can only be 1 or 2. If it is 1 then branches will be
12182 preferred over conditional code, if it is 2, then the opposite will
12185 @item -mflush-trap=@var{number}
12186 @opindex mflush-trap=@var{number}
12187 Specifies the trap number to use to flush the cache. The default is
12188 12. Valid numbers are between 0 and 15 inclusive.
12190 @item -mno-flush-trap
12191 @opindex mno-flush-trap
12192 Specifies that the cache cannot be flushed by using a trap.
12194 @item -mflush-func=@var{name}
12195 @opindex mflush-func=@var{name}
12196 Specifies the name of the operating system function to call to flush
12197 the cache. The default is @emph{_flush_cache}, but a function call
12198 will only be used if a trap is not available.
12200 @item -mno-flush-func
12201 @opindex mno-flush-func
12202 Indicates that there is no OS function for flushing the cache.
12206 @node M680x0 Options
12207 @subsection M680x0 Options
12208 @cindex M680x0 options
12210 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12211 The default settings depend on which architecture was selected when
12212 the compiler was configured; the defaults for the most common choices
12216 @item -march=@var{arch}
12218 Generate code for a specific M680x0 or ColdFire instruction set
12219 architecture. Permissible values of @var{arch} for M680x0
12220 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12221 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12222 architectures are selected according to Freescale's ISA classification
12223 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12224 @samp{isab} and @samp{isac}.
12226 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12227 code for a ColdFire target. The @var{arch} in this macro is one of the
12228 @option{-march} arguments given above.
12230 When used together, @option{-march} and @option{-mtune} select code
12231 that runs on a family of similar processors but that is optimized
12232 for a particular microarchitecture.
12234 @item -mcpu=@var{cpu}
12236 Generate code for a specific M680x0 or ColdFire processor.
12237 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12238 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12239 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12240 below, which also classifies the CPUs into families:
12242 @multitable @columnfractions 0.20 0.80
12243 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12244 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12245 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12246 @item @samp{5206e} @tab @samp{5206e}
12247 @item @samp{5208} @tab @samp{5207} @samp{5208}
12248 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12249 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12250 @item @samp{5216} @tab @samp{5214} @samp{5216}
12251 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12252 @item @samp{5225} @tab @samp{5224} @samp{5225}
12253 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12254 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12255 @item @samp{5249} @tab @samp{5249}
12256 @item @samp{5250} @tab @samp{5250}
12257 @item @samp{5271} @tab @samp{5270} @samp{5271}
12258 @item @samp{5272} @tab @samp{5272}
12259 @item @samp{5275} @tab @samp{5274} @samp{5275}
12260 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12261 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12262 @item @samp{5307} @tab @samp{5307}
12263 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12264 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12265 @item @samp{5407} @tab @samp{5407}
12266 @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}
12269 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12270 @var{arch} is compatible with @var{cpu}. Other combinations of
12271 @option{-mcpu} and @option{-march} are rejected.
12273 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12274 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12275 where the value of @var{family} is given by the table above.
12277 @item -mtune=@var{tune}
12279 Tune the code for a particular microarchitecture, within the
12280 constraints set by @option{-march} and @option{-mcpu}.
12281 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12282 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12283 and @samp{cpu32}. The ColdFire microarchitectures
12284 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12286 You can also use @option{-mtune=68020-40} for code that needs
12287 to run relatively well on 68020, 68030 and 68040 targets.
12288 @option{-mtune=68020-60} is similar but includes 68060 targets
12289 as well. These two options select the same tuning decisions as
12290 @option{-m68020-40} and @option{-m68020-60} respectively.
12292 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12293 when tuning for 680x0 architecture @var{arch}. It also defines
12294 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12295 option is used. If gcc is tuning for a range of architectures,
12296 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12297 it defines the macros for every architecture in the range.
12299 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12300 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12301 of the arguments given above.
12307 Generate output for a 68000. This is the default
12308 when the compiler is configured for 68000-based systems.
12309 It is equivalent to @option{-march=68000}.
12311 Use this option for microcontrollers with a 68000 or EC000 core,
12312 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12316 Generate output for a 68010. This is the default
12317 when the compiler is configured for 68010-based systems.
12318 It is equivalent to @option{-march=68010}.
12324 Generate output for a 68020. This is the default
12325 when the compiler is configured for 68020-based systems.
12326 It is equivalent to @option{-march=68020}.
12330 Generate output for a 68030. This is the default when the compiler is
12331 configured for 68030-based systems. It is equivalent to
12332 @option{-march=68030}.
12336 Generate output for a 68040. This is the default when the compiler is
12337 configured for 68040-based systems. It is equivalent to
12338 @option{-march=68040}.
12340 This option inhibits the use of 68881/68882 instructions that have to be
12341 emulated by software on the 68040. Use this option if your 68040 does not
12342 have code to emulate those instructions.
12346 Generate output for a 68060. This is the default when the compiler is
12347 configured for 68060-based systems. It is equivalent to
12348 @option{-march=68060}.
12350 This option inhibits the use of 68020 and 68881/68882 instructions that
12351 have to be emulated by software on the 68060. Use this option if your 68060
12352 does not have code to emulate those instructions.
12356 Generate output for a CPU32. This is the default
12357 when the compiler is configured for CPU32-based systems.
12358 It is equivalent to @option{-march=cpu32}.
12360 Use this option for microcontrollers with a
12361 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12362 68336, 68340, 68341, 68349 and 68360.
12366 Generate output for a 520X ColdFire CPU@. This is the default
12367 when the compiler is configured for 520X-based systems.
12368 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12369 in favor of that option.
12371 Use this option for microcontroller with a 5200 core, including
12372 the MCF5202, MCF5203, MCF5204 and MCF5206.
12376 Generate output for a 5206e ColdFire CPU@. The option is now
12377 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12381 Generate output for a member of the ColdFire 528X family.
12382 The option is now deprecated in favor of the equivalent
12383 @option{-mcpu=528x}.
12387 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12388 in favor of the equivalent @option{-mcpu=5307}.
12392 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12393 in favor of the equivalent @option{-mcpu=5407}.
12397 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12398 This includes use of hardware floating point instructions.
12399 The option is equivalent to @option{-mcpu=547x}, and is now
12400 deprecated in favor of that option.
12404 Generate output for a 68040, without using any of the new instructions.
12405 This results in code which can run relatively efficiently on either a
12406 68020/68881 or a 68030 or a 68040. The generated code does use the
12407 68881 instructions that are emulated on the 68040.
12409 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12413 Generate output for a 68060, without using any of the new instructions.
12414 This results in code which can run relatively efficiently on either a
12415 68020/68881 or a 68030 or a 68040. The generated code does use the
12416 68881 instructions that are emulated on the 68060.
12418 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12422 @opindex mhard-float
12424 Generate floating-point instructions. This is the default for 68020
12425 and above, and for ColdFire devices that have an FPU@. It defines the
12426 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12427 on ColdFire targets.
12430 @opindex msoft-float
12431 Do not generate floating-point instructions; use library calls instead.
12432 This is the default for 68000, 68010, and 68832 targets. It is also
12433 the default for ColdFire devices that have no FPU.
12439 Generate (do not generate) ColdFire hardware divide and remainder
12440 instructions. If @option{-march} is used without @option{-mcpu},
12441 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12442 architectures. Otherwise, the default is taken from the target CPU
12443 (either the default CPU, or the one specified by @option{-mcpu}). For
12444 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12445 @option{-mcpu=5206e}.
12447 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12451 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12452 Additionally, parameters passed on the stack are also aligned to a
12453 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12457 Do not consider type @code{int} to be 16 bits wide. This is the default.
12460 @itemx -mno-bitfield
12461 @opindex mnobitfield
12462 @opindex mno-bitfield
12463 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12464 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12468 Do use the bit-field instructions. The @option{-m68020} option implies
12469 @option{-mbitfield}. This is the default if you use a configuration
12470 designed for a 68020.
12474 Use a different function-calling convention, in which functions
12475 that take a fixed number of arguments return with the @code{rtd}
12476 instruction, which pops their arguments while returning. This
12477 saves one instruction in the caller since there is no need to pop
12478 the arguments there.
12480 This calling convention is incompatible with the one normally
12481 used on Unix, so you cannot use it if you need to call libraries
12482 compiled with the Unix compiler.
12484 Also, you must provide function prototypes for all functions that
12485 take variable numbers of arguments (including @code{printf});
12486 otherwise incorrect code will be generated for calls to those
12489 In addition, seriously incorrect code will result if you call a
12490 function with too many arguments. (Normally, extra arguments are
12491 harmlessly ignored.)
12493 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12494 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12498 Do not use the calling conventions selected by @option{-mrtd}.
12499 This is the default.
12502 @itemx -mno-align-int
12503 @opindex malign-int
12504 @opindex mno-align-int
12505 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12506 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12507 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12508 Aligning variables on 32-bit boundaries produces code that runs somewhat
12509 faster on processors with 32-bit busses at the expense of more memory.
12511 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12512 align structures containing the above types differently than
12513 most published application binary interface specifications for the m68k.
12517 Use the pc-relative addressing mode of the 68000 directly, instead of
12518 using a global offset table. At present, this option implies @option{-fpic},
12519 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12520 not presently supported with @option{-mpcrel}, though this could be supported for
12521 68020 and higher processors.
12523 @item -mno-strict-align
12524 @itemx -mstrict-align
12525 @opindex mno-strict-align
12526 @opindex mstrict-align
12527 Do not (do) assume that unaligned memory references will be handled by
12531 Generate code that allows the data segment to be located in a different
12532 area of memory from the text segment. This allows for execute in place in
12533 an environment without virtual memory management. This option implies
12536 @item -mno-sep-data
12537 Generate code that assumes that the data segment follows the text segment.
12538 This is the default.
12540 @item -mid-shared-library
12541 Generate code that supports shared libraries via the library ID method.
12542 This allows for execute in place and shared libraries in an environment
12543 without virtual memory management. This option implies @option{-fPIC}.
12545 @item -mno-id-shared-library
12546 Generate code that doesn't assume ID based shared libraries are being used.
12547 This is the default.
12549 @item -mshared-library-id=n
12550 Specified the identification number of the ID based shared library being
12551 compiled. Specifying a value of 0 will generate more compact code, specifying
12552 other values will force the allocation of that number to the current
12553 library but is no more space or time efficient than omitting this option.
12559 When generating position-independent code for ColdFire, generate code
12560 that works if the GOT has more than 8192 entries. This code is
12561 larger and slower than code generated without this option. On M680x0
12562 processors, this option is not needed; @option{-fPIC} suffices.
12564 GCC normally uses a single instruction to load values from the GOT@.
12565 While this is relatively efficient, it only works if the GOT
12566 is smaller than about 64k. Anything larger causes the linker
12567 to report an error such as:
12569 @cindex relocation truncated to fit (ColdFire)
12571 relocation truncated to fit: R_68K_GOT16O foobar
12574 If this happens, you should recompile your code with @option{-mxgot}.
12575 It should then work with very large GOTs. However, code generated with
12576 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12577 the value of a global symbol.
12579 Note that some linkers, including newer versions of the GNU linker,
12580 can create multiple GOTs and sort GOT entries. If you have such a linker,
12581 you should only need to use @option{-mxgot} when compiling a single
12582 object file that accesses more than 8192 GOT entries. Very few do.
12584 These options have no effect unless GCC is generating
12585 position-independent code.
12589 @node M68hc1x Options
12590 @subsection M68hc1x Options
12591 @cindex M68hc1x options
12593 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12594 microcontrollers. The default values for these options depends on
12595 which style of microcontroller was selected when the compiler was configured;
12596 the defaults for the most common choices are given below.
12603 Generate output for a 68HC11. This is the default
12604 when the compiler is configured for 68HC11-based systems.
12610 Generate output for a 68HC12. This is the default
12611 when the compiler is configured for 68HC12-based systems.
12617 Generate output for a 68HCS12.
12619 @item -mauto-incdec
12620 @opindex mauto-incdec
12621 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12628 Enable the use of 68HC12 min and max instructions.
12631 @itemx -mno-long-calls
12632 @opindex mlong-calls
12633 @opindex mno-long-calls
12634 Treat all calls as being far away (near). If calls are assumed to be
12635 far away, the compiler will use the @code{call} instruction to
12636 call a function and the @code{rtc} instruction for returning.
12640 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12642 @item -msoft-reg-count=@var{count}
12643 @opindex msoft-reg-count
12644 Specify the number of pseudo-soft registers which are used for the
12645 code generation. The maximum number is 32. Using more pseudo-soft
12646 register may or may not result in better code depending on the program.
12647 The default is 4 for 68HC11 and 2 for 68HC12.
12651 @node MCore Options
12652 @subsection MCore Options
12653 @cindex MCore options
12655 These are the @samp{-m} options defined for the Motorola M*Core
12661 @itemx -mno-hardlit
12663 @opindex mno-hardlit
12664 Inline constants into the code stream if it can be done in two
12665 instructions or less.
12671 Use the divide instruction. (Enabled by default).
12673 @item -mrelax-immediate
12674 @itemx -mno-relax-immediate
12675 @opindex mrelax-immediate
12676 @opindex mno-relax-immediate
12677 Allow arbitrary sized immediates in bit operations.
12679 @item -mwide-bitfields
12680 @itemx -mno-wide-bitfields
12681 @opindex mwide-bitfields
12682 @opindex mno-wide-bitfields
12683 Always treat bit-fields as int-sized.
12685 @item -m4byte-functions
12686 @itemx -mno-4byte-functions
12687 @opindex m4byte-functions
12688 @opindex mno-4byte-functions
12689 Force all functions to be aligned to a four byte boundary.
12691 @item -mcallgraph-data
12692 @itemx -mno-callgraph-data
12693 @opindex mcallgraph-data
12694 @opindex mno-callgraph-data
12695 Emit callgraph information.
12698 @itemx -mno-slow-bytes
12699 @opindex mslow-bytes
12700 @opindex mno-slow-bytes
12701 Prefer word access when reading byte quantities.
12703 @item -mlittle-endian
12704 @itemx -mbig-endian
12705 @opindex mlittle-endian
12706 @opindex mbig-endian
12707 Generate code for a little endian target.
12713 Generate code for the 210 processor.
12717 Assume that run-time support has been provided and so omit the
12718 simulator library (@file{libsim.a)} from the linker command line.
12720 @item -mstack-increment=@var{size}
12721 @opindex mstack-increment
12722 Set the maximum amount for a single stack increment operation. Large
12723 values can increase the speed of programs which contain functions
12724 that need a large amount of stack space, but they can also trigger a
12725 segmentation fault if the stack is extended too much. The default
12731 @subsection MIPS Options
12732 @cindex MIPS options
12738 Generate big-endian code.
12742 Generate little-endian code. This is the default for @samp{mips*el-*-*}
12745 @item -march=@var{arch}
12747 Generate code that will run on @var{arch}, which can be the name of a
12748 generic MIPS ISA, or the name of a particular processor.
12750 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
12751 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
12752 The processor names are:
12753 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
12754 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
12755 @samp{5kc}, @samp{5kf},
12757 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
12758 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
12759 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
12760 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
12761 @samp{loongson2e}, @samp{loongson2f},
12765 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
12766 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
12767 @samp{rm7000}, @samp{rm9000},
12768 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
12771 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
12772 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
12774 The special value @samp{from-abi} selects the
12775 most compatible architecture for the selected ABI (that is,
12776 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
12778 Native Linux/GNU toolchains also support the value @samp{native},
12779 which selects the best architecture option for the host processor.
12780 @option{-march=native} has no effect if GCC does not recognize
12783 In processor names, a final @samp{000} can be abbreviated as @samp{k}
12784 (for example, @samp{-march=r2k}). Prefixes are optional, and
12785 @samp{vr} may be written @samp{r}.
12787 Names of the form @samp{@var{n}f2_1} refer to processors with
12788 FPUs clocked at half the rate of the core, names of the form
12789 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
12790 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12791 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12792 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12793 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12794 accepted as synonyms for @samp{@var{n}f1_1}.
12796 GCC defines two macros based on the value of this option. The first
12797 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12798 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12799 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12800 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12801 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12803 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12804 above. In other words, it will have the full prefix and will not
12805 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12806 the macro names the resolved architecture (either @samp{"mips1"} or
12807 @samp{"mips3"}). It names the default architecture when no
12808 @option{-march} option is given.
12810 @item -mtune=@var{arch}
12812 Optimize for @var{arch}. Among other things, this option controls
12813 the way instructions are scheduled, and the perceived cost of arithmetic
12814 operations. The list of @var{arch} values is the same as for
12817 When this option is not used, GCC will optimize for the processor
12818 specified by @option{-march}. By using @option{-march} and
12819 @option{-mtune} together, it is possible to generate code that will
12820 run on a family of processors, but optimize the code for one
12821 particular member of that family.
12823 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12824 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12825 @samp{-march} ones described above.
12829 Equivalent to @samp{-march=mips1}.
12833 Equivalent to @samp{-march=mips2}.
12837 Equivalent to @samp{-march=mips3}.
12841 Equivalent to @samp{-march=mips4}.
12845 Equivalent to @samp{-march=mips32}.
12849 Equivalent to @samp{-march=mips32r2}.
12853 Equivalent to @samp{-march=mips64}.
12857 Equivalent to @samp{-march=mips64r2}.
12862 @opindex mno-mips16
12863 Generate (do not generate) MIPS16 code. If GCC is targetting a
12864 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12866 MIPS16 code generation can also be controlled on a per-function basis
12867 by means of @code{mips16} and @code{nomips16} attributes.
12868 @xref{Function Attributes}, for more information.
