1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007
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,
15 2007 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{-dr} 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.
151 * Running Protoize:: Automatically adding or removing function prototypes.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{]} --target-help @gol
167 --version @@@var{file}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @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++ -Wno-deprecated -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 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
232 -Wchar-subscripts -Wclobbered -Wcomment @gol
233 -Wconversion -Wcoverage-mismatch -Wno-deprecated-declarations @gol
234 -Wdisabled-optimization -Wno-div-by-zero @gol
235 -Wempty-body -Wno-endif-labels @gol
236 -Werror -Werror=* @gol
237 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
238 -Wno-format-extra-args -Wformat-nonliteral @gol
239 -Wformat-security -Wformat-y2k @gol
240 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
241 -Wimport -Wno-import -Winit-self -Winline @gol
242 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
243 -Winvalid-pch -Wlarger-than-@var{len} -Wunsafe-loop-optimizations @gol
244 -Wlogical-op -Wlong-long @gol
245 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
246 -Wmissing-format-attribute -Wmissing-include-dirs @gol
247 -Wmissing-noreturn @gol
248 -Wno-multichar -Wnonnull -Wno-overflow @gol
249 -Woverlength-strings -Wpacked -Wpadded @gol
250 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
251 -Wredundant-decls @gol
252 -Wreturn-type -Wsequence-point -Wshadow @gol
253 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
254 -Wstrict-aliasing -Wstrict-aliasing=n @gol
255 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
256 -Wswitch -Wswitch-default -Wswitch-enum @gol
257 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
258 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
259 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
260 -Wunused-value -Wunused-variable @gol
261 -Wvariadic-macros -Wvla @gol
262 -Wvolatile-register-var -Wwrite-strings}
264 @item C-only Warning Options
265 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
266 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
267 -Wold-style-declaration -Wold-style-definition @gol
268 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
269 -Wdeclaration-after-statement -Wpointer-sign}
271 @item Debugging Options
272 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
273 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
274 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
275 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
276 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
277 -fdump-ipa-all -fdump-ipa-cgraph @gol
279 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
282 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
284 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-nrv -fdump-tree-vect @gol
293 -fdump-tree-sink @gol
294 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-salias @gol
296 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
298 -ftree-vectorizer-verbose=@var{n} @gol
299 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
300 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
301 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
302 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
303 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
304 -ftest-coverage -ftime-report -fvar-tracking @gol
305 -g -g@var{level} -gcoff -gdwarf-2 @gol
306 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
307 -fdebug-prefix-map=@var{old}=@var{new} @gol
308 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
309 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
310 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
311 -print-multi-directory -print-multi-lib @gol
312 -print-prog-name=@var{program} -print-search-dirs -Q @gol
313 -print-sysroot-headers-suffix @gol
316 @item Optimization Options
317 @xref{Optimize Options,,Options that Control Optimization}.
318 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
319 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
320 -fassociative-math -freciprocal-math @gol
321 -fmudflap -fmudflapth -fmudflapir @gol
322 -fbranch-probabilities -fprofile-values -fvpt @gol
323 -fbranch-target-load-optimize @gol
324 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
325 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
326 -fcse-skip-blocks -fcx-limited-range -fdata-sections -fdce @gol
327 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fearly-inlining @gol
328 -fexpensive-optimizations -ffast-math -ffloat-store @gol
329 -fforce-addr -fforward-propagate -ffunction-sections @gol
330 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
331 -fcrossjumping -fif-conversion -fif-conversion2 @gol
332 -finline-functions -finline-functions-called-once @gol
333 -finline-small-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
334 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
335 -fmodulo-sched -fmodulo-sched-allow-regmoves -fno-branch-count-reg @gol
336 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
337 -fno-function-cse -fno-guess-branch-probability @gol
338 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
339 -funsafe-math-optimizations -funsafe-loop-optimizations @gol
340 -ffinite-math-only -fno-signed-zeros @gol
341 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
342 -fomit-frame-pointer -foptimize-register-move @gol
343 -foptimize-sibling-calls -fpredictive-commoning -fprefetch-loop-arrays @gol
344 -fprofile-generate -fprofile-use @gol
345 -fregmove -frename-registers @gol
346 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
347 -frerun-cse-after-loop @gol
348 -frounding-math -frtl-abstract-sequences @gol
349 -fschedule-insns -fschedule-insns2 @gol
350 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
351 -fsched-spec-load-dangerous @gol
352 -fsched-stalled-insns[=@var{n}] -fsched-stalled-insns-dep[=@var{n}] @gol
353 -fsched2-use-superblocks @gol
354 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
355 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
356 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
357 -fstrict-aliasing -fstrict-overflow -ftracer -fthread-jumps @gol
358 -funroll-all-loops -funroll-loops -fpeel-loops @gol
359 -fsplit-ivs-in-unroller -funswitch-loops @gol
360 -fvariable-expansion-in-unroller -ftree-reassoc @gol
361 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
362 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
363 -fcheck-data-deps -ftree-parallelize-loops=@var{n} @gol
364 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
365 -ftree-ch -ftree-sra -ftree-ter -ftree-fre -ftree-vectorize @gol
366 -ftree-vect-loop-version -fvect-cost-model -ftree-salias -fipa-pta -fweb @gol
367 -ftree-copy-prop -ftree-store-ccp -ftree-vrp @gol
368 -funit-at-a-time -fwhole-program @gol
369 --param @var{name}=@var{value}
370 -O -O0 -O1 -O2 -O3 -Os}
372 @item Preprocessor Options
373 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
374 @gccoptlist{-A@var{question}=@var{answer} @gol
375 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
376 -C -dD -dI -dM -dN @gol
377 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
378 -idirafter @var{dir} @gol
379 -include @var{file} -imacros @var{file} @gol
380 -iprefix @var{file} -iwithprefix @var{dir} @gol
381 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
382 -imultilib @var{dir} -isysroot @var{dir} @gol
383 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
384 -P -fworking-directory -remap @gol
385 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
386 -Xpreprocessor @var{option}}
388 @item Assembler Option
389 @xref{Assembler Options,,Passing Options to the Assembler}.
390 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
393 @xref{Link Options,,Options for Linking}.
394 @gccoptlist{@var{object-file-name} -l@var{library} @gol
395 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
396 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
397 -Wl,@var{option} -Xlinker @var{option} @gol
400 @item Directory Options
401 @xref{Directory Options,,Options for Directory Search}.
402 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
403 -specs=@var{file} -I- --sysroot=@var{dir}}
406 @c I wrote this xref this way to avoid overfull hbox. -- rms
407 @xref{Target Options}.
408 @gccoptlist{-V @var{version} -b @var{machine}}
410 @item Machine Dependent Options
411 @xref{Submodel Options,,Hardware Models and Configurations}.
412 @c This list is ordered alphanumerically by subsection name.
413 @c Try and put the significant identifier (CPU or system) first,
414 @c so users have a clue at guessing where the ones they want will be.
417 @gccoptlist{-EB -EL @gol
418 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
419 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
422 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
423 -mabi=@var{name} @gol
424 -mapcs-stack-check -mno-apcs-stack-check @gol
425 -mapcs-float -mno-apcs-float @gol
426 -mapcs-reentrant -mno-apcs-reentrant @gol
427 -msched-prolog -mno-sched-prolog @gol
428 -mlittle-endian -mbig-endian -mwords-little-endian @gol
429 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
430 -mthumb-interwork -mno-thumb-interwork @gol
431 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
432 -mstructure-size-boundary=@var{n} @gol
433 -mabort-on-noreturn @gol
434 -mlong-calls -mno-long-calls @gol
435 -msingle-pic-base -mno-single-pic-base @gol
436 -mpic-register=@var{reg} @gol
437 -mnop-fun-dllimport @gol
438 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
439 -mpoke-function-name @gol
441 -mtpcs-frame -mtpcs-leaf-frame @gol
442 -mcaller-super-interworking -mcallee-super-interworking @gol
446 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
447 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
449 @emph{Blackfin Options}
450 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
451 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
452 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
453 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
454 -mno-id-shared-library -mshared-library-id=@var{n} @gol
455 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
456 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
457 -mfast-fp -minline-plt}
460 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
461 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
462 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
463 -mstack-align -mdata-align -mconst-align @gol
464 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
465 -melf -maout -melinux -mlinux -sim -sim2 @gol
466 -mmul-bug-workaround -mno-mul-bug-workaround}
469 @gccoptlist{-mmac -mpush-args}
471 @emph{Darwin Options}
472 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
473 -arch_only -bind_at_load -bundle -bundle_loader @gol
474 -client_name -compatibility_version -current_version @gol
476 -dependency-file -dylib_file -dylinker_install_name @gol
477 -dynamic -dynamiclib -exported_symbols_list @gol
478 -filelist -flat_namespace -force_cpusubtype_ALL @gol
479 -force_flat_namespace -headerpad_max_install_names @gol
481 -image_base -init -install_name -keep_private_externs @gol
482 -multi_module -multiply_defined -multiply_defined_unused @gol
483 -noall_load -no_dead_strip_inits_and_terms @gol
484 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
485 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
486 -private_bundle -read_only_relocs -sectalign @gol
487 -sectobjectsymbols -whyload -seg1addr @gol
488 -sectcreate -sectobjectsymbols -sectorder @gol
489 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
490 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
491 -segprot -segs_read_only_addr -segs_read_write_addr @gol
492 -single_module -static -sub_library -sub_umbrella @gol
493 -twolevel_namespace -umbrella -undefined @gol
494 -unexported_symbols_list -weak_reference_mismatches @gol
495 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
496 -mkernel -mone-byte-bool}
498 @emph{DEC Alpha Options}
499 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
500 -mieee -mieee-with-inexact -mieee-conformant @gol
501 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
502 -mtrap-precision=@var{mode} -mbuild-constants @gol
503 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
504 -mbwx -mmax -mfix -mcix @gol
505 -mfloat-vax -mfloat-ieee @gol
506 -mexplicit-relocs -msmall-data -mlarge-data @gol
507 -msmall-text -mlarge-text @gol
508 -mmemory-latency=@var{time}}
510 @emph{DEC Alpha/VMS Options}
511 @gccoptlist{-mvms-return-codes}
514 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
515 -mhard-float -msoft-float @gol
516 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
517 -mdouble -mno-double @gol
518 -mmedia -mno-media -mmuladd -mno-muladd @gol
519 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
520 -mlinked-fp -mlong-calls -malign-labels @gol
521 -mlibrary-pic -macc-4 -macc-8 @gol
522 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
523 -moptimize-membar -mno-optimize-membar @gol
524 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
525 -mvliw-branch -mno-vliw-branch @gol
526 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
527 -mno-nested-cond-exec -mtomcat-stats @gol
531 @emph{GNU/Linux Options}
532 @gccoptlist{-muclibc}
534 @emph{H8/300 Options}
535 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
538 @gccoptlist{-march=@var{architecture-type} @gol
539 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
540 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
541 -mfixed-range=@var{register-range} @gol
542 -mjump-in-delay -mlinker-opt -mlong-calls @gol
543 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
544 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
545 -mno-jump-in-delay -mno-long-load-store @gol
546 -mno-portable-runtime -mno-soft-float @gol
547 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
548 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
549 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
550 -munix=@var{unix-std} -nolibdld -static -threads}
552 @emph{i386 and x86-64 Options}
553 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
554 -mfpmath=@var{unit} @gol
555 -masm=@var{dialect} -mno-fancy-math-387 @gol
556 -mno-fp-ret-in-387 -msoft-float @gol
557 -mno-wide-multiply -mrtd -malign-double @gol
558 -mpreferred-stack-boundary=@var{num} -mcx16 -msahf -mrecip @gol
559 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 @gol
560 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
561 -mthreads -mno-align-stringops -minline-all-stringops @gol
562 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
563 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
564 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
565 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
566 -mcmodel=@var{code-model} @gol
567 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
568 -mfused-madd -mno-fused-madd}
571 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
572 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
573 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
574 -minline-float-divide-max-throughput @gol
575 -minline-int-divide-min-latency @gol
576 -minline-int-divide-max-throughput @gol
577 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
578 -mno-dwarf2-asm -mearly-stop-bits @gol
579 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
580 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
581 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
582 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
583 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
584 -mno-sched-prefer-non-data-spec-insns @gol
585 -mno-sched-prefer-non-control-spec-insns @gol
586 -mno-sched-count-spec-in-critical-path}
588 @emph{M32R/D Options}
589 @gccoptlist{-m32r2 -m32rx -m32r @gol
591 -malign-loops -mno-align-loops @gol
592 -missue-rate=@var{number} @gol
593 -mbranch-cost=@var{number} @gol
594 -mmodel=@var{code-size-model-type} @gol
595 -msdata=@var{sdata-type} @gol
596 -mno-flush-func -mflush-func=@var{name} @gol
597 -mno-flush-trap -mflush-trap=@var{number} @gol
601 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
603 @emph{M680x0 Options}
604 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
605 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
606 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
607 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
608 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
609 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
610 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
611 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
613 @emph{M68hc1x Options}
614 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
615 -mauto-incdec -minmax -mlong-calls -mshort @gol
616 -msoft-reg-count=@var{count}}
619 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
620 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
621 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
622 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
623 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
626 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
627 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
628 -mips16 -mno-mips16 -mflip-mips16 @gol
629 -minterlink-mips16 -mno-interlink-mips16 @gol
630 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
631 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
632 -mfp32 -mfp64 -mhard-float -msoft-float @gol
633 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
634 -msmartmips -mno-smartmips @gol
635 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
636 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
637 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
638 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
639 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
640 -membedded-data -mno-embedded-data @gol
641 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
642 -mcode-readable=@var{setting} @gol
643 -msplit-addresses -mno-split-addresses @gol
644 -mexplicit-relocs -mno-explicit-relocs @gol
645 -mcheck-zero-division -mno-check-zero-division @gol
646 -mdivide-traps -mdivide-breaks @gol
647 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
648 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
649 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
650 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130 @gol
651 -mfix-sb1 -mno-fix-sb1 @gol
652 -mflush-func=@var{func} -mno-flush-func @gol
653 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
654 -mfp-exceptions -mno-fp-exceptions @gol
655 -mvr4130-align -mno-vr4130-align}
658 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
659 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
660 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
661 -mno-base-addresses -msingle-exit -mno-single-exit}
663 @emph{MN10300 Options}
664 @gccoptlist{-mmult-bug -mno-mult-bug @gol
665 -mam33 -mno-am33 @gol
666 -mam33-2 -mno-am33-2 @gol
667 -mreturn-pointer-on-d0 @gol
671 @gccoptlist{-mno-crt0 -mbacc -msim @gol
672 -march=@var{cpu-type} }
674 @emph{PDP-11 Options}
675 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
676 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
677 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
678 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
679 -mbranch-expensive -mbranch-cheap @gol
680 -msplit -mno-split -munix-asm -mdec-asm}
682 @emph{PowerPC Options}
683 See RS/6000 and PowerPC Options.
685 @emph{RS/6000 and PowerPC Options}
686 @gccoptlist{-mcpu=@var{cpu-type} @gol
687 -mtune=@var{cpu-type} @gol
688 -mpower -mno-power -mpower2 -mno-power2 @gol
689 -mpowerpc -mpowerpc64 -mno-powerpc @gol
690 -maltivec -mno-altivec @gol
691 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
692 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
693 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
694 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mdfp -mno-dfp @gol
695 -mnew-mnemonics -mold-mnemonics @gol
696 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
697 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
698 -malign-power -malign-natural @gol
699 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
700 -mstring -mno-string -mupdate -mno-update @gol
701 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
702 -mstrict-align -mno-strict-align -mrelocatable @gol
703 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
704 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
705 -mdynamic-no-pic -maltivec -mswdiv @gol
706 -mprioritize-restricted-insns=@var{priority} @gol
707 -msched-costly-dep=@var{dependence_type} @gol
708 -minsert-sched-nops=@var{scheme} @gol
709 -mcall-sysv -mcall-netbsd @gol
710 -maix-struct-return -msvr4-struct-return @gol
711 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
712 -misel -mno-isel @gol
713 -misel=yes -misel=no @gol
715 -mspe=yes -mspe=no @gol
717 -mvrsave -mno-vrsave @gol
718 -mmulhw -mno-mulhw @gol
719 -mdlmzb -mno-dlmzb @gol
720 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
721 -mprototype -mno-prototype @gol
722 -msim -mmvme -mads -myellowknife -memb -msdata @gol
723 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
725 @emph{S/390 and zSeries Options}
726 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
727 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
728 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
729 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
730 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
731 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
732 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
735 @gccoptlist{-meb -mel @gol
739 -mscore5 -mscore5u -mscore7 -mscore7d}
742 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
743 -m4-nofpu -m4-single-only -m4-single -m4 @gol
744 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
745 -m5-64media -m5-64media-nofpu @gol
746 -m5-32media -m5-32media-nofpu @gol
747 -m5-compact -m5-compact-nofpu @gol
748 -mb -ml -mdalign -mrelax @gol
749 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
750 -mieee -misize -minline-ic_invalidate -mpadstruct -mspace @gol
751 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
752 -mdivsi3_libfunc=@var{name} @gol
753 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
757 @gccoptlist{-mcpu=@var{cpu-type} @gol
758 -mtune=@var{cpu-type} @gol
759 -mcmodel=@var{code-model} @gol
760 -m32 -m64 -mapp-regs -mno-app-regs @gol
761 -mfaster-structs -mno-faster-structs @gol
762 -mfpu -mno-fpu -mhard-float -msoft-float @gol
763 -mhard-quad-float -msoft-quad-float @gol
764 -mimpure-text -mno-impure-text -mlittle-endian @gol
765 -mstack-bias -mno-stack-bias @gol
766 -munaligned-doubles -mno-unaligned-doubles @gol
767 -mv8plus -mno-v8plus -mvis -mno-vis
768 -threads -pthreads -pthread}
771 @gccoptlist{-mwarn-reloc -merror-reloc @gol
772 -msafe-dma -munsafe-dma @gol
774 -msmall-mem -mlarge-mem -mstdmain @gol
775 -mfixed-range=@var{register-range}}
777 @emph{System V Options}
778 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
780 @emph{TMS320C3x/C4x Options}
781 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
782 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
783 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
784 -mparallel-insns -mparallel-mpy -mpreserve-float}
787 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
788 -mprolog-function -mno-prolog-function -mspace @gol
789 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
790 -mapp-regs -mno-app-regs @gol
791 -mdisable-callt -mno-disable-callt @gol
797 @gccoptlist{-mg -mgnu -munix}
799 @emph{VxWorks Options}
800 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
801 -Xbind-lazy -Xbind-now}
803 @emph{x86-64 Options}
804 See i386 and x86-64 Options.
806 @emph{Xstormy16 Options}
809 @emph{Xtensa Options}
810 @gccoptlist{-mconst16 -mno-const16 @gol
811 -mfused-madd -mno-fused-madd @gol
812 -mtext-section-literals -mno-text-section-literals @gol
813 -mtarget-align -mno-target-align @gol
814 -mlongcalls -mno-longcalls}
816 @emph{zSeries Options}
817 See S/390 and zSeries Options.
819 @item Code Generation Options
820 @xref{Code Gen Options,,Options for Code Generation Conventions}.
821 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
822 -ffixed-@var{reg} -fexceptions @gol
823 -fnon-call-exceptions -funwind-tables @gol
824 -fasynchronous-unwind-tables @gol
825 -finhibit-size-directive -finstrument-functions @gol
826 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
827 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
828 -fno-common -fno-ident @gol
829 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
830 -fno-jump-tables @gol
831 -frecord-gcc-switches @gol
832 -freg-struct-return -fshort-enums @gol
833 -fshort-double -fshort-wchar @gol
834 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
835 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
836 -fno-stack-limit -fargument-alias -fargument-noalias @gol
837 -fargument-noalias-global -fargument-noalias-anything @gol
838 -fleading-underscore -ftls-model=@var{model} @gol
839 -ftrapv -fwrapv -fbounds-check @gol
844 * Overall Options:: Controlling the kind of output:
845 an executable, object files, assembler files,
846 or preprocessed source.
847 * C Dialect Options:: Controlling the variant of C language compiled.
848 * C++ Dialect Options:: Variations on C++.
849 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
851 * Language Independent Options:: Controlling how diagnostics should be
853 * Warning Options:: How picky should the compiler be?
854 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
855 * Optimize Options:: How much optimization?
856 * Preprocessor Options:: Controlling header files and macro definitions.
857 Also, getting dependency information for Make.
858 * Assembler Options:: Passing options to the assembler.
859 * Link Options:: Specifying libraries and so on.
860 * Directory Options:: Where to find header files and libraries.
861 Where to find the compiler executable files.
862 * Spec Files:: How to pass switches to sub-processes.
863 * Target Options:: Running a cross-compiler, or an old version of GCC.
866 @node Overall Options
867 @section Options Controlling the Kind of Output
869 Compilation can involve up to four stages: preprocessing, compilation
870 proper, assembly and linking, always in that order. GCC is capable of
871 preprocessing and compiling several files either into several
872 assembler input files, or into one assembler input file; then each
873 assembler input file produces an object file, and linking combines all
874 the object files (those newly compiled, and those specified as input)
875 into an executable file.
877 @cindex file name suffix
878 For any given input file, the file name suffix determines what kind of
883 C source code which must be preprocessed.
886 C source code which should not be preprocessed.
889 C++ source code which should not be preprocessed.
892 Objective-C source code. Note that you must link with the @file{libobjc}
893 library to make an Objective-C program work.
896 Objective-C source code which should not be preprocessed.
900 Objective-C++ source code. Note that you must link with the @file{libobjc}
901 library to make an Objective-C++ program work. Note that @samp{.M} refers
902 to a literal capital M@.
905 Objective-C++ source code which should not be preprocessed.
908 C, C++, Objective-C or Objective-C++ header file to be turned into a
913 @itemx @var{file}.cxx
914 @itemx @var{file}.cpp
915 @itemx @var{file}.CPP
916 @itemx @var{file}.c++
918 C++ source code which must be preprocessed. Note that in @samp{.cxx},
919 the last two letters must both be literally @samp{x}. Likewise,
920 @samp{.C} refers to a literal capital C@.
924 Objective-C++ source code which must be preprocessed.
927 Objective-C++ source code which should not be preprocessed.
932 @itemx @var{file}.hxx
933 @itemx @var{file}.hpp
934 @itemx @var{file}.HPP
935 @itemx @var{file}.h++
936 @itemx @var{file}.tcc
937 C++ header file to be turned into a precompiled header.
940 @itemx @var{file}.for
941 @itemx @var{file}.FOR
942 Fixed form Fortran source code which should not be preprocessed.
945 @itemx @var{file}.fpp
946 @itemx @var{file}.FPP
947 Fixed form Fortran source code which must be preprocessed (with the traditional
951 @itemx @var{file}.f95
952 Free form Fortran source code which should not be preprocessed.
955 @itemx @var{file}.F95
956 Free form Fortran source code which must be preprocessed (with the
957 traditional preprocessor).
959 @c FIXME: Descriptions of Java file types.
966 Ada source code file which contains a library unit declaration (a
967 declaration of a package, subprogram, or generic, or a generic
968 instantiation), or a library unit renaming declaration (a package,
969 generic, or subprogram renaming declaration). Such files are also
972 @itemx @var{file}.adb
973 Ada source code file containing a library unit body (a subprogram or
974 package body). Such files are also called @dfn{bodies}.
976 @c GCC also knows about some suffixes for languages not yet included:
988 Assembler code which must be preprocessed.
991 An object file to be fed straight into linking.
992 Any file name with no recognized suffix is treated this way.
996 You can specify the input language explicitly with the @option{-x} option:
999 @item -x @var{language}
1000 Specify explicitly the @var{language} for the following input files
1001 (rather than letting the compiler choose a default based on the file
1002 name suffix). This option applies to all following input files until
1003 the next @option{-x} option. Possible values for @var{language} are:
1005 c c-header c-cpp-output
1006 c++ c++-header c++-cpp-output
1007 objective-c objective-c-header objective-c-cpp-output
1008 objective-c++ objective-c++-header objective-c++-cpp-output
1009 assembler assembler-with-cpp
1017 Turn off any specification of a language, so that subsequent files are
1018 handled according to their file name suffixes (as they are if @option{-x}
1019 has not been used at all).
1021 @item -pass-exit-codes
1022 @opindex pass-exit-codes
1023 Normally the @command{gcc} program will exit with the code of 1 if any
1024 phase of the compiler returns a non-success return code. If you specify
1025 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1026 numerically highest error produced by any phase that returned an error
1027 indication. The C, C++, and Fortran frontends return 4, if an internal
1028 compiler error is encountered.
1031 If you only want some of the stages of compilation, you can use
1032 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1033 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1034 @command{gcc} is to stop. Note that some combinations (for example,
1035 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1040 Compile or assemble the source files, but do not link. The linking
1041 stage simply is not done. The ultimate output is in the form of an
1042 object file for each source file.
1044 By default, the object file name for a source file is made by replacing
1045 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1047 Unrecognized input files, not requiring compilation or assembly, are
1052 Stop after the stage of compilation proper; do not assemble. The output
1053 is in the form of an assembler code file for each non-assembler input
1056 By default, the assembler file name for a source file is made by
1057 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1059 Input files that don't require compilation are ignored.
1063 Stop after the preprocessing stage; do not run the compiler proper. The
1064 output is in the form of preprocessed source code, which is sent to the
1067 Input files which don't require preprocessing are ignored.
1069 @cindex output file option
1072 Place output in file @var{file}. This applies regardless to whatever
1073 sort of output is being produced, whether it be an executable file,
1074 an object file, an assembler file or preprocessed C code.
1076 If @option{-o} is not specified, the default is to put an executable
1077 file in @file{a.out}, the object file for
1078 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1079 assembler file in @file{@var{source}.s}, a precompiled header file in
1080 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1085 Print (on standard error output) the commands executed to run the stages
1086 of compilation. Also print the version number of the compiler driver
1087 program and of the preprocessor and the compiler proper.
1091 Like @option{-v} except the commands are not executed and all command
1092 arguments are quoted. This is useful for shell scripts to capture the
1093 driver-generated command lines.
1097 Use pipes rather than temporary files for communication between the
1098 various stages of compilation. This fails to work on some systems where
1099 the assembler is unable to read from a pipe; but the GNU assembler has
1104 If you are compiling multiple source files, this option tells the driver
1105 to pass all the source files to the compiler at once (for those
1106 languages for which the compiler can handle this). This will allow
1107 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1108 language for which this is supported is C@. If you pass source files for
1109 multiple languages to the driver, using this option, the driver will invoke
1110 the compiler(s) that support IMA once each, passing each compiler all the
1111 source files appropriate for it. For those languages that do not support
1112 IMA this option will be ignored, and the compiler will be invoked once for
1113 each source file in that language. If you use this option in conjunction
1114 with @option{-save-temps}, the compiler will generate multiple
1116 (one for each source file), but only one (combined) @file{.o} or
1121 Print (on the standard output) a description of the command line options
1122 understood by @command{gcc}. If the @option{-v} option is also specified
1123 then @option{--help} will also be passed on to the various processes
1124 invoked by @command{gcc}, so that they can display the command line options
1125 they accept. If the @option{-Wextra} option has also been specified
1126 (prior to the @option{--help} option), then command line options which
1127 have no documentation associated with them will also be displayed.
1130 @opindex target-help
1131 Print (on the standard output) a description of target-specific command
1132 line options for each tool. For some targets extra target-specific
1133 information may also be printed.
1135 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1136 Print (on the standard output) a description of the command line
1137 options understood by the compiler that fit into a specific class.
1138 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1139 @samp{params}, or @var{language}:
1142 @item @samp{optimizers}
1143 This will display all of the optimization options supported by the
1146 @item @samp{warnings}
1147 This will display all of the options controlling warning messages
1148 produced by the compiler.
1151 This will display target-specific options. Unlike the
1152 @option{--target-help} option however, target-specific options of the
1153 linker and assembler will not be displayed. This is because those
1154 tools do not currently support the extended @option{--help=} syntax.
1157 This will display the values recognized by the @option{--param}
1160 @item @var{language}
1161 This will display the options supported for @var{language}, where
1162 @var{language} is the name of one of the languages supported in this
1166 This will display the options that are common to all languages.
1169 It is possible to further refine the output of the @option{--help=}
1170 option by adding a comma separated list of qualifiers after the
1171 class. These can be any from the following list:
1174 @item @samp{undocumented}
1175 Display only those options which are undocumented.
1178 Display options which take an argument that appears after an equal
1179 sign in the same continuous piece of text, such as:
1180 @samp{--help=target}.
1182 @item @samp{separate}
1183 Display options which take an argument that appears as a separate word
1184 following the original option, such as: @samp{-o output-file}.
1187 Thus for example to display all the undocumented target-specific
1188 switches supported by the compiler the following can be used:
1191 --help=target,undocumented
1194 The sense of a qualifier can be inverted by prefixing it with the
1195 @var{^} character, so for example to display all binary warning
1196 options (i.e. ones that are either on or off and that do not take an
1197 argument), which have a description the following can be used:
1200 --help=warnings,^joined,^undocumented
1203 A class can also be used as a qualifier, although this usually
1204 restricts the output by so much that there is nothing to display. One
1205 case where it does work however is when one of the classes is
1206 @var{target}. So for example to display all the target-specific
1207 optimization options the following can be used:
1210 --help=target,optimizers
1213 The @option{--help=} option can be repeated on the command line. Each
1214 successive use will display its requested class of options, skipping
1215 those that have already been displayed.
1217 If the @option{-Q} option appears on the command line before the
1218 @option{--help=} option, then the descriptive text displayed by
1219 @option{--help=} is changed. Instead of describing the displayed
1220 options, an indication is given as to whether the option is enabled,
1221 disabled or set to a specific value (assuming that the compiler
1222 knows this at the point where the @option{--help=} option is used).
1224 Here is a truncated example from the ARM port of @command{gcc}:
1227 % gcc -Q -mabi=2 --help=target -c
1228 The following options are target specific:
1230 -mabort-on-noreturn [disabled]
1234 The output is sensitive to the effects of previous command line
1235 options, so for example it is possible to find out which optimizations
1236 are enabled at @option{-O2} by using:
1239 -O2 --help=optimizers
1242 Alternatively you can discover which binary optimizations are enabled
1243 by @option{-O3} by using:
1246 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1247 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1248 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1253 Display the version number and copyrights of the invoked GCC@.
1255 @include @value{srcdir}/../libiberty/at-file.texi
1259 @section Compiling C++ Programs
1261 @cindex suffixes for C++ source
1262 @cindex C++ source file suffixes
1263 C++ source files conventionally use one of the suffixes @samp{.C},
1264 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1265 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1266 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1267 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1268 files with these names and compiles them as C++ programs even if you
1269 call the compiler the same way as for compiling C programs (usually
1270 with the name @command{gcc}).
1274 However, the use of @command{gcc} does not add the C++ library.
1275 @command{g++} is a program that calls GCC and treats @samp{.c},
1276 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1277 files unless @option{-x} is used, and automatically specifies linking
1278 against the C++ library. This program is also useful when
1279 precompiling a C header file with a @samp{.h} extension for use in C++
1280 compilations. On many systems, @command{g++} is also installed with
1281 the name @command{c++}.
1283 @cindex invoking @command{g++}
1284 When you compile C++ programs, you may specify many of the same
1285 command-line options that you use for compiling programs in any
1286 language; or command-line options meaningful for C and related
1287 languages; or options that are meaningful only for C++ programs.
1288 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1289 explanations of options for languages related to C@.
1290 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1291 explanations of options that are meaningful only for C++ programs.
1293 @node C Dialect Options
1294 @section Options Controlling C Dialect
1295 @cindex dialect options
1296 @cindex language dialect options
1297 @cindex options, dialect
1299 The following options control the dialect of C (or languages derived
1300 from C, such as C++, Objective-C and Objective-C++) that the compiler
1304 @cindex ANSI support
1308 In C mode, support all ISO C90 programs. In C++ mode,
1309 remove GNU extensions that conflict with ISO C++.
1311 This turns off certain features of GCC that are incompatible with ISO
1312 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1313 such as the @code{asm} and @code{typeof} keywords, and
1314 predefined macros such as @code{unix} and @code{vax} that identify the
1315 type of system you are using. It also enables the undesirable and
1316 rarely used ISO trigraph feature. For the C compiler,
1317 it disables recognition of C++ style @samp{//} comments as well as
1318 the @code{inline} keyword.
1320 The alternate keywords @code{__asm__}, @code{__extension__},
1321 @code{__inline__} and @code{__typeof__} continue to work despite
1322 @option{-ansi}. You would not want to use them in an ISO C program, of
1323 course, but it is useful to put them in header files that might be included
1324 in compilations done with @option{-ansi}. Alternate predefined macros
1325 such as @code{__unix__} and @code{__vax__} are also available, with or
1326 without @option{-ansi}.
1328 The @option{-ansi} option does not cause non-ISO programs to be
1329 rejected gratuitously. For that, @option{-pedantic} is required in
1330 addition to @option{-ansi}. @xref{Warning Options}.
1332 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1333 option is used. Some header files may notice this macro and refrain
1334 from declaring certain functions or defining certain macros that the
1335 ISO standard doesn't call for; this is to avoid interfering with any
1336 programs that might use these names for other things.
1338 Functions which would normally be built in but do not have semantics
1339 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1340 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1341 built-in functions provided by GCC}, for details of the functions
1346 Determine the language standard. This option is currently only
1347 supported when compiling C or C++. A value for this option must be
1348 provided; possible values are
1353 ISO C90 (same as @option{-ansi}).
1355 @item iso9899:199409
1356 ISO C90 as modified in amendment 1.
1362 ISO C99. Note that this standard is not yet fully supported; see
1363 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1364 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1367 Default, ISO C90 plus GNU extensions (including some C99 features).
1371 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1372 this will become the default. The name @samp{gnu9x} is deprecated.
1375 The 1998 ISO C++ standard plus amendments.
1378 The same as @option{-std=c++98} plus GNU extensions. This is the
1379 default for C++ code.
1382 The working draft of the upcoming ISO C++0x standard. This option
1383 enables experimental features that are likely to be included in
1384 C++0x. The working draft is constantly changing, and any feature that is
1385 enabled by this flag may be removed from future versions of GCC if it is
1386 not part of the C++0x standard.
1389 The same as @option{-std=c++0x} plus GNU extensions. As with
1390 @option{-std=c++0x}, this option enables experimental features that may
1391 be removed in future versions of GCC.
1394 Even when this option is not specified, you can still use some of the
1395 features of newer standards in so far as they do not conflict with
1396 previous C standards. For example, you may use @code{__restrict__} even
1397 when @option{-std=c99} is not specified.
1399 The @option{-std} options specifying some version of ISO C have the same
1400 effects as @option{-ansi}, except that features that were not in ISO C90
1401 but are in the specified version (for example, @samp{//} comments and
1402 the @code{inline} keyword in ISO C99) are not disabled.
1404 @xref{Standards,,Language Standards Supported by GCC}, for details of
1405 these standard versions.
1407 @item -fgnu89-inline
1408 @opindex fgnu89-inline
1409 The option @option{-fgnu89-inline} tells GCC to use the traditional
1410 GNU semantics for @code{inline} functions when in C99 mode.
1411 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1412 is accepted and ignored by GCC versions 4.1.3 up to but not including
1413 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1414 C99 mode. Using this option is roughly equivalent to adding the
1415 @code{gnu_inline} function attribute to all inline functions
1416 (@pxref{Function Attributes}).
1418 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1419 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1420 specifies the default behavior). This option was first supported in
1421 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1423 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1424 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1425 in effect for @code{inline} functions. @xref{Common Predefined
1426 Macros,,,cpp,The C Preprocessor}.
1428 @item -aux-info @var{filename}
1430 Output to the given filename prototyped declarations for all functions
1431 declared and/or defined in a translation unit, including those in header
1432 files. This option is silently ignored in any language other than C@.
1434 Besides declarations, the file indicates, in comments, the origin of
1435 each declaration (source file and line), whether the declaration was
1436 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1437 @samp{O} for old, respectively, in the first character after the line
1438 number and the colon), and whether it came from a declaration or a
1439 definition (@samp{C} or @samp{F}, respectively, in the following
1440 character). In the case of function definitions, a K&R-style list of
1441 arguments followed by their declarations is also provided, inside
1442 comments, after the declaration.
1446 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1447 keyword, so that code can use these words as identifiers. You can use
1448 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1449 instead. @option{-ansi} implies @option{-fno-asm}.
1451 In C++, this switch only affects the @code{typeof} keyword, since
1452 @code{asm} and @code{inline} are standard keywords. You may want to
1453 use the @option{-fno-gnu-keywords} flag instead, which has the same
1454 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1455 switch only affects the @code{asm} and @code{typeof} keywords, since
1456 @code{inline} is a standard keyword in ISO C99.
1459 @itemx -fno-builtin-@var{function}
1460 @opindex fno-builtin
1461 @cindex built-in functions
1462 Don't recognize built-in functions that do not begin with
1463 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1464 functions provided by GCC}, for details of the functions affected,
1465 including those which are not built-in functions when @option{-ansi} or
1466 @option{-std} options for strict ISO C conformance are used because they
1467 do not have an ISO standard meaning.
1469 GCC normally generates special code to handle certain built-in functions
1470 more efficiently; for instance, calls to @code{alloca} may become single
1471 instructions that adjust the stack directly, and calls to @code{memcpy}
1472 may become inline copy loops. The resulting code is often both smaller
1473 and faster, but since the function calls no longer appear as such, you
1474 cannot set a breakpoint on those calls, nor can you change the behavior
1475 of the functions by linking with a different library. In addition,
1476 when a function is recognized as a built-in function, GCC may use
1477 information about that function to warn about problems with calls to
1478 that function, or to generate more efficient code, even if the
1479 resulting code still contains calls to that function. For example,
1480 warnings are given with @option{-Wformat} for bad calls to
1481 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1482 known not to modify global memory.
1484 With the @option{-fno-builtin-@var{function}} option
1485 only the built-in function @var{function} is
1486 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1487 function is named this is not built-in in this version of GCC, this
1488 option is ignored. There is no corresponding
1489 @option{-fbuiltin-@var{function}} option; if you wish to enable
1490 built-in functions selectively when using @option{-fno-builtin} or
1491 @option{-ffreestanding}, you may define macros such as:
1494 #define abs(n) __builtin_abs ((n))
1495 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1500 @cindex hosted environment
1502 Assert that compilation takes place in a hosted environment. This implies
1503 @option{-fbuiltin}. A hosted environment is one in which the
1504 entire standard library is available, and in which @code{main} has a return
1505 type of @code{int}. Examples are nearly everything except a kernel.
1506 This is equivalent to @option{-fno-freestanding}.
1508 @item -ffreestanding
1509 @opindex ffreestanding
1510 @cindex hosted environment
1512 Assert that compilation takes place in a freestanding environment. This
1513 implies @option{-fno-builtin}. A freestanding environment
1514 is one in which the standard library may not exist, and program startup may
1515 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1516 This is equivalent to @option{-fno-hosted}.
1518 @xref{Standards,,Language Standards Supported by GCC}, for details of
1519 freestanding and hosted environments.
1523 @cindex openmp parallel
1524 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1525 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1526 compiler generates parallel code according to the OpenMP Application
1527 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1529 @item -fms-extensions
1530 @opindex fms-extensions
1531 Accept some non-standard constructs used in Microsoft header files.
1533 Some cases of unnamed fields in structures and unions are only
1534 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1535 fields within structs/unions}, for details.
1539 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1540 options for strict ISO C conformance) implies @option{-trigraphs}.
1542 @item -no-integrated-cpp
1543 @opindex no-integrated-cpp
1544 Performs a compilation in two passes: preprocessing and compiling. This
1545 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1546 @option{-B} option. The user supplied compilation step can then add in
1547 an additional preprocessing step after normal preprocessing but before
1548 compiling. The default is to use the integrated cpp (internal cpp)
1550 The semantics of this option will change if "cc1", "cc1plus", and
1551 "cc1obj" are merged.
1553 @cindex traditional C language
1554 @cindex C language, traditional
1556 @itemx -traditional-cpp
1557 @opindex traditional-cpp
1558 @opindex traditional
1559 Formerly, these options caused GCC to attempt to emulate a pre-standard
1560 C compiler. They are now only supported with the @option{-E} switch.
1561 The preprocessor continues to support a pre-standard mode. See the GNU
1562 CPP manual for details.
1564 @item -fcond-mismatch
1565 @opindex fcond-mismatch
1566 Allow conditional expressions with mismatched types in the second and
1567 third arguments. The value of such an expression is void. This option
1568 is not supported for C++.
1570 @item -flax-vector-conversions
1571 @opindex flax-vector-conversions
1572 Allow implicit conversions between vectors with differing numbers of
1573 elements and/or incompatible element types. This option should not be
1576 @item -funsigned-char
1577 @opindex funsigned-char
1578 Let the type @code{char} be unsigned, like @code{unsigned char}.
1580 Each kind of machine has a default for what @code{char} should
1581 be. It is either like @code{unsigned char} by default or like
1582 @code{signed char} by default.
1584 Ideally, a portable program should always use @code{signed char} or
1585 @code{unsigned char} when it depends on the signedness of an object.
1586 But many programs have been written to use plain @code{char} and
1587 expect it to be signed, or expect it to be unsigned, depending on the
1588 machines they were written for. This option, and its inverse, let you
1589 make such a program work with the opposite default.
1591 The type @code{char} is always a distinct type from each of
1592 @code{signed char} or @code{unsigned char}, even though its behavior
1593 is always just like one of those two.
1596 @opindex fsigned-char
1597 Let the type @code{char} be signed, like @code{signed char}.
1599 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1600 the negative form of @option{-funsigned-char}. Likewise, the option
1601 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1603 @item -fsigned-bitfields
1604 @itemx -funsigned-bitfields
1605 @itemx -fno-signed-bitfields
1606 @itemx -fno-unsigned-bitfields
1607 @opindex fsigned-bitfields
1608 @opindex funsigned-bitfields
1609 @opindex fno-signed-bitfields
1610 @opindex fno-unsigned-bitfields
1611 These options control whether a bit-field is signed or unsigned, when the
1612 declaration does not use either @code{signed} or @code{unsigned}. By
1613 default, such a bit-field is signed, because this is consistent: the
1614 basic integer types such as @code{int} are signed types.
1617 @node C++ Dialect Options
1618 @section Options Controlling C++ Dialect
1620 @cindex compiler options, C++
1621 @cindex C++ options, command line
1622 @cindex options, C++
1623 This section describes the command-line options that are only meaningful
1624 for C++ programs; but you can also use most of the GNU compiler options
1625 regardless of what language your program is in. For example, you
1626 might compile a file @code{firstClass.C} like this:
1629 g++ -g -frepo -O -c firstClass.C
1633 In this example, only @option{-frepo} is an option meant
1634 only for C++ programs; you can use the other options with any
1635 language supported by GCC@.
1637 Here is a list of options that are @emph{only} for compiling C++ programs:
1641 @item -fabi-version=@var{n}
1642 @opindex fabi-version
1643 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1644 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1645 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1646 the version that conforms most closely to the C++ ABI specification.
1647 Therefore, the ABI obtained using version 0 will change as ABI bugs
1650 The default is version 2.
1652 @item -fno-access-control
1653 @opindex fno-access-control
1654 Turn off all access checking. This switch is mainly useful for working
1655 around bugs in the access control code.
1659 Check that the pointer returned by @code{operator new} is non-null
1660 before attempting to modify the storage allocated. This check is
1661 normally unnecessary because the C++ standard specifies that
1662 @code{operator new} will only return @code{0} if it is declared
1663 @samp{throw()}, in which case the compiler will always check the
1664 return value even without this option. In all other cases, when
1665 @code{operator new} has a non-empty exception specification, memory
1666 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1667 @samp{new (nothrow)}.
1669 @item -fconserve-space
1670 @opindex fconserve-space
1671 Put uninitialized or runtime-initialized global variables into the
1672 common segment, as C does. This saves space in the executable at the
1673 cost of not diagnosing duplicate definitions. If you compile with this
1674 flag and your program mysteriously crashes after @code{main()} has
1675 completed, you may have an object that is being destroyed twice because
1676 two definitions were merged.
1678 This option is no longer useful on most targets, now that support has
1679 been added for putting variables into BSS without making them common.
1681 @item -ffriend-injection
1682 @opindex ffriend-injection
1683 Inject friend functions into the enclosing namespace, so that they are
1684 visible outside the scope of the class in which they are declared.
1685 Friend functions were documented to work this way in the old Annotated
1686 C++ Reference Manual, and versions of G++ before 4.1 always worked
1687 that way. However, in ISO C++ a friend function which is not declared
1688 in an enclosing scope can only be found using argument dependent
1689 lookup. This option causes friends to be injected as they were in
1692 This option is for compatibility, and may be removed in a future
1695 @item -fno-elide-constructors
1696 @opindex fno-elide-constructors
1697 The C++ standard allows an implementation to omit creating a temporary
1698 which is only used to initialize another object of the same type.
1699 Specifying this option disables that optimization, and forces G++ to
1700 call the copy constructor in all cases.
1702 @item -fno-enforce-eh-specs
1703 @opindex fno-enforce-eh-specs
1704 Don't generate code to check for violation of exception specifications
1705 at runtime. This option violates the C++ standard, but may be useful
1706 for reducing code size in production builds, much like defining
1707 @samp{NDEBUG}. This does not give user code permission to throw
1708 exceptions in violation of the exception specifications; the compiler
1709 will still optimize based on the specifications, so throwing an
1710 unexpected exception will result in undefined behavior.
1713 @itemx -fno-for-scope
1715 @opindex fno-for-scope
1716 If @option{-ffor-scope} is specified, the scope of variables declared in
1717 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1718 as specified by the C++ standard.
1719 If @option{-fno-for-scope} is specified, the scope of variables declared in
1720 a @i{for-init-statement} extends to the end of the enclosing scope,
1721 as was the case in old versions of G++, and other (traditional)
1722 implementations of C++.
1724 The default if neither flag is given to follow the standard,
1725 but to allow and give a warning for old-style code that would
1726 otherwise be invalid, or have different behavior.
1728 @item -fno-gnu-keywords
1729 @opindex fno-gnu-keywords
1730 Do not recognize @code{typeof} as a keyword, so that code can use this
1731 word as an identifier. You can use the keyword @code{__typeof__} instead.
1732 @option{-ansi} implies @option{-fno-gnu-keywords}.
1734 @item -fno-implicit-templates
1735 @opindex fno-implicit-templates
1736 Never emit code for non-inline templates which are instantiated
1737 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1738 @xref{Template Instantiation}, for more information.
1740 @item -fno-implicit-inline-templates
1741 @opindex fno-implicit-inline-templates
1742 Don't emit code for implicit instantiations of inline templates, either.
1743 The default is to handle inlines differently so that compiles with and
1744 without optimization will need the same set of explicit instantiations.
1746 @item -fno-implement-inlines
1747 @opindex fno-implement-inlines
1748 To save space, do not emit out-of-line copies of inline functions
1749 controlled by @samp{#pragma implementation}. This will cause linker
1750 errors if these functions are not inlined everywhere they are called.
1752 @item -fms-extensions
1753 @opindex fms-extensions
1754 Disable pedantic warnings about constructs used in MFC, such as implicit
1755 int and getting a pointer to member function via non-standard syntax.
1757 @item -fno-nonansi-builtins
1758 @opindex fno-nonansi-builtins
1759 Disable built-in declarations of functions that are not mandated by
1760 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1761 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1763 @item -fno-operator-names
1764 @opindex fno-operator-names
1765 Do not treat the operator name keywords @code{and}, @code{bitand},
1766 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1767 synonyms as keywords.
1769 @item -fno-optional-diags
1770 @opindex fno-optional-diags
1771 Disable diagnostics that the standard says a compiler does not need to
1772 issue. Currently, the only such diagnostic issued by G++ is the one for
1773 a name having multiple meanings within a class.
1776 @opindex fpermissive
1777 Downgrade some diagnostics about nonconformant code from errors to
1778 warnings. Thus, using @option{-fpermissive} will allow some
1779 nonconforming code to compile.
1783 Enable automatic template instantiation at link time. This option also
1784 implies @option{-fno-implicit-templates}. @xref{Template
1785 Instantiation}, for more information.
1789 Disable generation of information about every class with virtual
1790 functions for use by the C++ runtime type identification features
1791 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1792 of the language, you can save some space by using this flag. Note that
1793 exception handling uses the same information, but it will generate it as
1794 needed. The @samp{dynamic_cast} operator can still be used for casts that
1795 do not require runtime type information, i.e. casts to @code{void *} or to
1796 unambiguous base classes.
1800 Emit statistics about front-end processing at the end of the compilation.
1801 This information is generally only useful to the G++ development team.
1803 @item -ftemplate-depth-@var{n}
1804 @opindex ftemplate-depth
1805 Set the maximum instantiation depth for template classes to @var{n}.
1806 A limit on the template instantiation depth is needed to detect
1807 endless recursions during template class instantiation. ANSI/ISO C++
1808 conforming programs must not rely on a maximum depth greater than 17.
1810 @item -fno-threadsafe-statics
1811 @opindex fno-threadsafe-statics
1812 Do not emit the extra code to use the routines specified in the C++
1813 ABI for thread-safe initialization of local statics. You can use this
1814 option to reduce code size slightly in code that doesn't need to be
1817 @item -fuse-cxa-atexit
1818 @opindex fuse-cxa-atexit
1819 Register destructors for objects with static storage duration with the
1820 @code{__cxa_atexit} function rather than the @code{atexit} function.
1821 This option is required for fully standards-compliant handling of static
1822 destructors, but will only work if your C library supports
1823 @code{__cxa_atexit}.
1825 @item -fno-use-cxa-get-exception-ptr
1826 @opindex fno-use-cxa-get-exception-ptr
1827 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1828 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1829 if the runtime routine is not available.
1831 @item -fvisibility-inlines-hidden
1832 @opindex fvisibility-inlines-hidden
1833 This switch declares that the user does not attempt to compare
1834 pointers to inline methods where the addresses of the two functions
1835 were taken in different shared objects.
1837 The effect of this is that GCC may, effectively, mark inline methods with
1838 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1839 appear in the export table of a DSO and do not require a PLT indirection
1840 when used within the DSO@. Enabling this option can have a dramatic effect
1841 on load and link times of a DSO as it massively reduces the size of the
1842 dynamic export table when the library makes heavy use of templates.
1844 The behavior of this switch is not quite the same as marking the
1845 methods as hidden directly, because it does not affect static variables
1846 local to the function or cause the compiler to deduce that
1847 the function is defined in only one shared object.
1849 You may mark a method as having a visibility explicitly to negate the
1850 effect of the switch for that method. For example, if you do want to
1851 compare pointers to a particular inline method, you might mark it as
1852 having default visibility. Marking the enclosing class with explicit
1853 visibility will have no effect.
1855 Explicitly instantiated inline methods are unaffected by this option
1856 as their linkage might otherwise cross a shared library boundary.
1857 @xref{Template Instantiation}.
1859 @item -fvisibility-ms-compat
1860 @opindex fvisibility-ms-compat
1861 This flag attempts to use visibility settings to make GCC's C++
1862 linkage model compatible with that of Microsoft Visual Studio.
1864 The flag makes these changes to GCC's linkage model:
1868 It sets the default visibility to @code{hidden}, like
1869 @option{-fvisibility=hidden}.
1872 Types, but not their members, are not hidden by default.
1875 The One Definition Rule is relaxed for types without explicit
1876 visibility specifications which are defined in more than one different
1877 shared object: those declarations are permitted if they would have
1878 been permitted when this option was not used.
1881 In new code it is better to use @option{-fvisibility=hidden} and
1882 export those classes which are intended to be externally visible.
1883 Unfortunately it is possible for code to rely, perhaps accidentally,
1884 on the Visual Studio behavior.
1886 Among the consequences of these changes are that static data members
1887 of the same type with the same name but defined in different shared
1888 objects will be different, so changing one will not change the other;
1889 and that pointers to function members defined in different shared
1890 objects may not compare equal. When this flag is given, it is a
1891 violation of the ODR to define types with the same name differently.
1895 Do not use weak symbol support, even if it is provided by the linker.
1896 By default, G++ will use weak symbols if they are available. This
1897 option exists only for testing, and should not be used by end-users;
1898 it will result in inferior code and has no benefits. This option may
1899 be removed in a future release of G++.
1903 Do not search for header files in the standard directories specific to
1904 C++, but do still search the other standard directories. (This option
1905 is used when building the C++ library.)
1908 In addition, these optimization, warning, and code generation options
1909 have meanings only for C++ programs:
1912 @item -fno-default-inline
1913 @opindex fno-default-inline
1914 Do not assume @samp{inline} for functions defined inside a class scope.
1915 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1916 functions will have linkage like inline functions; they just won't be
1919 @item -Wabi @r{(C++ only)}
1921 Warn when G++ generates code that is probably not compatible with the
1922 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1923 all such cases, there are probably some cases that are not warned about,
1924 even though G++ is generating incompatible code. There may also be
1925 cases where warnings are emitted even though the code that is generated
1928 You should rewrite your code to avoid these warnings if you are
1929 concerned about the fact that code generated by G++ may not be binary
1930 compatible with code generated by other compilers.
1932 The known incompatibilities at this point include:
1937 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1938 pack data into the same byte as a base class. For example:
1941 struct A @{ virtual void f(); int f1 : 1; @};
1942 struct B : public A @{ int f2 : 1; @};
1946 In this case, G++ will place @code{B::f2} into the same byte
1947 as@code{A::f1}; other compilers will not. You can avoid this problem
1948 by explicitly padding @code{A} so that its size is a multiple of the
1949 byte size on your platform; that will cause G++ and other compilers to
1950 layout @code{B} identically.
1953 Incorrect handling of tail-padding for virtual bases. G++ does not use
1954 tail padding when laying out virtual bases. For example:
1957 struct A @{ virtual void f(); char c1; @};
1958 struct B @{ B(); char c2; @};
1959 struct C : public A, public virtual B @{@};
1963 In this case, G++ will not place @code{B} into the tail-padding for
1964 @code{A}; other compilers will. You can avoid this problem by
1965 explicitly padding @code{A} so that its size is a multiple of its
1966 alignment (ignoring virtual base classes); that will cause G++ and other
1967 compilers to layout @code{C} identically.
1970 Incorrect handling of bit-fields with declared widths greater than that
1971 of their underlying types, when the bit-fields appear in a union. For
1975 union U @{ int i : 4096; @};
1979 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1980 union too small by the number of bits in an @code{int}.
1983 Empty classes can be placed at incorrect offsets. For example:
1993 struct C : public B, public A @{@};
1997 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1998 it should be placed at offset zero. G++ mistakenly believes that the
1999 @code{A} data member of @code{B} is already at offset zero.
2002 Names of template functions whose types involve @code{typename} or
2003 template template parameters can be mangled incorrectly.
2006 template <typename Q>
2007 void f(typename Q::X) @{@}
2009 template <template <typename> class Q>
2010 void f(typename Q<int>::X) @{@}
2014 Instantiations of these templates may be mangled incorrectly.
2018 @item -Wctor-dtor-privacy @r{(C++ only)}
2019 @opindex Wctor-dtor-privacy
2020 Warn when a class seems unusable because all the constructors or
2021 destructors in that class are private, and it has neither friends nor
2022 public static member functions.
2024 @item -Wnon-virtual-dtor @r{(C++ only)}
2025 @opindex Wnon-virtual-dtor
2026 Warn when a class has virtual functions and accessible non-virtual
2027 destructor, in which case it would be possible but unsafe to delete
2028 an instance of a derived class through a pointer to the base class.
2029 This warning is also enabled if -Weffc++ is specified.
2031 @item -Wreorder @r{(C++ only)}
2033 @cindex reordering, warning
2034 @cindex warning for reordering of member initializers
2035 Warn when the order of member initializers given in the code does not
2036 match the order in which they must be executed. For instance:
2042 A(): j (0), i (1) @{ @}
2046 The compiler will rearrange the member initializers for @samp{i}
2047 and @samp{j} to match the declaration order of the members, emitting
2048 a warning to that effect. This warning is enabled by @option{-Wall}.
2051 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2054 @item -Weffc++ @r{(C++ only)}
2056 Warn about violations of the following style guidelines from Scott Meyers'
2057 @cite{Effective C++} book:
2061 Item 11: Define a copy constructor and an assignment operator for classes
2062 with dynamically allocated memory.
2065 Item 12: Prefer initialization to assignment in constructors.
2068 Item 14: Make destructors virtual in base classes.
2071 Item 15: Have @code{operator=} return a reference to @code{*this}.
2074 Item 23: Don't try to return a reference when you must return an object.
2078 Also warn about violations of the following style guidelines from
2079 Scott Meyers' @cite{More Effective C++} book:
2083 Item 6: Distinguish between prefix and postfix forms of increment and
2084 decrement operators.
2087 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2091 When selecting this option, be aware that the standard library
2092 headers do not obey all of these guidelines; use @samp{grep -v}
2093 to filter out those warnings.
2095 @item -Wno-deprecated @r{(C++ only)}
2096 @opindex Wno-deprecated
2097 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
2099 @item -Wstrict-null-sentinel @r{(C++ only)}
2100 @opindex Wstrict-null-sentinel
2101 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2102 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2103 to @code{__null}. Although it is a null pointer constant not a null pointer,
2104 it is guaranteed to of the same size as a pointer. But this use is
2105 not portable across different compilers.
2107 @item -Wno-non-template-friend @r{(C++ only)}
2108 @opindex Wno-non-template-friend
2109 Disable warnings when non-templatized friend functions are declared
2110 within a template. Since the advent of explicit template specification
2111 support in G++, if the name of the friend is an unqualified-id (i.e.,
2112 @samp{friend foo(int)}), the C++ language specification demands that the
2113 friend declare or define an ordinary, nontemplate function. (Section
2114 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2115 could be interpreted as a particular specialization of a templatized
2116 function. Because this non-conforming behavior is no longer the default
2117 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2118 check existing code for potential trouble spots and is on by default.
2119 This new compiler behavior can be turned off with
2120 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2121 but disables the helpful warning.
2123 @item -Wold-style-cast @r{(C++ only)}
2124 @opindex Wold-style-cast
2125 Warn if an old-style (C-style) cast to a non-void type is used within
2126 a C++ program. The new-style casts (@samp{dynamic_cast},
2127 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2128 less vulnerable to unintended effects and much easier to search for.
2130 @item -Woverloaded-virtual @r{(C++ only)}
2131 @opindex Woverloaded-virtual
2132 @cindex overloaded virtual fn, warning
2133 @cindex warning for overloaded virtual fn
2134 Warn when a function declaration hides virtual functions from a
2135 base class. For example, in:
2142 struct B: public A @{
2147 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2155 will fail to compile.
2157 @item -Wno-pmf-conversions @r{(C++ only)}
2158 @opindex Wno-pmf-conversions
2159 Disable the diagnostic for converting a bound pointer to member function
2162 @item -Wsign-promo @r{(C++ only)}
2163 @opindex Wsign-promo
2164 Warn when overload resolution chooses a promotion from unsigned or
2165 enumerated type to a signed type, over a conversion to an unsigned type of
2166 the same size. Previous versions of G++ would try to preserve
2167 unsignedness, but the standard mandates the current behavior.
