1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-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-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
239 -Wformat-security -Wformat-y2k @gol
240 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
241 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
242 -Wimport -Wno-import -Winit-self -Winline @gol
243 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
244 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
245 -Wlogical-op -Wlong-long @gol
246 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
247 -Wmissing-format-attribute -Wmissing-include-dirs @gol
248 -Wmissing-noreturn -Wno-mudflap @gol
249 -Wno-multichar -Wnonnull -Wno-overflow @gol
250 -Woverlength-strings -Wpacked -Wpadded @gol
251 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
252 -Wredundant-decls @gol
253 -Wreturn-type -Wsequence-point -Wshadow @gol
254 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
255 -Wstrict-aliasing -Wstrict-aliasing=n @gol
256 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
257 -Wswitch -Wswitch-default -Wswitch-enum @gol
258 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
259 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
260 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
261 -Wunused-value -Wunused-variable @gol
262 -Wvariadic-macros -Wvla @gol
263 -Wvolatile-register-var -Wwrite-strings}
265 @item C and Objective-C-only Warning Options
266 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
267 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
268 -Wold-style-declaration -Wold-style-definition @gol
269 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
270 -Wdeclaration-after-statement -Wpointer-sign}
272 @item Debugging Options
273 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
274 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
275 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
276 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
277 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
278 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
280 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-optimized@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 -fno-merge-debug-strings -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}.
319 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
320 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
321 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
322 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
323 -fcheck-data-deps -fcprop-registers -fcrossjumping -fcse-follow-jumps @gol
324 -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
325 -fdata-sections -fdce -fdce @gol
326 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
327 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
328 -ffinite-math-only -ffloat-store -fforward-propagate @gol
329 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
330 -fgcse-sm -fif-conversion -fif-conversion2 -finline-functions @gol
331 -finline-functions-called-once -finline-limit=@var{n} @gol
332 -finline-small-functions -fipa-cp -fipa-marix-reorg -fipa-pta @gol
333 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
334 -fipa-type-escape -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
335 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
336 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
337 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
338 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
339 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
340 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
341 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
342 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
343 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
344 -fprofile-generate -fprofile-use -fprofile-values -freciprocal-math @gol
345 -fregmove -frename-registers -freorder-blocks @gol
346 -freorder-blocks-and-partition -freorder-functions @gol
347 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
348 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks @gol
349 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
350 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
351 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
352 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
353 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
354 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer -ftree-ccp @gol
355 -ftree-ch -ftree-copy-prop -ftree-copyrename -ftree-dce @gol
356 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
357 -ftree-loop-distribution @gol
358 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
359 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc -ftree-salias @gol
360 -ftree-sink -ftree-sra -ftree-store-ccp -ftree-ter @gol
361 -ftree-vect-loop-version -ftree-vectorize -ftree-vrp -funit-at-a-time @gol
362 -funroll-all-loops -funroll-loops -funsafe-loop-optimizations @gol
363 -funsafe-math-optimizations -funswitch-loops @gol
364 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
366 --param @var{name}=@var{value}
367 -O -O0 -O1 -O2 -O3 -Os}
369 @item Preprocessor Options
370 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
371 @gccoptlist{-A@var{question}=@var{answer} @gol
372 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
373 -C -dD -dI -dM -dN @gol
374 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
375 -idirafter @var{dir} @gol
376 -include @var{file} -imacros @var{file} @gol
377 -iprefix @var{file} -iwithprefix @var{dir} @gol
378 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
379 -imultilib @var{dir} -isysroot @var{dir} @gol
380 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
381 -P -fworking-directory -remap @gol
382 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
383 -Xpreprocessor @var{option}}
385 @item Assembler Option
386 @xref{Assembler Options,,Passing Options to the Assembler}.
387 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
390 @xref{Link Options,,Options for Linking}.
391 @gccoptlist{@var{object-file-name} -l@var{library} @gol
392 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
393 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
394 -Wl,@var{option} -Xlinker @var{option} @gol
397 @item Directory Options
398 @xref{Directory Options,,Options for Directory Search}.
399 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
400 -specs=@var{file} -I- --sysroot=@var{dir}}
403 @c I wrote this xref this way to avoid overfull hbox. -- rms
404 @xref{Target Options}.
405 @gccoptlist{-V @var{version} -b @var{machine}}
407 @item Machine Dependent Options
408 @xref{Submodel Options,,Hardware Models and Configurations}.
409 @c This list is ordered alphanumerically by subsection name.
410 @c Try and put the significant identifier (CPU or system) first,
411 @c so users have a clue at guessing where the ones they want will be.
414 @gccoptlist{-EB -EL @gol
415 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
416 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
419 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
420 -mabi=@var{name} @gol
421 -mapcs-stack-check -mno-apcs-stack-check @gol
422 -mapcs-float -mno-apcs-float @gol
423 -mapcs-reentrant -mno-apcs-reentrant @gol
424 -msched-prolog -mno-sched-prolog @gol
425 -mlittle-endian -mbig-endian -mwords-little-endian @gol
426 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
427 -mthumb-interwork -mno-thumb-interwork @gol
428 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
429 -mstructure-size-boundary=@var{n} @gol
430 -mabort-on-noreturn @gol
431 -mlong-calls -mno-long-calls @gol
432 -msingle-pic-base -mno-single-pic-base @gol
433 -mpic-register=@var{reg} @gol
434 -mnop-fun-dllimport @gol
435 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
436 -mpoke-function-name @gol
438 -mtpcs-frame -mtpcs-leaf-frame @gol
439 -mcaller-super-interworking -mcallee-super-interworking @gol
443 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
444 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
446 @emph{Blackfin Options}
447 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
448 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
449 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
450 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
451 -mno-id-shared-library -mshared-library-id=@var{n} @gol
452 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
453 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
454 -mfast-fp -minline-plt}
457 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
458 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
459 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
460 -mstack-align -mdata-align -mconst-align @gol
461 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
462 -melf -maout -melinux -mlinux -sim -sim2 @gol
463 -mmul-bug-workaround -mno-mul-bug-workaround}
466 @gccoptlist{-mmac -mpush-args}
468 @emph{Darwin Options}
469 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
470 -arch_only -bind_at_load -bundle -bundle_loader @gol
471 -client_name -compatibility_version -current_version @gol
473 -dependency-file -dylib_file -dylinker_install_name @gol
474 -dynamic -dynamiclib -exported_symbols_list @gol
475 -filelist -flat_namespace -force_cpusubtype_ALL @gol
476 -force_flat_namespace -headerpad_max_install_names @gol
478 -image_base -init -install_name -keep_private_externs @gol
479 -multi_module -multiply_defined -multiply_defined_unused @gol
480 -noall_load -no_dead_strip_inits_and_terms @gol
481 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
482 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
483 -private_bundle -read_only_relocs -sectalign @gol
484 -sectobjectsymbols -whyload -seg1addr @gol
485 -sectcreate -sectobjectsymbols -sectorder @gol
486 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
487 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
488 -segprot -segs_read_only_addr -segs_read_write_addr @gol
489 -single_module -static -sub_library -sub_umbrella @gol
490 -twolevel_namespace -umbrella -undefined @gol
491 -unexported_symbols_list -weak_reference_mismatches @gol
492 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
493 -mkernel -mone-byte-bool}
495 @emph{DEC Alpha Options}
496 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
497 -mieee -mieee-with-inexact -mieee-conformant @gol
498 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
499 -mtrap-precision=@var{mode} -mbuild-constants @gol
500 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
501 -mbwx -mmax -mfix -mcix @gol
502 -mfloat-vax -mfloat-ieee @gol
503 -mexplicit-relocs -msmall-data -mlarge-data @gol
504 -msmall-text -mlarge-text @gol
505 -mmemory-latency=@var{time}}
507 @emph{DEC Alpha/VMS Options}
508 @gccoptlist{-mvms-return-codes}
511 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
512 -mhard-float -msoft-float @gol
513 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
514 -mdouble -mno-double @gol
515 -mmedia -mno-media -mmuladd -mno-muladd @gol
516 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
517 -mlinked-fp -mlong-calls -malign-labels @gol
518 -mlibrary-pic -macc-4 -macc-8 @gol
519 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
520 -moptimize-membar -mno-optimize-membar @gol
521 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
522 -mvliw-branch -mno-vliw-branch @gol
523 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
524 -mno-nested-cond-exec -mtomcat-stats @gol
528 @emph{GNU/Linux Options}
529 @gccoptlist{-muclibc}
531 @emph{H8/300 Options}
532 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
535 @gccoptlist{-march=@var{architecture-type} @gol
536 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
537 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
538 -mfixed-range=@var{register-range} @gol
539 -mjump-in-delay -mlinker-opt -mlong-calls @gol
540 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
541 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
542 -mno-jump-in-delay -mno-long-load-store @gol
543 -mno-portable-runtime -mno-soft-float @gol
544 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
545 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
546 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
547 -munix=@var{unix-std} -nolibdld -static -threads}
549 @emph{i386 and x86-64 Options}
550 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
551 -mfpmath=@var{unit} @gol
552 -masm=@var{dialect} -mno-fancy-math-387 @gol
553 -mno-fp-ret-in-387 -msoft-float @gol
554 -mno-wide-multiply -mrtd -malign-double @gol
555 -mpreferred-stack-boundary=@var{num} -mcx16 -msahf -mrecip @gol
556 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 @gol
557 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
558 -mthreads -mno-align-stringops -minline-all-stringops @gol
559 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
560 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
561 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
562 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
563 -mcmodel=@var{code-model} @gol
564 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
565 -mfused-madd -mno-fused-madd}
568 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
569 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
570 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
571 -minline-float-divide-max-throughput @gol
572 -minline-int-divide-min-latency @gol
573 -minline-int-divide-max-throughput @gol
574 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
575 -mno-dwarf2-asm -mearly-stop-bits @gol
576 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
577 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
578 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
579 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
580 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
581 -mno-sched-prefer-non-data-spec-insns @gol
582 -mno-sched-prefer-non-control-spec-insns @gol
583 -mno-sched-count-spec-in-critical-path}
585 @emph{M32R/D Options}
586 @gccoptlist{-m32r2 -m32rx -m32r @gol
588 -malign-loops -mno-align-loops @gol
589 -missue-rate=@var{number} @gol
590 -mbranch-cost=@var{number} @gol
591 -mmodel=@var{code-size-model-type} @gol
592 -msdata=@var{sdata-type} @gol
593 -mno-flush-func -mflush-func=@var{name} @gol
594 -mno-flush-trap -mflush-trap=@var{number} @gol
598 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
600 @emph{M680x0 Options}
601 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
602 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
603 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
604 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
605 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
606 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
607 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
608 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
610 @emph{M68hc1x Options}
611 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
612 -mauto-incdec -minmax -mlong-calls -mshort @gol
613 -msoft-reg-count=@var{count}}
616 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
617 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
618 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
619 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
620 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
623 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
624 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
625 -mips16 -mno-mips16 -mflip-mips16 @gol
626 -minterlink-mips16 -mno-interlink-mips16 @gol
627 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
628 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
629 -mfp32 -mfp64 -mhard-float -msoft-float @gol
630 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
631 -msmartmips -mno-smartmips @gol
632 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
633 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
634 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
635 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
636 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
637 -membedded-data -mno-embedded-data @gol
638 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
639 -mcode-readable=@var{setting} @gol
640 -msplit-addresses -mno-split-addresses @gol
641 -mexplicit-relocs -mno-explicit-relocs @gol
642 -mcheck-zero-division -mno-check-zero-division @gol
643 -mdivide-traps -mdivide-breaks @gol
644 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
645 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
646 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
647 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130 @gol
648 -mfix-sb1 -mno-fix-sb1 @gol
649 -mflush-func=@var{func} -mno-flush-func @gol
650 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
651 -mfp-exceptions -mno-fp-exceptions @gol
652 -mvr4130-align -mno-vr4130-align}
655 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
656 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
657 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
658 -mno-base-addresses -msingle-exit -mno-single-exit}
660 @emph{MN10300 Options}
661 @gccoptlist{-mmult-bug -mno-mult-bug @gol
662 -mam33 -mno-am33 @gol
663 -mam33-2 -mno-am33-2 @gol
664 -mreturn-pointer-on-d0 @gol
668 @gccoptlist{-mno-crt0 -mbacc -msim @gol
669 -march=@var{cpu-type} }
671 @emph{PDP-11 Options}
672 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
673 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
674 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
675 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
676 -mbranch-expensive -mbranch-cheap @gol
677 -msplit -mno-split -munix-asm -mdec-asm}
679 @emph{PowerPC Options}
680 See RS/6000 and PowerPC Options.
682 @emph{RS/6000 and PowerPC Options}
683 @gccoptlist{-mcpu=@var{cpu-type} @gol
684 -mtune=@var{cpu-type} @gol
685 -mpower -mno-power -mpower2 -mno-power2 @gol
686 -mpowerpc -mpowerpc64 -mno-powerpc @gol
687 -maltivec -mno-altivec @gol
688 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
689 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
690 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
691 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
692 -mnew-mnemonics -mold-mnemonics @gol
693 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
694 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
695 -malign-power -malign-natural @gol
696 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
697 -mstring -mno-string -mupdate -mno-update @gol
698 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
699 -mstrict-align -mno-strict-align -mrelocatable @gol
700 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
701 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
702 -mdynamic-no-pic -maltivec -mswdiv @gol
703 -mprioritize-restricted-insns=@var{priority} @gol
704 -msched-costly-dep=@var{dependence_type} @gol
705 -minsert-sched-nops=@var{scheme} @gol
706 -mcall-sysv -mcall-netbsd @gol
707 -maix-struct-return -msvr4-struct-return @gol
708 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
709 -misel -mno-isel @gol
710 -misel=yes -misel=no @gol
712 -mspe=yes -mspe=no @gol
714 -mvrsave -mno-vrsave @gol
715 -mmulhw -mno-mulhw @gol
716 -mdlmzb -mno-dlmzb @gol
717 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
718 -mprototype -mno-prototype @gol
719 -msim -mmvme -mads -myellowknife -memb -msdata @gol
720 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
722 @emph{S/390 and zSeries Options}
723 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
724 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
725 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
726 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
727 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
728 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
729 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
732 @gccoptlist{-meb -mel @gol
736 -mscore5 -mscore5u -mscore7 -mscore7d}
739 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
740 -m4-nofpu -m4-single-only -m4-single -m4 @gol
741 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
742 -m5-64media -m5-64media-nofpu @gol
743 -m5-32media -m5-32media-nofpu @gol
744 -m5-compact -m5-compact-nofpu @gol
745 -mb -ml -mdalign -mrelax @gol
746 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
747 -mieee -misize -minline-ic_invalidate -mpadstruct -mspace @gol
748 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
749 -mdivsi3_libfunc=@var{name} @gol
750 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
754 @gccoptlist{-mcpu=@var{cpu-type} @gol
755 -mtune=@var{cpu-type} @gol
756 -mcmodel=@var{code-model} @gol
757 -m32 -m64 -mapp-regs -mno-app-regs @gol
758 -mfaster-structs -mno-faster-structs @gol
759 -mfpu -mno-fpu -mhard-float -msoft-float @gol
760 -mhard-quad-float -msoft-quad-float @gol
761 -mimpure-text -mno-impure-text -mlittle-endian @gol
762 -mstack-bias -mno-stack-bias @gol
763 -munaligned-doubles -mno-unaligned-doubles @gol
764 -mv8plus -mno-v8plus -mvis -mno-vis
765 -threads -pthreads -pthread}
768 @gccoptlist{-mwarn-reloc -merror-reloc @gol
769 -msafe-dma -munsafe-dma @gol
771 -msmall-mem -mlarge-mem -mstdmain @gol
772 -mfixed-range=@var{register-range}}
774 @emph{System V Options}
775 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
778 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
779 -mprolog-function -mno-prolog-function -mspace @gol
780 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
781 -mapp-regs -mno-app-regs @gol
782 -mdisable-callt -mno-disable-callt @gol
788 @gccoptlist{-mg -mgnu -munix}
790 @emph{VxWorks Options}
791 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
792 -Xbind-lazy -Xbind-now}
794 @emph{x86-64 Options}
795 See i386 and x86-64 Options.
797 @emph{Xstormy16 Options}
800 @emph{Xtensa Options}
801 @gccoptlist{-mconst16 -mno-const16 @gol
802 -mfused-madd -mno-fused-madd @gol
803 -mserialize-volatile -mno-serialize-volatile @gol
804 -mtext-section-literals -mno-text-section-literals @gol
805 -mtarget-align -mno-target-align @gol
806 -mlongcalls -mno-longcalls}
808 @emph{zSeries Options}
809 See S/390 and zSeries Options.
811 @item Code Generation Options
812 @xref{Code Gen Options,,Options for Code Generation Conventions}.