12870 @item -mflip-mips16
12871 @opindex mflip-mips16
12872 Generate MIPS16 code on alternating functions. This option is provided
12873 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12874 not intended for ordinary use in compiling user code.
12876 @item -minterlink-mips16
12877 @itemx -mno-interlink-mips16
12878 @opindex minterlink-mips16
12879 @opindex mno-interlink-mips16
12880 Require (do not require) that non-MIPS16 code be link-compatible with
12883 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12884 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12885 therefore disables direct jumps unless GCC knows that the target of the
12886 jump is not MIPS16.
12898 Generate code for the given ABI@.
12900 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12901 generates 64-bit code when you select a 64-bit architecture, but you
12902 can use @option{-mgp32} to get 32-bit code instead.
12904 For information about the O64 ABI, see
12905 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12907 GCC supports a variant of the o32 ABI in which floating-point registers
12908 are 64 rather than 32 bits wide. You can select this combination with
12909 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12910 and @samp{mfhc1} instructions and is therefore only supported for
12911 MIPS32R2 processors.
12913 The register assignments for arguments and return values remain the
12914 same, but each scalar value is passed in a single 64-bit register
12915 rather than a pair of 32-bit registers. For example, scalar
12916 floating-point values are returned in @samp{$f0} only, not a
12917 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12918 remains the same, but all 64 bits are saved.
12921 @itemx -mno-abicalls
12923 @opindex mno-abicalls
12924 Generate (do not generate) code that is suitable for SVR4-style
12925 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12930 Generate (do not generate) code that is fully position-independent,
12931 and that can therefore be linked into shared libraries. This option
12932 only affects @option{-mabicalls}.
12934 All @option{-mabicalls} code has traditionally been position-independent,
12935 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12936 as an extension, the GNU toolchain allows executables to use absolute
12937 accesses for locally-binding symbols. It can also use shorter GP
12938 initialization sequences and generate direct calls to locally-defined
12939 functions. This mode is selected by @option{-mno-shared}.
12941 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12942 objects that can only be linked by the GNU linker. However, the option
12943 does not affect the ABI of the final executable; it only affects the ABI
12944 of relocatable objects. Using @option{-mno-shared} will generally make
12945 executables both smaller and quicker.
12947 @option{-mshared} is the default.
12953 Assume (do not assume) that the static and dynamic linkers
12954 support PLTs and copy relocations. This option only affects
12955 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
12956 has no effect without @samp{-msym32}.
12958 You can make @option{-mplt} the default by configuring
12959 GCC with @option{--with-mips-plt}. The default is
12960 @option{-mno-plt} otherwise.
12966 Lift (do not lift) the usual restrictions on the size of the global
12969 GCC normally uses a single instruction to load values from the GOT@.
12970 While this is relatively efficient, it will only work if the GOT
12971 is smaller than about 64k. Anything larger will cause the linker
12972 to report an error such as:
12974 @cindex relocation truncated to fit (MIPS)
12976 relocation truncated to fit: R_MIPS_GOT16 foobar
12979 If this happens, you should recompile your code with @option{-mxgot}.
12980 It should then work with very large GOTs, although it will also be
12981 less efficient, since it will take three instructions to fetch the
12982 value of a global symbol.
12984 Note that some linkers can create multiple GOTs. If you have such a
12985 linker, you should only need to use @option{-mxgot} when a single object
12986 file accesses more than 64k's worth of GOT entries. Very few do.
12988 These options have no effect unless GCC is generating position
12993 Assume that general-purpose registers are 32 bits wide.
12997 Assume that general-purpose registers are 64 bits wide.
13001 Assume that floating-point registers are 32 bits wide.
13005 Assume that floating-point registers are 64 bits wide.
13008 @opindex mhard-float
13009 Use floating-point coprocessor instructions.
13012 @opindex msoft-float
13013 Do not use floating-point coprocessor instructions. Implement
13014 floating-point calculations using library calls instead.
13016 @item -msingle-float
13017 @opindex msingle-float
13018 Assume that the floating-point coprocessor only supports single-precision
13021 @item -mdouble-float
13022 @opindex mdouble-float
13023 Assume that the floating-point coprocessor supports double-precision
13024 operations. This is the default.
13030 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13031 implement atomic memory built-in functions. When neither option is
13032 specified, GCC will use the instructions if the target architecture
13035 @option{-mllsc} is useful if the runtime environment can emulate the
13036 instructions and @option{-mno-llsc} can be useful when compiling for
13037 nonstandard ISAs. You can make either option the default by
13038 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13039 respectively. @option{--with-llsc} is the default for some
13040 configurations; see the installation documentation for details.
13046 Use (do not use) revision 1 of the MIPS DSP ASE@.
13047 @xref{MIPS DSP Built-in Functions}. This option defines the
13048 preprocessor macro @samp{__mips_dsp}. It also defines
13049 @samp{__mips_dsp_rev} to 1.
13055 Use (do not use) revision 2 of the MIPS DSP ASE@.
13056 @xref{MIPS DSP Built-in Functions}. This option defines the
13057 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13058 It also defines @samp{__mips_dsp_rev} to 2.
13061 @itemx -mno-smartmips
13062 @opindex msmartmips
13063 @opindex mno-smartmips
13064 Use (do not use) the MIPS SmartMIPS ASE.
13066 @item -mpaired-single
13067 @itemx -mno-paired-single
13068 @opindex mpaired-single
13069 @opindex mno-paired-single
13070 Use (do not use) paired-single floating-point instructions.
13071 @xref{MIPS Paired-Single Support}. This option requires
13072 hardware floating-point support to be enabled.
13078 Use (do not use) MIPS Digital Media Extension instructions.
13079 This option can only be used when generating 64-bit code and requires
13080 hardware floating-point support to be enabled.
13085 @opindex mno-mips3d
13086 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13087 The option @option{-mips3d} implies @option{-mpaired-single}.
13093 Use (do not use) MT Multithreading instructions.
13097 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13098 an explanation of the default and the way that the pointer size is
13103 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13105 The default size of @code{int}s, @code{long}s and pointers depends on
13106 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13107 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13108 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13109 or the same size as integer registers, whichever is smaller.
13115 Assume (do not assume) that all symbols have 32-bit values, regardless
13116 of the selected ABI@. This option is useful in combination with
13117 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13118 to generate shorter and faster references to symbolic addresses.
13122 Put definitions of externally-visible data in a small data section
13123 if that data is no bigger than @var{num} bytes. GCC can then access
13124 the data more efficiently; see @option{-mgpopt} for details.
13126 The default @option{-G} option depends on the configuration.
13128 @item -mlocal-sdata
13129 @itemx -mno-local-sdata
13130 @opindex mlocal-sdata
13131 @opindex mno-local-sdata
13132 Extend (do not extend) the @option{-G} behavior to local data too,
13133 such as to static variables in C@. @option{-mlocal-sdata} is the
13134 default for all configurations.
13136 If the linker complains that an application is using too much small data,
13137 you might want to try rebuilding the less performance-critical parts with
13138 @option{-mno-local-sdata}. You might also want to build large
13139 libraries with @option{-mno-local-sdata}, so that the libraries leave
13140 more room for the main program.
13142 @item -mextern-sdata
13143 @itemx -mno-extern-sdata
13144 @opindex mextern-sdata
13145 @opindex mno-extern-sdata
13146 Assume (do not assume) that externally-defined data will be in
13147 a small data section if that data is within the @option{-G} limit.
13148 @option{-mextern-sdata} is the default for all configurations.
13150 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13151 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13152 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13153 is placed in a small data section. If @var{Var} is defined by another
13154 module, you must either compile that module with a high-enough
13155 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13156 definition. If @var{Var} is common, you must link the application
13157 with a high-enough @option{-G} setting.
13159 The easiest way of satisfying these restrictions is to compile
13160 and link every module with the same @option{-G} option. However,
13161 you may wish to build a library that supports several different
13162 small data limits. You can do this by compiling the library with
13163 the highest supported @option{-G} setting and additionally using
13164 @option{-mno-extern-sdata} to stop the library from making assumptions
13165 about externally-defined data.
13171 Use (do not use) GP-relative accesses for symbols that are known to be
13172 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13173 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13176 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13177 might not hold the value of @code{_gp}. For example, if the code is
13178 part of a library that might be used in a boot monitor, programs that
13179 call boot monitor routines will pass an unknown value in @code{$gp}.
13180 (In such situations, the boot monitor itself would usually be compiled
13181 with @option{-G0}.)
13183 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13184 @option{-mno-extern-sdata}.
13186 @item -membedded-data
13187 @itemx -mno-embedded-data
13188 @opindex membedded-data
13189 @opindex mno-embedded-data
13190 Allocate variables to the read-only data section first if possible, then
13191 next in the small data section if possible, otherwise in data. This gives
13192 slightly slower code than the default, but reduces the amount of RAM required
13193 when executing, and thus may be preferred for some embedded systems.
13195 @item -muninit-const-in-rodata
13196 @itemx -mno-uninit-const-in-rodata
13197 @opindex muninit-const-in-rodata
13198 @opindex mno-uninit-const-in-rodata
13199 Put uninitialized @code{const} variables in the read-only data section.
13200 This option is only meaningful in conjunction with @option{-membedded-data}.
13202 @item -mcode-readable=@var{setting}
13203 @opindex mcode-readable
13204 Specify whether GCC may generate code that reads from executable sections.
13205 There are three possible settings:
13208 @item -mcode-readable=yes
13209 Instructions may freely access executable sections. This is the
13212 @item -mcode-readable=pcrel
13213 MIPS16 PC-relative load instructions can access executable sections,
13214 but other instructions must not do so. This option is useful on 4KSc
13215 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13216 It is also useful on processors that can be configured to have a dual
13217 instruction/data SRAM interface and that, like the M4K, automatically
13218 redirect PC-relative loads to the instruction RAM.
13220 @item -mcode-readable=no
13221 Instructions must not access executable sections. This option can be
13222 useful on targets that are configured to have a dual instruction/data
13223 SRAM interface but that (unlike the M4K) do not automatically redirect
13224 PC-relative loads to the instruction RAM.
13227 @item -msplit-addresses
13228 @itemx -mno-split-addresses
13229 @opindex msplit-addresses
13230 @opindex mno-split-addresses
13231 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13232 relocation operators. This option has been superseded by
13233 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13235 @item -mexplicit-relocs
13236 @itemx -mno-explicit-relocs
13237 @opindex mexplicit-relocs
13238 @opindex mno-explicit-relocs
13239 Use (do not use) assembler relocation operators when dealing with symbolic
13240 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13241 is to use assembler macros instead.
13243 @option{-mexplicit-relocs} is the default if GCC was configured
13244 to use an assembler that supports relocation operators.
13246 @item -mcheck-zero-division
13247 @itemx -mno-check-zero-division
13248 @opindex mcheck-zero-division
13249 @opindex mno-check-zero-division
13250 Trap (do not trap) on integer division by zero.
13252 The default is @option{-mcheck-zero-division}.
13254 @item -mdivide-traps
13255 @itemx -mdivide-breaks
13256 @opindex mdivide-traps
13257 @opindex mdivide-breaks
13258 MIPS systems check for division by zero by generating either a
13259 conditional trap or a break instruction. Using traps results in
13260 smaller code, but is only supported on MIPS II and later. Also, some
13261 versions of the Linux kernel have a bug that prevents trap from
13262 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13263 allow conditional traps on architectures that support them and
13264 @option{-mdivide-breaks} to force the use of breaks.
13266 The default is usually @option{-mdivide-traps}, but this can be
13267 overridden at configure time using @option{--with-divide=breaks}.
13268 Divide-by-zero checks can be completely disabled using
13269 @option{-mno-check-zero-division}.
13274 @opindex mno-memcpy
13275 Force (do not force) the use of @code{memcpy()} for non-trivial block
13276 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13277 most constant-sized copies.
13280 @itemx -mno-long-calls
13281 @opindex mlong-calls
13282 @opindex mno-long-calls
13283 Disable (do not disable) use of the @code{jal} instruction. Calling
13284 functions using @code{jal} is more efficient but requires the caller
13285 and callee to be in the same 256 megabyte segment.
13287 This option has no effect on abicalls code. The default is
13288 @option{-mno-long-calls}.
13294 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13295 instructions, as provided by the R4650 ISA@.
13298 @itemx -mno-fused-madd
13299 @opindex mfused-madd
13300 @opindex mno-fused-madd
13301 Enable (disable) use of the floating point multiply-accumulate
13302 instructions, when they are available. The default is
13303 @option{-mfused-madd}.
13305 When multiply-accumulate instructions are used, the intermediate
13306 product is calculated to infinite precision and is not subject to
13307 the FCSR Flush to Zero bit. This may be undesirable in some
13312 Tell the MIPS assembler to not run its preprocessor over user
13313 assembler files (with a @samp{.s} suffix) when assembling them.
13316 @itemx -mno-fix-r4000
13317 @opindex mfix-r4000
13318 @opindex mno-fix-r4000
13319 Work around certain R4000 CPU errata:
13322 A double-word or a variable shift may give an incorrect result if executed
13323 immediately after starting an integer division.
13325 A double-word or a variable shift may give an incorrect result if executed
13326 while an integer multiplication is in progress.
13328 An integer division may give an incorrect result if started in a delay slot
13329 of a taken branch or a jump.
13333 @itemx -mno-fix-r4400
13334 @opindex mfix-r4400
13335 @opindex mno-fix-r4400
13336 Work around certain R4400 CPU errata:
13339 A double-word or a variable shift may give an incorrect result if executed
13340 immediately after starting an integer division.
13344 @itemx -mno-fix-r10000
13345 @opindex mfix-r10000
13346 @opindex mno-fix-r10000
13347 Work around certain R10000 errata:
13350 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13351 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13354 This option can only be used if the target architecture supports
13355 branch-likely instructions. @option{-mfix-r10000} is the default when
13356 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13360 @itemx -mno-fix-vr4120
13361 @opindex mfix-vr4120
13362 Work around certain VR4120 errata:
13365 @code{dmultu} does not always produce the correct result.
13367 @code{div} and @code{ddiv} do not always produce the correct result if one
13368 of the operands is negative.
13370 The workarounds for the division errata rely on special functions in
13371 @file{libgcc.a}. At present, these functions are only provided by
13372 the @code{mips64vr*-elf} configurations.
13374 Other VR4120 errata require a nop to be inserted between certain pairs of
13375 instructions. These errata are handled by the assembler, not by GCC itself.
13378 @opindex mfix-vr4130
13379 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
13380 workarounds are implemented by the assembler rather than by GCC,
13381 although GCC will avoid using @code{mflo} and @code{mfhi} if the
13382 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
13383 instructions are available instead.
13386 @itemx -mno-fix-sb1
13388 Work around certain SB-1 CPU core errata.
13389 (This flag currently works around the SB-1 revision 2
13390 ``F1'' and ``F2'' floating point errata.)
13392 @item -mr10k-cache-barrier=@var{setting}
13393 @opindex mr10k-cache-barrier
13394 Specify whether GCC should insert cache barriers to avoid the
13395 side-effects of speculation on R10K processors.
13397 In common with many processors, the R10K tries to predict the outcome
13398 of a conditional branch and speculatively executes instructions from
13399 the ``taken'' branch. It later aborts these instructions if the
13400 predicted outcome was wrong. However, on the R10K, even aborted
13401 instructions can have side effects.
13403 This problem only affects kernel stores and, depending on the system,
13404 kernel loads. As an example, a speculatively-executed store may load
13405 the target memory into cache and mark the cache line as dirty, even if
13406 the store itself is later aborted. If a DMA operation writes to the
13407 same area of memory before the ``dirty'' line is flushed, the cached
13408 data will overwrite the DMA-ed data. See the R10K processor manual
13409 for a full description, including other potential problems.
13411 One workaround is to insert cache barrier instructions before every memory
13412 access that might be speculatively executed and that might have side
13413 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
13414 controls GCC's implementation of this workaround. It assumes that
13415 aborted accesses to any byte in the following regions will not have
13420 the memory occupied by the current function's stack frame;
13423 the memory occupied by an incoming stack argument;
13426 the memory occupied by an object with a link-time-constant address.
13429 It is the kernel's responsibility to ensure that speculative
13430 accesses to these regions are indeed safe.
13432 If the input program contains a function declaration such as:
13438 then the implementation of @code{foo} must allow @code{j foo} and
13439 @code{jal foo} to be executed speculatively. GCC honors this
13440 restriction for functions it compiles itself. It expects non-GCC
13441 functions (such as hand-written assembly code) to do the same.
13443 The option has three forms:
13446 @item -mr10k-cache-barrier=load-store
13447 Insert a cache barrier before a load or store that might be
13448 speculatively executed and that might have side effects even
13451 @item -mr10k-cache-barrier=store
13452 Insert a cache barrier before a store that might be speculatively
13453 executed and that might have side effects even if aborted.
13455 @item -mr10k-cache-barrier=none
13456 Disable the insertion of cache barriers. This is the default setting.
13459 @item -mflush-func=@var{func}
13460 @itemx -mno-flush-func
13461 @opindex mflush-func
13462 Specifies the function to call to flush the I and D caches, or to not
13463 call any such function. If called, the function must take the same
13464 arguments as the common @code{_flush_func()}, that is, the address of the
13465 memory range for which the cache is being flushed, the size of the
13466 memory range, and the number 3 (to flush both caches). The default
13467 depends on the target GCC was configured for, but commonly is either
13468 @samp{_flush_func} or @samp{__cpu_flush}.