2172 A& operator = (int);
2182 In this example, G++ will synthesize a default @samp{A& operator =
2183 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2186 @node Objective-C and Objective-C++ Dialect Options
2187 @section Options Controlling Objective-C and Objective-C++ Dialects
2189 @cindex compiler options, Objective-C and Objective-C++
2190 @cindex Objective-C and Objective-C++ options, command line
2191 @cindex options, Objective-C and Objective-C++
2192 (NOTE: This manual does not describe the Objective-C and Objective-C++
2193 languages themselves. See @xref{Standards,,Language Standards
2194 Supported by GCC}, for references.)
2196 This section describes the command-line options that are only meaningful
2197 for Objective-C and Objective-C++ programs, but you can also use most of
2198 the language-independent GNU compiler options.
2199 For example, you might compile a file @code{some_class.m} like this:
2202 gcc -g -fgnu-runtime -O -c some_class.m
2206 In this example, @option{-fgnu-runtime} is an option meant only for
2207 Objective-C and Objective-C++ programs; you can use the other options with
2208 any language supported by GCC@.
2210 Note that since Objective-C is an extension of the C language, Objective-C
2211 compilations may also use options specific to the C front-end (e.g.,
2212 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2213 C++-specific options (e.g., @option{-Wabi}).
2215 Here is a list of options that are @emph{only} for compiling Objective-C
2216 and Objective-C++ programs:
2219 @item -fconstant-string-class=@var{class-name}
2220 @opindex fconstant-string-class
2221 Use @var{class-name} as the name of the class to instantiate for each
2222 literal string specified with the syntax @code{@@"@dots{}"}. The default
2223 class name is @code{NXConstantString} if the GNU runtime is being used, and
2224 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2225 @option{-fconstant-cfstrings} option, if also present, will override the
2226 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2227 to be laid out as constant CoreFoundation strings.
2230 @opindex fgnu-runtime
2231 Generate object code compatible with the standard GNU Objective-C
2232 runtime. This is the default for most types of systems.
2234 @item -fnext-runtime
2235 @opindex fnext-runtime
2236 Generate output compatible with the NeXT runtime. This is the default
2237 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2238 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2241 @item -fno-nil-receivers
2242 @opindex fno-nil-receivers
2243 Assume that all Objective-C message dispatches (e.g.,
2244 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2245 is not @code{nil}. This allows for more efficient entry points in the runtime
2246 to be used. Currently, this option is only available in conjunction with
2247 the NeXT runtime on Mac OS X 10.3 and later.
2249 @item -fobjc-call-cxx-cdtors
2250 @opindex fobjc-call-cxx-cdtors
2251 For each Objective-C class, check if any of its instance variables is a
2252 C++ object with a non-trivial default constructor. If so, synthesize a
2253 special @code{- (id) .cxx_construct} instance method that will run
2254 non-trivial default constructors on any such instance variables, in order,
2255 and then return @code{self}. Similarly, check if any instance variable
2256 is a C++ object with a non-trivial destructor, and if so, synthesize a
2257 special @code{- (void) .cxx_destruct} method that will run
2258 all such default destructors, in reverse order.
2260 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2261 thusly generated will only operate on instance variables declared in the
2262 current Objective-C class, and not those inherited from superclasses. It
2263 is the responsibility of the Objective-C runtime to invoke all such methods
2264 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2265 will be invoked by the runtime immediately after a new object
2266 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2267 be invoked immediately before the runtime deallocates an object instance.
2269 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2270 support for invoking the @code{- (id) .cxx_construct} and
2271 @code{- (void) .cxx_destruct} methods.
2273 @item -fobjc-direct-dispatch
2274 @opindex fobjc-direct-dispatch
2275 Allow fast jumps to the message dispatcher. On Darwin this is
2276 accomplished via the comm page.
2278 @item -fobjc-exceptions
2279 @opindex fobjc-exceptions
2280 Enable syntactic support for structured exception handling in Objective-C,
2281 similar to what is offered by C++ and Java. This option is
2282 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2291 @@catch (AnObjCClass *exc) @{
2298 @@catch (AnotherClass *exc) @{
2301 @@catch (id allOthers) @{
2311 The @code{@@throw} statement may appear anywhere in an Objective-C or
2312 Objective-C++ program; when used inside of a @code{@@catch} block, the
2313 @code{@@throw} may appear without an argument (as shown above), in which case
2314 the object caught by the @code{@@catch} will be rethrown.
2316 Note that only (pointers to) Objective-C objects may be thrown and
2317 caught using this scheme. When an object is thrown, it will be caught
2318 by the nearest @code{@@catch} clause capable of handling objects of that type,
2319 analogously to how @code{catch} blocks work in C++ and Java. A
2320 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2321 any and all Objective-C exceptions not caught by previous @code{@@catch}
2324 The @code{@@finally} clause, if present, will be executed upon exit from the
2325 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2326 regardless of whether any exceptions are thrown, caught or rethrown
2327 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2328 of the @code{finally} clause in Java.
2330 There are several caveats to using the new exception mechanism:
2334 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2335 idioms provided by the @code{NSException} class, the new
2336 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2337 systems, due to additional functionality needed in the (NeXT) Objective-C
2341 As mentioned above, the new exceptions do not support handling
2342 types other than Objective-C objects. Furthermore, when used from
2343 Objective-C++, the Objective-C exception model does not interoperate with C++
2344 exceptions at this time. This means you cannot @code{@@throw} an exception
2345 from Objective-C and @code{catch} it in C++, or vice versa
2346 (i.e., @code{throw @dots{} @@catch}).
2349 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2350 blocks for thread-safe execution:
2353 @@synchronized (ObjCClass *guard) @{
2358 Upon entering the @code{@@synchronized} block, a thread of execution shall
2359 first check whether a lock has been placed on the corresponding @code{guard}
2360 object by another thread. If it has, the current thread shall wait until
2361 the other thread relinquishes its lock. Once @code{guard} becomes available,
2362 the current thread will place its own lock on it, execute the code contained in
2363 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2364 making @code{guard} available to other threads).
2366 Unlike Java, Objective-C does not allow for entire methods to be marked
2367 @code{@@synchronized}. Note that throwing exceptions out of
2368 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2369 to be unlocked properly.
2373 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2375 @item -freplace-objc-classes
2376 @opindex freplace-objc-classes
2377 Emit a special marker instructing @command{ld(1)} not to statically link in
2378 the resulting object file, and allow @command{dyld(1)} to load it in at
2379 run time instead. This is used in conjunction with the Fix-and-Continue
2380 debugging mode, where the object file in question may be recompiled and
2381 dynamically reloaded in the course of program execution, without the need
2382 to restart the program itself. Currently, Fix-and-Continue functionality
2383 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2388 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2389 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2390 compile time) with static class references that get initialized at load time,
2391 which improves run-time performance. Specifying the @option{-fzero-link} flag
2392 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2393 to be retained. This is useful in Zero-Link debugging mode, since it allows
2394 for individual class implementations to be modified during program execution.
2398 Dump interface declarations for all classes seen in the source file to a
2399 file named @file{@var{sourcename}.decl}.
2401 @item -Wassign-intercept
2402 @opindex Wassign-intercept
2403 Warn whenever an Objective-C assignment is being intercepted by the
2407 @opindex Wno-protocol
2408 If a class is declared to implement a protocol, a warning is issued for
2409 every method in the protocol that is not implemented by the class. The
2410 default behavior is to issue a warning for every method not explicitly
2411 implemented in the class, even if a method implementation is inherited
2412 from the superclass. If you use the @option{-Wno-protocol} option, then
2413 methods inherited from the superclass are considered to be implemented,
2414 and no warning is issued for them.
2418 Warn if multiple methods of different types for the same selector are
2419 found during compilation. The check is performed on the list of methods
2420 in the final stage of compilation. Additionally, a check is performed
2421 for each selector appearing in a @code{@@selector(@dots{})}
2422 expression, and a corresponding method for that selector has been found
2423 during compilation. Because these checks scan the method table only at
2424 the end of compilation, these warnings are not produced if the final
2425 stage of compilation is not reached, for example because an error is
2426 found during compilation, or because the @option{-fsyntax-only} option is
2429 @item -Wstrict-selector-match
2430 @opindex Wstrict-selector-match
2431 Warn if multiple methods with differing argument and/or return types are
2432 found for a given selector when attempting to send a message using this
2433 selector to a receiver of type @code{id} or @code{Class}. When this flag
2434 is off (which is the default behavior), the compiler will omit such warnings
2435 if any differences found are confined to types which share the same size
2438 @item -Wundeclared-selector
2439 @opindex Wundeclared-selector
2440 Warn if a @code{@@selector(@dots{})} expression referring to an
2441 undeclared selector is found. A selector is considered undeclared if no
2442 method with that name has been declared before the
2443 @code{@@selector(@dots{})} expression, either explicitly in an
2444 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2445 an @code{@@implementation} section. This option always performs its
2446 checks as soon as a @code{@@selector(@dots{})} expression is found,
2447 while @option{-Wselector} only performs its checks in the final stage of
2448 compilation. This also enforces the coding style convention
2449 that methods and selectors must be declared before being used.
2451 @item -print-objc-runtime-info
2452 @opindex print-objc-runtime-info
2453 Generate C header describing the largest structure that is passed by
2458 @node Language Independent Options
2459 @section Options to Control Diagnostic Messages Formatting
2460 @cindex options to control diagnostics formatting
2461 @cindex diagnostic messages
2462 @cindex message formatting
2464 Traditionally, diagnostic messages have been formatted irrespective of
2465 the output device's aspect (e.g.@: its width, @dots{}). The options described
2466 below can be used to control the diagnostic messages formatting
2467 algorithm, e.g.@: how many characters per line, how often source location
2468 information should be reported. Right now, only the C++ front end can
2469 honor these options. However it is expected, in the near future, that
2470 the remaining front ends would be able to digest them correctly.
2473 @item -fmessage-length=@var{n}
2474 @opindex fmessage-length
2475 Try to format error messages so that they fit on lines of about @var{n}
2476 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2477 the front ends supported by GCC@. If @var{n} is zero, then no
2478 line-wrapping will be done; each error message will appear on a single
2481 @opindex fdiagnostics-show-location
2482 @item -fdiagnostics-show-location=once
2483 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2484 reporter to emit @emph{once} source location information; that is, in
2485 case the message is too long to fit on a single physical line and has to
2486 be wrapped, the source location won't be emitted (as prefix) again,
2487 over and over, in subsequent continuation lines. This is the default
2490 @item -fdiagnostics-show-location=every-line
2491 Only meaningful in line-wrapping mode. Instructs the diagnostic
2492 messages reporter to emit the same source location information (as
2493 prefix) for physical lines that result from the process of breaking
2494 a message which is too long to fit on a single line.
2496 @item -fdiagnostics-show-option
2497 @opindex fdiagnostics-show-option
2498 This option instructs the diagnostic machinery to add text to each
2499 diagnostic emitted, which indicates which command line option directly
2500 controls that diagnostic, when such an option is known to the
2501 diagnostic machinery.
2503 @item -Wcoverage-mismatch
2504 @opindex Wcoverage-mismatch
2505 Warn if feedback profiles do not match when using the
2506 @option{-fprofile-use} option.
2507 If a source file was changed between @option{-fprofile-gen} and
2508 @option{-fprofile-use}, the files with the profile feedback can fail
2509 to match the source file and GCC can not use the profile feedback
2510 information. By default, GCC emits an error message in this case.
2511 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2512 error. GCC does not use appropriate feedback profiles, so using this
2513 option can result in poorly optimized code. This option is useful
2514 only in the case of very minor changes such as bug fixes to an
2519 @node Warning Options
2520 @section Options to Request or Suppress Warnings
2521 @cindex options to control warnings
2522 @cindex warning messages
2523 @cindex messages, warning
2524 @cindex suppressing warnings
2526 Warnings are diagnostic messages that report constructions which
2527 are not inherently erroneous but which are risky or suggest there
2528 may have been an error.
2530 The following language-independent options do not enable specific
2531 warnings but control the kinds of diagnostics produced by GCC.
2534 @cindex syntax checking
2536 @opindex fsyntax-only
2537 Check the code for syntax errors, but don't do anything beyond that.
2541 Inhibit all warning messages.
2545 Make all warnings into errors.
2549 Make the specified warning into an error. The specifier for a warning
2550 is appended, for example @option{-Werror=switch} turns the warnings
2551 controlled by @option{-Wswitch} into errors. This switch takes a
2552 negative form, to be used to negate @option{-Werror} for specific
2553 warnings, for example @option{-Wno-error=switch} makes
2554 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2555 is in effect. You can use the @option{-fdiagnostics-show-option}
2556 option to have each controllable warning amended with the option which
2557 controls it, to determine what to use with this option.
2559 Note that specifying @option{-Werror=}@var{foo} automatically implies
2560 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2563 @item -Wfatal-errors
2564 @opindex Wfatal-errors
2565 This option causes the compiler to abort compilation on the first error
2566 occurred rather than trying to keep going and printing further error
2571 You can request many specific warnings with options beginning
2572 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2573 implicit declarations. Each of these specific warning options also
2574 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2575 example, @option{-Wno-implicit}. This manual lists only one of the
2576 two forms, whichever is not the default. For further,
2577 language-specific options also refer to @ref{C++ Dialect Options} and
2578 @ref{Objective-C and Objective-C++ Dialect Options}.
2583 Issue all the warnings demanded by strict ISO C and ISO C++;
2584 reject all programs that use forbidden extensions, and some other
2585 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2586 version of the ISO C standard specified by any @option{-std} option used.
2588 Valid ISO C and ISO C++ programs should compile properly with or without
2589 this option (though a rare few will require @option{-ansi} or a
2590 @option{-std} option specifying the required version of ISO C)@. However,
2591 without this option, certain GNU extensions and traditional C and C++
2592 features are supported as well. With this option, they are rejected.
2594 @option{-pedantic} does not cause warning messages for use of the
2595 alternate keywords whose names begin and end with @samp{__}. Pedantic
2596 warnings are also disabled in the expression that follows
2597 @code{__extension__}. However, only system header files should use
2598 these escape routes; application programs should avoid them.
2599 @xref{Alternate Keywords}.
2601 Some users try to use @option{-pedantic} to check programs for strict ISO
2602 C conformance. They soon find that it does not do quite what they want:
2603 it finds some non-ISO practices, but not all---only those for which
2604 ISO C @emph{requires} a diagnostic, and some others for which
2605 diagnostics have been added.
2607 A feature to report any failure to conform to ISO C might be useful in
2608 some instances, but would require considerable additional work and would
2609 be quite different from @option{-pedantic}. We don't have plans to
2610 support such a feature in the near future.
2612 Where the standard specified with @option{-std} represents a GNU
2613 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2614 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2615 extended dialect is based. Warnings from @option{-pedantic} are given
2616 where they are required by the base standard. (It would not make sense
2617 for such warnings to be given only for features not in the specified GNU
2618 C dialect, since by definition the GNU dialects of C include all
2619 features the compiler supports with the given option, and there would be
2620 nothing to warn about.)
2622 @item -pedantic-errors
2623 @opindex pedantic-errors
2624 Like @option{-pedantic}, except that errors are produced rather than
2629 This enables all the warnings about constructions that some users
2630 consider questionable, and that are easy to avoid (or modify to
2631 prevent the warning), even in conjunction with macros. This also
2632 enables some language-specific warnings described in @ref{C++ Dialect
2633 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2635 @option{-Wall} turns on the following warning flags:
2637 @gccoptlist{-Waddress @gol
2638 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2640 -Wchar-subscripts @gol
2642 -Wimplicit-function-declaration @gol
2645 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2646 -Wmissing-braces @gol
2652 -Wsequence-point @gol
2653 -Wsign-compare @r{(only in C++)} @gol
2654 -Wstrict-aliasing @gol
2655 -Wstrict-overflow @gol
2658 -Wuninitialized @r{(only with} @option{-O1} @r{and above)} @gol
2659 -Wunknown-pragmas @gol
2660 -Wunused-function @gol
2663 -Wunused-variable @gol
2666 Note that some warning flags are not implied by @option{-Wall}. Some of
2667 them warn about constructions that users generally do not consider
2668 questionable, but which occasionally you might wish to check for;
2669 others warn about constructions that are necessary or hard to avoid in
2670 some cases, and there is no simple way to modify the code to suppress
2671 the warning. Some of them are enabled by @option{-Wextra} but many of
2672 them must be enabled individually.
2677 This enables some extra warning flags that are not enabled by
2678 @option{-Wall}. (This option used to be called @option{-W}. The older
2679 name is still supported, but the newer name is more descriptive.)
2681 @gccoptlist{-Wclobbered @gol
2683 -Wmissing-field-initializers @gol
2684 -Wmissing-parameter-type @r{(C only)} @gol
2685 -Wold-style-declaration @r{(C only)} @gol
2686 -Woverride-init @gol
2689 -Wuninitialized @r{(only with} @option{-O1} @r{and above)} @gol
2690 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2693 The option @option{-Wextra} also prints warning messages for the
2699 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2700 @samp{>}, or @samp{>=}.
2703 (C++ only) An enumerator and a non-enumerator both appear in a
2704 conditional expression.
2707 (C++ only) A non-static reference or non-static @samp{const} member
2708 appears in a class without constructors.
2711 (C++ only) Ambiguous virtual bases.
2714 (C++ only) Subscripting an array which has been declared @samp{register}.
2717 (C++ only) Taking the address of a variable which has been declared
2721 (C++ only) A base class is not initialized in a derived class' copy
2728 Inhibit warning messages about the use of @samp{#import}.
2730 @item -Wchar-subscripts
2731 @opindex Wchar-subscripts
2732 Warn if an array subscript has type @code{char}. This is a common cause
2733 of error, as programmers often forget that this type is signed on some
2735 This warning is enabled by @option{-Wall}.
2739 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2740 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2741 This warning is enabled by @option{-Wall}.
2745 @opindex ffreestanding
2746 @opindex fno-builtin
2747 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2748 the arguments supplied have types appropriate to the format string
2749 specified, and that the conversions specified in the format string make
2750 sense. This includes standard functions, and others specified by format
2751 attributes (@pxref{Function Attributes}), in the @code{printf},
2752 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2753 not in the C standard) families (or other target-specific families).
2754 Which functions are checked without format attributes having been
2755 specified depends on the standard version selected, and such checks of
2756 functions without the attribute specified are disabled by
2757 @option{-ffreestanding} or @option{-fno-builtin}.
2759 The formats are checked against the format features supported by GNU
2760 libc version 2.2. These include all ISO C90 and C99 features, as well
2761 as features from the Single Unix Specification and some BSD and GNU
2762 extensions. Other library implementations may not support all these
2763 features; GCC does not support warning about features that go beyond a
2764 particular library's limitations. However, if @option{-pedantic} is used
2765 with @option{-Wformat}, warnings will be given about format features not
2766 in the selected standard version (but not for @code{strfmon} formats,
2767 since those are not in any version of the C standard). @xref{C Dialect
2768 Options,,Options Controlling C Dialect}.
2770 Since @option{-Wformat} also checks for null format arguments for
2771 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2773 @option{-Wformat} is included in @option{-Wall}. For more control over some
2774 aspects of format checking, the options @option{-Wformat-y2k},
2775 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2776 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2777 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2780 @opindex Wformat-y2k
2781 If @option{-Wformat} is specified, also warn about @code{strftime}
2782 formats which may yield only a two-digit year.
2784 @item -Wno-format-extra-args
2785 @opindex Wno-format-extra-args
2786 If @option{-Wformat} is specified, do not warn about excess arguments to a
2787 @code{printf} or @code{scanf} format function. The C standard specifies
2788 that such arguments are ignored.
2790 Where the unused arguments lie between used arguments that are
2791 specified with @samp{$} operand number specifications, normally
2792 warnings are still given, since the implementation could not know what
2793 type to pass to @code{va_arg} to skip the unused arguments. However,
2794 in the case of @code{scanf} formats, this option will suppress the
2795 warning if the unused arguments are all pointers, since the Single
2796 Unix Specification says that such unused arguments are allowed.
2798 @item -Wno-format-zero-length
2799 @opindex Wno-format-zero-length
2800 If @option{-Wformat} is specified, do not warn about zero-length formats.
2801 The C standard specifies that zero-length formats are allowed.
2803 @item -Wformat-nonliteral
2804 @opindex Wformat-nonliteral
2805 If @option{-Wformat} is specified, also warn if the format string is not a
2806 string literal and so cannot be checked, unless the format function
2807 takes its format arguments as a @code{va_list}.
2809 @item -Wformat-security
2810 @opindex Wformat-security
2811 If @option{-Wformat} is specified, also warn about uses of format
2812 functions that represent possible security problems. At present, this
2813 warns about calls to @code{printf} and @code{scanf} functions where the
2814 format string is not a string literal and there are no format arguments,
2815 as in @code{printf (foo);}. This may be a security hole if the format
2816 string came from untrusted input and contains @samp{%n}. (This is
2817 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2818 in future warnings may be added to @option{-Wformat-security} that are not
2819 included in @option{-Wformat-nonliteral}.)
2823 Enable @option{-Wformat} plus format checks not included in
2824 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2825 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2829 Warn about passing a null pointer for arguments marked as
2830 requiring a non-null value by the @code{nonnull} function attribute.
2832 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2833 can be disabled with the @option{-Wno-nonnull} option.
2835 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2837 Warn about uninitialized variables which are initialized with themselves.
2838 Note this option can only be used with the @option{-Wuninitialized} option,
2839 which in turn only works with @option{-O1} and above.
2841 For example, GCC will warn about @code{i} being uninitialized in the
2842 following snippet only when @option{-Winit-self} has been specified:
2853 @item -Wimplicit-int
2854 @opindex Wimplicit-int
2855 Warn when a declaration does not specify a type.
2856 This warning is enabled by @option{-Wall}.
2858 @item -Wimplicit-function-declaration
2859 @opindex Wimplicit-function-declaration
2860 @opindex Wno-implicit-function-declaration
2861 Give a warning whenever a function is used before being declared. In
2862 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2863 enabled by default and it is made into an error by
2864 @option{-pedantic-errors}. This warning is also enabled by
2869 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2870 This warning is enabled by @option{-Wall}.
2874 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2875 function with external linkage, returning int, taking either zero
2876 arguments, two, or three arguments of appropriate types.
2877 This warning is enabled by @option{-Wall}.
2879 @item -Wmissing-braces
2880 @opindex Wmissing-braces
2881 Warn if an aggregate or union initializer is not fully bracketed. In
2882 the following example, the initializer for @samp{a} is not fully
2883 bracketed, but that for @samp{b} is fully bracketed.
2886 int a[2][2] = @{ 0, 1, 2, 3 @};
2887 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2890 This warning is enabled by @option{-Wall}.
2892 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2893 @opindex Wmissing-include-dirs
2894 Warn if a user-supplied include directory does not exist.
2897 @opindex Wparentheses
2898 Warn if parentheses are omitted in certain contexts, such
2899 as when there is an assignment in a context where a truth value
2900 is expected, or when operators are nested whose precedence people
2901 often get confused about.
2903 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2904 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2905 interpretation from that of ordinary mathematical notation.
2907 Also warn about constructions where there may be confusion to which
2908 @code{if} statement an @code{else} branch belongs. Here is an example of
2923 In C/C++, every @code{else} branch belongs to the innermost possible
2924 @code{if} statement, which in this example is @code{if (b)}. This is
2925 often not what the programmer expected, as illustrated in the above
2926 example by indentation the programmer chose. When there is the
2927 potential for this confusion, GCC will issue a warning when this flag
2928 is specified. To eliminate the warning, add explicit braces around
2929 the innermost @code{if} statement so there is no way the @code{else}
2930 could belong to the enclosing @code{if}. The resulting code would
2947 This warning is enabled by @option{-Wall}.
2949 @item -Wsequence-point
2950 @opindex Wsequence-point
2951 Warn about code that may have undefined semantics because of violations
2952 of sequence point rules in the C and C++ standards.
2954 The C and C++ standards defines the order in which expressions in a C/C++
2955 program are evaluated in terms of @dfn{sequence points}, which represent
2956 a partial ordering between the execution of parts of the program: those
2957 executed before the sequence point, and those executed after it. These
2958 occur after the evaluation of a full expression (one which is not part
2959 of a larger expression), after the evaluation of the first operand of a
2960 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2961 function is called (but after the evaluation of its arguments and the
2962 expression denoting the called function), and in certain other places.
2963 Other than as expressed by the sequence point rules, the order of
2964 evaluation of subexpressions of an expression is not specified. All
2965 these rules describe only a partial order rather than a total order,
2966 since, for example, if two functions are called within one expression
2967 with no sequence point between them, the order in which the functions
2968 are called is not specified. However, the standards committee have
2969 ruled that function calls do not overlap.
2971 It is not specified when between sequence points modifications to the
2972 values of objects take effect. Programs whose behavior depends on this
2973 have undefined behavior; the C and C++ standards specify that ``Between
2974 the previous and next sequence point an object shall have its stored
2975 value modified at most once by the evaluation of an expression.
2976 Furthermore, the prior value shall be read only to determine the value
2977 to be stored.''. If a program breaks these rules, the results on any
2978 particular implementation are entirely unpredictable.
2980 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2981 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2982 diagnosed by this option, and it may give an occasional false positive
2983 result, but in general it has been found fairly effective at detecting
2984 this sort of problem in programs.
2986 The standard is worded confusingly, therefore there is some debate
2987 over the precise meaning of the sequence point rules in subtle cases.
2988 Links to discussions of the problem, including proposed formal
2989 definitions, may be found on the GCC readings page, at
2990 @w{@uref{http://gcc.gnu.org/readings.html}}.
2992 This warning is enabled by @option{-Wall} for C and C++.
2995 @opindex Wreturn-type
2996 @opindex Wno-return-type
2997 Warn whenever a function is defined with a return-type that defaults
2998 to @code{int}. Also warn about any @code{return} statement with no
2999 return-value in a function whose return-type is not @code{void}
3000 (falling off the end of the function body is considered returning
3001 without a value), and about a @code{return} statement with a
3002 expression in a function whose return-type is @code{void}.
3004 Also warn if the return type of a function has a type qualifier
3005 such as @code{const}. For ISO C such a type qualifier has no effect,
3006 since the value returned by a function is not an lvalue.
3007 For C++, the warning is only emitted for scalar types or @code{void}.
3008 ISO C prohibits qualified @code{void} return types on function
3009 definitions, so such return types always receive a warning
3010 even without this option.
3012 For C++, a function without return type always produces a diagnostic
3013 message, even when @option{-Wno-return-type} is specified. The only
3014 exceptions are @samp{main} and functions defined in system headers.
3016 This warning is enabled by @option{-Wall}.
3020 Warn whenever a @code{switch} statement has an index of enumerated type
3021 and lacks a @code{case} for one or more of the named codes of that
3022 enumeration. (The presence of a @code{default} label prevents this
3023 warning.) @code{case} labels outside the enumeration range also
3024 provoke warnings when this option is used.
3025 This warning is enabled by @option{-Wall}.
3027 @item -Wswitch-default
3028 @opindex Wswitch-switch
3029 Warn whenever a @code{switch} statement does not have a @code{default}
3033 @opindex Wswitch-enum
3034 Warn whenever a @code{switch} statement has an index of enumerated type
3035 and lacks a @code{case} for one or more of the named codes of that
3036 enumeration. @code{case} labels outside the enumeration range also
3037 provoke warnings when this option is used.
3041 Warn if any trigraphs are encountered that might change the meaning of
3042 the program (trigraphs within comments are not warned about).
3043 This warning is enabled by @option{-Wall}.
3045 @item -Wunused-function
3046 @opindex Wunused-function
3047 Warn whenever a static function is declared but not defined or a
3048 non-inline static function is unused.
3049 This warning is enabled by @option{-Wall}.
3051 @item -Wunused-label
3052 @opindex Wunused-label
3053 Warn whenever a label is declared but not used.
3054 This warning is enabled by @option{-Wall}.
3056 To suppress this warning use the @samp{unused} attribute
3057 (@pxref{Variable Attributes}).
3059 @item -Wunused-parameter
3060 @opindex Wunused-parameter
3061 Warn whenever a function parameter is unused aside from its declaration.
3063 To suppress this warning use the @samp{unused} attribute
3064 (@pxref{Variable Attributes}).
3066 @item -Wunused-variable
3067 @opindex Wunused-variable
3068 Warn whenever a local variable or non-constant static variable is unused
3069 aside from its declaration.
3070 This warning is enabled by @option{-Wall}.
3072 To suppress this warning use the @samp{unused} attribute
3073 (@pxref{Variable Attributes}).
3075 @item -Wunused-value
3076 @opindex Wunused-value
3077 Warn whenever a statement computes a result that is explicitly not
3078 used. To suppress this warning cast the unused expression to
3079 @samp{void}. This includes an expression-statement or the left-hand
3080 side of a comma expression that contains no side effects. For example,
3081 an expression such as @samp{x[i,j]} will cause a warning, while
3082 @samp{x[(void)i,j]} will not.
3084 This warning is enabled by @option{-Wall}.
3088 All the above @option{-Wunused} options combined.
3090 In order to get a warning about an unused function parameter, you must
3091 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3092 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3094 @item -Wuninitialized
3095 @opindex Wuninitialized
3096 Warn if an automatic variable is used without first being initialized or
3097 if a variable may be clobbered by a @code{setjmp} call.
3099 These warnings are possible only in optimizing compilation,
3100 because they require data flow information that is computed only
3101 when optimizing. If you do not specify @option{-O}, you will not get
3102 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
3103 requiring @option{-O}.
3105 If you want to warn about code which uses the uninitialized value of the
3106 variable in its own initializer, use the @option{-Winit-self} option.
3108 These warnings occur for individual uninitialized or clobbered
3109 elements of structure, union or array variables as well as for
3110 variables which are uninitialized or clobbered as a whole. They do
3111 not occur for variables or elements declared @code{volatile}. Because
3112 these warnings depend on optimization, the exact variables or elements
3113 for which there are warnings will depend on the precise optimization
3114 options and version of GCC used.
3116 Note that there may be no warning about a variable that is used only
3117 to compute a value that itself is never used, because such
3118 computations may be deleted by data flow analysis before the warnings
3121 These warnings are made optional because GCC is not smart
3122 enough to see all the reasons why the code might be correct
3123 despite appearing to have an error. Here is one example of how
3144 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3145 always initialized, but GCC doesn't know this. Here is
3146 another common case:
3151 if (change_y) save_y = y, y = new_y;
3153 if (change_y) y = save_y;
3158 This has no bug because @code{save_y} is used only if it is set.
3160 @cindex @code{longjmp} warnings
3161 This option also warns when a non-volatile automatic variable might be
3162 changed by a call to @code{longjmp}. These warnings as well are possible
3163 only in optimizing compilation.
3165 The compiler sees only the calls to @code{setjmp}. It cannot know
3166 where @code{longjmp} will be called; in fact, a signal handler could
3167 call it at any point in the code. As a result, you may get a warning
3168 even when there is in fact no problem because @code{longjmp} cannot
3169 in fact be called at the place which would cause a problem.
3171 Some spurious warnings can be avoided if you declare all the functions
3172 you use that never return as @code{noreturn}. @xref{Function
3175 This warning is enabled by @option{-Wall} or @option{-Wextra} in
3176 optimizing compilations (@option{-O1} and above).
3178 @item -Wunknown-pragmas
3179 @opindex Wunknown-pragmas
3180 @cindex warning for unknown pragmas
3181 @cindex unknown pragmas, warning
3182 @cindex pragmas, warning of unknown
3183 Warn when a #pragma directive is encountered which is not understood by
3184 GCC@. If this command line option is used, warnings will even be issued
3185 for unknown pragmas in system header files. This is not the case if
3186 the warnings were only enabled by the @option{-Wall} command line option.
3189 @opindex Wno-pragmas
3191 Do not warn about misuses of pragmas, such as incorrect parameters,
3192 invalid syntax, or conflicts between pragmas. See also
3193 @samp{-Wunknown-pragmas}.
3195 @item -Wstrict-aliasing
3196 @opindex Wstrict-aliasing
3197 This option is only active when @option{-fstrict-aliasing} is active.
3198 It warns about code which might break the strict aliasing rules that the
3199 compiler is using for optimization. The warning does not catch all
3200 cases, but does attempt to catch the more common pitfalls. It is
3201 included in @option{-Wall}.
3202 It is equivalent to @option{-Wstrict-aliasing=3}
3204 @item -Wstrict-aliasing=n
3205 @opindex Wstrict-aliasing=n
3206 This option is only active when @option{-fstrict-aliasing} is active.
3207 It warns about code which might break the strict aliasing rules that the
3208 compiler is using for optimization.
3209 Higher levels correspond to higher accuracy (fewer false positives).
3210 Higher levels also correspond to more effort, similar to the way -O works.
3211 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3214 Level 1: Most aggressive, quick, least accurate.
3215 Possibly useful when higher levels
3216 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3217 false negatives. However, it has many false positives.
3218 Warns for all pointer conversions between possibly incompatible types,
3219 even if never dereferenced. Runs in the frontend only.
3221 Level 2: Aggressive, quick, not too precise.
3222 May still have many false positives (not as many as level 1 though),
3223 and few false negatives (but possibly more than level 1).
3224 Unlike level 1, it only warns when an address is taken. Warns about
3225 incomplete types. Runs in the frontend only.
3227 Level 3 (default for @option{-Wstrict-aliasing}):
3228 Should have very few false positives and few false
3229 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3230 Takes care of the common punn+dereference pattern in the frontend:
3231 @code{*(int*)&some_float}.
3232 If optimization is enabled, it also runs in the backend, where it deals
3233 with multiple statement cases using flow-sensitive points-to information.
3234 Only warns when the converted pointer is dereferenced.
3235 Does not warn about incomplete types.
3237 @item -Wstrict-overflow
3238 @item -Wstrict-overflow=@var{n}
3239 @opindex Wstrict-overflow
3240 This option is only active when @option{-fstrict-overflow} is active.
3241 It warns about cases where the compiler optimizes based on the
3242 assumption that signed overflow does not occur. Note that it does not
3243 warn about all cases where the code might overflow: it only warns
3244 about cases where the compiler implements some optimization. Thus
3245 this warning depends on the optimization level.
3247 An optimization which assumes that signed overflow does not occur is
3248 perfectly safe if the values of the variables involved are such that
3249 overflow never does, in fact, occur. Therefore this warning can
3250 easily give a false positive: a warning about code which is not
3251 actually a problem. To help focus on important issues, several
3252 warning levels are defined. No warnings are issued for the use of
3253 undefined signed overflow when estimating how many iterations a loop
3254 will require, in particular when determining whether a loop will be
3258 @item -Wstrict-overflow=1
3259 Warn about cases which are both questionable and easy to avoid. For
3260 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3261 compiler will simplify this to @code{1}. This level of
3262 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3263 are not, and must be explicitly requested.
3265 @item -Wstrict-overflow=2
3266 Also warn about other cases where a comparison is simplified to a
3267 constant. For example: @code{abs (x) >= 0}. This can only be
3268 simplified when @option{-fstrict-overflow} is in effect, because
3269 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3270 zero. @option{-Wstrict-overflow} (with no level) is the same as
3271 @option{-Wstrict-overflow=2}.
3273 @item -Wstrict-overflow=3
3274 Also warn about other cases where a comparison is simplified. For
3275 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3277 @item -Wstrict-overflow=4
3278 Also warn about other simplifications not covered by the above cases.
3279 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3281 @item -Wstrict-overflow=5
3282 Also warn about cases where the compiler reduces the magnitude of a
3283 constant involved in a comparison. For example: @code{x + 2 > y} will
3284 be simplified to @code{x + 1 >= y}. This is reported only at the
3285 highest warning level because this simplification applies to many
3286 comparisons, so this warning level will give a very large number of
3290 @item -Warray-bounds
3291 @opindex Wno-array-bounds
3292 @opindex Warray-bounds
3293 This option is only active when @option{-ftree-vrp} is active
3294 (default for -O2 and above). It warns about subscripts to arrays
3295 that are always out of bounds. This warning is enabled by @option{-Wall}.
3297 @item -Wno-div-by-zero
3298 @opindex Wno-div-by-zero
3299 @opindex Wdiv-by-zero
3300 Do not warn about compile-time integer division by zero. Floating point
3301 division by zero is not warned about, as it can be a legitimate way of
3302 obtaining infinities and NaNs.
3304 @item -Wsystem-headers
3305 @opindex Wsystem-headers
3306 @cindex warnings from system headers
3307 @cindex system headers, warnings from
3308 Print warning messages for constructs found in system header files.
3309 Warnings from system headers are normally suppressed, on the assumption
3310 that they usually do not indicate real problems and would only make the
3311 compiler output harder to read. Using this command line option tells
3312 GCC to emit warnings from system headers as if they occurred in user
3313 code. However, note that using @option{-Wall} in conjunction with this
3314 option will @emph{not} warn about unknown pragmas in system
3315 headers---for that, @option{-Wunknown-pragmas} must also be used.
3318 @opindex Wfloat-equal
3319 Warn if floating point values are used in equality comparisons.
3321 The idea behind this is that sometimes it is convenient (for the
3322 programmer) to consider floating-point values as approximations to
3323 infinitely precise real numbers. If you are doing this, then you need
3324 to compute (by analyzing the code, or in some other way) the maximum or
3325 likely maximum error that the computation introduces, and allow for it
3326 when performing comparisons (and when producing output, but that's a
3327 different problem). In particular, instead of testing for equality, you
3328 would check to see whether the two values have ranges that overlap; and
3329 this is done with the relational operators, so equality comparisons are
3332 @item -Wtraditional @r{(C only)}
3333 @opindex Wtraditional
3334 Warn about certain constructs that behave differently in traditional and
3335 ISO C@. Also warn about ISO C constructs that have no traditional C
3336 equivalent, and/or problematic constructs which should be avoided.
3340 Macro parameters that appear within string literals in the macro body.
3341 In traditional C macro replacement takes place within string literals,
3342 but does not in ISO C@.
3345 In traditional C, some preprocessor directives did not exist.
3346 Traditional preprocessors would only consider a line to be a directive
3347 if the @samp{#} appeared in column 1 on the line. Therefore
3348 @option{-Wtraditional} warns about directives that traditional C
3349 understands but would ignore because the @samp{#} does not appear as the
3350 first character on the line. It also suggests you hide directives like
3351 @samp{#pragma} not understood by traditional C by indenting them. Some
3352 traditional implementations would not recognize @samp{#elif}, so it
3353 suggests avoiding it altogether.
3356 A function-like macro that appears without arguments.
3359 The unary plus operator.
3362 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3363 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3364 constants.) Note, these suffixes appear in macros defined in the system
3365 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3366 Use of these macros in user code might normally lead to spurious
3367 warnings, however GCC's integrated preprocessor has enough context to
3368 avoid warning in these cases.
3371 A function declared external in one block and then used after the end of
3375 A @code{switch} statement has an operand of type @code{long}.
3378 A non-@code{static} function declaration follows a @code{static} one.
3379 This construct is not accepted by some traditional C compilers.
3382 The ISO type of an integer constant has a different width or
3383 signedness from its traditional type. This warning is only issued if
3384 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3385 typically represent bit patterns, are not warned about.
3388 Usage of ISO string concatenation is detected.
3391 Initialization of automatic aggregates.
3394 Identifier conflicts with labels. Traditional C lacks a separate
3395 namespace for labels.
3398 Initialization of unions. If the initializer is zero, the warning is
3399 omitted. This is done under the assumption that the zero initializer in
3400 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3401 initializer warnings and relies on default initialization to zero in the
3405 Conversions by prototypes between fixed/floating point values and vice
3406 versa. The absence of these prototypes when compiling with traditional
3407 C would cause serious problems. This is a subset of the possible
3408 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3411 Use of ISO C style function definitions. This warning intentionally is
3412 @emph{not} issued for prototype declarations or variadic functions
3413 because these ISO C features will appear in your code when using
3414 libiberty's traditional C compatibility macros, @code{PARAMS} and
3415 @code{VPARAMS}. This warning is also bypassed for nested functions
3416 because that feature is already a GCC extension and thus not relevant to
3417 traditional C compatibility.
3420 @item -Wtraditional-conversion @r{(C only)}
3421 @opindex Wtraditional-conversion
3422 Warn if a prototype causes a type conversion that is different from what
3423 would happen to the same argument in the absence of a prototype. This
3424 includes conversions of fixed point to floating and vice versa, and
3425 conversions changing the width or signedness of a fixed point argument
3426 except when the same as the default promotion.
3428 @item -Wdeclaration-after-statement @r{(C only)}
3429 @opindex Wdeclaration-after-statement
3430 Warn when a declaration is found after a statement in a block. This
3431 construct, known from C++, was introduced with ISO C99 and is by default
3432 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3433 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3437 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3439 @item -Wno-endif-labels
3440 @opindex Wno-endif-labels
3441 @opindex Wendif-labels
3442 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3446 Warn whenever a local variable shadows another local variable, parameter or
3447 global variable or whenever a built-in function is shadowed.
3449 @item -Wlarger-than-@var{len}
3450 @opindex Wlarger-than
3451 Warn whenever an object of larger than @var{len} bytes is defined.
3453 @item -Wunsafe-loop-optimizations
3454 @opindex Wunsafe-loop-optimizations
3455 Warn if the loop cannot be optimized because the compiler could not
3456 assume anything on the bounds of the loop indices. With
3457 @option{-funsafe-loop-optimizations} warn if the compiler made
3460 @item -Wpointer-arith
3461 @opindex Wpointer-arith
3462 Warn about anything that depends on the ``size of'' a function type or
3463 of @code{void}. GNU C assigns these types a size of 1, for
3464 convenience in calculations with @code{void *} pointers and pointers
3465 to functions. In C++, warn also when an arithmetic operation involves
3466 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3469 @opindex Wtype-limits
3470 @opindex Wno-type-limits
3471 Warn if a comparison is always true or always false due to the limited
3472 range of the data type, but do not warn for constant expressions. For
3473 example, warn if an unsigned variable is compared against zero with
3474 @samp{<} or @samp{>=}. This warning is also enabled by
3477 @item -Wbad-function-cast @r{(C only)}
3478 @opindex Wbad-function-cast
3479 Warn whenever a function call is cast to a non-matching type.
3480 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3483 Warn about ISO C constructs that are outside of the common subset of
3484 ISO C and ISO C++, e.g.@: request for implicit conversion from
3485 @code{void *} to a pointer to non-@code{void} type.
3487 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3488 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3489 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3490 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3494 Warn whenever a pointer is cast so as to remove a type qualifier from
3495 the target type. For example, warn if a @code{const char *} is cast
3496 to an ordinary @code{char *}.
3499 @opindex Wcast-align
3500 Warn whenever a pointer is cast such that the required alignment of the
3501 target is increased. For example, warn if a @code{char *} is cast to
3502 an @code{int *} on machines where integers can only be accessed at
3503 two- or four-byte boundaries.
3505 @item -Wwrite-strings
3506 @opindex Wwrite-strings
3507 When compiling C, give string constants the type @code{const
3508 char[@var{length}]} so that
3509 copying the address of one into a non-@code{const} @code{char *}
3510 pointer will get a warning; when compiling C++, warn about the
3511 deprecated conversion from string literals to @code{char *}. This
3512 warning, by default, is enabled for C++ programs.
3513 These warnings will help you find at
3514 compile time code that can try to write into a string constant, but
3515 only if you have been very careful about using @code{const} in
3516 declarations and prototypes. Otherwise, it will just be a nuisance;
3517 this is why we did not make @option{-Wall} request these warnings.
3521 Warn for variables that might be changed by @samp{longjmp} or
3522 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3525 @opindex Wconversion
3526 @opindex Wno-conversion
3527 Warn for implicit conversions that may alter a value. This includes
3528 conversions between real and integer, like @code{abs (x)} when
3529 @code{x} is @code{double}; conversions between signed and unsigned,
3530 like @code{unsigned ui = -1}; and conversions to smaller types, like
3531 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3532 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3533 changed by the conversion like in @code{abs (2.0)}. Warnings about
3534 conversions between signed and unsigned integers can be disabled by
3535 using @option{-Wno-sign-conversion}.
3537 For C++, also warn for conversions between @code{NULL} and non-pointer
3538 types; confusing overload resolution for user-defined conversions; and
3539 conversions that will never use a type conversion operator:
3540 conversions to @code{void}, the same type, a base class or a reference
3541 to them. Warnings about conversions between signed and unsigned
3542 integers are disabled by default in C++ unless
3543 @option{-Wsign-conversion} is explicitly enabled.
3546 @opindex Wempty-body
3547 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3548 while} statement. Additionally, in C++, warn when an empty body occurs
3549 in a @samp{while} or @samp{for} statement with no whitespacing before
3550 the semicolon. This warning is also enabled by @option{-Wextra}.
3552 @item -Wsign-compare
3553 @opindex Wsign-compare
3554 @cindex warning for comparison of signed and unsigned values
3555 @cindex comparison of signed and unsigned values, warning
3556 @cindex signed and unsigned values, comparison warning
3557 Warn when a comparison between signed and unsigned values could produce
3558 an incorrect result when the signed value is converted to unsigned.
3559 This warning is also enabled by @option{-Wextra}; to get the other warnings
3560 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3562 @item -Wsign-conversion
3563 @opindex Wsign-conversion
3564 @opindex Wno-sign-conversion
3565 Warn for implicit conversions that may change the sign of an integer
3566 value, like assigning a signed integer expression to an unsigned
3567 integer variable. An explicit cast silences the warning. In C, this
3568 option is enabled also by @option{-Wconversion}.
3572 @opindex Wno-address
3573 Warn about suspicious uses of memory addresses. These include using
3574 the address of a function in a conditional expression, such as
3575 @code{void func(void); if (func)}, and comparisons against the memory
3576 address of a string literal, such as @code{if (x == "abc")}. Such
3577 uses typically indicate a programmer error: the address of a function
3578 always evaluates to true, so their use in a conditional usually
3579 indicate that the programmer forgot the parentheses in a function
3580 call; and comparisons against string literals result in unspecified
3581 behavior and are not portable in C, so they usually indicate that the
3582 programmer intended to use @code{strcmp}. This warning is enabled by
3586 @opindex Wlogical-op
3587 @opindex Wno-logical-op
3588 Warn about suspicious uses of logical operators in expressions.
3589 This includes using logical operators in contexts where a
3590 bit-wise operator is likely to be expected.
3592 @item -Waggregate-return
3593 @opindex Waggregate-return
3594 Warn if any functions that return structures or unions are defined or
3595 called. (In languages where you can return an array, this also elicits
3598 @item -Wno-attributes
3599 @opindex Wno-attributes
3600 @opindex Wattributes
3601 Do not warn if an unexpected @code{__attribute__} is used, such as
3602 unrecognized attributes, function attributes applied to variables,
3603 etc. This will not stop errors for incorrect use of supported
3606 @item -Wstrict-prototypes @r{(C only)}
3607 @opindex Wstrict-prototypes
3608 Warn if a function is declared or defined without specifying the
3609 argument types. (An old-style function definition is permitted without
3610 a warning if preceded by a declaration which specifies the argument
3613 @item -Wold-style-declaration @r{(C only)}
3614 @opindex Wold-style-declaration
3615 Warn for obsolescent usages, according to the C Standard, in a
3616 declaration. For example, warn if storage-class specifiers like
3617 @code{static} are not the first things in a declaration. This warning
3618 is also enabled by @option{-Wextra}.
3620 @item -Wold-style-definition @r{(C only)}
3621 @opindex Wold-style-definition
3622 Warn if an old-style function definition is used. A warning is given
3623 even if there is a previous prototype.
3625 @item -Wmissing-parameter-type @r{(C only)}
3626 @opindex Wmissing-parameter-type
3627 A function parameter is declared without a type specifier in K&R-style
3634 This warning is also enabled by @option{-Wextra}.
3636 @item -Wmissing-prototypes @r{(C only)}
3637 @opindex Wmissing-prototypes
3638 Warn if a global function is defined without a previous prototype
3639 declaration. This warning is issued even if the definition itself
3640 provides a prototype. The aim is to detect global functions that fail
3641 to be declared in header files.
3643 @item -Wmissing-declarations @r{(C and C++ only)}
3644 @opindex Wmissing-declarations
3645 Warn if a global function is defined without a previous declaration.
3646 Do so even if the definition itself provides a prototype.
3647 Use this option to detect global functions that are not declared in
3648 header files. In C++, no warnings are issued for function templates,
3649 or for inline functions, or for functions in anonymous namespaces.
3651 @item -Wmissing-field-initializers
3652 @opindex Wmissing-field-initializers
3655 Warn if a structure's initializer has some fields missing. For
3656 example, the following code would cause such a warning, because
3657 @code{x.h} is implicitly zero:
3660 struct s @{ int f, g, h; @};
3661 struct s x = @{ 3, 4 @};
3664 This option does not warn about designated initializers, so the following
3665 modification would not trigger a warning:
3668 struct s @{ int f, g, h; @};
3669 struct s x = @{ .f = 3, .g = 4 @};
3672 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3673 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3675 @item -Wmissing-noreturn
3676 @opindex Wmissing-noreturn
3677 Warn about functions which might be candidates for attribute @code{noreturn}.
3678 Note these are only possible candidates, not absolute ones. Care should
3679 be taken to manually verify functions actually do not ever return before
3680 adding the @code{noreturn} attribute, otherwise subtle code generation
3681 bugs could be introduced. You will not get a warning for @code{main} in
3682 hosted C environments.
3684 @item -Wmissing-format-attribute
3685 @opindex Wmissing-format-attribute
3687 Warn about function pointers which might be candidates for @code{format}
3688 attributes. Note these are only possible candidates, not absolute ones.
3689 GCC will guess that function pointers with @code{format} attributes that
3690 are used in assignment, initialization, parameter passing or return
3691 statements should have a corresponding @code{format} attribute in the
3692 resulting type. I.e.@: the left-hand side of the assignment or
3693 initialization, the type of the parameter variable, or the return type
3694 of the containing function respectively should also have a @code{format}
3695 attribute to avoid the warning.
3697 GCC will also warn about function definitions which might be
3698 candidates for @code{format} attributes. Again, these are only
3699 possible candidates. GCC will guess that @code{format} attributes
3700 might be appropriate for any function that calls a function like
3701 @code{vprintf} or @code{vscanf}, but this might not always be the
3702 case, and some functions for which @code{format} attributes are
3703 appropriate may not be detected.
3705 @item -Wno-multichar
3706 @opindex Wno-multichar
3708 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3709 Usually they indicate a typo in the user's code, as they have
3710 implementation-defined values, and should not be used in portable code.
3712 @item -Wnormalized=<none|id|nfc|nfkc>
3713 @opindex Wnormalized
3716 @cindex character set, input normalization
3717 In ISO C and ISO C++, two identifiers are different if they are
3718 different sequences of characters. However, sometimes when characters
3719 outside the basic ASCII character set are used, you can have two
3720 different character sequences that look the same. To avoid confusion,
3721 the ISO 10646 standard sets out some @dfn{normalization rules} which
3722 when applied ensure that two sequences that look the same are turned into
3723 the same sequence. GCC can warn you if you are using identifiers which
3724 have not been normalized; this option controls that warning.
3726 There are four levels of warning that GCC supports. The default is
3727 @option{-Wnormalized=nfc}, which warns about any identifier which is
3728 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3729 recommended form for most uses.
3731 Unfortunately, there are some characters which ISO C and ISO C++ allow
3732 in identifiers that when turned into NFC aren't allowable as
3733 identifiers. That is, there's no way to use these symbols in portable
3734 ISO C or C++ and have all your identifiers in NFC.
3735 @option{-Wnormalized=id} suppresses the warning for these characters.
3736 It is hoped that future versions of the standards involved will correct
3737 this, which is why this option is not the default.
3739 You can switch the warning off for all characters by writing
3740 @option{-Wnormalized=none}. You would only want to do this if you
3741 were using some other normalization scheme (like ``D''), because
3742 otherwise you can easily create bugs that are literally impossible to see.
3744 Some characters in ISO 10646 have distinct meanings but look identical
3745 in some fonts or display methodologies, especially once formatting has
3746 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3747 LETTER N'', will display just like a regular @code{n} which has been
3748 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3749 normalization scheme to convert all these into a standard form as
3750 well, and GCC will warn if your code is not in NFKC if you use
3751 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3752 about every identifier that contains the letter O because it might be
3753 confused with the digit 0, and so is not the default, but may be
3754 useful as a local coding convention if the programming environment is
3755 unable to be fixed to display these characters distinctly.
3757 @item -Wno-deprecated-declarations
3758 @opindex Wno-deprecated-declarations
3759 Do not warn about uses of functions (@pxref{Function Attributes}),
3760 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3761 Attributes}) marked as deprecated by using the @code{deprecated}
3765 @opindex Wno-overflow
3766 Do not warn about compile-time overflow in constant expressions.
3768 @item -Woverride-init
3769 @opindex Woverride-init
3772 Warn if an initialized field without side effects is overridden when
3773 using designated initializers (@pxref{Designated Inits, , Designated
3776 This warning is included in @option{-Wextra}. To get other
3777 @option{-Wextra} warnings without this one, use @samp{-Wextra
3778 -Wno-override-init}.
3782 Warn if a structure is given the packed attribute, but the packed
3783 attribute has no effect on the layout or size of the structure.
3784 Such structures may be mis-aligned for little benefit. For
3785 instance, in this code, the variable @code{f.x} in @code{struct bar}
3786 will be misaligned even though @code{struct bar} does not itself
3787 have the packed attribute:
3794 @} __attribute__((packed));
3804 Warn if padding is included in a structure, either to align an element
3805 of the structure or to align the whole structure. Sometimes when this
3806 happens it is possible to rearrange the fields of the structure to
3807 reduce the padding and so make the structure smaller.
3809 @item -Wredundant-decls
3810 @opindex Wredundant-decls
3811 Warn if anything is declared more than once in the same scope, even in
3812 cases where multiple declaration is valid and changes nothing.
3814 @item -Wnested-externs @r{(C only)}
3815 @opindex Wnested-externs
3816 Warn if an @code{extern} declaration is encountered within a function.
3818 @item -Wunreachable-code
3819 @opindex Wunreachable-code
3820 Warn if the compiler detects that code will never be executed.
3822 This option is intended to warn when the compiler detects that at
3823 least a whole line of source code will never be executed, because
3824 some condition is never satisfied or because it is after a
3825 procedure that never returns.
3827 It is possible for this option to produce a warning even though there
3828 are circumstances under which part of the affected line can be executed,
3829 so care should be taken when removing apparently-unreachable code.
3831 For instance, when a function is inlined, a warning may mean that the
3832 line is unreachable in only one inlined copy of the function.
3834 This option is not made part of @option{-Wall} because in a debugging
3835 version of a program there is often substantial code which checks
3836 correct functioning of the program and is, hopefully, unreachable
3837 because the program does work. Another common use of unreachable
3838 code is to provide behavior which is selectable at compile-time.
3842 Warn if a function can not be inlined and it was declared as inline.
3843 Even with this option, the compiler will not warn about failures to
3844 inline functions declared in system headers.
3846 The compiler uses a variety of heuristics to determine whether or not
3847 to inline a function. For example, the compiler takes into account
3848 the size of the function being inlined and the amount of inlining
3849 that has already been done in the current function. Therefore,
3850 seemingly insignificant changes in the source program can cause the
3851 warnings produced by @option{-Winline} to appear or disappear.
3853 @item -Wno-invalid-offsetof @r{(C++ only)}
3854 @opindex Wno-invalid-offsetof
3855 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3856 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3857 to a non-POD type is undefined. In existing C++ implementations,
3858 however, @samp{offsetof} typically gives meaningful results even when
3859 applied to certain kinds of non-POD types. (Such as a simple
3860 @samp{struct} that fails to be a POD type only by virtue of having a
3861 constructor.) This flag is for users who are aware that they are
3862 writing nonportable code and who have deliberately chosen to ignore the
3865 The restrictions on @samp{offsetof} may be relaxed in a future version
3866 of the C++ standard.
3868 @item -Wno-int-to-pointer-cast @r{(C only)}
3869 @opindex Wno-int-to-pointer-cast
3870 Suppress warnings from casts to pointer type of an integer of a
3873 @item -Wno-pointer-to-int-cast @r{(C only)}
3874 @opindex Wno-pointer-to-int-cast
3875 Suppress warnings from casts from a pointer to an integer type of a
3879 @opindex Winvalid-pch
3880 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3881 the search path but can't be used.
3885 @opindex Wno-long-long
3886 Warn if @samp{long long} type is used. This is default. To inhibit
3887 the warning messages, use @option{-Wno-long-long}. Flags
3888 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3889 only when @option{-pedantic} flag is used.
3891 @item -Wvariadic-macros
3892 @opindex Wvariadic-macros
3893 @opindex Wno-variadic-macros
3894 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3895 alternate syntax when in pedantic ISO C99 mode. This is default.
3896 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3901 Warn if variable length array is used in the code.
3902 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
3903 the variable length array.
3905 @item -Wvolatile-register-var
3906 @opindex Wvolatile-register-var
3907 @opindex Wno-volatile-register-var
3908 Warn if a register variable is declared volatile. The volatile
3909 modifier does not inhibit all optimizations that may eliminate reads
3910 and/or writes to register variables.
3912 @item -Wdisabled-optimization
3913 @opindex Wdisabled-optimization
3914 Warn if a requested optimization pass is disabled. This warning does
3915 not generally indicate that there is anything wrong with your code; it
3916 merely indicates that GCC's optimizers were unable to handle the code
3917 effectively. Often, the problem is that your code is too big or too
3918 complex; GCC will refuse to optimize programs when the optimization
3919 itself is likely to take inordinate amounts of time.
3921 @item -Wpointer-sign
3922 @opindex Wpointer-sign
3923 @opindex Wno-pointer-sign
3924 Warn for pointer argument passing or assignment with different signedness.
3925 This option is only supported for C and Objective-C@. It is implied by
3926 @option{-Wall} and by @option{-pedantic}, which can be disabled with
3927 @option{-Wno-pointer-sign}.
3929 @item -Wstack-protector
3930 @opindex Wstack-protector
3931 This option is only active when @option{-fstack-protector} is active. It
3932 warns about functions that will not be protected against stack smashing.
3934 @item -Woverlength-strings
3935 @opindex Woverlength-strings
3936 Warn about string constants which are longer than the ``minimum
3937 maximum'' length specified in the C standard. Modern compilers
3938 generally allow string constants which are much longer than the
3939 standard's minimum limit, but very portable programs should avoid
3940 using longer strings.
3942 The limit applies @emph{after} string constant concatenation, and does
3943 not count the trailing NUL@. In C89, the limit was 509 characters; in
3944 C99, it was raised to 4095. C++98 does not specify a normative
3945 minimum maximum, so we do not diagnose overlength strings in C++@.
3947 This option is implied by @option{-pedantic}, and can be disabled with
3948 @option{-Wno-overlength-strings}.
3951 @node Debugging Options
3952 @section Options for Debugging Your Program or GCC
3953 @cindex options, debugging
3954 @cindex debugging information options
3956 GCC has various special options that are used for debugging
3957 either your program or GCC:
3962 Produce debugging information in the operating system's native format
3963 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3966 On most systems that use stabs format, @option{-g} enables use of extra
3967 debugging information that only GDB can use; this extra information
3968 makes debugging work better in GDB but will probably make other debuggers
3970 refuse to read the program. If you want to control for certain whether
3971 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3972 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3974 GCC allows you to use @option{-g} with
3975 @option{-O}. The shortcuts taken by optimized code may occasionally
3976 produce surprising results: some variables you declared may not exist
3977 at all; flow of control may briefly move where you did not expect it;
3978 some statements may not be executed because they compute constant
3979 results or their values were already at hand; some statements may
3980 execute in different places because they were moved out of loops.