813 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
814 -ffixed-@var{reg} -fexceptions @gol
815 -fnon-call-exceptions -funwind-tables @gol
816 -fasynchronous-unwind-tables @gol
817 -finhibit-size-directive -finstrument-functions @gol
818 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
819 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
820 -fno-common -fno-ident @gol
821 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
822 -fno-jump-tables @gol
823 -frecord-gcc-switches @gol
824 -freg-struct-return -fshort-enums @gol
825 -fshort-double -fshort-wchar @gol
826 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
827 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
828 -fno-stack-limit -fargument-alias -fargument-noalias @gol
829 -fargument-noalias-global -fargument-noalias-anything @gol
830 -fleading-underscore -ftls-model=@var{model} @gol
831 -ftrapv -fwrapv -fbounds-check @gol
836 * Overall Options:: Controlling the kind of output:
837 an executable, object files, assembler files,
838 or preprocessed source.
839 * C Dialect Options:: Controlling the variant of C language compiled.
840 * C++ Dialect Options:: Variations on C++.
841 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
843 * Language Independent Options:: Controlling how diagnostics should be
845 * Warning Options:: How picky should the compiler be?
846 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
847 * Optimize Options:: How much optimization?
848 * Preprocessor Options:: Controlling header files and macro definitions.
849 Also, getting dependency information for Make.
850 * Assembler Options:: Passing options to the assembler.
851 * Link Options:: Specifying libraries and so on.
852 * Directory Options:: Where to find header files and libraries.
853 Where to find the compiler executable files.
854 * Spec Files:: How to pass switches to sub-processes.
855 * Target Options:: Running a cross-compiler, or an old version of GCC.
858 @node Overall Options
859 @section Options Controlling the Kind of Output
861 Compilation can involve up to four stages: preprocessing, compilation
862 proper, assembly and linking, always in that order. GCC is capable of
863 preprocessing and compiling several files either into several
864 assembler input files, or into one assembler input file; then each
865 assembler input file produces an object file, and linking combines all
866 the object files (those newly compiled, and those specified as input)
867 into an executable file.
869 @cindex file name suffix
870 For any given input file, the file name suffix determines what kind of
875 C source code which must be preprocessed.
878 C source code which should not be preprocessed.
881 C++ source code which should not be preprocessed.
884 Objective-C source code. Note that you must link with the @file{libobjc}
885 library to make an Objective-C program work.
888 Objective-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. Note that @samp{.M} refers
894 to a literal capital M@.
897 Objective-C++ source code which should not be preprocessed.
900 C, C++, Objective-C or Objective-C++ header file to be turned into a
905 @itemx @var{file}.cxx
906 @itemx @var{file}.cpp
907 @itemx @var{file}.CPP
908 @itemx @var{file}.c++
910 C++ source code which must be preprocessed. Note that in @samp{.cxx},
911 the last two letters must both be literally @samp{x}. Likewise,
912 @samp{.C} refers to a literal capital C@.
916 Objective-C++ source code which must be preprocessed.
919 Objective-C++ source code which should not be preprocessed.
924 @itemx @var{file}.hxx
925 @itemx @var{file}.hpp
926 @itemx @var{file}.HPP
927 @itemx @var{file}.h++
928 @itemx @var{file}.tcc
929 C++ header file to be turned into a precompiled header.
932 @itemx @var{file}.for
933 @itemx @var{file}.FOR
934 Fixed form Fortran source code which should not be preprocessed.
937 @itemx @var{file}.fpp
938 @itemx @var{file}.FPP
939 Fixed form Fortran source code which must be preprocessed (with the traditional
943 @itemx @var{file}.f95
944 Free form Fortran source code which should not be preprocessed.
947 @itemx @var{file}.F95
948 Free form Fortran source code which must be preprocessed (with the
949 traditional preprocessor).
951 @c FIXME: Descriptions of Java file types.
958 Ada source code file which contains a library unit declaration (a
959 declaration of a package, subprogram, or generic, or a generic
960 instantiation), or a library unit renaming declaration (a package,
961 generic, or subprogram renaming declaration). Such files are also
964 @itemx @var{file}.adb
965 Ada source code file containing a library unit body (a subprogram or
966 package body). Such files are also called @dfn{bodies}.
968 @c GCC also knows about some suffixes for languages not yet included:
980 Assembler code which must be preprocessed.
983 An object file to be fed straight into linking.
984 Any file name with no recognized suffix is treated this way.
988 You can specify the input language explicitly with the @option{-x} option:
991 @item -x @var{language}
992 Specify explicitly the @var{language} for the following input files
993 (rather than letting the compiler choose a default based on the file
994 name suffix). This option applies to all following input files until
995 the next @option{-x} option. Possible values for @var{language} are:
997 c c-header c-cpp-output
998 c++ c++-header c++-cpp-output
999 objective-c objective-c-header objective-c-cpp-output
1000 objective-c++ objective-c++-header objective-c++-cpp-output
1001 assembler assembler-with-cpp
1008 Turn off any specification of a language, so that subsequent files are
1009 handled according to their file name suffixes (as they are if @option{-x}
1010 has not been used at all).
1012 @item -pass-exit-codes
1013 @opindex pass-exit-codes
1014 Normally the @command{gcc} program will exit with the code of 1 if any
1015 phase of the compiler returns a non-success return code. If you specify
1016 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1017 numerically highest error produced by any phase that returned an error
1018 indication. The C, C++, and Fortran frontends return 4, if an internal
1019 compiler error is encountered.
1022 If you only want some of the stages of compilation, you can use
1023 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1024 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1025 @command{gcc} is to stop. Note that some combinations (for example,
1026 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1031 Compile or assemble the source files, but do not link. The linking
1032 stage simply is not done. The ultimate output is in the form of an
1033 object file for each source file.
1035 By default, the object file name for a source file is made by replacing
1036 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1038 Unrecognized input files, not requiring compilation or assembly, are
1043 Stop after the stage of compilation proper; do not assemble. The output
1044 is in the form of an assembler code file for each non-assembler input
1047 By default, the assembler file name for a source file is made by
1048 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1050 Input files that don't require compilation are ignored.
1054 Stop after the preprocessing stage; do not run the compiler proper. The
1055 output is in the form of preprocessed source code, which is sent to the
1058 Input files which don't require preprocessing are ignored.
1060 @cindex output file option
1063 Place output in file @var{file}. This applies regardless to whatever
1064 sort of output is being produced, whether it be an executable file,
1065 an object file, an assembler file or preprocessed C code.
1067 If @option{-o} is not specified, the default is to put an executable
1068 file in @file{a.out}, the object file for
1069 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1070 assembler file in @file{@var{source}.s}, a precompiled header file in
1071 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1076 Print (on standard error output) the commands executed to run the stages
1077 of compilation. Also print the version number of the compiler driver
1078 program and of the preprocessor and the compiler proper.
1082 Like @option{-v} except the commands are not executed and all command
1083 arguments are quoted. This is useful for shell scripts to capture the
1084 driver-generated command lines.
1088 Use pipes rather than temporary files for communication between the
1089 various stages of compilation. This fails to work on some systems where
1090 the assembler is unable to read from a pipe; but the GNU assembler has
1095 If you are compiling multiple source files, this option tells the driver
1096 to pass all the source files to the compiler at once (for those
1097 languages for which the compiler can handle this). This will allow
1098 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1099 language for which this is supported is C@. If you pass source files for
1100 multiple languages to the driver, using this option, the driver will invoke
1101 the compiler(s) that support IMA once each, passing each compiler all the
1102 source files appropriate for it. For those languages that do not support
1103 IMA this option will be ignored, and the compiler will be invoked once for
1104 each source file in that language. If you use this option in conjunction
1105 with @option{-save-temps}, the compiler will generate multiple
1107 (one for each source file), but only one (combined) @file{.o} or
1112 Print (on the standard output) a description of the command line options
1113 understood by @command{gcc}. If the @option{-v} option is also specified
1114 then @option{--help} will also be passed on to the various processes
1115 invoked by @command{gcc}, so that they can display the command line options
1116 they accept. If the @option{-Wextra} option has also been specified
1117 (prior to the @option{--help} option), then command line options which
1118 have no documentation associated with them will also be displayed.
1121 @opindex target-help
1122 Print (on the standard output) a description of target-specific command
1123 line options for each tool. For some targets extra target-specific
1124 information may also be printed.
1126 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1127 Print (on the standard output) a description of the command line
1128 options understood by the compiler that fit into a specific class.
1129 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1130 @samp{params}, or @var{language}:
1133 @item @samp{optimizers}
1134 This will display all of the optimization options supported by the
1137 @item @samp{warnings}
1138 This will display all of the options controlling warning messages
1139 produced by the compiler.
1142 This will display target-specific options. Unlike the
1143 @option{--target-help} option however, target-specific options of the
1144 linker and assembler will not be displayed. This is because those
1145 tools do not currently support the extended @option{--help=} syntax.
1148 This will display the values recognized by the @option{--param}
1151 @item @var{language}
1152 This will display the options supported for @var{language}, where
1153 @var{language} is the name of one of the languages supported in this
1157 This will display the options that are common to all languages.
1160 It is possible to further refine the output of the @option{--help=}
1161 option by adding a comma separated list of qualifiers after the
1162 class. These can be any from the following list:
1165 @item @samp{undocumented}
1166 Display only those options which are undocumented.
1169 Display options which take an argument that appears after an equal
1170 sign in the same continuous piece of text, such as:
1171 @samp{--help=target}.
1173 @item @samp{separate}
1174 Display options which take an argument that appears as a separate word
1175 following the original option, such as: @samp{-o output-file}.
1178 Thus for example to display all the undocumented target-specific
1179 switches supported by the compiler the following can be used:
1182 --help=target,undocumented
1185 The sense of a qualifier can be inverted by prefixing it with the
1186 @var{^} character, so for example to display all binary warning
1187 options (i.e., ones that are either on or off and that do not take an
1188 argument), which have a description the following can be used:
1191 --help=warnings,^joined,^undocumented
1194 A class can also be used as a qualifier, although this usually
1195 restricts the output by so much that there is nothing to display. One
1196 case where it does work however is when one of the classes is
1197 @var{target}. So for example to display all the target-specific
1198 optimization options the following can be used:
1201 --help=target,optimizers
1204 The @option{--help=} option can be repeated on the command line. Each
1205 successive use will display its requested class of options, skipping
1206 those that have already been displayed.
1208 If the @option{-Q} option appears on the command line before the
1209 @option{--help=} option, then the descriptive text displayed by
1210 @option{--help=} is changed. Instead of describing the displayed
1211 options, an indication is given as to whether the option is enabled,
1212 disabled or set to a specific value (assuming that the compiler
1213 knows this at the point where the @option{--help=} option is used).
1215 Here is a truncated example from the ARM port of @command{gcc}:
1218 % gcc -Q -mabi=2 --help=target -c
1219 The following options are target specific:
1221 -mabort-on-noreturn [disabled]
1225 The output is sensitive to the effects of previous command line
1226 options, so for example it is possible to find out which optimizations
1227 are enabled at @option{-O2} by using:
1230 -O2 --help=optimizers
1233 Alternatively you can discover which binary optimizations are enabled
1234 by @option{-O3} by using:
1237 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1238 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1239 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1244 Display the version number and copyrights of the invoked GCC@.
1246 @include @value{srcdir}/../libiberty/at-file.texi
1250 @section Compiling C++ Programs
1252 @cindex suffixes for C++ source
1253 @cindex C++ source file suffixes
1254 C++ source files conventionally use one of the suffixes @samp{.C},
1255 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1256 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1257 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1258 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1259 files with these names and compiles them as C++ programs even if you
1260 call the compiler the same way as for compiling C programs (usually
1261 with the name @command{gcc}).
1265 However, the use of @command{gcc} does not add the C++ library.
1266 @command{g++} is a program that calls GCC and treats @samp{.c},
1267 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1268 files unless @option{-x} is used, and automatically specifies linking
1269 against the C++ library. This program is also useful when
1270 precompiling a C header file with a @samp{.h} extension for use in C++
1271 compilations. On many systems, @command{g++} is also installed with
1272 the name @command{c++}.
1274 @cindex invoking @command{g++}
1275 When you compile C++ programs, you may specify many of the same
1276 command-line options that you use for compiling programs in any
1277 language; or command-line options meaningful for C and related
1278 languages; or options that are meaningful only for C++ programs.
1279 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1280 explanations of options for languages related to C@.
1281 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1282 explanations of options that are meaningful only for C++ programs.
1284 @node C Dialect Options
1285 @section Options Controlling C Dialect
1286 @cindex dialect options
1287 @cindex language dialect options
1288 @cindex options, dialect
1290 The following options control the dialect of C (or languages derived
1291 from C, such as C++, Objective-C and Objective-C++) that the compiler
1295 @cindex ANSI support
1299 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1300 equivalent to @samp{-std=c++98}.
1302 This turns off certain features of GCC that are incompatible with ISO
1303 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1304 such as the @code{asm} and @code{typeof} keywords, and
1305 predefined macros such as @code{unix} and @code{vax} that identify the
1306 type of system you are using. It also enables the undesirable and
1307 rarely used ISO trigraph feature. For the C compiler,
1308 it disables recognition of C++ style @samp{//} comments as well as
1309 the @code{inline} keyword.
1311 The alternate keywords @code{__asm__}, @code{__extension__},
1312 @code{__inline__} and @code{__typeof__} continue to work despite
1313 @option{-ansi}. You would not want to use them in an ISO C program, of
1314 course, but it is useful to put them in header files that might be included
1315 in compilations done with @option{-ansi}. Alternate predefined macros
1316 such as @code{__unix__} and @code{__vax__} are also available, with or
1317 without @option{-ansi}.