13470 @item mbranch-cost=@var{num}
13471 @opindex mbranch-cost
13472 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13473 This cost is only a heuristic and is not guaranteed to produce
13474 consistent results across releases. A zero cost redundantly selects
13475 the default, which is based on the @option{-mtune} setting.
13477 @item -mbranch-likely
13478 @itemx -mno-branch-likely
13479 @opindex mbranch-likely
13480 @opindex mno-branch-likely
13481 Enable or disable use of Branch Likely instructions, regardless of the
13482 default for the selected architecture. By default, Branch Likely
13483 instructions may be generated if they are supported by the selected
13484 architecture. An exception is for the MIPS32 and MIPS64 architectures
13485 and processors which implement those architectures; for those, Branch
13486 Likely instructions will not be generated by default because the MIPS32
13487 and MIPS64 architectures specifically deprecate their use.
13489 @item -mfp-exceptions
13490 @itemx -mno-fp-exceptions
13491 @opindex mfp-exceptions
13492 Specifies whether FP exceptions are enabled. This affects how we schedule
13493 FP instructions for some processors. The default is that FP exceptions are
13496 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
13497 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
13500 @item -mvr4130-align
13501 @itemx -mno-vr4130-align
13502 @opindex mvr4130-align
13503 The VR4130 pipeline is two-way superscalar, but can only issue two
13504 instructions together if the first one is 8-byte aligned. When this
13505 option is enabled, GCC will align pairs of instructions that it
13506 thinks should execute in parallel.
13508 This option only has an effect when optimizing for the VR4130.
13509 It normally makes code faster, but at the expense of making it bigger.
13510 It is enabled by default at optimization level @option{-O3}.
13514 @subsection MMIX Options
13515 @cindex MMIX Options
13517 These options are defined for the MMIX:
13521 @itemx -mno-libfuncs
13523 @opindex mno-libfuncs
13524 Specify that intrinsic library functions are being compiled, passing all
13525 values in registers, no matter the size.
13528 @itemx -mno-epsilon
13530 @opindex mno-epsilon
13531 Generate floating-point comparison instructions that compare with respect
13532 to the @code{rE} epsilon register.
13534 @item -mabi=mmixware
13536 @opindex mabi=mmixware
13538 Generate code that passes function parameters and return values that (in
13539 the called function) are seen as registers @code{$0} and up, as opposed to
13540 the GNU ABI which uses global registers @code{$231} and up.
13542 @item -mzero-extend
13543 @itemx -mno-zero-extend
13544 @opindex mzero-extend
13545 @opindex mno-zero-extend
13546 When reading data from memory in sizes shorter than 64 bits, use (do not
13547 use) zero-extending load instructions by default, rather than
13548 sign-extending ones.
13551 @itemx -mno-knuthdiv
13553 @opindex mno-knuthdiv
13554 Make the result of a division yielding a remainder have the same sign as
13555 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13556 remainder follows the sign of the dividend. Both methods are
13557 arithmetically valid, the latter being almost exclusively used.
13559 @item -mtoplevel-symbols
13560 @itemx -mno-toplevel-symbols
13561 @opindex mtoplevel-symbols
13562 @opindex mno-toplevel-symbols
13563 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13564 code can be used with the @code{PREFIX} assembly directive.
13568 Generate an executable in the ELF format, rather than the default
13569 @samp{mmo} format used by the @command{mmix} simulator.
13571 @item -mbranch-predict
13572 @itemx -mno-branch-predict
13573 @opindex mbranch-predict
13574 @opindex mno-branch-predict
13575 Use (do not use) the probable-branch instructions, when static branch
13576 prediction indicates a probable branch.
13578 @item -mbase-addresses
13579 @itemx -mno-base-addresses
13580 @opindex mbase-addresses
13581 @opindex mno-base-addresses
13582 Generate (do not generate) code that uses @emph{base addresses}. Using a
13583 base address automatically generates a request (handled by the assembler
13584 and the linker) for a constant to be set up in a global register. The
13585 register is used for one or more base address requests within the range 0
13586 to 255 from the value held in the register. The generally leads to short
13587 and fast code, but the number of different data items that can be
13588 addressed is limited. This means that a program that uses lots of static
13589 data may require @option{-mno-base-addresses}.
13591 @item -msingle-exit
13592 @itemx -mno-single-exit
13593 @opindex msingle-exit
13594 @opindex mno-single-exit
13595 Force (do not force) generated code to have a single exit point in each
13599 @node MN10300 Options
13600 @subsection MN10300 Options
13601 @cindex MN10300 options
13603 These @option{-m} options are defined for Matsushita MN10300 architectures:
13608 Generate code to avoid bugs in the multiply instructions for the MN10300
13609 processors. This is the default.
13611 @item -mno-mult-bug
13612 @opindex mno-mult-bug
13613 Do not generate code to avoid bugs in the multiply instructions for the
13614 MN10300 processors.
13618 Generate code which uses features specific to the AM33 processor.
13622 Do not generate code which uses features specific to the AM33 processor. This
13625 @item -mreturn-pointer-on-d0
13626 @opindex mreturn-pointer-on-d0
13627 When generating a function which returns a pointer, return the pointer
13628 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13629 only in a0, and attempts to call such functions without a prototype
13630 would result in errors. Note that this option is on by default; use
13631 @option{-mno-return-pointer-on-d0} to disable it.
13635 Do not link in the C run-time initialization object file.
13639 Indicate to the linker that it should perform a relaxation optimization pass
13640 to shorten branches, calls and absolute memory addresses. This option only
13641 has an effect when used on the command line for the final link step.
13643 This option makes symbolic debugging impossible.
13646 @node PDP-11 Options
13647 @subsection PDP-11 Options
13648 @cindex PDP-11 Options
13650 These options are defined for the PDP-11:
13655 Use hardware FPP floating point. This is the default. (FIS floating
13656 point on the PDP-11/40 is not supported.)
13659 @opindex msoft-float
13660 Do not use hardware floating point.
13664 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13668 Return floating-point results in memory. This is the default.
13672 Generate code for a PDP-11/40.
13676 Generate code for a PDP-11/45. This is the default.
13680 Generate code for a PDP-11/10.
13682 @item -mbcopy-builtin
13683 @opindex mbcopy-builtin
13684 Use inline @code{movmemhi} patterns for copying memory. This is the
13689 Do not use inline @code{movmemhi} patterns for copying memory.
13695 Use 16-bit @code{int}. This is the default.
13701 Use 32-bit @code{int}.
13704 @itemx -mno-float32
13706 @opindex mno-float32
13707 Use 64-bit @code{float}. This is the default.
13710 @itemx -mno-float64
13712 @opindex mno-float64
13713 Use 32-bit @code{float}.
13717 Use @code{abshi2} pattern. This is the default.
13721 Do not use @code{abshi2} pattern.
13723 @item -mbranch-expensive
13724 @opindex mbranch-expensive
13725 Pretend that branches are expensive. This is for experimenting with
13726 code generation only.
13728 @item -mbranch-cheap
13729 @opindex mbranch-cheap
13730 Do not pretend that branches are expensive. This is the default.
13734 Generate code for a system with split I&D@.
13738 Generate code for a system without split I&D@. This is the default.
13742 Use Unix assembler syntax. This is the default when configured for
13743 @samp{pdp11-*-bsd}.
13747 Use DEC assembler syntax. This is the default when configured for any
13748 PDP-11 target other than @samp{pdp11-*-bsd}.
13751 @node picoChip Options
13752 @subsection picoChip Options
13753 @cindex picoChip options
13755 These @samp{-m} options are defined for picoChip implementations:
13759 @item -mae=@var{ae_type}
13761 Set the instruction set, register set, and instruction scheduling
13762 parameters for array element type @var{ae_type}. Supported values
13763 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
13765 @option{-mae=ANY} selects a completely generic AE type. Code
13766 generated with this option will run on any of the other AE types. The
13767 code will not be as efficient as it would be if compiled for a specific
13768 AE type, and some types of operation (e.g., multiplication) will not
13769 work properly on all types of AE.
13771 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
13772 for compiled code, and is the default.
13774 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
13775 option may suffer from poor performance of byte (char) manipulation,
13776 since the DSP AE does not provide hardware support for byte load/stores.
13778 @item -msymbol-as-address
13779 Enable the compiler to directly use a symbol name as an address in a
13780 load/store instruction, without first loading it into a
13781 register. Typically, the use of this option will generate larger
13782 programs, which run faster than when the option isn't used. However, the
13783 results vary from program to program, so it is left as a user option,
13784 rather than being permanently enabled.
13786 @item -mno-inefficient-warnings
13787 Disables warnings about the generation of inefficient code. These
13788 warnings can be generated, for example, when compiling code which
13789 performs byte-level memory operations on the MAC AE type. The MAC AE has
13790 no hardware support for byte-level memory operations, so all byte
13791 load/stores must be synthesized from word load/store operations. This is
13792 inefficient and a warning will be generated indicating to the programmer
13793 that they should rewrite the code to avoid byte operations, or to target
13794 an AE type which has the necessary hardware support. This option enables
13795 the warning to be turned off.
13799 @node PowerPC Options
13800 @subsection PowerPC Options
13801 @cindex PowerPC options
13803 These are listed under @xref{RS/6000 and PowerPC Options}.
13805 @node RS/6000 and PowerPC Options
13806 @subsection IBM RS/6000 and PowerPC Options
13807 @cindex RS/6000 and PowerPC Options
13808 @cindex IBM RS/6000 and PowerPC Options
13810 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
13817 @itemx -mno-powerpc
13818 @itemx -mpowerpc-gpopt
13819 @itemx -mno-powerpc-gpopt
13820 @itemx -mpowerpc-gfxopt
13821 @itemx -mno-powerpc-gfxopt
13823 @itemx -mno-powerpc64
13827 @itemx -mno-popcntb
13835 @itemx -mno-hard-dfp
13839 @opindex mno-power2
13841 @opindex mno-powerpc
13842 @opindex mpowerpc-gpopt
13843 @opindex mno-powerpc-gpopt
13844 @opindex mpowerpc-gfxopt
13845 @opindex mno-powerpc-gfxopt
13846 @opindex mpowerpc64
13847 @opindex mno-powerpc64
13851 @opindex mno-popcntb
13857 @opindex mno-mfpgpr
13859 @opindex mno-hard-dfp
13860 GCC supports two related instruction set architectures for the
13861 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
13862 instructions supported by the @samp{rios} chip set used in the original
13863 RS/6000 systems and the @dfn{PowerPC} instruction set is the
13864 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
13865 the IBM 4xx, 6xx, and follow-on microprocessors.
13867 Neither architecture is a subset of the other. However there is a
13868 large common subset of instructions supported by both. An MQ
13869 register is included in processors supporting the POWER architecture.
13871 You use these options to specify which instructions are available on the
13872 processor you are using. The default value of these options is
13873 determined when configuring GCC@. Specifying the
13874 @option{-mcpu=@var{cpu_type}} overrides the specification of these
13875 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
13876 rather than the options listed above.
13878 The @option{-mpower} option allows GCC to generate instructions that
13879 are found only in the POWER architecture and to use the MQ register.
13880 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
13881 to generate instructions that are present in the POWER2 architecture but
13882 not the original POWER architecture.
13884 The @option{-mpowerpc} option allows GCC to generate instructions that
13885 are found only in the 32-bit subset of the PowerPC architecture.
13886 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
13887 GCC to use the optional PowerPC architecture instructions in the
13888 General Purpose group, including floating-point square root. Specifying
13889 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
13890 use the optional PowerPC architecture instructions in the Graphics
13891 group, including floating-point select.
13893 The @option{-mmfcrf} option allows GCC to generate the move from
13894 condition register field instruction implemented on the POWER4
13895 processor and other processors that support the PowerPC V2.01
13897 The @option{-mpopcntb} option allows GCC to generate the popcount and
13898 double precision FP reciprocal estimate instruction implemented on the
13899 POWER5 processor and other processors that support the PowerPC V2.02
13901 The @option{-mfprnd} option allows GCC to generate the FP round to
13902 integer instructions implemented on the POWER5+ processor and other
13903 processors that support the PowerPC V2.03 architecture.
13904 The @option{-mcmpb} option allows GCC to generate the compare bytes
13905 instruction implemented on the POWER6 processor and other processors
13906 that support the PowerPC V2.05 architecture.
13907 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
13908 general purpose register instructions implemented on the POWER6X
13909 processor and other processors that support the extended PowerPC V2.05
13911 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
13912 point instructions implemented on some POWER processors.
13914 The @option{-mpowerpc64} option allows GCC to generate the additional
13915 64-bit instructions that are found in the full PowerPC64 architecture
13916 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
13917 @option{-mno-powerpc64}.
13919 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
13920 will use only the instructions in the common subset of both
13921 architectures plus some special AIX common-mode calls, and will not use
13922 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
13923 permits GCC to use any instruction from either architecture and to
13924 allow use of the MQ register; specify this for the Motorola MPC601.
13926 @item -mnew-mnemonics
13927 @itemx -mold-mnemonics
13928 @opindex mnew-mnemonics
13929 @opindex mold-mnemonics
13930 Select which mnemonics to use in the generated assembler code. With
13931 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
13932 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
13933 assembler mnemonics defined for the POWER architecture. Instructions
13934 defined in only one architecture have only one mnemonic; GCC uses that
13935 mnemonic irrespective of which of these options is specified.
13937 GCC defaults to the mnemonics appropriate for the architecture in
13938 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13939 value of these option. Unless you are building a cross-compiler, you
13940 should normally not specify either @option{-mnew-mnemonics} or
13941 @option{-mold-mnemonics}, but should instead accept the default.
13943 @item -mcpu=@var{cpu_type}
13945 Set architecture type, register usage, choice of mnemonics, and
13946 instruction scheduling parameters for machine type @var{cpu_type}.
13947 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13948 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13949 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13950 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13951 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13952 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13953 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13954 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13955 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
13956 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13957 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13959 @option{-mcpu=common} selects a completely generic processor. Code
13960 generated under this option will run on any POWER or PowerPC processor.
13961 GCC will use only the instructions in the common subset of both
13962 architectures, and will not use the MQ register. GCC assumes a generic
13963 processor model for scheduling purposes.
13965 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13966 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13967 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13968 types, with an appropriate, generic processor model assumed for
13969 scheduling purposes.
13971 The other options specify a specific processor. Code generated under
13972 those options will run best on that processor, and may not run at all on
13975 The @option{-mcpu} options automatically enable or disable the
13978 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13979 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13980 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
13981 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr}
13983 The particular options set for any particular CPU will vary between
13984 compiler versions, depending on what setting seems to produce optimal
13985 code for that CPU; it doesn't necessarily reflect the actual hardware's
13986 capabilities. If you wish to set an individual option to a particular
13987 value, you may specify it after the @option{-mcpu} option, like
13988 @samp{-mcpu=970 -mno-altivec}.
13990 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13991 not enabled or disabled by the @option{-mcpu} option at present because
13992 AIX does not have full support for these options. You may still
13993 enable or disable them individually if you're sure it'll work in your
13996 @item -mtune=@var{cpu_type}
13998 Set the instruction scheduling parameters for machine type
13999 @var{cpu_type}, but do not set the architecture type, register usage, or
14000 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14001 values for @var{cpu_type} are used for @option{-mtune} as for
14002 @option{-mcpu}. If both are specified, the code generated will use the
14003 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14004 scheduling parameters set by @option{-mtune}.
14010 Generate code to compute division as reciprocal estimate and iterative
14011 refinement, creating opportunities for increased throughput. This
14012 feature requires: optional PowerPC Graphics instruction set for single
14013 precision and FRE instruction for double precision, assuming divides
14014 cannot generate user-visible traps, and the domain values not include
14015 Infinities, denormals or zero denominator.
14018 @itemx -mno-altivec
14020 @opindex mno-altivec
14021 Generate code that uses (does not use) AltiVec instructions, and also
14022 enable the use of built-in functions that allow more direct access to
14023 the AltiVec instruction set. You may also need to set
14024 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14030 @opindex mno-vrsave
14031 Generate VRSAVE instructions when generating AltiVec code.
14033 @item -mgen-cell-microcode
14034 @opindex mgen-cell-microcode
14035 Generate Cell microcode instructions
14037 @item -mwarn-cell-microcode
14038 @opindex mwarn-cell-microcode
14039 Warning when a Cell microcode instruction is going to emitted. An example
14040 of a Cell microcode instruction is a variable shift.
14043 @opindex msecure-plt
14044 Generate code that allows ld and ld.so to build executables and shared
14045 libraries with non-exec .plt and .got sections. This is a PowerPC
14046 32-bit SYSV ABI option.
14050 Generate code that uses a BSS .plt section that ld.so fills in, and
14051 requires .plt and .got sections that are both writable and executable.
14052 This is a PowerPC 32-bit SYSV ABI option.
14058 This switch enables or disables the generation of ISEL instructions.
14060 @item -misel=@var{yes/no}
14061 This switch has been deprecated. Use @option{-misel} and
14062 @option{-mno-isel} instead.
14068 This switch enables or disables the generation of SPE simd
14074 @opindex mno-paired
14075 This switch enables or disables the generation of PAIRED simd
14078 @item -mspe=@var{yes/no}
14079 This option has been deprecated. Use @option{-mspe} and
14080 @option{-mno-spe} instead.
14082 @item -mfloat-gprs=@var{yes/single/double/no}
14083 @itemx -mfloat-gprs
14084 @opindex mfloat-gprs
14085 This switch enables or disables the generation of floating point
14086 operations on the general purpose registers for architectures that
14089 The argument @var{yes} or @var{single} enables the use of
14090 single-precision floating point operations.
14092 The argument @var{double} enables the use of single and
14093 double-precision floating point operations.