3982 Nevertheless it proves possible to debug optimized output. This makes
3983 it reasonable to use the optimizer for programs that might have bugs.
3985 The following options are useful when GCC is generated with the
3986 capability for more than one debugging format.
3990 Produce debugging information for use by GDB@. This means to use the
3991 most expressive format available (DWARF 2, stabs, or the native format
3992 if neither of those are supported), including GDB extensions if at all
3997 Produce debugging information in stabs format (if that is supported),
3998 without GDB extensions. This is the format used by DBX on most BSD
3999 systems. On MIPS, Alpha and System V Release 4 systems this option
4000 produces stabs debugging output which is not understood by DBX or SDB@.
4001 On System V Release 4 systems this option requires the GNU assembler.
4003 @item -feliminate-unused-debug-symbols
4004 @opindex feliminate-unused-debug-symbols
4005 Produce debugging information in stabs format (if that is supported),
4006 for only symbols that are actually used.
4008 @item -femit-class-debug-always
4009 Instead of emitting debugging information for a C++ class in only one
4010 object file, emit it in all object files using the class. This option
4011 should be used only with debuggers that are unable to handle the way GCC
4012 normally emits debugging information for classes because using this
4013 option will increase the size of debugging information by as much as a
4018 Produce debugging information in stabs format (if that is supported),
4019 using GNU extensions understood only by the GNU debugger (GDB)@. The
4020 use of these extensions is likely to make other debuggers crash or
4021 refuse to read the program.
4025 Produce debugging information in COFF format (if that is supported).
4026 This is the format used by SDB on most System V systems prior to
4031 Produce debugging information in XCOFF format (if that is supported).
4032 This is the format used by the DBX debugger on IBM RS/6000 systems.
4036 Produce debugging information in XCOFF format (if that is supported),
4037 using GNU extensions understood only by the GNU debugger (GDB)@. The
4038 use of these extensions is likely to make other debuggers crash or
4039 refuse to read the program, and may cause assemblers other than the GNU
4040 assembler (GAS) to fail with an error.
4044 Produce debugging information in DWARF version 2 format (if that is
4045 supported). This is the format used by DBX on IRIX 6. With this
4046 option, GCC uses features of DWARF version 3 when they are useful;
4047 version 3 is upward compatible with version 2, but may still cause
4048 problems for older debuggers.
4052 Produce debugging information in VMS debug format (if that is
4053 supported). This is the format used by DEBUG on VMS systems.
4056 @itemx -ggdb@var{level}
4057 @itemx -gstabs@var{level}
4058 @itemx -gcoff@var{level}
4059 @itemx -gxcoff@var{level}
4060 @itemx -gvms@var{level}
4061 Request debugging information and also use @var{level} to specify how
4062 much information. The default level is 2.
4064 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4067 Level 1 produces minimal information, enough for making backtraces in
4068 parts of the program that you don't plan to debug. This includes
4069 descriptions of functions and external variables, but no information
4070 about local variables and no line numbers.
4072 Level 3 includes extra information, such as all the macro definitions
4073 present in the program. Some debuggers support macro expansion when
4074 you use @option{-g3}.
4076 @option{-gdwarf-2} does not accept a concatenated debug level, because
4077 GCC used to support an option @option{-gdwarf} that meant to generate
4078 debug information in version 1 of the DWARF format (which is very
4079 different from version 2), and it would have been too confusing. That
4080 debug format is long obsolete, but the option cannot be changed now.
4081 Instead use an additional @option{-g@var{level}} option to change the
4082 debug level for DWARF2.
4084 @item -feliminate-dwarf2-dups
4085 @opindex feliminate-dwarf2-dups
4086 Compress DWARF2 debugging information by eliminating duplicated
4087 information about each symbol. This option only makes sense when
4088 generating DWARF2 debugging information with @option{-gdwarf-2}.
4090 @item -femit-struct-debug-baseonly
4091 Emit debug information for struct-like types
4092 only when the base name of the compilation source file
4093 matches the base name of file in which the struct was defined.
4095 This option substantially reduces the size of debugging information,
4096 but at significant potential loss in type information to the debugger.
4097 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4098 See @option{-femit-struct-debug-detailed} for more detailed control.
4100 This option works only with DWARF 2.
4102 @item -femit-struct-debug-reduced
4103 Emit debug information for struct-like types
4104 only when the base name of the compilation source file
4105 matches the base name of file in which the type was defined,
4106 unless the struct is a template or defined in a system header.
4108 This option significantly reduces the size of debugging information,
4109 with some potential loss in type information to the debugger.
4110 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4111 See @option{-femit-struct-debug-detailed} for more detailed control.
4113 This option works only with DWARF 2.
4115 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4116 Specify the struct-like types
4117 for which the compiler will generate debug information.
4118 The intent is to reduce duplicate struct debug information
4119 between different object files within the same program.
4121 This option is a detailed version of
4122 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4123 which will serve for most needs.
4125 A specification has the syntax
4126 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4128 The optional first word limits the specification to
4129 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4130 A struct type is used directly when it is the type of a variable, member.
4131 Indirect uses arise through pointers to structs.
4132 That is, when use of an incomplete struct would be legal, the use is indirect.
4134 @samp{struct one direct; struct two * indirect;}.
4136 The optional second word limits the specification to
4137 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4138 Generic structs are a bit complicated to explain.
4139 For C++, these are non-explicit specializations of template classes,
4140 or non-template classes within the above.
4141 Other programming languages have generics,
4142 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4144 The third word specifies the source files for those
4145 structs for which the compiler will emit debug information.
4146 The values @samp{none} and @samp{any} have the normal meaning.
4147 The value @samp{base} means that
4148 the base of name of the file in which the type declaration appears
4149 must match the base of the name of the main compilation file.
4150 In practice, this means that
4151 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4152 but types declared in other header will not.
4153 The value @samp{sys} means those types satisfying @samp{base}
4154 or declared in system or compiler headers.
4156 You may need to experiment to determine the best settings for your application.
4158 The default is @samp{-femit-struct-debug-detailed=all}.
4160 This option works only with DWARF 2.
4162 @item -fdebug-prefix-map=@var{old}=@var{new}
4163 @opindex fdebug-prefix-map
4164 When compiling files in directory @file{@var{old}}, record debugging
4165 information describing them as in @file{@var{new}} instead.
4167 @cindex @command{prof}
4170 Generate extra code to write profile information suitable for the
4171 analysis program @command{prof}. You must use this option when compiling
4172 the source files you want data about, and you must also use it when
4175 @cindex @command{gprof}
4178 Generate extra code to write profile information suitable for the
4179 analysis program @command{gprof}. You must use this option when compiling
4180 the source files you want data about, and you must also use it when
4185 Makes the compiler print out each function name as it is compiled, and
4186 print some statistics about each pass when it finishes.
4189 @opindex ftime-report
4190 Makes the compiler print some statistics about the time consumed by each
4191 pass when it finishes.
4194 @opindex fmem-report
4195 Makes the compiler print some statistics about permanent memory
4196 allocation when it finishes.
4198 @item -fpre-ipa-mem-report
4199 @opindex fpre-ipa-mem-report
4200 @item -fpost-ipa-mem-report
4201 @opindex fpost-ipa-mem-report
4202 Makes the compiler print some statistics about permanent memory
4203 allocation before or after interprocedural optimization.
4205 @item -fprofile-arcs
4206 @opindex fprofile-arcs
4207 Add code so that program flow @dfn{arcs} are instrumented. During
4208 execution the program records how many times each branch and call is
4209 executed and how many times it is taken or returns. When the compiled
4210 program exits it saves this data to a file called
4211 @file{@var{auxname}.gcda} for each source file. The data may be used for
4212 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4213 test coverage analysis (@option{-ftest-coverage}). Each object file's
4214 @var{auxname} is generated from the name of the output file, if
4215 explicitly specified and it is not the final executable, otherwise it is
4216 the basename of the source file. In both cases any suffix is removed
4217 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4218 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4219 @xref{Cross-profiling}.
4221 @cindex @command{gcov}
4225 This option is used to compile and link code instrumented for coverage
4226 analysis. The option is a synonym for @option{-fprofile-arcs}
4227 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4228 linking). See the documentation for those options for more details.
4233 Compile the source files with @option{-fprofile-arcs} plus optimization
4234 and code generation options. For test coverage analysis, use the
4235 additional @option{-ftest-coverage} option. You do not need to profile
4236 every source file in a program.
4239 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4240 (the latter implies the former).
4243 Run the program on a representative workload to generate the arc profile
4244 information. This may be repeated any number of times. You can run
4245 concurrent instances of your program, and provided that the file system
4246 supports locking, the data files will be correctly updated. Also
4247 @code{fork} calls are detected and correctly handled (double counting
4251 For profile-directed optimizations, compile the source files again with
4252 the same optimization and code generation options plus
4253 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4254 Control Optimization}).
4257 For test coverage analysis, use @command{gcov} to produce human readable
4258 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4259 @command{gcov} documentation for further information.
4263 With @option{-fprofile-arcs}, for each function of your program GCC
4264 creates a program flow graph, then finds a spanning tree for the graph.
4265 Only arcs that are not on the spanning tree have to be instrumented: the
4266 compiler adds code to count the number of times that these arcs are
4267 executed. When an arc is the only exit or only entrance to a block, the
4268 instrumentation code can be added to the block; otherwise, a new basic
4269 block must be created to hold the instrumentation code.
4272 @item -ftest-coverage
4273 @opindex ftest-coverage
4274 Produce a notes file that the @command{gcov} code-coverage utility
4275 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4276 show program coverage. Each source file's note file is called
4277 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4278 above for a description of @var{auxname} and instructions on how to
4279 generate test coverage data. Coverage data will match the source files
4280 more closely, if you do not optimize.
4282 @item -fdbg-cnt-list
4283 @opindex fdbg-cnt-list
4284 Print the name and the counter upperbound for all debug counters.
4286 @item -fdbg-cnt=@var{counter-value-list}
4288 Set the internal debug counter upperbound. @var{counter-value-list}
4289 is a comma-separated list of @var{name}:@var{value} pairs
4290 which sets the upperbound of each debug counter @var{name} to @var{value}.
4291 All debug counters have the initial upperbound of @var{UINT_MAX},
4292 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4293 e.g. With -fdbg-cnt=dce:10,tail_call:0
4294 dbg_cnt(dce) will return true only for first 10 invocations
4295 and dbg_cnt(tail_call) will return false always.
4297 @item -d@var{letters}
4298 @item -fdump-rtl-@var{pass}
4300 Says to make debugging dumps during compilation at times specified by
4301 @var{letters}. This is used for debugging the RTL-based passes of the
4302 compiler. The file names for most of the dumps are made by appending a
4303 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4304 from the name of the output file, if explicitly specified and it is not
4305 an executable, otherwise it is the basename of the source file. These
4306 switches may have different effects when @option{-E} is used for
4309 Most debug dumps can be enabled either passing a letter to the @option{-d}
4310 option, or with a long @option{-fdump-rtl} switch; here are the possible
4311 letters for use in @var{letters} and @var{pass}, and their meanings:
4316 Annotate the assembler output with miscellaneous debugging information.
4319 @itemx -fdump-rtl-bbro
4321 @opindex fdump-rtl-bbro
4322 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4325 @itemx -fdump-rtl-combine
4327 @opindex fdump-rtl-combine
4328 Dump after the RTL instruction combination pass, to the file
4329 @file{@var{file}.129r.combine}.
4332 @itemx -fdump-rtl-ce1
4333 @itemx -fdump-rtl-ce2
4335 @opindex fdump-rtl-ce1
4336 @opindex fdump-rtl-ce2
4337 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
4338 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
4339 and @option{-fdump-rtl-ce2} enable dumping after the second if
4340 conversion, to the file @file{@var{file}.130r.ce2}.
4343 @itemx -fdump-rtl-btl
4344 @itemx -fdump-rtl-dbr
4346 @opindex fdump-rtl-btl
4347 @opindex fdump-rtl-dbr
4348 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
4349 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
4350 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
4351 scheduling, to @file{@var{file}.36.dbr}.
4355 Dump all macro definitions, at the end of preprocessing, in addition to
4359 @itemx -fdump-rtl-ce3
4361 @opindex fdump-rtl-ce3
4362 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4365 @itemx -fdump-rtl-cfg
4366 @itemx -fdump-rtl-life
4368 @opindex fdump-rtl-cfg
4369 @opindex fdump-rtl-life
4370 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
4371 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
4372 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4373 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4376 @itemx -fdump-rtl-greg
4378 @opindex fdump-rtl-greg
4379 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4382 @itemx -fdump-rtl-gcse
4383 @itemx -fdump-rtl-bypass
4385 @opindex fdump-rtl-gcse
4386 @opindex fdump-rtl-bypass
4387 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4388 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
4389 enable dumping after jump bypassing and control flow optimizations, to
4390 @file{@var{file}.115r.bypass}.
4393 @itemx -fdump-rtl-eh
4395 @opindex fdump-rtl-eh
4396 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4399 @itemx -fdump-rtl-sibling
4401 @opindex fdump-rtl-sibling
4402 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4405 @itemx -fdump-rtl-jump
4407 @opindex fdump-rtl-jump
4408 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4411 @itemx -fdump-rtl-stack
4413 @opindex fdump-rtl-stack
4414 Dump after conversion from GCC's "flat register file" registers to the
4415 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4418 @itemx -fdump-rtl-lreg
4420 @opindex fdump-rtl-lreg
4421 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4424 @itemx -fdump-rtl-loop2
4426 @opindex fdump-rtl-loop2
4427 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4428 loop optimization pass, to @file{@var{file}.119r.loop2},
4429 @file{@var{file}.120r.loop2_init},
4430 @file{@var{file}.121r.loop2_invariant}, and
4431 @file{@var{file}.125r.loop2_done}.
4434 @itemx -fdump-rtl-sms
4436 @opindex fdump-rtl-sms
4437 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4440 @itemx -fdump-rtl-mach
4442 @opindex fdump-rtl-mach
4443 Dump after performing the machine dependent reorganization pass, to
4444 @file{@var{file}.155r.mach} if that pass exists.
4447 @itemx -fdump-rtl-rnreg
4449 @opindex fdump-rtl-rnreg
4450 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4453 @itemx -fdump-rtl-regmove
4455 @opindex fdump-rtl-regmove
4456 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4459 @itemx -fdump-rtl-postreload
4461 @opindex fdump-rtl-postreload
4462 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4465 @itemx -fdump-rtl-expand
4467 @opindex fdump-rtl-expand
4468 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4471 @itemx -fdump-rtl-sched2
4473 @opindex fdump-rtl-sched2
4474 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4477 @itemx -fdump-rtl-cse
4479 @opindex fdump-rtl-cse
4480 Dump after CSE (including the jump optimization that sometimes follows
4481 CSE), to @file{@var{file}.113r.cse}.
4484 @itemx -fdump-rtl-sched1
4486 @opindex fdump-rtl-sched1
4487 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4490 @itemx -fdump-rtl-cse2
4492 @opindex fdump-rtl-cse2
4493 Dump after the second CSE pass (including the jump optimization that
4494 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4497 @itemx -fdump-rtl-tracer
4499 @opindex fdump-rtl-tracer
4500 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4503 @itemx -fdump-rtl-vpt
4504 @itemx -fdump-rtl-vartrack
4506 @opindex fdump-rtl-vpt
4507 @opindex fdump-rtl-vartrack
4508 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
4509 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
4510 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4511 to @file{@var{file}.154r.vartrack}.
4514 @itemx -fdump-rtl-flow2
4516 @opindex fdump-rtl-flow2
4517 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4520 @itemx -fdump-rtl-peephole2
4522 @opindex fdump-rtl-peephole2
4523 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4526 @itemx -fdump-rtl-web
4528 @opindex fdump-rtl-web
4529 Dump after live range splitting, to @file{@var{file}.126r.web}.
4532 @itemx -fdump-rtl-all
4534 @opindex fdump-rtl-all
4535 Produce all the dumps listed above.
4539 Produce a core dump whenever an error occurs.
4543 Print statistics on memory usage, at the end of the run, to
4548 Annotate the assembler output with a comment indicating which
4549 pattern and alternative was used. The length of each instruction is
4554 Dump the RTL in the assembler output as a comment before each instruction.
4555 Also turns on @option{-dp} annotation.
4559 For each of the other indicated dump files (either with @option{-d} or
4560 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4561 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4565 Just generate RTL for a function instead of compiling it. Usually used
4566 with @samp{r} (@option{-fdump-rtl-expand}).
4570 Dump debugging information during parsing, to standard error.
4574 @opindex fdump-noaddr
4575 When doing debugging dumps (see @option{-d} option above), suppress
4576 address output. This makes it more feasible to use diff on debugging
4577 dumps for compiler invocations with different compiler binaries and/or
4578 different text / bss / data / heap / stack / dso start locations.
4580 @item -fdump-unnumbered
4581 @opindex fdump-unnumbered
4582 When doing debugging dumps (see @option{-d} option above), suppress instruction
4583 numbers and address output. This makes it more feasible to
4584 use diff on debugging dumps for compiler invocations with different
4585 options, in particular with and without @option{-g}.
4587 @item -fdump-translation-unit @r{(C++ only)}
4588 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4589 @opindex fdump-translation-unit
4590 Dump a representation of the tree structure for the entire translation
4591 unit to a file. The file name is made by appending @file{.tu} to the
4592 source file name. If the @samp{-@var{options}} form is used, @var{options}
4593 controls the details of the dump as described for the
4594 @option{-fdump-tree} options.
4596 @item -fdump-class-hierarchy @r{(C++ only)}
4597 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4598 @opindex fdump-class-hierarchy
4599 Dump a representation of each class's hierarchy and virtual function
4600 table layout to a file. The file name is made by appending @file{.class}
4601 to the source file name. If the @samp{-@var{options}} form is used,
4602 @var{options} controls the details of the dump as described for the
4603 @option{-fdump-tree} options.
4605 @item -fdump-ipa-@var{switch}
4607 Control the dumping at various stages of inter-procedural analysis
4608 language tree to a file. The file name is generated by appending a switch
4609 specific suffix to the source file name. The following dumps are possible:
4613 Enables all inter-procedural analysis dumps; currently the only produced
4614 dump is the @samp{cgraph} dump.
4617 Dumps information about call-graph optimization, unused function removal,
4618 and inlining decisions.
4621 @item -fdump-tree-@var{switch}
4622 @itemx -fdump-tree-@var{switch}-@var{options}
4624 Control the dumping at various stages of processing the intermediate
4625 language tree to a file. The file name is generated by appending a switch
4626 specific suffix to the source file name. If the @samp{-@var{options}}
4627 form is used, @var{options} is a list of @samp{-} separated options that
4628 control the details of the dump. Not all options are applicable to all
4629 dumps, those which are not meaningful will be ignored. The following
4630 options are available
4634 Print the address of each node. Usually this is not meaningful as it
4635 changes according to the environment and source file. Its primary use
4636 is for tying up a dump file with a debug environment.
4638 Inhibit dumping of members of a scope or body of a function merely
4639 because that scope has been reached. Only dump such items when they
4640 are directly reachable by some other path. When dumping pretty-printed
4641 trees, this option inhibits dumping the bodies of control structures.
4643 Print a raw representation of the tree. By default, trees are
4644 pretty-printed into a C-like representation.
4646 Enable more detailed dumps (not honored by every dump option).
4648 Enable dumping various statistics about the pass (not honored by every dump
4651 Enable showing basic block boundaries (disabled in raw dumps).
4653 Enable showing virtual operands for every statement.
4655 Enable showing line numbers for statements.
4657 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4659 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4662 The following tree dumps are possible:
4666 Dump before any tree based optimization, to @file{@var{file}.original}.
4669 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4672 Dump after function inlining, to @file{@var{file}.inlined}.
4675 @opindex fdump-tree-gimple
4676 Dump each function before and after the gimplification pass to a file. The
4677 file name is made by appending @file{.gimple} to the source file name.
4680 @opindex fdump-tree-cfg
4681 Dump the control flow graph of each function to a file. The file name is
4682 made by appending @file{.cfg} to the source file name.
4685 @opindex fdump-tree-vcg
4686 Dump the control flow graph of each function to a file in VCG format. The
4687 file name is made by appending @file{.vcg} to the source file name. Note
4688 that if the file contains more than one function, the generated file cannot
4689 be used directly by VCG@. You will need to cut and paste each function's
4690 graph into its own separate file first.
4693 @opindex fdump-tree-ch
4694 Dump each function after copying loop headers. The file name is made by
4695 appending @file{.ch} to the source file name.
4698 @opindex fdump-tree-ssa
4699 Dump SSA related information to a file. The file name is made by appending
4700 @file{.ssa} to the source file name.
4703 @opindex fdump-tree-salias
4704 Dump structure aliasing variable information to a file. This file name
4705 is made by appending @file{.salias} to the source file name.
4708 @opindex fdump-tree-alias
4709 Dump aliasing information for each function. The file name is made by
4710 appending @file{.alias} to the source file name.
4713 @opindex fdump-tree-ccp
4714 Dump each function after CCP@. The file name is made by appending
4715 @file{.ccp} to the source file name.
4718 @opindex fdump-tree-storeccp
4719 Dump each function after STORE-CCP. The file name is made by appending
4720 @file{.storeccp} to the source file name.
4723 @opindex fdump-tree-pre
4724 Dump trees after partial redundancy elimination. The file name is made
4725 by appending @file{.pre} to the source file name.
4728 @opindex fdump-tree-fre
4729 Dump trees after full redundancy elimination. The file name is made
4730 by appending @file{.fre} to the source file name.
4733 @opindex fdump-tree-copyprop
4734 Dump trees after copy propagation. The file name is made
4735 by appending @file{.copyprop} to the source file name.
4737 @item store_copyprop
4738 @opindex fdump-tree-store_copyprop
4739 Dump trees after store copy-propagation. The file name is made
4740 by appending @file{.store_copyprop} to the source file name.
4743 @opindex fdump-tree-dce
4744 Dump each function after dead code elimination. The file name is made by
4745 appending @file{.dce} to the source file name.
4748 @opindex fdump-tree-mudflap
4749 Dump each function after adding mudflap instrumentation. The file name is
4750 made by appending @file{.mudflap} to the source file name.
4753 @opindex fdump-tree-sra
4754 Dump each function after performing scalar replacement of aggregates. The
4755 file name is made by appending @file{.sra} to the source file name.
4758 @opindex fdump-tree-sink
4759 Dump each function after performing code sinking. The file name is made
4760 by appending @file{.sink} to the source file name.
4763 @opindex fdump-tree-dom
4764 Dump each function after applying dominator tree optimizations. The file
4765 name is made by appending @file{.dom} to the source file name.
4768 @opindex fdump-tree-dse
4769 Dump each function after applying dead store elimination. The file
4770 name is made by appending @file{.dse} to the source file name.
4773 @opindex fdump-tree-phiopt
4774 Dump each function after optimizing PHI nodes into straightline code. The file
4775 name is made by appending @file{.phiopt} to the source file name.
4778 @opindex fdump-tree-forwprop
4779 Dump each function after forward propagating single use variables. The file
4780 name is made by appending @file{.forwprop} to the source file name.
4783 @opindex fdump-tree-copyrename
4784 Dump each function after applying the copy rename optimization. The file
4785 name is made by appending @file{.copyrename} to the source file name.
4788 @opindex fdump-tree-nrv
4789 Dump each function after applying the named return value optimization on
4790 generic trees. The file name is made by appending @file{.nrv} to the source
4794 @opindex fdump-tree-vect
4795 Dump each function after applying vectorization of loops. The file name is
4796 made by appending @file{.vect} to the source file name.
4799 @opindex fdump-tree-vrp
4800 Dump each function after Value Range Propagation (VRP). The file name
4801 is made by appending @file{.vrp} to the source file name.
4804 @opindex fdump-tree-all
4805 Enable all the available tree dumps with the flags provided in this option.
4808 @item -ftree-vectorizer-verbose=@var{n}
4809 @opindex ftree-vectorizer-verbose
4810 This option controls the amount of debugging output the vectorizer prints.
4811 This information is written to standard error, unless
4812 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4813 in which case it is output to the usual dump listing file, @file{.vect}.
4814 For @var{n}=0 no diagnostic information is reported.
4815 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4816 and the total number of loops that got vectorized.
4817 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4818 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4819 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4820 level that @option{-fdump-tree-vect-stats} uses.
4821 Higher verbosity levels mean either more information dumped for each
4822 reported loop, or same amount of information reported for more loops:
4823 If @var{n}=3, alignment related information is added to the reports.
4824 If @var{n}=4, data-references related information (e.g. memory dependences,
4825 memory access-patterns) is added to the reports.
4826 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4827 that did not pass the first analysis phase (i.e. may not be countable, or
4828 may have complicated control-flow).
4829 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4830 For @var{n}=7, all the information the vectorizer generates during its
4831 analysis and transformation is reported. This is the same verbosity level
4832 that @option{-fdump-tree-vect-details} uses.
4834 @item -frandom-seed=@var{string}
4835 @opindex frandom-string
4836 This option provides a seed that GCC uses when it would otherwise use
4837 random numbers. It is used to generate certain symbol names
4838 that have to be different in every compiled file. It is also used to
4839 place unique stamps in coverage data files and the object files that
4840 produce them. You can use the @option{-frandom-seed} option to produce
4841 reproducibly identical object files.
4843 The @var{string} should be different for every file you compile.
4845 @item -fsched-verbose=@var{n}
4846 @opindex fsched-verbose
4847 On targets that use instruction scheduling, this option controls the
4848 amount of debugging output the scheduler prints. This information is
4849 written to standard error, unless @option{-dS} or @option{-dR} is
4850 specified, in which case it is output to the usual dump
4851 listing file, @file{.sched} or @file{.sched2} respectively. However
4852 for @var{n} greater than nine, the output is always printed to standard
4855 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4856 same information as @option{-dRS}. For @var{n} greater than one, it
4857 also output basic block probabilities, detailed ready list information
4858 and unit/insn info. For @var{n} greater than two, it includes RTL
4859 at abort point, control-flow and regions info. And for @var{n} over
4860 four, @option{-fsched-verbose} also includes dependence info.
4864 Store the usual ``temporary'' intermediate files permanently; place them
4865 in the current directory and name them based on the source file. Thus,
4866 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4867 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4868 preprocessed @file{foo.i} output file even though the compiler now
4869 normally uses an integrated preprocessor.
4871 When used in combination with the @option{-x} command line option,
4872 @option{-save-temps} is sensible enough to avoid over writing an
4873 input source file with the same extension as an intermediate file.
4874 The corresponding intermediate file may be obtained by renaming the
4875 source file before using @option{-save-temps}.
4879 Report the CPU time taken by each subprocess in the compilation
4880 sequence. For C source files, this is the compiler proper and assembler
4881 (plus the linker if linking is done). The output looks like this:
4888 The first number on each line is the ``user time'', that is time spent
4889 executing the program itself. The second number is ``system time'',
4890 time spent executing operating system routines on behalf of the program.
4891 Both numbers are in seconds.
4893 @item -fvar-tracking
4894 @opindex fvar-tracking
4895 Run variable tracking pass. It computes where variables are stored at each
4896 position in code. Better debugging information is then generated
4897 (if the debugging information format supports this information).
4899 It is enabled by default when compiling with optimization (@option{-Os},
4900 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4901 the debug info format supports it.
4903 @item -print-file-name=@var{library}
4904 @opindex print-file-name
4905 Print the full absolute name of the library file @var{library} that
4906 would be used when linking---and don't do anything else. With this
4907 option, GCC does not compile or link anything; it just prints the
4910 @item -print-multi-directory
4911 @opindex print-multi-directory
4912 Print the directory name corresponding to the multilib selected by any
4913 other switches present in the command line. This directory is supposed
4914 to exist in @env{GCC_EXEC_PREFIX}.
4916 @item -print-multi-lib
4917 @opindex print-multi-lib
4918 Print the mapping from multilib directory names to compiler switches
4919 that enable them. The directory name is separated from the switches by
4920 @samp{;}, and each switch starts with an @samp{@@} instead of the
4921 @samp{-}, without spaces between multiple switches. This is supposed to
4922 ease shell-processing.
4924 @item -print-prog-name=@var{program}
4925 @opindex print-prog-name
4926 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4928 @item -print-libgcc-file-name
4929 @opindex print-libgcc-file-name
4930 Same as @option{-print-file-name=libgcc.a}.
4932 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4933 but you do want to link with @file{libgcc.a}. You can do
4936 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4939 @item -print-search-dirs
4940 @opindex print-search-dirs
4941 Print the name of the configured installation directory and a list of
4942 program and library directories @command{gcc} will search---and don't do anything else.
4944 This is useful when @command{gcc} prints the error message
4945 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4946 To resolve this you either need to put @file{cpp0} and the other compiler
4947 components where @command{gcc} expects to find them, or you can set the environment
4948 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4949 Don't forget the trailing @samp{/}.
4950 @xref{Environment Variables}.
4952 @item -print-sysroot-headers-suffix
4953 @opindex print-sysroot-headers-suffix
4954 Print the suffix added to the target sysroot when searching for
4955 headers, or give an error if the compiler is not configured with such
4956 a suffix---and don't do anything else.
4959 @opindex dumpmachine
4960 Print the compiler's target machine (for example,
4961 @samp{i686-pc-linux-gnu})---and don't do anything else.
4964 @opindex dumpversion
4965 Print the compiler version (for example, @samp{3.0})---and don't do
4970 Print the compiler's built-in specs---and don't do anything else. (This
4971 is used when GCC itself is being built.) @xref{Spec Files}.
4973 @item -feliminate-unused-debug-types
4974 @opindex feliminate-unused-debug-types
4975 Normally, when producing DWARF2 output, GCC will emit debugging
4976 information for all types declared in a compilation
4977 unit, regardless of whether or not they are actually used
4978 in that compilation unit. Sometimes this is useful, such as
4979 if, in the debugger, you want to cast a value to a type that is
4980 not actually used in your program (but is declared). More often,
4981 however, this results in a significant amount of wasted space.
4982 With this option, GCC will avoid producing debug symbol output
4983 for types that are nowhere used in the source file being compiled.
4986 @node Optimize Options
4987 @section Options That Control Optimization
4988 @cindex optimize options
4989 @cindex options, optimization
4991 These options control various sorts of optimizations.
4993 Without any optimization option, the compiler's goal is to reduce the
4994 cost of compilation and to make debugging produce the expected
4995 results. Statements are independent: if you stop the program with a
4996 breakpoint between statements, you can then assign a new value to any
4997 variable or change the program counter to any other statement in the
4998 function and get exactly the results you would expect from the source
5001 Turning on optimization flags makes the compiler attempt to improve
5002 the performance and/or code size at the expense of compilation time
5003 and possibly the ability to debug the program.
5005 The compiler performs optimization based on the knowledge it has of
5006 the program. Optimization levels @option{-O} and above, in
5007 particular, enable @emph{unit-at-a-time} mode, which allows the
5008 compiler to consider information gained from later functions in
5009 the file when compiling a function. Compiling multiple files at
5010 once to a single output file in @emph{unit-at-a-time} mode allows
5011 the compiler to use information gained from all of the files when
5012 compiling each of them.
5014 Not all optimizations are controlled directly by a flag. Only
5015 optimizations that have a flag are listed.
5022 Optimize. Optimizing compilation takes somewhat more time, and a lot
5023 more memory for a large function.
5025 With @option{-O}, the compiler tries to reduce code size and execution
5026 time, without performing any optimizations that take a great deal of
5029 @option{-O} turns on the following optimization flags:
5030 @gccoptlist{-fdefer-pop @gol
5031 -fdelayed-branch @gol
5032 -fguess-branch-probability @gol
5033 -fcprop-registers @gol
5036 -fif-conversion @gol
5037 -fif-conversion2 @gol
5038 -fsplit-wide-types @gol
5041 -ftree-dominator-opts @gol
5045 -ftree-copyrename @gol
5048 -funit-at-a-time @gol
5049 -finline-small-functions @gol
5052 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5053 where doing so does not interfere with debugging.
5057 Optimize even more. GCC performs nearly all supported optimizations
5058 that do not involve a space-speed tradeoff. The compiler does not
5059 perform loop unrolling or function inlining when you specify @option{-O2}.
5060 As compared to @option{-O}, this option increases both compilation time
5061 and the performance of the generated code.
5063 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5064 also turns on the following optimization flags:
5065 @gccoptlist{-fthread-jumps @gol
5067 -foptimize-sibling-calls @gol
5068 -fcse-follow-jumps -fcse-skip-blocks @gol
5069 -fgcse -fgcse-lm @gol
5070 -fexpensive-optimizations @gol
5071 -frerun-cse-after-loop @gol
5074 -fschedule-insns -fschedule-insns2 @gol
5075 -fsched-interblock -fsched-spec @gol
5077 -fstrict-aliasing -fstrict-overflow @gol
5078 -fdelete-null-pointer-checks @gol
5079 -freorder-blocks -freorder-functions @gol
5080 -falign-functions -falign-jumps @gol
5081 -falign-loops -falign-labels @gol
5085 Please note the warning under @option{-fgcse} about
5086 invoking @option{-O2} on programs that use computed gotos.
5090 Optimize yet more. @option{-O3} turns on all optimizations specified by
5091 @option{-O2} and also turns on the @option{-finline-functions},
5092 @option{-funswitch-loops}, @option{-fpredictive-commoning} and
5093 @option{-fgcse-after-reload} options.
5097 Reduce compilation time and make debugging produce the expected
5098 results. This is the default.
5102 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5103 do not typically increase code size. It also performs further
5104 optimizations designed to reduce code size.
5106 @option{-Os} disables the following optimization flags:
5107 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5108 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5109 -fprefetch-loop-arrays -ftree-vect-loop-version}
5111 If you use multiple @option{-O} options, with or without level numbers,
5112 the last such option is the one that is effective.
5115 Options of the form @option{-f@var{flag}} specify machine-independent
5116 flags. Most flags have both positive and negative forms; the negative
5117 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5118 below, only one of the forms is listed---the one you typically will
5119 use. You can figure out the other form by either removing @samp{no-}
5122 The following options control specific optimizations. They are either
5123 activated by @option{-O} options or are related to ones that are. You
5124 can use the following flags in the rare cases when ``fine-tuning'' of
5125 optimizations to be performed is desired.
5128 @item -fno-default-inline
5129 @opindex fno-default-inline
5130 Do not make member functions inline by default merely because they are
5131 defined inside the class scope (C++ only). Otherwise, when you specify
5132 @w{@option{-O}}, member functions defined inside class scope are compiled
5133 inline by default; i.e., you don't need to add @samp{inline} in front of
5134 the member function name.
5136 @item -fno-defer-pop
5137 @opindex fno-defer-pop
5138 Always pop the arguments to each function call as soon as that function
5139 returns. For machines which must pop arguments after a function call,
5140 the compiler normally lets arguments accumulate on the stack for several
5141 function calls and pops them all at once.
5143 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5146 @opindex fforce-addr
5147 Force memory address constants to be copied into registers before
5148 doing arithmetic on them.
5150 @item -fforward-propagate
5151 @opindex fforward-propagate
5152 Perform a forward propagation pass on RTL. The pass tries to combine two
5153 instructions and checks if the result can be simplified. If loop unrolling
5154 is active, two passes are performed and the second is scheduled after
5157 This option is enabled by default at optimization levels @option{-O2},
5158 @option{-O3}, @option{-Os}.
5160 @item -fomit-frame-pointer
5161 @opindex fomit-frame-pointer
5162 Don't keep the frame pointer in a register for functions that
5163 don't need one. This avoids the instructions to save, set up and
5164 restore frame pointers; it also makes an extra register available
5165 in many functions. @strong{It also makes debugging impossible on
5168 On some machines, such as the VAX, this flag has no effect, because
5169 the standard calling sequence automatically handles the frame pointer
5170 and nothing is saved by pretending it doesn't exist. The
5171 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5172 whether a target machine supports this flag. @xref{Registers,,Register
5173 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5175 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5177 @item -foptimize-sibling-calls
5178 @opindex foptimize-sibling-calls
5179 Optimize sibling and tail recursive calls.
5181 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5185 Don't pay attention to the @code{inline} keyword. Normally this option
5186 is used to keep the compiler from expanding any functions inline.
5187 Note that if you are not optimizing, no functions can be expanded inline.
5189 @item -finline-small-functions
5190 @opindex finline-small-functions
5191 Integrate functions into their callers when their body is smaller than expected
5192 function call code (so overall size of program gets smaller). The compiler
5193 heuristically decides which functions are simple enough to be worth integrating
5196 Enabled at level @option{-O2}.
5198 @item -finline-functions
5199 @opindex finline-functions
5200 Integrate all simple functions into their callers. The compiler
5201 heuristically decides which functions are simple enough to be worth
5202 integrating in this way.
5204 If all calls to a given function are integrated, and the function is
5205 declared @code{static}, then the function is normally not output as
5206 assembler code in its own right.
5208 Enabled at level @option{-O3}.
5210 @item -finline-functions-called-once
5211 @opindex finline-functions-called-once
5212 Consider all @code{static} functions called once for inlining into their
5213 caller even if they are not marked @code{inline}. If a call to a given
5214 function is integrated, then the function is not output as assembler code
5217 Enabled if @option{-funit-at-a-time} is enabled.
5219 @item -fearly-inlining
5220 @opindex fearly-inlining
5221 Inline functions marked by @code{always_inline} and functions whose body seems
5222 smaller than the function call overhead early before doing
5223 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5224 makes profiling significantly cheaper and usually inlining faster on programs
5225 having large chains of nested wrapper functions.
5229 @item -finline-limit=@var{n}
5230 @opindex finline-limit
5231 By default, GCC limits the size of functions that can be inlined. This flag
5232 allows the control of this limit for functions that are explicitly marked as
5233 inline (i.e., marked with the inline keyword or defined within the class
5234 definition in c++). @var{n} is the size of functions that can be inlined in
5235 number of pseudo instructions (not counting parameter handling). The default
5236 value of @var{n} is 600.
5237 Increasing this value can result in more inlined code at
5238 the cost of compilation time and memory consumption. Decreasing usually makes
5239 the compilation faster and less code will be inlined (which presumably
5240 means slower programs). This option is particularly useful for programs that
5241 use inlining heavily such as those based on recursive templates with C++.
5243 Inlining is actually controlled by a number of parameters, which may be
5244 specified individually by using @option{--param @var{name}=@var{value}}.
5245 The @option{-finline-limit=@var{n}} option sets some of these parameters
5249 @item max-inline-insns-single
5250 is set to @var{n}/2.
5251 @item max-inline-insns-auto
5252 is set to @var{n}/2.
5253 @item min-inline-insns
5254 is set to 130 or @var{n}/4, whichever is smaller.
5255 @item max-inline-insns-rtl
5259 See below for a documentation of the individual
5260 parameters controlling inlining.
5262 @emph{Note:} pseudo instruction represents, in this particular context, an
5263 abstract measurement of function's size. In no way does it represent a count
5264 of assembly instructions and as such its exact meaning might change from one
5265 release to an another.
5267 @item -fkeep-inline-functions
5268 @opindex fkeep-inline-functions
5269 In C, emit @code{static} functions that are declared @code{inline}
5270 into the object file, even if the function has been inlined into all
5271 of its callers. This switch does not affect functions using the
5272 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5273 inline functions into the object file.
5275 @item -fkeep-static-consts
5276 @opindex fkeep-static-consts
5277 Emit variables declared @code{static const} when optimization isn't turned
5278 on, even if the variables aren't referenced.
5280 GCC enables this option by default. If you want to force the compiler to
5281 check if the variable was referenced, regardless of whether or not
5282 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5284 @item -fmerge-constants
5285 @opindex fmerge-constants
5286 Attempt to merge identical constants (string constants and floating point
5287 constants) across compilation units.
5289 This option is the default for optimized compilation if the assembler and
5290 linker support it. Use @option{-fno-merge-constants} to inhibit this
5293 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5295 @item -fmerge-all-constants
5296 @opindex fmerge-all-constants
5297 Attempt to merge identical constants and identical variables.
5299 This option implies @option{-fmerge-constants}. In addition to
5300 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5301 arrays or initialized constant variables with integral or floating point
5302 types. Languages like C or C++ require each non-automatic variable to
5303 have distinct location, so using this option will result in non-conforming
5306 @item -fmodulo-sched
5307 @opindex fmodulo-sched
5308 Perform swing modulo scheduling immediately before the first scheduling
5309 pass. This pass looks at innermost loops and reorders their
5310 instructions by overlapping different iterations.
5312 @item -fmodulo-sched-allow-regmoves
5313 @opindex fmodulo-sched-allow-regmoves
5314 Perform more aggressive SMS based modulo scheduling with register moves
5315 allowed. By setting this flag certain anti-dependences edges will be
5316 deleted which will trigger the generation of reg-moves based on the
5317 life-range analysis. This option is effective only with
5318 @option{-fmodulo-sched} enabled.
5320 @item -fno-branch-count-reg
5321 @opindex fno-branch-count-reg
5322 Do not use ``decrement and branch'' instructions on a count register,
5323 but instead generate a sequence of instructions that decrement a
5324 register, compare it against zero, then branch based upon the result.
5325 This option is only meaningful on architectures that support such
5326 instructions, which include x86, PowerPC, IA-64 and S/390.
5328 The default is @option{-fbranch-count-reg}.
5330 @item -fno-function-cse
5331 @opindex fno-function-cse
5332 Do not put function addresses in registers; make each instruction that
5333 calls a constant function contain the function's address explicitly.
5335 This option results in less efficient code, but some strange hacks
5336 that alter the assembler output may be confused by the optimizations
5337 performed when this option is not used.
5339 The default is @option{-ffunction-cse}
5341 @item -fno-zero-initialized-in-bss
5342 @opindex fno-zero-initialized-in-bss
5343 If the target supports a BSS section, GCC by default puts variables that
5344 are initialized to zero into BSS@. This can save space in the resulting
5347 This option turns off this behavior because some programs explicitly
5348 rely on variables going to the data section. E.g., so that the
5349 resulting executable can find the beginning of that section and/or make
5350 assumptions based on that.
5352 The default is @option{-fzero-initialized-in-bss}.
5354 @item -fmudflap -fmudflapth -fmudflapir
5358 @cindex bounds checking
5360 For front-ends that support it (C and C++), instrument all risky
5361 pointer/array dereferencing operations, some standard library
5362 string/heap functions, and some other associated constructs with
5363 range/validity tests. Modules so instrumented should be immune to
5364 buffer overflows, invalid heap use, and some other classes of C/C++
5365 programming errors. The instrumentation relies on a separate runtime
5366 library (@file{libmudflap}), which will be linked into a program if
5367 @option{-fmudflap} is given at link time. Run-time behavior of the
5368 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5369 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5372 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5373 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5374 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5375 instrumentation should ignore pointer reads. This produces less
5376 instrumentation (and therefore faster execution) and still provides
5377 some protection against outright memory corrupting writes, but allows
5378 erroneously read data to propagate within a program.
5380 @item -fthread-jumps
5381 @opindex fthread-jumps
5382 Perform optimizations where we check to see if a jump branches to a
5383 location where another comparison subsumed by the first is found. If
5384 so, the first branch is redirected to either the destination of the
5385 second branch or a point immediately following it, depending on whether
5386 the condition is known to be true or false.
5388 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5390 @item -fsplit-wide-types
5391 @opindex fsplit-wide-types
5392 When using a type that occupies multiple registers, such as @code{long
5393 long} on a 32-bit system, split the registers apart and allocate them
5394 independently. This normally generates better code for those types,
5395 but may make debugging more difficult.
5397 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5400 @item -fcse-follow-jumps
5401 @opindex fcse-follow-jumps
5402 In common subexpression elimination (CSE), scan through jump instructions
5403 when the target of the jump is not reached by any other path. For
5404 example, when CSE encounters an @code{if} statement with an
5405 @code{else} clause, CSE will follow the jump when the condition
5408 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5410 @item -fcse-skip-blocks
5411 @opindex fcse-skip-blocks
5412 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5413 follow jumps which conditionally skip over blocks. When CSE
5414 encounters a simple @code{if} statement with no else clause,
5415 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5416 body of the @code{if}.
5418 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5420 @item -frerun-cse-after-loop
5421 @opindex frerun-cse-after-loop
5422 Re-run common subexpression elimination after loop optimizations has been
5425 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5429 Perform a global common subexpression elimination pass.
5430 This pass also performs global constant and copy propagation.
5432 @emph{Note:} When compiling a program using computed gotos, a GCC
5433 extension, you may get better runtime performance if you disable
5434 the global common subexpression elimination pass by adding
5435 @option{-fno-gcse} to the command line.
5437 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5441 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5442 attempt to move loads which are only killed by stores into themselves. This
5443 allows a loop containing a load/store sequence to be changed to a load outside
5444 the loop, and a copy/store within the loop.
5446 Enabled by default when gcse is enabled.
5450 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5451 global common subexpression elimination. This pass will attempt to move
5452 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5453 loops containing a load/store sequence can be changed to a load before
5454 the loop and a store after the loop.
5456 Not enabled at any optimization level.
5460 When @option{-fgcse-las} is enabled, the global common subexpression
5461 elimination pass eliminates redundant loads that come after stores to the
5462 same memory location (both partial and full redundancies).
5464 Not enabled at any optimization level.
5466 @item -fgcse-after-reload
5467 @opindex fgcse-after-reload
5468 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5469 pass is performed after reload. The purpose of this pass is to cleanup
5472 @item -funsafe-loop-optimizations
5473 @opindex funsafe-loop-optimizations
5474 If given, the loop optimizer will assume that loop indices do not
5475 overflow, and that the loops with nontrivial exit condition are not
5476 infinite. This enables a wider range of loop optimizations even if
5477 the loop optimizer itself cannot prove that these assumptions are valid.
5478 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5479 if it finds this kind of loop.
5481 @item -fcrossjumping
5482 @opindex fcrossjumping
5483 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5484 resulting code may or may not perform better than without cross-jumping.
5486 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5490 Perform dead code elimination (DCE) on RTL.
5491 Enabled by default at @option{-O} and higher.
5495 Perform dead store elimination (DSE) on RTL.
5496 Enabled by default at @option{-O} and higher.
5498 @item -fif-conversion
5499 @opindex fif-conversion
5500 Attempt to transform conditional jumps into branch-less equivalents. This
5501 include use of conditional moves, min, max, set flags and abs instructions, and
5502 some tricks doable by standard arithmetics. The use of conditional execution
5503 on chips where it is available is controlled by @code{if-conversion2}.
5505 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5507 @item -fif-conversion2
5508 @opindex fif-conversion2
5509 Use conditional execution (where available) to transform conditional jumps into
5510 branch-less equivalents.
5512 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5514 @item -fdelete-null-pointer-checks
5515 @opindex fdelete-null-pointer-checks
5516 Use global dataflow analysis to identify and eliminate useless checks
5517 for null pointers. The compiler assumes that dereferencing a null
5518 pointer would have halted the program. If a pointer is checked after
5519 it has already been dereferenced, it cannot be null.
5521 In some environments, this assumption is not true, and programs can
5522 safely dereference null pointers. Use
5523 @option{-fno-delete-null-pointer-checks} to disable this optimization
5524 for programs which depend on that behavior.
5526 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5528 @item -fexpensive-optimizations
5529 @opindex fexpensive-optimizations
5530 Perform a number of minor optimizations that are relatively expensive.
5532 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5534 @item -foptimize-register-move
5536 @opindex foptimize-register-move
5538 Attempt to reassign register numbers in move instructions and as
5539 operands of other simple instructions in order to maximize the amount of
5540 register tying. This is especially helpful on machines with two-operand
5543 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5546 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5548 @item -fdelayed-branch
5549 @opindex fdelayed-branch
5550 If supported for the target machine, attempt to reorder instructions
5551 to exploit instruction slots available after delayed branch
5554 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5556 @item -fschedule-insns
5557 @opindex fschedule-insns
5558 If supported for the target machine, attempt to reorder instructions to
5559 eliminate execution stalls due to required data being unavailable. This
5560 helps machines that have slow floating point or memory load instructions
5561 by allowing other instructions to be issued until the result of the load
5562 or floating point instruction is required.
5564 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5566 @item -fschedule-insns2
5567 @opindex fschedule-insns2
5568 Similar to @option{-fschedule-insns}, but requests an additional pass of
5569 instruction scheduling after register allocation has been done. This is
5570 especially useful on machines with a relatively small number of
5571 registers and where memory load instructions take more than one cycle.
5573 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5575 @item -fno-sched-interblock
5576 @opindex fno-sched-interblock
5577 Don't schedule instructions across basic blocks. This is normally
5578 enabled by default when scheduling before register allocation, i.e.@:
5579 with @option{-fschedule-insns} or at @option{-O2} or higher.
5581 @item -fno-sched-spec
5582 @opindex fno-sched-spec
5583 Don't allow speculative motion of non-load instructions. This is normally
5584 enabled by default when scheduling before register allocation, i.e.@:
5585 with @option{-fschedule-insns} or at @option{-O2} or higher.
5587 @item -fsched-spec-load
5588 @opindex fsched-spec-load
5589 Allow speculative motion of some load instructions. This only makes
5590 sense when scheduling before register allocation, i.e.@: with
5591 @option{-fschedule-insns} or at @option{-O2} or higher.
5593 @item -fsched-spec-load-dangerous
5594 @opindex fsched-spec-load-dangerous
5595 Allow speculative motion of more load instructions. This only makes
5596 sense when scheduling before register allocation, i.e.@: with
5597 @option{-fschedule-insns} or at @option{-O2} or higher.
5599 @item -fsched-stalled-insns
5600 @itemx -fsched-stalled-insns=@var{n}
5601 @opindex fsched-stalled-insns
5602 Define how many insns (if any) can be moved prematurely from the queue
5603 of stalled insns into the ready list, during the second scheduling pass.
5604 @option{-fno-fsched-stalled-insns} and @option{-fsched-stalled-insns=0}
5605 are equivalent and mean that no insns will be moved prematurely.
5606 If @var{n} is unspecified then there is no limit on how many queued
5607 insns can be moved prematurely.
5609 @item -fsched-stalled-insns-dep
5610 @itemx -fsched-stalled-insns-dep=@var{n}
5611 @opindex fsched-stalled-insns-dep
5612 Define how many insn groups (cycles) will be examined for a dependency
5613 on a stalled insn that is candidate for premature removal from the queue
5614 of stalled insns. This has an effect only during the second scheduling pass,
5615 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
5616 @option{-fno-sched-stalled-insns-dep} is equivalent to
5617 @option{-fsched-stalled-insns-dep=0}.
5618 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5619 @option{-fsched-stalled-insns-dep=1}.
5621 @item -fsched2-use-superblocks
5622 @opindex fsched2-use-superblocks
5623 When scheduling after register allocation, do use superblock scheduling
5624 algorithm. Superblock scheduling allows motion across basic block boundaries
5625 resulting on faster schedules. This option is experimental, as not all machine
5626 descriptions used by GCC model the CPU closely enough to avoid unreliable
5627 results from the algorithm.
5629 This only makes sense when scheduling after register allocation, i.e.@: with
5630 @option{-fschedule-insns2} or at @option{-O2} or higher.
5632 @item -fsched2-use-traces
5633 @opindex fsched2-use-traces
5634 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5635 allocation and additionally perform code duplication in order to increase the
5636 size of superblocks using tracer pass. See @option{-ftracer} for details on
5639 This mode should produce faster but significantly longer programs. Also
5640 without @option{-fbranch-probabilities} the traces constructed may not
5641 match the reality and hurt the performance. This only makes
5642 sense when scheduling after register allocation, i.e.@: with
5643 @option{-fschedule-insns2} or at @option{-O2} or higher.
5647 Eliminate redundant sign extension instructions and move the non-redundant
5648 ones to optimal placement using lazy code motion (LCM).
5650 @item -freschedule-modulo-scheduled-loops
5651 @opindex freschedule-modulo-scheduled-loops
5652 The modulo scheduling comes before the traditional scheduling, if a loop
5653 was modulo scheduled we may want to prevent the later scheduling passes
5654 from changing its schedule, we use this option to control that.
5656 @item -fcaller-saves
5657 @opindex fcaller-saves
5658 Enable values to be allocated in registers that will be clobbered by
5659 function calls, by emitting extra instructions to save and restore the
5660 registers around such calls. Such allocation is done only when it
5661 seems to result in better code than would otherwise be produced.
5663 This option is always enabled by default on certain machines, usually
5664 those which have no call-preserved registers to use instead.
5666 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5668 @item -ftree-reassoc
5669 @opindex ftree-reassoc
5670 Perform reassociation on trees. This flag is enabled by default
5671 at @option{-O} and higher.
5675 Perform partial redundancy elimination (PRE) on trees. This flag is
5676 enabled by default at @option{-O2} and @option{-O3}.
5680 Perform full redundancy elimination (FRE) on trees. The difference
5681 between FRE and PRE is that FRE only considers expressions
5682 that are computed on all paths leading to the redundant computation.
5683 This analysis is faster than PRE, though it exposes fewer redundancies.
5684 This flag is enabled by default at @option{-O} and higher.
5686 @item -ftree-copy-prop
5687 @opindex ftree-copy-prop
5688 Perform copy propagation on trees. This pass eliminates unnecessary
5689 copy operations. This flag is enabled by default at @option{-O} and
5693 @opindex ftree-salias
5694 Perform structural alias analysis on trees. This flag
5695 is enabled by default at @option{-O} and higher.
5699 Perform interprocedural pointer analysis.
5703 Perform forward store motion on trees. This flag is
5704 enabled by default at @option{-O} and higher.
5708 Perform sparse conditional constant propagation (CCP) on trees. This
5709 pass only operates on local scalar variables and is enabled by default
5710 at @option{-O} and higher.
5712 @item -ftree-store-ccp
5713 @opindex ftree-store-ccp
5714 Perform sparse conditional constant propagation (CCP) on trees. This
5715 pass operates on both local scalar variables and memory stores and
5716 loads (global variables, structures, arrays, etc). This flag is
5717 enabled by default at @option{-O2} and higher.
5721 Perform dead code elimination (DCE) on trees. This flag is enabled by
5722 default at @option{-O} and higher.
5724 @item -ftree-dominator-opts
5725 @opindex ftree-dominator-opts
5726 Perform a variety of simple scalar cleanups (constant/copy
5727 propagation, redundancy elimination, range propagation and expression
5728 simplification) based on a dominator tree traversal. This also
5729 performs jump threading (to reduce jumps to jumps). This flag is
5730 enabled by default at @option{-O} and higher.
5734 Perform dead store elimination (DSE) on trees. A dead store is a store into
5735 a memory location which will later be overwritten by another store without
5736 any intervening loads. In this case the earlier store can be deleted. This
5737 flag is enabled by default at @option{-O} and higher.
5741 Perform loop header copying on trees. This is beneficial since it increases
5742 effectiveness of code motion optimizations. It also saves one jump. This flag
5743 is enabled by default at @option{-O} and higher. It is not enabled
5744 for @option{-Os}, since it usually increases code size.
5746 @item -ftree-loop-optimize
5747 @opindex ftree-loop-optimize
5748 Perform loop optimizations on trees. This flag is enabled by default
5749 at @option{-O} and higher.
5751 @item -ftree-loop-linear
5752 @opindex ftree-loop-linear
5753 Perform linear loop transformations on tree. This flag can improve cache
5754 performance and allow further loop optimizations to take place.
5756 @item -fcheck-data-deps
5757 @opindex fcheck-data-deps
5758 Compare the results of several data dependence analyzers. This option
5759 is used for debugging the data dependence analyzers.
5761 @item -ftree-loop-im
5762 @opindex ftree-loop-im
5763 Perform loop invariant motion on trees. This pass moves only invariants that
5764 would be hard to handle at RTL level (function calls, operations that expand to
5765 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5766 operands of conditions that are invariant out of the loop, so that we can use
5767 just trivial invariantness analysis in loop unswitching. The pass also includes
5770 @item -ftree-loop-ivcanon
5771 @opindex ftree-loop-ivcanon
5772 Create a canonical counter for number of iterations in the loop for that
5773 determining number of iterations requires complicated analysis. Later
5774 optimizations then may determine the number easily. Useful especially
5775 in connection with unrolling.
5779 Perform induction variable optimizations (strength reduction, induction
5780 variable merging and induction variable elimination) on trees.
5782 @item -ftree-parallelize-loops=n
5783 @opindex ftree-parallelize-loops
5784 Parallelize loops, i.e., split their iteration space to run in n threads.
5785 This is only possible for loops whose iterations are independent
5786 and can be arbitrarily reordered. The optimization is only
5787 profitable on multiprocessor machines, for loops that are CPU-intensive,
5788 rather than constrained e.g. by memory bandwidth.
5792 Perform scalar replacement of aggregates. This pass replaces structure
5793 references with scalars to prevent committing structures to memory too
5794 early. This flag is enabled by default at @option{-O} and higher.
5796 @item -ftree-copyrename
5797 @opindex ftree-copyrename
5798 Perform copy renaming on trees. This pass attempts to rename compiler
5799 temporaries to other variables at copy locations, usually resulting in
5800 variable names which more closely resemble the original variables. This flag
5801 is enabled by default at @option{-O} and higher.
5805 Perform temporary expression replacement during the SSA->normal phase. Single
5806 use/single def temporaries are replaced at their use location with their
5807 defining expression. This results in non-GIMPLE code, but gives the expanders
5808 much more complex trees to work on resulting in better RTL generation. This is
5809 enabled by default at @option{-O} and higher.
5811 @item -ftree-vectorize
5812 @opindex ftree-vectorize
5813 Perform loop vectorization on trees.
5815 @item -ftree-vect-loop-version
5816 @opindex ftree-vect-loop-version
5817 Perform loop versioning when doing loop vectorization on trees. When a loop
5818 appears to be vectorizable except that data alignment or data dependence cannot
5819 be determined at compile time then vectorized and non-vectorized versions of
5820 the loop are generated along with runtime checks for alignment or dependence
5821 to control which version is executed. This option is enabled by default
5822 except at level @option{-Os} where it is disabled.
5824 @item -fvect-cost-model
5825 @opindex fvect-cost-model
5826 Enable cost model for vectorization.
5830 Perform Value Range Propagation on trees. This is similar to the
5831 constant propagation pass, but instead of values, ranges of values are
5832 propagated. This allows the optimizers to remove unnecessary range
5833 checks like array bound checks and null pointer checks. This is
5834 enabled by default at @option{-O2} and higher. Null pointer check
5835 elimination is only done if @option{-fdelete-null-pointer-checks} is
5840 Perform tail duplication to enlarge superblock size. This transformation
5841 simplifies the control flow of the function allowing other optimizations to do
5844 @item -funroll-loops
5845 @opindex funroll-loops
5846 Unroll loops whose number of iterations can be determined at compile
5847 time or upon entry to the loop. @option{-funroll-loops} implies
5848 @option{-frerun-cse-after-loop}. This option makes code larger,
5849 and may or may not make it run faster.
5851 @item -funroll-all-loops
5852 @opindex funroll-all-loops
5853 Unroll all loops, even if their number of iterations is uncertain when
5854 the loop is entered. This usually makes programs run more slowly.
5855 @option{-funroll-all-loops} implies the same options as
5856 @option{-funroll-loops},
5858 @item -fsplit-ivs-in-unroller
5859 @opindex fsplit-ivs-in-unroller
5860 Enables expressing of values of induction variables in later iterations
5861 of the unrolled loop using the value in the first iteration. This breaks
5862 long dependency chains, thus improving efficiency of the scheduling passes.