1319 The @option{-ansi} option does not cause non-ISO programs to be
1320 rejected gratuitously. For that, @option{-pedantic} is required in
1321 addition to @option{-ansi}. @xref{Warning Options}.
1323 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1324 option is used. Some header files may notice this macro and refrain
1325 from declaring certain functions or defining certain macros that the
1326 ISO standard doesn't call for; this is to avoid interfering with any
1327 programs that might use these names for other things.
1329 Functions that would normally be built in but do not have semantics
1330 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1331 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1332 built-in functions provided by GCC}, for details of the functions
1337 Determine the language standard. @xref{Standards,,Language Standards
1338 Supported by GCC}, for details of these standard versions. This option
1339 is currently only supported when compiling C or C++.
1341 The compiler can accept several base standards, such as @samp{c89} or
1342 @samp{c++98}, and GNU dialects of those standards, such as
1343 @samp{gnu89} or @samp{gnu++98}. By specifing a base standard, the
1344 compiler will accept all programs following that standard and those
1345 using GNU extensions that do not contradict it. For example,
1346 @samp{-std=c89} turns off certain features of GCC that are
1347 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1348 keywords, but not other GNU extensions that do not have a meaning in
1349 ISO C90, such as omitting the middle term of a @code{?:}
1350 expression. On the other hand, by specifing a GNU dialect of a
1351 standard, all features the compiler support are enabled, even when
1352 those features change the meaning of the base standard and some
1353 strict-conforming programs may be rejected. The particular standard
1354 is used by @option{-pedantic} to identify which features are GNU
1355 extensions given that version of the standard. For example
1356 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1357 comments, while @samp{-std=gnu99 -pedantic} would not.
1359 A value for this option must be provided; possible values are
1364 Support all ISO C90 programs (certain GNU extensions that conflict
1365 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1367 @item iso9899:199409
1368 ISO C90 as modified in amendment 1.
1374 ISO C99. Note that this standard is not yet fully supported; see
1375 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1376 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1379 GNU dialect of ISO C90 (including some C99 features). This
1380 is the default for C code.
1384 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1385 this will become the default. The name @samp{gnu9x} is deprecated.
1388 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1392 GNU dialect of @option{-std=c++98}. This is the default for
1396 The working draft of the upcoming ISO C++0x standard. This option
1397 enables experimental features that are likely to be included in
1398 C++0x. The working draft is constantly changing, and any feature that is
1399 enabled by this flag may be removed from future versions of GCC if it is
1400 not part of the C++0x standard.
1403 GNU dialect of @option{-std=c++0x}. This option enables
1404 experimental features that may be removed in future versions of GCC.
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/}}. This option
1528 implies @option{-pthread}, and thus is only supported on targets that
1529 have support for @option{-pthread}.
1531 @item -fms-extensions
1532 @opindex fms-extensions
1533 Accept some non-standard constructs used in Microsoft header files.
1535 Some cases of unnamed fields in structures and unions are only
1536 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1537 fields within structs/unions}, for details.
1541 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1542 options for strict ISO C conformance) implies @option{-trigraphs}.
1544 @item -no-integrated-cpp
1545 @opindex no-integrated-cpp
1546 Performs a compilation in two passes: preprocessing and compiling. This
1547 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1548 @option{-B} option. The user supplied compilation step can then add in
1549 an additional preprocessing step after normal preprocessing but before
1550 compiling. The default is to use the integrated cpp (internal cpp)
1552 The semantics of this option will change if "cc1", "cc1plus", and
1553 "cc1obj" are merged.
1555 @cindex traditional C language
1556 @cindex C language, traditional
1558 @itemx -traditional-cpp
1559 @opindex traditional-cpp
1560 @opindex traditional
1561 Formerly, these options caused GCC to attempt to emulate a pre-standard
1562 C compiler. They are now only supported with the @option{-E} switch.
1563 The preprocessor continues to support a pre-standard mode. See the GNU
1564 CPP manual for details.
1566 @item -fcond-mismatch
1567 @opindex fcond-mismatch
1568 Allow conditional expressions with mismatched types in the second and
1569 third arguments. The value of such an expression is void. This option
1570 is not supported for C++.
1572 @item -flax-vector-conversions
1573 @opindex flax-vector-conversions
1574 Allow implicit conversions between vectors with differing numbers of
1575 elements and/or incompatible element types. This option should not be
1578 @item -funsigned-char
1579 @opindex funsigned-char
1580 Let the type @code{char} be unsigned, like @code{unsigned char}.
1582 Each kind of machine has a default for what @code{char} should
1583 be. It is either like @code{unsigned char} by default or like
1584 @code{signed char} by default.
1586 Ideally, a portable program should always use @code{signed char} or
1587 @code{unsigned char} when it depends on the signedness of an object.
1588 But many programs have been written to use plain @code{char} and
1589 expect it to be signed, or expect it to be unsigned, depending on the
1590 machines they were written for. This option, and its inverse, let you
1591 make such a program work with the opposite default.
1593 The type @code{char} is always a distinct type from each of
1594 @code{signed char} or @code{unsigned char}, even though its behavior
1595 is always just like one of those two.
1598 @opindex fsigned-char
1599 Let the type @code{char} be signed, like @code{signed char}.
1601 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1602 the negative form of @option{-funsigned-char}. Likewise, the option
1603 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1605 @item -fsigned-bitfields
1606 @itemx -funsigned-bitfields
1607 @itemx -fno-signed-bitfields
1608 @itemx -fno-unsigned-bitfields
1609 @opindex fsigned-bitfields
1610 @opindex funsigned-bitfields
1611 @opindex fno-signed-bitfields
1612 @opindex fno-unsigned-bitfields
1613 These options control whether a bit-field is signed or unsigned, when the
1614 declaration does not use either @code{signed} or @code{unsigned}. By
1615 default, such a bit-field is signed, because this is consistent: the
1616 basic integer types such as @code{int} are signed types.
1619 @node C++ Dialect Options
1620 @section Options Controlling C++ Dialect
1622 @cindex compiler options, C++
1623 @cindex C++ options, command line
1624 @cindex options, C++
1625 This section describes the command-line options that are only meaningful
1626 for C++ programs; but you can also use most of the GNU compiler options
1627 regardless of what language your program is in. For example, you
1628 might compile a file @code{firstClass.C} like this:
1631 g++ -g -frepo -O -c firstClass.C
1635 In this example, only @option{-frepo} is an option meant
1636 only for C++ programs; you can use the other options with any
1637 language supported by GCC@.
1639 Here is a list of options that are @emph{only} for compiling C++ programs:
1643 @item -fabi-version=@var{n}
1644 @opindex fabi-version
1645 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1646 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1647 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1648 the version that conforms most closely to the C++ ABI specification.
1649 Therefore, the ABI obtained using version 0 will change as ABI bugs
1652 The default is version 2.
1654 @item -fno-access-control
1655 @opindex fno-access-control
1656 Turn off all access checking. This switch is mainly useful for working
1657 around bugs in the access control code.
1661 Check that the pointer returned by @code{operator new} is non-null
1662 before attempting to modify the storage allocated. This check is
1663 normally unnecessary because the C++ standard specifies that
1664 @code{operator new} will only return @code{0} if it is declared
1665 @samp{throw()}, in which case the compiler will always check the
1666 return value even without this option. In all other cases, when
1667 @code{operator new} has a non-empty exception specification, memory
1668 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1669 @samp{new (nothrow)}.
1671 @item -fconserve-space
1672 @opindex fconserve-space
1673 Put uninitialized or runtime-initialized global variables into the
1674 common segment, as C does. This saves space in the executable at the
1675 cost of not diagnosing duplicate definitions. If you compile with this
1676 flag and your program mysteriously crashes after @code{main()} has
1677 completed, you may have an object that is being destroyed twice because
1678 two definitions were merged.
1680 This option is no longer useful on most targets, now that support has
1681 been added for putting variables into BSS without making them common.
1683 @item -ffriend-injection
1684 @opindex ffriend-injection
1685 Inject friend functions into the enclosing namespace, so that they are
1686 visible outside the scope of the class in which they are declared.
1687 Friend functions were documented to work this way in the old Annotated
1688 C++ Reference Manual, and versions of G++ before 4.1 always worked
1689 that way. However, in ISO C++ a friend function which is not declared
1690 in an enclosing scope can only be found using argument dependent
1691 lookup. This option causes friends to be injected as they were in
1694 This option is for compatibility, and may be removed in a future
1697 @item -fno-elide-constructors
1698 @opindex fno-elide-constructors
1699 The C++ standard allows an implementation to omit creating a temporary
1700 which is only used to initialize another object of the same type.
1701 Specifying this option disables that optimization, and forces G++ to
1702 call the copy constructor in all cases.
1704 @item -fno-enforce-eh-specs
1705 @opindex fno-enforce-eh-specs
1706 Don't generate code to check for violation of exception specifications
1707 at runtime. This option violates the C++ standard, but may be useful
1708 for reducing code size in production builds, much like defining
1709 @samp{NDEBUG}. This does not give user code permission to throw
1710 exceptions in violation of the exception specifications; the compiler
1711 will still optimize based on the specifications, so throwing an
1712 unexpected exception will result in undefined behavior.
1715 @itemx -fno-for-scope
1717 @opindex fno-for-scope
1718 If @option{-ffor-scope} is specified, the scope of variables declared in
1719 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1720 as specified by the C++ standard.
1721 If @option{-fno-for-scope} is specified, the scope of variables declared in
1722 a @i{for-init-statement} extends to the end of the enclosing scope,
1723 as was the case in old versions of G++, and other (traditional)
1724 implementations of C++.
1726 The default if neither flag is given to follow the standard,
1727 but to allow and give a warning for old-style code that would
1728 otherwise be invalid, or have different behavior.
1730 @item -fno-gnu-keywords
1731 @opindex fno-gnu-keywords
1732 Do not recognize @code{typeof} as a keyword, so that code can use this
1733 word as an identifier. You can use the keyword @code{__typeof__} instead.
1734 @option{-ansi} implies @option{-fno-gnu-keywords}.
1736 @item -fno-implicit-templates
1737 @opindex fno-implicit-templates
1738 Never emit code for non-inline templates which are instantiated
1739 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1740 @xref{Template Instantiation}, for more information.
1742 @item -fno-implicit-inline-templates
1743 @opindex fno-implicit-inline-templates
1744 Don't emit code for implicit instantiations of inline templates, either.
1745 The default is to handle inlines differently so that compiles with and
1746 without optimization will need the same set of explicit instantiations.
1748 @item -fno-implement-inlines
1749 @opindex fno-implement-inlines
1750 To save space, do not emit out-of-line copies of inline functions
1751 controlled by @samp{#pragma implementation}. This will cause linker
1752 errors if these functions are not inlined everywhere they are called.
1754 @item -fms-extensions
1755 @opindex fms-extensions
1756 Disable pedantic warnings about constructs used in MFC, such as implicit
1757 int and getting a pointer to member function via non-standard syntax.
1759 @item -fno-nonansi-builtins
1760 @opindex fno-nonansi-builtins
1761 Disable built-in declarations of functions that are not mandated by
1762 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1763 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1765 @item -fno-operator-names
1766 @opindex fno-operator-names
1767 Do not treat the operator name keywords @code{and}, @code{bitand},
1768 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1769 synonyms as keywords.
1771 @item -fno-optional-diags
1772 @opindex fno-optional-diags
1773 Disable diagnostics that the standard says a compiler does not need to
1774 issue. Currently, the only such diagnostic issued by G++ is the one for
1775 a name having multiple meanings within a class.
1778 @opindex fpermissive
1779 Downgrade some diagnostics about nonconformant code from errors to
1780 warnings. Thus, using @option{-fpermissive} will allow some
1781 nonconforming code to compile.
1785 Enable automatic template instantiation at link time. This option also
1786 implies @option{-fno-implicit-templates}. @xref{Template
1787 Instantiation}, for more information.
1791 Disable generation of information about every class with virtual
1792 functions for use by the C++ runtime type identification features
1793 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1794 of the language, you can save some space by using this flag. Note that
1795 exception handling uses the same information, but it will generate it as
1796 needed. The @samp{dynamic_cast} operator can still be used for casts that
1797 do not require runtime type information, i.e.@: casts to @code{void *} or to
1798 unambiguous base classes.
1802 Emit statistics about front-end processing at the end of the compilation.
1803 This information is generally only useful to the G++ development team.
1805 @item -ftemplate-depth-@var{n}
1806 @opindex ftemplate-depth
1807 Set the maximum instantiation depth for template classes to @var{n}.
1808 A limit on the template instantiation depth is needed to detect
1809 endless recursions during template class instantiation. ANSI/ISO C++
1810 conforming programs must not rely on a maximum depth greater than 17.
1812 @item -fno-threadsafe-statics
1813 @opindex fno-threadsafe-statics
1814 Do not emit the extra code to use the routines specified in the C++
1815 ABI for thread-safe initialization of local statics. You can use this
1816 option to reduce code size slightly in code that doesn't need to be
1819 @item -fuse-cxa-atexit
1820 @opindex fuse-cxa-atexit
1821 Register destructors for objects with static storage duration with the
1822 @code{__cxa_atexit} function rather than the @code{atexit} function.
1823 This option is required for fully standards-compliant handling of static
1824 destructors, but will only work if your C library supports
1825 @code{__cxa_atexit}.
1827 @item -fno-use-cxa-get-exception-ptr
1828 @opindex fno-use-cxa-get-exception-ptr
1829 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1830 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1831 if the runtime routine is not available.
1833 @item -fvisibility-inlines-hidden
1834 @opindex fvisibility-inlines-hidden
1835 This switch declares that the user does not attempt to compare
1836 pointers to inline methods where the addresses of the two functions
1837 were taken in different shared objects.
1839 The effect of this is that GCC may, effectively, mark inline methods with
1840 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1841 appear in the export table of a DSO and do not require a PLT indirection
1842 when used within the DSO@. Enabling this option can have a dramatic effect
1843 on load and link times of a DSO as it massively reduces the size of the
1844 dynamic export table when the library makes heavy use of templates.
1846 The behavior of this switch is not quite the same as marking the
1847 methods as hidden directly, because it does not affect static variables
1848 local to the function or cause the compiler to deduce that
1849 the function is defined in only one shared object.
1851 You may mark a method as having a visibility explicitly to negate the
1852 effect of the switch for that method. For example, if you do want to
1853 compare pointers to a particular inline method, you might mark it as
1854 having default visibility. Marking the enclosing class with explicit
1855 visibility will have no effect.
1857 Explicitly instantiated inline methods are unaffected by this option
1858 as their linkage might otherwise cross a shared library boundary.
1859 @xref{Template Instantiation}.
1861 @item -fvisibility-ms-compat
1862 @opindex fvisibility-ms-compat
1863 This flag attempts to use visibility settings to make GCC's C++
1864 linkage model compatible with that of Microsoft Visual Studio.
1866 The flag makes these changes to GCC's linkage model:
1870 It sets the default visibility to @code{hidden}, like
1871 @option{-fvisibility=hidden}.