14095 The argument @var{no} disables floating point operations on the
14096 general purpose registers.
14098 This option is currently only available on the MPC854x.
14104 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14105 targets (including GNU/Linux). The 32-bit environment sets int, long
14106 and pointer to 32 bits and generates code that runs on any PowerPC
14107 variant. The 64-bit environment sets int to 32 bits and long and
14108 pointer to 64 bits, and generates code for PowerPC64, as for
14109 @option{-mpowerpc64}.
14112 @itemx -mno-fp-in-toc
14113 @itemx -mno-sum-in-toc
14114 @itemx -mminimal-toc
14116 @opindex mno-fp-in-toc
14117 @opindex mno-sum-in-toc
14118 @opindex mminimal-toc
14119 Modify generation of the TOC (Table Of Contents), which is created for
14120 every executable file. The @option{-mfull-toc} option is selected by
14121 default. In that case, GCC will allocate at least one TOC entry for
14122 each unique non-automatic variable reference in your program. GCC
14123 will also place floating-point constants in the TOC@. However, only
14124 16,384 entries are available in the TOC@.
14126 If you receive a linker error message that saying you have overflowed
14127 the available TOC space, you can reduce the amount of TOC space used
14128 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14129 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14130 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14131 generate code to calculate the sum of an address and a constant at
14132 run-time instead of putting that sum into the TOC@. You may specify one
14133 or both of these options. Each causes GCC to produce very slightly
14134 slower and larger code at the expense of conserving TOC space.
14136 If you still run out of space in the TOC even when you specify both of
14137 these options, specify @option{-mminimal-toc} instead. This option causes
14138 GCC to make only one TOC entry for every file. When you specify this
14139 option, GCC will produce code that is slower and larger but which
14140 uses extremely little TOC space. You may wish to use this option
14141 only on files that contain less frequently executed code.
14147 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14148 @code{long} type, and the infrastructure needed to support them.
14149 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14150 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14151 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14154 @itemx -mno-xl-compat
14155 @opindex mxl-compat
14156 @opindex mno-xl-compat
14157 Produce code that conforms more closely to IBM XL compiler semantics
14158 when using AIX-compatible ABI@. Pass floating-point arguments to
14159 prototyped functions beyond the register save area (RSA) on the stack
14160 in addition to argument FPRs. Do not assume that most significant
14161 double in 128-bit long double value is properly rounded when comparing
14162 values and converting to double. Use XL symbol names for long double
14165 The AIX calling convention was extended but not initially documented to
14166 handle an obscure K&R C case of calling a function that takes the
14167 address of its arguments with fewer arguments than declared. IBM XL
14168 compilers access floating point arguments which do not fit in the
14169 RSA from the stack when a subroutine is compiled without
14170 optimization. Because always storing floating-point arguments on the
14171 stack is inefficient and rarely needed, this option is not enabled by
14172 default and only is necessary when calling subroutines compiled by IBM
14173 XL compilers without optimization.
14177 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14178 application written to use message passing with special startup code to
14179 enable the application to run. The system must have PE installed in the
14180 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14181 must be overridden with the @option{-specs=} option to specify the
14182 appropriate directory location. The Parallel Environment does not
14183 support threads, so the @option{-mpe} option and the @option{-pthread}
14184 option are incompatible.
14186 @item -malign-natural
14187 @itemx -malign-power
14188 @opindex malign-natural
14189 @opindex malign-power
14190 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14191 @option{-malign-natural} overrides the ABI-defined alignment of larger
14192 types, such as floating-point doubles, on their natural size-based boundary.
14193 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14194 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14196 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14200 @itemx -mhard-float
14201 @opindex msoft-float
14202 @opindex mhard-float
14203 Generate code that does not use (uses) the floating-point register set.
14204 Software floating point emulation is provided if you use the
14205 @option{-msoft-float} option, and pass the option to GCC when linking.
14207 @item -msingle-float
14208 @itemx -mdouble-float
14209 @opindex msingle-float
14210 @opindex mdouble-float
14211 Generate code for single or double-precision floating point operations.
14212 @option{-mdouble-float} implies @option{-msingle-float}.
14215 @opindex msimple-fpu
14216 Do not generate sqrt and div instructions for hardware floating point unit.
14220 Specify type of floating point unit. Valid values are @var{sp_lite}
14221 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14222 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14223 and @var{dp_full} (equivalent to -mdouble-float).
14226 @opindex mxilinx-fpu
14227 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14230 @itemx -mno-multiple
14232 @opindex mno-multiple
14233 Generate code that uses (does not use) the load multiple word
14234 instructions and the store multiple word instructions. These
14235 instructions are generated by default on POWER systems, and not
14236 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14237 endian PowerPC systems, since those instructions do not work when the
14238 processor is in little endian mode. The exceptions are PPC740 and
14239 PPC750 which permit the instructions usage in little endian mode.
14244 @opindex mno-string
14245 Generate code that uses (does not use) the load string instructions
14246 and the store string word instructions to save multiple registers and
14247 do small block moves. These instructions are generated by default on
14248 POWER systems, and not generated on PowerPC systems. Do not use
14249 @option{-mstring} on little endian PowerPC systems, since those
14250 instructions do not work when the processor is in little endian mode.
14251 The exceptions are PPC740 and PPC750 which permit the instructions
14252 usage in little endian mode.
14257 @opindex mno-update
14258 Generate code that uses (does not use) the load or store instructions
14259 that update the base register to the address of the calculated memory
14260 location. These instructions are generated by default. If you use
14261 @option{-mno-update}, there is a small window between the time that the
14262 stack pointer is updated and the address of the previous frame is
14263 stored, which means code that walks the stack frame across interrupts or
14264 signals may get corrupted data.
14266 @item -mavoid-indexed-addresses
14267 @item -mno-avoid-indexed-addresses
14268 @opindex mavoid-indexed-addresses
14269 @opindex mno-avoid-indexed-addresses
14270 Generate code that tries to avoid (not avoid) the use of indexed load
14271 or store instructions. These instructions can incur a performance
14272 penalty on Power6 processors in certain situations, such as when
14273 stepping through large arrays that cross a 16M boundary. This option
14274 is enabled by default when targetting Power6 and disabled otherwise.
14277 @itemx -mno-fused-madd
14278 @opindex mfused-madd
14279 @opindex mno-fused-madd
14280 Generate code that uses (does not use) the floating point multiply and
14281 accumulate instructions. These instructions are generated by default if
14282 hardware floating is used.
14288 Generate code that uses (does not use) the half-word multiply and
14289 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
14290 These instructions are generated by default when targetting those
14297 Generate code that uses (does not use) the string-search @samp{dlmzb}
14298 instruction on the IBM 405, 440 and 464 processors. This instruction is
14299 generated by default when targetting those processors.
14301 @item -mno-bit-align
14303 @opindex mno-bit-align
14304 @opindex mbit-align
14305 On System V.4 and embedded PowerPC systems do not (do) force structures
14306 and unions that contain bit-fields to be aligned to the base type of the
14309 For example, by default a structure containing nothing but 8
14310 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
14311 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
14312 the structure would be aligned to a 1 byte boundary and be one byte in
14315 @item -mno-strict-align
14316 @itemx -mstrict-align
14317 @opindex mno-strict-align
14318 @opindex mstrict-align
14319 On System V.4 and embedded PowerPC systems do not (do) assume that
14320 unaligned memory references will be handled by the system.
14322 @item -mrelocatable
14323 @itemx -mno-relocatable
14324 @opindex mrelocatable
14325 @opindex mno-relocatable
14326 On embedded PowerPC systems generate code that allows (does not allow)
14327 the program to be relocated to a different address at runtime. If you
14328 use @option{-mrelocatable} on any module, all objects linked together must
14329 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14331 @item -mrelocatable-lib
14332 @itemx -mno-relocatable-lib
14333 @opindex mrelocatable-lib
14334 @opindex mno-relocatable-lib
14335 On embedded PowerPC systems generate code that allows (does not allow)
14336 the program to be relocated to a different address at runtime. Modules
14337 compiled with @option{-mrelocatable-lib} can be linked with either modules
14338 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
14339 with modules compiled with the @option{-mrelocatable} options.
14345 On System V.4 and embedded PowerPC systems do not (do) assume that
14346 register 2 contains a pointer to a global area pointing to the addresses
14347 used in the program.
14350 @itemx -mlittle-endian
14352 @opindex mlittle-endian
14353 On System V.4 and embedded PowerPC systems compile code for the
14354 processor in little endian mode. The @option{-mlittle-endian} option is
14355 the same as @option{-mlittle}.
14358 @itemx -mbig-endian
14360 @opindex mbig-endian
14361 On System V.4 and embedded PowerPC systems compile code for the
14362 processor in big endian mode. The @option{-mbig-endian} option is
14363 the same as @option{-mbig}.
14365 @item -mdynamic-no-pic
14366 @opindex mdynamic-no-pic
14367 On Darwin and Mac OS X systems, compile code so that it is not
14368 relocatable, but that its external references are relocatable. The
14369 resulting code is suitable for applications, but not shared
14372 @item -mprioritize-restricted-insns=@var{priority}
14373 @opindex mprioritize-restricted-insns
14374 This option controls the priority that is assigned to
14375 dispatch-slot restricted instructions during the second scheduling
14376 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
14377 @var{no/highest/second-highest} priority to dispatch slot restricted
14380 @item -msched-costly-dep=@var{dependence_type}
14381 @opindex msched-costly-dep
14382 This option controls which dependences are considered costly
14383 by the target during instruction scheduling. The argument
14384 @var{dependence_type} takes one of the following values:
14385 @var{no}: no dependence is costly,
14386 @var{all}: all dependences are costly,
14387 @var{true_store_to_load}: a true dependence from store to load is costly,
14388 @var{store_to_load}: any dependence from store to load is costly,
14389 @var{number}: any dependence which latency >= @var{number} is costly.
14391 @item -minsert-sched-nops=@var{scheme}
14392 @opindex minsert-sched-nops
14393 This option controls which nop insertion scheme will be used during
14394 the second scheduling pass. The argument @var{scheme} takes one of the
14396 @var{no}: Don't insert nops.
14397 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
14398 according to the scheduler's grouping.
14399 @var{regroup_exact}: Insert nops to force costly dependent insns into
14400 separate groups. Insert exactly as many nops as needed to force an insn
14401 to a new group, according to the estimated processor grouping.
14402 @var{number}: Insert nops to force costly dependent insns into
14403 separate groups. Insert @var{number} nops to force an insn to a new group.
14406 @opindex mcall-sysv
14407 On System V.4 and embedded PowerPC systems compile code using calling
14408 conventions that adheres to the March 1995 draft of the System V
14409 Application Binary Interface, PowerPC processor supplement. This is the
14410 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
14412 @item -mcall-sysv-eabi
14414 @opindex mcall-sysv-eabi
14415 @opindex mcall-eabi
14416 Specify both @option{-mcall-sysv} and @option{-meabi} options.
14418 @item -mcall-sysv-noeabi
14419 @opindex mcall-sysv-noeabi
14420 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
14422 @item -mcall-aixdesc
14424 On System V.4 and embedded PowerPC systems compile code for the AIX
14428 @opindex mcall-linux
14429 On System V.4 and embedded PowerPC systems compile code for the
14430 Linux-based GNU system.
14434 On System V.4 and embedded PowerPC systems compile code for the
14435 Hurd-based GNU system.
14437 @item -mcall-freebsd
14438 @opindex mcall-freebsd
14439 On System V.4 and embedded PowerPC systems compile code for the
14440 FreeBSD operating system.
14442 @item -mcall-netbsd
14443 @opindex mcall-netbsd
14444 On System V.4 and embedded PowerPC systems compile code for the
14445 NetBSD operating system.
14447 @item -mcall-openbsd
14448 @opindex mcall-netbsd
14449 On System V.4 and embedded PowerPC systems compile code for the
14450 OpenBSD operating system.
14452 @item -maix-struct-return
14453 @opindex maix-struct-return
14454 Return all structures in memory (as specified by the AIX ABI)@.
14456 @item -msvr4-struct-return
14457 @opindex msvr4-struct-return
14458 Return structures smaller than 8 bytes in registers (as specified by the
14461 @item -mabi=@var{abi-type}
14463 Extend the current ABI with a particular extension, or remove such extension.
14464 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
14465 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
14469 Extend the current ABI with SPE ABI extensions. This does not change
14470 the default ABI, instead it adds the SPE ABI extensions to the current
14474 @opindex mabi=no-spe
14475 Disable Booke SPE ABI extensions for the current ABI@.
14477 @item -mabi=ibmlongdouble
14478 @opindex mabi=ibmlongdouble
14479 Change the current ABI to use IBM extended precision long double.
14480 This is a PowerPC 32-bit SYSV ABI option.
14482 @item -mabi=ieeelongdouble
14483 @opindex mabi=ieeelongdouble
14484 Change the current ABI to use IEEE extended precision long double.
14485 This is a PowerPC 32-bit Linux ABI option.
14488 @itemx -mno-prototype
14489 @opindex mprototype
14490 @opindex mno-prototype
14491 On System V.4 and embedded PowerPC systems assume that all calls to
14492 variable argument functions are properly prototyped. Otherwise, the
14493 compiler must insert an instruction before every non prototyped call to
14494 set or clear bit 6 of the condition code register (@var{CR}) to
14495 indicate whether floating point values were passed in the floating point
14496 registers in case the function takes a variable arguments. With
14497 @option{-mprototype}, only calls to prototyped variable argument functions
14498 will set or clear the bit.
14502 On embedded PowerPC systems, assume that the startup module is called
14503 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
14504 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
14509 On embedded PowerPC systems, assume that the startup module is called
14510 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
14515 On embedded PowerPC systems, assume that the startup module is called
14516 @file{crt0.o} and the standard C libraries are @file{libads.a} and
14519 @item -myellowknife
14520 @opindex myellowknife
14521 On embedded PowerPC systems, assume that the startup module is called
14522 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14527 On System V.4 and embedded PowerPC systems, specify that you are
14528 compiling for a VxWorks system.
14532 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14533 header to indicate that @samp{eabi} extended relocations are used.
14539 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14540 Embedded Applications Binary Interface (eabi) which is a set of
14541 modifications to the System V.4 specifications. Selecting @option{-meabi}
14542 means that the stack is aligned to an 8 byte boundary, a function
14543 @code{__eabi} is called to from @code{main} to set up the eabi
14544 environment, and the @option{-msdata} option can use both @code{r2} and
14545 @code{r13} to point to two separate small data areas. Selecting
14546 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14547 do not call an initialization function from @code{main}, and the
14548 @option{-msdata} option will only use @code{r13} to point to a single
14549 small data area. The @option{-meabi} option is on by default if you
14550 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14553 @opindex msdata=eabi
14554 On System V.4 and embedded PowerPC systems, put small initialized
14555 @code{const} global and static data in the @samp{.sdata2} section, which
14556 is pointed to by register @code{r2}. Put small initialized
14557 non-@code{const} global and static data in the @samp{.sdata} section,
14558 which is pointed to by register @code{r13}. Put small uninitialized
14559 global and static data in the @samp{.sbss} section, which is adjacent to
14560 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14561 incompatible with the @option{-mrelocatable} option. The
14562 @option{-msdata=eabi} option also sets the @option{-memb} option.
14565 @opindex msdata=sysv
14566 On System V.4 and embedded PowerPC systems, put small global and static
14567 data in the @samp{.sdata} section, which is pointed to by register
14568 @code{r13}. Put small uninitialized global and static data in the
14569 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14570 The @option{-msdata=sysv} option is incompatible with the
14571 @option{-mrelocatable} option.
14573 @item -msdata=default
14575 @opindex msdata=default
14577 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14578 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14579 same as @option{-msdata=sysv}.
14582 @opindex msdata=data
14583 On System V.4 and embedded PowerPC systems, put small global
14584 data in the @samp{.sdata} section. Put small uninitialized global
14585 data in the @samp{.sbss} section. Do not use register @code{r13}
14586 to address small data however. This is the default behavior unless
14587 other @option{-msdata} options are used.
14591 @opindex msdata=none
14593 On embedded PowerPC systems, put all initialized global and static data
14594 in the @samp{.data} section, and all uninitialized data in the
14595 @samp{.bss} section.
14599 @cindex smaller data references (PowerPC)
14600 @cindex .sdata/.sdata2 references (PowerPC)
14601 On embedded PowerPC systems, put global and static items less than or
14602 equal to @var{num} bytes into the small data or bss sections instead of
14603 the normal data or bss section. By default, @var{num} is 8. The
14604 @option{-G @var{num}} switch is also passed to the linker.
14605 All modules should be compiled with the same @option{-G @var{num}} value.
14608 @itemx -mno-regnames
14610 @opindex mno-regnames
14611 On System V.4 and embedded PowerPC systems do (do not) emit register
14612 names in the assembly language output using symbolic forms.
14615 @itemx -mno-longcall
14617 @opindex mno-longcall
14618 By default assume that all calls are far away so that a longer more
14619 expensive calling sequence is required. This is required for calls
14620 further than 32 megabytes (33,554,432 bytes) from the current location.
14621 A short call will be generated if the compiler knows
14622 the call cannot be that far away. This setting can be overridden by
14623 the @code{shortcall} function attribute, or by @code{#pragma
14626 Some linkers are capable of detecting out-of-range calls and generating
14627 glue code on the fly. On these systems, long calls are unnecessary and
14628 generate slower code. As of this writing, the AIX linker can do this,
14629 as can the GNU linker for PowerPC/64. It is planned to add this feature
14630 to the GNU linker for 32-bit PowerPC systems as well.