5864 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5865 same effect. However in cases the loop body is more complicated than
5866 a single basic block, this is not reliable. It also does not work at all
5867 on some of the architectures due to restrictions in the CSE pass.
5869 This optimization is enabled by default.
5871 @item -fvariable-expansion-in-unroller
5872 @opindex fvariable-expansion-in-unroller
5873 With this option, the compiler will create multiple copies of some
5874 local variables when unrolling a loop which can result in superior code.
5876 @item -fpredictive-commoning
5877 @opindex fpredictive-commoning
5878 Perform predictive commoning optimization, i.e., reusing computations
5879 (especially memory loads and stores) performed in previous
5880 iterations of loops.
5882 This option is enabled at level @option{-O3}.
5884 @item -fprefetch-loop-arrays
5885 @opindex fprefetch-loop-arrays
5886 If supported by the target machine, generate instructions to prefetch
5887 memory to improve the performance of loops that access large arrays.
5889 This option may generate better or worse code; results are highly
5890 dependent on the structure of loops within the source code.
5892 Disabled at level @option{-Os}.
5895 @itemx -fno-peephole2
5896 @opindex fno-peephole
5897 @opindex fno-peephole2
5898 Disable any machine-specific peephole optimizations. The difference
5899 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5900 are implemented in the compiler; some targets use one, some use the
5901 other, a few use both.
5903 @option{-fpeephole} is enabled by default.
5904 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5906 @item -fno-guess-branch-probability
5907 @opindex fno-guess-branch-probability
5908 Do not guess branch probabilities using heuristics.
5910 GCC will use heuristics to guess branch probabilities if they are
5911 not provided by profiling feedback (@option{-fprofile-arcs}). These
5912 heuristics are based on the control flow graph. If some branch probabilities
5913 are specified by @samp{__builtin_expect}, then the heuristics will be
5914 used to guess branch probabilities for the rest of the control flow graph,
5915 taking the @samp{__builtin_expect} info into account. The interactions
5916 between the heuristics and @samp{__builtin_expect} can be complex, and in
5917 some cases, it may be useful to disable the heuristics so that the effects
5918 of @samp{__builtin_expect} are easier to understand.
5920 The default is @option{-fguess-branch-probability} at levels
5921 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5923 @item -freorder-blocks
5924 @opindex freorder-blocks
5925 Reorder basic blocks in the compiled function in order to reduce number of
5926 taken branches and improve code locality.
5928 Enabled at levels @option{-O2}, @option{-O3}.
5930 @item -freorder-blocks-and-partition
5931 @opindex freorder-blocks-and-partition
5932 In addition to reordering basic blocks in the compiled function, in order
5933 to reduce number of taken branches, partitions hot and cold basic blocks
5934 into separate sections of the assembly and .o files, to improve
5935 paging and cache locality performance.
5937 This optimization is automatically turned off in the presence of
5938 exception handling, for linkonce sections, for functions with a user-defined
5939 section attribute and on any architecture that does not support named
5942 @item -freorder-functions
5943 @opindex freorder-functions
5944 Reorder functions in the object file in order to
5945 improve code locality. This is implemented by using special
5946 subsections @code{.text.hot} for most frequently executed functions and
5947 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5948 the linker so object file format must support named sections and linker must
5949 place them in a reasonable way.
5951 Also profile feedback must be available in to make this option effective. See
5952 @option{-fprofile-arcs} for details.
5954 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5956 @item -fstrict-aliasing
5957 @opindex fstrict-aliasing
5958 Allows the compiler to assume the strictest aliasing rules applicable to
5959 the language being compiled. For C (and C++), this activates
5960 optimizations based on the type of expressions. In particular, an
5961 object of one type is assumed never to reside at the same address as an
5962 object of a different type, unless the types are almost the same. For
5963 example, an @code{unsigned int} can alias an @code{int}, but not a
5964 @code{void*} or a @code{double}. A character type may alias any other
5967 Pay special attention to code like this:
5980 The practice of reading from a different union member than the one most
5981 recently written to (called ``type-punning'') is common. Even with
5982 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5983 is accessed through the union type. So, the code above will work as
5984 expected. However, this code might not:
5995 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5997 @item -fstrict-overflow
5998 @opindex fstrict-overflow
5999 Allow the compiler to assume strict signed overflow rules, depending
6000 on the language being compiled. For C (and C++) this means that
6001 overflow when doing arithmetic with signed numbers is undefined, which
6002 means that the compiler may assume that it will not happen. This
6003 permits various optimizations. For example, the compiler will assume
6004 that an expression like @code{i + 10 > i} will always be true for
6005 signed @code{i}. This assumption is only valid if signed overflow is
6006 undefined, as the expression is false if @code{i + 10} overflows when
6007 using twos complement arithmetic. When this option is in effect any
6008 attempt to determine whether an operation on signed numbers will
6009 overflow must be written carefully to not actually involve overflow.
6011 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6012 that signed overflow is fully defined: it wraps. When
6013 @option{-fwrapv} is used, there is no difference between
6014 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
6015 @option{-fwrapv} certain types of overflow are permitted. For
6016 example, if the compiler gets an overflow when doing arithmetic on
6017 constants, the overflowed value can still be used with
6018 @option{-fwrapv}, but not otherwise.
6020 The @option{-fstrict-overflow} option is enabled at levels
6021 @option{-O2}, @option{-O3}, @option{-Os}.
6023 @item -falign-functions
6024 @itemx -falign-functions=@var{n}
6025 @opindex falign-functions
6026 Align the start of functions to the next power-of-two greater than
6027 @var{n}, skipping up to @var{n} bytes. For instance,
6028 @option{-falign-functions=32} aligns functions to the next 32-byte
6029 boundary, but @option{-falign-functions=24} would align to the next
6030 32-byte boundary only if this can be done by skipping 23 bytes or less.
6032 @option{-fno-align-functions} and @option{-falign-functions=1} are
6033 equivalent and mean that functions will not be aligned.
6035 Some assemblers only support this flag when @var{n} is a power of two;
6036 in that case, it is rounded up.
6038 If @var{n} is not specified or is zero, use a machine-dependent default.
6040 Enabled at levels @option{-O2}, @option{-O3}.
6042 @item -falign-labels
6043 @itemx -falign-labels=@var{n}
6044 @opindex falign-labels
6045 Align all branch targets to a power-of-two boundary, skipping up to
6046 @var{n} bytes like @option{-falign-functions}. This option can easily
6047 make code slower, because it must insert dummy operations for when the
6048 branch target is reached in the usual flow of the code.
6050 @option{-fno-align-labels} and @option{-falign-labels=1} are
6051 equivalent and mean that labels will not be aligned.
6053 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6054 are greater than this value, then their values are used instead.
6056 If @var{n} is not specified or is zero, use a machine-dependent default
6057 which is very likely to be @samp{1}, meaning no alignment.
6059 Enabled at levels @option{-O2}, @option{-O3}.
6062 @itemx -falign-loops=@var{n}
6063 @opindex falign-loops
6064 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6065 like @option{-falign-functions}. The hope is that the loop will be
6066 executed many times, which will make up for any execution of the dummy
6069 @option{-fno-align-loops} and @option{-falign-loops=1} are
6070 equivalent and mean that loops will not be aligned.
6072 If @var{n} is not specified or is zero, use a machine-dependent default.
6074 Enabled at levels @option{-O2}, @option{-O3}.
6077 @itemx -falign-jumps=@var{n}
6078 @opindex falign-jumps
6079 Align branch targets to a power-of-two boundary, for branch targets
6080 where the targets can only be reached by jumping, skipping up to @var{n}
6081 bytes like @option{-falign-functions}. In this case, no dummy operations
6084 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6085 equivalent and mean that loops will not be aligned.
6087 If @var{n} is not specified or is zero, use a machine-dependent default.
6089 Enabled at levels @option{-O2}, @option{-O3}.
6091 @item -funit-at-a-time
6092 @opindex funit-at-a-time
6093 Parse the whole compilation unit before starting to produce code.
6094 This allows some extra optimizations to take place but consumes
6095 more memory (in general). There are some compatibility issues
6096 with @emph{unit-at-a-time} mode:
6099 enabling @emph{unit-at-a-time} mode may change the order
6100 in which functions, variables, and top-level @code{asm} statements
6101 are emitted, and will likely break code relying on some particular
6102 ordering. The majority of such top-level @code{asm} statements,
6103 though, can be replaced by @code{section} attributes. The
6104 @option{fno-toplevel-reorder} option may be used to keep the ordering
6105 used in the input file, at the cost of some optimizations.
6108 @emph{unit-at-a-time} mode removes unreferenced static variables
6109 and functions. This may result in undefined references
6110 when an @code{asm} statement refers directly to variables or functions
6111 that are otherwise unused. In that case either the variable/function
6112 shall be listed as an operand of the @code{asm} statement operand or,
6113 in the case of top-level @code{asm} statements the attribute @code{used}
6114 shall be used on the declaration.
6117 Static functions now can use non-standard passing conventions that
6118 may break @code{asm} statements calling functions directly. Again,
6119 attribute @code{used} will prevent this behavior.
6122 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
6123 but this scheme may not be supported by future releases of GCC@.
6125 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6127 @item -fno-toplevel-reorder
6128 Do not reorder top-level functions, variables, and @code{asm}
6129 statements. Output them in the same order that they appear in the
6130 input file. When this option is used, unreferenced static variables
6131 will not be removed. This option is intended to support existing code
6132 which relies on a particular ordering. For new code, it is better to
6137 Constructs webs as commonly used for register allocation purposes and assign
6138 each web individual pseudo register. This allows the register allocation pass
6139 to operate on pseudos directly, but also strengthens several other optimization
6140 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6141 however, make debugging impossible, since variables will no longer stay in a
6144 Enabled by default with @option{-funroll-loops}.
6146 @item -fwhole-program
6147 @opindex fwhole-program
6148 Assume that the current compilation unit represents whole program being
6149 compiled. All public functions and variables with the exception of @code{main}
6150 and those merged by attribute @code{externally_visible} become static functions
6151 and in a affect gets more aggressively optimized by interprocedural optimizers.
6152 While this option is equivalent to proper use of @code{static} keyword for
6153 programs consisting of single file, in combination with option
6154 @option{--combine} this flag can be used to compile most of smaller scale C
6155 programs since the functions and variables become local for the whole combined
6156 compilation unit, not for the single source file itself.
6158 This option is not supported for Fortran programs.
6160 @item -fcprop-registers
6161 @opindex fcprop-registers
6162 After register allocation and post-register allocation instruction splitting,
6163 we perform a copy-propagation pass to try to reduce scheduling dependencies
6164 and occasionally eliminate the copy.
6166 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6168 @item -fprofile-generate
6169 @opindex fprofile-generate
6171 Enable options usually used for instrumenting application to produce
6172 profile useful for later recompilation with profile feedback based
6173 optimization. You must use @option{-fprofile-generate} both when
6174 compiling and when linking your program.
6176 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6179 @opindex fprofile-use
6180 Enable profile feedback directed optimizations, and optimizations
6181 generally profitable only with profile feedback available.
6183 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6184 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6186 By default, GCC emits an error message if the feedback profiles do not
6187 match the source code. This error can be turned into a warning by using
6188 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6192 The following options control compiler behavior regarding floating
6193 point arithmetic. These options trade off between speed and
6194 correctness. All must be specifically enabled.
6198 @opindex ffloat-store
6199 Do not store floating point variables in registers, and inhibit other
6200 options that might change whether a floating point value is taken from a
6203 @cindex floating point precision
6204 This option prevents undesirable excess precision on machines such as
6205 the 68000 where the floating registers (of the 68881) keep more
6206 precision than a @code{double} is supposed to have. Similarly for the
6207 x86 architecture. For most programs, the excess precision does only
6208 good, but a few programs rely on the precise definition of IEEE floating
6209 point. Use @option{-ffloat-store} for such programs, after modifying
6210 them to store all pertinent intermediate computations into variables.
6214 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
6215 @option{-fno-trapping-math}, @option{-ffinite-math-only},
6216 @option{-fno-rounding-math}, @option{-fno-signaling-nans},
6217 @option{-fno-signed-zeros} and @option{fcx-limited-range}.
6219 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6221 This option is not turned on by any @option{-O} option since
6222 it can result in incorrect output for programs which depend on
6223 an exact implementation of IEEE or ISO rules/specifications for
6224 math functions. It may, however, yield faster code for programs
6225 that do not require the guarantees of these specifications.
6227 @item -fno-math-errno
6228 @opindex fno-math-errno
6229 Do not set ERRNO after calling math functions that are executed
6230 with a single instruction, e.g., sqrt. A program that relies on
6231 IEEE exceptions for math error handling may want to use this flag
6232 for speed while maintaining IEEE arithmetic compatibility.
6234 This option is not turned on by any @option{-O} option since
6235 it can result in incorrect output for programs which depend on
6236 an exact implementation of IEEE or ISO rules/specifications for
6237 math functions. It may, however, yield faster code for programs
6238 that do not require the guarantees of these specifications.
6240 The default is @option{-fmath-errno}.
6242 On Darwin systems, the math library never sets @code{errno}. There is
6243 therefore no reason for the compiler to consider the possibility that
6244 it might, and @option{-fno-math-errno} is the default.
6246 @item -funsafe-math-optimizations
6247 @opindex funsafe-math-optimizations
6249 Allow optimizations for floating-point arithmetic that (a) assume
6250 that arguments and results are valid and (b) may violate IEEE or
6251 ANSI standards. When used at link-time, it may include libraries
6252 or startup files that change the default FPU control word or other
6253 similar optimizations.
6255 This option is not turned on by any @option{-O} option since
6256 it can result in incorrect output for programs which depend on
6257 an exact implementation of IEEE or ISO rules/specifications for
6258 math functions. It may, however, yield faster code for programs
6259 that do not require the guarantees of these specifications.
6260 Enables @option{-freciprocal-math} and @option{-fassociative-math}.
6262 The default is @option{-fno-unsafe-math-optimizations}.
6264 @item -fassociative-math
6265 @opindex fassociative-math
6267 Allow re-association of operands in series of floating-point operations.
6268 This violates the ISO C and C++ language standard by possibly changing
6269 computation result. NOTE: re-ordering may change the sign of zero as
6270 well as ignore NaNs and inhibit or create underflow or overflow (and
6271 thus cannot be used on a code which relies on rounding behavior like
6272 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6273 and thus may not be used when ordered comparisons are required.
6274 This flag doesn't make much sense without @option{-fno-signed-zeros}
6275 or @option{-fno-trapping-math} or with @option{-frounding-math}.
6277 The default is @option{-fno-associative-math}.
6279 @item -freciprocal-math
6280 @opindex freciprocal-math
6282 Allow the reciprocal of a value to be used instead of dividing by
6283 the value if this enables optimizations. For example @code{x / y}
6284 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6285 is subject to common subexpression elimination. Note that this loses
6286 precision and increases the number of flops operating on the value.
6288 The default is @option{-fno-reciprocal-math}.
6290 @item -ffinite-math-only
6291 @opindex ffinite-math-only
6292 Allow optimizations for floating-point arithmetic that assume
6293 that arguments and results are not NaNs or +-Infs.
6295 This option is not turned on by any @option{-O} option since
6296 it can result in incorrect output for programs which depend on
6297 an exact implementation of IEEE or ISO rules/specifications for
6298 math functions. It may, however, yield faster code for programs
6299 that do not require the guarantees of these specifications.
6301 The default is @option{-fno-finite-math-only}.
6303 @item -fno-signed-zeros
6304 @opindex fno-signed-zeros
6305 Allow optimizations for floating point arithmetic that ignore the
6306 signedness of zero. IEEE arithmetic specifies the behavior of
6307 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6308 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6309 This option implies that the sign of a zero result isn't significant.
6311 The default is @option{-fsigned-zeros}.
6313 @item -fno-trapping-math
6314 @opindex fno-trapping-math
6315 Compile code assuming that floating-point operations cannot generate
6316 user-visible traps. These traps include division by zero, overflow,
6317 underflow, inexact result and invalid operation. This option implies
6318 @option{-fno-signaling-nans}. Setting this option may allow faster
6319 code if one relies on ``non-stop'' IEEE arithmetic, for example.
6321 This option should never be turned on by any @option{-O} option since
6322 it can result in incorrect output for programs which depend on
6323 an exact implementation of IEEE or ISO rules/specifications for
6326 The default is @option{-ftrapping-math}.
6328 @item -frounding-math
6329 @opindex frounding-math
6330 Disable transformations and optimizations that assume default floating
6331 point rounding behavior. This is round-to-zero for all floating point
6332 to integer conversions, and round-to-nearest for all other arithmetic
6333 truncations. This option should be specified for programs that change
6334 the FP rounding mode dynamically, or that may be executed with a
6335 non-default rounding mode. This option disables constant folding of
6336 floating point expressions at compile-time (which may be affected by
6337 rounding mode) and arithmetic transformations that are unsafe in the
6338 presence of sign-dependent rounding modes.
6340 The default is @option{-fno-rounding-math}.
6342 This option is experimental and does not currently guarantee to
6343 disable all GCC optimizations that are affected by rounding mode.
6344 Future versions of GCC may provide finer control of this setting
6345 using C99's @code{FENV_ACCESS} pragma. This command line option
6346 will be used to specify the default state for @code{FENV_ACCESS}.
6348 @item -frtl-abstract-sequences
6349 @opindex frtl-abstract-sequences
6350 It is a size optimization method. This option is to find identical
6351 sequences of code, which can be turned into pseudo-procedures and
6352 then replace all occurrences with calls to the newly created
6353 subroutine. It is kind of an opposite of @option{-finline-functions}.
6354 This optimization runs at RTL level.
6356 @item -fsignaling-nans
6357 @opindex fsignaling-nans
6358 Compile code assuming that IEEE signaling NaNs may generate user-visible
6359 traps during floating-point operations. Setting this option disables
6360 optimizations that may change the number of exceptions visible with
6361 signaling NaNs. This option implies @option{-ftrapping-math}.
6363 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6366 The default is @option{-fno-signaling-nans}.
6368 This option is experimental and does not currently guarantee to
6369 disable all GCC optimizations that affect signaling NaN behavior.
6371 @item -fsingle-precision-constant
6372 @opindex fsingle-precision-constant
6373 Treat floating point constant as single precision constant instead of
6374 implicitly converting it to double precision constant.
6376 @item -fcx-limited-range
6377 @opindex fcx-limited-range
6378 When enabled, this option states that a range reduction step is not
6379 needed when performing complex division. The default is
6380 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
6382 This option controls the default setting of the ISO C99
6383 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6388 The following options control optimizations that may improve
6389 performance, but are not enabled by any @option{-O} options. This
6390 section includes experimental options that may produce broken code.
6393 @item -fbranch-probabilities
6394 @opindex fbranch-probabilities
6395 After running a program compiled with @option{-fprofile-arcs}
6396 (@pxref{Debugging Options,, Options for Debugging Your Program or
6397 @command{gcc}}), you can compile it a second time using
6398 @option{-fbranch-probabilities}, to improve optimizations based on
6399 the number of times each branch was taken. When the program
6400 compiled with @option{-fprofile-arcs} exits it saves arc execution
6401 counts to a file called @file{@var{sourcename}.gcda} for each source
6402 file. The information in this data file is very dependent on the
6403 structure of the generated code, so you must use the same source code
6404 and the same optimization options for both compilations.
6406 With @option{-fbranch-probabilities}, GCC puts a
6407 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6408 These can be used to improve optimization. Currently, they are only
6409 used in one place: in @file{reorg.c}, instead of guessing which path a
6410 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6411 exactly determine which path is taken more often.
6413 @item -fprofile-values
6414 @opindex fprofile-values
6415 If combined with @option{-fprofile-arcs}, it adds code so that some
6416 data about values of expressions in the program is gathered.
6418 With @option{-fbranch-probabilities}, it reads back the data gathered
6419 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6420 notes to instructions for their later usage in optimizations.
6422 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6426 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6427 a code to gather information about values of expressions.
6429 With @option{-fbranch-probabilities}, it reads back the data gathered
6430 and actually performs the optimizations based on them.
6431 Currently the optimizations include specialization of division operation
6432 using the knowledge about the value of the denominator.
6434 @item -frename-registers
6435 @opindex frename-registers
6436 Attempt to avoid false dependencies in scheduled code by making use
6437 of registers left over after register allocation. This optimization
6438 will most benefit processors with lots of registers. Depending on the
6439 debug information format adopted by the target, however, it can
6440 make debugging impossible, since variables will no longer stay in
6441 a ``home register''.
6443 Enabled by default with @option{-funroll-loops}.
6447 Perform tail duplication to enlarge superblock size. This transformation
6448 simplifies the control flow of the function allowing other optimizations to do
6451 Enabled with @option{-fprofile-use}.
6453 @item -funroll-loops
6454 @opindex funroll-loops
6455 Unroll loops whose number of iterations can be determined at compile time or
6456 upon entry to the loop. @option{-funroll-loops} implies
6457 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6458 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6459 small constant number of iterations). This option makes code larger, and may
6460 or may not make it run faster.
6462 Enabled with @option{-fprofile-use}.
6464 @item -funroll-all-loops
6465 @opindex funroll-all-loops
6466 Unroll all loops, even if their number of iterations is uncertain when
6467 the loop is entered. This usually makes programs run more slowly.
6468 @option{-funroll-all-loops} implies the same options as
6469 @option{-funroll-loops}.
6472 @opindex fpeel-loops
6473 Peels the loops for that there is enough information that they do not
6474 roll much (from profile feedback). It also turns on complete loop peeling
6475 (i.e.@: complete removal of loops with small constant number of iterations).
6477 Enabled with @option{-fprofile-use}.
6479 @item -fmove-loop-invariants
6480 @opindex fmove-loop-invariants
6481 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6482 at level @option{-O1}
6484 @item -funswitch-loops
6485 @opindex funswitch-loops
6486 Move branches with loop invariant conditions out of the loop, with duplicates
6487 of the loop on both branches (modified according to result of the condition).
6489 @item -ffunction-sections
6490 @itemx -fdata-sections
6491 @opindex ffunction-sections
6492 @opindex fdata-sections
6493 Place each function or data item into its own section in the output
6494 file if the target supports arbitrary sections. The name of the
6495 function or the name of the data item determines the section's name
6498 Use these options on systems where the linker can perform optimizations
6499 to improve locality of reference in the instruction space. Most systems
6500 using the ELF object format and SPARC processors running Solaris 2 have
6501 linkers with such optimizations. AIX may have these optimizations in
6504 Only use these options when there are significant benefits from doing
6505 so. When you specify these options, the assembler and linker will
6506 create larger object and executable files and will also be slower.
6507 You will not be able to use @code{gprof} on all systems if you
6508 specify this option and you may have problems with debugging if
6509 you specify both this option and @option{-g}.
6511 @item -fbranch-target-load-optimize
6512 @opindex fbranch-target-load-optimize
6513 Perform branch target register load optimization before prologue / epilogue
6515 The use of target registers can typically be exposed only during reload,
6516 thus hoisting loads out of loops and doing inter-block scheduling needs
6517 a separate optimization pass.
6519 @item -fbranch-target-load-optimize2
6520 @opindex fbranch-target-load-optimize2
6521 Perform branch target register load optimization after prologue / epilogue
6524 @item -fbtr-bb-exclusive
6525 @opindex fbtr-bb-exclusive
6526 When performing branch target register load optimization, don't reuse
6527 branch target registers in within any basic block.
6529 @item -fstack-protector
6530 @opindex fstack-protector
6531 Emit extra code to check for buffer overflows, such as stack smashing
6532 attacks. This is done by adding a guard variable to functions with
6533 vulnerable objects. This includes functions that call alloca, and
6534 functions with buffers larger than 8 bytes. The guards are initialized
6535 when a function is entered and then checked when the function exits.
6536 If a guard check fails, an error message is printed and the program exits.
6538 @item -fstack-protector-all
6539 @opindex fstack-protector-all
6540 Like @option{-fstack-protector} except that all functions are protected.
6542 @item -fsection-anchors
6543 @opindex fsection-anchors
6544 Try to reduce the number of symbolic address calculations by using
6545 shared ``anchor'' symbols to address nearby objects. This transformation
6546 can help to reduce the number of GOT entries and GOT accesses on some
6549 For example, the implementation of the following function @code{foo}:
6553 int foo (void) @{ return a + b + c; @}
6556 would usually calculate the addresses of all three variables, but if you
6557 compile it with @option{-fsection-anchors}, it will access the variables
6558 from a common anchor point instead. The effect is similar to the
6559 following pseudocode (which isn't valid C):
6564 register int *xr = &x;
6565 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6569 Not all targets support this option.
6571 @item --param @var{name}=@var{value}
6573 In some places, GCC uses various constants to control the amount of
6574 optimization that is done. For example, GCC will not inline functions
6575 that contain more that a certain number of instructions. You can
6576 control some of these constants on the command-line using the
6577 @option{--param} option.
6579 The names of specific parameters, and the meaning of the values, are
6580 tied to the internals of the compiler, and are subject to change
6581 without notice in future releases.
6583 In each case, the @var{value} is an integer. The allowable choices for
6584 @var{name} are given in the following table:
6587 @item salias-max-implicit-fields
6588 The maximum number of fields in a variable without direct
6589 structure accesses for which structure aliasing will consider trying
6590 to track each field. The default is 5
6592 @item salias-max-array-elements
6593 The maximum number of elements an array can have and its elements
6594 still be tracked individually by structure aliasing. The default is 4
6596 @item sra-max-structure-size
6597 The maximum structure size, in bytes, at which the scalar replacement
6598 of aggregates (SRA) optimization will perform block copies. The
6599 default value, 0, implies that GCC will select the most appropriate
6602 @item sra-field-structure-ratio
6603 The threshold ratio (as a percentage) between instantiated fields and
6604 the complete structure size. We say that if the ratio of the number
6605 of bytes in instantiated fields to the number of bytes in the complete
6606 structure exceeds this parameter, then block copies are not used. The
6609 @item max-crossjump-edges
6610 The maximum number of incoming edges to consider for crossjumping.
6611 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6612 the number of edges incoming to each block. Increasing values mean
6613 more aggressive optimization, making the compile time increase with
6614 probably small improvement in executable size.
6616 @item min-crossjump-insns
6617 The minimum number of instructions which must be matched at the end
6618 of two blocks before crossjumping will be performed on them. This
6619 value is ignored in the case where all instructions in the block being
6620 crossjumped from are matched. The default value is 5.
6622 @item max-grow-copy-bb-insns
6623 The maximum code size expansion factor when copying basic blocks
6624 instead of jumping. The expansion is relative to a jump instruction.
6625 The default value is 8.
6627 @item max-goto-duplication-insns
6628 The maximum number of instructions to duplicate to a block that jumps
6629 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6630 passes, GCC factors computed gotos early in the compilation process,
6631 and unfactors them as late as possible. Only computed jumps at the
6632 end of a basic blocks with no more than max-goto-duplication-insns are
6633 unfactored. The default value is 8.
6635 @item max-delay-slot-insn-search
6636 The maximum number of instructions to consider when looking for an
6637 instruction to fill a delay slot. If more than this arbitrary number of
6638 instructions is searched, the time savings from filling the delay slot
6639 will be minimal so stop searching. Increasing values mean more
6640 aggressive optimization, making the compile time increase with probably
6641 small improvement in executable run time.
6643 @item max-delay-slot-live-search
6644 When trying to fill delay slots, the maximum number of instructions to
6645 consider when searching for a block with valid live register
6646 information. Increasing this arbitrarily chosen value means more
6647 aggressive optimization, increasing the compile time. This parameter
6648 should be removed when the delay slot code is rewritten to maintain the
6651 @item max-gcse-memory
6652 The approximate maximum amount of memory that will be allocated in
6653 order to perform the global common subexpression elimination
6654 optimization. If more memory than specified is required, the
6655 optimization will not be done.
6657 @item max-gcse-passes
6658 The maximum number of passes of GCSE to run. The default is 1.
6660 @item max-pending-list-length
6661 The maximum number of pending dependencies scheduling will allow
6662 before flushing the current state and starting over. Large functions
6663 with few branches or calls can create excessively large lists which
6664 needlessly consume memory and resources.
6666 @item max-inline-insns-single
6667 Several parameters control the tree inliner used in gcc.
6668 This number sets the maximum number of instructions (counted in GCC's
6669 internal representation) in a single function that the tree inliner
6670 will consider for inlining. This only affects functions declared
6671 inline and methods implemented in a class declaration (C++).
6672 The default value is 450.
6674 @item max-inline-insns-auto
6675 When you use @option{-finline-functions} (included in @option{-O3}),
6676 a lot of functions that would otherwise not be considered for inlining
6677 by the compiler will be investigated. To those functions, a different
6678 (more restrictive) limit compared to functions declared inline can
6680 The default value is 90.
6682 @item large-function-insns
6683 The limit specifying really large functions. For functions larger than this
6684 limit after inlining inlining is constrained by
6685 @option{--param large-function-growth}. This parameter is useful primarily
6686 to avoid extreme compilation time caused by non-linear algorithms used by the
6688 This parameter is ignored when @option{-funit-at-a-time} is not used.
6689 The default value is 2700.
6691 @item large-function-growth
6692 Specifies maximal growth of large function caused by inlining in percents.
6693 This parameter is ignored when @option{-funit-at-a-time} is not used.
6694 The default value is 100 which limits large function growth to 2.0 times
6697 @item large-unit-insns
6698 The limit specifying large translation unit. Growth caused by inlining of
6699 units larger than this limit is limited by @option{--param inline-unit-growth}.
6700 For small units this might be too tight (consider unit consisting of function A
6701 that is inline and B that just calls A three time. If B is small relative to
6702 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6703 large units consisting of small inlininable functions however the overall unit
6704 growth limit is needed to avoid exponential explosion of code size. Thus for
6705 smaller units, the size is increased to @option{--param large-unit-insns}
6706 before applying @option{--param inline-unit-growth}. The default is 10000
6708 @item inline-unit-growth
6709 Specifies maximal overall growth of the compilation unit caused by inlining.
6710 This parameter is ignored when @option{-funit-at-a-time} is not used.
6711 The default value is 30 which limits unit growth to 1.3 times the original
6714 @item large-stack-frame
6715 The limit specifying large stack frames. While inlining the algorithm is trying
6716 to not grow past this limit too much. Default value is 256 bytes.
6718 @item large-stack-frame-growth
6719 Specifies maximal growth of large stack frames caused by inlining in percents.
6720 The default value is 1000 which limits large stack frame growth to 11 times
6723 @item max-inline-insns-recursive
6724 @itemx max-inline-insns-recursive-auto
6725 Specifies maximum number of instructions out-of-line copy of self recursive inline
6726 function can grow into by performing recursive inlining.
6728 For functions declared inline @option{--param max-inline-insns-recursive} is
6729 taken into account. For function not declared inline, recursive inlining
6730 happens only when @option{-finline-functions} (included in @option{-O3}) is
6731 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6732 default value is 450.
6734 @item max-inline-recursive-depth
6735 @itemx max-inline-recursive-depth-auto
6736 Specifies maximum recursion depth used by the recursive inlining.
6738 For functions declared inline @option{--param max-inline-recursive-depth} is
6739 taken into account. For function not declared inline, recursive inlining
6740 happens only when @option{-finline-functions} (included in @option{-O3}) is
6741 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6742 default value is 450.
6744 @item min-inline-recursive-probability
6745 Recursive inlining is profitable only for function having deep recursion
6746 in average and can hurt for function having little recursion depth by
6747 increasing the prologue size or complexity of function body to other
6750 When profile feedback is available (see @option{-fprofile-generate}) the actual
6751 recursion depth can be guessed from probability that function will recurse via
6752 given call expression. This parameter limits inlining only to call expression
6753 whose probability exceeds given threshold (in percents). The default value is
6756 @item inline-call-cost
6757 Specify cost of call instruction relative to simple arithmetics operations
6758 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6759 functions and at the same time increases size of leaf function that is believed to
6760 reduce function size by being inlined. In effect it increases amount of
6761 inlining for code having large abstraction penalty (many functions that just
6762 pass the arguments to other functions) and decrease inlining for code with low
6763 abstraction penalty. The default value is 12.
6765 @item min-vect-loop-bound
6766 The minimum number of iterations under which a loop will not get vectorized
6767 when @option{-ftree-vectorize} is used. The number of iterations after
6768 vectorization needs to be greater than the value specified by this option
6769 to allow vectorization. The default value is 0.
6771 @item max-unrolled-insns
6772 The maximum number of instructions that a loop should have if that loop
6773 is unrolled, and if the loop is unrolled, it determines how many times
6774 the loop code is unrolled.
6776 @item max-average-unrolled-insns
6777 The maximum number of instructions biased by probabilities of their execution
6778 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6779 it determines how many times the loop code is unrolled.
6781 @item max-unroll-times
6782 The maximum number of unrollings of a single loop.
6784 @item max-peeled-insns
6785 The maximum number of instructions that a loop should have if that loop
6786 is peeled, and if the loop is peeled, it determines how many times
6787 the loop code is peeled.
6789 @item max-peel-times
6790 The maximum number of peelings of a single loop.
6792 @item max-completely-peeled-insns
6793 The maximum number of insns of a completely peeled loop.
6795 @item max-completely-peel-times
6796 The maximum number of iterations of a loop to be suitable for complete peeling.
6798 @item max-unswitch-insns
6799 The maximum number of insns of an unswitched loop.
6801 @item max-unswitch-level
6802 The maximum number of branches unswitched in a single loop.
6805 The minimum cost of an expensive expression in the loop invariant motion.
6807 @item iv-consider-all-candidates-bound
6808 Bound on number of candidates for induction variables below that
6809 all candidates are considered for each use in induction variable
6810 optimizations. Only the most relevant candidates are considered
6811 if there are more candidates, to avoid quadratic time complexity.
6813 @item iv-max-considered-uses
6814 The induction variable optimizations give up on loops that contain more
6815 induction variable uses.
6817 @item iv-always-prune-cand-set-bound
6818 If number of candidates in the set is smaller than this value,
6819 we always try to remove unnecessary ivs from the set during its
6820 optimization when a new iv is added to the set.
6822 @item scev-max-expr-size
6823 Bound on size of expressions used in the scalar evolutions analyzer.
6824 Large expressions slow the analyzer.
6826 @item omega-max-vars
6827 The maximum number of variables in an Omega constraint system.
6828 The default value is 128.
6830 @item omega-max-geqs
6831 The maximum number of inequalities in an Omega constraint system.
6832 The default value is 256.
6835 The maximum number of equalities in an Omega constraint system.
6836 The default value is 128.
6838 @item omega-max-wild-cards
6839 The maximum number of wildcard variables that the Omega solver will
6840 be able to insert. The default value is 18.
6842 @item omega-hash-table-size
6843 The size of the hash table in the Omega solver. The default value is
6846 @item omega-max-keys
6847 The maximal number of keys used by the Omega solver. The default
6850 @item omega-eliminate-redundant-constraints
6851 When set to 1, use expensive methods to eliminate all redundant
6852 constraints. The default value is 0.
6854 @item vect-max-version-for-alignment-checks
6855 The maximum number of runtime checks that can be performed when
6856 doing loop versioning for alignment in the vectorizer. See option
6857 ftree-vect-loop-version for more information.
6859 @item vect-max-version-for-alias-checks
6860 The maximum number of runtime checks that can be performed when
6861 doing loop versioning for alias in the vectorizer. See option
6862 ftree-vect-loop-version for more information.
6864 @item max-iterations-to-track
6866 The maximum number of iterations of a loop the brute force algorithm
6867 for analysis of # of iterations of the loop tries to evaluate.
6869 @item hot-bb-count-fraction
6870 Select fraction of the maximal count of repetitions of basic block in program
6871 given basic block needs to have to be considered hot.
6873 @item hot-bb-frequency-fraction
6874 Select fraction of the maximal frequency of executions of basic block in
6875 function given basic block needs to have to be considered hot
6877 @item max-predicted-iterations
6878 The maximum number of loop iterations we predict statically. This is useful
6879 in cases where function contain single loop with known bound and other loop
6880 with unknown. We predict the known number of iterations correctly, while
6881 the unknown number of iterations average to roughly 10. This means that the
6882 loop without bounds would appear artificially cold relative to the other one.
6884 @item align-threshold
6886 Select fraction of the maximal frequency of executions of basic block in
6887 function given basic block will get aligned.
6889 @item align-loop-iterations
6891 A loop expected to iterate at lest the selected number of iterations will get
6894 @item tracer-dynamic-coverage
6895 @itemx tracer-dynamic-coverage-feedback
6897 This value is used to limit superblock formation once the given percentage of
6898 executed instructions is covered. This limits unnecessary code size
6901 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6902 feedback is available. The real profiles (as opposed to statically estimated
6903 ones) are much less balanced allowing the threshold to be larger value.
6905 @item tracer-max-code-growth
6906 Stop tail duplication once code growth has reached given percentage. This is
6907 rather hokey argument, as most of the duplicates will be eliminated later in
6908 cross jumping, so it may be set to much higher values than is the desired code
6911 @item tracer-min-branch-ratio
6913 Stop reverse growth when the reverse probability of best edge is less than this
6914 threshold (in percent).
6916 @item tracer-min-branch-ratio
6917 @itemx tracer-min-branch-ratio-feedback
6919 Stop forward growth if the best edge do have probability lower than this
6922 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6923 compilation for profile feedback and one for compilation without. The value
6924 for compilation with profile feedback needs to be more conservative (higher) in
6925 order to make tracer effective.
6927 @item max-cse-path-length
6929 Maximum number of basic blocks on path that cse considers. The default is 10.
6932 The maximum instructions CSE process before flushing. The default is 1000.
6934 @item max-aliased-vops
6936 Maximum number of virtual operands per function allowed to represent
6937 aliases before triggering the alias partitioning heuristic. Alias
6938 partitioning reduces compile times and memory consumption needed for
6939 aliasing at the expense of precision loss in alias information. The
6940 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
6943 Notice that if a function contains more memory statements than the
6944 value of this parameter, it is not really possible to achieve this
6945 reduction. In this case, the compiler will use the number of memory
6946 statements as the value for @option{max-aliased-vops}.
6948 @item avg-aliased-vops
6950 Average number of virtual operands per statement allowed to represent
6951 aliases before triggering the alias partitioning heuristic. This
6952 works in conjunction with @option{max-aliased-vops}. If a function
6953 contains more than @option{max-aliased-vops} virtual operators, then
6954 memory symbols will be grouped into memory partitions until either the
6955 total number of virtual operators is below @option{max-aliased-vops}
6956 or the average number of virtual operators per memory statement is
6957 below @option{avg-aliased-vops}. The default value for this parameter
6958 is 1 for -O1 and -O2, and 3 for -O3.
6960 @item ggc-min-expand
6962 GCC uses a garbage collector to manage its own memory allocation. This
6963 parameter specifies the minimum percentage by which the garbage
6964 collector's heap should be allowed to expand between collections.
6965 Tuning this may improve compilation speed; it has no effect on code
6968 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6969 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6970 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6971 GCC is not able to calculate RAM on a particular platform, the lower
6972 bound of 30% is used. Setting this parameter and
6973 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6974 every opportunity. This is extremely slow, but can be useful for
6977 @item ggc-min-heapsize
6979 Minimum size of the garbage collector's heap before it begins bothering
6980 to collect garbage. The first collection occurs after the heap expands
6981 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6982 tuning this may improve compilation speed, and has no effect on code
6985 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6986 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6987 with a lower bound of 4096 (four megabytes) and an upper bound of
6988 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6989 particular platform, the lower bound is used. Setting this parameter
6990 very large effectively disables garbage collection. Setting this
6991 parameter and @option{ggc-min-expand} to zero causes a full collection
6992 to occur at every opportunity.
6994 @item max-reload-search-insns
6995 The maximum number of instruction reload should look backward for equivalent
6996 register. Increasing values mean more aggressive optimization, making the
6997 compile time increase with probably slightly better performance. The default
7000 @item max-cselib-memory-locations
7001 The maximum number of memory locations cselib should take into account.
7002 Increasing values mean more aggressive optimization, making the compile time
7003 increase with probably slightly better performance. The default value is 500.
7005 @item max-flow-memory-locations
7006 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
7007 The default value is 100.
7009 @item reorder-blocks-duplicate
7010 @itemx reorder-blocks-duplicate-feedback
7012 Used by basic block reordering pass to decide whether to use unconditional
7013 branch or duplicate the code on its destination. Code is duplicated when its
7014 estimated size is smaller than this value multiplied by the estimated size of
7015 unconditional jump in the hot spots of the program.
7017 The @option{reorder-block-duplicate-feedback} is used only when profile
7018 feedback is available and may be set to higher values than
7019 @option{reorder-block-duplicate} since information about the hot spots is more
7022 @item max-sched-ready-insns
7023 The maximum number of instructions ready to be issued the scheduler should
7024 consider at any given time during the first scheduling pass. Increasing
7025 values mean more thorough searches, making the compilation time increase
7026 with probably little benefit. The default value is 100.
7028 @item max-sched-region-blocks
7029 The maximum number of blocks in a region to be considered for
7030 interblock scheduling. The default value is 10.
7032 @item max-sched-region-insns
7033 The maximum number of insns in a region to be considered for
7034 interblock scheduling. The default value is 100.
7037 The minimum probability (in percents) of reaching a source block
7038 for interblock speculative scheduling. The default value is 40.
7040 @item max-sched-extend-regions-iters
7041 The maximum number of iterations through CFG to extend regions.
7042 0 - disable region extension,
7043 N - do at most N iterations.
7044 The default value is 0.
7046 @item max-sched-insn-conflict-delay
7047 The maximum conflict delay for an insn to be considered for speculative motion.
7048 The default value is 3.
7050 @item sched-spec-prob-cutoff
7051 The minimal probability of speculation success (in percents), so that
7052 speculative insn will be scheduled.
7053 The default value is 40.
7055 @item max-last-value-rtl
7057 The maximum size measured as number of RTLs that can be recorded in an expression
7058 in combiner for a pseudo register as last known value of that register. The default
7061 @item integer-share-limit
7062 Small integer constants can use a shared data structure, reducing the
7063 compiler's memory usage and increasing its speed. This sets the maximum
7064 value of a shared integer constant's. The default value is 256.
7066 @item min-virtual-mappings
7067 Specifies the minimum number of virtual mappings in the incremental
7068 SSA updater that should be registered to trigger the virtual mappings
7069 heuristic defined by virtual-mappings-ratio. The default value is
7072 @item virtual-mappings-ratio
7073 If the number of virtual mappings is virtual-mappings-ratio bigger
7074 than the number of virtual symbols to be updated, then the incremental
7075 SSA updater switches to a full update for those symbols. The default
7078 @item ssp-buffer-size
7079 The minimum size of buffers (i.e. arrays) that will receive stack smashing
7080 protection when @option{-fstack-protection} is used.
7082 @item max-jump-thread-duplication-stmts
7083 Maximum number of statements allowed in a block that needs to be
7084 duplicated when threading jumps.
7086 @item max-fields-for-field-sensitive
7087 Maximum number of fields in a structure we will treat in
7088 a field sensitive manner during pointer analysis.
7090 @item prefetch-latency
7091 Estimate on average number of instructions that are executed before
7092 prefetch finishes. The distance we prefetch ahead is proportional
7093 to this constant. Increasing this number may also lead to less
7094 streams being prefetched (see @option{simultaneous-prefetches}).
7096 @item simultaneous-prefetches
7097 Maximum number of prefetches that can run at the same time.
7099 @item l1-cache-line-size
7100 The size of cache line in L1 cache, in bytes.
7103 The size of L1 cache, in kilobytes.
7106 The size of L2 cache, in kilobytes.
7108 @item use-canonical-types
7109 Whether the compiler should use the ``canonical'' type system. By
7110 default, this should always be 1, which uses a more efficient internal
7111 mechanism for comparing types in C++ and Objective-C++. However, if
7112 bugs in the canonical type system are causing compilation failures,
7113 set this value to 0 to disable canonical types.
7118 @node Preprocessor Options
7119 @section Options Controlling the Preprocessor
7120 @cindex preprocessor options
7121 @cindex options, preprocessor
7123 These options control the C preprocessor, which is run on each C source
7124 file before actual compilation.
7126 If you use the @option{-E} option, nothing is done except preprocessing.
7127 Some of these options make sense only together with @option{-E} because
7128 they cause the preprocessor output to be unsuitable for actual
7133 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7134 and pass @var{option} directly through to the preprocessor. If
7135 @var{option} contains commas, it is split into multiple options at the
7136 commas. However, many options are modified, translated or interpreted
7137 by the compiler driver before being passed to the preprocessor, and
7138 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7139 interface is undocumented and subject to change, so whenever possible
7140 you should avoid using @option{-Wp} and let the driver handle the
7143 @item -Xpreprocessor @var{option}
7144 @opindex preprocessor
7145 Pass @var{option} as an option to the preprocessor. You can use this to
7146 supply system-specific preprocessor options which GCC does not know how to
7149 If you want to pass an option that takes an argument, you must use
7150 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7153 @include cppopts.texi
7155 @node Assembler Options
7156 @section Passing Options to the Assembler
7158 @c prevent bad page break with this line
7159 You can pass options to the assembler.
7162 @item -Wa,@var{option}
7164 Pass @var{option} as an option to the assembler. If @var{option}
7165 contains commas, it is split into multiple options at the commas.
7167 @item -Xassembler @var{option}
7169 Pass @var{option} as an option to the assembler. You can use this to
7170 supply system-specific assembler options which GCC does not know how to
7173 If you want to pass an option that takes an argument, you must use
7174 @option{-Xassembler} twice, once for the option and once for the argument.
7179 @section Options for Linking
7180 @cindex link options
7181 @cindex options, linking
7183 These options come into play when the compiler links object files into
7184 an executable output file. They are meaningless if the compiler is
7185 not doing a link step.
7189 @item @var{object-file-name}
7190 A file name that does not end in a special recognized suffix is
7191 considered to name an object file or library. (Object files are
7192 distinguished from libraries by the linker according to the file
7193 contents.) If linking is done, these object files are used as input
7202 If any of these options is used, then the linker is not run, and
7203 object file names should not be used as arguments. @xref{Overall
7207 @item -l@var{library}
7208 @itemx -l @var{library}
7210 Search the library named @var{library} when linking. (The second
7211 alternative with the library as a separate argument is only for
7212 POSIX compliance and is not recommended.)
7214 It makes a difference where in the command you write this option; the
7215 linker searches and processes libraries and object files in the order they
7216 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7217 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7218 to functions in @samp{z}, those functions may not be loaded.
7220 The linker searches a standard list of directories for the library,
7221 which is actually a file named @file{lib@var{library}.a}. The linker
7222 then uses this file as if it had been specified precisely by name.
7224 The directories searched include several standard system directories
7225 plus any that you specify with @option{-L}.
7227 Normally the files found this way are library files---archive files
7228 whose members are object files. The linker handles an archive file by
7229 scanning through it for members which define symbols that have so far
7230 been referenced but not defined. But if the file that is found is an
7231 ordinary object file, it is linked in the usual fashion. The only
7232 difference between using an @option{-l} option and specifying a file name
7233 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7234 and searches several directories.
7238 You need this special case of the @option{-l} option in order to
7239 link an Objective-C or Objective-C++ program.
7242 @opindex nostartfiles
7243 Do not use the standard system startup files when linking.
7244 The standard system libraries are used normally, unless @option{-nostdlib}
7245 or @option{-nodefaultlibs} is used.
7247 @item -nodefaultlibs
7248 @opindex nodefaultlibs
7249 Do not use the standard system libraries when linking.
7250 Only the libraries you specify will be passed to the linker.
7251 The standard startup files are used normally, unless @option{-nostartfiles}
7252 is used. The compiler may generate calls to @code{memcmp},
7253 @code{memset}, @code{memcpy} and @code{memmove}.
7254 These entries are usually resolved by entries in
7255 libc. These entry points should be supplied through some other
7256 mechanism when this option is specified.
7260 Do not use the standard system startup files or libraries when linking.
7261 No startup files and only the libraries you specify will be passed to
7262 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7263 @code{memcpy} and @code{memmove}.
7264 These entries are usually resolved by entries in
7265 libc. These entry points should be supplied through some other
7266 mechanism when this option is specified.
7268 @cindex @option{-lgcc}, use with @option{-nostdlib}
7269 @cindex @option{-nostdlib} and unresolved references
7270 @cindex unresolved references and @option{-nostdlib}
7271 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7272 @cindex @option{-nodefaultlibs} and unresolved references
7273 @cindex unresolved references and @option{-nodefaultlibs}
7274 One of the standard libraries bypassed by @option{-nostdlib} and
7275 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7276 that GCC uses to overcome shortcomings of particular machines, or special
7277 needs for some languages.
7278 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7279 Collection (GCC) Internals},
7280 for more discussion of @file{libgcc.a}.)
7281 In most cases, you need @file{libgcc.a} even when you want to avoid
7282 other standard libraries. In other words, when you specify @option{-nostdlib}
7283 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7284 This ensures that you have no unresolved references to internal GCC
7285 library subroutines. (For example, @samp{__main}, used to ensure C++
7286 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7287 GNU Compiler Collection (GCC) Internals}.)
7291 Produce a position independent executable on targets which support it.
7292 For predictable results, you must also specify the same set of options
7293 that were used to generate code (@option{-fpie}, @option{-fPIE},
7294 or model suboptions) when you specify this option.
7298 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7299 that support it. This instructs the linker to add all symbols, not
7300 only used ones, to the dynamic symbol table. This option is needed
7301 for some uses of @code{dlopen} or to allow obtaining backtraces
7302 from within a program.
7306 Remove all symbol table and relocation information from the executable.
7310 On systems that support dynamic linking, this prevents linking with the shared
7311 libraries. On other systems, this option has no effect.
7315 Produce a shared object which can then be linked with other objects to
7316 form an executable. Not all systems support this option. For predictable
7317 results, you must also specify the same set of options that were used to
7318 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7319 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7320 needs to build supplementary stub code for constructors to work. On
7321 multi-libbed systems, @samp{gcc -shared} must select the correct support
7322 libraries to link against. Failing to supply the correct flags may lead
7323 to subtle defects. Supplying them in cases where they are not necessary
7326 @item -shared-libgcc
7327 @itemx -static-libgcc
7328 @opindex shared-libgcc
7329 @opindex static-libgcc
7330 On systems that provide @file{libgcc} as a shared library, these options
7331 force the use of either the shared or static version respectively.
7332 If no shared version of @file{libgcc} was built when the compiler was
7333 configured, these options have no effect.
7335 There are several situations in which an application should use the
7336 shared @file{libgcc} instead of the static version. The most common
7337 of these is when the application wishes to throw and catch exceptions
7338 across different shared libraries. In that case, each of the libraries
7339 as well as the application itself should use the shared @file{libgcc}.
7341 Therefore, the G++ and GCJ drivers automatically add
7342 @option{-shared-libgcc} whenever you build a shared library or a main
7343 executable, because C++ and Java programs typically use exceptions, so
7344 this is the right thing to do.
7346 If, instead, you use the GCC driver to create shared libraries, you may
7347 find that they will not always be linked with the shared @file{libgcc}.
7348 If GCC finds, at its configuration time, that you have a non-GNU linker
7349 or a GNU linker that does not support option @option{--eh-frame-hdr},
7350 it will link the shared version of @file{libgcc} into shared libraries
7351 by default. Otherwise, it will take advantage of the linker and optimize
7352 away the linking with the shared version of @file{libgcc}, linking with
7353 the static version of libgcc by default. This allows exceptions to
7354 propagate through such shared libraries, without incurring relocation
7355 costs at library load time.
7357 However, if a library or main executable is supposed to throw or catch
7358 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7359 for the languages used in the program, or using the option
7360 @option{-shared-libgcc}, such that it is linked with the shared
7365 Bind references to global symbols when building a shared object. Warn
7366 about any unresolved references (unless overridden by the link editor
7367 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7370 @item -Xlinker @var{option}
7372 Pass @var{option} as an option to the linker. You can use this to
7373 supply system-specific linker options which GCC does not know how to
7376 If you want to pass an option that takes an argument, you must use
7377 @option{-Xlinker} twice, once for the option and once for the argument.
7378 For example, to pass @option{-assert definitions}, you must write
7379 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7380 @option{-Xlinker "-assert definitions"}, because this passes the entire
7381 string as a single argument, which is not what the linker expects.
7383 @item -Wl,@var{option}
7385 Pass @var{option} as an option to the linker. If @var{option} contains
7386 commas, it is split into multiple options at the commas.
7388 @item -u @var{symbol}
7390 Pretend the symbol @var{symbol} is undefined, to force linking of
7391 library modules to define it. You can use @option{-u} multiple times with
7392 different symbols to force loading of additional library modules.
7395 @node Directory Options
7396 @section Options for Directory Search
7397 @cindex directory options
7398 @cindex options, directory search
7401 These options specify directories to search for header files, for
7402 libraries and for parts of the compiler:
7407 Add the directory @var{dir} to the head of the list of directories to be
7408 searched for header files. This can be used to override a system header
7409 file, substituting your own version, since these directories are
7410 searched before the system header file directories. However, you should
7411 not use this option to add directories that contain vendor-supplied
7412 system header files (use @option{-isystem} for that). If you use more than
7413 one @option{-I} option, the directories are scanned in left-to-right
7414 order; the standard system directories come after.
7416 If a standard system include directory, or a directory specified with
7417 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7418 option will be ignored. The directory will still be searched but as a
7419 system directory at its normal position in the system include chain.
7420 This is to ensure that GCC's procedure to fix buggy system headers and
7421 the ordering for the include_next directive are not inadvertently changed.
7422 If you really need to change the search order for system directories,
7423 use the @option{-nostdinc} and/or @option{-isystem} options.
7425 @item -iquote@var{dir}
7427 Add the directory @var{dir} to the head of the list of directories to
7428 be searched for header files only for the case of @samp{#include
7429 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7430 otherwise just like @option{-I}.
7434 Add directory @var{dir} to the list of directories to be searched
7437 @item -B@var{prefix}
7439 This option specifies where to find the executables, libraries,
7440 include files, and data files of the compiler itself.
7442 The compiler driver program runs one or more of the subprograms
7443 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7444 @var{prefix} as a prefix for each program it tries to run, both with and
7445 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7447 For each subprogram to be run, the compiler driver first tries the
7448 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7449 was not specified, the driver tries two standard prefixes, which are
7450 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7451 those results in a file name that is found, the unmodified program
7452 name is searched for using the directories specified in your
7453 @env{PATH} environment variable.
7455 The compiler will check to see if the path provided by the @option{-B}
7456 refers to a directory, and if necessary it will add a directory
7457 separator character at the end of the path.
7459 @option{-B} prefixes that effectively specify directory names also apply
7460 to libraries in the linker, because the compiler translates these
7461 options into @option{-L} options for the linker. They also apply to
7462 includes files in the preprocessor, because the compiler translates these
7463 options into @option{-isystem} options for the preprocessor. In this case,
7464 the compiler appends @samp{include} to the prefix.
7466 The run-time support file @file{libgcc.a} can also be searched for using
7467 the @option{-B} prefix, if needed. If it is not found there, the two
7468 standard prefixes above are tried, and that is all. The file is left
7469 out of the link if it is not found by those means.
7471 Another way to specify a prefix much like the @option{-B} prefix is to use
7472 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7475 As a special kludge, if the path provided by @option{-B} is
7476 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7477 9, then it will be replaced by @file{[dir/]include}. This is to help
7478 with boot-strapping the compiler.
7480 @item -specs=@var{file}
7482 Process @var{file} after the compiler reads in the standard @file{specs}
7483 file, in order to override the defaults that the @file{gcc} driver
7484 program uses when determining what switches to pass to @file{cc1},
7485 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7486 @option{-specs=@var{file}} can be specified on the command line, and they
7487 are processed in order, from left to right.
7489 @item --sysroot=@var{dir}
7491 Use @var{dir} as the logical root directory for headers and libraries.
7492 For example, if the compiler would normally search for headers in
7493 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7494 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7496 If you use both this option and the @option{-isysroot} option, then
7497 the @option{--sysroot} option will apply to libraries, but the
7498 @option{-isysroot} option will apply to header files.
7500 The GNU linker (beginning with version 2.16) has the necessary support
7501 for this option. If your linker does not support this option, the
7502 header file aspect of @option{--sysroot} will still work, but the
7503 library aspect will not.
7507 This option has been deprecated. Please use @option{-iquote} instead for
7508 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7509 Any directories you specify with @option{-I} options before the @option{-I-}
7510 option are searched only for the case of @samp{#include "@var{file}"};
7511 they are not searched for @samp{#include <@var{file}>}.
7513 If additional directories are specified with @option{-I} options after
7514 the @option{-I-}, these directories are searched for all @samp{#include}
7515 directives. (Ordinarily @emph{all} @option{-I} directories are used
7518 In addition, the @option{-I-} option inhibits the use of the current
7519 directory (where the current input file came from) as the first search
7520 directory for @samp{#include "@var{file}"}. There is no way to
7521 override this effect of @option{-I-}. With @option{-I.} you can specify
7522 searching the directory which was current when the compiler was
7523 invoked. That is not exactly the same as what the preprocessor does
7524 by default, but it is often satisfactory.
7526 @option{-I-} does not inhibit the use of the standard system directories
7527 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7534 @section Specifying subprocesses and the switches to pass to them
7537 @command{gcc} is a driver program. It performs its job by invoking a
7538 sequence of other programs to do the work of compiling, assembling and
7539 linking. GCC interprets its command-line parameters and uses these to
7540 deduce which programs it should invoke, and which command-line options
7541 it ought to place on their command lines. This behavior is controlled
7542 by @dfn{spec strings}. In most cases there is one spec string for each
7543 program that GCC can invoke, but a few programs have multiple spec
7544 strings to control their behavior. The spec strings built into GCC can
7545 be overridden by using the @option{-specs=} command-line switch to specify
7548 @dfn{Spec files} are plaintext files that are used to construct spec
7549 strings. They consist of a sequence of directives separated by blank
7550 lines. The type of directive is determined by the first non-whitespace
7551 character on the line and it can be one of the following:
7554 @item %@var{command}
7555 Issues a @var{command} to the spec file processor. The commands that can
7559 @item %include <@var{file}>
7561 Search for @var{file} and insert its text at the current point in the
7564 @item %include_noerr <@var{file}>
7565 @cindex %include_noerr
7566 Just like @samp{%include}, but do not generate an error message if the include
7567 file cannot be found.
7569 @item %rename @var{old_name} @var{new_name}
7571 Rename the spec string @var{old_name} to @var{new_name}.
7575 @item *[@var{spec_name}]:
7576 This tells the compiler to create, override or delete the named spec
7577 string. All lines after this directive up to the next directive or
7578 blank line are considered to be the text for the spec string. If this
7579 results in an empty string then the spec will be deleted. (Or, if the
7580 spec did not exist, then nothing will happened.) Otherwise, if the spec
7581 does not currently exist a new spec will be created. If the spec does
7582 exist then its contents will be overridden by the text of this
7583 directive, unless the first character of that text is the @samp{+}
7584 character, in which case the text will be appended to the spec.
7586 @item [@var{suffix}]:
7587 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7588 and up to the next directive or blank line are considered to make up the
7589 spec string for the indicated suffix. When the compiler encounters an
7590 input file with the named suffix, it will processes the spec string in
7591 order to work out how to compile that file. For example:
7598 This says that any input file whose name ends in @samp{.ZZ} should be
7599 passed to the program @samp{z-compile}, which should be invoked with the
7600 command-line switch @option{-input} and with the result of performing the
7601 @samp{%i} substitution. (See below.)