1874 Types, but not their members, are not hidden by default.
1877 The One Definition Rule is relaxed for types without explicit
1878 visibility specifications which are defined in more than one different
1879 shared object: those declarations are permitted if they would have
1880 been permitted when this option was not used.
1883 In new code it is better to use @option{-fvisibility=hidden} and
1884 export those classes which are intended to be externally visible.
1885 Unfortunately it is possible for code to rely, perhaps accidentally,
1886 on the Visual Studio behavior.
1888 Among the consequences of these changes are that static data members
1889 of the same type with the same name but defined in different shared
1890 objects will be different, so changing one will not change the other;
1891 and that pointers to function members defined in different shared
1892 objects may not compare equal. When this flag is given, it is a
1893 violation of the ODR to define types with the same name differently.
1897 Do not use weak symbol support, even if it is provided by the linker.
1898 By default, G++ will use weak symbols if they are available. This
1899 option exists only for testing, and should not be used by end-users;
1900 it will result in inferior code and has no benefits. This option may
1901 be removed in a future release of G++.
1905 Do not search for header files in the standard directories specific to
1906 C++, but do still search the other standard directories. (This option
1907 is used when building the C++ library.)
1910 In addition, these optimization, warning, and code generation options
1911 have meanings only for C++ programs:
1914 @item -fno-default-inline
1915 @opindex fno-default-inline
1916 Do not assume @samp{inline} for functions defined inside a class scope.
1917 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1918 functions will have linkage like inline functions; they just won't be
1921 @item -Wabi @r{(C++ and Objective-C++ only)}
1924 Warn when G++ generates code that is probably not compatible with the
1925 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1926 all such cases, there are probably some cases that are not warned about,
1927 even though G++ is generating incompatible code. There may also be
1928 cases where warnings are emitted even though the code that is generated
1931 You should rewrite your code to avoid these warnings if you are
1932 concerned about the fact that code generated by G++ may not be binary
1933 compatible with code generated by other compilers.
1935 The known incompatibilities at this point include:
1940 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1941 pack data into the same byte as a base class. For example:
1944 struct A @{ virtual void f(); int f1 : 1; @};
1945 struct B : public A @{ int f2 : 1; @};
1949 In this case, G++ will place @code{B::f2} into the same byte
1950 as@code{A::f1}; other compilers will not. You can avoid this problem
1951 by explicitly padding @code{A} so that its size is a multiple of the
1952 byte size on your platform; that will cause G++ and other compilers to
1953 layout @code{B} identically.
1956 Incorrect handling of tail-padding for virtual bases. G++ does not use
1957 tail padding when laying out virtual bases. For example:
1960 struct A @{ virtual void f(); char c1; @};
1961 struct B @{ B(); char c2; @};
1962 struct C : public A, public virtual B @{@};
1966 In this case, G++ will not place @code{B} into the tail-padding for
1967 @code{A}; other compilers will. You can avoid this problem by
1968 explicitly padding @code{A} so that its size is a multiple of its
1969 alignment (ignoring virtual base classes); that will cause G++ and other
1970 compilers to layout @code{C} identically.
1973 Incorrect handling of bit-fields with declared widths greater than that
1974 of their underlying types, when the bit-fields appear in a union. For
1978 union U @{ int i : 4096; @};
1982 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1983 union too small by the number of bits in an @code{int}.
1986 Empty classes can be placed at incorrect offsets. For example:
1996 struct C : public B, public A @{@};
2000 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2001 it should be placed at offset zero. G++ mistakenly believes that the
2002 @code{A} data member of @code{B} is already at offset zero.
2005 Names of template functions whose types involve @code{typename} or
2006 template template parameters can be mangled incorrectly.
2009 template <typename Q>
2010 void f(typename Q::X) @{@}
2012 template <template <typename> class Q>
2013 void f(typename Q<int>::X) @{@}
2017 Instantiations of these templates may be mangled incorrectly.
2021 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2022 @opindex Wctor-dtor-privacy
2023 @opindex Wno-ctor-dtor-privacy
2024 Warn when a class seems unusable because all the constructors or
2025 destructors in that class are private, and it has neither friends nor
2026 public static member functions.
2028 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2029 @opindex Wnon-virtual-dtor
2030 @opindex Wno-non-virtual-dtor
2031 Warn when a class has virtual functions and accessible non-virtual
2032 destructor, in which case it would be possible but unsafe to delete
2033 an instance of a derived class through a pointer to the base class.
2034 This warning is also enabled if -Weffc++ is specified.
2036 @item -Wreorder @r{(C++ and Objective-C++ only)}
2038 @opindex Wno-reorder
2039 @cindex reordering, warning
2040 @cindex warning for reordering of member initializers
2041 Warn when the order of member initializers given in the code does not
2042 match the order in which they must be executed. For instance:
2048 A(): j (0), i (1) @{ @}
2052 The compiler will rearrange the member initializers for @samp{i}
2053 and @samp{j} to match the declaration order of the members, emitting
2054 a warning to that effect. This warning is enabled by @option{-Wall}.
2057 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2060 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2063 Warn about violations of the following style guidelines from Scott Meyers'
2064 @cite{Effective C++} book:
2068 Item 11: Define a copy constructor and an assignment operator for classes
2069 with dynamically allocated memory.
2072 Item 12: Prefer initialization to assignment in constructors.
2075 Item 14: Make destructors virtual in base classes.
2078 Item 15: Have @code{operator=} return a reference to @code{*this}.
2081 Item 23: Don't try to return a reference when you must return an object.
2085 Also warn about violations of the following style guidelines from
2086 Scott Meyers' @cite{More Effective C++} book:
2090 Item 6: Distinguish between prefix and postfix forms of increment and
2091 decrement operators.
2094 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2098 When selecting this option, be aware that the standard library
2099 headers do not obey all of these guidelines; use @samp{grep -v}
2100 to filter out those warnings.
2102 @item -Wno-deprecated @r{(C++ and Objective-C++ only)}
2103 @opindex Wno-deprecated
2104 @opindex Wdeprecated
2105 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
2107 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2108 @opindex Wstrict-null-sentinel
2109 @opindex Wno-strict-null-sentinel
2110 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2111 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2112 to @code{__null}. Although it is a null pointer constant not a null pointer,
2113 it is guaranteed to of the same size as a pointer. But this use is
2114 not portable across different compilers.
2116 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2117 @opindex Wno-non-template-friend
2118 @opindex Wnon-template-friend
2119 Disable warnings when non-templatized friend functions are declared
2120 within a template. Since the advent of explicit template specification
2121 support in G++, if the name of the friend is an unqualified-id (i.e.,
2122 @samp{friend foo(int)}), the C++ language specification demands that the
2123 friend declare or define an ordinary, nontemplate function. (Section
2124 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2125 could be interpreted as a particular specialization of a templatized
2126 function. Because this non-conforming behavior is no longer the default
2127 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2128 check existing code for potential trouble spots and is on by default.
2129 This new compiler behavior can be turned off with
2130 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2131 but disables the helpful warning.
2133 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2134 @opindex Wold-style-cast
2135 @opindex Wno-old-style-cast
2136 Warn if an old-style (C-style) cast to a non-void type is used within
2137 a C++ program. The new-style casts (@samp{dynamic_cast},
2138 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2139 less vulnerable to unintended effects and much easier to search for.
2141 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2142 @opindex Woverloaded-virtual
2143 @opindex Wno-overloaded-virtual
2144 @cindex overloaded virtual fn, warning
2145 @cindex warning for overloaded virtual fn
2146 Warn when a function declaration hides virtual functions from a
2147 base class. For example, in:
2154 struct B: public A @{
2159 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2167 will fail to compile.
2169 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2170 @opindex Wno-pmf-conversions
2171 @opindex Wpmf-conversions
2172 Disable the diagnostic for converting a bound pointer to member function
2175 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2176 @opindex Wsign-promo
2177 @opindex Wno-sign-promo
2178 Warn when overload resolution chooses a promotion from unsigned or
2179 enumerated type to a signed type, over a conversion to an unsigned type of
2180 the same size. Previous versions of G++ would try to preserve
2181 unsignedness, but the standard mandates the current behavior.
2186 A& operator = (int);
2196 In this example, G++ will synthesize a default @samp{A& operator =
2197 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2200 @node Objective-C and Objective-C++ Dialect Options
2201 @section Options Controlling Objective-C and Objective-C++ Dialects
2203 @cindex compiler options, Objective-C and Objective-C++
2204 @cindex Objective-C and Objective-C++ options, command line
2205 @cindex options, Objective-C and Objective-C++
2206 (NOTE: This manual does not describe the Objective-C and Objective-C++
2207 languages themselves. See @xref{Standards,,Language Standards
2208 Supported by GCC}, for references.)
2210 This section describes the command-line options that are only meaningful
2211 for Objective-C and Objective-C++ programs, but you can also use most of
2212 the language-independent GNU compiler options.
2213 For example, you might compile a file @code{some_class.m} like this:
2216 gcc -g -fgnu-runtime -O -c some_class.m
2220 In this example, @option{-fgnu-runtime} is an option meant only for
2221 Objective-C and Objective-C++ programs; you can use the other options with
2222 any language supported by GCC@.
2224 Note that since Objective-C is an extension of the C language, Objective-C
2225 compilations may also use options specific to the C front-end (e.g.,
2226 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2227 C++-specific options (e.g., @option{-Wabi}).
2229 Here is a list of options that are @emph{only} for compiling Objective-C
2230 and Objective-C++ programs:
2233 @item -fconstant-string-class=@var{class-name}
2234 @opindex fconstant-string-class
2235 Use @var{class-name} as the name of the class to instantiate for each
2236 literal string specified with the syntax @code{@@"@dots{}"}. The default
2237 class name is @code{NXConstantString} if the GNU runtime is being used, and
2238 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2239 @option{-fconstant-cfstrings} option, if also present, will override the
2240 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2241 to be laid out as constant CoreFoundation strings.
2244 @opindex fgnu-runtime
2245 Generate object code compatible with the standard GNU Objective-C
2246 runtime. This is the default for most types of systems.
2248 @item -fnext-runtime
2249 @opindex fnext-runtime
2250 Generate output compatible with the NeXT runtime. This is the default
2251 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2252 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2255 @item -fno-nil-receivers
2256 @opindex fno-nil-receivers
2257 Assume that all Objective-C message dispatches (e.g.,
2258 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2259 is not @code{nil}. This allows for more efficient entry points in the runtime
2260 to be used. Currently, this option is only available in conjunction with
2261 the NeXT runtime on Mac OS X 10.3 and later.
2263 @item -fobjc-call-cxx-cdtors
2264 @opindex fobjc-call-cxx-cdtors
2265 For each Objective-C class, check if any of its instance variables is a
2266 C++ object with a non-trivial default constructor. If so, synthesize a
2267 special @code{- (id) .cxx_construct} instance method that will run
2268 non-trivial default constructors on any such instance variables, in order,
2269 and then return @code{self}. Similarly, check if any instance variable
2270 is a C++ object with a non-trivial destructor, and if so, synthesize a
2271 special @code{- (void) .cxx_destruct} method that will run
2272 all such default destructors, in reverse order.
2274 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2275 thusly generated will only operate on instance variables declared in the
2276 current Objective-C class, and not those inherited from superclasses. It
2277 is the responsibility of the Objective-C runtime to invoke all such methods
2278 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2279 will be invoked by the runtime immediately after a new object
2280 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2281 be invoked immediately before the runtime deallocates an object instance.
2283 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2284 support for invoking the @code{- (id) .cxx_construct} and
2285 @code{- (void) .cxx_destruct} methods.
2287 @item -fobjc-direct-dispatch
2288 @opindex fobjc-direct-dispatch
2289 Allow fast jumps to the message dispatcher. On Darwin this is
2290 accomplished via the comm page.
2292 @item -fobjc-exceptions
2293 @opindex fobjc-exceptions
2294 Enable syntactic support for structured exception handling in Objective-C,
2295 similar to what is offered by C++ and Java. This option is
2296 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2305 @@catch (AnObjCClass *exc) @{
2312 @@catch (AnotherClass *exc) @{
2315 @@catch (id allOthers) @{
2325 The @code{@@throw} statement may appear anywhere in an Objective-C or
2326 Objective-C++ program; when used inside of a @code{@@catch} block, the
2327 @code{@@throw} may appear without an argument (as shown above), in which case
2328 the object caught by the @code{@@catch} will be rethrown.
2330 Note that only (pointers to) Objective-C objects may be thrown and
2331 caught using this scheme. When an object is thrown, it will be caught
2332 by the nearest @code{@@catch} clause capable of handling objects of that type,
2333 analogously to how @code{catch} blocks work in C++ and Java. A
2334 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2335 any and all Objective-C exceptions not caught by previous @code{@@catch}
2338 The @code{@@finally} clause, if present, will be executed upon exit from the
2339 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2340 regardless of whether any exceptions are thrown, caught or rethrown
2341 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2342 of the @code{finally} clause in Java.
2344 There are several caveats to using the new exception mechanism:
2348 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2349 idioms provided by the @code{NSException} class, the new
2350 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2351 systems, due to additional functionality needed in the (NeXT) Objective-C
2355 As mentioned above, the new exceptions do not support handling
2356 types other than Objective-C objects. Furthermore, when used from
2357 Objective-C++, the Objective-C exception model does not interoperate with C++
2358 exceptions at this time. This means you cannot @code{@@throw} an exception
2359 from Objective-C and @code{catch} it in C++, or vice versa
2360 (i.e., @code{throw @dots{} @@catch}).
2363 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2364 blocks for thread-safe execution:
2367 @@synchronized (ObjCClass *guard) @{
2372 Upon entering the @code{@@synchronized} block, a thread of execution shall
2373 first check whether a lock has been placed on the corresponding @code{guard}
2374 object by another thread. If it has, the current thread shall wait until
2375 the other thread relinquishes its lock. Once @code{guard} becomes available,
2376 the current thread will place its own lock on it, execute the code contained in
2377 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2378 making @code{guard} available to other threads).
2380 Unlike Java, Objective-C does not allow for entire methods to be marked
2381 @code{@@synchronized}. Note that throwing exceptions out of
2382 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2383 to be unlocked properly.
2387 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2389 @item -freplace-objc-classes
2390 @opindex freplace-objc-classes
2391 Emit a special marker instructing @command{ld(1)} not to statically link in
2392 the resulting object file, and allow @command{dyld(1)} to load it in at
2393 run time instead. This is used in conjunction with the Fix-and-Continue
2394 debugging mode, where the object file in question may be recompiled and
2395 dynamically reloaded in the course of program execution, without the need
2396 to restart the program itself. Currently, Fix-and-Continue functionality
2397 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2402 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2403 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2404 compile time) with static class references that get initialized at load time,
2405 which improves run-time performance. Specifying the @option{-fzero-link} flag
2406 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2407 to be retained. This is useful in Zero-Link debugging mode, since it allows
2408 for individual class implementations to be modified during program execution.
2412 Dump interface declarations for all classes seen in the source file to a
2413 file named @file{@var{sourcename}.decl}.