14632 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14633 callee, L42'', plus a ``branch island'' (glue code). The two target
14634 addresses represent the callee and the ``branch island''. The
14635 Darwin/PPC linker will prefer the first address and generate a ``bl
14636 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14637 otherwise, the linker will generate ``bl L42'' to call the ``branch
14638 island''. The ``branch island'' is appended to the body of the
14639 calling function; it computes the full 32-bit address of the callee
14642 On Mach-O (Darwin) systems, this option directs the compiler emit to
14643 the glue for every direct call, and the Darwin linker decides whether
14644 to use or discard it.
14646 In the future, we may cause GCC to ignore all longcall specifications
14647 when the linker is known to generate glue.
14649 @item -mtls-markers
14650 @itemx -mno-tls-markers
14651 @opindex mtls-markers
14652 @opindex mno-tls-markers
14653 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
14654 specifying the function argument. The relocation allows ld to
14655 reliably associate function call with argument setup instructions for
14656 TLS optimization, which in turn allows gcc to better schedule the
14661 Adds support for multithreading with the @dfn{pthreads} library.
14662 This option sets flags for both the preprocessor and linker.
14666 @node S/390 and zSeries Options
14667 @subsection S/390 and zSeries Options
14668 @cindex S/390 and zSeries Options
14670 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
14674 @itemx -msoft-float
14675 @opindex mhard-float
14676 @opindex msoft-float
14677 Use (do not use) the hardware floating-point instructions and registers
14678 for floating-point operations. When @option{-msoft-float} is specified,
14679 functions in @file{libgcc.a} will be used to perform floating-point
14680 operations. When @option{-mhard-float} is specified, the compiler
14681 generates IEEE floating-point instructions. This is the default.
14684 @itemx -mno-hard-dfp
14686 @opindex mno-hard-dfp
14687 Use (do not use) the hardware decimal-floating-point instructions for
14688 decimal-floating-point operations. When @option{-mno-hard-dfp} is
14689 specified, functions in @file{libgcc.a} will be used to perform
14690 decimal-floating-point operations. When @option{-mhard-dfp} is
14691 specified, the compiler generates decimal-floating-point hardware
14692 instructions. This is the default for @option{-march=z9-ec} or higher.
14694 @item -mlong-double-64
14695 @itemx -mlong-double-128
14696 @opindex mlong-double-64
14697 @opindex mlong-double-128
14698 These switches control the size of @code{long double} type. A size
14699 of 64bit makes the @code{long double} type equivalent to the @code{double}
14700 type. This is the default.
14703 @itemx -mno-backchain
14704 @opindex mbackchain
14705 @opindex mno-backchain
14706 Store (do not store) the address of the caller's frame as backchain pointer
14707 into the callee's stack frame.
14708 A backchain may be needed to allow debugging using tools that do not understand
14709 DWARF-2 call frame information.
14710 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
14711 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
14712 the backchain is placed into the topmost word of the 96/160 byte register
14715 In general, code compiled with @option{-mbackchain} is call-compatible with
14716 code compiled with @option{-mmo-backchain}; however, use of the backchain
14717 for debugging purposes usually requires that the whole binary is built with
14718 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
14719 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14720 to build a linux kernel use @option{-msoft-float}.
14722 The default is to not maintain the backchain.
14724 @item -mpacked-stack
14725 @itemx -mno-packed-stack
14726 @opindex mpacked-stack
14727 @opindex mno-packed-stack
14728 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
14729 specified, the compiler uses the all fields of the 96/160 byte register save
14730 area only for their default purpose; unused fields still take up stack space.
14731 When @option{-mpacked-stack} is specified, register save slots are densely
14732 packed at the top of the register save area; unused space is reused for other
14733 purposes, allowing for more efficient use of the available stack space.
14734 However, when @option{-mbackchain} is also in effect, the topmost word of
14735 the save area is always used to store the backchain, and the return address
14736 register is always saved two words below the backchain.
14738 As long as the stack frame backchain is not used, code generated with
14739 @option{-mpacked-stack} is call-compatible with code generated with
14740 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
14741 S/390 or zSeries generated code that uses the stack frame backchain at run
14742 time, not just for debugging purposes. Such code is not call-compatible
14743 with code compiled with @option{-mpacked-stack}. Also, note that the
14744 combination of @option{-mbackchain},
14745 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14746 to build a linux kernel use @option{-msoft-float}.
14748 The default is to not use the packed stack layout.
14751 @itemx -mno-small-exec
14752 @opindex msmall-exec
14753 @opindex mno-small-exec
14754 Generate (or do not generate) code using the @code{bras} instruction
14755 to do subroutine calls.
14756 This only works reliably if the total executable size does not
14757 exceed 64k. The default is to use the @code{basr} instruction instead,
14758 which does not have this limitation.
14764 When @option{-m31} is specified, generate code compliant to the
14765 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
14766 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
14767 particular to generate 64-bit instructions. For the @samp{s390}
14768 targets, the default is @option{-m31}, while the @samp{s390x}
14769 targets default to @option{-m64}.
14775 When @option{-mzarch} is specified, generate code using the
14776 instructions available on z/Architecture.
14777 When @option{-mesa} is specified, generate code using the
14778 instructions available on ESA/390. Note that @option{-mesa} is
14779 not possible with @option{-m64}.
14780 When generating code compliant to the GNU/Linux for S/390 ABI,
14781 the default is @option{-mesa}. When generating code compliant
14782 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
14788 Generate (or do not generate) code using the @code{mvcle} instruction
14789 to perform block moves. When @option{-mno-mvcle} is specified,
14790 use a @code{mvc} loop instead. This is the default unless optimizing for
14797 Print (or do not print) additional debug information when compiling.
14798 The default is to not print debug information.
14800 @item -march=@var{cpu-type}
14802 Generate code that will run on @var{cpu-type}, which is the name of a system
14803 representing a certain processor type. Possible values for
14804 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
14805 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
14806 When generating code using the instructions available on z/Architecture,
14807 the default is @option{-march=z900}. Otherwise, the default is
14808 @option{-march=g5}.
14810 @item -mtune=@var{cpu-type}
14812 Tune to @var{cpu-type} everything applicable about the generated code,
14813 except for the ABI and the set of available instructions.
14814 The list of @var{cpu-type} values is the same as for @option{-march}.
14815 The default is the value used for @option{-march}.
14818 @itemx -mno-tpf-trace
14819 @opindex mtpf-trace
14820 @opindex mno-tpf-trace
14821 Generate code that adds (does not add) in TPF OS specific branches to trace
14822 routines in the operating system. This option is off by default, even
14823 when compiling for the TPF OS@.
14826 @itemx -mno-fused-madd
14827 @opindex mfused-madd
14828 @opindex mno-fused-madd
14829 Generate code that uses (does not use) the floating point multiply and
14830 accumulate instructions. These instructions are generated by default if
14831 hardware floating point is used.
14833 @item -mwarn-framesize=@var{framesize}
14834 @opindex mwarn-framesize
14835 Emit a warning if the current function exceeds the given frame size. Because
14836 this is a compile time check it doesn't need to be a real problem when the program
14837 runs. It is intended to identify functions which most probably cause
14838 a stack overflow. It is useful to be used in an environment with limited stack
14839 size e.g.@: the linux kernel.
14841 @item -mwarn-dynamicstack
14842 @opindex mwarn-dynamicstack
14843 Emit a warning if the function calls alloca or uses dynamically
14844 sized arrays. This is generally a bad idea with a limited stack size.
14846 @item -mstack-guard=@var{stack-guard}
14847 @itemx -mstack-size=@var{stack-size}
14848 @opindex mstack-guard
14849 @opindex mstack-size
14850 If these options are provided the s390 back end emits additional instructions in
14851 the function prologue which trigger a trap if the stack size is @var{stack-guard}
14852 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
14853 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
14854 the frame size of the compiled function is chosen.
14855 These options are intended to be used to help debugging stack overflow problems.
14856 The additionally emitted code causes only little overhead and hence can also be
14857 used in production like systems without greater performance degradation. The given
14858 values have to be exact powers of 2 and @var{stack-size} has to be greater than
14859 @var{stack-guard} without exceeding 64k.
14860 In order to be efficient the extra code makes the assumption that the stack starts
14861 at an address aligned to the value given by @var{stack-size}.
14862 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
14865 @node Score Options
14866 @subsection Score Options
14867 @cindex Score Options
14869 These options are defined for Score implementations:
14874 Compile code for big endian mode. This is the default.
14878 Compile code for little endian mode.
14882 Disable generate bcnz instruction.
14886 Enable generate unaligned load and store instruction.
14890 Enable the use of multiply-accumulate instructions. Disabled by default.
14894 Specify the SCORE5 as the target architecture.
14898 Specify the SCORE5U of the target architecture.
14902 Specify the SCORE7 as the target architecture. This is the default.
14906 Specify the SCORE7D as the target architecture.
14910 @subsection SH Options
14912 These @samp{-m} options are defined for the SH implementations:
14917 Generate code for the SH1.
14921 Generate code for the SH2.
14924 Generate code for the SH2e.
14928 Generate code for the SH3.
14932 Generate code for the SH3e.
14936 Generate code for the SH4 without a floating-point unit.
14938 @item -m4-single-only
14939 @opindex m4-single-only
14940 Generate code for the SH4 with a floating-point unit that only
14941 supports single-precision arithmetic.
14945 Generate code for the SH4 assuming the floating-point unit is in
14946 single-precision mode by default.
14950 Generate code for the SH4.
14954 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
14955 floating-point unit is not used.
14957 @item -m4a-single-only
14958 @opindex m4a-single-only
14959 Generate code for the SH4a, in such a way that no double-precision
14960 floating point operations are used.
14963 @opindex m4a-single
14964 Generate code for the SH4a assuming the floating-point unit is in
14965 single-precision mode by default.
14969 Generate code for the SH4a.
14973 Same as @option{-m4a-nofpu}, except that it implicitly passes
14974 @option{-dsp} to the assembler. GCC doesn't generate any DSP
14975 instructions at the moment.
14979 Compile code for the processor in big endian mode.
14983 Compile code for the processor in little endian mode.
14987 Align doubles at 64-bit boundaries. Note that this changes the calling
14988 conventions, and thus some functions from the standard C library will
14989 not work unless you recompile it first with @option{-mdalign}.
14993 Shorten some address references at link time, when possible; uses the
14994 linker option @option{-relax}.
14998 Use 32-bit offsets in @code{switch} tables. The default is to use
15003 Enable the use of bit manipulation instructions on SH2A.
15007 Enable the use of the instruction @code{fmovd}.
15011 Comply with the calling conventions defined by Renesas.
15015 Comply with the calling conventions defined by Renesas.
15019 Comply with the calling conventions defined for GCC before the Renesas
15020 conventions were available. This option is the default for all
15021 targets of the SH toolchain except for @samp{sh-symbianelf}.
15024 @opindex mnomacsave
15025 Mark the @code{MAC} register as call-clobbered, even if
15026 @option{-mhitachi} is given.
15030 Increase IEEE-compliance of floating-point code.
15031 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15032 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15033 comparisons of NANs / infinities incurs extra overhead in every
15034 floating point comparison, therefore the default is set to
15035 @option{-ffinite-math-only}.
15037 @item -minline-ic_invalidate
15038 @opindex minline-ic_invalidate
15039 Inline code to invalidate instruction cache entries after setting up
15040 nested function trampolines.
15041 This option has no effect if -musermode is in effect and the selected
15042 code generation option (e.g. -m4) does not allow the use of the icbi
15044 If the selected code generation option does not allow the use of the icbi
15045 instruction, and -musermode is not in effect, the inlined code will
15046 manipulate the instruction cache address array directly with an associative
15047 write. This not only requires privileged mode, but it will also
15048 fail if the cache line had been mapped via the TLB and has become unmapped.
15052 Dump instruction size and location in the assembly code.
15055 @opindex mpadstruct
15056 This option is deprecated. It pads structures to multiple of 4 bytes,
15057 which is incompatible with the SH ABI@.
15061 Optimize for space instead of speed. Implied by @option{-Os}.
15064 @opindex mprefergot
15065 When generating position-independent code, emit function calls using
15066 the Global Offset Table instead of the Procedure Linkage Table.
15070 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15071 if the inlined code would not work in user mode.
15072 This is the default when the target is @code{sh-*-linux*}.
15074 @item -multcost=@var{number}
15075 @opindex multcost=@var{number}
15076 Set the cost to assume for a multiply insn.
15078 @item -mdiv=@var{strategy}
15079 @opindex mdiv=@var{strategy}
15080 Set the division strategy to use for SHmedia code. @var{strategy} must be
15081 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15082 inv:call2, inv:fp .
15083 "fp" performs the operation in floating point. This has a very high latency,
15084 but needs only a few instructions, so it might be a good choice if
15085 your code has enough easily exploitable ILP to allow the compiler to
15086 schedule the floating point instructions together with other instructions.
15087 Division by zero causes a floating point exception.
15088 "inv" uses integer operations to calculate the inverse of the divisor,
15089 and then multiplies the dividend with the inverse. This strategy allows
15090 cse and hoisting of the inverse calculation. Division by zero calculates
15091 an unspecified result, but does not trap.
15092 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15093 have been found, or if the entire operation has been hoisted to the same
15094 place, the last stages of the inverse calculation are intertwined with the
15095 final multiply to reduce the overall latency, at the expense of using a few
15096 more instructions, and thus offering fewer scheduling opportunities with
15098 "call" calls a library function that usually implements the inv:minlat
15100 This gives high code density for m5-*media-nofpu compilations.
15101 "call2" uses a different entry point of the same library function, where it
15102 assumes that a pointer to a lookup table has already been set up, which
15103 exposes the pointer load to cse / code hoisting optimizations.
15104 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15105 code generation, but if the code stays unoptimized, revert to the "call",
15106 "call2", or "fp" strategies, respectively. Note that the
15107 potentially-trapping side effect of division by zero is carried by a
15108 separate instruction, so it is possible that all the integer instructions
15109 are hoisted out, but the marker for the side effect stays where it is.
15110 A recombination to fp operations or a call is not possible in that case.
15111 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15112 that the inverse calculation was nor separated from the multiply, they speed
15113 up division where the dividend fits into 20 bits (plus sign where applicable),
15114 by inserting a test to skip a number of operations in this case; this test
15115 slows down the case of larger dividends. inv20u assumes the case of a such
15116 a small dividend to be unlikely, and inv20l assumes it to be likely.
15118 @item -mdivsi3_libfunc=@var{name}
15119 @opindex mdivsi3_libfunc=@var{name}
15120 Set the name of the library function used for 32 bit signed division to
15121 @var{name}. This only affect the name used in the call and inv:call
15122 division strategies, and the compiler will still expect the same
15123 sets of input/output/clobbered registers as if this option was not present.
15125 @item -mfixed-range=@var{register-range}
15126 @opindex mfixed-range
15127 Generate code treating the given register range as fixed registers.
15128 A fixed register is one that the register allocator can not use. This is
15129 useful when compiling kernel code. A register range is specified as
15130 two registers separated by a dash. Multiple register ranges can be
15131 specified separated by a comma.
15133 @item -madjust-unroll
15134 @opindex madjust-unroll
15135 Throttle unrolling to avoid thrashing target registers.
15136 This option only has an effect if the gcc code base supports the
15137 TARGET_ADJUST_UNROLL_MAX target hook.
15139 @item -mindexed-addressing
15140 @opindex mindexed-addressing
15141 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15142 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15143 semantics for the indexed addressing mode. The architecture allows the
15144 implementation of processors with 64 bit MMU, which the OS could use to
15145 get 32 bit addressing, but since no current hardware implementation supports
15146 this or any other way to make the indexed addressing mode safe to use in
15147 the 32 bit ABI, the default is -mno-indexed-addressing.
15149 @item -mgettrcost=@var{number}
15150 @opindex mgettrcost=@var{number}
15151 Set the cost assumed for the gettr instruction to @var{number}.
15152 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15156 Assume pt* instructions won't trap. This will generally generate better
15157 scheduled code, but is unsafe on current hardware. The current architecture
15158 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
15159 This has the unintentional effect of making it unsafe to schedule ptabs /
15160 ptrel before a branch, or hoist it out of a loop. For example,
15161 __do_global_ctors, a part of libgcc that runs constructors at program
15162 startup, calls functions in a list which is delimited by @minus{}1. With the
15163 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
15164 That means that all the constructors will be run a bit quicker, but when
15165 the loop comes to the end of the list, the program crashes because ptabs
15166 loads @minus{}1 into a target register. Since this option is unsafe for any
15167 hardware implementing the current architecture specification, the default
15168 is -mno-pt-fixed. Unless the user specifies a specific cost with
15169 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
15170 this deters register allocation using target registers for storing
15173 @item -minvalid-symbols
15174 @opindex minvalid-symbols
15175 Assume symbols might be invalid. Ordinary function symbols generated by
15176 the compiler will always be valid to load with movi/shori/ptabs or
15177 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
15178 to generate symbols that will cause ptabs / ptrel to trap.
15179 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
15180 It will then prevent cross-basic-block cse, hoisting and most scheduling
15181 of symbol loads. The default is @option{-mno-invalid-symbols}.
15184 @node SPARC Options
15185 @subsection SPARC Options
15186 @cindex SPARC options
15188 These @samp{-m} options are supported on the SPARC:
15191 @item -mno-app-regs
15193 @opindex mno-app-regs
15195 Specify @option{-mapp-regs} to generate output using the global registers
15196 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
15199 To be fully SVR4 ABI compliant at the cost of some performance loss,
15200 specify @option{-mno-app-regs}. You should compile libraries and system
15201 software with this option.