7603 As an alternative to providing a spec string, the text that follows a
7604 suffix directive can be one of the following:
7607 @item @@@var{language}
7608 This says that the suffix is an alias for a known @var{language}. This is
7609 similar to using the @option{-x} command-line switch to GCC to specify a
7610 language explicitly. For example:
7617 Says that .ZZ files are, in fact, C++ source files.
7620 This causes an error messages saying:
7623 @var{name} compiler not installed on this system.
7627 GCC already has an extensive list of suffixes built into it.
7628 This directive will add an entry to the end of the list of suffixes, but
7629 since the list is searched from the end backwards, it is effectively
7630 possible to override earlier entries using this technique.
7634 GCC has the following spec strings built into it. Spec files can
7635 override these strings or create their own. Note that individual
7636 targets can also add their own spec strings to this list.
7639 asm Options to pass to the assembler
7640 asm_final Options to pass to the assembler post-processor
7641 cpp Options to pass to the C preprocessor
7642 cc1 Options to pass to the C compiler
7643 cc1plus Options to pass to the C++ compiler
7644 endfile Object files to include at the end of the link
7645 link Options to pass to the linker
7646 lib Libraries to include on the command line to the linker
7647 libgcc Decides which GCC support library to pass to the linker
7648 linker Sets the name of the linker
7649 predefines Defines to be passed to the C preprocessor
7650 signed_char Defines to pass to CPP to say whether @code{char} is signed
7652 startfile Object files to include at the start of the link
7655 Here is a small example of a spec file:
7661 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7664 This example renames the spec called @samp{lib} to @samp{old_lib} and
7665 then overrides the previous definition of @samp{lib} with a new one.
7666 The new definition adds in some extra command-line options before
7667 including the text of the old definition.
7669 @dfn{Spec strings} are a list of command-line options to be passed to their
7670 corresponding program. In addition, the spec strings can contain
7671 @samp{%}-prefixed sequences to substitute variable text or to
7672 conditionally insert text into the command line. Using these constructs
7673 it is possible to generate quite complex command lines.
7675 Here is a table of all defined @samp{%}-sequences for spec
7676 strings. Note that spaces are not generated automatically around the
7677 results of expanding these sequences. Therefore you can concatenate them
7678 together or combine them with constant text in a single argument.
7682 Substitute one @samp{%} into the program name or argument.
7685 Substitute the name of the input file being processed.
7688 Substitute the basename of the input file being processed.
7689 This is the substring up to (and not including) the last period
7690 and not including the directory.
7693 This is the same as @samp{%b}, but include the file suffix (text after
7697 Marks the argument containing or following the @samp{%d} as a
7698 temporary file name, so that that file will be deleted if GCC exits
7699 successfully. Unlike @samp{%g}, this contributes no text to the
7702 @item %g@var{suffix}
7703 Substitute a file name that has suffix @var{suffix} and is chosen
7704 once per compilation, and mark the argument in the same way as
7705 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7706 name is now chosen in a way that is hard to predict even when previously
7707 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7708 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7709 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7710 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7711 was simply substituted with a file name chosen once per compilation,
7712 without regard to any appended suffix (which was therefore treated
7713 just like ordinary text), making such attacks more likely to succeed.
7715 @item %u@var{suffix}
7716 Like @samp{%g}, but generates a new temporary file name even if
7717 @samp{%u@var{suffix}} was already seen.
7719 @item %U@var{suffix}
7720 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7721 new one if there is no such last file name. In the absence of any
7722 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7723 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7724 would involve the generation of two distinct file names, one
7725 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7726 simply substituted with a file name chosen for the previous @samp{%u},
7727 without regard to any appended suffix.
7729 @item %j@var{suffix}
7730 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7731 writable, and if save-temps is off; otherwise, substitute the name
7732 of a temporary file, just like @samp{%u}. This temporary file is not
7733 meant for communication between processes, but rather as a junk
7736 @item %|@var{suffix}
7737 @itemx %m@var{suffix}
7738 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7739 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7740 all. These are the two most common ways to instruct a program that it
7741 should read from standard input or write to standard output. If you
7742 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7743 construct: see for example @file{f/lang-specs.h}.
7745 @item %.@var{SUFFIX}
7746 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7747 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7748 terminated by the next space or %.
7751 Marks the argument containing or following the @samp{%w} as the
7752 designated output file of this compilation. This puts the argument
7753 into the sequence of arguments that @samp{%o} will substitute later.
7756 Substitutes the names of all the output files, with spaces
7757 automatically placed around them. You should write spaces
7758 around the @samp{%o} as well or the results are undefined.
7759 @samp{%o} is for use in the specs for running the linker.
7760 Input files whose names have no recognized suffix are not compiled
7761 at all, but they are included among the output files, so they will
7765 Substitutes the suffix for object files. Note that this is
7766 handled specially when it immediately follows @samp{%g, %u, or %U},
7767 because of the need for those to form complete file names. The
7768 handling is such that @samp{%O} is treated exactly as if it had already
7769 been substituted, except that @samp{%g, %u, and %U} do not currently
7770 support additional @var{suffix} characters following @samp{%O} as they would
7771 following, for example, @samp{.o}.
7774 Substitutes the standard macro predefinitions for the
7775 current target machine. Use this when running @code{cpp}.
7778 Like @samp{%p}, but puts @samp{__} before and after the name of each
7779 predefined macro, except for macros that start with @samp{__} or with
7780 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
7784 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
7785 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
7786 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
7787 and @option{-imultilib} as necessary.
7790 Current argument is the name of a library or startup file of some sort.
7791 Search for that file in a standard list of directories and substitute
7792 the full name found.
7795 Print @var{str} as an error message. @var{str} is terminated by a newline.
7796 Use this when inconsistent options are detected.
7799 Substitute the contents of spec string @var{name} at this point.
7802 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
7804 @item %x@{@var{option}@}
7805 Accumulate an option for @samp{%X}.
7808 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
7812 Output the accumulated assembler options specified by @option{-Wa}.
7815 Output the accumulated preprocessor options specified by @option{-Wp}.
7818 Process the @code{asm} spec. This is used to compute the
7819 switches to be passed to the assembler.
7822 Process the @code{asm_final} spec. This is a spec string for
7823 passing switches to an assembler post-processor, if such a program is
7827 Process the @code{link} spec. This is the spec for computing the
7828 command line passed to the linker. Typically it will make use of the
7829 @samp{%L %G %S %D and %E} sequences.
7832 Dump out a @option{-L} option for each directory that GCC believes might
7833 contain startup files. If the target supports multilibs then the
7834 current multilib directory will be prepended to each of these paths.
7837 Process the @code{lib} spec. This is a spec string for deciding which
7838 libraries should be included on the command line to the linker.
7841 Process the @code{libgcc} spec. This is a spec string for deciding
7842 which GCC support library should be included on the command line to the linker.
7845 Process the @code{startfile} spec. This is a spec for deciding which
7846 object files should be the first ones passed to the linker. Typically
7847 this might be a file named @file{crt0.o}.
7850 Process the @code{endfile} spec. This is a spec string that specifies
7851 the last object files that will be passed to the linker.
7854 Process the @code{cpp} spec. This is used to construct the arguments
7855 to be passed to the C preprocessor.
7858 Process the @code{cc1} spec. This is used to construct the options to be
7859 passed to the actual C compiler (@samp{cc1}).
7862 Process the @code{cc1plus} spec. This is used to construct the options to be
7863 passed to the actual C++ compiler (@samp{cc1plus}).
7866 Substitute the variable part of a matched option. See below.
7867 Note that each comma in the substituted string is replaced by
7871 Remove all occurrences of @code{-S} from the command line. Note---this
7872 command is position dependent. @samp{%} commands in the spec string
7873 before this one will see @code{-S}, @samp{%} commands in the spec string
7874 after this one will not.
7876 @item %:@var{function}(@var{args})
7877 Call the named function @var{function}, passing it @var{args}.
7878 @var{args} is first processed as a nested spec string, then split
7879 into an argument vector in the usual fashion. The function returns
7880 a string which is processed as if it had appeared literally as part
7881 of the current spec.
7883 The following built-in spec functions are provided:
7887 The @code{getenv} spec function takes two arguments: an environment
7888 variable name and a string. If the environment variable is not
7889 defined, a fatal error is issued. Otherwise, the return value is the
7890 value of the environment variable concatenated with the string. For
7891 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
7894 %:getenv(TOPDIR /include)
7897 expands to @file{/path/to/top/include}.
7899 @item @code{if-exists}
7900 The @code{if-exists} spec function takes one argument, an absolute
7901 pathname to a file. If the file exists, @code{if-exists} returns the
7902 pathname. Here is a small example of its usage:
7906 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7909 @item @code{if-exists-else}
7910 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7911 spec function, except that it takes two arguments. The first argument is
7912 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7913 returns the pathname. If it does not exist, it returns the second argument.
7914 This way, @code{if-exists-else} can be used to select one file or another,
7915 based on the existence of the first. Here is a small example of its usage:
7919 crt0%O%s %:if-exists(crti%O%s) \
7920 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7923 @item @code{replace-outfile}
7924 The @code{replace-outfile} spec function takes two arguments. It looks for the
7925 first argument in the outfiles array and replaces it with the second argument. Here
7926 is a small example of its usage:
7929 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
7932 @item @code{print-asm-header}
7933 The @code{print-asm-header} function takes no arguments and simply
7934 prints a banner like:
7940 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
7943 It is used to separate compiler options from assembler options
7944 in the @option{--target-help} output.
7948 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7949 If that switch was not specified, this substitutes nothing. Note that
7950 the leading dash is omitted when specifying this option, and it is
7951 automatically inserted if the substitution is performed. Thus the spec
7952 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7953 and would output the command line option @option{-foo}.
7955 @item %W@{@code{S}@}
7956 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7959 @item %@{@code{S}*@}
7960 Substitutes all the switches specified to GCC whose names start
7961 with @code{-S}, but which also take an argument. This is used for
7962 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7963 GCC considers @option{-o foo} as being
7964 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7965 text, including the space. Thus two arguments would be generated.
7967 @item %@{@code{S}*&@code{T}*@}
7968 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7969 (the order of @code{S} and @code{T} in the spec is not significant).
7970 There can be any number of ampersand-separated variables; for each the
7971 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7973 @item %@{@code{S}:@code{X}@}
7974 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7976 @item %@{!@code{S}:@code{X}@}
7977 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7979 @item %@{@code{S}*:@code{X}@}
7980 Substitutes @code{X} if one or more switches whose names start with
7981 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7982 once, no matter how many such switches appeared. However, if @code{%*}
7983 appears somewhere in @code{X}, then @code{X} will be substituted once
7984 for each matching switch, with the @code{%*} replaced by the part of
7985 that switch that matched the @code{*}.
7987 @item %@{.@code{S}:@code{X}@}
7988 Substitutes @code{X}, if processing a file with suffix @code{S}.
7990 @item %@{!.@code{S}:@code{X}@}
7991 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7993 @item %@{,@code{S}:@code{X}@}
7994 Substitutes @code{X}, if processing a file for language @code{S}.
7996 @item %@{!,@code{S}:@code{X}@}
7997 Substitutes @code{X}, if not processing a file for language @code{S}.
7999 @item %@{@code{S}|@code{P}:@code{X}@}
8000 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8001 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8002 @code{*} sequences as well, although they have a stronger binding than
8003 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8004 alternatives must be starred, and only the first matching alternative
8007 For example, a spec string like this:
8010 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8013 will output the following command-line options from the following input
8014 command-line options:
8019 -d fred.c -foo -baz -boggle
8020 -d jim.d -bar -baz -boggle
8023 @item %@{S:X; T:Y; :D@}
8025 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8026 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8027 be as many clauses as you need. This may be combined with @code{.},
8028 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8033 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8034 construct may contain other nested @samp{%} constructs or spaces, or
8035 even newlines. They are processed as usual, as described above.
8036 Trailing white space in @code{X} is ignored. White space may also
8037 appear anywhere on the left side of the colon in these constructs,
8038 except between @code{.} or @code{*} and the corresponding word.
8040 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8041 handled specifically in these constructs. If another value of
8042 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8043 @option{-W} switch is found later in the command line, the earlier
8044 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8045 just one letter, which passes all matching options.
8047 The character @samp{|} at the beginning of the predicate text is used to
8048 indicate that a command should be piped to the following command, but
8049 only if @option{-pipe} is specified.
8051 It is built into GCC which switches take arguments and which do not.
8052 (You might think it would be useful to generalize this to allow each
8053 compiler's spec to say which switches take arguments. But this cannot
8054 be done in a consistent fashion. GCC cannot even decide which input
8055 files have been specified without knowing which switches take arguments,
8056 and it must know which input files to compile in order to tell which
8059 GCC also knows implicitly that arguments starting in @option{-l} are to be
8060 treated as compiler output files, and passed to the linker in their
8061 proper position among the other output files.
8063 @c man begin OPTIONS
8065 @node Target Options
8066 @section Specifying Target Machine and Compiler Version
8067 @cindex target options
8068 @cindex cross compiling
8069 @cindex specifying machine version
8070 @cindex specifying compiler version and target machine
8071 @cindex compiler version, specifying
8072 @cindex target machine, specifying
8074 The usual way to run GCC is to run the executable called @file{gcc}, or
8075 @file{<machine>-gcc} when cross-compiling, or
8076 @file{<machine>-gcc-<version>} to run a version other than the one that
8077 was installed last. Sometimes this is inconvenient, so GCC provides
8078 options that will switch to another cross-compiler or version.
8081 @item -b @var{machine}
8083 The argument @var{machine} specifies the target machine for compilation.
8085 The value to use for @var{machine} is the same as was specified as the
8086 machine type when configuring GCC as a cross-compiler. For
8087 example, if a cross-compiler was configured with @samp{configure
8088 arm-elf}, meaning to compile for an arm processor with elf binaries,
8089 then you would specify @option{-b arm-elf} to run that cross compiler.
8090 Because there are other options beginning with @option{-b}, the
8091 configuration must contain a hyphen.
8093 @item -V @var{version}
8095 The argument @var{version} specifies which version of GCC to run.
8096 This is useful when multiple versions are installed. For example,
8097 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8100 The @option{-V} and @option{-b} options work by running the
8101 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8102 use them if you can just run that directly.
8104 @node Submodel Options
8105 @section Hardware Models and Configurations
8106 @cindex submodel options
8107 @cindex specifying hardware config
8108 @cindex hardware models and configurations, specifying
8109 @cindex machine dependent options
8111 Earlier we discussed the standard option @option{-b} which chooses among
8112 different installed compilers for completely different target
8113 machines, such as VAX vs.@: 68000 vs.@: 80386.
8115 In addition, each of these target machine types can have its own
8116 special options, starting with @samp{-m}, to choose among various
8117 hardware models or configurations---for example, 68010 vs 68020,
8118 floating coprocessor or none. A single installed version of the
8119 compiler can compile for any model or configuration, according to the
8122 Some configurations of the compiler also support additional special
8123 options, usually for compatibility with other compilers on the same
8126 @c This list is ordered alphanumerically by subsection name.
8127 @c It should be the same order and spelling as these options are listed
8128 @c in Machine Dependent Options
8134 * Blackfin Options::
8138 * DEC Alpha Options::
8139 * DEC Alpha/VMS Options::
8141 * GNU/Linux Options::
8144 * i386 and x86-64 Options::
8157 * RS/6000 and PowerPC Options::
8158 * S/390 and zSeries Options::
8163 * System V Options::
8164 * TMS320C3x/C4x Options::
8169 * Xstormy16 Options::
8175 @subsection ARC Options
8178 These options are defined for ARC implementations:
8183 Compile code for little endian mode. This is the default.
8187 Compile code for big endian mode.
8190 @opindex mmangle-cpu
8191 Prepend the name of the cpu to all public symbol names.
8192 In multiple-processor systems, there are many ARC variants with different
8193 instruction and register set characteristics. This flag prevents code
8194 compiled for one cpu to be linked with code compiled for another.
8195 No facility exists for handling variants that are ``almost identical''.
8196 This is an all or nothing option.
8198 @item -mcpu=@var{cpu}
8200 Compile code for ARC variant @var{cpu}.
8201 Which variants are supported depend on the configuration.
8202 All variants support @option{-mcpu=base}, this is the default.
8204 @item -mtext=@var{text-section}
8205 @itemx -mdata=@var{data-section}
8206 @itemx -mrodata=@var{readonly-data-section}
8210 Put functions, data, and readonly data in @var{text-section},
8211 @var{data-section}, and @var{readonly-data-section} respectively
8212 by default. This can be overridden with the @code{section} attribute.
8213 @xref{Variable Attributes}.
8218 @subsection ARM Options
8221 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8225 @item -mabi=@var{name}
8227 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8228 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8231 @opindex mapcs-frame
8232 Generate a stack frame that is compliant with the ARM Procedure Call
8233 Standard for all functions, even if this is not strictly necessary for
8234 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8235 with this option will cause the stack frames not to be generated for
8236 leaf functions. The default is @option{-mno-apcs-frame}.
8240 This is a synonym for @option{-mapcs-frame}.
8243 @c not currently implemented
8244 @item -mapcs-stack-check
8245 @opindex mapcs-stack-check
8246 Generate code to check the amount of stack space available upon entry to
8247 every function (that actually uses some stack space). If there is
8248 insufficient space available then either the function
8249 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8250 called, depending upon the amount of stack space required. The run time
8251 system is required to provide these functions. The default is
8252 @option{-mno-apcs-stack-check}, since this produces smaller code.
8254 @c not currently implemented
8256 @opindex mapcs-float
8257 Pass floating point arguments using the float point registers. This is
8258 one of the variants of the APCS@. This option is recommended if the
8259 target hardware has a floating point unit or if a lot of floating point
8260 arithmetic is going to be performed by the code. The default is
8261 @option{-mno-apcs-float}, since integer only code is slightly increased in
8262 size if @option{-mapcs-float} is used.
8264 @c not currently implemented
8265 @item -mapcs-reentrant
8266 @opindex mapcs-reentrant
8267 Generate reentrant, position independent code. The default is
8268 @option{-mno-apcs-reentrant}.
8271 @item -mthumb-interwork
8272 @opindex mthumb-interwork
8273 Generate code which supports calling between the ARM and Thumb
8274 instruction sets. Without this option the two instruction sets cannot
8275 be reliably used inside one program. The default is
8276 @option{-mno-thumb-interwork}, since slightly larger code is generated
8277 when @option{-mthumb-interwork} is specified.
8279 @item -mno-sched-prolog
8280 @opindex mno-sched-prolog
8281 Prevent the reordering of instructions in the function prolog, or the
8282 merging of those instruction with the instructions in the function's
8283 body. This means that all functions will start with a recognizable set
8284 of instructions (or in fact one of a choice from a small set of
8285 different function prologues), and this information can be used to
8286 locate the start if functions inside an executable piece of code. The
8287 default is @option{-msched-prolog}.
8290 @opindex mhard-float
8291 Generate output containing floating point instructions. This is the
8295 @opindex msoft-float
8296 Generate output containing library calls for floating point.
8297 @strong{Warning:} the requisite libraries are not available for all ARM
8298 targets. Normally the facilities of the machine's usual C compiler are
8299 used, but this cannot be done directly in cross-compilation. You must make
8300 your own arrangements to provide suitable library functions for
8303 @option{-msoft-float} changes the calling convention in the output file;
8304 therefore, it is only useful if you compile @emph{all} of a program with
8305 this option. In particular, you need to compile @file{libgcc.a}, the
8306 library that comes with GCC, with @option{-msoft-float} in order for
8309 @item -mfloat-abi=@var{name}
8311 Specifies which ABI to use for floating point values. Permissible values
8312 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8314 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
8315 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
8316 of floating point instructions, but still uses the soft-float calling
8319 @item -mlittle-endian
8320 @opindex mlittle-endian
8321 Generate code for a processor running in little-endian mode. This is
8322 the default for all standard configurations.
8325 @opindex mbig-endian
8326 Generate code for a processor running in big-endian mode; the default is
8327 to compile code for a little-endian processor.
8329 @item -mwords-little-endian
8330 @opindex mwords-little-endian
8331 This option only applies when generating code for big-endian processors.
8332 Generate code for a little-endian word order but a big-endian byte
8333 order. That is, a byte order of the form @samp{32107654}. Note: this
8334 option should only be used if you require compatibility with code for
8335 big-endian ARM processors generated by versions of the compiler prior to
8338 @item -mcpu=@var{name}
8340 This specifies the name of the target ARM processor. GCC uses this name
8341 to determine what kind of instructions it can emit when generating
8342 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8343 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8344 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8345 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8346 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8347 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8348 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8349 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8350 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8351 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8352 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8353 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8354 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8355 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8356 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3},
8357 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8359 @itemx -mtune=@var{name}
8361 This option is very similar to the @option{-mcpu=} option, except that
8362 instead of specifying the actual target processor type, and hence
8363 restricting which instructions can be used, it specifies that GCC should
8364 tune the performance of the code as if the target were of the type
8365 specified in this option, but still choosing the instructions that it
8366 will generate based on the cpu specified by a @option{-mcpu=} option.
8367 For some ARM implementations better performance can be obtained by using
8370 @item -march=@var{name}
8372 This specifies the name of the target ARM architecture. GCC uses this
8373 name to determine what kind of instructions it can emit when generating
8374 assembly code. This option can be used in conjunction with or instead
8375 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8376 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8377 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8378 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv7}, @samp{armv7-a},
8379 @samp{armv7-r}, @samp{armv7-m}, @samp{iwmmxt}, @samp{ep9312}.
8381 @item -mfpu=@var{name}
8382 @itemx -mfpe=@var{number}
8383 @itemx -mfp=@var{number}
8387 This specifies what floating point hardware (or hardware emulation) is
8388 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8389 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
8390 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8391 with older versions of GCC@.
8393 If @option{-msoft-float} is specified this specifies the format of
8394 floating point values.
8396 @item -mstructure-size-boundary=@var{n}
8397 @opindex mstructure-size-boundary
8398 The size of all structures and unions will be rounded up to a multiple
8399 of the number of bits set by this option. Permissible values are 8, 32
8400 and 64. The default value varies for different toolchains. For the COFF
8401 targeted toolchain the default value is 8. A value of 64 is only allowed
8402 if the underlying ABI supports it.
8404 Specifying the larger number can produce faster, more efficient code, but
8405 can also increase the size of the program. Different values are potentially
8406 incompatible. Code compiled with one value cannot necessarily expect to
8407 work with code or libraries compiled with another value, if they exchange
8408 information using structures or unions.
8410 @item -mabort-on-noreturn
8411 @opindex mabort-on-noreturn
8412 Generate a call to the function @code{abort} at the end of a
8413 @code{noreturn} function. It will be executed if the function tries to
8417 @itemx -mno-long-calls
8418 @opindex mlong-calls
8419 @opindex mno-long-calls
8420 Tells the compiler to perform function calls by first loading the
8421 address of the function into a register and then performing a subroutine
8422 call on this register. This switch is needed if the target function
8423 will lie outside of the 64 megabyte addressing range of the offset based
8424 version of subroutine call instruction.
8426 Even if this switch is enabled, not all function calls will be turned
8427 into long calls. The heuristic is that static functions, functions
8428 which have the @samp{short-call} attribute, functions that are inside
8429 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8430 definitions have already been compiled within the current compilation
8431 unit, will not be turned into long calls. The exception to this rule is
8432 that weak function definitions, functions with the @samp{long-call}
8433 attribute or the @samp{section} attribute, and functions that are within
8434 the scope of a @samp{#pragma long_calls} directive, will always be
8435 turned into long calls.
8437 This feature is not enabled by default. Specifying
8438 @option{-mno-long-calls} will restore the default behavior, as will
8439 placing the function calls within the scope of a @samp{#pragma
8440 long_calls_off} directive. Note these switches have no effect on how
8441 the compiler generates code to handle function calls via function
8444 @item -mnop-fun-dllimport
8445 @opindex mnop-fun-dllimport
8446 Disable support for the @code{dllimport} attribute.
8448 @item -msingle-pic-base
8449 @opindex msingle-pic-base
8450 Treat the register used for PIC addressing as read-only, rather than
8451 loading it in the prologue for each function. The run-time system is
8452 responsible for initializing this register with an appropriate value
8453 before execution begins.
8455 @item -mpic-register=@var{reg}
8456 @opindex mpic-register
8457 Specify the register to be used for PIC addressing. The default is R10
8458 unless stack-checking is enabled, when R9 is used.
8460 @item -mcirrus-fix-invalid-insns
8461 @opindex mcirrus-fix-invalid-insns
8462 @opindex mno-cirrus-fix-invalid-insns
8463 Insert NOPs into the instruction stream to in order to work around
8464 problems with invalid Maverick instruction combinations. This option
8465 is only valid if the @option{-mcpu=ep9312} option has been used to
8466 enable generation of instructions for the Cirrus Maverick floating
8467 point co-processor. This option is not enabled by default, since the
8468 problem is only present in older Maverick implementations. The default
8469 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8472 @item -mpoke-function-name
8473 @opindex mpoke-function-name
8474 Write the name of each function into the text section, directly
8475 preceding the function prologue. The generated code is similar to this:
8479 .ascii "arm_poke_function_name", 0
8482 .word 0xff000000 + (t1 - t0)
8483 arm_poke_function_name
8485 stmfd sp!, @{fp, ip, lr, pc@}
8489 When performing a stack backtrace, code can inspect the value of
8490 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8491 location @code{pc - 12} and the top 8 bits are set, then we know that
8492 there is a function name embedded immediately preceding this location
8493 and has length @code{((pc[-3]) & 0xff000000)}.
8497 Generate code for the Thumb instruction set. The default is to
8498 use the 32-bit ARM instruction set.
8499 This option automatically enables either 16-bit Thumb-1 or
8500 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8501 and @option{-march=@var{name}} options.
8504 @opindex mtpcs-frame
8505 Generate a stack frame that is compliant with the Thumb Procedure Call
8506 Standard for all non-leaf functions. (A leaf function is one that does
8507 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8509 @item -mtpcs-leaf-frame
8510 @opindex mtpcs-leaf-frame
8511 Generate a stack frame that is compliant with the Thumb Procedure Call
8512 Standard for all leaf functions. (A leaf function is one that does
8513 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8515 @item -mcallee-super-interworking
8516 @opindex mcallee-super-interworking
8517 Gives all externally visible functions in the file being compiled an ARM
8518 instruction set header which switches to Thumb mode before executing the
8519 rest of the function. This allows these functions to be called from
8520 non-interworking code.
8522 @item -mcaller-super-interworking
8523 @opindex mcaller-super-interworking
8524 Allows calls via function pointers (including virtual functions) to
8525 execute correctly regardless of whether the target code has been
8526 compiled for interworking or not. There is a small overhead in the cost
8527 of executing a function pointer if this option is enabled.
8529 @item -mtp=@var{name}
8531 Specify the access model for the thread local storage pointer. The valid
8532 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8533 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8534 (supported in the arm6k architecture), and @option{auto}, which uses the
8535 best available method for the selected processor. The default setting is
8541 @subsection AVR Options
8544 These options are defined for AVR implementations:
8547 @item -mmcu=@var{mcu}
8549 Specify ATMEL AVR instruction set or MCU type.
8551 Instruction set avr1 is for the minimal AVR core, not supported by the C
8552 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8553 attiny11, attiny12, attiny15, attiny28).
8555 Instruction set avr2 (default) is for the classic AVR core with up to
8556 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8557 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8558 at90c8534, at90s8535).
8560 Instruction set avr3 is for the classic AVR core with up to 128K program
8561 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8563 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8564 memory space (MCU types: atmega8, atmega83, atmega85).
8566 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8567 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8568 atmega64, atmega128, at43usb355, at94k).
8572 Output instruction sizes to the asm file.
8574 @item -minit-stack=@var{N}
8575 @opindex minit-stack
8576 Specify the initial stack address, which may be a symbol or numeric value,
8577 @samp{__stack} is the default.
8579 @item -mno-interrupts
8580 @opindex mno-interrupts
8581 Generated code is not compatible with hardware interrupts.
8582 Code size will be smaller.
8584 @item -mcall-prologues
8585 @opindex mcall-prologues
8586 Functions prologues/epilogues expanded as call to appropriate
8587 subroutines. Code size will be smaller.
8589 @item -mno-tablejump
8590 @opindex mno-tablejump
8591 Do not generate tablejump insns which sometimes increase code size.
8594 @opindex mtiny-stack
8595 Change only the low 8 bits of the stack pointer.
8599 Assume int to be 8 bit integer. This affects the sizes of all types: A
8600 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8601 and long long will be 4 bytes. Please note that this option does not
8602 comply to the C standards, but it will provide you with smaller code
8606 @node Blackfin Options
8607 @subsection Blackfin Options
8608 @cindex Blackfin Options
8611 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
8613 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
8614 can be one of @samp{bf522}, @samp{bf525}, @samp{bf527},
8615 @samp{bf531}, @samp{bf532}, @samp{bf533}, @samp{bf534},
8616 @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
8617 @samp{bf542}, @samp{bf544}, @samp{bf548}, @samp{bf549},
8619 The optional @var{sirevision} specifies the silicon revision of the target
8620 Blackfin processor. Any workarounds available for the targeted silicon revision
8621 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
8622 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
8623 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
8624 hexadecimal digits representing the major and minor numbers in the silicon
8625 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
8626 is not defined. If @var{sirevision} is @samp{any}, the
8627 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
8628 If this optional @var{sirevision} is not used, GCC assumes the latest known
8629 silicon revision of the targeted Blackfin processor.
8631 Support for @samp{bf561} is incomplete. For @samp{bf561},
8632 Only the processor macro is defined.
8633 Without this option, @samp{bf532} is used as the processor by default.
8634 The corresponding predefined processor macros for @var{cpu} is to
8635 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
8636 provided by libgloss to be linked in if @option{-msim} is not given.
8640 Specifies that the program will be run on the simulator. This causes
8641 the simulator BSP provided by libgloss to be linked in. This option
8642 has effect only for @samp{bfin-elf} toolchain.
8644 @item -momit-leaf-frame-pointer
8645 @opindex momit-leaf-frame-pointer
8646 Don't keep the frame pointer in a register for leaf functions. This
8647 avoids the instructions to save, set up and restore frame pointers and
8648 makes an extra register available in leaf functions. The option
8649 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8650 which might make debugging harder.
8652 @item -mspecld-anomaly
8653 @opindex mspecld-anomaly
8654 When enabled, the compiler will ensure that the generated code does not
8655 contain speculative loads after jump instructions. If this option is used,
8656 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
8658 @item -mno-specld-anomaly
8659 @opindex mno-specld-anomaly
8660 Don't generate extra code to prevent speculative loads from occurring.
8662 @item -mcsync-anomaly
8663 @opindex mcsync-anomaly
8664 When enabled, the compiler will ensure that the generated code does not
8665 contain CSYNC or SSYNC instructions too soon after conditional branches.
8666 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
8668 @item -mno-csync-anomaly
8669 @opindex mno-csync-anomaly
8670 Don't generate extra code to prevent CSYNC or SSYNC instructions from
8671 occurring too soon after a conditional branch.
8675 When enabled, the compiler is free to take advantage of the knowledge that
8676 the entire program fits into the low 64k of memory.
8679 @opindex mno-low-64k
8680 Assume that the program is arbitrarily large. This is the default.
8682 @item -mstack-check-l1
8683 @opindex mstack-check-l1
8684 Do stack checking using information placed into L1 scratchpad memory by the
8687 @item -mid-shared-library
8688 @opindex mid-shared-library
8689 Generate code that supports shared libraries via the library ID method.
8690 This allows for execute in place and shared libraries in an environment
8691 without virtual memory management. This option implies @option{-fPIC}.
8693 @item -mno-id-shared-library
8694 @opindex mno-id-shared-library
8695 Generate code that doesn't assume ID based shared libraries are being used.
8696 This is the default.
8698 @item -mleaf-id-shared-library
8699 @opindex mleaf-id-shared-library
8700 Generate code that supports shared libraries via the library ID method,
8701 but assumes that this library or executable won't link against any other
8702 ID shared libraries. That allows the compiler to use faster code for jumps
8705 @item -mno-leaf-id-shared-library
8706 @opindex mno-leaf-id-shared-library
8707 Do not assume that the code being compiled won't link against any ID shared
8708 libraries. Slower code will be generated for jump and call insns.
8710 @item -mshared-library-id=n
8711 @opindex mshared-library-id
8712 Specified the identification number of the ID based shared library being
8713 compiled. Specifying a value of 0 will generate more compact code, specifying
8714 other values will force the allocation of that number to the current
8715 library but is no more space or time efficient than omitting this option.
8719 Generate code that allows the data segment to be located in a different
8720 area of memory from the text segment. This allows for execute in place in
8721 an environment without virtual memory management by eliminating relocations
8722 against the text section.
8725 @opindex mno-sep-data
8726 Generate code that assumes that the data segment follows the text segment.
8727 This is the default.
8730 @itemx -mno-long-calls
8731 @opindex mlong-calls
8732 @opindex mno-long-calls
8733 Tells the compiler to perform function calls by first loading the
8734 address of the function into a register and then performing a subroutine
8735 call on this register. This switch is needed if the target function
8736 will lie outside of the 24 bit addressing range of the offset based
8737 version of subroutine call instruction.
8739 This feature is not enabled by default. Specifying
8740 @option{-mno-long-calls} will restore the default behavior. Note these
8741 switches have no effect on how the compiler generates code to handle
8742 function calls via function pointers.
8746 Link with the fast floating-point library. This library relaxes some of
8747 the IEEE floating-point standard's rules for checking inputs against
8748 Not-a-Number (NAN), in the interest of performance.
8751 @opindex minline-plt
8752 Enable inlining of PLT entries in function calls to functions that are
8753 not known to bind locally. It has no effect without @option{-mfdpic}.
8757 @subsection CRIS Options
8758 @cindex CRIS Options
8760 These options are defined specifically for the CRIS ports.
8763 @item -march=@var{architecture-type}
8764 @itemx -mcpu=@var{architecture-type}
8767 Generate code for the specified architecture. The choices for
8768 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
8769 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
8770 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
8773 @item -mtune=@var{architecture-type}
8775 Tune to @var{architecture-type} everything applicable about the generated
8776 code, except for the ABI and the set of available instructions. The
8777 choices for @var{architecture-type} are the same as for
8778 @option{-march=@var{architecture-type}}.
8780 @item -mmax-stack-frame=@var{n}
8781 @opindex mmax-stack-frame
8782 Warn when the stack frame of a function exceeds @var{n} bytes.
8784 @item -melinux-stacksize=@var{n}
8785 @opindex melinux-stacksize
8786 Only available with the @samp{cris-axis-aout} target. Arranges for
8787 indications in the program to the kernel loader that the stack of the
8788 program should be set to @var{n} bytes.
8794 The options @option{-metrax4} and @option{-metrax100} are synonyms for
8795 @option{-march=v3} and @option{-march=v8} respectively.
8797 @item -mmul-bug-workaround
8798 @itemx -mno-mul-bug-workaround
8799 @opindex mmul-bug-workaround
8800 @opindex mno-mul-bug-workaround
8801 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
8802 models where it applies. This option is active by default.
8806 Enable CRIS-specific verbose debug-related information in the assembly
8807 code. This option also has the effect to turn off the @samp{#NO_APP}
8808 formatted-code indicator to the assembler at the beginning of the
8813 Do not use condition-code results from previous instruction; always emit
8814 compare and test instructions before use of condition codes.
8816 @item -mno-side-effects
8817 @opindex mno-side-effects
8818 Do not emit instructions with side-effects in addressing modes other than
8822 @itemx -mno-stack-align
8824 @itemx -mno-data-align
8825 @itemx -mconst-align
8826 @itemx -mno-const-align
8827 @opindex mstack-align
8828 @opindex mno-stack-align
8829 @opindex mdata-align
8830 @opindex mno-data-align
8831 @opindex mconst-align
8832 @opindex mno-const-align
8833 These options (no-options) arranges (eliminate arrangements) for the
8834 stack-frame, individual data and constants to be aligned for the maximum
8835 single data access size for the chosen CPU model. The default is to
8836 arrange for 32-bit alignment. ABI details such as structure layout are
8837 not affected by these options.
8845 Similar to the stack- data- and const-align options above, these options
8846 arrange for stack-frame, writable data and constants to all be 32-bit,
8847 16-bit or 8-bit aligned. The default is 32-bit alignment.
8849 @item -mno-prologue-epilogue
8850 @itemx -mprologue-epilogue
8851 @opindex mno-prologue-epilogue
8852 @opindex mprologue-epilogue
8853 With @option{-mno-prologue-epilogue}, the normal function prologue and
8854 epilogue that sets up the stack-frame are omitted and no return
8855 instructions or return sequences are generated in the code. Use this
8856 option only together with visual inspection of the compiled code: no
8857 warnings or errors are generated when call-saved registers must be saved,
8858 or storage for local variable needs to be allocated.
8864 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
8865 instruction sequences that load addresses for functions from the PLT part
8866 of the GOT rather than (traditional on other architectures) calls to the
8867 PLT@. The default is @option{-mgotplt}.
8871 Legacy no-op option only recognized with the cris-axis-aout target.
8875 Legacy no-op option only recognized with the cris-axis-elf and
8876 cris-axis-linux-gnu targets.
8880 Only recognized with the cris-axis-aout target, where it selects a
8881 GNU/linux-like multilib, include files and instruction set for
8886 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
8890 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
8891 to link with input-output functions from a simulator library. Code,
8892 initialized data and zero-initialized data are allocated consecutively.
8896 Like @option{-sim}, but pass linker options to locate initialized data at
8897 0x40000000 and zero-initialized data at 0x80000000.
8901 @subsection CRX Options
8904 These options are defined specifically for the CRX ports.
8910 Enable the use of multiply-accumulate instructions. Disabled by default.
8914 Push instructions will be used to pass outgoing arguments when functions
8915 are called. Enabled by default.
8918 @node Darwin Options
8919 @subsection Darwin Options
8920 @cindex Darwin options
8922 These options are defined for all architectures running the Darwin operating
8925 FSF GCC on Darwin does not create ``fat'' object files; it will create
8926 an object file for the single architecture that it was built to
8927 target. Apple's GCC on Darwin does create ``fat'' files if multiple
8928 @option{-arch} options are used; it does so by running the compiler or
8929 linker multiple times and joining the results together with
8932 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
8933 @samp{i686}) is determined by the flags that specify the ISA
8934 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
8935 @option{-force_cpusubtype_ALL} option can be used to override this.
8937 The Darwin tools vary in their behavior when presented with an ISA
8938 mismatch. The assembler, @file{as}, will only permit instructions to
8939 be used that are valid for the subtype of the file it is generating,
8940 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
8941 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8942 and print an error if asked to create a shared library with a less
8943 restrictive subtype than its input files (for instance, trying to put
8944 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8945 for executables, @file{ld}, will quietly give the executable the most
8946 restrictive subtype of any of its input files.
8951 Add the framework directory @var{dir} to the head of the list of
8952 directories to be searched for header files. These directories are
8953 interleaved with those specified by @option{-I} options and are
8954 scanned in a left-to-right order.
8956 A framework directory is a directory with frameworks in it. A
8957 framework is a directory with a @samp{"Headers"} and/or
8958 @samp{"PrivateHeaders"} directory contained directly in it that ends
8959 in @samp{".framework"}. The name of a framework is the name of this
8960 directory excluding the @samp{".framework"}. Headers associated with
8961 the framework are found in one of those two directories, with
8962 @samp{"Headers"} being searched first. A subframework is a framework
8963 directory that is in a framework's @samp{"Frameworks"} directory.
8964 Includes of subframework headers can only appear in a header of a
8965 framework that contains the subframework, or in a sibling subframework
8966 header. Two subframeworks are siblings if they occur in the same
8967 framework. A subframework should not have the same name as a
8968 framework, a warning will be issued if this is violated. Currently a
8969 subframework cannot have subframeworks, in the future, the mechanism
8970 may be extended to support this. The standard frameworks can be found
8971 in @samp{"/System/Library/Frameworks"} and
8972 @samp{"/Library/Frameworks"}. An example include looks like
8973 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8974 the name of the framework and header.h is found in the
8975 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8977 @item -iframework@var{dir}
8979 Like @option{-F} except the directory is a treated as a system
8980 directory. The main difference between this @option{-iframework} and
8981 @option{-F} is that with @option{-iframework} the compiler does not
8982 warn about constructs contained within header files found via
8983 @var{dir}. This option is valid only for the C family of languages.
8987 Emit debugging information for symbols that are used. For STABS
8988 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8989 This is by default ON@.
8993 Emit debugging information for all symbols and types.
8995 @item -mmacosx-version-min=@var{version}
8996 The earliest version of MacOS X that this executable will run on
8997 is @var{version}. Typical values of @var{version} include @code{10.1},
8998 @code{10.2}, and @code{10.3.9}.
9000 If the compiler was built to use the system's headers by default,
9001 then the default for this option is the system version on which the
9002 compiler is running, otherwise the default is to make choices which
9003 are compatible with as many systems and code bases as possible.
9007 Enable kernel development mode. The @option{-mkernel} option sets
9008 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9009 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9010 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9011 applicable. This mode also sets @option{-mno-altivec},
9012 @option{-msoft-float}, @option{-fno-builtin} and
9013 @option{-mlong-branch} for PowerPC targets.
9015 @item -mone-byte-bool
9016 @opindex mone-byte-bool
9017 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9018 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9019 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9020 option has no effect on x86.
9022 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9023 to generate code that is not binary compatible with code generated
9024 without that switch. Using this switch may require recompiling all
9025 other modules in a program, including system libraries. Use this
9026 switch to conform to a non-default data model.
9028 @item -mfix-and-continue
9029 @itemx -ffix-and-continue
9030 @itemx -findirect-data
9031 @opindex mfix-and-continue
9032 @opindex ffix-and-continue
9033 @opindex findirect-data
9034 Generate code suitable for fast turn around development. Needed to
9035 enable gdb to dynamically load @code{.o} files into already running
9036 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9037 are provided for backwards compatibility.
9041 Loads all members of static archive libraries.
9042 See man ld(1) for more information.
9044 @item -arch_errors_fatal
9045 @opindex arch_errors_fatal
9046 Cause the errors having to do with files that have the wrong architecture
9050 @opindex bind_at_load
9051 Causes the output file to be marked such that the dynamic linker will
9052 bind all undefined references when the file is loaded or launched.
9056 Produce a Mach-o bundle format file.
9057 See man ld(1) for more information.
9059 @item -bundle_loader @var{executable}
9060 @opindex bundle_loader
9061 This option specifies the @var{executable} that will be loading the build
9062 output file being linked. See man ld(1) for more information.
9066 When passed this option, GCC will produce a dynamic library instead of
9067 an executable when linking, using the Darwin @file{libtool} command.
9069 @item -force_cpusubtype_ALL
9070 @opindex force_cpusubtype_ALL
9071 This causes GCC's output file to have the @var{ALL} subtype, instead of
9072 one controlled by the @option{-mcpu} or @option{-march} option.
9074 @item -allowable_client @var{client_name}
9076 @itemx -compatibility_version
9077 @itemx -current_version
9079 @itemx -dependency-file
9081 @itemx -dylinker_install_name
9083 @itemx -exported_symbols_list
9085 @itemx -flat_namespace
9086 @itemx -force_flat_namespace
9087 @itemx -headerpad_max_install_names
9090 @itemx -install_name
9091 @itemx -keep_private_externs
9092 @itemx -multi_module
9093 @itemx -multiply_defined
9094 @itemx -multiply_defined_unused
9096 @itemx -no_dead_strip_inits_and_terms
9097 @itemx -nofixprebinding
9100 @itemx -noseglinkedit
9101 @itemx -pagezero_size
9103 @itemx -prebind_all_twolevel_modules
9104 @itemx -private_bundle
9105 @itemx -read_only_relocs
9107 @itemx -sectobjectsymbols
9111 @itemx -sectobjectsymbols
9114 @itemx -segs_read_only_addr
9115 @itemx -segs_read_write_addr
9116 @itemx -seg_addr_table
9117 @itemx -seg_addr_table_filename
9120 @itemx -segs_read_only_addr
9121 @itemx -segs_read_write_addr
9122 @itemx -single_module
9125 @itemx -sub_umbrella
9126 @itemx -twolevel_namespace
9129 @itemx -unexported_symbols_list
9130 @itemx -weak_reference_mismatches
9133 @opindex allowable_client
9134 @opindex client_name
9135 @opindex compatibility_version
9136 @opindex current_version
9138 @opindex dependency-file
9140 @opindex dylinker_install_name
9142 @opindex exported_symbols_list
9144 @opindex flat_namespace
9145 @opindex force_flat_namespace
9146 @opindex headerpad_max_install_names
9149 @opindex install_name
9150 @opindex keep_private_externs
9151 @opindex multi_module
9152 @opindex multiply_defined
9153 @opindex multiply_defined_unused
9155 @opindex no_dead_strip_inits_and_terms
9156 @opindex nofixprebinding
9157 @opindex nomultidefs
9159 @opindex noseglinkedit
9160 @opindex pagezero_size
9162 @opindex prebind_all_twolevel_modules
9163 @opindex private_bundle
9164 @opindex read_only_relocs
9166 @opindex sectobjectsymbols
9170 @opindex sectobjectsymbols
9173 @opindex segs_read_only_addr
9174 @opindex segs_read_write_addr
9175 @opindex seg_addr_table
9176 @opindex seg_addr_table_filename
9177 @opindex seglinkedit
9179 @opindex segs_read_only_addr
9180 @opindex segs_read_write_addr
9181 @opindex single_module
9183 @opindex sub_library
9184 @opindex sub_umbrella
9185 @opindex twolevel_namespace
9188 @opindex unexported_symbols_list
9189 @opindex weak_reference_mismatches
9190 @opindex whatsloaded
9192 These options are passed to the Darwin linker. The Darwin linker man page
9193 describes them in detail.
9196 @node DEC Alpha Options
9197 @subsection DEC Alpha Options
9199 These @samp{-m} options are defined for the DEC Alpha implementations:
9202 @item -mno-soft-float
9204 @opindex mno-soft-float
9205 @opindex msoft-float
9206 Use (do not use) the hardware floating-point instructions for
9207 floating-point operations. When @option{-msoft-float} is specified,
9208 functions in @file{libgcc.a} will be used to perform floating-point
9209 operations. Unless they are replaced by routines that emulate the
9210 floating-point operations, or compiled in such a way as to call such
9211 emulations routines, these routines will issue floating-point
9212 operations. If you are compiling for an Alpha without floating-point
9213 operations, you must ensure that the library is built so as not to call
9216 Note that Alpha implementations without floating-point operations are
9217 required to have floating-point registers.
9222 @opindex mno-fp-regs
9223 Generate code that uses (does not use) the floating-point register set.
9224 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9225 register set is not used, floating point operands are passed in integer
9226 registers as if they were integers and floating-point results are passed
9227 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9228 so any function with a floating-point argument or return value called by code
9229 compiled with @option{-mno-fp-regs} must also be compiled with that
9232 A typical use of this option is building a kernel that does not use,
9233 and hence need not save and restore, any floating-point registers.
9237 The Alpha architecture implements floating-point hardware optimized for
9238 maximum performance. It is mostly compliant with the IEEE floating
9239 point standard. However, for full compliance, software assistance is
9240 required. This option generates code fully IEEE compliant code
9241 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9242 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9243 defined during compilation. The resulting code is less efficient but is
9244 able to correctly support denormalized numbers and exceptional IEEE
9245 values such as not-a-number and plus/minus infinity. Other Alpha
9246 compilers call this option @option{-ieee_with_no_inexact}.
9248 @item -mieee-with-inexact
9249 @opindex mieee-with-inexact
9250 This is like @option{-mieee} except the generated code also maintains
9251 the IEEE @var{inexact-flag}. Turning on this option causes the
9252 generated code to implement fully-compliant IEEE math. In addition to
9253 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9254 macro. On some Alpha implementations the resulting code may execute
9255 significantly slower than the code generated by default. Since there is
9256 very little code that depends on the @var{inexact-flag}, you should
9257 normally not specify this option. Other Alpha compilers call this
9258 option @option{-ieee_with_inexact}.
9260 @item -mfp-trap-mode=@var{trap-mode}
9261 @opindex mfp-trap-mode
9262 This option controls what floating-point related traps are enabled.
9263 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9264 The trap mode can be set to one of four values:
9268 This is the default (normal) setting. The only traps that are enabled
9269 are the ones that cannot be disabled in software (e.g., division by zero
9273 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9277 Like @samp{u}, but the instructions are marked to be safe for software
9278 completion (see Alpha architecture manual for details).
9281 Like @samp{su}, but inexact traps are enabled as well.
9284 @item -mfp-rounding-mode=@var{rounding-mode}
9285 @opindex mfp-rounding-mode
9286 Selects the IEEE rounding mode. Other Alpha compilers call this option
9287 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9292 Normal IEEE rounding mode. Floating point numbers are rounded towards
9293 the nearest machine number or towards the even machine number in case
9297 Round towards minus infinity.
9300 Chopped rounding mode. Floating point numbers are rounded towards zero.
9303 Dynamic rounding mode. A field in the floating point control register
9304 (@var{fpcr}, see Alpha architecture reference manual) controls the
9305 rounding mode in effect. The C library initializes this register for
9306 rounding towards plus infinity. Thus, unless your program modifies the
9307 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9310 @item -mtrap-precision=@var{trap-precision}
9311 @opindex mtrap-precision
9312 In the Alpha architecture, floating point traps are imprecise. This
9313 means without software assistance it is impossible to recover from a
9314 floating trap and program execution normally needs to be terminated.
9315 GCC can generate code that can assist operating system trap handlers
9316 in determining the exact location that caused a floating point trap.
9317 Depending on the requirements of an application, different levels of
9318 precisions can be selected:
9322 Program precision. This option is the default and means a trap handler
9323 can only identify which program caused a floating point exception.
9326 Function precision. The trap handler can determine the function that
9327 caused a floating point exception.
9330 Instruction precision. The trap handler can determine the exact
9331 instruction that caused a floating point exception.
9334 Other Alpha compilers provide the equivalent options called
9335 @option{-scope_safe} and @option{-resumption_safe}.
9337 @item -mieee-conformant
9338 @opindex mieee-conformant
9339 This option marks the generated code as IEEE conformant. You must not
9340 use this option unless you also specify @option{-mtrap-precision=i} and either
9341 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9342 is to emit the line @samp{.eflag 48} in the function prologue of the
9343 generated assembly file. Under DEC Unix, this has the effect that
9344 IEEE-conformant math library routines will be linked in.
9346 @item -mbuild-constants
9347 @opindex mbuild-constants
9348 Normally GCC examines a 32- or 64-bit integer constant to
9349 see if it can construct it from smaller constants in two or three
9350 instructions. If it cannot, it will output the constant as a literal and
9351 generate code to load it from the data segment at runtime.
9353 Use this option to require GCC to construct @emph{all} integer constants
9354 using code, even if it takes more instructions (the maximum is six).
9356 You would typically use this option to build a shared library dynamic
9357 loader. Itself a shared library, it must relocate itself in memory
9358 before it can find the variables and constants in its own data segment.
9364 Select whether to generate code to be assembled by the vendor-supplied
9365 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9383 Indicate whether GCC should generate code to use the optional BWX,
9384 CIX, FIX and MAX instruction sets. The default is to use the instruction
9385 sets supported by the CPU type specified via @option{-mcpu=} option or that
9386 of the CPU on which GCC was built if none was specified.
9391 @opindex mfloat-ieee
9392 Generate code that uses (does not use) VAX F and G floating point
9393 arithmetic instead of IEEE single and double precision.
9395 @item -mexplicit-relocs
9396 @itemx -mno-explicit-relocs
9397 @opindex mexplicit-relocs
9398 @opindex mno-explicit-relocs
9399 Older Alpha assemblers provided no way to generate symbol relocations
9400 except via assembler macros. Use of these macros does not allow
9401 optimal instruction scheduling. GNU binutils as of version 2.12
9402 supports a new syntax that allows the compiler to explicitly mark
9403 which relocations should apply to which instructions. This option
9404 is mostly useful for debugging, as GCC detects the capabilities of
9405 the assembler when it is built and sets the default accordingly.
9409 @opindex msmall-data
9410 @opindex mlarge-data
9411 When @option{-mexplicit-relocs} is in effect, static data is
9412 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9413 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9414 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9415 16-bit relocations off of the @code{$gp} register. This limits the
9416 size of the small data area to 64KB, but allows the variables to be
9417 directly accessed via a single instruction.
9419 The default is @option{-mlarge-data}. With this option the data area
9420 is limited to just below 2GB@. Programs that require more than 2GB of
9421 data must use @code{malloc} or @code{mmap} to allocate the data in the
9422 heap instead of in the program's data segment.
9424 When generating code for shared libraries, @option{-fpic} implies
9425 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9429 @opindex msmall-text
9430 @opindex mlarge-text
9431 When @option{-msmall-text} is used, the compiler assumes that the
9432 code of the entire program (or shared library) fits in 4MB, and is
9433 thus reachable with a branch instruction. When @option{-msmall-data}
9434 is used, the compiler can assume that all local symbols share the
9435 same @code{$gp} value, and thus reduce the number of instructions
9436 required for a function call from 4 to 1.
9438 The default is @option{-mlarge-text}.
9440 @item -mcpu=@var{cpu_type}
9442 Set the instruction set and instruction scheduling parameters for
9443 machine type @var{cpu_type}. You can specify either the @samp{EV}
9444 style name or the corresponding chip number. GCC supports scheduling
9445 parameters for the EV4, EV5 and EV6 family of processors and will
9446 choose the default values for the instruction set from the processor
9447 you specify. If you do not specify a processor type, GCC will default
9448 to the processor on which the compiler was built.
9450 Supported values for @var{cpu_type} are
9456 Schedules as an EV4 and has no instruction set extensions.
9460 Schedules as an EV5 and has no instruction set extensions.
9464 Schedules as an EV5 and supports the BWX extension.
9469 Schedules as an EV5 and supports the BWX and MAX extensions.
9473 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9477 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9480 @item -mtune=@var{cpu_type}
9482 Set only the instruction scheduling parameters for machine type
9483 @var{cpu_type}. The instruction set is not changed.
9485 @item -mmemory-latency=@var{time}
9486 @opindex mmemory-latency
9487 Sets the latency the scheduler should assume for typical memory
9488 references as seen by the application. This number is highly
9489 dependent on the memory access patterns used by the application
9490 and the size of the external cache on the machine.
9492 Valid options for @var{time} are
9496 A decimal number representing clock cycles.
9502 The compiler contains estimates of the number of clock cycles for
9503 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9504 (also called Dcache, Scache, and Bcache), as well as to main memory.
9505 Note that L3 is only valid for EV5.
9510 @node DEC Alpha/VMS Options
9511 @subsection DEC Alpha/VMS Options
9513 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9516 @item -mvms-return-codes
9517 @opindex mvms-return-codes
9518 Return VMS condition codes from main. The default is to return POSIX
9519 style condition (e.g.@: error) codes.
9523 @subsection FRV Options
9530 Only use the first 32 general purpose registers.
9535 Use all 64 general purpose registers.
9540 Use only the first 32 floating point registers.
9545 Use all 64 floating point registers
9548 @opindex mhard-float
9550 Use hardware instructions for floating point operations.
9553 @opindex msoft-float
9555 Use library routines for floating point operations.
9560 Dynamically allocate condition code registers.
9565 Do not try to dynamically allocate condition code registers, only
9566 use @code{icc0} and @code{fcc0}.
9571 Change ABI to use double word insns.
9576 Do not use double word instructions.
9581 Use floating point double instructions.
9586 Do not use floating point double instructions.
9591 Use media instructions.
9596 Do not use media instructions.
9601 Use multiply and add/subtract instructions.
9606 Do not use multiply and add/subtract instructions.
9611 Select the FDPIC ABI, that uses function descriptors to represent
9612 pointers to functions. Without any PIC/PIE-related options, it
9613 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9614 assumes GOT entries and small data are within a 12-bit range from the
9615 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9616 are computed with 32 bits.
9619 @opindex minline-plt
9621 Enable inlining of PLT entries in function calls to functions that are
9622 not known to bind locally. It has no effect without @option{-mfdpic}.
9623 It's enabled by default if optimizing for speed and compiling for
9624 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9625 optimization option such as @option{-O3} or above is present in the
9631 Assume a large TLS segment when generating thread-local code.
9636 Do not assume a large TLS segment when generating thread-local code.
9641 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9642 that is known to be in read-only sections. It's enabled by default,
9643 except for @option{-fpic} or @option{-fpie}: even though it may help
9644 make the global offset table smaller, it trades 1 instruction for 4.
9645 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9646 one of which may be shared by multiple symbols, and it avoids the need
9647 for a GOT entry for the referenced symbol, so it's more likely to be a
9648 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
9650 @item -multilib-library-pic
9651 @opindex multilib-library-pic
9653 Link with the (library, not FD) pic libraries. It's implied by
9654 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
9655 @option{-fpic} without @option{-mfdpic}. You should never have to use
9661 Follow the EABI requirement of always creating a frame pointer whenever
9662 a stack frame is allocated. This option is enabled by default and can
9663 be disabled with @option{-mno-linked-fp}.
9666 @opindex mlong-calls
9668 Use indirect addressing to call functions outside the current
9669 compilation unit. This allows the functions to be placed anywhere
9670 within the 32-bit address space.
9672 @item -malign-labels
9673 @opindex malign-labels
9675 Try to align labels to an 8-byte boundary by inserting nops into the
9676 previous packet. This option only has an effect when VLIW packing
9677 is enabled. It doesn't create new packets; it merely adds nops to
9681 @opindex mlibrary-pic
9683 Generate position-independent EABI code.
9688 Use only the first four media accumulator registers.
9693 Use all eight media accumulator registers.
9698 Pack VLIW instructions.
9703 Do not pack VLIW instructions.
9708 Do not mark ABI switches in e_flags.
9713 Enable the use of conditional-move instructions (default).
9715 This switch is mainly for debugging the compiler and will likely be removed
9716 in a future version.
9718 @item -mno-cond-move
9719 @opindex mno-cond-move
9721 Disable the use of conditional-move instructions.
9723 This switch is mainly for debugging the compiler and will likely be removed
9724 in a future version.
9729 Enable the use of conditional set instructions (default).
9731 This switch is mainly for debugging the compiler and will likely be removed
9732 in a future version.
9737 Disable the use of conditional set instructions.
9739 This switch is mainly for debugging the compiler and will likely be removed
9740 in a future version.
9745 Enable the use of conditional execution (default).
9747 This switch is mainly for debugging the compiler and will likely be removed
9748 in a future version.
9750 @item -mno-cond-exec
9751 @opindex mno-cond-exec
9753 Disable the use of conditional execution.
9755 This switch is mainly for debugging the compiler and will likely be removed
9756 in a future version.
9759 @opindex mvliw-branch
9761 Run a pass to pack branches into VLIW instructions (default).
9763 This switch is mainly for debugging the compiler and will likely be removed
9764 in a future version.
9766 @item -mno-vliw-branch
9767 @opindex mno-vliw-branch
9769 Do not run a pass to pack branches into VLIW instructions.
9771 This switch is mainly for debugging the compiler and will likely be removed
9772 in a future version.