2415 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2416 @opindex Wassign-intercept
2417 @opindex Wno-assign-intercept
2418 Warn whenever an Objective-C assignment is being intercepted by the
2421 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2422 @opindex Wno-protocol
2424 If a class is declared to implement a protocol, a warning is issued for
2425 every method in the protocol that is not implemented by the class. The
2426 default behavior is to issue a warning for every method not explicitly
2427 implemented in the class, even if a method implementation is inherited
2428 from the superclass. If you use the @option{-Wno-protocol} option, then
2429 methods inherited from the superclass are considered to be implemented,
2430 and no warning is issued for them.
2432 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2434 @opindex Wno-selector
2435 Warn if multiple methods of different types for the same selector are
2436 found during compilation. The check is performed on the list of methods
2437 in the final stage of compilation. Additionally, a check is performed
2438 for each selector appearing in a @code{@@selector(@dots{})}
2439 expression, and a corresponding method for that selector has been found
2440 during compilation. Because these checks scan the method table only at
2441 the end of compilation, these warnings are not produced if the final
2442 stage of compilation is not reached, for example because an error is
2443 found during compilation, or because the @option{-fsyntax-only} option is
2446 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2447 @opindex Wstrict-selector-match
2448 @opindex Wno-strict-selector-match
2449 Warn if multiple methods with differing argument and/or return types are
2450 found for a given selector when attempting to send a message using this
2451 selector to a receiver of type @code{id} or @code{Class}. When this flag
2452 is off (which is the default behavior), the compiler will omit such warnings
2453 if any differences found are confined to types which share the same size
2456 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2457 @opindex Wundeclared-selector
2458 @opindex Wno-undeclared-selector
2459 Warn if a @code{@@selector(@dots{})} expression referring to an
2460 undeclared selector is found. A selector is considered undeclared if no
2461 method with that name has been declared before the
2462 @code{@@selector(@dots{})} expression, either explicitly in an
2463 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2464 an @code{@@implementation} section. This option always performs its
2465 checks as soon as a @code{@@selector(@dots{})} expression is found,
2466 while @option{-Wselector} only performs its checks in the final stage of
2467 compilation. This also enforces the coding style convention
2468 that methods and selectors must be declared before being used.
2470 @item -print-objc-runtime-info
2471 @opindex print-objc-runtime-info
2472 Generate C header describing the largest structure that is passed by
2477 @node Language Independent Options
2478 @section Options to Control Diagnostic Messages Formatting
2479 @cindex options to control diagnostics formatting
2480 @cindex diagnostic messages
2481 @cindex message formatting
2483 Traditionally, diagnostic messages have been formatted irrespective of
2484 the output device's aspect (e.g.@: its width, @dots{}). The options described
2485 below can be used to control the diagnostic messages formatting
2486 algorithm, e.g.@: how many characters per line, how often source location
2487 information should be reported. Right now, only the C++ front end can
2488 honor these options. However it is expected, in the near future, that
2489 the remaining front ends would be able to digest them correctly.
2492 @item -fmessage-length=@var{n}
2493 @opindex fmessage-length
2494 Try to format error messages so that they fit on lines of about @var{n}
2495 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2496 the front ends supported by GCC@. If @var{n} is zero, then no
2497 line-wrapping will be done; each error message will appear on a single
2500 @opindex fdiagnostics-show-location
2501 @item -fdiagnostics-show-location=once
2502 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2503 reporter to emit @emph{once} source location information; that is, in
2504 case the message is too long to fit on a single physical line and has to
2505 be wrapped, the source location won't be emitted (as prefix) again,
2506 over and over, in subsequent continuation lines. This is the default
2509 @item -fdiagnostics-show-location=every-line
2510 Only meaningful in line-wrapping mode. Instructs the diagnostic
2511 messages reporter to emit the same source location information (as
2512 prefix) for physical lines that result from the process of breaking
2513 a message which is too long to fit on a single line.
2515 @item -fdiagnostics-show-option
2516 @opindex fdiagnostics-show-option
2517 This option instructs the diagnostic machinery to add text to each
2518 diagnostic emitted, which indicates which command line option directly
2519 controls that diagnostic, when such an option is known to the
2520 diagnostic machinery.
2522 @item -Wcoverage-mismatch
2523 @opindex Wcoverage-mismatch
2524 Warn if feedback profiles do not match when using the
2525 @option{-fprofile-use} option.
2526 If a source file was changed between @option{-fprofile-gen} and
2527 @option{-fprofile-use}, the files with the profile feedback can fail
2528 to match the source file and GCC can not use the profile feedback
2529 information. By default, GCC emits an error message in this case.
2530 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2531 error. GCC does not use appropriate feedback profiles, so using this
2532 option can result in poorly optimized code. This option is useful
2533 only in the case of very minor changes such as bug fixes to an
2538 @node Warning Options
2539 @section Options to Request or Suppress Warnings
2540 @cindex options to control warnings
2541 @cindex warning messages
2542 @cindex messages, warning
2543 @cindex suppressing warnings
2545 Warnings are diagnostic messages that report constructions which
2546 are not inherently erroneous but which are risky or suggest there
2547 may have been an error.
2549 The following language-independent options do not enable specific
2550 warnings but control the kinds of diagnostics produced by GCC.
2553 @cindex syntax checking
2555 @opindex fsyntax-only
2556 Check the code for syntax errors, but don't do anything beyond that.
2560 Inhibit all warning messages.
2565 Make all warnings into errors.
2570 Make the specified warning into an error. The specifier for a warning
2571 is appended, for example @option{-Werror=switch} turns the warnings
2572 controlled by @option{-Wswitch} into errors. This switch takes a
2573 negative form, to be used to negate @option{-Werror} for specific
2574 warnings, for example @option{-Wno-error=switch} makes
2575 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2576 is in effect. You can use the @option{-fdiagnostics-show-option}
2577 option to have each controllable warning amended with the option which
2578 controls it, to determine what to use with this option.
2580 Note that specifying @option{-Werror=}@var{foo} automatically implies
2581 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2584 @item -Wfatal-errors
2585 @opindex Wfatal-errors
2586 @opindex Wno-fatal-errors
2587 This option causes the compiler to abort compilation on the first error
2588 occurred rather than trying to keep going and printing further error
2593 You can request many specific warnings with options beginning
2594 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2595 implicit declarations. Each of these specific warning options also
2596 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2597 example, @option{-Wno-implicit}. This manual lists only one of the
2598 two forms, whichever is not the default. For further,
2599 language-specific options also refer to @ref{C++ Dialect Options} and
2600 @ref{Objective-C and Objective-C++ Dialect Options}.
2605 Issue all the warnings demanded by strict ISO C and ISO C++;
2606 reject all programs that use forbidden extensions, and some other
2607 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2608 version of the ISO C standard specified by any @option{-std} option used.
2610 Valid ISO C and ISO C++ programs should compile properly with or without
2611 this option (though a rare few will require @option{-ansi} or a
2612 @option{-std} option specifying the required version of ISO C)@. However,
2613 without this option, certain GNU extensions and traditional C and C++
2614 features are supported as well. With this option, they are rejected.
2616 @option{-pedantic} does not cause warning messages for use of the
2617 alternate keywords whose names begin and end with @samp{__}. Pedantic
2618 warnings are also disabled in the expression that follows
2619 @code{__extension__}. However, only system header files should use
2620 these escape routes; application programs should avoid them.
2621 @xref{Alternate Keywords}.
2623 Some users try to use @option{-pedantic} to check programs for strict ISO
2624 C conformance. They soon find that it does not do quite what they want:
2625 it finds some non-ISO practices, but not all---only those for which
2626 ISO C @emph{requires} a diagnostic, and some others for which
2627 diagnostics have been added.
2629 A feature to report any failure to conform to ISO C might be useful in
2630 some instances, but would require considerable additional work and would
2631 be quite different from @option{-pedantic}. We don't have plans to
2632 support such a feature in the near future.
2634 Where the standard specified with @option{-std} represents a GNU
2635 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2636 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2637 extended dialect is based. Warnings from @option{-pedantic} are given
2638 where they are required by the base standard. (It would not make sense
2639 for such warnings to be given only for features not in the specified GNU
2640 C dialect, since by definition the GNU dialects of C include all
2641 features the compiler supports with the given option, and there would be
2642 nothing to warn about.)
2644 @item -pedantic-errors
2645 @opindex pedantic-errors
2646 Like @option{-pedantic}, except that errors are produced rather than
2652 This enables all the warnings about constructions that some users
2653 consider questionable, and that are easy to avoid (or modify to
2654 prevent the warning), even in conjunction with macros. This also
2655 enables some language-specific warnings described in @ref{C++ Dialect
2656 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2658 @option{-Wall} turns on the following warning flags:
2660 @gccoptlist{-Waddress @gol
2661 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2663 -Wchar-subscripts @gol
2665 -Wimplicit-function-declaration @gol
2668 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2669 -Wmissing-braces @gol
2675 -Wsequence-point @gol
2676 -Wsign-compare @r{(only in C++)} @gol
2677 -Wstrict-aliasing @gol
2678 -Wstrict-overflow=1 @gol
2681 -Wuninitialized @r{(only with} @option{-O1} @r{and above)} @gol
2682 -Wunknown-pragmas @gol
2683 -Wunused-function @gol
2686 -Wunused-variable @gol
2687 -Wvolatile-register-var @gol
2690 Note that some warning flags are not implied by @option{-Wall}. Some of
2691 them warn about constructions that users generally do not consider
2692 questionable, but which occasionally you might wish to check for;
2693 others warn about constructions that are necessary or hard to avoid in
2694 some cases, and there is no simple way to modify the code to suppress
2695 the warning. Some of them are enabled by @option{-Wextra} but many of
2696 them must be enabled individually.
2702 This enables some extra warning flags that are not enabled by
2703 @option{-Wall}. (This option used to be called @option{-W}. The older
2704 name is still supported, but the newer name is more descriptive.)
2706 @gccoptlist{-Wclobbered @gol
2708 -Wignored-qualifiers @gol
2709 -Wmissing-field-initializers @gol
2710 -Wmissing-parameter-type @r{(C only)} @gol
2711 -Wold-style-declaration @r{(C only)} @gol
2712 -Woverride-init @gol
2715 -Wuninitialized @r{(only with} @option{-O1} @r{and above)} @gol
2716 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2719 The option @option{-Wextra} also prints warning messages for the
2725 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2726 @samp{>}, or @samp{>=}.
2729 (C++ only) An enumerator and a non-enumerator both appear in a
2730 conditional expression.
2733 (C++ only) A non-static reference or non-static @samp{const} member
2734 appears in a class without constructors.
2737 (C++ only) Ambiguous virtual bases.
2740 (C++ only) Subscripting an array which has been declared @samp{register}.
2743 (C++ only) Taking the address of a variable which has been declared
2747 (C++ only) A base class is not initialized in a derived class' copy
2755 Inhibit warning messages about the use of @samp{#import}.
2757 @item -Wchar-subscripts
2758 @opindex Wchar-subscripts
2759 @opindex Wno-char-subscripts
2760 Warn if an array subscript has type @code{char}. This is a common cause
2761 of error, as programmers often forget that this type is signed on some
2763 This warning is enabled by @option{-Wall}.
2767 @opindex Wno-comment
2768 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2769 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2770 This warning is enabled by @option{-Wall}.
2775 @opindex ffreestanding
2776 @opindex fno-builtin
2777 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2778 the arguments supplied have types appropriate to the format string
2779 specified, and that the conversions specified in the format string make
2780 sense. This includes standard functions, and others specified by format
2781 attributes (@pxref{Function Attributes}), in the @code{printf},
2782 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2783 not in the C standard) families (or other target-specific families).
2784 Which functions are checked without format attributes having been
2785 specified depends on the standard version selected, and such checks of
2786 functions without the attribute specified are disabled by
2787 @option{-ffreestanding} or @option{-fno-builtin}.
2789 The formats are checked against the format features supported by GNU
2790 libc version 2.2. These include all ISO C90 and C99 features, as well
2791 as features from the Single Unix Specification and some BSD and GNU
2792 extensions. Other library implementations may not support all these
2793 features; GCC does not support warning about features that go beyond a
2794 particular library's limitations. However, if @option{-pedantic} is used
2795 with @option{-Wformat}, warnings will be given about format features not
2796 in the selected standard version (but not for @code{strfmon} formats,
2797 since those are not in any version of the C standard). @xref{C Dialect
2798 Options,,Options Controlling C Dialect}.
2800 Since @option{-Wformat} also checks for null format arguments for
2801 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2803 @option{-Wformat} is included in @option{-Wall}. For more control over some
2804 aspects of format checking, the options @option{-Wformat-y2k},
2805 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2806 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2807 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2810 @opindex Wformat-y2k
2811 @opindex Wno-format-y2k
2812 If @option{-Wformat} is specified, also warn about @code{strftime}
2813 formats which may yield only a two-digit year.
2815 @item -Wno-format-contains-nul
2816 @opindex Wno-format-contains-nul
2817 @opindex Wformat-contains-nul
2818 If @option{-Wformat} is specified, do not warn about format strings that
2821 @item -Wno-format-extra-args
2822 @opindex Wno-format-extra-args
2823 @opindex Wformat-extra-args
2824 If @option{-Wformat} is specified, do not warn about excess arguments to a
2825 @code{printf} or @code{scanf} format function. The C standard specifies
2826 that such arguments are ignored.
2828 Where the unused arguments lie between used arguments that are
2829 specified with @samp{$} operand number specifications, normally
2830 warnings are still given, since the implementation could not know what
2831 type to pass to @code{va_arg} to skip the unused arguments. However,
2832 in the case of @code{scanf} formats, this option will suppress the
2833 warning if the unused arguments are all pointers, since the Single
2834 Unix Specification says that such unused arguments are allowed.
2836 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2837 @opindex Wno-format-zero-length
2838 @opindex Wformat-zero-length
2839 If @option{-Wformat} is specified, do not warn about zero-length formats.
2840 The C standard specifies that zero-length formats are allowed.
2842 @item -Wformat-nonliteral
2843 @opindex Wformat-nonliteral
2844 @opindex Wno-format-nonliteral
2845 If @option{-Wformat} is specified, also warn if the format string is not a
2846 string literal and so cannot be checked, unless the format function
2847 takes its format arguments as a @code{va_list}.
2849 @item -Wformat-security
2850 @opindex Wformat-security
2851 @opindex Wno-format-security
2852 If @option{-Wformat} is specified, also warn about uses of format
2853 functions that represent possible security problems. At present, this
2854 warns about calls to @code{printf} and @code{scanf} functions where the
2855 format string is not a string literal and there are no format arguments,
2856 as in @code{printf (foo);}. This may be a security hole if the format
2857 string came from untrusted input and contains @samp{%n}. (This is
2858 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2859 in future warnings may be added to @option{-Wformat-security} that are not
2860 included in @option{-Wformat-nonliteral}.)
2864 @opindex Wno-format=2
2865 Enable @option{-Wformat} plus format checks not included in
2866 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2867 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2869 @item -Wnonnull @r{(C and Objective-C only)}
2871 @opindex Wno-nonnull
2872 Warn about passing a null pointer for arguments marked as
2873 requiring a non-null value by the @code{nonnull} function attribute.
2875 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2876 can be disabled with the @option{-Wno-nonnull} option.
2878 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2880 @opindex Wno-init-self
2881 Warn about uninitialized variables which are initialized with themselves.
2882 Note this option can only be used with the @option{-Wuninitialized} option,
2883 which in turn only works with @option{-O1} and above.