15204 @itemx -mhard-float
15206 @opindex mhard-float
15207 Generate output containing floating point instructions. This is the
15211 @itemx -msoft-float
15213 @opindex msoft-float
15214 Generate output containing library calls for floating point.
15215 @strong{Warning:} the requisite libraries are not available for all SPARC
15216 targets. Normally the facilities of the machine's usual C compiler are
15217 used, but this cannot be done directly in cross-compilation. You must make
15218 your own arrangements to provide suitable library functions for
15219 cross-compilation. The embedded targets @samp{sparc-*-aout} and
15220 @samp{sparclite-*-*} do provide software floating point support.
15222 @option{-msoft-float} changes the calling convention in the output file;
15223 therefore, it is only useful if you compile @emph{all} of a program with
15224 this option. In particular, you need to compile @file{libgcc.a}, the
15225 library that comes with GCC, with @option{-msoft-float} in order for
15228 @item -mhard-quad-float
15229 @opindex mhard-quad-float
15230 Generate output containing quad-word (long double) floating point
15233 @item -msoft-quad-float
15234 @opindex msoft-quad-float
15235 Generate output containing library calls for quad-word (long double)
15236 floating point instructions. The functions called are those specified
15237 in the SPARC ABI@. This is the default.
15239 As of this writing, there are no SPARC implementations that have hardware
15240 support for the quad-word floating point instructions. They all invoke
15241 a trap handler for one of these instructions, and then the trap handler
15242 emulates the effect of the instruction. Because of the trap handler overhead,
15243 this is much slower than calling the ABI library routines. Thus the
15244 @option{-msoft-quad-float} option is the default.
15246 @item -mno-unaligned-doubles
15247 @itemx -munaligned-doubles
15248 @opindex mno-unaligned-doubles
15249 @opindex munaligned-doubles
15250 Assume that doubles have 8 byte alignment. This is the default.
15252 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15253 alignment only if they are contained in another type, or if they have an
15254 absolute address. Otherwise, it assumes they have 4 byte alignment.
15255 Specifying this option avoids some rare compatibility problems with code
15256 generated by other compilers. It is not the default because it results
15257 in a performance loss, especially for floating point code.
15259 @item -mno-faster-structs
15260 @itemx -mfaster-structs
15261 @opindex mno-faster-structs
15262 @opindex mfaster-structs
15263 With @option{-mfaster-structs}, the compiler assumes that structures
15264 should have 8 byte alignment. This enables the use of pairs of
15265 @code{ldd} and @code{std} instructions for copies in structure
15266 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15267 However, the use of this changed alignment directly violates the SPARC
15268 ABI@. Thus, it's intended only for use on targets where the developer
15269 acknowledges that their resulting code will not be directly in line with
15270 the rules of the ABI@.
15272 @item -mimpure-text
15273 @opindex mimpure-text
15274 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15275 the compiler to not pass @option{-z text} to the linker when linking a
15276 shared object. Using this option, you can link position-dependent
15277 code into a shared object.
15279 @option{-mimpure-text} suppresses the ``relocations remain against
15280 allocatable but non-writable sections'' linker error message.
15281 However, the necessary relocations will trigger copy-on-write, and the
15282 shared object is not actually shared across processes. Instead of
15283 using @option{-mimpure-text}, you should compile all source code with
15284 @option{-fpic} or @option{-fPIC}.
15286 This option is only available on SunOS and Solaris.
15288 @item -mcpu=@var{cpu_type}
15290 Set the instruction set, register set, and instruction scheduling parameters
15291 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
15292 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
15293 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
15294 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
15295 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
15297 Default instruction scheduling parameters are used for values that select
15298 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
15299 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
15301 Here is a list of each supported architecture and their supported
15306 v8: supersparc, hypersparc
15307 sparclite: f930, f934, sparclite86x
15309 v9: ultrasparc, ultrasparc3, niagara, niagara2
15312 By default (unless configured otherwise), GCC generates code for the V7
15313 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
15314 additionally optimizes it for the Cypress CY7C602 chip, as used in the
15315 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
15316 SPARCStation 1, 2, IPX etc.
15318 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
15319 architecture. The only difference from V7 code is that the compiler emits
15320 the integer multiply and integer divide instructions which exist in SPARC-V8
15321 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
15322 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
15325 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
15326 the SPARC architecture. This adds the integer multiply, integer divide step
15327 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
15328 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
15329 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
15330 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
15331 MB86934 chip, which is the more recent SPARClite with FPU@.
15333 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
15334 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
15335 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
15336 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
15337 optimizes it for the TEMIC SPARClet chip.
15339 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
15340 architecture. This adds 64-bit integer and floating-point move instructions,
15341 3 additional floating-point condition code registers and conditional move
15342 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
15343 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
15344 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
15345 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
15346 @option{-mcpu=niagara}, the compiler additionally optimizes it for
15347 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
15348 additionally optimizes it for Sun UltraSPARC T2 chips.
15350 @item -mtune=@var{cpu_type}
15352 Set the instruction scheduling parameters for machine type
15353 @var{cpu_type}, but do not set the instruction set or register set that the
15354 option @option{-mcpu=@var{cpu_type}} would.
15356 The same values for @option{-mcpu=@var{cpu_type}} can be used for
15357 @option{-mtune=@var{cpu_type}}, but the only useful values are those
15358 that select a particular cpu implementation. Those are @samp{cypress},
15359 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
15360 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
15361 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
15366 @opindex mno-v8plus
15367 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
15368 difference from the V8 ABI is that the global and out registers are
15369 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
15370 mode for all SPARC-V9 processors.
15376 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
15377 Visual Instruction Set extensions. The default is @option{-mno-vis}.
15380 These @samp{-m} options are supported in addition to the above
15381 on SPARC-V9 processors in 64-bit environments:
15384 @item -mlittle-endian
15385 @opindex mlittle-endian
15386 Generate code for a processor running in little-endian mode. It is only
15387 available for a few configurations and most notably not on Solaris and Linux.
15393 Generate code for a 32-bit or 64-bit environment.
15394 The 32-bit environment sets int, long and pointer to 32 bits.
15395 The 64-bit environment sets int to 32 bits and long and pointer
15398 @item -mcmodel=medlow
15399 @opindex mcmodel=medlow
15400 Generate code for the Medium/Low code model: 64-bit addresses, programs
15401 must be linked in the low 32 bits of memory. Programs can be statically
15402 or dynamically linked.
15404 @item -mcmodel=medmid
15405 @opindex mcmodel=medmid
15406 Generate code for the Medium/Middle code model: 64-bit addresses, programs
15407 must be linked in the low 44 bits of memory, the text and data segments must
15408 be less than 2GB in size and the data segment must be located within 2GB of
15411 @item -mcmodel=medany
15412 @opindex mcmodel=medany
15413 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
15414 may be linked anywhere in memory, the text and data segments must be less
15415 than 2GB in size and the data segment must be located within 2GB of the
15418 @item -mcmodel=embmedany
15419 @opindex mcmodel=embmedany
15420 Generate code for the Medium/Anywhere code model for embedded systems:
15421 64-bit addresses, the text and data segments must be less than 2GB in
15422 size, both starting anywhere in memory (determined at link time). The
15423 global register %g4 points to the base of the data segment. Programs
15424 are statically linked and PIC is not supported.
15427 @itemx -mno-stack-bias
15428 @opindex mstack-bias
15429 @opindex mno-stack-bias
15430 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
15431 frame pointer if present, are offset by @minus{}2047 which must be added back
15432 when making stack frame references. This is the default in 64-bit mode.
15433 Otherwise, assume no such offset is present.
15436 These switches are supported in addition to the above on Solaris:
15441 Add support for multithreading using the Solaris threads library. This
15442 option sets flags for both the preprocessor and linker. This option does
15443 not affect the thread safety of object code produced by the compiler or
15444 that of libraries supplied with it.
15448 Add support for multithreading using the POSIX threads library. This
15449 option sets flags for both the preprocessor and linker. This option does
15450 not affect the thread safety of object code produced by the compiler or
15451 that of libraries supplied with it.
15455 This is a synonym for @option{-pthreads}.
15459 @subsection SPU Options
15460 @cindex SPU options
15462 These @samp{-m} options are supported on the SPU:
15466 @itemx -merror-reloc
15467 @opindex mwarn-reloc
15468 @opindex merror-reloc
15470 The loader for SPU does not handle dynamic relocations. By default, GCC
15471 will give an error when it generates code that requires a dynamic
15472 relocation. @option{-mno-error-reloc} disables the error,
15473 @option{-mwarn-reloc} will generate a warning instead.
15476 @itemx -munsafe-dma
15478 @opindex munsafe-dma
15480 Instructions which initiate or test completion of DMA must not be
15481 reordered with respect to loads and stores of the memory which is being
15482 accessed. Users typically address this problem using the volatile
15483 keyword, but that can lead to inefficient code in places where the
15484 memory is known to not change. Rather than mark the memory as volatile
15485 we treat the DMA instructions as potentially effecting all memory. With
15486 @option{-munsafe-dma} users must use the volatile keyword to protect
15489 @item -mbranch-hints
15490 @opindex mbranch-hints
15492 By default, GCC will generate a branch hint instruction to avoid
15493 pipeline stalls for always taken or probably taken branches. A hint
15494 will not be generated closer than 8 instructions away from its branch.
15495 There is little reason to disable them, except for debugging purposes,
15496 or to make an object a little bit smaller.
15500 @opindex msmall-mem
15501 @opindex mlarge-mem
15503 By default, GCC generates code assuming that addresses are never larger
15504 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
15505 a full 32 bit address.
15510 By default, GCC links against startup code that assumes the SPU-style
15511 main function interface (which has an unconventional parameter list).
15512 With @option{-mstdmain}, GCC will link your program against startup
15513 code that assumes a C99-style interface to @code{main}, including a
15514 local copy of @code{argv} strings.
15516 @item -mfixed-range=@var{register-range}
15517 @opindex mfixed-range
15518 Generate code treating the given register range as fixed registers.
15519 A fixed register is one that the register allocator can not use. This is
15520 useful when compiling kernel code. A register range is specified as
15521 two registers separated by a dash. Multiple register ranges can be
15522 specified separated by a comma.
15525 @itemx -mdual-nops=@var{n}
15526 @opindex mdual-nops
15527 By default, GCC will insert nops to increase dual issue when it expects
15528 it to increase performance. @var{n} can be a value from 0 to 10. A
15529 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
15530 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
15532 @item -mhint-max-nops=@var{n}
15533 @opindex mhint-max-nops
15534 Maximum number of nops to insert for a branch hint. A branch hint must
15535 be at least 8 instructions away from the branch it is effecting. GCC
15536 will insert up to @var{n} nops to enforce this, otherwise it will not
15537 generate the branch hint.
15539 @item -mhint-max-distance=@var{n}
15540 @opindex mhint-max-distance
15541 The encoding of the branch hint instruction limits the hint to be within
15542 256 instructions of the branch it is effecting. By default, GCC makes
15543 sure it is within 125.
15546 @opindex msafe-hints
15547 Work around a hardware bug which causes the SPU to stall indefinitely.
15548 By default, GCC will insert the @code{hbrp} instruction to make sure
15549 this stall won't happen.
15553 @node System V Options
15554 @subsection Options for System V
15556 These additional options are available on System V Release 4 for
15557 compatibility with other compilers on those systems:
15562 Create a shared object.
15563 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15567 Identify the versions of each tool used by the compiler, in a
15568 @code{.ident} assembler directive in the output.
15572 Refrain from adding @code{.ident} directives to the output file (this is
15575 @item -YP,@var{dirs}
15577 Search the directories @var{dirs}, and no others, for libraries
15578 specified with @option{-l}.
15580 @item -Ym,@var{dir}
15582 Look in the directory @var{dir} to find the M4 preprocessor.
15583 The assembler uses this option.
15584 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15585 @c the generic assembler that comes with Solaris takes just -Ym.
15589 @subsection V850 Options
15590 @cindex V850 Options
15592 These @samp{-m} options are defined for V850 implementations:
15596 @itemx -mno-long-calls
15597 @opindex mlong-calls
15598 @opindex mno-long-calls
15599 Treat all calls as being far away (near). If calls are assumed to be
15600 far away, the compiler will always load the functions address up into a
15601 register, and call indirect through the pointer.
15607 Do not optimize (do optimize) basic blocks that use the same index
15608 pointer 4 or more times to copy pointer into the @code{ep} register, and
15609 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15610 option is on by default if you optimize.
15612 @item -mno-prolog-function
15613 @itemx -mprolog-function
15614 @opindex mno-prolog-function
15615 @opindex mprolog-function
15616 Do not use (do use) external functions to save and restore registers
15617 at the prologue and epilogue of a function. The external functions
15618 are slower, but use less code space if more than one function saves
15619 the same number of registers. The @option{-mprolog-function} option
15620 is on by default if you optimize.
15624 Try to make the code as small as possible. At present, this just turns
15625 on the @option{-mep} and @option{-mprolog-function} options.
15627 @item -mtda=@var{n}
15629 Put static or global variables whose size is @var{n} bytes or less into
15630 the tiny data area that register @code{ep} points to. The tiny data
15631 area can hold up to 256 bytes in total (128 bytes for byte references).
15633 @item -msda=@var{n}
15635 Put static or global variables whose size is @var{n} bytes or less into
15636 the small data area that register @code{gp} points to. The small data
15637 area can hold up to 64 kilobytes.
15639 @item -mzda=@var{n}
15641 Put static or global variables whose size is @var{n} bytes or less into
15642 the first 32 kilobytes of memory.
15646 Specify that the target processor is the V850.
15649 @opindex mbig-switch
15650 Generate code suitable for big switch tables. Use this option only if
15651 the assembler/linker complain about out of range branches within a switch
15656 This option will cause r2 and r5 to be used in the code generated by
15657 the compiler. This setting is the default.
15659 @item -mno-app-regs
15660 @opindex mno-app-regs
15661 This option will cause r2 and r5 to be treated as fixed registers.
15665 Specify that the target processor is the V850E1. The preprocessor
15666 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
15667 this option is used.
15671 Specify that the target processor is the V850E@. The preprocessor
15672 constant @samp{__v850e__} will be defined if this option is used.
15674 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
15675 are defined then a default target processor will be chosen and the
15676 relevant @samp{__v850*__} preprocessor constant will be defined.
15678 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
15679 defined, regardless of which processor variant is the target.
15681 @item -mdisable-callt
15682 @opindex mdisable-callt
15683 This option will suppress generation of the CALLT instruction for the
15684 v850e and v850e1 flavors of the v850 architecture. The default is
15685 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
15690 @subsection VAX Options
15691 @cindex VAX options
15693 These @samp{-m} options are defined for the VAX:
15698 Do not output certain jump instructions (@code{aobleq} and so on)
15699 that the Unix assembler for the VAX cannot handle across long
15704 Do output those jump instructions, on the assumption that you
15705 will assemble with the GNU assembler.
15709 Output code for g-format floating point numbers instead of d-format.
15712 @node VxWorks Options
15713 @subsection VxWorks Options
15714 @cindex VxWorks Options
15716 The options in this section are defined for all VxWorks targets.
15717 Options specific to the target hardware are listed with the other
15718 options for that target.
15723 GCC can generate code for both VxWorks kernels and real time processes
15724 (RTPs). This option switches from the former to the latter. It also
15725 defines the preprocessor macro @code{__RTP__}.
15728 @opindex non-static
15729 Link an RTP executable against shared libraries rather than static
15730 libraries. The options @option{-static} and @option{-shared} can
15731 also be used for RTPs (@pxref{Link Options}); @option{-static}
15738 These options are passed down to the linker. They are defined for
15739 compatibility with Diab.
15742 @opindex Xbind-lazy
15743 Enable lazy binding of function calls. This option is equivalent to
15744 @option{-Wl,-z,now} and is defined for compatibility with Diab.
15748 Disable lazy binding of function calls. This option is the default and
15749 is defined for compatibility with Diab.
15752 @node x86-64 Options
15753 @subsection x86-64 Options
15754 @cindex x86-64 options
15756 These are listed under @xref{i386 and x86-64 Options}.
15758 @node i386 and x86-64 Windows Options
15759 @subsection i386 and x86-64 Windows Options
15760 @cindex i386 and x86-64 Windows Options
15762 These additional options are available for Windows targets:
15767 This option is available for Cygwin and MinGW targets. It
15768 specifies that a console application is to be generated, by
15769 instructing the linker to set the PE header subsystem type
15770 required for console applications.
15771 This is the default behavior for Cygwin and MinGW targets.
15775 This option is available for Cygwin targets. It specifies that
15776 the Cygwin internal interface is to be used for predefined
15777 preprocessor macros, C runtime libraries and related linker
15778 paths and options. For Cygwin targets this is the default behavior.
15779 This option is deprecated and will be removed in a future release.
15782 @opindex mno-cygwin
15783 This option is available for Cygwin targets. It specifies that
15784 the MinGW internal interface is to be used instead of Cygwin's, by
15785 setting MinGW-related predefined macros and linker paths and default
15787 This option is deprecated and will be removed in a future release.
15791 This option is available for Cygwin and MinGW targets. It
15792 specifies that a DLL - a dynamic link library - is to be
15793 generated, enabling the selection of the required runtime
15794 startup object and entry point.
15796 @item -mnop-fun-dllimport
15797 @opindex mnop-fun-dllimport
15798 This option is available for Cygwin and MinGW targets. It
15799 specifies that the dllimport attribute should be ignored.
15803 This option is available for MinGW targets. It specifies
15804 that MinGW-specific thread support is to be used.