9774 @item -mmulti-cond-exec
9775 @opindex mmulti-cond-exec
9777 Enable optimization of @code{&&} and @code{||} in conditional execution
9780 This switch is mainly for debugging the compiler and will likely be removed
9781 in a future version.
9783 @item -mno-multi-cond-exec
9784 @opindex mno-multi-cond-exec
9786 Disable optimization of @code{&&} and @code{||} in conditional execution.
9788 This switch is mainly for debugging the compiler and will likely be removed
9789 in a future version.
9791 @item -mnested-cond-exec
9792 @opindex mnested-cond-exec
9794 Enable nested conditional execution optimizations (default).
9796 This switch is mainly for debugging the compiler and will likely be removed
9797 in a future version.
9799 @item -mno-nested-cond-exec
9800 @opindex mno-nested-cond-exec
9802 Disable nested conditional execution optimizations.
9804 This switch is mainly for debugging the compiler and will likely be removed
9805 in a future version.
9807 @item -moptimize-membar
9808 @opindex moptimize-membar
9810 This switch removes redundant @code{membar} instructions from the
9811 compiler generated code. It is enabled by default.
9813 @item -mno-optimize-membar
9814 @opindex mno-optimize-membar
9816 This switch disables the automatic removal of redundant @code{membar}
9817 instructions from the generated code.
9819 @item -mtomcat-stats
9820 @opindex mtomcat-stats
9822 Cause gas to print out tomcat statistics.
9824 @item -mcpu=@var{cpu}
9827 Select the processor type for which to generate code. Possible values are
9828 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
9829 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
9833 @node GNU/Linux Options
9834 @subsection GNU/Linux Options
9836 These @samp{-m} options are defined for GNU/Linux targets:
9841 Use the GNU C library instead of uClibc. This is the default except
9842 on @samp{*-*-linux-*uclibc*} targets.
9846 Use uClibc instead of the GNU C library. This is the default on
9847 @samp{*-*-linux-*uclibc*} targets.
9850 @node H8/300 Options
9851 @subsection H8/300 Options
9853 These @samp{-m} options are defined for the H8/300 implementations:
9858 Shorten some address references at link time, when possible; uses the
9859 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9860 ld, Using ld}, for a fuller description.
9864 Generate code for the H8/300H@.
9868 Generate code for the H8S@.
9872 Generate code for the H8S and H8/300H in the normal mode. This switch
9873 must be used either with @option{-mh} or @option{-ms}.
9877 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9881 Make @code{int} data 32 bits by default.
9885 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9886 The default for the H8/300H and H8S is to align longs and floats on 4
9888 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9889 This option has no effect on the H8/300.
9893 @subsection HPPA Options
9894 @cindex HPPA Options
9896 These @samp{-m} options are defined for the HPPA family of computers:
9899 @item -march=@var{architecture-type}
9901 Generate code for the specified architecture. The choices for
9902 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
9903 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
9904 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
9905 architecture option for your machine. Code compiled for lower numbered
9906 architectures will run on higher numbered architectures, but not the
9910 @itemx -mpa-risc-1-1
9911 @itemx -mpa-risc-2-0
9912 @opindex mpa-risc-1-0
9913 @opindex mpa-risc-1-1
9914 @opindex mpa-risc-2-0
9915 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
9918 @opindex mbig-switch
9919 Generate code suitable for big switch tables. Use this option only if
9920 the assembler/linker complain about out of range branches within a switch
9923 @item -mjump-in-delay
9924 @opindex mjump-in-delay
9925 Fill delay slots of function calls with unconditional jump instructions
9926 by modifying the return pointer for the function call to be the target
9927 of the conditional jump.
9929 @item -mdisable-fpregs
9930 @opindex mdisable-fpregs
9931 Prevent floating point registers from being used in any manner. This is
9932 necessary for compiling kernels which perform lazy context switching of
9933 floating point registers. If you use this option and attempt to perform
9934 floating point operations, the compiler will abort.
9936 @item -mdisable-indexing
9937 @opindex mdisable-indexing
9938 Prevent the compiler from using indexing address modes. This avoids some
9939 rather obscure problems when compiling MIG generated code under MACH@.
9941 @item -mno-space-regs
9942 @opindex mno-space-regs
9943 Generate code that assumes the target has no space registers. This allows
9944 GCC to generate faster indirect calls and use unscaled index address modes.
9946 Such code is suitable for level 0 PA systems and kernels.
9948 @item -mfast-indirect-calls
9949 @opindex mfast-indirect-calls
9950 Generate code that assumes calls never cross space boundaries. This
9951 allows GCC to emit code which performs faster indirect calls.
9953 This option will not work in the presence of shared libraries or nested
9956 @item -mfixed-range=@var{register-range}
9957 @opindex mfixed-range
9958 Generate code treating the given register range as fixed registers.
9959 A fixed register is one that the register allocator can not use. This is
9960 useful when compiling kernel code. A register range is specified as
9961 two registers separated by a dash. Multiple register ranges can be
9962 specified separated by a comma.
9964 @item -mlong-load-store
9965 @opindex mlong-load-store
9966 Generate 3-instruction load and store sequences as sometimes required by
9967 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9970 @item -mportable-runtime
9971 @opindex mportable-runtime
9972 Use the portable calling conventions proposed by HP for ELF systems.
9976 Enable the use of assembler directives only GAS understands.
9978 @item -mschedule=@var{cpu-type}
9980 Schedule code according to the constraints for the machine type
9981 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9982 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9983 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9984 proper scheduling option for your machine. The default scheduling is
9988 @opindex mlinker-opt
9989 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9990 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9991 linkers in which they give bogus error messages when linking some programs.
9994 @opindex msoft-float
9995 Generate output containing library calls for floating point.
9996 @strong{Warning:} the requisite libraries are not available for all HPPA
9997 targets. Normally the facilities of the machine's usual C compiler are
9998 used, but this cannot be done directly in cross-compilation. You must make
9999 your own arrangements to provide suitable library functions for
10000 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
10001 does provide software floating point support.
10003 @option{-msoft-float} changes the calling convention in the output file;
10004 therefore, it is only useful if you compile @emph{all} of a program with
10005 this option. In particular, you need to compile @file{libgcc.a}, the
10006 library that comes with GCC, with @option{-msoft-float} in order for
10011 Generate the predefine, @code{_SIO}, for server IO@. The default is
10012 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10013 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10014 options are available under HP-UX and HI-UX@.
10018 Use GNU ld specific options. This passes @option{-shared} to ld when
10019 building a shared library. It is the default when GCC is configured,
10020 explicitly or implicitly, with the GNU linker. This option does not
10021 have any affect on which ld is called, it only changes what parameters
10022 are passed to that ld. The ld that is called is determined by the
10023 @option{--with-ld} configure option, GCC's program search path, and
10024 finally by the user's @env{PATH}. The linker used by GCC can be printed
10025 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10026 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
10030 Use HP ld specific options. This passes @option{-b} to ld when building
10031 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10032 links. It is the default when GCC is configured, explicitly or
10033 implicitly, with the HP linker. This option does not have any affect on
10034 which ld is called, it only changes what parameters are passed to that
10035 ld. The ld that is called is determined by the @option{--with-ld}
10036 configure option, GCC's program search path, and finally by the user's
10037 @env{PATH}. The linker used by GCC can be printed using @samp{which
10038 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10039 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
10042 @opindex mno-long-calls
10043 Generate code that uses long call sequences. This ensures that a call
10044 is always able to reach linker generated stubs. The default is to generate
10045 long calls only when the distance from the call site to the beginning
10046 of the function or translation unit, as the case may be, exceeds a
10047 predefined limit set by the branch type being used. The limits for
10048 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10049 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10052 Distances are measured from the beginning of functions when using the
10053 @option{-ffunction-sections} option, or when using the @option{-mgas}
10054 and @option{-mno-portable-runtime} options together under HP-UX with
10057 It is normally not desirable to use this option as it will degrade
10058 performance. However, it may be useful in large applications,
10059 particularly when partial linking is used to build the application.
10061 The types of long calls used depends on the capabilities of the
10062 assembler and linker, and the type of code being generated. The
10063 impact on systems that support long absolute calls, and long pic
10064 symbol-difference or pc-relative calls should be relatively small.
10065 However, an indirect call is used on 32-bit ELF systems in pic code
10066 and it is quite long.
10068 @item -munix=@var{unix-std}
10070 Generate compiler predefines and select a startfile for the specified
10071 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10072 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10073 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10074 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10075 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10078 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10079 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10080 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10081 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10082 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10083 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10085 It is @emph{important} to note that this option changes the interfaces
10086 for various library routines. It also affects the operational behavior
10087 of the C library. Thus, @emph{extreme} care is needed in using this
10090 Library code that is intended to operate with more than one UNIX
10091 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10092 as appropriate. Most GNU software doesn't provide this capability.
10096 Suppress the generation of link options to search libdld.sl when the
10097 @option{-static} option is specified on HP-UX 10 and later.
10101 The HP-UX implementation of setlocale in libc has a dependency on
10102 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10103 when the @option{-static} option is specified, special link options
10104 are needed to resolve this dependency.
10106 On HP-UX 10 and later, the GCC driver adds the necessary options to
10107 link with libdld.sl when the @option{-static} option is specified.
10108 This causes the resulting binary to be dynamic. On the 64-bit port,
10109 the linkers generate dynamic binaries by default in any case. The
10110 @option{-nolibdld} option can be used to prevent the GCC driver from
10111 adding these link options.
10115 Add support for multithreading with the @dfn{dce thread} library
10116 under HP-UX@. This option sets flags for both the preprocessor and
10120 @node i386 and x86-64 Options
10121 @subsection Intel 386 and AMD x86-64 Options
10122 @cindex i386 Options
10123 @cindex x86-64 Options
10124 @cindex Intel 386 Options
10125 @cindex AMD x86-64 Options
10127 These @samp{-m} options are defined for the i386 and x86-64 family of
10131 @item -mtune=@var{cpu-type}
10133 Tune to @var{cpu-type} everything applicable about the generated code, except
10134 for the ABI and the set of available instructions. The choices for
10135 @var{cpu-type} are:
10138 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10139 If you know the CPU on which your code will run, then you should use
10140 the corresponding @option{-mtune} option instead of
10141 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10142 of your application will have, then you should use this option.
10144 As new processors are deployed in the marketplace, the behavior of this
10145 option will change. Therefore, if you upgrade to a newer version of
10146 GCC, the code generated option will change to reflect the processors
10147 that were most common when that version of GCC was released.
10149 There is no @option{-march=generic} option because @option{-march}
10150 indicates the instruction set the compiler can use, and there is no
10151 generic instruction set applicable to all processors. In contrast,
10152 @option{-mtune} indicates the processor (or, in this case, collection of
10153 processors) for which the code is optimized.
10155 This selects the CPU to tune for at compilation time by determining
10156 the processor type of the compiling machine. Using @option{-mtune=native}
10157 will produce code optimized for the local machine under the constraints
10158 of the selected instruction set. Using @option{-march=native} will
10159 enable all instruction subsets supported by the local machine (hence
10160 the result might not run on different machines).
10162 Original Intel's i386 CPU@.
10164 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10165 @item i586, pentium
10166 Intel Pentium CPU with no MMX support.
10168 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10170 Intel PentiumPro CPU@.
10172 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10173 instruction set will be used, so the code will run on all i686 family chips.
10175 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10176 @item pentium3, pentium3m
10177 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10180 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10181 support. Used by Centrino notebooks.
10182 @item pentium4, pentium4m
10183 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10185 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10188 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10189 SSE2 and SSE3 instruction set support.
10191 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10192 instruction set support.
10194 AMD K6 CPU with MMX instruction set support.
10196 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
10197 @item athlon, athlon-tbird
10198 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
10200 @item athlon-4, athlon-xp, athlon-mp
10201 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
10202 instruction set support.
10203 @item k8, opteron, athlon64, athlon-fx
10204 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10205 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
10206 @item k8-sse3, opteron-sse3, athlon64-sse3
10207 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10208 @item amdfam10, barcelona
10209 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10210 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10211 instruction set extensions.)
10213 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10216 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
10217 instruction set support.
10219 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
10220 implemented for this chip.)
10222 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10223 implemented for this chip.)
10225 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10228 While picking a specific @var{cpu-type} will schedule things appropriately
10229 for that particular chip, the compiler will not generate any code that
10230 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10233 @item -march=@var{cpu-type}
10235 Generate instructions for the machine type @var{cpu-type}. The choices
10236 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10237 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10239 @item -mcpu=@var{cpu-type}
10241 A deprecated synonym for @option{-mtune}.
10243 @item -mfpmath=@var{unit}
10245 Generate floating point arithmetics for selected unit @var{unit}. The choices
10246 for @var{unit} are:
10250 Use the standard 387 floating point coprocessor present majority of chips and
10251 emulated otherwise. Code compiled with this option will run almost everywhere.
10252 The temporary results are computed in 80bit precision instead of precision
10253 specified by the type resulting in slightly different results compared to most
10254 of other chips. See @option{-ffloat-store} for more detailed description.
10256 This is the default choice for i386 compiler.
10259 Use scalar floating point instructions present in the SSE instruction set.
10260 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10261 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10262 instruction set supports only single precision arithmetics, thus the double and
10263 extended precision arithmetics is still done using 387. Later version, present
10264 only in Pentium4 and the future AMD x86-64 chips supports double precision
10267 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10268 or @option{-msse2} switches to enable SSE extensions and make this option
10269 effective. For the x86-64 compiler, these extensions are enabled by default.
10271 The resulting code should be considerably faster in the majority of cases and avoid
10272 the numerical instability problems of 387 code, but may break some existing
10273 code that expects temporaries to be 80bit.
10275 This is the default choice for the x86-64 compiler.
10278 Attempt to utilize both instruction sets at once. This effectively double the
10279 amount of available registers and on chips with separate execution units for
10280 387 and SSE the execution resources too. Use this option with care, as it is
10281 still experimental, because the GCC register allocator does not model separate
10282 functional units well resulting in instable performance.
10285 @item -masm=@var{dialect}
10286 @opindex masm=@var{dialect}
10287 Output asm instructions using selected @var{dialect}. Supported
10288 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10289 not support @samp{intel}.
10292 @itemx -mno-ieee-fp
10294 @opindex mno-ieee-fp
10295 Control whether or not the compiler uses IEEE floating point
10296 comparisons. These handle correctly the case where the result of a
10297 comparison is unordered.
10300 @opindex msoft-float
10301 Generate output containing library calls for floating point.
10302 @strong{Warning:} the requisite libraries are not part of GCC@.
10303 Normally the facilities of the machine's usual C compiler are used, but
10304 this can't be done directly in cross-compilation. You must make your
10305 own arrangements to provide suitable library functions for
10308 On machines where a function returns floating point results in the 80387
10309 register stack, some floating point opcodes may be emitted even if
10310 @option{-msoft-float} is used.
10312 @item -mno-fp-ret-in-387
10313 @opindex mno-fp-ret-in-387
10314 Do not use the FPU registers for return values of functions.
10316 The usual calling convention has functions return values of types
10317 @code{float} and @code{double} in an FPU register, even if there
10318 is no FPU@. The idea is that the operating system should emulate
10321 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10322 in ordinary CPU registers instead.
10324 @item -mno-fancy-math-387
10325 @opindex mno-fancy-math-387
10326 Some 387 emulators do not support the @code{sin}, @code{cos} and
10327 @code{sqrt} instructions for the 387. Specify this option to avoid
10328 generating those instructions. This option is the default on FreeBSD,
10329 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10330 indicates that the target cpu will always have an FPU and so the
10331 instruction will not need emulation. As of revision 2.6.1, these
10332 instructions are not generated unless you also use the
10333 @option{-funsafe-math-optimizations} switch.
10335 @item -malign-double
10336 @itemx -mno-align-double
10337 @opindex malign-double
10338 @opindex mno-align-double
10339 Control whether GCC aligns @code{double}, @code{long double}, and
10340 @code{long long} variables on a two word boundary or a one word
10341 boundary. Aligning @code{double} variables on a two word boundary will
10342 produce code that runs somewhat faster on a @samp{Pentium} at the
10343 expense of more memory.
10345 On x86-64, @option{-malign-double} is enabled by default.
10347 @strong{Warning:} if you use the @option{-malign-double} switch,
10348 structures containing the above types will be aligned differently than
10349 the published application binary interface specifications for the 386
10350 and will not be binary compatible with structures in code compiled
10351 without that switch.
10353 @item -m96bit-long-double
10354 @itemx -m128bit-long-double
10355 @opindex m96bit-long-double
10356 @opindex m128bit-long-double
10357 These switches control the size of @code{long double} type. The i386
10358 application binary interface specifies the size to be 96 bits,
10359 so @option{-m96bit-long-double} is the default in 32 bit mode.
10361 Modern architectures (Pentium and newer) would prefer @code{long double}
10362 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10363 conforming to the ABI, this would not be possible. So specifying a
10364 @option{-m128bit-long-double} will align @code{long double}
10365 to a 16 byte boundary by padding the @code{long double} with an additional
10368 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10369 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10371 Notice that neither of these options enable any extra precision over the x87
10372 standard of 80 bits for a @code{long double}.
10374 @strong{Warning:} if you override the default value for your target ABI, the
10375 structures and arrays containing @code{long double} variables will change
10376 their size as well as function calling convention for function taking
10377 @code{long double} will be modified. Hence they will not be binary
10378 compatible with arrays or structures in code compiled without that switch.
10380 @item -mmlarge-data-threshold=@var{number}
10381 @opindex mlarge-data-threshold=@var{number}
10382 When @option{-mcmodel=medium} is specified, the data greater than
10383 @var{threshold} are placed in large data section. This value must be the
10384 same across all object linked into the binary and defaults to 65535.
10388 Use a different function-calling convention, in which functions that
10389 take a fixed number of arguments return with the @code{ret} @var{num}
10390 instruction, which pops their arguments while returning. This saves one
10391 instruction in the caller since there is no need to pop the arguments
10394 You can specify that an individual function is called with this calling
10395 sequence with the function attribute @samp{stdcall}. You can also
10396 override the @option{-mrtd} option by using the function attribute
10397 @samp{cdecl}. @xref{Function Attributes}.
10399 @strong{Warning:} this calling convention is incompatible with the one
10400 normally used on Unix, so you cannot use it if you need to call
10401 libraries compiled with the Unix compiler.
10403 Also, you must provide function prototypes for all functions that
10404 take variable numbers of arguments (including @code{printf});
10405 otherwise incorrect code will be generated for calls to those
10408 In addition, seriously incorrect code will result if you call a
10409 function with too many arguments. (Normally, extra arguments are
10410 harmlessly ignored.)
10412 @item -mregparm=@var{num}
10414 Control how many registers are used to pass integer arguments. By
10415 default, no registers are used to pass arguments, and at most 3
10416 registers can be used. You can control this behavior for a specific
10417 function by using the function attribute @samp{regparm}.
10418 @xref{Function Attributes}.
10420 @strong{Warning:} if you use this switch, and
10421 @var{num} is nonzero, then you must build all modules with the same
10422 value, including any libraries. This includes the system libraries and
10426 @opindex msseregparm
10427 Use SSE register passing conventions for float and double arguments
10428 and return values. You can control this behavior for a specific
10429 function by using the function attribute @samp{sseregparm}.
10430 @xref{Function Attributes}.
10432 @strong{Warning:} if you use this switch then you must build all
10433 modules with the same value, including any libraries. This includes
10434 the system libraries and startup modules.
10443 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10444 is specified, the significands of results of floating-point operations are
10445 rounded to 24 bits (single precision); @option{-mpc64} rounds the the
10446 significands of results of floating-point operations to 53 bits (double
10447 precision) and @option{-mpc80} rounds the significands of results of
10448 floating-point operations to 64 bits (extended double precision), which is
10449 the default. When this option is used, floating-point operations in higher
10450 precisions are not available to the programmer without setting the FPU
10451 control word explicitly.
10453 Setting the rounding of floating-point operations to less than the default
10454 80 bits can speed some programs by 2% or more. Note that some mathematical
10455 libraries assume that extended precision (80 bit) floating-point operations
10456 are enabled by default; routines in such libraries could suffer significant
10457 loss of accuracy, typically through so-called "catastrophic cancellation",
10458 when this option is used to set the precision to less than extended precision.
10460 @item -mstackrealign
10461 @opindex mstackrealign
10462 Realign the stack at entry. On the Intel x86, the
10463 @option{-mstackrealign} option will generate an alternate prologue and
10464 epilogue that realigns the runtime stack. This supports mixing legacy
10465 codes that keep a 4-byte aligned stack with modern codes that keep a
10466 16-byte stack for SSE compatibility. The alternate prologue and
10467 epilogue are slower and bigger than the regular ones, and the
10468 alternate prologue requires an extra scratch register; this lowers the
10469 number of registers available if used in conjunction with the
10470 @code{regparm} attribute. The @option{-mstackrealign} option is
10471 incompatible with the nested function prologue; this is considered a
10472 hard error. See also the attribute @code{force_align_arg_pointer},
10473 applicable to individual functions.
10475 @item -mpreferred-stack-boundary=@var{num}
10476 @opindex mpreferred-stack-boundary
10477 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
10478 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
10479 the default is 4 (16 bytes or 128 bits).
10481 On Pentium and PentiumPro, @code{double} and @code{long double} values
10482 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
10483 suffer significant run time performance penalties. On Pentium III, the
10484 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
10485 properly if it is not 16 byte aligned.
10487 To ensure proper alignment of this values on the stack, the stack boundary
10488 must be as aligned as that required by any value stored on the stack.
10489 Further, every function must be generated such that it keeps the stack
10490 aligned. Thus calling a function compiled with a higher preferred
10491 stack boundary from a function compiled with a lower preferred stack
10492 boundary will most likely misalign the stack. It is recommended that
10493 libraries that use callbacks always use the default setting.
10495 This extra alignment does consume extra stack space, and generally
10496 increases code size. Code that is sensitive to stack space usage, such
10497 as embedded systems and operating system kernels, may want to reduce the
10498 preferred alignment to @option{-mpreferred-stack-boundary=2}.
10532 These switches enable or disable the use of instructions in the MMX,
10533 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4A, SSE5, ABM or 3DNow! extended
10535 These extensions are also available as built-in functions: see
10536 @ref{X86 Built-in Functions}, for details of the functions enabled and
10537 disabled by these switches.
10539 To have SSE/SSE2 instructions generated automatically from floating-point
10540 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
10542 These options will enable GCC to use these extended instructions in
10543 generated code, even without @option{-mfpmath=sse}. Applications which
10544 perform runtime CPU detection must compile separate files for each
10545 supported architecture, using the appropriate flags. In particular,
10546 the file containing the CPU detection code should be compiled without
10551 This option will enable GCC to use CMPXCHG16B instruction in generated code.
10552 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
10553 data types. This is useful for high resolution counters that could be updated
10554 by multiple processors (or cores). This instruction is generated as part of
10555 atomic built-in functions: see @ref{Atomic Builtins} for details.
10559 This option will enable GCC to use SAHF instruction in generated 64-bit code.
10560 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
10561 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
10562 SAHF are load and store instructions, respectively, for certain status flags.
10563 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
10564 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
10568 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
10569 vectorized variants RCPPS and RSQRTPS) instead of DIVSS and SQRTSS (and their
10570 vectorized variants). These instructions will be generated only when
10571 @option{-funsafe-math-optimizations} is enabled.
10573 @item -mveclibabi=@var{type}
10574 @opindex mveclibabi
10575 Specifies the ABI type to use for vectorizing intrinsics using an
10576 external library. Supported types are @code{acml} for the AMD
10577 math core library style of interfacing. GCC will currently emit
10578 calls to @code{__vrd2_sin}, @code{__vrd2_cos}, @code{__vrd2_exp},
10579 @code{__vrd2_log}, @code{__vrd2_log2}, @code{__vrd2_log10},
10580 @code{__vrs4_sinf}, @code{__vrs4_cosf}, @code{__vrs4_expf},
10581 @code{__vrs4_logf}, @code{__vrs4_log2f}, @code{__vrs4_log10f}
10582 and @code{__vrs4_powf} when using this type and @option{-ftree-vectorize}
10583 is enabled. A ACML ABI compatible library will have to be specified
10587 @itemx -mno-push-args
10588 @opindex mpush-args
10589 @opindex mno-push-args
10590 Use PUSH operations to store outgoing parameters. This method is shorter
10591 and usually equally fast as method using SUB/MOV operations and is enabled
10592 by default. In some cases disabling it may improve performance because of
10593 improved scheduling and reduced dependencies.
10595 @item -maccumulate-outgoing-args
10596 @opindex maccumulate-outgoing-args
10597 If enabled, the maximum amount of space required for outgoing arguments will be
10598 computed in the function prologue. This is faster on most modern CPUs
10599 because of reduced dependencies, improved scheduling and reduced stack usage
10600 when preferred stack boundary is not equal to 2. The drawback is a notable
10601 increase in code size. This switch implies @option{-mno-push-args}.
10605 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
10606 on thread-safe exception handling must compile and link all code with the
10607 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10608 @option{-D_MT}; when linking, it links in a special thread helper library
10609 @option{-lmingwthrd} which cleans up per thread exception handling data.
10611 @item -mno-align-stringops
10612 @opindex mno-align-stringops
10613 Do not align destination of inlined string operations. This switch reduces
10614 code size and improves performance in case the destination is already aligned,
10615 but GCC doesn't know about it.
10617 @item -minline-all-stringops
10618 @opindex minline-all-stringops
10619 By default GCC inlines string operations only when destination is known to be
10620 aligned at least to 4 byte boundary. This enables more inlining, increase code
10621 size, but may improve performance of code that depends on fast memcpy, strlen
10622 and memset for short lengths.
10624 @item -minline-stringops-dynamically
10625 @opindex minline-stringops-dynamically
10626 For string operation of unknown size, inline runtime checks so for small
10627 blocks inline code is used, while for large blocks library call is used.
10629 @item -mstringop-strategy=@var{alg}
10630 @opindex mstringop-strategy=@var{alg}
10631 Overwrite internal decision heuristic about particular algorithm to inline
10632 string operation with. The allowed values are @code{rep_byte},
10633 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
10634 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
10635 expanding inline loop, @code{libcall} for always expanding library call.
10637 @item -momit-leaf-frame-pointer
10638 @opindex momit-leaf-frame-pointer
10639 Don't keep the frame pointer in a register for leaf functions. This
10640 avoids the instructions to save, set up and restore frame pointers and
10641 makes an extra register available in leaf functions. The option
10642 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10643 which might make debugging harder.
10645 @item -mtls-direct-seg-refs
10646 @itemx -mno-tls-direct-seg-refs
10647 @opindex mtls-direct-seg-refs
10648 Controls whether TLS variables may be accessed with offsets from the
10649 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
10650 or whether the thread base pointer must be added. Whether or not this
10651 is legal depends on the operating system, and whether it maps the
10652 segment to cover the entire TLS area.
10654 For systems that use GNU libc, the default is on.
10657 @itemx -mno-fused-madd
10658 @opindex mfused-madd
10659 Enable automatic generation of fused floating point multiply-add instructions
10660 if the ISA supports such instructions. The -mfused-madd option is on by
10661 default. The fused multiply-add instructions have a different
10662 rounding behavior compared to executing a multiply followed by an add.
10665 These @samp{-m} switches are supported in addition to the above
10666 on AMD x86-64 processors in 64-bit environments.
10673 Generate code for a 32-bit or 64-bit environment.
10674 The 32-bit environment sets int, long and pointer to 32 bits and
10675 generates code that runs on any i386 system.
10676 The 64-bit environment sets int to 32 bits and long and pointer
10677 to 64 bits and generates code for AMD's x86-64 architecture. For
10678 darwin only the -m64 option turns off the @option{-fno-pic} and
10679 @option{-mdynamic-no-pic} options.
10681 @item -mno-red-zone
10682 @opindex no-red-zone
10683 Do not use a so called red zone for x86-64 code. The red zone is mandated
10684 by the x86-64 ABI, it is a 128-byte area beyond the location of the
10685 stack pointer that will not be modified by signal or interrupt handlers
10686 and therefore can be used for temporary data without adjusting the stack
10687 pointer. The flag @option{-mno-red-zone} disables this red zone.
10689 @item -mcmodel=small
10690 @opindex mcmodel=small
10691 Generate code for the small code model: the program and its symbols must
10692 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
10693 Programs can be statically or dynamically linked. This is the default
10696 @item -mcmodel=kernel
10697 @opindex mcmodel=kernel
10698 Generate code for the kernel code model. The kernel runs in the
10699 negative 2 GB of the address space.
10700 This model has to be used for Linux kernel code.
10702 @item -mcmodel=medium
10703 @opindex mcmodel=medium
10704 Generate code for the medium model: The program is linked in the lower 2
10705 GB of the address space but symbols can be located anywhere in the
10706 address space. Programs can be statically or dynamically linked, but
10707 building of shared libraries are not supported with the medium model.
10709 @item -mcmodel=large
10710 @opindex mcmodel=large
10711 Generate code for the large model: This model makes no assumptions
10712 about addresses and sizes of sections.
10715 @node IA-64 Options
10716 @subsection IA-64 Options
10717 @cindex IA-64 Options
10719 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10723 @opindex mbig-endian
10724 Generate code for a big endian target. This is the default for HP-UX@.
10726 @item -mlittle-endian
10727 @opindex mlittle-endian
10728 Generate code for a little endian target. This is the default for AIX5
10734 @opindex mno-gnu-as
10735 Generate (or don't) code for the GNU assembler. This is the default.
10736 @c Also, this is the default if the configure option @option{--with-gnu-as}
10742 @opindex mno-gnu-ld
10743 Generate (or don't) code for the GNU linker. This is the default.
10744 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10749 Generate code that does not use a global pointer register. The result
10750 is not position independent code, and violates the IA-64 ABI@.
10752 @item -mvolatile-asm-stop
10753 @itemx -mno-volatile-asm-stop
10754 @opindex mvolatile-asm-stop
10755 @opindex mno-volatile-asm-stop
10756 Generate (or don't) a stop bit immediately before and after volatile asm
10759 @item -mregister-names
10760 @itemx -mno-register-names
10761 @opindex mregister-names
10762 @opindex mno-register-names
10763 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10764 the stacked registers. This may make assembler output more readable.
10770 Disable (or enable) optimizations that use the small data section. This may
10771 be useful for working around optimizer bugs.
10773 @item -mconstant-gp
10774 @opindex mconstant-gp
10775 Generate code that uses a single constant global pointer value. This is
10776 useful when compiling kernel code.
10780 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10781 This is useful when compiling firmware code.
10783 @item -minline-float-divide-min-latency
10784 @opindex minline-float-divide-min-latency
10785 Generate code for inline divides of floating point values
10786 using the minimum latency algorithm.
10788 @item -minline-float-divide-max-throughput
10789 @opindex minline-float-divide-max-throughput
10790 Generate code for inline divides of floating point values
10791 using the maximum throughput algorithm.
10793 @item -minline-int-divide-min-latency
10794 @opindex minline-int-divide-min-latency
10795 Generate code for inline divides of integer values
10796 using the minimum latency algorithm.
10798 @item -minline-int-divide-max-throughput
10799 @opindex minline-int-divide-max-throughput
10800 Generate code for inline divides of integer values
10801 using the maximum throughput algorithm.
10803 @item -minline-sqrt-min-latency
10804 @opindex minline-sqrt-min-latency
10805 Generate code for inline square roots
10806 using the minimum latency algorithm.
10808 @item -minline-sqrt-max-throughput
10809 @opindex minline-sqrt-max-throughput
10810 Generate code for inline square roots
10811 using the maximum throughput algorithm.
10813 @item -mno-dwarf2-asm
10814 @itemx -mdwarf2-asm
10815 @opindex mno-dwarf2-asm
10816 @opindex mdwarf2-asm
10817 Don't (or do) generate assembler code for the DWARF2 line number debugging
10818 info. This may be useful when not using the GNU assembler.
10820 @item -mearly-stop-bits
10821 @itemx -mno-early-stop-bits
10822 @opindex mearly-stop-bits
10823 @opindex mno-early-stop-bits
10824 Allow stop bits to be placed earlier than immediately preceding the
10825 instruction that triggered the stop bit. This can improve instruction
10826 scheduling, but does not always do so.
10828 @item -mfixed-range=@var{register-range}
10829 @opindex mfixed-range
10830 Generate code treating the given register range as fixed registers.
10831 A fixed register is one that the register allocator can not use. This is
10832 useful when compiling kernel code. A register range is specified as
10833 two registers separated by a dash. Multiple register ranges can be
10834 specified separated by a comma.
10836 @item -mtls-size=@var{tls-size}
10838 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
10841 @item -mtune=@var{cpu-type}
10843 Tune the instruction scheduling for a particular CPU, Valid values are
10844 itanium, itanium1, merced, itanium2, and mckinley.
10850 Add support for multithreading using the POSIX threads library. This
10851 option sets flags for both the preprocessor and linker. It does
10852 not affect the thread safety of object code produced by the compiler or
10853 that of libraries supplied with it. These are HP-UX specific flags.
10859 Generate code for a 32-bit or 64-bit environment.
10860 The 32-bit environment sets int, long and pointer to 32 bits.
10861 The 64-bit environment sets int to 32 bits and long and pointer
10862 to 64 bits. These are HP-UX specific flags.
10864 @item -mno-sched-br-data-spec
10865 @itemx -msched-br-data-spec
10866 @opindex mno-sched-br-data-spec
10867 @opindex msched-br-data-spec
10868 (Dis/En)able data speculative scheduling before reload.
10869 This will result in generation of the ld.a instructions and
10870 the corresponding check instructions (ld.c / chk.a).
10871 The default is 'disable'.
10873 @item -msched-ar-data-spec
10874 @itemx -mno-sched-ar-data-spec
10875 @opindex msched-ar-data-spec
10876 @opindex mno-sched-ar-data-spec
10877 (En/Dis)able data speculative scheduling after reload.
10878 This will result in generation of the ld.a instructions and
10879 the corresponding check instructions (ld.c / chk.a).
10880 The default is 'enable'.
10882 @item -mno-sched-control-spec
10883 @itemx -msched-control-spec
10884 @opindex mno-sched-control-spec
10885 @opindex msched-control-spec
10886 (Dis/En)able control speculative scheduling. This feature is
10887 available only during region scheduling (i.e. before reload).
10888 This will result in generation of the ld.s instructions and
10889 the corresponding check instructions chk.s .
10890 The default is 'disable'.
10892 @item -msched-br-in-data-spec
10893 @itemx -mno-sched-br-in-data-spec
10894 @opindex msched-br-in-data-spec
10895 @opindex mno-sched-br-in-data-spec
10896 (En/Dis)able speculative scheduling of the instructions that
10897 are dependent on the data speculative loads before reload.
10898 This is effective only with @option{-msched-br-data-spec} enabled.
10899 The default is 'enable'.
10901 @item -msched-ar-in-data-spec
10902 @itemx -mno-sched-ar-in-data-spec
10903 @opindex msched-ar-in-data-spec
10904 @opindex mno-sched-ar-in-data-spec
10905 (En/Dis)able speculative scheduling of the instructions that
10906 are dependent on the data speculative loads after reload.
10907 This is effective only with @option{-msched-ar-data-spec} enabled.
10908 The default is 'enable'.
10910 @item -msched-in-control-spec
10911 @itemx -mno-sched-in-control-spec
10912 @opindex msched-in-control-spec
10913 @opindex mno-sched-in-control-spec
10914 (En/Dis)able speculative scheduling of the instructions that
10915 are dependent on the control speculative loads.
10916 This is effective only with @option{-msched-control-spec} enabled.
10917 The default is 'enable'.
10920 @itemx -mno-sched-ldc
10921 @opindex msched-ldc
10922 @opindex mno-sched-ldc
10923 (En/Dis)able use of simple data speculation checks ld.c .
10924 If disabled, only chk.a instructions will be emitted to check
10925 data speculative loads.
10926 The default is 'enable'.
10928 @item -mno-sched-control-ldc
10929 @itemx -msched-control-ldc
10930 @opindex mno-sched-control-ldc
10931 @opindex msched-control-ldc
10932 (Dis/En)able use of ld.c instructions to check control speculative loads.
10933 If enabled, in case of control speculative load with no speculatively
10934 scheduled dependent instructions this load will be emitted as ld.sa and
10935 ld.c will be used to check it.
10936 The default is 'disable'.
10938 @item -mno-sched-spec-verbose
10939 @itemx -msched-spec-verbose
10940 @opindex mno-sched-spec-verbose
10941 @opindex msched-spec-verbose
10942 (Dis/En)able printing of the information about speculative motions.
10944 @item -mno-sched-prefer-non-data-spec-insns
10945 @itemx -msched-prefer-non-data-spec-insns
10946 @opindex mno-sched-prefer-non-data-spec-insns
10947 @opindex msched-prefer-non-data-spec-insns
10948 If enabled, data speculative instructions will be chosen for schedule
10949 only if there are no other choices at the moment. This will make
10950 the use of the data speculation much more conservative.
10951 The default is 'disable'.
10953 @item -mno-sched-prefer-non-control-spec-insns
10954 @itemx -msched-prefer-non-control-spec-insns
10955 @opindex mno-sched-prefer-non-control-spec-insns
10956 @opindex msched-prefer-non-control-spec-insns
10957 If enabled, control speculative instructions will be chosen for schedule
10958 only if there are no other choices at the moment. This will make
10959 the use of the control speculation much more conservative.
10960 The default is 'disable'.
10962 @item -mno-sched-count-spec-in-critical-path
10963 @itemx -msched-count-spec-in-critical-path
10964 @opindex mno-sched-count-spec-in-critical-path
10965 @opindex msched-count-spec-in-critical-path
10966 If enabled, speculative dependencies will be considered during
10967 computation of the instructions priorities. This will make the use of the
10968 speculation a bit more conservative.
10969 The default is 'disable'.
10974 @subsection M32C Options
10975 @cindex M32C options
10978 @item -mcpu=@var{name}
10980 Select the CPU for which code is generated. @var{name} may be one of
10981 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
10982 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
10983 the M32C/80 series.
10987 Specifies that the program will be run on the simulator. This causes
10988 an alternate runtime library to be linked in which supports, for
10989 example, file I/O. You must not use this option when generating
10990 programs that will run on real hardware; you must provide your own
10991 runtime library for whatever I/O functions are needed.
10993 @item -memregs=@var{number}
10995 Specifies the number of memory-based pseudo-registers GCC will use
10996 during code generation. These pseudo-registers will be used like real
10997 registers, so there is a tradeoff between GCC's ability to fit the
10998 code into available registers, and the performance penalty of using
10999 memory instead of registers. Note that all modules in a program must
11000 be compiled with the same value for this option. Because of that, you
11001 must not use this option with the default runtime libraries gcc
11006 @node M32R/D Options
11007 @subsection M32R/D Options
11008 @cindex M32R/D options
11010 These @option{-m} options are defined for Renesas M32R/D architectures:
11015 Generate code for the M32R/2@.
11019 Generate code for the M32R/X@.
11023 Generate code for the M32R@. This is the default.
11025 @item -mmodel=small
11026 @opindex mmodel=small
11027 Assume all objects live in the lower 16MB of memory (so that their addresses
11028 can be loaded with the @code{ld24} instruction), and assume all subroutines
11029 are reachable with the @code{bl} instruction.
11030 This is the default.
11032 The addressability of a particular object can be set with the
11033 @code{model} attribute.
11035 @item -mmodel=medium
11036 @opindex mmodel=medium
11037 Assume objects may be anywhere in the 32-bit address space (the compiler
11038 will generate @code{seth/add3} instructions to load their addresses), and
11039 assume all subroutines are reachable with the @code{bl} instruction.
11041 @item -mmodel=large
11042 @opindex mmodel=large
11043 Assume objects may be anywhere in the 32-bit address space (the compiler
11044 will generate @code{seth/add3} instructions to load their addresses), and
11045 assume subroutines may not be reachable with the @code{bl} instruction
11046 (the compiler will generate the much slower @code{seth/add3/jl}
11047 instruction sequence).
11050 @opindex msdata=none
11051 Disable use of the small data area. Variables will be put into
11052 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11053 @code{section} attribute has been specified).
11054 This is the default.
11056 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11057 Objects may be explicitly put in the small data area with the
11058 @code{section} attribute using one of these sections.
11060 @item -msdata=sdata
11061 @opindex msdata=sdata
11062 Put small global and static data in the small data area, but do not
11063 generate special code to reference them.
11066 @opindex msdata=use
11067 Put small global and static data in the small data area, and generate
11068 special instructions to reference them.
11072 @cindex smaller data references
11073 Put global and static objects less than or equal to @var{num} bytes
11074 into the small data or bss sections instead of the normal data or bss
11075 sections. The default value of @var{num} is 8.
11076 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11077 for this option to have any effect.
11079 All modules should be compiled with the same @option{-G @var{num}} value.
11080 Compiling with different values of @var{num} may or may not work; if it
11081 doesn't the linker will give an error message---incorrect code will not be
11086 Makes the M32R specific code in the compiler display some statistics
11087 that might help in debugging programs.
11089 @item -malign-loops
11090 @opindex malign-loops
11091 Align all loops to a 32-byte boundary.
11093 @item -mno-align-loops
11094 @opindex mno-align-loops
11095 Do not enforce a 32-byte alignment for loops. This is the default.
11097 @item -missue-rate=@var{number}
11098 @opindex missue-rate=@var{number}
11099 Issue @var{number} instructions per cycle. @var{number} can only be 1
11102 @item -mbranch-cost=@var{number}
11103 @opindex mbranch-cost=@var{number}
11104 @var{number} can only be 1 or 2. If it is 1 then branches will be
11105 preferred over conditional code, if it is 2, then the opposite will
11108 @item -mflush-trap=@var{number}
11109 @opindex mflush-trap=@var{number}
11110 Specifies the trap number to use to flush the cache. The default is
11111 12. Valid numbers are between 0 and 15 inclusive.
11113 @item -mno-flush-trap
11114 @opindex mno-flush-trap
11115 Specifies that the cache cannot be flushed by using a trap.
11117 @item -mflush-func=@var{name}
11118 @opindex mflush-func=@var{name}
11119 Specifies the name of the operating system function to call to flush
11120 the cache. The default is @emph{_flush_cache}, but a function call
11121 will only be used if a trap is not available.
11123 @item -mno-flush-func
11124 @opindex mno-flush-func
11125 Indicates that there is no OS function for flushing the cache.
11129 @node M680x0 Options
11130 @subsection M680x0 Options
11131 @cindex M680x0 options
11133 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11134 The default settings depend on which architecture was selected when
11135 the compiler was configured; the defaults for the most common choices
11139 @item -march=@var{arch}
11141 Generate code for a specific M680x0 or ColdFire instruction set
11142 architecture. Permissible values of @var{arch} for M680x0
11143 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11144 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11145 architectures are selected according to Freescale's ISA classification
11146 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11147 @samp{isab} and @samp{isac}.
11149 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11150 code for a ColdFire target. The @var{arch} in this macro is one of the
11151 @option{-march} arguments given above.
11153 When used together, @option{-march} and @option{-mtune} select code
11154 that runs on a family of similar processors but that is optimized
11155 for a particular microarchitecture.
11157 @item -mcpu=@var{cpu}
11159 Generate code for a specific M680x0 or ColdFire processor.
11160 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11161 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11162 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11163 below, which also classifies the CPUs into families:
11165 @multitable @columnfractions 0.20 0.80
11166 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11167 @item @samp{51qe} @tab @samp{51qe}
11168 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11169 @item @samp{5206e} @tab @samp{5206e}
11170 @item @samp{5208} @tab @samp{5207} @samp{5208}
11171 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11172 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11173 @item @samp{5216} @tab @samp{5214} @samp{5216}
11174 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11175 @item @samp{5225} @tab @samp{5224} @samp{5225}
11176 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11177 @item @samp{5249} @tab @samp{5249}
11178 @item @samp{5250} @tab @samp{5250}
11179 @item @samp{5271} @tab @samp{5270} @samp{5271}
11180 @item @samp{5272} @tab @samp{5272}
11181 @item @samp{5275} @tab @samp{5274} @samp{5275}
11182 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11183 @item @samp{5307} @tab @samp{5307}
11184 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11185 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11186 @item @samp{5407} @tab @samp{5407}
11187 @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}
11190 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11191 @var{arch} is compatible with @var{cpu}. Other combinations of
11192 @option{-mcpu} and @option{-march} are rejected.
11194 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11195 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11196 where the value of @var{family} is given by the table above.
11198 @item -mtune=@var{tune}
11200 Tune the code for a particular microarchitecture, within the
11201 constraints set by @option{-march} and @option{-mcpu}.
11202 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11203 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11204 and @samp{cpu32}. The ColdFire microarchitectures
11205 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11207 You can also use @option{-mtune=68020-40} for code that needs
11208 to run relatively well on 68020, 68030 and 68040 targets.
11209 @option{-mtune=68020-60} is similar but includes 68060 targets
11210 as well. These two options select the same tuning decisions as
11211 @option{-m68020-40} and @option{-m68020-60} respectively.
11213 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11214 when tuning for 680x0 architecture @var{arch}. It also defines
11215 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11216 option is used. If gcc is tuning for a range of architectures,
11217 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11218 it defines the macros for every architecture in the range.
11220 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11221 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11222 of the arguments given above.
11228 Generate output for a 68000. This is the default
11229 when the compiler is configured for 68000-based systems.
11230 It is equivalent to @option{-march=68000}.
11232 Use this option for microcontrollers with a 68000 or EC000 core,
11233 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11237 Generate output for a 68010. This is the default
11238 when the compiler is configured for 68010-based systems.
11239 It is equivalent to @option{-march=68010}.
11245 Generate output for a 68020. This is the default
11246 when the compiler is configured for 68020-based systems.
11247 It is equivalent to @option{-march=68020}.
11251 Generate output for a 68030. This is the default when the compiler is
11252 configured for 68030-based systems. It is equivalent to
11253 @option{-march=68030}.
11257 Generate output for a 68040. This is the default when the compiler is
11258 configured for 68040-based systems. It is equivalent to
11259 @option{-march=68040}.
11261 This option inhibits the use of 68881/68882 instructions that have to be
11262 emulated by software on the 68040. Use this option if your 68040 does not
11263 have code to emulate those instructions.
11267 Generate output for a 68060. This is the default when the compiler is
11268 configured for 68060-based systems. It is equivalent to
11269 @option{-march=68060}.
11271 This option inhibits the use of 68020 and 68881/68882 instructions that
11272 have to be emulated by software on the 68060. Use this option if your 68060
11273 does not have code to emulate those instructions.
11277 Generate output for a CPU32. This is the default
11278 when the compiler is configured for CPU32-based systems.
11279 It is equivalent to @option{-march=cpu32}.
11281 Use this option for microcontrollers with a
11282 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11283 68336, 68340, 68341, 68349 and 68360.
11287 Generate output for a 520X ColdFire CPU. This is the default
11288 when the compiler is configured for 520X-based systems.
11289 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11290 in favor of that option.
11292 Use this option for microcontroller with a 5200 core, including
11293 the MCF5202, MCF5203, MCF5204 and MCF5206.
11297 Generate output for a 5206e ColdFire CPU. The option is now
11298 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11302 Generate output for a member of the ColdFire 528X family.
11303 The option is now deprecated in favor of the equivalent
11304 @option{-mcpu=528x}.
11308 Generate output for a ColdFire 5307 CPU. The option is now deprecated
11309 in favor of the equivalent @option{-mcpu=5307}.
11313 Generate output for a ColdFire 5407 CPU. The option is now deprecated
11314 in favor of the equivalent @option{-mcpu=5407}.
11318 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11319 This includes use of hardware floating point instructions.
11320 The option is equivalent to @option{-mcpu=547x}, and is now
11321 deprecated in favor of that option.
11325 Generate output for a 68040, without using any of the new instructions.
11326 This results in code which can run relatively efficiently on either a
11327 68020/68881 or a 68030 or a 68040. The generated code does use the
11328 68881 instructions that are emulated on the 68040.
11330 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11334 Generate output for a 68060, without using any of the new instructions.
11335 This results in code which can run relatively efficiently on either a
11336 68020/68881 or a 68030 or a 68040. The generated code does use the
11337 68881 instructions that are emulated on the 68060.
11339 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11343 @opindex mhard-float
11345 Generate floating-point instructions. This is the default for 68020
11346 and above, and for ColdFire devices that have an FPU. It defines the
11347 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11348 on ColdFire targets.
11351 @opindex msoft-float
11352 Do not generate floating-point instructions; use library calls instead.
11353 This is the default for 68000, 68010, and 68832 targets. It is also
11354 the default for ColdFire devices that have no FPU.
11360 Generate (do not generate) ColdFire hardware divide and remainder
11361 instructions. If @option{-march} is used without @option{-mcpu},
11362 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11363 architectures. Otherwise, the default is taken from the target CPU
11364 (either the default CPU, or the one specified by @option{-mcpu}). For
11365 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11366 @option{-mcpu=5206e}.
11368 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11372 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11373 Additionally, parameters passed on the stack are also aligned to a
11374 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11378 Do not consider type @code{int} to be 16 bits wide. This is the default.
11381 @itemx -mno-bitfield
11382 @opindex mnobitfield
11383 @opindex mno-bitfield
11384 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11385 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11389 Do use the bit-field instructions. The @option{-m68020} option implies
11390 @option{-mbitfield}. This is the default if you use a configuration
11391 designed for a 68020.
11395 Use a different function-calling convention, in which functions
11396 that take a fixed number of arguments return with the @code{rtd}
11397 instruction, which pops their arguments while returning. This
11398 saves one instruction in the caller since there is no need to pop
11399 the arguments there.
11401 This calling convention is incompatible with the one normally
11402 used on Unix, so you cannot use it if you need to call libraries
11403 compiled with the Unix compiler.
11405 Also, you must provide function prototypes for all functions that
11406 take variable numbers of arguments (including @code{printf});
11407 otherwise incorrect code will be generated for calls to those
11410 In addition, seriously incorrect code will result if you call a
11411 function with too many arguments. (Normally, extra arguments are
11412 harmlessly ignored.)
11414 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
11415 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11419 Do not use the calling conventions selected by @option{-mrtd}.
11420 This is the default.
11423 @itemx -mno-align-int
11424 @opindex malign-int
11425 @opindex mno-align-int
11426 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
11427 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
11428 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
11429 Aligning variables on 32-bit boundaries produces code that runs somewhat
11430 faster on processors with 32-bit busses at the expense of more memory.
11432 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
11433 align structures containing the above types differently than
11434 most published application binary interface specifications for the m68k.
11438 Use the pc-relative addressing mode of the 68000 directly, instead of
11439 using a global offset table. At present, this option implies @option{-fpic},
11440 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
11441 not presently supported with @option{-mpcrel}, though this could be supported for
11442 68020 and higher processors.
11444 @item -mno-strict-align
11445 @itemx -mstrict-align
11446 @opindex mno-strict-align
11447 @opindex mstrict-align
11448 Do not (do) assume that unaligned memory references will be handled by
11452 Generate code that allows the data segment to be located in a different
11453 area of memory from the text segment. This allows for execute in place in
11454 an environment without virtual memory management. This option implies
11457 @item -mno-sep-data
11458 Generate code that assumes that the data segment follows the text segment.
11459 This is the default.
11461 @item -mid-shared-library
11462 Generate code that supports shared libraries via the library ID method.
11463 This allows for execute in place and shared libraries in an environment
11464 without virtual memory management. This option implies @option{-fPIC}.
11466 @item -mno-id-shared-library
11467 Generate code that doesn't assume ID based shared libraries are being used.
11468 This is the default.
11470 @item -mshared-library-id=n
11471 Specified the identification number of the ID based shared library being
11472 compiled. Specifying a value of 0 will generate more compact code, specifying
11473 other values will force the allocation of that number to the current
11474 library but is no more space or time efficient than omitting this option.
11478 @node M68hc1x Options
11479 @subsection M68hc1x Options
11480 @cindex M68hc1x options
11482 These are the @samp{-m} options defined for the 68hc11 and 68hc12
11483 microcontrollers. The default values for these options depends on
11484 which style of microcontroller was selected when the compiler was configured;
11485 the defaults for the most common choices are given below.
11492 Generate output for a 68HC11. This is the default
11493 when the compiler is configured for 68HC11-based systems.
11499 Generate output for a 68HC12. This is the default
11500 when the compiler is configured for 68HC12-based systems.
11506 Generate output for a 68HCS12.
11508 @item -mauto-incdec
11509 @opindex mauto-incdec
11510 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
11517 Enable the use of 68HC12 min and max instructions.
11520 @itemx -mno-long-calls
11521 @opindex mlong-calls
11522 @opindex mno-long-calls
11523 Treat all calls as being far away (near). If calls are assumed to be
11524 far away, the compiler will use the @code{call} instruction to
11525 call a function and the @code{rtc} instruction for returning.
11529 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11531 @item -msoft-reg-count=@var{count}
11532 @opindex msoft-reg-count
11533 Specify the number of pseudo-soft registers which are used for the
11534 code generation. The maximum number is 32. Using more pseudo-soft
11535 register may or may not result in better code depending on the program.
11536 The default is 4 for 68HC11 and 2 for 68HC12.
11540 @node MCore Options
11541 @subsection MCore Options
11542 @cindex MCore options
11544 These are the @samp{-m} options defined for the Motorola M*Core
11550 @itemx -mno-hardlit
11552 @opindex mno-hardlit
11553 Inline constants into the code stream if it can be done in two
11554 instructions or less.
11560 Use the divide instruction. (Enabled by default).
11562 @item -mrelax-immediate
11563 @itemx -mno-relax-immediate
11564 @opindex mrelax-immediate
11565 @opindex mno-relax-immediate
11566 Allow arbitrary sized immediates in bit operations.
11568 @item -mwide-bitfields
11569 @itemx -mno-wide-bitfields
11570 @opindex mwide-bitfields
11571 @opindex mno-wide-bitfields
11572 Always treat bit-fields as int-sized.
11574 @item -m4byte-functions
11575 @itemx -mno-4byte-functions
11576 @opindex m4byte-functions
11577 @opindex mno-4byte-functions
11578 Force all functions to be aligned to a four byte boundary.
11580 @item -mcallgraph-data
11581 @itemx -mno-callgraph-data
11582 @opindex mcallgraph-data
11583 @opindex mno-callgraph-data
11584 Emit callgraph information.
11587 @itemx -mno-slow-bytes
11588 @opindex mslow-bytes
11589 @opindex mno-slow-bytes
11590 Prefer word access when reading byte quantities.
11592 @item -mlittle-endian
11593 @itemx -mbig-endian
11594 @opindex mlittle-endian
11595 @opindex mbig-endian
11596 Generate code for a little endian target.
11602 Generate code for the 210 processor.
11606 @subsection MIPS Options
11607 @cindex MIPS options
11613 Generate big-endian code.
11617 Generate little-endian code. This is the default for @samp{mips*el-*-*}
11620 @item -march=@var{arch}
11622 Generate code that will run on @var{arch}, which can be the name of a
11623 generic MIPS ISA, or the name of a particular processor.
11625 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
11626 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
11627 The processor names are:
11628 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
11629 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
11630 @samp{5kc}, @samp{5kf},
11632 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
11633 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
11634 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
11635 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
11638 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
11639 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
11640 @samp{rm7000}, @samp{rm9000},
11643 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
11644 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
11645 The special value @samp{from-abi} selects the
11646 most compatible architecture for the selected ABI (that is,
11647 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
11649 In processor names, a final @samp{000} can be abbreviated as @samp{k}
11650 (for example, @samp{-march=r2k}). Prefixes are optional, and
11651 @samp{vr} may be written @samp{r}.
11653 Names of the form @samp{@var{n}f2_1} refer to processors with
11654 FPUs clocked at half the rate of the core, names of the form
11655 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
11656 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
11657 processors with FPUs clocked a ratio of 3:2 with respect to the core.
11658 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
11659 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
11660 accepted as synonyms for @samp{@var{n}f1_1}.
11662 GCC defines two macros based on the value of this option. The first
11663 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
11664 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
11665 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
11666 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
11667 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
11669 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
11670 above. In other words, it will have the full prefix and will not
11671 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
11672 the macro names the resolved architecture (either @samp{"mips1"} or
11673 @samp{"mips3"}). It names the default architecture when no
11674 @option{-march} option is given.
11676 @item -mtune=@var{arch}
11678 Optimize for @var{arch}. Among other things, this option controls
11679 the way instructions are scheduled, and the perceived cost of arithmetic
11680 operations. The list of @var{arch} values is the same as for
11683 When this option is not used, GCC will optimize for the processor
11684 specified by @option{-march}. By using @option{-march} and
11685 @option{-mtune} together, it is possible to generate code that will
11686 run on a family of processors, but optimize the code for one
11687 particular member of that family.
11689 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
11690 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
11691 @samp{-march} ones described above.
11695 Equivalent to @samp{-march=mips1}.
11699 Equivalent to @samp{-march=mips2}.
11703 Equivalent to @samp{-march=mips3}.
11707 Equivalent to @samp{-march=mips4}.
11711 Equivalent to @samp{-march=mips32}.
11715 Equivalent to @samp{-march=mips32r2}.
11719 Equivalent to @samp{-march=mips64}.
11724 @opindex mno-mips16
11725 Generate (do not generate) MIPS16 code. If GCC is targetting a
11726 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
11728 MIPS16 code generation can also be controlled on a per-function basis
11729 by means of @code{mips16} and @code{nomips16} attributes.
11730 @xref{Function Attributes}, for more information.
11732 @item -mflip-mips16
11733 @opindex mflip-mips16
11734 Generate MIPS16 code on alternating functions. This option is provided
11735 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
11736 not intended for ordinary use in compiling user code.
11738 @item -minterlink-mips16
11739 @itemx -mno-interlink-mips16
11740 @opindex minterlink-mips16
11741 @opindex mno-interlink-mips16
11742 Require (do not require) that non-MIPS16 code be link-compatible with
11745 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
11746 it must either use a call or an indirect jump. @option{-minterlink-mips16}
11747 therefore disables direct jumps unless GCC knows that the target of the
11748 jump is not MIPS16.
11760 Generate code for the given ABI@.
11762 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
11763 generates 64-bit code when you select a 64-bit architecture, but you
11764 can use @option{-mgp32} to get 32-bit code instead.
11766 For information about the O64 ABI, see
11767 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
11769 GCC supports a variant of the o32 ABI in which floating-point registers
11770 are 64 rather than 32 bits wide. You can select this combination with
11771 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
11772 and @samp{mfhc1} instructions and is therefore only supported for
11773 MIPS32R2 processors.
11775 The register assignments for arguments and return values remain the
11776 same, but each scalar value is passed in a single 64-bit register
11777 rather than a pair of 32-bit registers. For example, scalar
11778 floating-point values are returned in @samp{$f0} only, not a
11779 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
11780 remains the same, but all 64 bits are saved.
11783 @itemx -mno-abicalls
11785 @opindex mno-abicalls
11786 Generate (do not generate) code that is suitable for SVR4-style
11787 dynamic objects. @option{-mabicalls} is the default for SVR4-based
11792 Generate (do not generate) code that is fully position-independent,
11793 and that can therefore be linked into shared libraries. This option
11794 only affects @option{-mabicalls}.
11796 All @option{-mabicalls} code has traditionally been position-independent,
11797 regardless of options like @option{-fPIC} and @option{-fpic}. However,
11798 as an extension, the GNU toolchain allows executables to use absolute
11799 accesses for locally-binding symbols. It can also use shorter GP
11800 initialization sequences and generate direct calls to locally-defined
11801 functions. This mode is selected by @option{-mno-shared}.