2885 For example, GCC will warn about @code{i} being uninitialized in the
2886 following snippet only when @option{-Winit-self} has been specified:
2897 @item -Wimplicit-int @r{(C and Objective-C only)}
2898 @opindex Wimplicit-int
2899 @opindex Wno-implicit-int
2900 Warn when a declaration does not specify a type.
2901 This warning is enabled by @option{-Wall}.
2903 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2904 @opindex Wimplicit-function-declaration
2905 @opindex Wno-implicit-function-declaration
2906 Give a warning whenever a function is used before being declared. In
2907 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2908 enabled by default and it is made into an error by
2909 @option{-pedantic-errors}. This warning is also enabled by
2914 @opindex Wno-implicit
2915 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2916 This warning is enabled by @option{-Wall}.
2918 @item -Wignored-qualifiers @r{(C and C++ only)}
2919 @opindex Wignored-qualifiers
2920 @opindex Wno-ignored-qualifiers
2921 Warn if the return type of a function has a type qualifier
2922 such as @code{const}. For ISO C such a type qualifier has no effect,
2923 since the value returned by a function is not an lvalue.
2924 For C++, the warning is only emitted for scalar types or @code{void}.
2925 ISO C prohibits qualified @code{void} return types on function
2926 definitions, so such return types always receive a warning
2927 even without this option.
2929 This warning is also enabled by @option{-Wextra}.
2934 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2935 function with external linkage, returning int, taking either zero
2936 arguments, two, or three arguments of appropriate types.
2937 This warning is enabled by @option{-Wall}.
2939 @item -Wmissing-braces
2940 @opindex Wmissing-braces
2941 @opindex Wno-missing-braces
2942 Warn if an aggregate or union initializer is not fully bracketed. In
2943 the following example, the initializer for @samp{a} is not fully
2944 bracketed, but that for @samp{b} is fully bracketed.
2947 int a[2][2] = @{ 0, 1, 2, 3 @};
2948 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2951 This warning is enabled by @option{-Wall}.
2953 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2954 @opindex Wmissing-include-dirs
2955 @opindex Wno-missing-include-dirs
2956 Warn if a user-supplied include directory does not exist.
2959 @opindex Wparentheses
2960 @opindex Wno-parentheses
2961 Warn if parentheses are omitted in certain contexts, such
2962 as when there is an assignment in a context where a truth value
2963 is expected, or when operators are nested whose precedence people
2964 often get confused about.
2966 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2967 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2968 interpretation from that of ordinary mathematical notation.
2970 Also warn about constructions where there may be confusion to which
2971 @code{if} statement an @code{else} branch belongs. Here is an example of
2986 In C/C++, every @code{else} branch belongs to the innermost possible
2987 @code{if} statement, which in this example is @code{if (b)}. This is
2988 often not what the programmer expected, as illustrated in the above
2989 example by indentation the programmer chose. When there is the
2990 potential for this confusion, GCC will issue a warning when this flag
2991 is specified. To eliminate the warning, add explicit braces around
2992 the innermost @code{if} statement so there is no way the @code{else}
2993 could belong to the enclosing @code{if}. The resulting code would
3010 This warning is enabled by @option{-Wall}.
3012 @item -Wsequence-point
3013 @opindex Wsequence-point
3014 @opindex Wno-sequence-point
3015 Warn about code that may have undefined semantics because of violations
3016 of sequence point rules in the C and C++ standards.
3018 The C and C++ standards defines the order in which expressions in a C/C++
3019 program are evaluated in terms of @dfn{sequence points}, which represent
3020 a partial ordering between the execution of parts of the program: those
3021 executed before the sequence point, and those executed after it. These
3022 occur after the evaluation of a full expression (one which is not part
3023 of a larger expression), after the evaluation of the first operand of a
3024 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3025 function is called (but after the evaluation of its arguments and the
3026 expression denoting the called function), and in certain other places.
3027 Other than as expressed by the sequence point rules, the order of
3028 evaluation of subexpressions of an expression is not specified. All
3029 these rules describe only a partial order rather than a total order,
3030 since, for example, if two functions are called within one expression
3031 with no sequence point between them, the order in which the functions
3032 are called is not specified. However, the standards committee have
3033 ruled that function calls do not overlap.
3035 It is not specified when between sequence points modifications to the
3036 values of objects take effect. Programs whose behavior depends on this
3037 have undefined behavior; the C and C++ standards specify that ``Between
3038 the previous and next sequence point an object shall have its stored
3039 value modified at most once by the evaluation of an expression.
3040 Furthermore, the prior value shall be read only to determine the value
3041 to be stored.''. If a program breaks these rules, the results on any
3042 particular implementation are entirely unpredictable.
3044 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3045 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3046 diagnosed by this option, and it may give an occasional false positive
3047 result, but in general it has been found fairly effective at detecting
3048 this sort of problem in programs.
3050 The standard is worded confusingly, therefore there is some debate
3051 over the precise meaning of the sequence point rules in subtle cases.
3052 Links to discussions of the problem, including proposed formal
3053 definitions, may be found on the GCC readings page, at
3054 @w{@uref{http://gcc.gnu.org/readings.html}}.
3056 This warning is enabled by @option{-Wall} for C and C++.
3059 @opindex Wreturn-type
3060 @opindex Wno-return-type
3061 Warn whenever a function is defined with a return-type that defaults
3062 to @code{int}. Also warn about any @code{return} statement with no
3063 return-value in a function whose return-type is not @code{void}
3064 (falling off the end of the function body is considered returning
3065 without a value), and about a @code{return} statement with a
3066 expression in a function whose return-type is @code{void}.
3068 For C++, a function without return type always produces a diagnostic
3069 message, even when @option{-Wno-return-type} is specified. The only
3070 exceptions are @samp{main} and functions defined in system headers.
3072 This warning is enabled by @option{-Wall}.
3077 Warn whenever a @code{switch} statement has an index of enumerated type
3078 and lacks a @code{case} for one or more of the named codes of that
3079 enumeration. (The presence of a @code{default} label prevents this
3080 warning.) @code{case} labels outside the enumeration range also
3081 provoke warnings when this option is used.
3082 This warning is enabled by @option{-Wall}.
3084 @item -Wswitch-default
3085 @opindex Wswitch-default
3086 @opindex Wno-switch-default
3087 Warn whenever a @code{switch} statement does not have a @code{default}
3091 @opindex Wswitch-enum
3092 @opindex Wno-switch-enum
3093 Warn whenever a @code{switch} statement has an index of enumerated type
3094 and lacks a @code{case} for one or more of the named codes of that
3095 enumeration. @code{case} labels outside the enumeration range also
3096 provoke warnings when this option is used.
3100 @opindex Wno-trigraphs
3101 Warn if any trigraphs are encountered that might change the meaning of
3102 the program (trigraphs within comments are not warned about).
3103 This warning is enabled by @option{-Wall}.
3105 @item -Wunused-function
3106 @opindex Wunused-function
3107 @opindex Wno-unused-function
3108 Warn whenever a static function is declared but not defined or a
3109 non-inline static function is unused.
3110 This warning is enabled by @option{-Wall}.
3112 @item -Wunused-label
3113 @opindex Wunused-label
3114 @opindex Wno-unused-label
3115 Warn whenever a label is declared but not used.
3116 This warning is enabled by @option{-Wall}.
3118 To suppress this warning use the @samp{unused} attribute
3119 (@pxref{Variable Attributes}).
3121 @item -Wunused-parameter
3122 @opindex Wunused-parameter
3123 @opindex Wno-unused-parameter
3124 Warn whenever a function parameter is unused aside from its declaration.
3126 To suppress this warning use the @samp{unused} attribute
3127 (@pxref{Variable Attributes}).
3129 @item -Wunused-variable
3130 @opindex Wunused-variable
3131 @opindex Wno-unused-variable
3132 Warn whenever a local variable or non-constant static variable is unused
3133 aside from its declaration.
3134 This warning is enabled by @option{-Wall}.
3136 To suppress this warning use the @samp{unused} attribute
3137 (@pxref{Variable Attributes}).
3139 @item -Wunused-value
3140 @opindex Wunused-value
3141 @opindex Wno-unused-value
3142 Warn whenever a statement computes a result that is explicitly not
3143 used. To suppress this warning cast the unused expression to
3144 @samp{void}. This includes an expression-statement or the left-hand
3145 side of a comma expression that contains no side effects. For example,
3146 an expression such as @samp{x[i,j]} will cause a warning, while
3147 @samp{x[(void)i,j]} will not.
3149 This warning is enabled by @option{-Wall}.
3154 All the above @option{-Wunused} options combined.
3156 In order to get a warning about an unused function parameter, you must
3157 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3158 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3160 @item -Wuninitialized
3161 @opindex Wuninitialized
3162 @opindex Wno-uninitialized
3163 Warn if an automatic variable is used without first being initialized or
3164 if a variable may be clobbered by a @code{setjmp} call.
3166 These warnings are possible only in optimizing compilation,
3167 because they require data flow information that is computed only
3168 when optimizing. If you do not specify @option{-O}, you will not get
3169 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
3170 requiring @option{-O}.
3172 If you want to warn about code which uses the uninitialized value of the
3173 variable in its own initializer, use the @option{-Winit-self} option.
3175 These warnings occur for individual uninitialized or clobbered
3176 elements of structure, union or array variables as well as for
3177 variables which are uninitialized or clobbered as a whole. They do
3178 not occur for variables or elements declared @code{volatile}. Because
3179 these warnings depend on optimization, the exact variables or elements
3180 for which there are warnings will depend on the precise optimization
3181 options and version of GCC used.
3183 Note that there may be no warning about a variable that is used only
3184 to compute a value that itself is never used, because such
3185 computations may be deleted by data flow analysis before the warnings
3188 These warnings are made optional because GCC is not smart
3189 enough to see all the reasons why the code might be correct
3190 despite appearing to have an error. Here is one example of how
3211 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3212 always initialized, but GCC doesn't know this. Here is
3213 another common case:
3218 if (change_y) save_y = y, y = new_y;
3220 if (change_y) y = save_y;
3225 This has no bug because @code{save_y} is used only if it is set.
3227 @cindex @code{longjmp} warnings
3228 This option also warns when a non-volatile automatic variable might be
3229 changed by a call to @code{longjmp}. These warnings as well are possible
3230 only in optimizing compilation.
3232 The compiler sees only the calls to @code{setjmp}. It cannot know
3233 where @code{longjmp} will be called; in fact, a signal handler could
3234 call it at any point in the code. As a result, you may get a warning
3235 even when there is in fact no problem because @code{longjmp} cannot
3236 in fact be called at the place which would cause a problem.
3238 Some spurious warnings can be avoided if you declare all the functions
3239 you use that never return as @code{noreturn}. @xref{Function
3242 This warning is enabled by @option{-Wall} or @option{-Wextra} in
3243 optimizing compilations (@option{-O1} and above).
3245 @item -Wunknown-pragmas
3246 @opindex Wunknown-pragmas
3247 @opindex Wno-unknown-pragmas
3248 @cindex warning for unknown pragmas
3249 @cindex unknown pragmas, warning
3250 @cindex pragmas, warning of unknown
3251 Warn when a #pragma directive is encountered which is not understood by
3252 GCC@. If this command line option is used, warnings will even be issued
3253 for unknown pragmas in system header files. This is not the case if
3254 the warnings were only enabled by the @option{-Wall} command line option.
3257 @opindex Wno-pragmas
3259 Do not warn about misuses of pragmas, such as incorrect parameters,
3260 invalid syntax, or conflicts between pragmas. See also
3261 @samp{-Wunknown-pragmas}.
3263 @item -Wstrict-aliasing
3264 @opindex Wstrict-aliasing
3265 @opindex Wno-strict-aliasing
3266 This option is only active when @option{-fstrict-aliasing} is active.
3267 It warns about code which might break the strict aliasing rules that the
3268 compiler is using for optimization. The warning does not catch all
3269 cases, but does attempt to catch the more common pitfalls. It is
3270 included in @option{-Wall}.
3271 It is equivalent to @option{-Wstrict-aliasing=3}
3273 @item -Wstrict-aliasing=n
3274 @opindex Wstrict-aliasing=n
3275 @opindex Wno-strict-aliasing=n
3276 This option is only active when @option{-fstrict-aliasing} is active.
3277 It warns about code which might break the strict aliasing rules that the
3278 compiler is using for optimization.
3279 Higher levels correspond to higher accuracy (fewer false positives).
3280 Higher levels also correspond to more effort, similar to the way -O works.
3281 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3284 Level 1: Most aggressive, quick, least accurate.
3285 Possibly useful when higher levels
3286 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3287 false negatives. However, it has many false positives.
3288 Warns for all pointer conversions between possibly incompatible types,
3289 even if never dereferenced. Runs in the frontend only.
3291 Level 2: Aggressive, quick, not too precise.
3292 May still have many false positives (not as many as level 1 though),
3293 and few false negatives (but possibly more than level 1).
3294 Unlike level 1, it only warns when an address is taken. Warns about
3295 incomplete types. Runs in the frontend only.
3297 Level 3 (default for @option{-Wstrict-aliasing}):
3298 Should have very few false positives and few false
3299 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3300 Takes care of the common punn+dereference pattern in the frontend:
3301 @code{*(int*)&some_float}.
3302 If optimization is enabled, it also runs in the backend, where it deals
3303 with multiple statement cases using flow-sensitive points-to information.
3304 Only warns when the converted pointer is dereferenced.
3305 Does not warn about incomplete types.
3307 @item -Wstrict-overflow
3308 @itemx -Wstrict-overflow=@var{n}
3309 @opindex Wstrict-overflow
3310 @opindex Wno-strict-overflow
3311 This option is only active when @option{-fstrict-overflow} is active.
3312 It warns about cases where the compiler optimizes based on the
3313 assumption that signed overflow does not occur. Note that it does not
3314 warn about all cases where the code might overflow: it only warns
3315 about cases where the compiler implements some optimization. Thus
3316 this warning depends on the optimization level.
3318 An optimization which assumes that signed overflow does not occur is
3319 perfectly safe if the values of the variables involved are such that
3320 overflow never does, in fact, occur. Therefore this warning can
3321 easily give a false positive: a warning about code which is not
3322 actually a problem. To help focus on important issues, several
3323 warning levels are defined. No warnings are issued for the use of
3324 undefined signed overflow when estimating how many iterations a loop
3325 will require, in particular when determining whether a loop will be
3329 @item -Wstrict-overflow=1
3330 Warn about cases which are both questionable and easy to avoid. For
3331 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3332 compiler will simplify this to @code{1}. This level of
3333 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3334 are not, and must be explicitly requested.
3336 @item -Wstrict-overflow=2
3337 Also warn about other cases where a comparison is simplified to a
3338 constant. For example: @code{abs (x) >= 0}. This can only be
3339 simplified when @option{-fstrict-overflow} is in effect, because
3340 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3341 zero. @option{-Wstrict-overflow} (with no level) is the same as
3342 @option{-Wstrict-overflow=2}.