15808 This option is available for mingw-w64 targets. It specifies
15809 that the UNICODE macro is getting pre-defined and that the
15810 unicode capable runtime startup code is choosen.
15814 This option is available for Cygwin and MinGW targets. It
15815 specifies that the typical Windows pre-defined macros are to
15816 be set in the pre-processor, but does not influence the choice
15817 of runtime library/startup code.
15821 This option is available for Cygwin and MinGW targets. It
15822 specifies that a GUI application is to be generated by
15823 instructing the linker to set the PE header subsystem type
15826 @item -mpe-aligned-commons
15827 @opindex mpe-aligned-commons
15828 This option is available for Cygwin and MinGW targets. It
15829 specifies that the GNU extension to the PE file format that
15830 permits the correct alignment of COMMON variables should be
15831 used when generating code. It will be enabled by default if
15832 GCC detects that the target assembler found during configuration
15833 supports the feature.
15836 See also under @ref{i386 and x86-64 Options} for standard options.
15838 @node Xstormy16 Options
15839 @subsection Xstormy16 Options
15840 @cindex Xstormy16 Options
15842 These options are defined for Xstormy16:
15847 Choose startup files and linker script suitable for the simulator.
15850 @node Xtensa Options
15851 @subsection Xtensa Options
15852 @cindex Xtensa Options
15854 These options are supported for Xtensa targets:
15858 @itemx -mno-const16
15860 @opindex mno-const16
15861 Enable or disable use of @code{CONST16} instructions for loading
15862 constant values. The @code{CONST16} instruction is currently not a
15863 standard option from Tensilica. When enabled, @code{CONST16}
15864 instructions are always used in place of the standard @code{L32R}
15865 instructions. The use of @code{CONST16} is enabled by default only if
15866 the @code{L32R} instruction is not available.
15869 @itemx -mno-fused-madd
15870 @opindex mfused-madd
15871 @opindex mno-fused-madd
15872 Enable or disable use of fused multiply/add and multiply/subtract
15873 instructions in the floating-point option. This has no effect if the
15874 floating-point option is not also enabled. Disabling fused multiply/add
15875 and multiply/subtract instructions forces the compiler to use separate
15876 instructions for the multiply and add/subtract operations. This may be
15877 desirable in some cases where strict IEEE 754-compliant results are
15878 required: the fused multiply add/subtract instructions do not round the
15879 intermediate result, thereby producing results with @emph{more} bits of
15880 precision than specified by the IEEE standard. Disabling fused multiply
15881 add/subtract instructions also ensures that the program output is not
15882 sensitive to the compiler's ability to combine multiply and add/subtract
15885 @item -mserialize-volatile
15886 @itemx -mno-serialize-volatile
15887 @opindex mserialize-volatile
15888 @opindex mno-serialize-volatile
15889 When this option is enabled, GCC inserts @code{MEMW} instructions before
15890 @code{volatile} memory references to guarantee sequential consistency.
15891 The default is @option{-mserialize-volatile}. Use
15892 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
15894 @item -mtext-section-literals
15895 @itemx -mno-text-section-literals
15896 @opindex mtext-section-literals
15897 @opindex mno-text-section-literals
15898 Control the treatment of literal pools. The default is
15899 @option{-mno-text-section-literals}, which places literals in a separate
15900 section in the output file. This allows the literal pool to be placed
15901 in a data RAM/ROM, and it also allows the linker to combine literal
15902 pools from separate object files to remove redundant literals and
15903 improve code size. With @option{-mtext-section-literals}, the literals
15904 are interspersed in the text section in order to keep them as close as
15905 possible to their references. This may be necessary for large assembly
15908 @item -mtarget-align
15909 @itemx -mno-target-align
15910 @opindex mtarget-align
15911 @opindex mno-target-align
15912 When this option is enabled, GCC instructs the assembler to
15913 automatically align instructions to reduce branch penalties at the
15914 expense of some code density. The assembler attempts to widen density
15915 instructions to align branch targets and the instructions following call
15916 instructions. If there are not enough preceding safe density
15917 instructions to align a target, no widening will be performed. The
15918 default is @option{-mtarget-align}. These options do not affect the
15919 treatment of auto-aligned instructions like @code{LOOP}, which the
15920 assembler will always align, either by widening density instructions or
15921 by inserting no-op instructions.
15924 @itemx -mno-longcalls
15925 @opindex mlongcalls
15926 @opindex mno-longcalls
15927 When this option is enabled, GCC instructs the assembler to translate
15928 direct calls to indirect calls unless it can determine that the target
15929 of a direct call is in the range allowed by the call instruction. This
15930 translation typically occurs for calls to functions in other source
15931 files. Specifically, the assembler translates a direct @code{CALL}
15932 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
15933 The default is @option{-mno-longcalls}. This option should be used in
15934 programs where the call target can potentially be out of range. This
15935 option is implemented in the assembler, not the compiler, so the
15936 assembly code generated by GCC will still show direct call
15937 instructions---look at the disassembled object code to see the actual
15938 instructions. Note that the assembler will use an indirect call for
15939 every cross-file call, not just those that really will be out of range.
15942 @node zSeries Options
15943 @subsection zSeries Options
15944 @cindex zSeries options
15946 These are listed under @xref{S/390 and zSeries Options}.
15948 @node Code Gen Options
15949 @section Options for Code Generation Conventions
15950 @cindex code generation conventions
15951 @cindex options, code generation
15952 @cindex run-time options
15954 These machine-independent options control the interface conventions
15955 used in code generation.
15957 Most of them have both positive and negative forms; the negative form
15958 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
15959 one of the forms is listed---the one which is not the default. You
15960 can figure out the other form by either removing @samp{no-} or adding
15964 @item -fbounds-check
15965 @opindex fbounds-check
15966 For front-ends that support it, generate additional code to check that
15967 indices used to access arrays are within the declared range. This is
15968 currently only supported by the Java and Fortran front-ends, where
15969 this option defaults to true and false respectively.
15973 This option generates traps for signed overflow on addition, subtraction,
15974 multiplication operations.
15978 This option instructs the compiler to assume that signed arithmetic
15979 overflow of addition, subtraction and multiplication wraps around
15980 using twos-complement representation. This flag enables some optimizations
15981 and disables others. This option is enabled by default for the Java
15982 front-end, as required by the Java language specification.
15985 @opindex fexceptions
15986 Enable exception handling. Generates extra code needed to propagate
15987 exceptions. For some targets, this implies GCC will generate frame
15988 unwind information for all functions, which can produce significant data
15989 size overhead, although it does not affect execution. If you do not
15990 specify this option, GCC will enable it by default for languages like
15991 C++ which normally require exception handling, and disable it for
15992 languages like C that do not normally require it. However, you may need
15993 to enable this option when compiling C code that needs to interoperate
15994 properly with exception handlers written in C++. You may also wish to
15995 disable this option if you are compiling older C++ programs that don't
15996 use exception handling.
15998 @item -fnon-call-exceptions
15999 @opindex fnon-call-exceptions
16000 Generate code that allows trapping instructions to throw exceptions.
16001 Note that this requires platform-specific runtime support that does
16002 not exist everywhere. Moreover, it only allows @emph{trapping}
16003 instructions to throw exceptions, i.e.@: memory references or floating
16004 point instructions. It does not allow exceptions to be thrown from
16005 arbitrary signal handlers such as @code{SIGALRM}.
16007 @item -funwind-tables
16008 @opindex funwind-tables
16009 Similar to @option{-fexceptions}, except that it will just generate any needed
16010 static data, but will not affect the generated code in any other way.
16011 You will normally not enable this option; instead, a language processor
16012 that needs this handling would enable it on your behalf.
16014 @item -fasynchronous-unwind-tables
16015 @opindex fasynchronous-unwind-tables
16016 Generate unwind table in dwarf2 format, if supported by target machine. The
16017 table is exact at each instruction boundary, so it can be used for stack
16018 unwinding from asynchronous events (such as debugger or garbage collector).
16020 @item -fpcc-struct-return
16021 @opindex fpcc-struct-return
16022 Return ``short'' @code{struct} and @code{union} values in memory like
16023 longer ones, rather than in registers. This convention is less
16024 efficient, but it has the advantage of allowing intercallability between
16025 GCC-compiled files and files compiled with other compilers, particularly
16026 the Portable C Compiler (pcc).
16028 The precise convention for returning structures in memory depends
16029 on the target configuration macros.
16031 Short structures and unions are those whose size and alignment match
16032 that of some integer type.
16034 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16035 switch is not binary compatible with code compiled with the
16036 @option{-freg-struct-return} switch.
16037 Use it to conform to a non-default application binary interface.
16039 @item -freg-struct-return
16040 @opindex freg-struct-return
16041 Return @code{struct} and @code{union} values in registers when possible.
16042 This is more efficient for small structures than
16043 @option{-fpcc-struct-return}.
16045 If you specify neither @option{-fpcc-struct-return} nor
16046 @option{-freg-struct-return}, GCC defaults to whichever convention is
16047 standard for the target. If there is no standard convention, GCC
16048 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16049 the principal compiler. In those cases, we can choose the standard, and
16050 we chose the more efficient register return alternative.
16052 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16053 switch is not binary compatible with code compiled with the
16054 @option{-fpcc-struct-return} switch.
16055 Use it to conform to a non-default application binary interface.
16057 @item -fshort-enums
16058 @opindex fshort-enums
16059 Allocate to an @code{enum} type only as many bytes as it needs for the
16060 declared range of possible values. Specifically, the @code{enum} type
16061 will be equivalent to the smallest integer type which has enough room.
16063 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16064 code that is not binary compatible with code generated without that switch.
16065 Use it to conform to a non-default application binary interface.
16067 @item -fshort-double
16068 @opindex fshort-double
16069 Use the same size for @code{double} as for @code{float}.
16071 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16072 code that is not binary compatible with code generated without that switch.
16073 Use it to conform to a non-default application binary interface.
16075 @item -fshort-wchar
16076 @opindex fshort-wchar
16077 Override the underlying type for @samp{wchar_t} to be @samp{short
16078 unsigned int} instead of the default for the target. This option is
16079 useful for building programs to run under WINE@.
16081 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16082 code that is not binary compatible with code generated without that switch.
16083 Use it to conform to a non-default application binary interface.
16086 @opindex fno-common
16087 In C code, controls the placement of uninitialized global variables.
16088 Unix C compilers have traditionally permitted multiple definitions of
16089 such variables in different compilation units by placing the variables
16091 This is the behavior specified by @option{-fcommon}, and is the default
16092 for GCC on most targets.
16093 On the other hand, this behavior is not required by ISO C, and on some
16094 targets may carry a speed or code size penalty on variable references.
16095 The @option{-fno-common} option specifies that the compiler should place
16096 uninitialized global variables in the data section of the object file,
16097 rather than generating them as common blocks.
16098 This has the effect that if the same variable is declared
16099 (without @code{extern}) in two different compilations,
16100 you will get a multiple-definition error when you link them.
16101 In this case, you must compile with @option{-fcommon} instead.
16102 Compiling with @option{-fno-common} is useful on targets for which
16103 it provides better performance, or if you wish to verify that the
16104 program will work on other systems which always treat uninitialized
16105 variable declarations this way.
16109 Ignore the @samp{#ident} directive.
16111 @item -finhibit-size-directive
16112 @opindex finhibit-size-directive
16113 Don't output a @code{.size} assembler directive, or anything else that
16114 would cause trouble if the function is split in the middle, and the
16115 two halves are placed at locations far apart in memory. This option is
16116 used when compiling @file{crtstuff.c}; you should not need to use it
16119 @item -fverbose-asm
16120 @opindex fverbose-asm
16121 Put extra commentary information in the generated assembly code to
16122 make it more readable. This option is generally only of use to those
16123 who actually need to read the generated assembly code (perhaps while
16124 debugging the compiler itself).
16126 @option{-fno-verbose-asm}, the default, causes the
16127 extra information to be omitted and is useful when comparing two assembler
16130 @item -frecord-gcc-switches
16131 @opindex frecord-gcc-switches
16132 This switch causes the command line that was used to invoke the
16133 compiler to be recorded into the object file that is being created.
16134 This switch is only implemented on some targets and the exact format
16135 of the recording is target and binary file format dependent, but it
16136 usually takes the form of a section containing ASCII text. This
16137 switch is related to the @option{-fverbose-asm} switch, but that
16138 switch only records information in the assembler output file as
16139 comments, so it never reaches the object file.
16143 @cindex global offset table
16145 Generate position-independent code (PIC) suitable for use in a shared
16146 library, if supported for the target machine. Such code accesses all
16147 constant addresses through a global offset table (GOT)@. The dynamic
16148 loader resolves the GOT entries when the program starts (the dynamic
16149 loader is not part of GCC; it is part of the operating system). If
16150 the GOT size for the linked executable exceeds a machine-specific
16151 maximum size, you get an error message from the linker indicating that
16152 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
16153 instead. (These maximums are 8k on the SPARC and 32k
16154 on the m68k and RS/6000. The 386 has no such limit.)
16156 Position-independent code requires special support, and therefore works
16157 only on certain machines. For the 386, GCC supports PIC for System V
16158 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
16159 position-independent.
16161 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16166 If supported for the target machine, emit position-independent code,
16167 suitable for dynamic linking and avoiding any limit on the size of the
16168 global offset table. This option makes a difference on the m68k,
16169 PowerPC and SPARC@.
16171 Position-independent code requires special support, and therefore works
16172 only on certain machines.
16174 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16181 These options are similar to @option{-fpic} and @option{-fPIC}, but
16182 generated position independent code can be only linked into executables.
16183 Usually these options are used when @option{-pie} GCC option will be
16184 used during linking.
16186 @option{-fpie} and @option{-fPIE} both define the macros
16187 @code{__pie__} and @code{__PIE__}. The macros have the value 1
16188 for @option{-fpie} and 2 for @option{-fPIE}.
16190 @item -fno-jump-tables
16191 @opindex fno-jump-tables
16192 Do not use jump tables for switch statements even where it would be
16193 more efficient than other code generation strategies. This option is
16194 of use in conjunction with @option{-fpic} or @option{-fPIC} for
16195 building code which forms part of a dynamic linker and cannot
16196 reference the address of a jump table. On some targets, jump tables
16197 do not require a GOT and this option is not needed.
16199 @item -ffixed-@var{reg}
16201 Treat the register named @var{reg} as a fixed register; generated code
16202 should never refer to it (except perhaps as a stack pointer, frame
16203 pointer or in some other fixed role).
16205 @var{reg} must be the name of a register. The register names accepted
16206 are machine-specific and are defined in the @code{REGISTER_NAMES}
16207 macro in the machine description macro file.
16209 This flag does not have a negative form, because it specifies a
16212 @item -fcall-used-@var{reg}
16213 @opindex fcall-used
16214 Treat the register named @var{reg} as an allocable register that is
16215 clobbered by function calls. It may be allocated for temporaries or
16216 variables that do not live across a call. Functions compiled this way
16217 will not save and restore the register @var{reg}.
16219 It is an error to used this flag with the frame pointer or stack pointer.
16220 Use of this flag for other registers that have fixed pervasive roles in
16221 the machine's execution model will produce disastrous results.
16223 This flag does not have a negative form, because it specifies a
16226 @item -fcall-saved-@var{reg}
16227 @opindex fcall-saved
16228 Treat the register named @var{reg} as an allocable register saved by
16229 functions. It may be allocated even for temporaries or variables that
16230 live across a call. Functions compiled this way will save and restore
16231 the register @var{reg} if they use it.
16233 It is an error to used this flag with the frame pointer or stack pointer.
16234 Use of this flag for other registers that have fixed pervasive roles in
16235 the machine's execution model will produce disastrous results.
16237 A different sort of disaster will result from the use of this flag for
16238 a register in which function values may be returned.
16240 This flag does not have a negative form, because it specifies a
16243 @item -fpack-struct[=@var{n}]
16244 @opindex fpack-struct
16245 Without a value specified, pack all structure members together without
16246 holes. When a value is specified (which must be a small power of two), pack
16247 structure members according to this value, representing the maximum
16248 alignment (that is, objects with default alignment requirements larger than
16249 this will be output potentially unaligned at the next fitting location.
16251 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
16252 code that is not binary compatible with code generated without that switch.
16253 Additionally, it makes the code suboptimal.
16254 Use it to conform to a non-default application binary interface.
16256 @item -finstrument-functions
16257 @opindex finstrument-functions
16258 Generate instrumentation calls for entry and exit to functions. Just
16259 after function entry and just before function exit, the following
16260 profiling functions will be called with the address of the current
16261 function and its call site. (On some platforms,
16262 @code{__builtin_return_address} does not work beyond the current
16263 function, so the call site information may not be available to the
16264 profiling functions otherwise.)
16267 void __cyg_profile_func_enter (void *this_fn,
16269 void __cyg_profile_func_exit (void *this_fn,
16273 The first argument is the address of the start of the current function,
16274 which may be looked up exactly in the symbol table.
16276 This instrumentation is also done for functions expanded inline in other
16277 functions. The profiling calls will indicate where, conceptually, the
16278 inline function is entered and exited. This means that addressable
16279 versions of such functions must be available. If all your uses of a
16280 function are expanded inline, this may mean an additional expansion of
16281 code size. If you use @samp{extern inline} in your C code, an
16282 addressable version of such functions must be provided. (This is
16283 normally the case anyways, but if you get lucky and the optimizer always
16284 expands the functions inline, you might have gotten away without
16285 providing static copies.)