11803 @option{-mno-shared} depends on binutils 2.16 or higher and generates
11804 objects that can only be linked by the GNU linker. However, the option
11805 does not affect the ABI of the final executable; it only affects the ABI
11806 of relocatable objects. Using @option{-mno-shared} will generally make
11807 executables both smaller and quicker.
11809 @option{-mshared} is the default.
11815 Lift (do not lift) the usual restrictions on the size of the global
11818 GCC normally uses a single instruction to load values from the GOT@.
11819 While this is relatively efficient, it will only work if the GOT
11820 is smaller than about 64k. Anything larger will cause the linker
11821 to report an error such as:
11823 @cindex relocation truncated to fit (MIPS)
11825 relocation truncated to fit: R_MIPS_GOT16 foobar
11828 If this happens, you should recompile your code with @option{-mxgot}.
11829 It should then work with very large GOTs, although it will also be
11830 less efficient, since it will take three instructions to fetch the
11831 value of a global symbol.
11833 Note that some linkers can create multiple GOTs. If you have such a
11834 linker, you should only need to use @option{-mxgot} when a single object
11835 file accesses more than 64k's worth of GOT entries. Very few do.
11837 These options have no effect unless GCC is generating position
11842 Assume that general-purpose registers are 32 bits wide.
11846 Assume that general-purpose registers are 64 bits wide.
11850 Assume that floating-point registers are 32 bits wide.
11854 Assume that floating-point registers are 64 bits wide.
11857 @opindex mhard-float
11858 Use floating-point coprocessor instructions.
11861 @opindex msoft-float
11862 Do not use floating-point coprocessor instructions. Implement
11863 floating-point calculations using library calls instead.
11865 @item -msingle-float
11866 @opindex msingle-float
11867 Assume that the floating-point coprocessor only supports single-precision
11870 @item -mdouble-float
11871 @opindex mdouble-float
11872 Assume that the floating-point coprocessor supports double-precision
11873 operations. This is the default.
11879 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
11880 implement atomic memory built-in functions. When neither option is
11881 specified, GCC will use the instructions if the target architecture
11884 @option{-mllsc} is useful if the runtime environment can emulate the
11885 instructions and @option{-mno-llsc} can be useful when compiling for
11886 nonstandard ISAs. You can make either option the default by
11887 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
11888 respectively. @option{--with-llsc} is the default for some
11889 configurations; see the installation documentation for details.
11895 Use (do not use) revision 1 of the MIPS DSP ASE.
11896 @xref{MIPS DSP Built-in Functions}. This option defines the
11897 preprocessor macro @samp{__mips_dsp}. It also defines
11898 @samp{__mips_dsp_rev} to 1.
11904 Use (do not use) revision 2 of the MIPS DSP ASE.
11905 @xref{MIPS DSP Built-in Functions}. This option defines the
11906 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
11907 It also defines @samp{__mips_dsp_rev} to 2.
11910 @itemx -mno-smartmips
11911 @opindex msmartmips
11912 @opindex mno-smartmips
11913 Use (do not use) the MIPS SmartMIPS ASE.
11915 @item -mpaired-single
11916 @itemx -mno-paired-single
11917 @opindex mpaired-single
11918 @opindex mno-paired-single
11919 Use (do not use) paired-single floating-point instructions.
11920 @xref{MIPS Paired-Single Support}. This option requires
11921 hardware floating-point support to be enabled.
11927 Use (do not use) MIPS Digital Media Extension instructions.
11928 This option can only be used when generating 64-bit code and requires
11929 hardware floating-point support to be enabled.
11934 @opindex mno-mips3d
11935 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
11936 The option @option{-mips3d} implies @option{-mpaired-single}.
11942 Use (do not use) MT Multithreading instructions.
11946 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
11947 an explanation of the default and the way that the pointer size is
11952 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
11954 The default size of @code{int}s, @code{long}s and pointers depends on
11955 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
11956 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
11957 32-bit @code{long}s. Pointers are the same size as @code{long}s,
11958 or the same size as integer registers, whichever is smaller.
11964 Assume (do not assume) that all symbols have 32-bit values, regardless
11965 of the selected ABI@. This option is useful in combination with
11966 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
11967 to generate shorter and faster references to symbolic addresses.
11971 Put definitions of externally-visible data in a small data section
11972 if that data is no bigger than @var{num} bytes. GCC can then access
11973 the data more efficiently; see @option{-mgpopt} for details.
11975 The default @option{-G} option depends on the configuration.
11977 @item -mlocal-sdata
11978 @itemx -mno-local-sdata
11979 @opindex mlocal-sdata
11980 @opindex mno-local-sdata
11981 Extend (do not extend) the @option{-G} behavior to local data too,
11982 such as to static variables in C. @option{-mlocal-sdata} is the
11983 default for all configurations.
11985 If the linker complains that an application is using too much small data,
11986 you might want to try rebuilding the less performance-critical parts with
11987 @option{-mno-local-sdata}. You might also want to build large
11988 libraries with @option{-mno-local-sdata}, so that the libraries leave
11989 more room for the main program.
11991 @item -mextern-sdata
11992 @itemx -mno-extern-sdata
11993 @opindex mextern-sdata
11994 @opindex mno-extern-sdata
11995 Assume (do not assume) that externally-defined data will be in
11996 a small data section if that data is within the @option{-G} limit.
11997 @option{-mextern-sdata} is the default for all configurations.
11999 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12000 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12001 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12002 is placed in a small data section. If @var{Var} is defined by another
12003 module, you must either compile that module with a high-enough
12004 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12005 definition. If @var{Var} is common, you must link the application
12006 with a high-enough @option{-G} setting.
12008 The easiest way of satisfying these restrictions is to compile
12009 and link every module with the same @option{-G} option. However,
12010 you may wish to build a library that supports several different
12011 small data limits. You can do this by compiling the library with
12012 the highest supported @option{-G} setting and additionally using
12013 @option{-mno-extern-sdata} to stop the library from making assumptions
12014 about externally-defined data.
12020 Use (do not use) GP-relative accesses for symbols that are known to be
12021 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12022 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12025 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12026 might not hold the value of @code{_gp}. For example, if the code is
12027 part of a library that might be used in a boot monitor, programs that
12028 call boot monitor routines will pass an unknown value in @code{$gp}.
12029 (In such situations, the boot monitor itself would usually be compiled
12030 with @option{-G0}.)
12032 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12033 @option{-mno-extern-sdata}.
12035 @item -membedded-data
12036 @itemx -mno-embedded-data
12037 @opindex membedded-data
12038 @opindex mno-embedded-data
12039 Allocate variables to the read-only data section first if possible, then
12040 next in the small data section if possible, otherwise in data. This gives
12041 slightly slower code than the default, but reduces the amount of RAM required
12042 when executing, and thus may be preferred for some embedded systems.
12044 @item -muninit-const-in-rodata
12045 @itemx -mno-uninit-const-in-rodata
12046 @opindex muninit-const-in-rodata
12047 @opindex mno-uninit-const-in-rodata
12048 Put uninitialized @code{const} variables in the read-only data section.
12049 This option is only meaningful in conjunction with @option{-membedded-data}.
12051 @item -mcode-readable=@var{setting}
12052 @opindex mcode-readable
12053 Specify whether GCC may generate code that reads from executable sections.
12054 There are three possible settings:
12057 @item -mcode-readable=yes
12058 Instructions may freely access executable sections. This is the
12061 @item -mcode-readable=pcrel
12062 MIPS16 PC-relative load instructions can access executable sections,
12063 but other instructions must not do so. This option is useful on 4KSc
12064 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12065 It is also useful on processors that can be configured to have a dual
12066 instruction/data SRAM interface and that, like the M4K, automatically
12067 redirect PC-relative loads to the instruction RAM.
12069 @item -mcode-readable=no
12070 Instructions must not access executable sections. This option can be
12071 useful on targets that are configured to have a dual instruction/data
12072 SRAM interface but that (unlike the M4K) do not automatically redirect
12073 PC-relative loads to the instruction RAM.
12076 @item -msplit-addresses
12077 @itemx -mno-split-addresses
12078 @opindex msplit-addresses
12079 @opindex mno-split-addresses
12080 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12081 relocation operators. This option has been superseded by
12082 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12084 @item -mexplicit-relocs
12085 @itemx -mno-explicit-relocs
12086 @opindex mexplicit-relocs
12087 @opindex mno-explicit-relocs
12088 Use (do not use) assembler relocation operators when dealing with symbolic
12089 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12090 is to use assembler macros instead.
12092 @option{-mexplicit-relocs} is the default if GCC was configured
12093 to use an assembler that supports relocation operators.
12095 @item -mcheck-zero-division
12096 @itemx -mno-check-zero-division
12097 @opindex mcheck-zero-division
12098 @opindex mno-check-zero-division
12099 Trap (do not trap) on integer division by zero.
12101 The default is @option{-mcheck-zero-division}.
12103 @item -mdivide-traps
12104 @itemx -mdivide-breaks
12105 @opindex mdivide-traps
12106 @opindex mdivide-breaks
12107 MIPS systems check for division by zero by generating either a
12108 conditional trap or a break instruction. Using traps results in
12109 smaller code, but is only supported on MIPS II and later. Also, some
12110 versions of the Linux kernel have a bug that prevents trap from
12111 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12112 allow conditional traps on architectures that support them and
12113 @option{-mdivide-breaks} to force the use of breaks.
12115 The default is usually @option{-mdivide-traps}, but this can be
12116 overridden at configure time using @option{--with-divide=breaks}.
12117 Divide-by-zero checks can be completely disabled using
12118 @option{-mno-check-zero-division}.
12123 @opindex mno-memcpy
12124 Force (do not force) the use of @code{memcpy()} for non-trivial block
12125 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12126 most constant-sized copies.
12129 @itemx -mno-long-calls
12130 @opindex mlong-calls
12131 @opindex mno-long-calls
12132 Disable (do not disable) use of the @code{jal} instruction. Calling
12133 functions using @code{jal} is more efficient but requires the caller
12134 and callee to be in the same 256 megabyte segment.
12136 This option has no effect on abicalls code. The default is
12137 @option{-mno-long-calls}.
12143 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12144 instructions, as provided by the R4650 ISA@.
12147 @itemx -mno-fused-madd
12148 @opindex mfused-madd
12149 @opindex mno-fused-madd
12150 Enable (disable) use of the floating point multiply-accumulate
12151 instructions, when they are available. The default is
12152 @option{-mfused-madd}.
12154 When multiply-accumulate instructions are used, the intermediate
12155 product is calculated to infinite precision and is not subject to
12156 the FCSR Flush to Zero bit. This may be undesirable in some
12161 Tell the MIPS assembler to not run its preprocessor over user
12162 assembler files (with a @samp{.s} suffix) when assembling them.
12165 @itemx -mno-fix-r4000
12166 @opindex mfix-r4000
12167 @opindex mno-fix-r4000
12168 Work around certain R4000 CPU errata:
12171 A double-word or a variable shift may give an incorrect result if executed
12172 immediately after starting an integer division.
12174 A double-word or a variable shift may give an incorrect result if executed
12175 while an integer multiplication is in progress.
12177 An integer division may give an incorrect result if started in a delay slot
12178 of a taken branch or a jump.
12182 @itemx -mno-fix-r4400
12183 @opindex mfix-r4400
12184 @opindex mno-fix-r4400
12185 Work around certain R4400 CPU errata:
12188 A double-word or a variable shift may give an incorrect result if executed
12189 immediately after starting an integer division.
12193 @itemx -mno-fix-vr4120
12194 @opindex mfix-vr4120
12195 Work around certain VR4120 errata:
12198 @code{dmultu} does not always produce the correct result.
12200 @code{div} and @code{ddiv} do not always produce the correct result if one
12201 of the operands is negative.
12203 The workarounds for the division errata rely on special functions in
12204 @file{libgcc.a}. At present, these functions are only provided by
12205 the @code{mips64vr*-elf} configurations.
12207 Other VR4120 errata require a nop to be inserted between certain pairs of
12208 instructions. These errata are handled by the assembler, not by GCC itself.
12211 @opindex mfix-vr4130
12212 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
12213 workarounds are implemented by the assembler rather than by GCC,
12214 although GCC will avoid using @code{mflo} and @code{mfhi} if the
12215 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
12216 instructions are available instead.
12219 @itemx -mno-fix-sb1
12221 Work around certain SB-1 CPU core errata.
12222 (This flag currently works around the SB-1 revision 2
12223 ``F1'' and ``F2'' floating point errata.)
12225 @item -mflush-func=@var{func}
12226 @itemx -mno-flush-func
12227 @opindex mflush-func
12228 Specifies the function to call to flush the I and D caches, or to not
12229 call any such function. If called, the function must take the same
12230 arguments as the common @code{_flush_func()}, that is, the address of the
12231 memory range for which the cache is being flushed, the size of the
12232 memory range, and the number 3 (to flush both caches). The default
12233 depends on the target GCC was configured for, but commonly is either
12234 @samp{_flush_func} or @samp{__cpu_flush}.
12236 @item mbranch-cost=@var{num}
12237 @opindex mbranch-cost
12238 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12239 This cost is only a heuristic and is not guaranteed to produce
12240 consistent results across releases. A zero cost redundantly selects
12241 the default, which is based on the @option{-mtune} setting.
12243 @item -mbranch-likely
12244 @itemx -mno-branch-likely
12245 @opindex mbranch-likely
12246 @opindex mno-branch-likely
12247 Enable or disable use of Branch Likely instructions, regardless of the
12248 default for the selected architecture. By default, Branch Likely
12249 instructions may be generated if they are supported by the selected
12250 architecture. An exception is for the MIPS32 and MIPS64 architectures
12251 and processors which implement those architectures; for those, Branch
12252 Likely instructions will not be generated by default because the MIPS32
12253 and MIPS64 architectures specifically deprecate their use.
12255 @item -mfp-exceptions
12256 @itemx -mno-fp-exceptions
12257 @opindex mfp-exceptions
12258 Specifies whether FP exceptions are enabled. This affects how we schedule
12259 FP instructions for some processors. The default is that FP exceptions are
12262 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
12263 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
12266 @item -mvr4130-align
12267 @itemx -mno-vr4130-align
12268 @opindex mvr4130-align
12269 The VR4130 pipeline is two-way superscalar, but can only issue two
12270 instructions together if the first one is 8-byte aligned. When this
12271 option is enabled, GCC will align pairs of instructions that it
12272 thinks should execute in parallel.
12274 This option only has an effect when optimizing for the VR4130.
12275 It normally makes code faster, but at the expense of making it bigger.
12276 It is enabled by default at optimization level @option{-O3}.
12280 @subsection MMIX Options
12281 @cindex MMIX Options
12283 These options are defined for the MMIX:
12287 @itemx -mno-libfuncs
12289 @opindex mno-libfuncs
12290 Specify that intrinsic library functions are being compiled, passing all
12291 values in registers, no matter the size.
12294 @itemx -mno-epsilon
12296 @opindex mno-epsilon
12297 Generate floating-point comparison instructions that compare with respect
12298 to the @code{rE} epsilon register.
12300 @item -mabi=mmixware
12302 @opindex mabi-mmixware
12304 Generate code that passes function parameters and return values that (in
12305 the called function) are seen as registers @code{$0} and up, as opposed to
12306 the GNU ABI which uses global registers @code{$231} and up.
12308 @item -mzero-extend
12309 @itemx -mno-zero-extend
12310 @opindex mzero-extend
12311 @opindex mno-zero-extend
12312 When reading data from memory in sizes shorter than 64 bits, use (do not
12313 use) zero-extending load instructions by default, rather than
12314 sign-extending ones.
12317 @itemx -mno-knuthdiv
12319 @opindex mno-knuthdiv
12320 Make the result of a division yielding a remainder have the same sign as
12321 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
12322 remainder follows the sign of the dividend. Both methods are
12323 arithmetically valid, the latter being almost exclusively used.
12325 @item -mtoplevel-symbols
12326 @itemx -mno-toplevel-symbols
12327 @opindex mtoplevel-symbols
12328 @opindex mno-toplevel-symbols
12329 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
12330 code can be used with the @code{PREFIX} assembly directive.
12334 Generate an executable in the ELF format, rather than the default
12335 @samp{mmo} format used by the @command{mmix} simulator.
12337 @item -mbranch-predict
12338 @itemx -mno-branch-predict
12339 @opindex mbranch-predict
12340 @opindex mno-branch-predict
12341 Use (do not use) the probable-branch instructions, when static branch
12342 prediction indicates a probable branch.
12344 @item -mbase-addresses
12345 @itemx -mno-base-addresses
12346 @opindex mbase-addresses
12347 @opindex mno-base-addresses
12348 Generate (do not generate) code that uses @emph{base addresses}. Using a
12349 base address automatically generates a request (handled by the assembler
12350 and the linker) for a constant to be set up in a global register. The
12351 register is used for one or more base address requests within the range 0
12352 to 255 from the value held in the register. The generally leads to short
12353 and fast code, but the number of different data items that can be
12354 addressed is limited. This means that a program that uses lots of static
12355 data may require @option{-mno-base-addresses}.
12357 @item -msingle-exit
12358 @itemx -mno-single-exit
12359 @opindex msingle-exit
12360 @opindex mno-single-exit
12361 Force (do not force) generated code to have a single exit point in each
12365 @node MN10300 Options
12366 @subsection MN10300 Options
12367 @cindex MN10300 options
12369 These @option{-m} options are defined for Matsushita MN10300 architectures:
12374 Generate code to avoid bugs in the multiply instructions for the MN10300
12375 processors. This is the default.
12377 @item -mno-mult-bug
12378 @opindex mno-mult-bug
12379 Do not generate code to avoid bugs in the multiply instructions for the
12380 MN10300 processors.
12384 Generate code which uses features specific to the AM33 processor.
12388 Do not generate code which uses features specific to the AM33 processor. This
12391 @item -mreturn-pointer-on-d0
12392 @opindex mreturn-pointer-on-d0
12393 When generating a function which returns a pointer, return the pointer
12394 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
12395 only in a0, and attempts to call such functions without a prototype
12396 would result in errors. Note that this option is on by default; use
12397 @option{-mno-return-pointer-on-d0} to disable it.
12401 Do not link in the C run-time initialization object file.
12405 Indicate to the linker that it should perform a relaxation optimization pass
12406 to shorten branches, calls and absolute memory addresses. This option only
12407 has an effect when used on the command line for the final link step.
12409 This option makes symbolic debugging impossible.
12413 @subsection MT Options
12416 These @option{-m} options are defined for Morpho MT architectures:
12420 @item -march=@var{cpu-type}
12422 Generate code that will run on @var{cpu-type}, which is the name of a system
12423 representing a certain processor type. Possible values for
12424 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
12425 @samp{ms1-16-003} and @samp{ms2}.
12427 When this option is not used, the default is @option{-march=ms1-16-002}.
12431 Use byte loads and stores when generating code.
12435 Do not use byte loads and stores when generating code.
12439 Use simulator runtime
12443 Do not link in the C run-time initialization object file
12444 @file{crti.o}. Other run-time initialization and termination files
12445 such as @file{startup.o} and @file{exit.o} are still included on the
12446 linker command line.
12450 @node PDP-11 Options
12451 @subsection PDP-11 Options
12452 @cindex PDP-11 Options
12454 These options are defined for the PDP-11:
12459 Use hardware FPP floating point. This is the default. (FIS floating
12460 point on the PDP-11/40 is not supported.)
12463 @opindex msoft-float
12464 Do not use hardware floating point.
12468 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
12472 Return floating-point results in memory. This is the default.
12476 Generate code for a PDP-11/40.
12480 Generate code for a PDP-11/45. This is the default.
12484 Generate code for a PDP-11/10.
12486 @item -mbcopy-builtin
12487 @opindex bcopy-builtin
12488 Use inline @code{movmemhi} patterns for copying memory. This is the
12493 Do not use inline @code{movmemhi} patterns for copying memory.
12499 Use 16-bit @code{int}. This is the default.
12505 Use 32-bit @code{int}.
12508 @itemx -mno-float32
12510 @opindex mno-float32
12511 Use 64-bit @code{float}. This is the default.
12514 @itemx -mno-float64
12516 @opindex mno-float64
12517 Use 32-bit @code{float}.
12521 Use @code{abshi2} pattern. This is the default.
12525 Do not use @code{abshi2} pattern.
12527 @item -mbranch-expensive
12528 @opindex mbranch-expensive
12529 Pretend that branches are expensive. This is for experimenting with
12530 code generation only.
12532 @item -mbranch-cheap
12533 @opindex mbranch-cheap
12534 Do not pretend that branches are expensive. This is the default.
12538 Generate code for a system with split I&D@.
12542 Generate code for a system without split I&D@. This is the default.
12546 Use Unix assembler syntax. This is the default when configured for
12547 @samp{pdp11-*-bsd}.
12551 Use DEC assembler syntax. This is the default when configured for any
12552 PDP-11 target other than @samp{pdp11-*-bsd}.
12555 @node PowerPC Options
12556 @subsection PowerPC Options
12557 @cindex PowerPC options
12559 These are listed under @xref{RS/6000 and PowerPC Options}.
12561 @node RS/6000 and PowerPC Options
12562 @subsection IBM RS/6000 and PowerPC Options
12563 @cindex RS/6000 and PowerPC Options
12564 @cindex IBM RS/6000 and PowerPC Options
12566 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
12573 @itemx -mno-powerpc
12574 @itemx -mpowerpc-gpopt
12575 @itemx -mno-powerpc-gpopt
12576 @itemx -mpowerpc-gfxopt
12577 @itemx -mno-powerpc-gfxopt
12579 @itemx -mno-powerpc64
12583 @itemx -mno-popcntb
12595 @opindex mno-power2
12597 @opindex mno-powerpc
12598 @opindex mpowerpc-gpopt
12599 @opindex mno-powerpc-gpopt
12600 @opindex mpowerpc-gfxopt
12601 @opindex mno-powerpc-gfxopt
12602 @opindex mpowerpc64
12603 @opindex mno-powerpc64
12607 @opindex mno-popcntb
12613 @opindex mno-mfpgpr
12616 GCC supports two related instruction set architectures for the
12617 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
12618 instructions supported by the @samp{rios} chip set used in the original
12619 RS/6000 systems and the @dfn{PowerPC} instruction set is the
12620 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
12621 the IBM 4xx, 6xx, and follow-on microprocessors.
12623 Neither architecture is a subset of the other. However there is a
12624 large common subset of instructions supported by both. An MQ
12625 register is included in processors supporting the POWER architecture.
12627 You use these options to specify which instructions are available on the
12628 processor you are using. The default value of these options is
12629 determined when configuring GCC@. Specifying the
12630 @option{-mcpu=@var{cpu_type}} overrides the specification of these
12631 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
12632 rather than the options listed above.
12634 The @option{-mpower} option allows GCC to generate instructions that
12635 are found only in the POWER architecture and to use the MQ register.
12636 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
12637 to generate instructions that are present in the POWER2 architecture but
12638 not the original POWER architecture.
12640 The @option{-mpowerpc} option allows GCC to generate instructions that
12641 are found only in the 32-bit subset of the PowerPC architecture.
12642 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
12643 GCC to use the optional PowerPC architecture instructions in the
12644 General Purpose group, including floating-point square root. Specifying
12645 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
12646 use the optional PowerPC architecture instructions in the Graphics
12647 group, including floating-point select.
12649 The @option{-mmfcrf} option allows GCC to generate the move from
12650 condition register field instruction implemented on the POWER4
12651 processor and other processors that support the PowerPC V2.01
12653 The @option{-mpopcntb} option allows GCC to generate the popcount and
12654 double precision FP reciprocal estimate instruction implemented on the
12655 POWER5 processor and other processors that support the PowerPC V2.02
12657 The @option{-mfprnd} option allows GCC to generate the FP round to
12658 integer instructions implemented on the POWER5+ processor and other
12659 processors that support the PowerPC V2.03 architecture.
12660 The @option{-mcmpb} option allows GCC to generate the compare bytes
12661 instruction implemented on the POWER6 processor and other processors
12662 that support the PowerPC V2.05 architecture.
12663 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
12664 general purpose register instructions implemented on the POWER6X
12665 processor and other processors that support the extended PowerPC V2.05
12667 The @option{-mdfp} option allows GCC to generate the decimal floating
12668 point instructions implemented on some POWER processors.
12670 The @option{-mpowerpc64} option allows GCC to generate the additional
12671 64-bit instructions that are found in the full PowerPC64 architecture
12672 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
12673 @option{-mno-powerpc64}.
12675 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
12676 will use only the instructions in the common subset of both
12677 architectures plus some special AIX common-mode calls, and will not use
12678 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
12679 permits GCC to use any instruction from either architecture and to
12680 allow use of the MQ register; specify this for the Motorola MPC601.
12682 @item -mnew-mnemonics
12683 @itemx -mold-mnemonics
12684 @opindex mnew-mnemonics
12685 @opindex mold-mnemonics
12686 Select which mnemonics to use in the generated assembler code. With
12687 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
12688 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
12689 assembler mnemonics defined for the POWER architecture. Instructions
12690 defined in only one architecture have only one mnemonic; GCC uses that
12691 mnemonic irrespective of which of these options is specified.
12693 GCC defaults to the mnemonics appropriate for the architecture in
12694 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
12695 value of these option. Unless you are building a cross-compiler, you
12696 should normally not specify either @option{-mnew-mnemonics} or
12697 @option{-mold-mnemonics}, but should instead accept the default.
12699 @item -mcpu=@var{cpu_type}
12701 Set architecture type, register usage, choice of mnemonics, and
12702 instruction scheduling parameters for machine type @var{cpu_type}.
12703 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
12704 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
12705 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
12706 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
12707 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
12708 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
12709 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
12710 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
12711 @samp{power6x}, @samp{common}, @samp{powerpc}, @samp{powerpc64},
12712 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
12714 @option{-mcpu=common} selects a completely generic processor. Code
12715 generated under this option will run on any POWER or PowerPC processor.
12716 GCC will use only the instructions in the common subset of both
12717 architectures, and will not use the MQ register. GCC assumes a generic
12718 processor model for scheduling purposes.
12720 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
12721 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
12722 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
12723 types, with an appropriate, generic processor model assumed for
12724 scheduling purposes.
12726 The other options specify a specific processor. Code generated under
12727 those options will run best on that processor, and may not run at all on
12730 The @option{-mcpu} options automatically enable or disable the
12733 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
12734 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
12735 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
12737 The particular options set for any particular CPU will vary between
12738 compiler versions, depending on what setting seems to produce optimal
12739 code for that CPU; it doesn't necessarily reflect the actual hardware's
12740 capabilities. If you wish to set an individual option to a particular
12741 value, you may specify it after the @option{-mcpu} option, like
12742 @samp{-mcpu=970 -mno-altivec}.
12744 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
12745 not enabled or disabled by the @option{-mcpu} option at present because
12746 AIX does not have full support for these options. You may still
12747 enable or disable them individually if you're sure it'll work in your
12750 @item -mtune=@var{cpu_type}
12752 Set the instruction scheduling parameters for machine type
12753 @var{cpu_type}, but do not set the architecture type, register usage, or
12754 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
12755 values for @var{cpu_type} are used for @option{-mtune} as for
12756 @option{-mcpu}. If both are specified, the code generated will use the
12757 architecture, registers, and mnemonics set by @option{-mcpu}, but the
12758 scheduling parameters set by @option{-mtune}.
12764 Generate code to compute division as reciprocal estimate and iterative
12765 refinement, creating opportunities for increased throughput. This
12766 feature requires: optional PowerPC Graphics instruction set for single
12767 precision and FRE instruction for double precision, assuming divides
12768 cannot generate user-visible traps, and the domain values not include
12769 Infinities, denormals or zero denominator.
12772 @itemx -mno-altivec
12774 @opindex mno-altivec
12775 Generate code that uses (does not use) AltiVec instructions, and also
12776 enable the use of built-in functions that allow more direct access to
12777 the AltiVec instruction set. You may also need to set
12778 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
12784 @opindex mno-vrsave
12785 Generate VRSAVE instructions when generating AltiVec code.
12788 @opindex msecure-plt
12789 Generate code that allows ld and ld.so to build executables and shared
12790 libraries with non-exec .plt and .got sections. This is a PowerPC
12791 32-bit SYSV ABI option.
12795 Generate code that uses a BSS .plt section that ld.so fills in, and
12796 requires .plt and .got sections that are both writable and executable.
12797 This is a PowerPC 32-bit SYSV ABI option.
12803 This switch enables or disables the generation of ISEL instructions.
12805 @item -misel=@var{yes/no}
12806 This switch has been deprecated. Use @option{-misel} and
12807 @option{-mno-isel} instead.
12813 This switch enables or disables the generation of SPE simd
12819 @opindex mno-paired
12820 This switch enables or disables the generation of PAIRED simd
12823 @item -mspe=@var{yes/no}
12824 This option has been deprecated. Use @option{-mspe} and
12825 @option{-mno-spe} instead.
12827 @item -mfloat-gprs=@var{yes/single/double/no}
12828 @itemx -mfloat-gprs
12829 @opindex mfloat-gprs
12830 This switch enables or disables the generation of floating point
12831 operations on the general purpose registers for architectures that
12834 The argument @var{yes} or @var{single} enables the use of
12835 single-precision floating point operations.
12837 The argument @var{double} enables the use of single and
12838 double-precision floating point operations.
12840 The argument @var{no} disables floating point operations on the
12841 general purpose registers.
12843 This option is currently only available on the MPC854x.
12849 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
12850 targets (including GNU/Linux). The 32-bit environment sets int, long
12851 and pointer to 32 bits and generates code that runs on any PowerPC
12852 variant. The 64-bit environment sets int to 32 bits and long and
12853 pointer to 64 bits, and generates code for PowerPC64, as for
12854 @option{-mpowerpc64}.
12857 @itemx -mno-fp-in-toc
12858 @itemx -mno-sum-in-toc
12859 @itemx -mminimal-toc
12861 @opindex mno-fp-in-toc
12862 @opindex mno-sum-in-toc
12863 @opindex mminimal-toc
12864 Modify generation of the TOC (Table Of Contents), which is created for
12865 every executable file. The @option{-mfull-toc} option is selected by
12866 default. In that case, GCC will allocate at least one TOC entry for
12867 each unique non-automatic variable reference in your program. GCC
12868 will also place floating-point constants in the TOC@. However, only
12869 16,384 entries are available in the TOC@.
12871 If you receive a linker error message that saying you have overflowed
12872 the available TOC space, you can reduce the amount of TOC space used
12873 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
12874 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
12875 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
12876 generate code to calculate the sum of an address and a constant at
12877 run-time instead of putting that sum into the TOC@. You may specify one
12878 or both of these options. Each causes GCC to produce very slightly
12879 slower and larger code at the expense of conserving TOC space.
12881 If you still run out of space in the TOC even when you specify both of
12882 these options, specify @option{-mminimal-toc} instead. This option causes
12883 GCC to make only one TOC entry for every file. When you specify this
12884 option, GCC will produce code that is slower and larger but which
12885 uses extremely little TOC space. You may wish to use this option
12886 only on files that contain less frequently executed code.
12892 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
12893 @code{long} type, and the infrastructure needed to support them.
12894 Specifying @option{-maix64} implies @option{-mpowerpc64} and
12895 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
12896 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
12899 @itemx -mno-xl-compat
12900 @opindex mxl-compat
12901 @opindex mno-xl-compat
12902 Produce code that conforms more closely to IBM XL compiler semantics
12903 when using AIX-compatible ABI. Pass floating-point arguments to
12904 prototyped functions beyond the register save area (RSA) on the stack
12905 in addition to argument FPRs. Do not assume that most significant
12906 double in 128-bit long double value is properly rounded when comparing
12907 values and converting to double. Use XL symbol names for long double
12910 The AIX calling convention was extended but not initially documented to
12911 handle an obscure K&R C case of calling a function that takes the
12912 address of its arguments with fewer arguments than declared. IBM XL
12913 compilers access floating point arguments which do not fit in the
12914 RSA from the stack when a subroutine is compiled without
12915 optimization. Because always storing floating-point arguments on the
12916 stack is inefficient and rarely needed, this option is not enabled by
12917 default and only is necessary when calling subroutines compiled by IBM
12918 XL compilers without optimization.
12922 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
12923 application written to use message passing with special startup code to
12924 enable the application to run. The system must have PE installed in the
12925 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
12926 must be overridden with the @option{-specs=} option to specify the
12927 appropriate directory location. The Parallel Environment does not
12928 support threads, so the @option{-mpe} option and the @option{-pthread}
12929 option are incompatible.
12931 @item -malign-natural
12932 @itemx -malign-power
12933 @opindex malign-natural
12934 @opindex malign-power
12935 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
12936 @option{-malign-natural} overrides the ABI-defined alignment of larger
12937 types, such as floating-point doubles, on their natural size-based boundary.
12938 The option @option{-malign-power} instructs GCC to follow the ABI-specified
12939 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
12941 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
12945 @itemx -mhard-float
12946 @opindex msoft-float
12947 @opindex mhard-float
12948 Generate code that does not use (uses) the floating-point register set.
12949 Software floating point emulation is provided if you use the
12950 @option{-msoft-float} option, and pass the option to GCC when linking.
12953 @itemx -mno-multiple
12955 @opindex mno-multiple
12956 Generate code that uses (does not use) the load multiple word
12957 instructions and the store multiple word instructions. These
12958 instructions are generated by default on POWER systems, and not
12959 generated on PowerPC systems. Do not use @option{-mmultiple} on little
12960 endian PowerPC systems, since those instructions do not work when the
12961 processor is in little endian mode. The exceptions are PPC740 and
12962 PPC750 which permit the instructions usage in little endian mode.
12967 @opindex mno-string
12968 Generate code that uses (does not use) the load string instructions
12969 and the store string word instructions to save multiple registers and
12970 do small block moves. These instructions are generated by default on
12971 POWER systems, and not generated on PowerPC systems. Do not use
12972 @option{-mstring} on little endian PowerPC systems, since those
12973 instructions do not work when the processor is in little endian mode.
12974 The exceptions are PPC740 and PPC750 which permit the instructions
12975 usage in little endian mode.
12980 @opindex mno-update
12981 Generate code that uses (does not use) the load or store instructions
12982 that update the base register to the address of the calculated memory
12983 location. These instructions are generated by default. If you use
12984 @option{-mno-update}, there is a small window between the time that the
12985 stack pointer is updated and the address of the previous frame is
12986 stored, which means code that walks the stack frame across interrupts or
12987 signals may get corrupted data.
12990 @itemx -mno-fused-madd
12991 @opindex mfused-madd
12992 @opindex mno-fused-madd
12993 Generate code that uses (does not use) the floating point multiply and
12994 accumulate instructions. These instructions are generated by default if
12995 hardware floating is used.
13001 Generate code that uses (does not use) the half-word multiply and
13002 multiply-accumulate instructions on the IBM 405 and 440 processors.
13003 These instructions are generated by default when targetting those
13010 Generate code that uses (does not use) the string-search @samp{dlmzb}
13011 instruction on the IBM 405 and 440 processors. This instruction is
13012 generated by default when targetting those processors.
13014 @item -mno-bit-align
13016 @opindex mno-bit-align
13017 @opindex mbit-align
13018 On System V.4 and embedded PowerPC systems do not (do) force structures
13019 and unions that contain bit-fields to be aligned to the base type of the
13022 For example, by default a structure containing nothing but 8
13023 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13024 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13025 the structure would be aligned to a 1 byte boundary and be one byte in
13028 @item -mno-strict-align
13029 @itemx -mstrict-align
13030 @opindex mno-strict-align
13031 @opindex mstrict-align
13032 On System V.4 and embedded PowerPC systems do not (do) assume that
13033 unaligned memory references will be handled by the system.
13035 @item -mrelocatable
13036 @itemx -mno-relocatable
13037 @opindex mrelocatable
13038 @opindex mno-relocatable
13039 On embedded PowerPC systems generate code that allows (does not allow)
13040 the program to be relocated to a different address at runtime. If you
13041 use @option{-mrelocatable} on any module, all objects linked together must
13042 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
13044 @item -mrelocatable-lib
13045 @itemx -mno-relocatable-lib
13046 @opindex mrelocatable-lib
13047 @opindex mno-relocatable-lib
13048 On embedded PowerPC systems generate code that allows (does not allow)
13049 the program to be relocated to a different address at runtime. Modules
13050 compiled with @option{-mrelocatable-lib} can be linked with either modules
13051 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
13052 with modules compiled with the @option{-mrelocatable} options.
13058 On System V.4 and embedded PowerPC systems do not (do) assume that
13059 register 2 contains a pointer to a global area pointing to the addresses
13060 used in the program.
13063 @itemx -mlittle-endian
13065 @opindex mlittle-endian
13066 On System V.4 and embedded PowerPC systems compile code for the
13067 processor in little endian mode. The @option{-mlittle-endian} option is
13068 the same as @option{-mlittle}.
13071 @itemx -mbig-endian
13073 @opindex mbig-endian
13074 On System V.4 and embedded PowerPC systems compile code for the
13075 processor in big endian mode. The @option{-mbig-endian} option is
13076 the same as @option{-mbig}.
13078 @item -mdynamic-no-pic
13079 @opindex mdynamic-no-pic
13080 On Darwin and Mac OS X systems, compile code so that it is not
13081 relocatable, but that its external references are relocatable. The
13082 resulting code is suitable for applications, but not shared
13085 @item -mprioritize-restricted-insns=@var{priority}
13086 @opindex mprioritize-restricted-insns
13087 This option controls the priority that is assigned to
13088 dispatch-slot restricted instructions during the second scheduling
13089 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
13090 @var{no/highest/second-highest} priority to dispatch slot restricted
13093 @item -msched-costly-dep=@var{dependence_type}
13094 @opindex msched-costly-dep
13095 This option controls which dependences are considered costly
13096 by the target during instruction scheduling. The argument
13097 @var{dependence_type} takes one of the following values:
13098 @var{no}: no dependence is costly,
13099 @var{all}: all dependences are costly,
13100 @var{true_store_to_load}: a true dependence from store to load is costly,
13101 @var{store_to_load}: any dependence from store to load is costly,
13102 @var{number}: any dependence which latency >= @var{number} is costly.
13104 @item -minsert-sched-nops=@var{scheme}
13105 @opindex minsert-sched-nops
13106 This option controls which nop insertion scheme will be used during
13107 the second scheduling pass. The argument @var{scheme} takes one of the
13109 @var{no}: Don't insert nops.
13110 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
13111 according to the scheduler's grouping.
13112 @var{regroup_exact}: Insert nops to force costly dependent insns into
13113 separate groups. Insert exactly as many nops as needed to force an insn
13114 to a new group, according to the estimated processor grouping.
13115 @var{number}: Insert nops to force costly dependent insns into
13116 separate groups. Insert @var{number} nops to force an insn to a new group.
13119 @opindex mcall-sysv
13120 On System V.4 and embedded PowerPC systems compile code using calling
13121 conventions that adheres to the March 1995 draft of the System V
13122 Application Binary Interface, PowerPC processor supplement. This is the
13123 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
13125 @item -mcall-sysv-eabi
13126 @opindex mcall-sysv-eabi
13127 Specify both @option{-mcall-sysv} and @option{-meabi} options.
13129 @item -mcall-sysv-noeabi
13130 @opindex mcall-sysv-noeabi
13131 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
13133 @item -mcall-solaris
13134 @opindex mcall-solaris
13135 On System V.4 and embedded PowerPC systems compile code for the Solaris
13139 @opindex mcall-linux
13140 On System V.4 and embedded PowerPC systems compile code for the
13141 Linux-based GNU system.
13145 On System V.4 and embedded PowerPC systems compile code for the
13146 Hurd-based GNU system.
13148 @item -mcall-netbsd
13149 @opindex mcall-netbsd
13150 On System V.4 and embedded PowerPC systems compile code for the
13151 NetBSD operating system.
13153 @item -maix-struct-return
13154 @opindex maix-struct-return
13155 Return all structures in memory (as specified by the AIX ABI)@.
13157 @item -msvr4-struct-return
13158 @opindex msvr4-struct-return
13159 Return structures smaller than 8 bytes in registers (as specified by the
13162 @item -mabi=@var{abi-type}
13164 Extend the current ABI with a particular extension, or remove such extension.
13165 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
13166 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
13170 Extend the current ABI with SPE ABI extensions. This does not change
13171 the default ABI, instead it adds the SPE ABI extensions to the current
13175 @opindex mabi=no-spe
13176 Disable Booke SPE ABI extensions for the current ABI@.
13178 @item -mabi=ibmlongdouble
13179 @opindex mabi=ibmlongdouble
13180 Change the current ABI to use IBM extended precision long double.
13181 This is a PowerPC 32-bit SYSV ABI option.
13183 @item -mabi=ieeelongdouble
13184 @opindex mabi=ieeelongdouble
13185 Change the current ABI to use IEEE extended precision long double.
13186 This is a PowerPC 32-bit Linux ABI option.
13189 @itemx -mno-prototype
13190 @opindex mprototype
13191 @opindex mno-prototype
13192 On System V.4 and embedded PowerPC systems assume that all calls to
13193 variable argument functions are properly prototyped. Otherwise, the
13194 compiler must insert an instruction before every non prototyped call to
13195 set or clear bit 6 of the condition code register (@var{CR}) to
13196 indicate whether floating point values were passed in the floating point
13197 registers in case the function takes a variable arguments. With
13198 @option{-mprototype}, only calls to prototyped variable argument functions
13199 will set or clear the bit.
13203 On embedded PowerPC systems, assume that the startup module is called
13204 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
13205 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
13210 On embedded PowerPC systems, assume that the startup module is called
13211 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
13216 On embedded PowerPC systems, assume that the startup module is called
13217 @file{crt0.o} and the standard C libraries are @file{libads.a} and
13220 @item -myellowknife
13221 @opindex myellowknife
13222 On embedded PowerPC systems, assume that the startup module is called
13223 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
13228 On System V.4 and embedded PowerPC systems, specify that you are
13229 compiling for a VxWorks system.
13233 Specify that you are compiling for the WindISS simulation environment.
13237 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
13238 header to indicate that @samp{eabi} extended relocations are used.
13244 On System V.4 and embedded PowerPC systems do (do not) adhere to the
13245 Embedded Applications Binary Interface (eabi) which is a set of
13246 modifications to the System V.4 specifications. Selecting @option{-meabi}
13247 means that the stack is aligned to an 8 byte boundary, a function
13248 @code{__eabi} is called to from @code{main} to set up the eabi
13249 environment, and the @option{-msdata} option can use both @code{r2} and
13250 @code{r13} to point to two separate small data areas. Selecting
13251 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
13252 do not call an initialization function from @code{main}, and the
13253 @option{-msdata} option will only use @code{r13} to point to a single
13254 small data area. The @option{-meabi} option is on by default if you
13255 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
13258 @opindex msdata=eabi
13259 On System V.4 and embedded PowerPC systems, put small initialized
13260 @code{const} global and static data in the @samp{.sdata2} section, which
13261 is pointed to by register @code{r2}. Put small initialized
13262 non-@code{const} global and static data in the @samp{.sdata} section,
13263 which is pointed to by register @code{r13}. Put small uninitialized
13264 global and static data in the @samp{.sbss} section, which is adjacent to
13265 the @samp{.sdata} section. The @option{-msdata=eabi} option is
13266 incompatible with the @option{-mrelocatable} option. The
13267 @option{-msdata=eabi} option also sets the @option{-memb} option.
13270 @opindex msdata=sysv
13271 On System V.4 and embedded PowerPC systems, put small global and static
13272 data in the @samp{.sdata} section, which is pointed to by register
13273 @code{r13}. Put small uninitialized global and static data in the
13274 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
13275 The @option{-msdata=sysv} option is incompatible with the
13276 @option{-mrelocatable} option.
13278 @item -msdata=default
13280 @opindex msdata=default
13282 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
13283 compile code the same as @option{-msdata=eabi}, otherwise compile code the
13284 same as @option{-msdata=sysv}.
13287 @opindex msdata-data
13288 On System V.4 and embedded PowerPC systems, put small global
13289 data in the @samp{.sdata} section. Put small uninitialized global
13290 data in the @samp{.sbss} section. Do not use register @code{r13}
13291 to address small data however. This is the default behavior unless
13292 other @option{-msdata} options are used.
13296 @opindex msdata=none
13298 On embedded PowerPC systems, put all initialized global and static data
13299 in the @samp{.data} section, and all uninitialized data in the
13300 @samp{.bss} section.
13304 @cindex smaller data references (PowerPC)
13305 @cindex .sdata/.sdata2 references (PowerPC)
13306 On embedded PowerPC systems, put global and static items less than or
13307 equal to @var{num} bytes into the small data or bss sections instead of
13308 the normal data or bss section. By default, @var{num} is 8. The
13309 @option{-G @var{num}} switch is also passed to the linker.
13310 All modules should be compiled with the same @option{-G @var{num}} value.
13313 @itemx -mno-regnames
13315 @opindex mno-regnames
13316 On System V.4 and embedded PowerPC systems do (do not) emit register
13317 names in the assembly language output using symbolic forms.
13320 @itemx -mno-longcall
13322 @opindex mno-longcall
13323 By default assume that all calls are far away so that a longer more
13324 expensive calling sequence is required. This is required for calls
13325 further than 32 megabytes (33,554,432 bytes) from the current location.
13326 A short call will be generated if the compiler knows
13327 the call cannot be that far away. This setting can be overridden by
13328 the @code{shortcall} function attribute, or by @code{#pragma
13331 Some linkers are capable of detecting out-of-range calls and generating
13332 glue code on the fly. On these systems, long calls are unnecessary and
13333 generate slower code. As of this writing, the AIX linker can do this,
13334 as can the GNU linker for PowerPC/64. It is planned to add this feature
13335 to the GNU linker for 32-bit PowerPC systems as well.
13337 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
13338 callee, L42'', plus a ``branch island'' (glue code). The two target
13339 addresses represent the callee and the ``branch island''. The
13340 Darwin/PPC linker will prefer the first address and generate a ``bl
13341 callee'' if the PPC ``bl'' instruction will reach the callee directly;
13342 otherwise, the linker will generate ``bl L42'' to call the ``branch
13343 island''. The ``branch island'' is appended to the body of the
13344 calling function; it computes the full 32-bit address of the callee
13347 On Mach-O (Darwin) systems, this option directs the compiler emit to
13348 the glue for every direct call, and the Darwin linker decides whether
13349 to use or discard it.
13351 In the future, we may cause GCC to ignore all longcall specifications
13352 when the linker is known to generate glue.
13356 Adds support for multithreading with the @dfn{pthreads} library.
13357 This option sets flags for both the preprocessor and linker.
13361 @node S/390 and zSeries Options
13362 @subsection S/390 and zSeries Options
13363 @cindex S/390 and zSeries Options
13365 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
13369 @itemx -msoft-float
13370 @opindex mhard-float
13371 @opindex msoft-float
13372 Use (do not use) the hardware floating-point instructions and registers
13373 for floating-point operations. When @option{-msoft-float} is specified,
13374 functions in @file{libgcc.a} will be used to perform floating-point
13375 operations. When @option{-mhard-float} is specified, the compiler
13376 generates IEEE floating-point instructions. This is the default.
13378 @item -mlong-double-64
13379 @itemx -mlong-double-128
13380 @opindex mlong-double-64
13381 @opindex mlong-double-128
13382 These switches control the size of @code{long double} type. A size
13383 of 64bit makes the @code{long double} type equivalent to the @code{double}
13384 type. This is the default.
13387 @itemx -mno-backchain
13388 @opindex mbackchain
13389 @opindex mno-backchain
13390 Store (do not store) the address of the caller's frame as backchain pointer
13391 into the callee's stack frame.
13392 A backchain may be needed to allow debugging using tools that do not understand
13393 DWARF-2 call frame information.
13394 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
13395 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
13396 the backchain is placed into the topmost word of the 96/160 byte register
13399 In general, code compiled with @option{-mbackchain} is call-compatible with
13400 code compiled with @option{-mmo-backchain}; however, use of the backchain
13401 for debugging purposes usually requires that the whole binary is built with
13402 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
13403 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13404 to build a linux kernel use @option{-msoft-float}.
13406 The default is to not maintain the backchain.
13408 @item -mpacked-stack
13409 @item -mno-packed-stack
13410 @opindex mpacked-stack
13411 @opindex mno-packed-stack
13412 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
13413 specified, the compiler uses the all fields of the 96/160 byte register save
13414 area only for their default purpose; unused fields still take up stack space.
13415 When @option{-mpacked-stack} is specified, register save slots are densely
13416 packed at the top of the register save area; unused space is reused for other
13417 purposes, allowing for more efficient use of the available stack space.
13418 However, when @option{-mbackchain} is also in effect, the topmost word of
13419 the save area is always used to store the backchain, and the return address
13420 register is always saved two words below the backchain.
13422 As long as the stack frame backchain is not used, code generated with
13423 @option{-mpacked-stack} is call-compatible with code generated with
13424 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
13425 S/390 or zSeries generated code that uses the stack frame backchain at run
13426 time, not just for debugging purposes. Such code is not call-compatible
13427 with code compiled with @option{-mpacked-stack}. Also, note that the
13428 combination of @option{-mbackchain},
13429 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13430 to build a linux kernel use @option{-msoft-float}.
13432 The default is to not use the packed stack layout.
13435 @itemx -mno-small-exec
13436 @opindex msmall-exec
13437 @opindex mno-small-exec
13438 Generate (or do not generate) code using the @code{bras} instruction
13439 to do subroutine calls.
13440 This only works reliably if the total executable size does not
13441 exceed 64k. The default is to use the @code{basr} instruction instead,
13442 which does not have this limitation.
13448 When @option{-m31} is specified, generate code compliant to the
13449 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
13450 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
13451 particular to generate 64-bit instructions. For the @samp{s390}
13452 targets, the default is @option{-m31}, while the @samp{s390x}
13453 targets default to @option{-m64}.
13459 When @option{-mzarch} is specified, generate code using the
13460 instructions available on z/Architecture.
13461 When @option{-mesa} is specified, generate code using the
13462 instructions available on ESA/390. Note that @option{-mesa} is
13463 not possible with @option{-m64}.
13464 When generating code compliant to the GNU/Linux for S/390 ABI,
13465 the default is @option{-mesa}. When generating code compliant
13466 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
13472 Generate (or do not generate) code using the @code{mvcle} instruction
13473 to perform block moves. When @option{-mno-mvcle} is specified,
13474 use a @code{mvc} loop instead. This is the default unless optimizing for
13481 Print (or do not print) additional debug information when compiling.
13482 The default is to not print debug information.
13484 @item -march=@var{cpu-type}
13486 Generate code that will run on @var{cpu-type}, which is the name of a system
13487 representing a certain processor type. Possible values for
13488 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
13489 When generating code using the instructions available on z/Architecture,
13490 the default is @option{-march=z900}. Otherwise, the default is
13491 @option{-march=g5}.
13493 @item -mtune=@var{cpu-type}
13495 Tune to @var{cpu-type} everything applicable about the generated code,
13496 except for the ABI and the set of available instructions.
13497 The list of @var{cpu-type} values is the same as for @option{-march}.
13498 The default is the value used for @option{-march}.
13501 @itemx -mno-tpf-trace
13502 @opindex mtpf-trace
13503 @opindex mno-tpf-trace
13504 Generate code that adds (does not add) in TPF OS specific branches to trace
13505 routines in the operating system. This option is off by default, even
13506 when compiling for the TPF OS@.
13509 @itemx -mno-fused-madd
13510 @opindex mfused-madd
13511 @opindex mno-fused-madd
13512 Generate code that uses (does not use) the floating point multiply and
13513 accumulate instructions. These instructions are generated by default if
13514 hardware floating point is used.
13516 @item -mwarn-framesize=@var{framesize}
13517 @opindex mwarn-framesize
13518 Emit a warning if the current function exceeds the given frame size. Because
13519 this is a compile time check it doesn't need to be a real problem when the program
13520 runs. It is intended to identify functions which most probably cause
13521 a stack overflow. It is useful to be used in an environment with limited stack
13522 size e.g.@: the linux kernel.
13524 @item -mwarn-dynamicstack
13525 @opindex mwarn-dynamicstack
13526 Emit a warning if the function calls alloca or uses dynamically
13527 sized arrays. This is generally a bad idea with a limited stack size.
13529 @item -mstack-guard=@var{stack-guard}
13530 @item -mstack-size=@var{stack-size}
13531 @opindex mstack-guard
13532 @opindex mstack-size
13533 If these options are provided the s390 back end emits additional instructions in
13534 the function prologue which trigger a trap if the stack size is @var{stack-guard}
13535 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
13536 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
13537 the frame size of the compiled function is chosen.
13538 These options are intended to be used to help debugging stack overflow problems.
13539 The additionally emitted code causes only little overhead and hence can also be
13540 used in production like systems without greater performance degradation. The given
13541 values have to be exact powers of 2 and @var{stack-size} has to be greater than
13542 @var{stack-guard} without exceeding 64k.
13543 In order to be efficient the extra code makes the assumption that the stack starts
13544 at an address aligned to the value given by @var{stack-size}.
13545 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
13548 @node Score Options
13549 @subsection Score Options
13550 @cindex Score Options
13552 These options are defined for Score implementations:
13557 Compile code for big endian mode. This is the default.
13561 Compile code for little endian mode.
13565 Disable generate bcnz instruction.
13569 Enable generate unaligned load and store instruction.
13573 Enable the use of multiply-accumulate instructions. Disabled by default.
13577 Specify the SCORE5 as the target architecture.
13581 Specify the SCORE5U of the target architecture.
13585 Specify the SCORE7 as the target architecture. This is the default.
13589 Specify the SCORE7D as the target architecture.
13593 @subsection SH Options
13595 These @samp{-m} options are defined for the SH implementations:
13600 Generate code for the SH1.
13604 Generate code for the SH2.
13607 Generate code for the SH2e.
13611 Generate code for the SH3.
13615 Generate code for the SH3e.
13619 Generate code for the SH4 without a floating-point unit.
13621 @item -m4-single-only
13622 @opindex m4-single-only
13623 Generate code for the SH4 with a floating-point unit that only
13624 supports single-precision arithmetic.
13628 Generate code for the SH4 assuming the floating-point unit is in
13629 single-precision mode by default.
13633 Generate code for the SH4.
13637 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
13638 floating-point unit is not used.
13640 @item -m4a-single-only
13641 @opindex m4a-single-only
13642 Generate code for the SH4a, in such a way that no double-precision
13643 floating point operations are used.
13646 @opindex m4a-single
13647 Generate code for the SH4a assuming the floating-point unit is in
13648 single-precision mode by default.
13652 Generate code for the SH4a.
13656 Same as @option{-m4a-nofpu}, except that it implicitly passes
13657 @option{-dsp} to the assembler. GCC doesn't generate any DSP
13658 instructions at the moment.
13662 Compile code for the processor in big endian mode.
13666 Compile code for the processor in little endian mode.
13670 Align doubles at 64-bit boundaries. Note that this changes the calling
13671 conventions, and thus some functions from the standard C library will
13672 not work unless you recompile it first with @option{-mdalign}.
13676 Shorten some address references at link time, when possible; uses the
13677 linker option @option{-relax}.
13681 Use 32-bit offsets in @code{switch} tables. The default is to use
13686 Enable the use of the instruction @code{fmovd}.
13690 Comply with the calling conventions defined by Renesas.
13694 Comply with the calling conventions defined by Renesas.
13698 Comply with the calling conventions defined for GCC before the Renesas
13699 conventions were available. This option is the default for all
13700 targets of the SH toolchain except for @samp{sh-symbianelf}.
13703 @opindex mnomacsave
13704 Mark the @code{MAC} register as call-clobbered, even if
13705 @option{-mhitachi} is given.
13709 Increase IEEE-compliance of floating-point code.
13710 At the moment, this is equivalent to @option{-fno-finite-math-only}.
13711 When generating 16 bit SH opcodes, getting IEEE-conforming results for
13712 comparisons of NANs / infinities incurs extra overhead in every
13713 floating point comparison, therefore the default is set to
13714 @option{-ffinite-math-only}.
13716 @item -minline-ic_invalidate
13717 @opindex minline-ic_invalidate
13718 Inline code to invalidate instruction cache entries after setting up
13719 nested function trampolines.
13720 This option has no effect if -musermode is in effect and the selected
13721 code generation option (e.g. -m4) does not allow the use of the icbi
13723 If the selected code generation option does not allow the use of the icbi
13724 instruction, and -musermode is not in effect, the inlined code will
13725 manipulate the instruction cache address array directly with an associative
13726 write. This not only requires privileged mode, but it will also
13727 fail if the cache line had been mapped via the TLB and has become unmapped.
13731 Dump instruction size and location in the assembly code.
13734 @opindex mpadstruct
13735 This option is deprecated. It pads structures to multiple of 4 bytes,
13736 which is incompatible with the SH ABI@.
13740 Optimize for space instead of speed. Implied by @option{-Os}.
13743 @opindex mprefergot
13744 When generating position-independent code, emit function calls using
13745 the Global Offset Table instead of the Procedure Linkage Table.
13749 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
13750 if the inlined code would not work in user mode.
13751 This is the default when the target is @code{sh-*-linux*}.
13753 @item -multcost=@var{number}
13754 @opindex multcost=@var{number}
13755 Set the cost to assume for a multiply insn.
13757 @item -mdiv=@var{strategy}
13758 @opindex mdiv=@var{strategy}
13759 Set the division strategy to use for SHmedia code. @var{strategy} must be
13760 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
13761 inv:call2, inv:fp .
13762 "fp" performs the operation in floating point. This has a very high latency,
13763 but needs only a few instructions, so it might be a good choice if
13764 your code has enough easily exploitable ILP to allow the compiler to
13765 schedule the floating point instructions together with other instructions.
13766 Division by zero causes a floating point exception.
13767 "inv" uses integer operations to calculate the inverse of the divisor,
13768 and then multiplies the dividend with the inverse. This strategy allows
13769 cse and hoisting of the inverse calculation. Division by zero calculates
13770 an unspecified result, but does not trap.
13771 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
13772 have been found, or if the entire operation has been hoisted to the same
13773 place, the last stages of the inverse calculation are intertwined with the
13774 final multiply to reduce the overall latency, at the expense of using a few
13775 more instructions, and thus offering fewer scheduling opportunities with
13777 "call" calls a library function that usually implements the inv:minlat
13779 This gives high code density for m5-*media-nofpu compilations.
13780 "call2" uses a different entry point of the same library function, where it
13781 assumes that a pointer to a lookup table has already been set up, which
13782 exposes the pointer load to cse / code hoisting optimizations.
13783 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
13784 code generation, but if the code stays unoptimized, revert to the "call",
13785 "call2", or "fp" strategies, respectively. Note that the
13786 potentially-trapping side effect of division by zero is carried by a
13787 separate instruction, so it is possible that all the integer instructions
13788 are hoisted out, but the marker for the side effect stays where it is.
13789 A recombination to fp operations or a call is not possible in that case.
13790 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
13791 that the inverse calculation was nor separated from the multiply, they speed
13792 up division where the dividend fits into 20 bits (plus sign where applicable),
13793 by inserting a test to skip a number of operations in this case; this test
13794 slows down the case of larger dividends. inv20u assumes the case of a such
13795 a small dividend to be unlikely, and inv20l assumes it to be likely.