3344 @item -Wstrict-overflow=3
3345 Also warn about other cases where a comparison is simplified. For
3346 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3348 @item -Wstrict-overflow=4
3349 Also warn about other simplifications not covered by the above cases.
3350 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3352 @item -Wstrict-overflow=5
3353 Also warn about cases where the compiler reduces the magnitude of a
3354 constant involved in a comparison. For example: @code{x + 2 > y} will
3355 be simplified to @code{x + 1 >= y}. This is reported only at the
3356 highest warning level because this simplification applies to many
3357 comparisons, so this warning level will give a very large number of
3361 @item -Warray-bounds
3362 @opindex Wno-array-bounds
3363 @opindex Warray-bounds
3364 This option is only active when @option{-ftree-vrp} is active
3365 (default for -O2 and above). It warns about subscripts to arrays
3366 that are always out of bounds. This warning is enabled by @option{-Wall}.
3368 @item -Wno-div-by-zero
3369 @opindex Wno-div-by-zero
3370 @opindex Wdiv-by-zero
3371 Do not warn about compile-time integer division by zero. Floating point
3372 division by zero is not warned about, as it can be a legitimate way of
3373 obtaining infinities and NaNs.
3375 @item -Wsystem-headers
3376 @opindex Wsystem-headers
3377 @opindex Wno-system-headers
3378 @cindex warnings from system headers
3379 @cindex system headers, warnings from
3380 Print warning messages for constructs found in system header files.
3381 Warnings from system headers are normally suppressed, on the assumption
3382 that they usually do not indicate real problems and would only make the
3383 compiler output harder to read. Using this command line option tells
3384 GCC to emit warnings from system headers as if they occurred in user
3385 code. However, note that using @option{-Wall} in conjunction with this
3386 option will @emph{not} warn about unknown pragmas in system
3387 headers---for that, @option{-Wunknown-pragmas} must also be used.
3390 @opindex Wfloat-equal
3391 @opindex Wno-float-equal
3392 Warn if floating point values are used in equality comparisons.
3394 The idea behind this is that sometimes it is convenient (for the
3395 programmer) to consider floating-point values as approximations to
3396 infinitely precise real numbers. If you are doing this, then you need
3397 to compute (by analyzing the code, or in some other way) the maximum or
3398 likely maximum error that the computation introduces, and allow for it
3399 when performing comparisons (and when producing output, but that's a
3400 different problem). In particular, instead of testing for equality, you
3401 would check to see whether the two values have ranges that overlap; and
3402 this is done with the relational operators, so equality comparisons are
3405 @item -Wtraditional @r{(C and Objective-C only)}
3406 @opindex Wtraditional
3407 @opindex Wno-traditional
3408 Warn about certain constructs that behave differently in traditional and
3409 ISO C@. Also warn about ISO C constructs that have no traditional C
3410 equivalent, and/or problematic constructs which should be avoided.
3414 Macro parameters that appear within string literals in the macro body.
3415 In traditional C macro replacement takes place within string literals,
3416 but does not in ISO C@.
3419 In traditional C, some preprocessor directives did not exist.
3420 Traditional preprocessors would only consider a line to be a directive
3421 if the @samp{#} appeared in column 1 on the line. Therefore
3422 @option{-Wtraditional} warns about directives that traditional C
3423 understands but would ignore because the @samp{#} does not appear as the
3424 first character on the line. It also suggests you hide directives like
3425 @samp{#pragma} not understood by traditional C by indenting them. Some
3426 traditional implementations would not recognize @samp{#elif}, so it
3427 suggests avoiding it altogether.
3430 A function-like macro that appears without arguments.
3433 The unary plus operator.
3436 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3437 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3438 constants.) Note, these suffixes appear in macros defined in the system
3439 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3440 Use of these macros in user code might normally lead to spurious
3441 warnings, however GCC's integrated preprocessor has enough context to
3442 avoid warning in these cases.
3445 A function declared external in one block and then used after the end of
3449 A @code{switch} statement has an operand of type @code{long}.
3452 A non-@code{static} function declaration follows a @code{static} one.
3453 This construct is not accepted by some traditional C compilers.
3456 The ISO type of an integer constant has a different width or
3457 signedness from its traditional type. This warning is only issued if
3458 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3459 typically represent bit patterns, are not warned about.
3462 Usage of ISO string concatenation is detected.
3465 Initialization of automatic aggregates.
3468 Identifier conflicts with labels. Traditional C lacks a separate
3469 namespace for labels.
3472 Initialization of unions. If the initializer is zero, the warning is
3473 omitted. This is done under the assumption that the zero initializer in
3474 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3475 initializer warnings and relies on default initialization to zero in the
3479 Conversions by prototypes between fixed/floating point values and vice
3480 versa. The absence of these prototypes when compiling with traditional
3481 C would cause serious problems. This is a subset of the possible
3482 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3485 Use of ISO C style function definitions. This warning intentionally is
3486 @emph{not} issued for prototype declarations or variadic functions
3487 because these ISO C features will appear in your code when using
3488 libiberty's traditional C compatibility macros, @code{PARAMS} and
3489 @code{VPARAMS}. This warning is also bypassed for nested functions
3490 because that feature is already a GCC extension and thus not relevant to
3491 traditional C compatibility.
3494 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3495 @opindex Wtraditional-conversion
3496 @opindex Wno-traditional-conversion
3497 Warn if a prototype causes a type conversion that is different from what
3498 would happen to the same argument in the absence of a prototype. This
3499 includes conversions of fixed point to floating and vice versa, and
3500 conversions changing the width or signedness of a fixed point argument
3501 except when the same as the default promotion.
3503 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3504 @opindex Wdeclaration-after-statement
3505 @opindex Wno-declaration-after-statement
3506 Warn when a declaration is found after a statement in a block. This
3507 construct, known from C++, was introduced with ISO C99 and is by default
3508 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3509 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3514 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3516 @item -Wno-endif-labels
3517 @opindex Wno-endif-labels
3518 @opindex Wendif-labels
3519 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3524 Warn whenever a local variable shadows another local variable, parameter or
3525 global variable or whenever a built-in function is shadowed.
3527 @item -Wlarger-than=@var{len}
3528 @opindex Wlarger-than=@var{len}
3529 @opindex Wlarger-than-@var{len}
3530 Warn whenever an object of larger than @var{len} bytes is defined.
3532 @item -Wframe-larger-than=@var{len}
3533 @opindex Wframe-larger-than
3534 Warn whenever the size of a function frame is larger than @var{len} bytes.
3536 @item -Wunsafe-loop-optimizations
3537 @opindex Wunsafe-loop-optimizations
3538 @opindex Wno-unsafe-loop-optimizations
3539 Warn if the loop cannot be optimized because the compiler could not
3540 assume anything on the bounds of the loop indices. With
3541 @option{-funsafe-loop-optimizations} warn if the compiler made
3544 @item -Wpointer-arith
3545 @opindex Wpointer-arith
3546 @opindex Wno-pointer-arith
3547 Warn about anything that depends on the ``size of'' a function type or
3548 of @code{void}. GNU C assigns these types a size of 1, for
3549 convenience in calculations with @code{void *} pointers and pointers
3550 to functions. In C++, warn also when an arithmetic operation involves
3551 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3554 @opindex Wtype-limits
3555 @opindex Wno-type-limits
3556 Warn if a comparison is always true or always false due to the limited
3557 range of the data type, but do not warn for constant expressions. For
3558 example, warn if an unsigned variable is compared against zero with
3559 @samp{<} or @samp{>=}. This warning is also enabled by
3562 @item -Wbad-function-cast @r{(C and Objective-C only)}
3563 @opindex Wbad-function-cast
3564 @opindex Wno-bad-function-cast
3565 Warn whenever a function call is cast to a non-matching type.
3566 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3568 @item -Wc++-compat @r{(C and Objective-C only)}
3569 Warn about ISO C constructs that are outside of the common subset of
3570 ISO C and ISO C++, e.g.@: request for implicit conversion from
3571 @code{void *} to a pointer to non-@code{void} type.
3573 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3574 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3575 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3576 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3580 @opindex Wno-cast-qual
3581 Warn whenever a pointer is cast so as to remove a type qualifier from
3582 the target type. For example, warn if a @code{const char *} is cast
3583 to an ordinary @code{char *}.
3586 @opindex Wcast-align
3587 @opindex Wno-cast-align
3588 Warn whenever a pointer is cast such that the required alignment of the
3589 target is increased. For example, warn if a @code{char *} is cast to
3590 an @code{int *} on machines where integers can only be accessed at
3591 two- or four-byte boundaries.
3593 @item -Wwrite-strings
3594 @opindex Wwrite-strings
3595 @opindex Wno-write-strings
3596 When compiling C, give string constants the type @code{const
3597 char[@var{length}]} so that
3598 copying the address of one into a non-@code{const} @code{char *}
3599 pointer will get a warning; when compiling C++, warn about the
3600 deprecated conversion from string literals to @code{char *}. This
3601 warning, by default, is enabled for C++ programs.
3602 These warnings will help you find at
3603 compile time code that can try to write into a string constant, but
3604 only if you have been very careful about using @code{const} in
3605 declarations and prototypes. Otherwise, it will just be a nuisance;
3606 this is why we did not make @option{-Wall} request these warnings.
3610 @opindex Wno-clobbered
3611 Warn for variables that might be changed by @samp{longjmp} or
3612 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3615 @opindex Wconversion
3616 @opindex Wno-conversion
3617 Warn for implicit conversions that may alter a value. This includes
3618 conversions between real and integer, like @code{abs (x)} when
3619 @code{x} is @code{double}; conversions between signed and unsigned,
3620 like @code{unsigned ui = -1}; and conversions to smaller types, like
3621 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3622 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3623 changed by the conversion like in @code{abs (2.0)}. Warnings about
3624 conversions between signed and unsigned integers can be disabled by
3625 using @option{-Wno-sign-conversion}.
3627 For C++, also warn for conversions between @code{NULL} and non-pointer
3628 types; confusing overload resolution for user-defined conversions; and
3629 conversions that will never use a type conversion operator:
3630 conversions to @code{void}, the same type, a base class or a reference
3631 to them. Warnings about conversions between signed and unsigned
3632 integers are disabled by default in C++ unless
3633 @option{-Wsign-conversion} is explicitly enabled.
3636 @opindex Wempty-body
3637 @opindex Wno-empty-body
3638 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3639 while} statement. Additionally, in C++, warn when an empty body occurs
3640 in a @samp{while} or @samp{for} statement with no whitespacing before
3641 the semicolon. This warning is also enabled by @option{-Wextra}.
3643 @item -Wsign-compare
3644 @opindex Wsign-compare
3645 @opindex Wno-sign-compare
3646 @cindex warning for comparison of signed and unsigned values
3647 @cindex comparison of signed and unsigned values, warning
3648 @cindex signed and unsigned values, comparison warning
3649 Warn when a comparison between signed and unsigned values could produce
3650 an incorrect result when the signed value is converted to unsigned.
3651 This warning is also enabled by @option{-Wextra}; to get the other warnings
3652 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3654 @item -Wsign-conversion
3655 @opindex Wsign-conversion
3656 @opindex Wno-sign-conversion
3657 Warn for implicit conversions that may change the sign of an integer
3658 value, like assigning a signed integer expression to an unsigned
3659 integer variable. An explicit cast silences the warning. In C, this
3660 option is enabled also by @option{-Wconversion}.
3664 @opindex Wno-address
3665 Warn about suspicious uses of memory addresses. These include using
3666 the address of a function in a conditional expression, such as
3667 @code{void func(void); if (func)}, and comparisons against the memory
3668 address of a string literal, such as @code{if (x == "abc")}. Such
3669 uses typically indicate a programmer error: the address of a function
3670 always evaluates to true, so their use in a conditional usually
3671 indicate that the programmer forgot the parentheses in a function
3672 call; and comparisons against string literals result in unspecified
3673 behavior and are not portable in C, so they usually indicate that the
3674 programmer intended to use @code{strcmp}. This warning is enabled by
3678 @opindex Wlogical-op
3679 @opindex Wno-logical-op
3680 Warn about suspicious uses of logical operators in expressions.
3681 This includes using logical operators in contexts where a
3682 bit-wise operator is likely to be expected.
3684 @item -Waggregate-return
3685 @opindex Waggregate-return
3686 @opindex Wno-aggregate-return
3687 Warn if any functions that return structures or unions are defined or
3688 called. (In languages where you can return an array, this also elicits
3691 @item -Wno-attributes
3692 @opindex Wno-attributes
3693 @opindex Wattributes
3694 Do not warn if an unexpected @code{__attribute__} is used, such as
3695 unrecognized attributes, function attributes applied to variables,
3696 etc. This will not stop errors for incorrect use of supported
3699 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3700 @opindex Wstrict-prototypes
3701 @opindex Wno-strict-prototypes
3702 Warn if a function is declared or defined without specifying the
3703 argument types. (An old-style function definition is permitted without
3704 a warning if preceded by a declaration which specifies the argument
3707 @item -Wold-style-declaration @r{(C and Objective-C only)}
3708 @opindex Wold-style-declaration
3709 @opindex Wno-old-style-declaration
3710 Warn for obsolescent usages, according to the C Standard, in a
3711 declaration. For example, warn if storage-class specifiers like
3712 @code{static} are not the first things in a declaration. This warning
3713 is also enabled by @option{-Wextra}.
3715 @item -Wold-style-definition @r{(C and Objective-C only)}
3716 @opindex Wold-style-definition
3717 @opindex Wno-old-style-definition
3718 Warn if an old-style function definition is used. A warning is given
3719 even if there is a previous prototype.
3721 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3722 @opindex Wmissing-parameter-type
3723 @opindex Wno-missing-parameter-type
3724 A function parameter is declared without a type specifier in K&R-style
3731 This warning is also enabled by @option{-Wextra}.
3733 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3734 @opindex Wmissing-prototypes
3735 @opindex Wno-missing-prototypes
3736 Warn if a global function is defined without a previous prototype
3737 declaration. This warning is issued even if the definition itself
3738 provides a prototype. The aim is to detect global functions that fail
3739 to be declared in header files.
3741 @item -Wmissing-declarations
3742 @opindex Wmissing-declarations
3743 @opindex Wno-missing-declarations
3744 Warn if a global function is defined without a previous declaration.
3745 Do so even if the definition itself provides a prototype.
3746 Use this option to detect global functions that are not declared in
3747 header files. In C++, no warnings are issued for function templates,
3748 or for inline functions, or for functions in anonymous namespaces.
3750 @item -Wmissing-field-initializers
3751 @opindex Wmissing-field-initializers
3752 @opindex Wno-missing-field-initializers
3756 Warn if a structure's initializer has some fields missing. For
3757 example, the following code would cause such a warning, because
3758 @code{x.h} is implicitly zero:
3761 struct s @{ int f, g, h; @};
3762 struct s x = @{ 3, 4 @};
3765 This option does not warn about designated initializers, so the following
3766 modification would not trigger a warning:
3769 struct s @{ int f, g, h; @};
3770 struct s x = @{ .f = 3, .g = 4 @};
3773 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3774 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3776 @item -Wmissing-noreturn
3777 @opindex Wmissing-noreturn
3778 @opindex Wno-missing-noreturn
3779 Warn about functions which might be candidates for attribute @code{noreturn}.
3780 Note these are only possible candidates, not absolute ones. Care should
3781 be taken to manually verify functions actually do not ever return before
3782 adding the @code{noreturn} attribute, otherwise subtle code generation
3783 bugs could be introduced. You will not get a warning for @code{main} in
3784 hosted C environments.