16287 A function may be given the attribute @code{no_instrument_function}, in
16288 which case this instrumentation will not be done. This can be used, for
16289 example, for the profiling functions listed above, high-priority
16290 interrupt routines, and any functions from which the profiling functions
16291 cannot safely be called (perhaps signal handlers, if the profiling
16292 routines generate output or allocate memory).
16294 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
16295 @opindex finstrument-functions-exclude-file-list
16297 Set the list of functions that are excluded from instrumentation (see
16298 the description of @code{-finstrument-functions}). If the file that
16299 contains a function definition matches with one of @var{file}, then
16300 that function is not instrumented. The match is done on substrings:
16301 if the @var{file} parameter is a substring of the file name, it is
16302 considered to be a match.
16305 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
16306 will exclude any inline function defined in files whose pathnames
16307 contain @code{/bits/stl} or @code{include/sys}.
16309 If, for some reason, you want to include letter @code{','} in one of
16310 @var{sym}, write @code{'\,'}. For example,
16311 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
16312 (note the single quote surrounding the option).
16314 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
16315 @opindex finstrument-functions-exclude-function-list
16317 This is similar to @code{-finstrument-functions-exclude-file-list},
16318 but this option sets the list of function names to be excluded from
16319 instrumentation. The function name to be matched is its user-visible
16320 name, such as @code{vector<int> blah(const vector<int> &)}, not the
16321 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
16322 match is done on substrings: if the @var{sym} parameter is a substring
16323 of the function name, it is considered to be a match. For C99 and C++
16324 extended identifiers, the function name must be given in UTF-8, not
16325 using universal character names.
16327 @item -fstack-check
16328 @opindex fstack-check
16329 Generate code to verify that you do not go beyond the boundary of the
16330 stack. You should specify this flag if you are running in an
16331 environment with multiple threads, but only rarely need to specify it in
16332 a single-threaded environment since stack overflow is automatically
16333 detected on nearly all systems if there is only one stack.
16335 Note that this switch does not actually cause checking to be done; the
16336 operating system or the language runtime must do that. The switch causes
16337 generation of code to ensure that they see the stack being extended.
16339 You can additionally specify a string parameter: @code{no} means no
16340 checking, @code{generic} means force the use of old-style checking,
16341 @code{specific} means use the best checking method and is equivalent
16342 to bare @option{-fstack-check}.
16344 Old-style checking is a generic mechanism that requires no specific
16345 target support in the compiler but comes with the following drawbacks:
16349 Modified allocation strategy for large objects: they will always be
16350 allocated dynamically if their size exceeds a fixed threshold.
16353 Fixed limit on the size of the static frame of functions: when it is
16354 topped by a particular function, stack checking is not reliable and
16355 a warning is issued by the compiler.
16358 Inefficiency: because of both the modified allocation strategy and the
16359 generic implementation, the performances of the code are hampered.
16362 Note that old-style stack checking is also the fallback method for
16363 @code{specific} if no target support has been added in the compiler.
16365 @item -fstack-limit-register=@var{reg}
16366 @itemx -fstack-limit-symbol=@var{sym}
16367 @itemx -fno-stack-limit
16368 @opindex fstack-limit-register
16369 @opindex fstack-limit-symbol
16370 @opindex fno-stack-limit
16371 Generate code to ensure that the stack does not grow beyond a certain value,
16372 either the value of a register or the address of a symbol. If the stack
16373 would grow beyond the value, a signal is raised. For most targets,
16374 the signal is raised before the stack overruns the boundary, so
16375 it is possible to catch the signal without taking special precautions.
16377 For instance, if the stack starts at absolute address @samp{0x80000000}
16378 and grows downwards, you can use the flags
16379 @option{-fstack-limit-symbol=__stack_limit} and
16380 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
16381 of 128KB@. Note that this may only work with the GNU linker.
16383 @cindex aliasing of parameters
16384 @cindex parameters, aliased
16385 @item -fargument-alias
16386 @itemx -fargument-noalias
16387 @itemx -fargument-noalias-global
16388 @itemx -fargument-noalias-anything
16389 @opindex fargument-alias
16390 @opindex fargument-noalias
16391 @opindex fargument-noalias-global
16392 @opindex fargument-noalias-anything
16393 Specify the possible relationships among parameters and between
16394 parameters and global data.
16396 @option{-fargument-alias} specifies that arguments (parameters) may
16397 alias each other and may alias global storage.@*
16398 @option{-fargument-noalias} specifies that arguments do not alias
16399 each other, but may alias global storage.@*
16400 @option{-fargument-noalias-global} specifies that arguments do not
16401 alias each other and do not alias global storage.
16402 @option{-fargument-noalias-anything} specifies that arguments do not
16403 alias any other storage.
16405 Each language will automatically use whatever option is required by
16406 the language standard. You should not need to use these options yourself.
16408 @item -fleading-underscore
16409 @opindex fleading-underscore
16410 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
16411 change the way C symbols are represented in the object file. One use
16412 is to help link with legacy assembly code.
16414 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
16415 generate code that is not binary compatible with code generated without that
16416 switch. Use it to conform to a non-default application binary interface.
16417 Not all targets provide complete support for this switch.
16419 @item -ftls-model=@var{model}
16420 @opindex ftls-model
16421 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
16422 The @var{model} argument should be one of @code{global-dynamic},
16423 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
16425 The default without @option{-fpic} is @code{initial-exec}; with
16426 @option{-fpic} the default is @code{global-dynamic}.
16428 @item -fvisibility=@var{default|internal|hidden|protected}
16429 @opindex fvisibility
16430 Set the default ELF image symbol visibility to the specified option---all
16431 symbols will be marked with this unless overridden within the code.
16432 Using this feature can very substantially improve linking and
16433 load times of shared object libraries, produce more optimized
16434 code, provide near-perfect API export and prevent symbol clashes.
16435 It is @strong{strongly} recommended that you use this in any shared objects
16438 Despite the nomenclature, @code{default} always means public ie;
16439 available to be linked against from outside the shared object.
16440 @code{protected} and @code{internal} are pretty useless in real-world
16441 usage so the only other commonly used option will be @code{hidden}.
16442 The default if @option{-fvisibility} isn't specified is
16443 @code{default}, i.e., make every
16444 symbol public---this causes the same behavior as previous versions of
16447 A good explanation of the benefits offered by ensuring ELF
16448 symbols have the correct visibility is given by ``How To Write
16449 Shared Libraries'' by Ulrich Drepper (which can be found at
16450 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
16451 solution made possible by this option to marking things hidden when
16452 the default is public is to make the default hidden and mark things
16453 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
16454 and @code{__attribute__ ((visibility("default")))} instead of
16455 @code{__declspec(dllexport)} you get almost identical semantics with
16456 identical syntax. This is a great boon to those working with
16457 cross-platform projects.
16459 For those adding visibility support to existing code, you may find
16460 @samp{#pragma GCC visibility} of use. This works by you enclosing
16461 the declarations you wish to set visibility for with (for example)
16462 @samp{#pragma GCC visibility push(hidden)} and
16463 @samp{#pragma GCC visibility pop}.
16464 Bear in mind that symbol visibility should be viewed @strong{as
16465 part of the API interface contract} and thus all new code should
16466 always specify visibility when it is not the default ie; declarations
16467 only for use within the local DSO should @strong{always} be marked explicitly
16468 as hidden as so to avoid PLT indirection overheads---making this
16469 abundantly clear also aids readability and self-documentation of the code.
16470 Note that due to ISO C++ specification requirements, operator new and
16471 operator delete must always be of default visibility.
16473 Be aware that headers from outside your project, in particular system
16474 headers and headers from any other library you use, may not be
16475 expecting to be compiled with visibility other than the default. You
16476 may need to explicitly say @samp{#pragma GCC visibility push(default)}
16477 before including any such headers.
16479 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
16480 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
16481 no modifications. However, this means that calls to @samp{extern}
16482 functions with no explicit visibility will use the PLT, so it is more
16483 effective to use @samp{__attribute ((visibility))} and/or
16484 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
16485 declarations should be treated as hidden.
16487 Note that @samp{-fvisibility} does affect C++ vague linkage
16488 entities. This means that, for instance, an exception class that will
16489 be thrown between DSOs must be explicitly marked with default
16490 visibility so that the @samp{type_info} nodes will be unified between
16493 An overview of these techniques, their benefits and how to use them
16494 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
16500 @node Environment Variables
16501 @section Environment Variables Affecting GCC
16502 @cindex environment variables
16504 @c man begin ENVIRONMENT
16505 This section describes several environment variables that affect how GCC
16506 operates. Some of them work by specifying directories or prefixes to use
16507 when searching for various kinds of files. Some are used to specify other
16508 aspects of the compilation environment.
16510 Note that you can also specify places to search using options such as
16511 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
16512 take precedence over places specified using environment variables, which
16513 in turn take precedence over those specified by the configuration of GCC@.
16514 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
16515 GNU Compiler Collection (GCC) Internals}.
16520 @c @itemx LC_COLLATE
16522 @c @itemx LC_MONETARY
16523 @c @itemx LC_NUMERIC
16528 @c @findex LC_COLLATE
16529 @findex LC_MESSAGES
16530 @c @findex LC_MONETARY
16531 @c @findex LC_NUMERIC
16535 These environment variables control the way that GCC uses
16536 localization information that allow GCC to work with different
16537 national conventions. GCC inspects the locale categories
16538 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
16539 so. These locale categories can be set to any value supported by your
16540 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
16541 Kingdom encoded in UTF-8.
16543 The @env{LC_CTYPE} environment variable specifies character
16544 classification. GCC uses it to determine the character boundaries in
16545 a string; this is needed for some multibyte encodings that contain quote
16546 and escape characters that would otherwise be interpreted as a string
16549 The @env{LC_MESSAGES} environment variable specifies the language to
16550 use in diagnostic messages.
16552 If the @env{LC_ALL} environment variable is set, it overrides the value
16553 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
16554 and @env{LC_MESSAGES} default to the value of the @env{LANG}
16555 environment variable. If none of these variables are set, GCC
16556 defaults to traditional C English behavior.
16560 If @env{TMPDIR} is set, it specifies the directory to use for temporary
16561 files. GCC uses temporary files to hold the output of one stage of
16562 compilation which is to be used as input to the next stage: for example,
16563 the output of the preprocessor, which is the input to the compiler
16566 @item GCC_EXEC_PREFIX
16567 @findex GCC_EXEC_PREFIX
16568 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
16569 names of the subprograms executed by the compiler. No slash is added
16570 when this prefix is combined with the name of a subprogram, but you can
16571 specify a prefix that ends with a slash if you wish.
16573 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
16574 an appropriate prefix to use based on the pathname it was invoked with.
16576 If GCC cannot find the subprogram using the specified prefix, it
16577 tries looking in the usual places for the subprogram.
16579 The default value of @env{GCC_EXEC_PREFIX} is
16580 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
16581 the installed compiler. In many cases @var{prefix} is the value
16582 of @code{prefix} when you ran the @file{configure} script.
16584 Other prefixes specified with @option{-B} take precedence over this prefix.
16586 This prefix is also used for finding files such as @file{crt0.o} that are
16589 In addition, the prefix is used in an unusual way in finding the
16590 directories to search for header files. For each of the standard
16591 directories whose name normally begins with @samp{/usr/local/lib/gcc}
16592 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
16593 replacing that beginning with the specified prefix to produce an
16594 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
16595 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
16596 These alternate directories are searched first; the standard directories
16597 come next. If a standard directory begins with the configured
16598 @var{prefix} then the value of @var{prefix} is replaced by
16599 @env{GCC_EXEC_PREFIX} when looking for header files.
16601 @item COMPILER_PATH
16602 @findex COMPILER_PATH
16603 The value of @env{COMPILER_PATH} is a colon-separated list of
16604 directories, much like @env{PATH}. GCC tries the directories thus
16605 specified when searching for subprograms, if it can't find the
16606 subprograms using @env{GCC_EXEC_PREFIX}.
16609 @findex LIBRARY_PATH
16610 The value of @env{LIBRARY_PATH} is a colon-separated list of
16611 directories, much like @env{PATH}. When configured as a native compiler,
16612 GCC tries the directories thus specified when searching for special
16613 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
16614 using GCC also uses these directories when searching for ordinary
16615 libraries for the @option{-l} option (but directories specified with
16616 @option{-L} come first).
16620 @cindex locale definition
16621 This variable is used to pass locale information to the compiler. One way in
16622 which this information is used is to determine the character set to be used
16623 when character literals, string literals and comments are parsed in C and C++.
16624 When the compiler is configured to allow multibyte characters,
16625 the following values for @env{LANG} are recognized:
16629 Recognize JIS characters.
16631 Recognize SJIS characters.
16633 Recognize EUCJP characters.
16636 If @env{LANG} is not defined, or if it has some other value, then the
16637 compiler will use mblen and mbtowc as defined by the default locale to
16638 recognize and translate multibyte characters.
16642 Some additional environments variables affect the behavior of the
16645 @include cppenv.texi
16649 @node Precompiled Headers
16650 @section Using Precompiled Headers
16651 @cindex precompiled headers
16652 @cindex speed of compilation
16654 Often large projects have many header files that are included in every
16655 source file. The time the compiler takes to process these header files
16656 over and over again can account for nearly all of the time required to
16657 build the project. To make builds faster, GCC allows users to
16658 `precompile' a header file; then, if builds can use the precompiled
16659 header file they will be much faster.
16661 To create a precompiled header file, simply compile it as you would any
16662 other file, if necessary using the @option{-x} option to make the driver
16663 treat it as a C or C++ header file. You will probably want to use a
16664 tool like @command{make} to keep the precompiled header up-to-date when
16665 the headers it contains change.
16667 A precompiled header file will be searched for when @code{#include} is
16668 seen in the compilation. As it searches for the included file
16669 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
16670 compiler looks for a precompiled header in each directory just before it
16671 looks for the include file in that directory. The name searched for is
16672 the name specified in the @code{#include} with @samp{.gch} appended. If
16673 the precompiled header file can't be used, it is ignored.
16675 For instance, if you have @code{#include "all.h"}, and you have
16676 @file{all.h.gch} in the same directory as @file{all.h}, then the
16677 precompiled header file will be used if possible, and the original
16678 header will be used otherwise.
16680 Alternatively, you might decide to put the precompiled header file in a
16681 directory and use @option{-I} to ensure that directory is searched
16682 before (or instead of) the directory containing the original header.
16683 Then, if you want to check that the precompiled header file is always
16684 used, you can put a file of the same name as the original header in this
16685 directory containing an @code{#error} command.
16687 This also works with @option{-include}. So yet another way to use
16688 precompiled headers, good for projects not designed with precompiled
16689 header files in mind, is to simply take most of the header files used by
16690 a project, include them from another header file, precompile that header
16691 file, and @option{-include} the precompiled header. If the header files
16692 have guards against multiple inclusion, they will be skipped because
16693 they've already been included (in the precompiled header).
16695 If you need to precompile the same header file for different
16696 languages, targets, or compiler options, you can instead make a
16697 @emph{directory} named like @file{all.h.gch}, and put each precompiled
16698 header in the directory, perhaps using @option{-o}. It doesn't matter
16699 what you call the files in the directory, every precompiled header in
16700 the directory will be considered. The first precompiled header
16701 encountered in the directory that is valid for this compilation will
16702 be used; they're searched in no particular order.
16704 There are many other possibilities, limited only by your imagination,
16705 good sense, and the constraints of your build system.
16707 A precompiled header file can be used only when these conditions apply:
16711 Only one precompiled header can be used in a particular compilation.
16714 A precompiled header can't be used once the first C token is seen. You
16715 can have preprocessor directives before a precompiled header; you can
16716 even include a precompiled header from inside another header, so long as
16717 there are no C tokens before the @code{#include}.
16720 The precompiled header file must be produced for the same language as
16721 the current compilation. You can't use a C precompiled header for a C++
16725 The precompiled header file must have been produced by the same compiler
16726 binary as the current compilation is using.
16729 Any macros defined before the precompiled header is included must
16730 either be defined in the same way as when the precompiled header was
16731 generated, or must not affect the precompiled header, which usually
16732 means that they don't appear in the precompiled header at all.
16734 The @option{-D} option is one way to define a macro before a
16735 precompiled header is included; using a @code{#define} can also do it.
16736 There are also some options that define macros implicitly, like
16737 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
16740 @item If debugging information is output when using the precompiled
16741 header, using @option{-g} or similar, the same kind of debugging information
16742 must have been output when building the precompiled header. However,
16743 a precompiled header built using @option{-g} can be used in a compilation
16744 when no debugging information is being output.
16746 @item The same @option{-m} options must generally be used when building
16747 and using the precompiled header. @xref{Submodel Options},
16748 for any cases where this rule is relaxed.
16750 @item Each of the following options must be the same when building and using
16751 the precompiled header:
16753 @gccoptlist{-fexceptions}
16756 Some other command-line options starting with @option{-f},
16757 @option{-p}, or @option{-O} must be defined in the same way as when
16758 the precompiled header was generated. At present, it's not clear
16759 which options are safe to change and which are not; the safest choice
16760 is to use exactly the same options when generating and using the
16761 precompiled header. The following are known to be safe:
16763 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
16764 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
16765 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
16770 For all of these except the last, the compiler will automatically
16771 ignore the precompiled header if the conditions aren't met. If you
16772 find an option combination that doesn't work and doesn't cause the
16773 precompiled header to be ignored, please consider filing a bug report,
16776 If you do use differing options when generating and using the
16777 precompiled header, the actual behavior will be a mixture of the
16778 behavior for the options. For instance, if you use @option{-g} to
16779 generate the precompiled header but not when using it, you may or may
16780 not get debugging information for routines in the precompiled header.