13797 @item -mdivsi3_libfunc=@var{name}
13798 @opindex mdivsi3_libfunc=@var{name}
13799 Set the name of the library function used for 32 bit signed division to
13800 @var{name}. This only affect the name used in the call and inv:call
13801 division strategies, and the compiler will still expect the same
13802 sets of input/output/clobbered registers as if this option was not present.
13804 @item -madjust-unroll
13805 @opindex madjust-unroll
13806 Throttle unrolling to avoid thrashing target registers.
13807 This option only has an effect if the gcc code base supports the
13808 TARGET_ADJUST_UNROLL_MAX target hook.
13810 @item -mindexed-addressing
13811 @opindex mindexed-addressing
13812 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
13813 This is only safe if the hardware and/or OS implement 32 bit wrap-around
13814 semantics for the indexed addressing mode. The architecture allows the
13815 implementation of processors with 64 bit MMU, which the OS could use to
13816 get 32 bit addressing, but since no current hardware implementation supports
13817 this or any other way to make the indexed addressing mode safe to use in
13818 the 32 bit ABI, the default is -mno-indexed-addressing.
13820 @item -mgettrcost=@var{number}
13821 @opindex mgettrcost=@var{number}
13822 Set the cost assumed for the gettr instruction to @var{number}.
13823 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
13827 Assume pt* instructions won't trap. This will generally generate better
13828 scheduled code, but is unsafe on current hardware. The current architecture
13829 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
13830 This has the unintentional effect of making it unsafe to schedule ptabs /
13831 ptrel before a branch, or hoist it out of a loop. For example,
13832 __do_global_ctors, a part of libgcc that runs constructors at program
13833 startup, calls functions in a list which is delimited by @minus{}1. With the
13834 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
13835 That means that all the constructors will be run a bit quicker, but when
13836 the loop comes to the end of the list, the program crashes because ptabs
13837 loads @minus{}1 into a target register. Since this option is unsafe for any
13838 hardware implementing the current architecture specification, the default
13839 is -mno-pt-fixed. Unless the user specifies a specific cost with
13840 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
13841 this deters register allocation using target registers for storing
13844 @item -minvalid-symbols
13845 @opindex minvalid-symbols
13846 Assume symbols might be invalid. Ordinary function symbols generated by
13847 the compiler will always be valid to load with movi/shori/ptabs or
13848 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
13849 to generate symbols that will cause ptabs / ptrel to trap.
13850 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
13851 It will then prevent cross-basic-block cse, hoisting and most scheduling
13852 of symbol loads. The default is @option{-mno-invalid-symbols}.
13855 @node SPARC Options
13856 @subsection SPARC Options
13857 @cindex SPARC options
13859 These @samp{-m} options are supported on the SPARC:
13862 @item -mno-app-regs
13864 @opindex mno-app-regs
13866 Specify @option{-mapp-regs} to generate output using the global registers
13867 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
13870 To be fully SVR4 ABI compliant at the cost of some performance loss,
13871 specify @option{-mno-app-regs}. You should compile libraries and system
13872 software with this option.
13875 @itemx -mhard-float
13877 @opindex mhard-float
13878 Generate output containing floating point instructions. This is the
13882 @itemx -msoft-float
13884 @opindex msoft-float
13885 Generate output containing library calls for floating point.
13886 @strong{Warning:} the requisite libraries are not available for all SPARC
13887 targets. Normally the facilities of the machine's usual C compiler are
13888 used, but this cannot be done directly in cross-compilation. You must make
13889 your own arrangements to provide suitable library functions for
13890 cross-compilation. The embedded targets @samp{sparc-*-aout} and
13891 @samp{sparclite-*-*} do provide software floating point support.
13893 @option{-msoft-float} changes the calling convention in the output file;
13894 therefore, it is only useful if you compile @emph{all} of a program with
13895 this option. In particular, you need to compile @file{libgcc.a}, the
13896 library that comes with GCC, with @option{-msoft-float} in order for
13899 @item -mhard-quad-float
13900 @opindex mhard-quad-float
13901 Generate output containing quad-word (long double) floating point
13904 @item -msoft-quad-float
13905 @opindex msoft-quad-float
13906 Generate output containing library calls for quad-word (long double)
13907 floating point instructions. The functions called are those specified
13908 in the SPARC ABI@. This is the default.
13910 As of this writing, there are no SPARC implementations that have hardware
13911 support for the quad-word floating point instructions. They all invoke
13912 a trap handler for one of these instructions, and then the trap handler
13913 emulates the effect of the instruction. Because of the trap handler overhead,
13914 this is much slower than calling the ABI library routines. Thus the
13915 @option{-msoft-quad-float} option is the default.
13917 @item -mno-unaligned-doubles
13918 @itemx -munaligned-doubles
13919 @opindex mno-unaligned-doubles
13920 @opindex munaligned-doubles
13921 Assume that doubles have 8 byte alignment. This is the default.
13923 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
13924 alignment only if they are contained in another type, or if they have an
13925 absolute address. Otherwise, it assumes they have 4 byte alignment.
13926 Specifying this option avoids some rare compatibility problems with code
13927 generated by other compilers. It is not the default because it results
13928 in a performance loss, especially for floating point code.
13930 @item -mno-faster-structs
13931 @itemx -mfaster-structs
13932 @opindex mno-faster-structs
13933 @opindex mfaster-structs
13934 With @option{-mfaster-structs}, the compiler assumes that structures
13935 should have 8 byte alignment. This enables the use of pairs of
13936 @code{ldd} and @code{std} instructions for copies in structure
13937 assignment, in place of twice as many @code{ld} and @code{st} pairs.
13938 However, the use of this changed alignment directly violates the SPARC
13939 ABI@. Thus, it's intended only for use on targets where the developer
13940 acknowledges that their resulting code will not be directly in line with
13941 the rules of the ABI@.
13943 @item -mimpure-text
13944 @opindex mimpure-text
13945 @option{-mimpure-text}, used in addition to @option{-shared}, tells
13946 the compiler to not pass @option{-z text} to the linker when linking a
13947 shared object. Using this option, you can link position-dependent
13948 code into a shared object.
13950 @option{-mimpure-text} suppresses the ``relocations remain against
13951 allocatable but non-writable sections'' linker error message.
13952 However, the necessary relocations will trigger copy-on-write, and the
13953 shared object is not actually shared across processes. Instead of
13954 using @option{-mimpure-text}, you should compile all source code with
13955 @option{-fpic} or @option{-fPIC}.
13957 This option is only available on SunOS and Solaris.
13959 @item -mcpu=@var{cpu_type}
13961 Set the instruction set, register set, and instruction scheduling parameters
13962 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
13963 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
13964 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
13965 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
13966 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
13968 Default instruction scheduling parameters are used for values that select
13969 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
13970 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
13972 Here is a list of each supported architecture and their supported
13977 v8: supersparc, hypersparc
13978 sparclite: f930, f934, sparclite86x
13980 v9: ultrasparc, ultrasparc3, niagara, niagara2
13983 By default (unless configured otherwise), GCC generates code for the V7
13984 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
13985 additionally optimizes it for the Cypress CY7C602 chip, as used in the
13986 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
13987 SPARCStation 1, 2, IPX etc.
13989 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
13990 architecture. The only difference from V7 code is that the compiler emits
13991 the integer multiply and integer divide instructions which exist in SPARC-V8
13992 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
13993 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
13996 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
13997 the SPARC architecture. This adds the integer multiply, integer divide step
13998 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
13999 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14000 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14001 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14002 MB86934 chip, which is the more recent SPARClite with FPU@.
14004 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14005 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
14006 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
14007 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
14008 optimizes it for the TEMIC SPARClet chip.
14010 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
14011 architecture. This adds 64-bit integer and floating-point move instructions,
14012 3 additional floating-point condition code registers and conditional move
14013 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
14014 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
14015 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
14016 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
14017 @option{-mcpu=niagara}, the compiler additionally optimizes it for
14018 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
14019 additionally optimizes it for Sun UltraSPARC T2 chips.
14021 @item -mtune=@var{cpu_type}
14023 Set the instruction scheduling parameters for machine type
14024 @var{cpu_type}, but do not set the instruction set or register set that the
14025 option @option{-mcpu=@var{cpu_type}} would.
14027 The same values for @option{-mcpu=@var{cpu_type}} can be used for
14028 @option{-mtune=@var{cpu_type}}, but the only useful values are those
14029 that select a particular cpu implementation. Those are @samp{cypress},
14030 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
14031 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
14032 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
14037 @opindex mno-v8plus
14038 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
14039 difference from the V8 ABI is that the global and out registers are
14040 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
14041 mode for all SPARC-V9 processors.
14047 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
14048 Visual Instruction Set extensions. The default is @option{-mno-vis}.
14051 These @samp{-m} options are supported in addition to the above
14052 on SPARC-V9 processors in 64-bit environments:
14055 @item -mlittle-endian
14056 @opindex mlittle-endian
14057 Generate code for a processor running in little-endian mode. It is only
14058 available for a few configurations and most notably not on Solaris and Linux.
14064 Generate code for a 32-bit or 64-bit environment.
14065 The 32-bit environment sets int, long and pointer to 32 bits.
14066 The 64-bit environment sets int to 32 bits and long and pointer
14069 @item -mcmodel=medlow
14070 @opindex mcmodel=medlow
14071 Generate code for the Medium/Low code model: 64-bit addresses, programs
14072 must be linked in the low 32 bits of memory. Programs can be statically
14073 or dynamically linked.
14075 @item -mcmodel=medmid
14076 @opindex mcmodel=medmid
14077 Generate code for the Medium/Middle code model: 64-bit addresses, programs
14078 must be linked in the low 44 bits of memory, the text and data segments must
14079 be less than 2GB in size and the data segment must be located within 2GB of
14082 @item -mcmodel=medany
14083 @opindex mcmodel=medany
14084 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
14085 may be linked anywhere in memory, the text and data segments must be less
14086 than 2GB in size and the data segment must be located within 2GB of the
14089 @item -mcmodel=embmedany
14090 @opindex mcmodel=embmedany
14091 Generate code for the Medium/Anywhere code model for embedded systems:
14092 64-bit addresses, the text and data segments must be less than 2GB in
14093 size, both starting anywhere in memory (determined at link time). The
14094 global register %g4 points to the base of the data segment. Programs
14095 are statically linked and PIC is not supported.
14098 @itemx -mno-stack-bias
14099 @opindex mstack-bias
14100 @opindex mno-stack-bias
14101 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
14102 frame pointer if present, are offset by @minus{}2047 which must be added back
14103 when making stack frame references. This is the default in 64-bit mode.
14104 Otherwise, assume no such offset is present.
14107 These switches are supported in addition to the above on Solaris:
14112 Add support for multithreading using the Solaris threads library. This
14113 option sets flags for both the preprocessor and linker. This option does
14114 not affect the thread safety of object code produced by the compiler or
14115 that of libraries supplied with it.
14119 Add support for multithreading using the POSIX threads library. This
14120 option sets flags for both the preprocessor and linker. This option does
14121 not affect the thread safety of object code produced by the compiler or
14122 that of libraries supplied with it.
14126 This is a synonym for @option{-pthreads}.
14130 @subsection SPU Options
14131 @cindex SPU options
14133 These @samp{-m} options are supported on the SPU:
14137 @itemx -merror-reloc
14138 @opindex mwarn-reloc
14139 @opindex merror-reloc
14141 The loader for SPU does not handle dynamic relocations. By default, GCC
14142 will give an error when it generates code that requires a dynamic
14143 relocation. @option{-mno-error-reloc} disables the error,
14144 @option{-mwarn-reloc} will generate a warning instead.
14147 @itemx -munsafe-dma
14149 @opindex munsafe-dma
14151 Instructions which initiate or test completion of DMA must not be
14152 reordered with respect to loads and stores of the memory which is being
14153 accessed. Users typically address this problem using the volatile
14154 keyword, but that can lead to inefficient code in places where the
14155 memory is known to not change. Rather than mark the memory as volatile
14156 we treat the DMA instructions as potentially effecting all memory. With
14157 @option{-munsafe-dma} users must use the volatile keyword to protect
14160 @item -mbranch-hints
14161 @opindex mbranch-hints
14163 By default, GCC will generate a branch hint instruction to avoid
14164 pipeline stalls for always taken or probably taken branches. A hint
14165 will not be generated closer than 8 instructions away from its branch.
14166 There is little reason to disable them, except for debugging purposes,
14167 or to make an object a little bit smaller.
14171 @opindex msmall-mem
14172 @opindex mlarge-mem
14174 By default, GCC generates code assuming that addresses are never larger
14175 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
14176 a full 32 bit address.
14181 By default, GCC links against startup code that assumes the SPU-style
14182 main function interface (which has an unconventional parameter list).
14183 With @option{-mstdmain}, GCC will link your program against startup
14184 code that assumes a C99-style interface to @code{main}, including a
14185 local copy of @code{argv} strings.
14187 @item -mfixed-range=@var{register-range}
14188 @opindex mfixed-range
14189 Generate code treating the given register range as fixed registers.
14190 A fixed register is one that the register allocator can not use. This is
14191 useful when compiling kernel code. A register range is specified as
14192 two registers separated by a dash. Multiple register ranges can be
14193 specified separated by a comma.
14197 @node System V Options
14198 @subsection Options for System V
14200 These additional options are available on System V Release 4 for
14201 compatibility with other compilers on those systems:
14206 Create a shared object.
14207 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
14211 Identify the versions of each tool used by the compiler, in a
14212 @code{.ident} assembler directive in the output.
14216 Refrain from adding @code{.ident} directives to the output file (this is
14219 @item -YP,@var{dirs}
14221 Search the directories @var{dirs}, and no others, for libraries
14222 specified with @option{-l}.
14224 @item -Ym,@var{dir}
14226 Look in the directory @var{dir} to find the M4 preprocessor.
14227 The assembler uses this option.
14228 @c This is supposed to go with a -Yd for predefined M4 macro files, but
14229 @c the generic assembler that comes with Solaris takes just -Ym.
14232 @node TMS320C3x/C4x Options
14233 @subsection TMS320C3x/C4x Options
14234 @cindex TMS320C3x/C4x Options
14236 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
14240 @item -mcpu=@var{cpu_type}
14242 Set the instruction set, register set, and instruction scheduling
14243 parameters for machine type @var{cpu_type}. Supported values for
14244 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
14245 @samp{c44}. The default is @samp{c40} to generate code for the
14250 @itemx -msmall-memory
14252 @opindex mbig-memory
14254 @opindex msmall-memory
14256 Generates code for the big or small memory model. The small memory
14257 model assumed that all data fits into one 64K word page. At run-time
14258 the data page (DP) register must be set to point to the 64K page
14259 containing the .bss and .data program sections. The big memory model is
14260 the default and requires reloading of the DP register for every direct
14267 Allow (disallow) allocation of general integer operands into the block
14268 count register BK@.
14274 Enable (disable) generation of code using decrement and branch,
14275 DBcond(D), instructions. This is enabled by default for the C4x. To be
14276 on the safe side, this is disabled for the C3x, since the maximum
14277 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
14278 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
14279 that it can utilize the decrement and branch instruction, but will give
14280 up if there is more than one memory reference in the loop. Thus a loop
14281 where the loop counter is decremented can generate slightly more
14282 efficient code, in cases where the RPTB instruction cannot be utilized.
14284 @item -mdp-isr-reload
14286 @opindex mdp-isr-reload
14288 Force the DP register to be saved on entry to an interrupt service
14289 routine (ISR), reloaded to point to the data section, and restored on
14290 exit from the ISR@. This should not be required unless someone has
14291 violated the small memory model by modifying the DP register, say within
14298 For the C3x use the 24-bit MPYI instruction for integer multiplies
14299 instead of a library call to guarantee 32-bit results. Note that if one
14300 of the operands is a constant, then the multiplication will be performed
14301 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
14302 then squaring operations are performed inline instead of a library call.
14305 @itemx -mno-fast-fix
14307 @opindex mno-fast-fix
14308 The C3x/C4x FIX instruction to convert a floating point value to an
14309 integer value chooses the nearest integer less than or equal to the
14310 floating point value rather than to the nearest integer. Thus if the
14311 floating point number is negative, the result will be incorrectly
14312 truncated an additional code is necessary to detect and correct this
14313 case. This option can be used to disable generation of the additional
14314 code required to correct the result.
14320 Enable (disable) generation of repeat block sequences using the RPTB
14321 instruction for zero overhead looping. The RPTB construct is only used
14322 for innermost loops that do not call functions or jump across the loop
14323 boundaries. There is no advantage having nested RPTB loops due to the
14324 overhead required to save and restore the RC, RS, and RE registers.
14325 This is enabled by default with @option{-O2}.
14327 @item -mrpts=@var{count}
14331 Enable (disable) the use of the single instruction repeat instruction
14332 RPTS@. If a repeat block contains a single instruction, and the loop
14333 count can be guaranteed to be less than the value @var{count}, GCC will
14334 emit a RPTS instruction instead of a RPTB@. If no value is specified,
14335 then a RPTS will be emitted even if the loop count cannot be determined
14336 at compile time. Note that the repeated instruction following RPTS does
14337 not have to be reloaded from memory each iteration, thus freeing up the
14338 CPU buses for operands. However, since interrupts are blocked by this
14339 instruction, it is disabled by default.
14341 @item -mloop-unsigned
14342 @itemx -mno-loop-unsigned
14343 @opindex mloop-unsigned
14344 @opindex mno-loop-unsigned
14345 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
14346 is @math{2^{31} + 1} since these instructions test if the iteration count is
14347 negative to terminate the loop. If the iteration count is unsigned
14348 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
14349 exceeded. This switch allows an unsigned iteration count.
14353 Try to emit an assembler syntax that the TI assembler (asm30) is happy
14354 with. This also enforces compatibility with the API employed by the TI
14355 C3x C compiler. For example, long doubles are passed as structures
14356 rather than in floating point registers.
14362 Generate code that uses registers (stack) for passing arguments to functions.
14363 By default, arguments are passed in registers where possible rather
14364 than by pushing arguments on to the stack.
14366 @item -mparallel-insns
14367 @itemx -mno-parallel-insns
14368 @opindex mparallel-insns
14369 @opindex mno-parallel-insns
14370 Allow the generation of parallel instructions. This is enabled by
14371 default with @option{-O2}.
14373 @item -mparallel-mpy
14374 @itemx -mno-parallel-mpy
14375 @opindex mparallel-mpy
14376 @opindex mno-parallel-mpy
14377 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
14378 provided @option{-mparallel-insns} is also specified. These instructions have
14379 tight register constraints which can pessimize the code generation
14380 of large functions.
14385 @subsection V850 Options
14386 @cindex V850 Options
14388 These @samp{-m} options are defined for V850 implementations:
14392 @itemx -mno-long-calls
14393 @opindex mlong-calls
14394 @opindex mno-long-calls
14395 Treat all calls as being far away (near). If calls are assumed to be
14396 far away, the compiler will always load the functions address up into a
14397 register, and call indirect through the pointer.
14403 Do not optimize (do optimize) basic blocks that use the same index
14404 pointer 4 or more times to copy pointer into the @code{ep} register, and
14405 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
14406 option is on by default if you optimize.
14408 @item -mno-prolog-function
14409 @itemx -mprolog-function
14410 @opindex mno-prolog-function
14411 @opindex mprolog-function
14412 Do not use (do use) external functions to save and restore registers
14413 at the prologue and epilogue of a function. The external functions
14414 are slower, but use less code space if more than one function saves
14415 the same number of registers. The @option{-mprolog-function} option
14416 is on by default if you optimize.
14420 Try to make the code as small as possible. At present, this just turns
14421 on the @option{-mep} and @option{-mprolog-function} options.
14423 @item -mtda=@var{n}
14425 Put static or global variables whose size is @var{n} bytes or less into
14426 the tiny data area that register @code{ep} points to. The tiny data
14427 area can hold up to 256 bytes in total (128 bytes for byte references).
14429 @item -msda=@var{n}
14431 Put static or global variables whose size is @var{n} bytes or less into
14432 the small data area that register @code{gp} points to. The small data
14433 area can hold up to 64 kilobytes.
14435 @item -mzda=@var{n}
14437 Put static or global variables whose size is @var{n} bytes or less into
14438 the first 32 kilobytes of memory.
14442 Specify that the target processor is the V850.
14445 @opindex mbig-switch
14446 Generate code suitable for big switch tables. Use this option only if
14447 the assembler/linker complain about out of range branches within a switch
14452 This option will cause r2 and r5 to be used in the code generated by
14453 the compiler. This setting is the default.
14455 @item -mno-app-regs
14456 @opindex mno-app-regs
14457 This option will cause r2 and r5 to be treated as fixed registers.
14461 Specify that the target processor is the V850E1. The preprocessor
14462 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
14463 this option is used.
14467 Specify that the target processor is the V850E@. The preprocessor
14468 constant @samp{__v850e__} will be defined if this option is used.
14470 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
14471 are defined then a default target processor will be chosen and the
14472 relevant @samp{__v850*__} preprocessor constant will be defined.
14474 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
14475 defined, regardless of which processor variant is the target.
14477 @item -mdisable-callt
14478 @opindex mdisable-callt
14479 This option will suppress generation of the CALLT instruction for the
14480 v850e and v850e1 flavors of the v850 architecture. The default is
14481 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
14486 @subsection VAX Options
14487 @cindex VAX options
14489 These @samp{-m} options are defined for the VAX:
14494 Do not output certain jump instructions (@code{aobleq} and so on)
14495 that the Unix assembler for the VAX cannot handle across long
14500 Do output those jump instructions, on the assumption that you
14501 will assemble with the GNU assembler.
14505 Output code for g-format floating point numbers instead of d-format.
14508 @node VxWorks Options
14509 @subsection VxWorks Options
14510 @cindex VxWorks Options
14512 The options in this section are defined for all VxWorks targets.
14513 Options specific to the target hardware are listed with the other
14514 options for that target.
14519 GCC can generate code for both VxWorks kernels and real time processes
14520 (RTPs). This option switches from the former to the latter. It also
14521 defines the preprocessor macro @code{__RTP__}.
14524 @opindex non-static
14525 Link an RTP executable against shared libraries rather than static
14526 libraries. The options @option{-static} and @option{-shared} can
14527 also be used for RTPs (@pxref{Link Options}); @option{-static}
14534 These options are passed down to the linker. They are defined for
14535 compatibility with Diab.
14538 @opindex Xbind-lazy
14539 Enable lazy binding of function calls. This option is equivalent to
14540 @option{-Wl,-z,now} and is defined for compatibility with Diab.
14544 Disable lazy binding of function calls. This option is the default and
14545 is defined for compatibility with Diab.
14548 @node x86-64 Options
14549 @subsection x86-64 Options
14550 @cindex x86-64 options
14552 These are listed under @xref{i386 and x86-64 Options}.
14554 @node Xstormy16 Options
14555 @subsection Xstormy16 Options
14556 @cindex Xstormy16 Options
14558 These options are defined for Xstormy16:
14563 Choose startup files and linker script suitable for the simulator.
14566 @node Xtensa Options
14567 @subsection Xtensa Options
14568 @cindex Xtensa Options
14570 These options are supported for Xtensa targets:
14574 @itemx -mno-const16
14576 @opindex mno-const16
14577 Enable or disable use of @code{CONST16} instructions for loading
14578 constant values. The @code{CONST16} instruction is currently not a
14579 standard option from Tensilica. When enabled, @code{CONST16}
14580 instructions are always used in place of the standard @code{L32R}
14581 instructions. The use of @code{CONST16} is enabled by default only if
14582 the @code{L32R} instruction is not available.
14585 @itemx -mno-fused-madd
14586 @opindex mfused-madd
14587 @opindex mno-fused-madd
14588 Enable or disable use of fused multiply/add and multiply/subtract
14589 instructions in the floating-point option. This has no effect if the
14590 floating-point option is not also enabled. Disabling fused multiply/add
14591 and multiply/subtract instructions forces the compiler to use separate
14592 instructions for the multiply and add/subtract operations. This may be
14593 desirable in some cases where strict IEEE 754-compliant results are
14594 required: the fused multiply add/subtract instructions do not round the
14595 intermediate result, thereby producing results with @emph{more} bits of
14596 precision than specified by the IEEE standard. Disabling fused multiply
14597 add/subtract instructions also ensures that the program output is not
14598 sensitive to the compiler's ability to combine multiply and add/subtract
14601 @item -mtext-section-literals
14602 @itemx -mno-text-section-literals
14603 @opindex mtext-section-literals
14604 @opindex mno-text-section-literals
14605 Control the treatment of literal pools. The default is
14606 @option{-mno-text-section-literals}, which places literals in a separate
14607 section in the output file. This allows the literal pool to be placed
14608 in a data RAM/ROM, and it also allows the linker to combine literal
14609 pools from separate object files to remove redundant literals and
14610 improve code size. With @option{-mtext-section-literals}, the literals
14611 are interspersed in the text section in order to keep them as close as
14612 possible to their references. This may be necessary for large assembly
14615 @item -mtarget-align
14616 @itemx -mno-target-align
14617 @opindex mtarget-align
14618 @opindex mno-target-align
14619 When this option is enabled, GCC instructs the assembler to
14620 automatically align instructions to reduce branch penalties at the
14621 expense of some code density. The assembler attempts to widen density
14622 instructions to align branch targets and the instructions following call
14623 instructions. If there are not enough preceding safe density
14624 instructions to align a target, no widening will be performed. The
14625 default is @option{-mtarget-align}. These options do not affect the
14626 treatment of auto-aligned instructions like @code{LOOP}, which the
14627 assembler will always align, either by widening density instructions or
14628 by inserting no-op instructions.
14631 @itemx -mno-longcalls
14632 @opindex mlongcalls
14633 @opindex mno-longcalls
14634 When this option is enabled, GCC instructs the assembler to translate
14635 direct calls to indirect calls unless it can determine that the target
14636 of a direct call is in the range allowed by the call instruction. This
14637 translation typically occurs for calls to functions in other source
14638 files. Specifically, the assembler translates a direct @code{CALL}
14639 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
14640 The default is @option{-mno-longcalls}. This option should be used in
14641 programs where the call target can potentially be out of range. This
14642 option is implemented in the assembler, not the compiler, so the
14643 assembly code generated by GCC will still show direct call
14644 instructions---look at the disassembled object code to see the actual
14645 instructions. Note that the assembler will use an indirect call for
14646 every cross-file call, not just those that really will be out of range.
14649 @node zSeries Options
14650 @subsection zSeries Options
14651 @cindex zSeries options
14653 These are listed under @xref{S/390 and zSeries Options}.
14655 @node Code Gen Options
14656 @section Options for Code Generation Conventions
14657 @cindex code generation conventions
14658 @cindex options, code generation
14659 @cindex run-time options
14661 These machine-independent options control the interface conventions
14662 used in code generation.
14664 Most of them have both positive and negative forms; the negative form
14665 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
14666 one of the forms is listed---the one which is not the default. You
14667 can figure out the other form by either removing @samp{no-} or adding
14671 @item -fbounds-check
14672 @opindex fbounds-check
14673 For front-ends that support it, generate additional code to check that
14674 indices used to access arrays are within the declared range. This is
14675 currently only supported by the Java and Fortran front-ends, where
14676 this option defaults to true and false respectively.
14680 This option generates traps for signed overflow on addition, subtraction,
14681 multiplication operations.
14685 This option instructs the compiler to assume that signed arithmetic
14686 overflow of addition, subtraction and multiplication wraps around
14687 using twos-complement representation. This flag enables some optimizations
14688 and disables others. This option is enabled by default for the Java
14689 front-end, as required by the Java language specification.
14692 @opindex fexceptions
14693 Enable exception handling. Generates extra code needed to propagate
14694 exceptions. For some targets, this implies GCC will generate frame
14695 unwind information for all functions, which can produce significant data
14696 size overhead, although it does not affect execution. If you do not
14697 specify this option, GCC will enable it by default for languages like
14698 C++ which normally require exception handling, and disable it for
14699 languages like C that do not normally require it. However, you may need
14700 to enable this option when compiling C code that needs to interoperate
14701 properly with exception handlers written in C++. You may also wish to
14702 disable this option if you are compiling older C++ programs that don't
14703 use exception handling.
14705 @item -fnon-call-exceptions
14706 @opindex fnon-call-exceptions
14707 Generate code that allows trapping instructions to throw exceptions.
14708 Note that this requires platform-specific runtime support that does
14709 not exist everywhere. Moreover, it only allows @emph{trapping}
14710 instructions to throw exceptions, i.e.@: memory references or floating
14711 point instructions. It does not allow exceptions to be thrown from
14712 arbitrary signal handlers such as @code{SIGALRM}.
14714 @item -funwind-tables
14715 @opindex funwind-tables
14716 Similar to @option{-fexceptions}, except that it will just generate any needed
14717 static data, but will not affect the generated code in any other way.
14718 You will normally not enable this option; instead, a language processor
14719 that needs this handling would enable it on your behalf.
14721 @item -fasynchronous-unwind-tables
14722 @opindex fasynchronous-unwind-tables
14723 Generate unwind table in dwarf2 format, if supported by target machine. The
14724 table is exact at each instruction boundary, so it can be used for stack
14725 unwinding from asynchronous events (such as debugger or garbage collector).
14727 @item -fpcc-struct-return
14728 @opindex fpcc-struct-return
14729 Return ``short'' @code{struct} and @code{union} values in memory like
14730 longer ones, rather than in registers. This convention is less
14731 efficient, but it has the advantage of allowing intercallability between
14732 GCC-compiled files and files compiled with other compilers, particularly
14733 the Portable C Compiler (pcc).
14735 The precise convention for returning structures in memory depends
14736 on the target configuration macros.
14738 Short structures and unions are those whose size and alignment match
14739 that of some integer type.
14741 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
14742 switch is not binary compatible with code compiled with the
14743 @option{-freg-struct-return} switch.
14744 Use it to conform to a non-default application binary interface.
14746 @item -freg-struct-return
14747 @opindex freg-struct-return
14748 Return @code{struct} and @code{union} values in registers when possible.
14749 This is more efficient for small structures than
14750 @option{-fpcc-struct-return}.
14752 If you specify neither @option{-fpcc-struct-return} nor
14753 @option{-freg-struct-return}, GCC defaults to whichever convention is
14754 standard for the target. If there is no standard convention, GCC
14755 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
14756 the principal compiler. In those cases, we can choose the standard, and
14757 we chose the more efficient register return alternative.
14759 @strong{Warning:} code compiled with the @option{-freg-struct-return}
14760 switch is not binary compatible with code compiled with the
14761 @option{-fpcc-struct-return} switch.
14762 Use it to conform to a non-default application binary interface.
14764 @item -fshort-enums
14765 @opindex fshort-enums
14766 Allocate to an @code{enum} type only as many bytes as it needs for the
14767 declared range of possible values. Specifically, the @code{enum} type
14768 will be equivalent to the smallest integer type which has enough room.
14770 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
14771 code that is not binary compatible with code generated without that switch.
14772 Use it to conform to a non-default application binary interface.
14774 @item -fshort-double
14775 @opindex fshort-double
14776 Use the same size for @code{double} as for @code{float}.
14778 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
14779 code that is not binary compatible with code generated without that switch.
14780 Use it to conform to a non-default application binary interface.
14782 @item -fshort-wchar
14783 @opindex fshort-wchar
14784 Override the underlying type for @samp{wchar_t} to be @samp{short
14785 unsigned int} instead of the default for the target. This option is
14786 useful for building programs to run under WINE@.
14788 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
14789 code that is not binary compatible with code generated without that switch.
14790 Use it to conform to a non-default application binary interface.
14793 @opindex fno-common
14794 In C, allocate even uninitialized global variables in the data section of the
14795 object file, rather than generating them as common blocks. This has the
14796 effect that if the same variable is declared (without @code{extern}) in
14797 two different compilations, you will get an error when you link them.
14798 The only reason this might be useful is if you wish to verify that the
14799 program will work on other systems which always work this way.
14803 Ignore the @samp{#ident} directive.
14805 @item -finhibit-size-directive
14806 @opindex finhibit-size-directive
14807 Don't output a @code{.size} assembler directive, or anything else that
14808 would cause trouble if the function is split in the middle, and the
14809 two halves are placed at locations far apart in memory. This option is
14810 used when compiling @file{crtstuff.c}; you should not need to use it
14813 @item -fverbose-asm
14814 @opindex fverbose-asm
14815 Put extra commentary information in the generated assembly code to
14816 make it more readable. This option is generally only of use to those
14817 who actually need to read the generated assembly code (perhaps while
14818 debugging the compiler itself).
14820 @option{-fno-verbose-asm}, the default, causes the
14821 extra information to be omitted and is useful when comparing two assembler
14824 @item -frecord-gcc-switches
14825 @opindex frecord-gcc-switches
14826 This switch causes the command line that was used to invoke the
14827 compiler to be recorded into the object file that is being created.
14828 This switch is only implemented on some targets and the exact format
14829 of the recording is target and binary file format dependent, but it
14830 usually takes the form of a section containing ASCII text. This
14831 switch is related to the @option{-fverbose-asm} switch, but that
14832 switch only records information in the assembler output file as
14833 comments, so it never reaches the object file.
14837 @cindex global offset table
14839 Generate position-independent code (PIC) suitable for use in a shared
14840 library, if supported for the target machine. Such code accesses all
14841 constant addresses through a global offset table (GOT)@. The dynamic
14842 loader resolves the GOT entries when the program starts (the dynamic
14843 loader is not part of GCC; it is part of the operating system). If
14844 the GOT size for the linked executable exceeds a machine-specific
14845 maximum size, you get an error message from the linker indicating that
14846 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14847 instead. (These maximums are 8k on the SPARC and 32k
14848 on the m68k and RS/6000. The 386 has no such limit.)
14850 Position-independent code requires special support, and therefore works
14851 only on certain machines. For the 386, GCC supports PIC for System V
14852 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14853 position-independent.
14855 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14860 If supported for the target machine, emit position-independent code,
14861 suitable for dynamic linking and avoiding any limit on the size of the
14862 global offset table. This option makes a difference on the m68k,
14863 PowerPC and SPARC@.
14865 Position-independent code requires special support, and therefore works
14866 only on certain machines.
14868 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14875 These options are similar to @option{-fpic} and @option{-fPIC}, but
14876 generated position independent code can be only linked into executables.
14877 Usually these options are used when @option{-pie} GCC option will be
14878 used during linking.
14880 @option{-fpie} and @option{-fPIE} both define the macros
14881 @code{__pie__} and @code{__PIE__}. The macros have the value 1
14882 for @option{-fpie} and 2 for @option{-fPIE}.
14884 @item -fno-jump-tables
14885 @opindex fno-jump-tables
14886 Do not use jump tables for switch statements even where it would be
14887 more efficient than other code generation strategies. This option is
14888 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14889 building code which forms part of a dynamic linker and cannot
14890 reference the address of a jump table. On some targets, jump tables
14891 do not require a GOT and this option is not needed.
14893 @item -ffixed-@var{reg}
14895 Treat the register named @var{reg} as a fixed register; generated code
14896 should never refer to it (except perhaps as a stack pointer, frame
14897 pointer or in some other fixed role).
14899 @var{reg} must be the name of a register. The register names accepted
14900 are machine-specific and are defined in the @code{REGISTER_NAMES}
14901 macro in the machine description macro file.
14903 This flag does not have a negative form, because it specifies a
14906 @item -fcall-used-@var{reg}
14907 @opindex fcall-used
14908 Treat the register named @var{reg} as an allocable register that is
14909 clobbered by function calls. It may be allocated for temporaries or
14910 variables that do not live across a call. Functions compiled this way
14911 will not save and restore the register @var{reg}.
14913 It is an error to used this flag with the frame pointer or stack pointer.
14914 Use of this flag for other registers that have fixed pervasive roles in
14915 the machine's execution model will produce disastrous results.
14917 This flag does not have a negative form, because it specifies a
14920 @item -fcall-saved-@var{reg}
14921 @opindex fcall-saved
14922 Treat the register named @var{reg} as an allocable register saved by
14923 functions. It may be allocated even for temporaries or variables that
14924 live across a call. Functions compiled this way will save and restore
14925 the register @var{reg} if they use it.
14927 It is an error to used this flag with the frame pointer or stack pointer.
14928 Use of this flag for other registers that have fixed pervasive roles in
14929 the machine's execution model will produce disastrous results.
14931 A different sort of disaster will result from the use of this flag for
14932 a register in which function values may be returned.
14934 This flag does not have a negative form, because it specifies a
14937 @item -fpack-struct[=@var{n}]
14938 @opindex fpack-struct
14939 Without a value specified, pack all structure members together without
14940 holes. When a value is specified (which must be a small power of two), pack
14941 structure members according to this value, representing the maximum
14942 alignment (that is, objects with default alignment requirements larger than
14943 this will be output potentially unaligned at the next fitting location.
14945 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14946 code that is not binary compatible with code generated without that switch.
14947 Additionally, it makes the code suboptimal.
14948 Use it to conform to a non-default application binary interface.
14950 @item -finstrument-functions
14951 @opindex finstrument-functions
14952 Generate instrumentation calls for entry and exit to functions. Just
14953 after function entry and just before function exit, the following
14954 profiling functions will be called with the address of the current
14955 function and its call site. (On some platforms,
14956 @code{__builtin_return_address} does not work beyond the current
14957 function, so the call site information may not be available to the
14958 profiling functions otherwise.)
14961 void __cyg_profile_func_enter (void *this_fn,
14963 void __cyg_profile_func_exit (void *this_fn,
14967 The first argument is the address of the start of the current function,
14968 which may be looked up exactly in the symbol table.
14970 This instrumentation is also done for functions expanded inline in other
14971 functions. The profiling calls will indicate where, conceptually, the
14972 inline function is entered and exited. This means that addressable
14973 versions of such functions must be available. If all your uses of a
14974 function are expanded inline, this may mean an additional expansion of
14975 code size. If you use @samp{extern inline} in your C code, an
14976 addressable version of such functions must be provided. (This is
14977 normally the case anyways, but if you get lucky and the optimizer always
14978 expands the functions inline, you might have gotten away without
14979 providing static copies.)
14981 A function may be given the attribute @code{no_instrument_function}, in
14982 which case this instrumentation will not be done. This can be used, for
14983 example, for the profiling functions listed above, high-priority
14984 interrupt routines, and any functions from which the profiling functions
14985 cannot safely be called (perhaps signal handlers, if the profiling
14986 routines generate output or allocate memory).
14988 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
14989 @opindex finstrument-functions-exclude-file-list
14991 Set the list of functions that are excluded from instrumentation (see
14992 the description of @code{-finstrument-functions}). If the file that
14993 contains a function definition matches with one of @var{file}, then
14994 that function is not instrumented. The match is done on substrings:
14995 if the @var{file} parameter is a substring of the file name, it is
14996 considered to be a match.
14999 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15000 will exclude any inline function defined in files whose pathnames
15001 contain @code{/bits/stl} or @code{include/sys}.
15003 If, for some reason, you want to include letter @code{','} in one of
15004 @var{sym}, write @code{'\,'}. For example,
15005 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15006 (note the single quote surrounding the option).
15008 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15009 @opindex finstrument-functions-exclude-function-list
15011 This is similar to @code{-finstrument-functions-exclude-file-list},
15012 but this option sets the list of function names to be excluded from
15013 instrumentation. The function name to be matched is its user-visible
15014 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15015 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15016 match is done on substrings: if the @var{sym} parameter is a substring
15017 of the function name, it is considered to be a match.
15019 @item -fstack-check
15020 @opindex fstack-check
15021 Generate code to verify that you do not go beyond the boundary of the
15022 stack. You should specify this flag if you are running in an
15023 environment with multiple threads, but only rarely need to specify it in
15024 a single-threaded environment since stack overflow is automatically
15025 detected on nearly all systems if there is only one stack.
15027 Note that this switch does not actually cause checking to be done; the
15028 operating system must do that. The switch causes generation of code
15029 to ensure that the operating system sees the stack being extended.
15031 @item -fstack-limit-register=@var{reg}
15032 @itemx -fstack-limit-symbol=@var{sym}
15033 @itemx -fno-stack-limit
15034 @opindex fstack-limit-register
15035 @opindex fstack-limit-symbol
15036 @opindex fno-stack-limit
15037 Generate code to ensure that the stack does not grow beyond a certain value,
15038 either the value of a register or the address of a symbol. If the stack
15039 would grow beyond the value, a signal is raised. For most targets,
15040 the signal is raised before the stack overruns the boundary, so
15041 it is possible to catch the signal without taking special precautions.
15043 For instance, if the stack starts at absolute address @samp{0x80000000}
15044 and grows downwards, you can use the flags
15045 @option{-fstack-limit-symbol=__stack_limit} and
15046 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15047 of 128KB@. Note that this may only work with the GNU linker.
15049 @cindex aliasing of parameters
15050 @cindex parameters, aliased
15051 @item -fargument-alias
15052 @itemx -fargument-noalias
15053 @itemx -fargument-noalias-global
15054 @itemx -fargument-noalias-anything
15055 @opindex fargument-alias
15056 @opindex fargument-noalias
15057 @opindex fargument-noalias-global
15058 @opindex fargument-noalias-anything
15059 Specify the possible relationships among parameters and between
15060 parameters and global data.
15062 @option{-fargument-alias} specifies that arguments (parameters) may
15063 alias each other and may alias global storage.@*
15064 @option{-fargument-noalias} specifies that arguments do not alias
15065 each other, but may alias global storage.@*
15066 @option{-fargument-noalias-global} specifies that arguments do not
15067 alias each other and do not alias global storage.
15068 @option{-fargument-noalias-anything} specifies that arguments do not
15069 alias any other storage.
15071 Each language will automatically use whatever option is required by
15072 the language standard. You should not need to use these options yourself.
15074 @item -fleading-underscore
15075 @opindex fleading-underscore
15076 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15077 change the way C symbols are represented in the object file. One use
15078 is to help link with legacy assembly code.
15080 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15081 generate code that is not binary compatible with code generated without that
15082 switch. Use it to conform to a non-default application binary interface.
15083 Not all targets provide complete support for this switch.
15085 @item -ftls-model=@var{model}
15086 @opindex ftls-model
15087 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15088 The @var{model} argument should be one of @code{global-dynamic},
15089 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15091 The default without @option{-fpic} is @code{initial-exec}; with
15092 @option{-fpic} the default is @code{global-dynamic}.
15094 @item -fvisibility=@var{default|internal|hidden|protected}
15095 @opindex fvisibility
15096 Set the default ELF image symbol visibility to the specified option---all
15097 symbols will be marked with this unless overridden within the code.
15098 Using this feature can very substantially improve linking and
15099 load times of shared object libraries, produce more optimized
15100 code, provide near-perfect API export and prevent symbol clashes.
15101 It is @strong{strongly} recommended that you use this in any shared objects
15104 Despite the nomenclature, @code{default} always means public ie;
15105 available to be linked against from outside the shared object.
15106 @code{protected} and @code{internal} are pretty useless in real-world
15107 usage so the only other commonly used option will be @code{hidden}.
15108 The default if @option{-fvisibility} isn't specified is
15109 @code{default}, i.e., make every
15110 symbol public---this causes the same behavior as previous versions of
15113 A good explanation of the benefits offered by ensuring ELF
15114 symbols have the correct visibility is given by ``How To Write
15115 Shared Libraries'' by Ulrich Drepper (which can be found at
15116 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
15117 solution made possible by this option to marking things hidden when
15118 the default is public is to make the default hidden and mark things
15119 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
15120 and @code{__attribute__ ((visibility("default")))} instead of
15121 @code{__declspec(dllexport)} you get almost identical semantics with
15122 identical syntax. This is a great boon to those working with
15123 cross-platform projects.
15125 For those adding visibility support to existing code, you may find
15126 @samp{#pragma GCC visibility} of use. This works by you enclosing
15127 the declarations you wish to set visibility for with (for example)
15128 @samp{#pragma GCC visibility push(hidden)} and
15129 @samp{#pragma GCC visibility pop}.
15130 Bear in mind that symbol visibility should be viewed @strong{as
15131 part of the API interface contract} and thus all new code should
15132 always specify visibility when it is not the default ie; declarations
15133 only for use within the local DSO should @strong{always} be marked explicitly
15134 as hidden as so to avoid PLT indirection overheads---making this
15135 abundantly clear also aids readability and self-documentation of the code.
15136 Note that due to ISO C++ specification requirements, operator new and
15137 operator delete must always be of default visibility.
15139 Be aware that headers from outside your project, in particular system
15140 headers and headers from any other library you use, may not be
15141 expecting to be compiled with visibility other than the default. You
15142 may need to explicitly say @samp{#pragma GCC visibility push(default)}
15143 before including any such headers.
15145 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
15146 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
15147 no modifications. However, this means that calls to @samp{extern}
15148 functions with no explicit visibility will use the PLT, so it is more
15149 effective to use @samp{__attribute ((visibility))} and/or
15150 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
15151 declarations should be treated as hidden.
15153 Note that @samp{-fvisibility} does affect C++ vague linkage
15154 entities. This means that, for instance, an exception class that will
15155 be thrown between DSOs must be explicitly marked with default
15156 visibility so that the @samp{type_info} nodes will be unified between
15159 An overview of these techniques, their benefits and how to use them
15160 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
15166 @node Environment Variables
15167 @section Environment Variables Affecting GCC
15168 @cindex environment variables
15170 @c man begin ENVIRONMENT
15171 This section describes several environment variables that affect how GCC
15172 operates. Some of them work by specifying directories or prefixes to use
15173 when searching for various kinds of files. Some are used to specify other
15174 aspects of the compilation environment.
15176 Note that you can also specify places to search using options such as
15177 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
15178 take precedence over places specified using environment variables, which
15179 in turn take precedence over those specified by the configuration of GCC@.
15180 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
15181 GNU Compiler Collection (GCC) Internals}.
15186 @c @itemx LC_COLLATE
15188 @c @itemx LC_MONETARY
15189 @c @itemx LC_NUMERIC
15194 @c @findex LC_COLLATE
15195 @findex LC_MESSAGES
15196 @c @findex LC_MONETARY
15197 @c @findex LC_NUMERIC
15201 These environment variables control the way that GCC uses
15202 localization information that allow GCC to work with different
15203 national conventions. GCC inspects the locale categories
15204 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
15205 so. These locale categories can be set to any value supported by your
15206 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
15207 Kingdom encoded in UTF-8.
15209 The @env{LC_CTYPE} environment variable specifies character
15210 classification. GCC uses it to determine the character boundaries in
15211 a string; this is needed for some multibyte encodings that contain quote
15212 and escape characters that would otherwise be interpreted as a string
15215 The @env{LC_MESSAGES} environment variable specifies the language to
15216 use in diagnostic messages.
15218 If the @env{LC_ALL} environment variable is set, it overrides the value
15219 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
15220 and @env{LC_MESSAGES} default to the value of the @env{LANG}
15221 environment variable. If none of these variables are set, GCC
15222 defaults to traditional C English behavior.
15226 If @env{TMPDIR} is set, it specifies the directory to use for temporary
15227 files. GCC uses temporary files to hold the output of one stage of
15228 compilation which is to be used as input to the next stage: for example,
15229 the output of the preprocessor, which is the input to the compiler
15232 @item GCC_EXEC_PREFIX
15233 @findex GCC_EXEC_PREFIX
15234 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
15235 names of the subprograms executed by the compiler. No slash is added
15236 when this prefix is combined with the name of a subprogram, but you can
15237 specify a prefix that ends with a slash if you wish.
15239 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
15240 an appropriate prefix to use based on the pathname it was invoked with.
15242 If GCC cannot find the subprogram using the specified prefix, it
15243 tries looking in the usual places for the subprogram.
15245 The default value of @env{GCC_EXEC_PREFIX} is
15246 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
15247 the installed compiler. In many cases @var{prefix} is the value
15248 of @code{prefix} when you ran the @file{configure} script.
15250 Other prefixes specified with @option{-B} take precedence over this prefix.
15252 This prefix is also used for finding files such as @file{crt0.o} that are
15255 In addition, the prefix is used in an unusual way in finding the
15256 directories to search for header files. For each of the standard
15257 directories whose name normally begins with @samp{/usr/local/lib/gcc}
15258 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
15259 replacing that beginning with the specified prefix to produce an
15260 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
15261 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
15262 These alternate directories are searched first; the standard directories
15263 come next. If a standard directory begins with the configured
15264 @var{prefix} then the value of @var{prefix} is replaced by
15265 @env{GCC_EXEC_PREFIX} when looking for header files.
15267 @item COMPILER_PATH
15268 @findex COMPILER_PATH
15269 The value of @env{COMPILER_PATH} is a colon-separated list of
15270 directories, much like @env{PATH}. GCC tries the directories thus
15271 specified when searching for subprograms, if it can't find the
15272 subprograms using @env{GCC_EXEC_PREFIX}.
15275 @findex LIBRARY_PATH
15276 The value of @env{LIBRARY_PATH} is a colon-separated list of
15277 directories, much like @env{PATH}. When configured as a native compiler,
15278 GCC tries the directories thus specified when searching for special
15279 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
15280 using GCC also uses these directories when searching for ordinary
15281 libraries for the @option{-l} option (but directories specified with
15282 @option{-L} come first).
15286 @cindex locale definition
15287 This variable is used to pass locale information to the compiler. One way in
15288 which this information is used is to determine the character set to be used
15289 when character literals, string literals and comments are parsed in C and C++.
15290 When the compiler is configured to allow multibyte characters,
15291 the following values for @env{LANG} are recognized:
15295 Recognize JIS characters.
15297 Recognize SJIS characters.
15299 Recognize EUCJP characters.
15302 If @env{LANG} is not defined, or if it has some other value, then the
15303 compiler will use mblen and mbtowc as defined by the default locale to
15304 recognize and translate multibyte characters.
15308 Some additional environments variables affect the behavior of the
15311 @include cppenv.texi
15315 @node Precompiled Headers
15316 @section Using Precompiled Headers
15317 @cindex precompiled headers
15318 @cindex speed of compilation
15320 Often large projects have many header files that are included in every
15321 source file. The time the compiler takes to process these header files
15322 over and over again can account for nearly all of the time required to
15323 build the project. To make builds faster, GCC allows users to
15324 `precompile' a header file; then, if builds can use the precompiled
15325 header file they will be much faster.
15327 To create a precompiled header file, simply compile it as you would any
15328 other file, if necessary using the @option{-x} option to make the driver
15329 treat it as a C or C++ header file. You will probably want to use a
15330 tool like @command{make} to keep the precompiled header up-to-date when
15331 the headers it contains change.
15333 A precompiled header file will be searched for when @code{#include} is
15334 seen in the compilation. As it searches for the included file
15335 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
15336 compiler looks for a precompiled header in each directory just before it
15337 looks for the include file in that directory. The name searched for is
15338 the name specified in the @code{#include} with @samp{.gch} appended. If
15339 the precompiled header file can't be used, it is ignored.
15341 For instance, if you have @code{#include "all.h"}, and you have
15342 @file{all.h.gch} in the same directory as @file{all.h}, then the
15343 precompiled header file will be used if possible, and the original
15344 header will be used otherwise.
15346 Alternatively, you might decide to put the precompiled header file in a
15347 directory and use @option{-I} to ensure that directory is searched
15348 before (or instead of) the directory containing the original header.
15349 Then, if you want to check that the precompiled header file is always
15350 used, you can put a file of the same name as the original header in this
15351 directory containing an @code{#error} command.
15353 This also works with @option{-include}. So yet another way to use
15354 precompiled headers, good for projects not designed with precompiled
15355 header files in mind, is to simply take most of the header files used by
15356 a project, include them from another header file, precompile that header
15357 file, and @option{-include} the precompiled header. If the header files
15358 have guards against multiple inclusion, they will be skipped because
15359 they've already been included (in the precompiled header).
15361 If you need to precompile the same header file for different
15362 languages, targets, or compiler options, you can instead make a
15363 @emph{directory} named like @file{all.h.gch}, and put each precompiled
15364 header in the directory, perhaps using @option{-o}. It doesn't matter
15365 what you call the files in the directory, every precompiled header in
15366 the directory will be considered. The first precompiled header
15367 encountered in the directory that is valid for this compilation will
15368 be used; they're searched in no particular order.
15370 There are many other possibilities, limited only by your imagination,
15371 good sense, and the constraints of your build system.
15373 A precompiled header file can be used only when these conditions apply:
15377 Only one precompiled header can be used in a particular compilation.
15380 A precompiled header can't be used once the first C token is seen. You
15381 can have preprocessor directives before a precompiled header; you can
15382 even include a precompiled header from inside another header, so long as
15383 there are no C tokens before the @code{#include}.
15386 The precompiled header file must be produced for the same language as
15387 the current compilation. You can't use a C precompiled header for a C++
15391 The precompiled header file must have been produced by the same compiler
15392 binary as the current compilation is using.
15395 Any macros defined before the precompiled header is included must
15396 either be defined in the same way as when the precompiled header was
15397 generated, or must not affect the precompiled header, which usually
15398 means that they don't appear in the precompiled header at all.
15400 The @option{-D} option is one way to define a macro before a
15401 precompiled header is included; using a @code{#define} can also do it.
15402 There are also some options that define macros implicitly, like
15403 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
15406 @item If debugging information is output when using the precompiled
15407 header, using @option{-g} or similar, the same kind of debugging information
15408 must have been output when building the precompiled header. However,
15409 a precompiled header built using @option{-g} can be used in a compilation
15410 when no debugging information is being output.
15412 @item The same @option{-m} options must generally be used when building
15413 and using the precompiled header. @xref{Submodel Options},
15414 for any cases where this rule is relaxed.
15416 @item Each of the following options must be the same when building and using
15417 the precompiled header:
15419 @gccoptlist{-fexceptions -funit-at-a-time}
15422 Some other command-line options starting with @option{-f},
15423 @option{-p}, or @option{-O} must be defined in the same way as when
15424 the precompiled header was generated. At present, it's not clear
15425 which options are safe to change and which are not; the safest choice
15426 is to use exactly the same options when generating and using the
15427 precompiled header. The following are known to be safe:
15429 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
15430 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
15431 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
15436 For all of these except the last, the compiler will automatically
15437 ignore the precompiled header if the conditions aren't met. If you
15438 find an option combination that doesn't work and doesn't cause the
15439 precompiled header to be ignored, please consider filing a bug report,
15442 If you do use differing options when generating and using the
15443 precompiled header, the actual behavior will be a mixture of the
15444 behavior for the options. For instance, if you use @option{-g} to
15445 generate the precompiled header but not when using it, you may or may
15446 not get debugging information for routines in the precompiled header.
15448 @node Running Protoize
15449 @section Running Protoize
15451 The program @code{protoize} is an optional part of GCC@. You can use
15452 it to add prototypes to a program, thus converting the program to ISO
15453 C in one respect. The companion program @code{unprotoize} does the
15454 reverse: it removes argument types from any prototypes that are found.
15456 When you run these programs, you must specify a set of source files as
15457 command line arguments. The conversion programs start out by compiling
15458 these files to see what functions they define. The information gathered
15459 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
15461 After scanning comes actual conversion. The specified files are all
15462 eligible to be converted; any files they include (whether sources or
15463 just headers) are eligible as well.
15465 But not all the eligible files are converted. By default,
15466 @code{protoize} and @code{unprotoize} convert only source and header
15467 files in the current directory. You can specify additional directories
15468 whose files should be converted with the @option{-d @var{directory}}
15469 option. You can also specify particular files to exclude with the
15470 @option{-x @var{file}} option. A file is converted if it is eligible, its
15471 directory name matches one of the specified directory names, and its
15472 name within the directory has not been excluded.
15474 Basic conversion with @code{protoize} consists of rewriting most
15475 function definitions and function declarations to specify the types of
15476 the arguments. The only ones not rewritten are those for varargs
15479 @code{protoize} optionally inserts prototype declarations at the
15480 beginning of the source file, to make them available for any calls that
15481 precede the function's definition. Or it can insert prototype
15482 declarations with block scope in the blocks where undeclared functions
15485 Basic conversion with @code{unprotoize} consists of rewriting most
15486 function declarations to remove any argument types, and rewriting
15487 function definitions to the old-style pre-ISO form.
15489 Both conversion programs print a warning for any function declaration or
15490 definition that they can't convert. You can suppress these warnings
15493 The output from @code{protoize} or @code{unprotoize} replaces the
15494 original source file. The original file is renamed to a name ending
15495 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
15496 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
15497 for DOS) file already exists, then the source file is simply discarded.
15499 @code{protoize} and @code{unprotoize} both depend on GCC itself to
15500 scan the program and collect information about the functions it uses.
15501 So neither of these programs will work until GCC is installed.
15503 Here is a table of the options you can use with @code{protoize} and
15504 @code{unprotoize}. Each option works with both programs unless
15508 @item -B @var{directory}
15509 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
15510 usual directory (normally @file{/usr/local/lib}). This file contains
15511 prototype information about standard system functions. This option
15512 applies only to @code{protoize}.
15514 @item -c @var{compilation-options}
15515 Use @var{compilation-options} as the options when running @command{gcc} to
15516 produce the @samp{.X} files. The special option @option{-aux-info} is
15517 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
15519 Note that the compilation options must be given as a single argument to
15520 @code{protoize} or @code{unprotoize}. If you want to specify several
15521 @command{gcc} options, you must quote the entire set of compilation options
15522 to make them a single word in the shell.
15524 There are certain @command{gcc} arguments that you cannot use, because they
15525 would produce the wrong kind of output. These include @option{-g},
15526 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
15527 the @var{compilation-options}, they are ignored.
15530 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
15531 systems) instead of @samp{.c}. This is convenient if you are converting
15532 a C program to C++. This option applies only to @code{protoize}.
15535 Add explicit global declarations. This means inserting explicit
15536 declarations at the beginning of each source file for each function
15537 that is called in the file and was not declared. These declarations
15538 precede the first function definition that contains a call to an
15539 undeclared function. This option applies only to @code{protoize}.
15541 @item -i @var{string}
15542 Indent old-style parameter declarations with the string @var{string}.
15543 This option applies only to @code{protoize}.
15545 @code{unprotoize} converts prototyped function definitions to old-style
15546 function definitions, where the arguments are declared between the
15547 argument list and the initial @samp{@{}. By default, @code{unprotoize}
15548 uses five spaces as the indentation. If you want to indent with just
15549 one space instead, use @option{-i " "}.
15552 Keep the @samp{.X} files. Normally, they are deleted after conversion
15556 Add explicit local declarations. @code{protoize} with @option{-l} inserts
15557 a prototype declaration for each function in each block which calls the
15558 function without any declaration. This option applies only to
15562 Make no real changes. This mode just prints information about the conversions
15563 that would have been done without @option{-n}.
15566 Make no @samp{.save} files. The original files are simply deleted.
15567 Use this option with caution.
15569 @item -p @var{program}
15570 Use the program @var{program} as the compiler. Normally, the name
15571 @file{gcc} is used.
15574 Work quietly. Most warnings are suppressed.
15577 Print the version number, just like @option{-v} for @command{gcc}.
15580 If you need special compiler options to compile one of your program's
15581 source files, then you should generate that file's @samp{.X} file
15582 specially, by running @command{gcc} on that source file with the
15583 appropriate options and the option @option{-aux-info}. Then run
15584 @code{protoize} on the entire set of files. @code{protoize} will use
15585 the existing @samp{.X} file because it is newer than the source file.
15589 gcc -Dfoo=bar file1.c -aux-info file1.X
15594 You need to include the special files along with the rest in the
15595 @code{protoize} command, even though their @samp{.X} files already
15596 exist, because otherwise they won't get converted.
15598 @xref{Protoize Caveats}, for more information on how to use
15599 @code{protoize} successfully.