3786 @item -Wmissing-format-attribute
3787 @opindex Wmissing-format-attribute
3788 @opindex Wno-missing-format-attribute
3791 Warn about function pointers which might be candidates for @code{format}
3792 attributes. Note these are only possible candidates, not absolute ones.
3793 GCC will guess that function pointers with @code{format} attributes that
3794 are used in assignment, initialization, parameter passing or return
3795 statements should have a corresponding @code{format} attribute in the
3796 resulting type. I.e.@: the left-hand side of the assignment or
3797 initialization, the type of the parameter variable, or the return type
3798 of the containing function respectively should also have a @code{format}
3799 attribute to avoid the warning.
3801 GCC will also warn about function definitions which might be
3802 candidates for @code{format} attributes. Again, these are only
3803 possible candidates. GCC will guess that @code{format} attributes
3804 might be appropriate for any function that calls a function like
3805 @code{vprintf} or @code{vscanf}, but this might not always be the
3806 case, and some functions for which @code{format} attributes are
3807 appropriate may not be detected.
3809 @item -Wno-multichar
3810 @opindex Wno-multichar
3812 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3813 Usually they indicate a typo in the user's code, as they have
3814 implementation-defined values, and should not be used in portable code.
3816 @item -Wnormalized=<none|id|nfc|nfkc>
3817 @opindex Wnormalized=
3820 @cindex character set, input normalization
3821 In ISO C and ISO C++, two identifiers are different if they are
3822 different sequences of characters. However, sometimes when characters
3823 outside the basic ASCII character set are used, you can have two
3824 different character sequences that look the same. To avoid confusion,
3825 the ISO 10646 standard sets out some @dfn{normalization rules} which
3826 when applied ensure that two sequences that look the same are turned into
3827 the same sequence. GCC can warn you if you are using identifiers which
3828 have not been normalized; this option controls that warning.
3830 There are four levels of warning that GCC supports. The default is
3831 @option{-Wnormalized=nfc}, which warns about any identifier which is
3832 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3833 recommended form for most uses.
3835 Unfortunately, there are some characters which ISO C and ISO C++ allow
3836 in identifiers that when turned into NFC aren't allowable as
3837 identifiers. That is, there's no way to use these symbols in portable
3838 ISO C or C++ and have all your identifiers in NFC@.
3839 @option{-Wnormalized=id} suppresses the warning for these characters.
3840 It is hoped that future versions of the standards involved will correct
3841 this, which is why this option is not the default.
3843 You can switch the warning off for all characters by writing
3844 @option{-Wnormalized=none}. You would only want to do this if you
3845 were using some other normalization scheme (like ``D''), because
3846 otherwise you can easily create bugs that are literally impossible to see.
3848 Some characters in ISO 10646 have distinct meanings but look identical
3849 in some fonts or display methodologies, especially once formatting has
3850 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3851 LETTER N'', will display just like a regular @code{n} which has been
3852 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3853 normalization scheme to convert all these into a standard form as
3854 well, and GCC will warn if your code is not in NFKC if you use
3855 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3856 about every identifier that contains the letter O because it might be
3857 confused with the digit 0, and so is not the default, but may be
3858 useful as a local coding convention if the programming environment is
3859 unable to be fixed to display these characters distinctly.
3861 @item -Wno-deprecated-declarations
3862 @opindex Wno-deprecated-declarations
3863 @opindex Wdeprecated-declarations
3864 Do not warn about uses of functions (@pxref{Function Attributes}),
3865 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3866 Attributes}) marked as deprecated by using the @code{deprecated}
3870 @opindex Wno-overflow
3872 Do not warn about compile-time overflow in constant expressions.
3874 @item -Woverride-init @r{(C and Objective-C only)}
3875 @opindex Woverride-init
3876 @opindex Wno-override-init
3880 Warn if an initialized field without side effects is overridden when
3881 using designated initializers (@pxref{Designated Inits, , Designated
3884 This warning is included in @option{-Wextra}. To get other
3885 @option{-Wextra} warnings without this one, use @samp{-Wextra
3886 -Wno-override-init}.
3891 Warn if a structure is given the packed attribute, but the packed
3892 attribute has no effect on the layout or size of the structure.
3893 Such structures may be mis-aligned for little benefit. For
3894 instance, in this code, the variable @code{f.x} in @code{struct bar}
3895 will be misaligned even though @code{struct bar} does not itself
3896 have the packed attribute:
3903 @} __attribute__((packed));
3914 Warn if padding is included in a structure, either to align an element
3915 of the structure or to align the whole structure. Sometimes when this
3916 happens it is possible to rearrange the fields of the structure to
3917 reduce the padding and so make the structure smaller.
3919 @item -Wredundant-decls
3920 @opindex Wredundant-decls
3921 @opindex Wno-redundant-decls
3922 Warn if anything is declared more than once in the same scope, even in
3923 cases where multiple declaration is valid and changes nothing.
3925 @item -Wnested-externs @r{(C and Objective-C only)}
3926 @opindex Wnested-externs
3927 @opindex Wno-nested-externs
3928 Warn if an @code{extern} declaration is encountered within a function.
3930 @item -Wunreachable-code
3931 @opindex Wunreachable-code
3932 @opindex Wno-unreachable-code
3933 Warn if the compiler detects that code will never be executed.
3935 This option is intended to warn when the compiler detects that at
3936 least a whole line of source code will never be executed, because
3937 some condition is never satisfied or because it is after a
3938 procedure that never returns.
3940 It is possible for this option to produce a warning even though there
3941 are circumstances under which part of the affected line can be executed,
3942 so care should be taken when removing apparently-unreachable code.
3944 For instance, when a function is inlined, a warning may mean that the
3945 line is unreachable in only one inlined copy of the function.
3947 This option is not made part of @option{-Wall} because in a debugging
3948 version of a program there is often substantial code which checks
3949 correct functioning of the program and is, hopefully, unreachable
3950 because the program does work. Another common use of unreachable
3951 code is to provide behavior which is selectable at compile-time.
3956 Warn if a function can not be inlined and it was declared as inline.
3957 Even with this option, the compiler will not warn about failures to
3958 inline functions declared in system headers.
3960 The compiler uses a variety of heuristics to determine whether or not
3961 to inline a function. For example, the compiler takes into account
3962 the size of the function being inlined and the amount of inlining
3963 that has already been done in the current function. Therefore,
3964 seemingly insignificant changes in the source program can cause the
3965 warnings produced by @option{-Winline} to appear or disappear.
3967 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
3968 @opindex Wno-invalid-offsetof
3969 @opindex Winvalid-offsetof
3970 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3971 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3972 to a non-POD type is undefined. In existing C++ implementations,
3973 however, @samp{offsetof} typically gives meaningful results even when
3974 applied to certain kinds of non-POD types. (Such as a simple
3975 @samp{struct} that fails to be a POD type only by virtue of having a
3976 constructor.) This flag is for users who are aware that they are
3977 writing nonportable code and who have deliberately chosen to ignore the
3980 The restrictions on @samp{offsetof} may be relaxed in a future version
3981 of the C++ standard.
3983 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
3984 @opindex Wno-int-to-pointer-cast
3985 @opindex Wint-to-pointer-cast
3986 Suppress warnings from casts to pointer type of an integer of a
3989 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
3990 @opindex Wno-pointer-to-int-cast
3991 @opindex Wpointer-to-int-cast
3992 Suppress warnings from casts from a pointer to an integer type of a
3996 @opindex Winvalid-pch
3997 @opindex Wno-invalid-pch
3998 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3999 the search path but can't be used.
4003 @opindex Wno-long-long
4004 Warn if @samp{long long} type is used. This is default. To inhibit
4005 the warning messages, use @option{-Wno-long-long}. Flags
4006 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4007 only when @option{-pedantic} flag is used.
4009 @item -Wvariadic-macros
4010 @opindex Wvariadic-macros
4011 @opindex Wno-variadic-macros
4012 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4013 alternate syntax when in pedantic ISO C99 mode. This is default.
4014 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4019 Warn if variable length array is used in the code.
4020 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4021 the variable length array.
4023 @item -Wvolatile-register-var
4024 @opindex Wvolatile-register-var
4025 @opindex Wno-volatile-register-var
4026 Warn if a register variable is declared volatile. The volatile
4027 modifier does not inhibit all optimizations that may eliminate reads
4028 and/or writes to register variables. This warning is enabled by
4031 @item -Wdisabled-optimization
4032 @opindex Wdisabled-optimization
4033 @opindex Wno-disabled-optimization
4034 Warn if a requested optimization pass is disabled. This warning does
4035 not generally indicate that there is anything wrong with your code; it
4036 merely indicates that GCC's optimizers were unable to handle the code
4037 effectively. Often, the problem is that your code is too big or too
4038 complex; GCC will refuse to optimize programs when the optimization
4039 itself is likely to take inordinate amounts of time.
4041 @item -Wpointer-sign @r{(C and Objective-C only)}
4042 @opindex Wpointer-sign
4043 @opindex Wno-pointer-sign
4044 Warn for pointer argument passing or assignment with different signedness.
4045 This option is only supported for C and Objective-C@. It is implied by
4046 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4047 @option{-Wno-pointer-sign}.
4049 @item -Wstack-protector
4050 @opindex Wstack-protector
4051 @opindex Wno-stack-protector
4052 This option is only active when @option{-fstack-protector} is active. It
4053 warns about functions that will not be protected against stack smashing.
4056 @opindex Wno-mudflap
4057 Suppress warnings about constructs that cannot be instrumented by
4060 @item -Woverlength-strings
4061 @opindex Woverlength-strings
4062 @opindex Wno-overlength-strings
4063 Warn about string constants which are longer than the ``minimum
4064 maximum'' length specified in the C standard. Modern compilers
4065 generally allow string constants which are much longer than the
4066 standard's minimum limit, but very portable programs should avoid
4067 using longer strings.
4069 The limit applies @emph{after} string constant concatenation, and does
4070 not count the trailing NUL@. In C89, the limit was 509 characters; in
4071 C99, it was raised to 4095. C++98 does not specify a normative
4072 minimum maximum, so we do not diagnose overlength strings in C++@.
4074 This option is implied by @option{-pedantic}, and can be disabled with
4075 @option{-Wno-overlength-strings}.
4078 @node Debugging Options
4079 @section Options for Debugging Your Program or GCC
4080 @cindex options, debugging
4081 @cindex debugging information options
4083 GCC has various special options that are used for debugging
4084 either your program or GCC:
4089 Produce debugging information in the operating system's native format
4090 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4093 On most systems that use stabs format, @option{-g} enables use of extra
4094 debugging information that only GDB can use; this extra information
4095 makes debugging work better in GDB but will probably make other debuggers
4097 refuse to read the program. If you want to control for certain whether
4098 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4099 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4101 GCC allows you to use @option{-g} with
4102 @option{-O}. The shortcuts taken by optimized code may occasionally
4103 produce surprising results: some variables you declared may not exist
4104 at all; flow of control may briefly move where you did not expect it;
4105 some statements may not be executed because they compute constant
4106 results or their values were already at hand; some statements may
4107 execute in different places because they were moved out of loops.
4109 Nevertheless it proves possible to debug optimized output. This makes
4110 it reasonable to use the optimizer for programs that might have bugs.
4112 The following options are useful when GCC is generated with the
4113 capability for more than one debugging format.
4117 Produce debugging information for use by GDB@. This means to use the
4118 most expressive format available (DWARF 2, stabs, or the native format
4119 if neither of those are supported), including GDB extensions if at all
4124 Produce debugging information in stabs format (if that is supported),
4125 without GDB extensions. This is the format used by DBX on most BSD
4126 systems. On MIPS, Alpha and System V Release 4 systems this option
4127 produces stabs debugging output which is not understood by DBX or SDB@.
4128 On System V Release 4 systems this option requires the GNU assembler.
4130 @item -feliminate-unused-debug-symbols
4131 @opindex feliminate-unused-debug-symbols
4132 Produce debugging information in stabs format (if that is supported),
4133 for only symbols that are actually used.
4135 @item -femit-class-debug-always
4136 Instead of emitting debugging information for a C++ class in only one
4137 object file, emit it in all object files using the class. This option
4138 should be used only with debuggers that are unable to handle the way GCC
4139 normally emits debugging information for classes because using this
4140 option will increase the size of debugging information by as much as a
4145 Produce debugging information in stabs format (if that is supported),
4146 using GNU extensions understood only by the GNU debugger (GDB)@. The
4147 use of these extensions is likely to make other debuggers crash or
4148 refuse to read the program.
4152 Produce debugging information in COFF format (if that is supported).
4153 This is the format used by SDB on most System V systems prior to
4158 Produce debugging information in XCOFF format (if that is supported).
4159 This is the format used by the DBX debugger on IBM RS/6000 systems.
4163 Produce debugging information in XCOFF format (if that is supported),
4164 using GNU extensions understood only by the GNU debugger (GDB)@. The
4165 use of these extensions is likely to make other debuggers crash or
4166 refuse to read the program, and may cause assemblers other than the GNU
4167 assembler (GAS) to fail with an error.
4171 Produce debugging information in DWARF version 2 format (if that is
4172 supported). This is the format used by DBX on IRIX 6. With this
4173 option, GCC uses features of DWARF version 3 when they are useful;
4174 version 3 is upward compatible with version 2, but may still cause
4175 problems for older debuggers.
4179 Produce debugging information in VMS debug format (if that is
4180 supported). This is the format used by DEBUG on VMS systems.
4183 @itemx -ggdb@var{level}
4184 @itemx -gstabs@var{level}
4185 @itemx -gcoff@var{level}
4186 @itemx -gxcoff@var{level}
4187 @itemx -gvms@var{level}
4188 Request debugging information and also use @var{level} to specify how
4189 much information. The default level is 2.
4191 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4194 Level 1 produces minimal information, enough for making backtraces in
4195 parts of the program that you don't plan to debug. This includes
4196 descriptions of functions and external variables, but no information
4197 about local variables and no line numbers.
4199 Level 3 includes extra information, such as all the macro definitions
4200 present in the program. Some debuggers support macro expansion when
4201 you use @option{-g3}.
4203 @option{-gdwarf-2} does not accept a concatenated debug level, because
4204 GCC used to support an option @option{-gdwarf} that meant to generate
4205 debug information in version 1 of the DWARF format (which is very
4206 different from version 2), and it would have been too confusing. That
4207 debug format is long obsolete, but the option cannot be changed now.
4208 Instead use an additional @option{-g@var{level}} option to change the
4209 debug level for DWARF2.
4211 @item -feliminate-dwarf2-dups
4212 @opindex feliminate-dwarf2-dups
4213 Compress DWARF2 debugging information by eliminating duplicated
4214 information about each symbol. This option only makes sense when
4215 generating DWARF2 debugging information with @option{-gdwarf-2}.
4217 @item -femit-struct-debug-baseonly
4218 Emit debug information for struct-like types
4219 only when the base name of the compilation source file
4220 matches the base name of file in which the struct was defined.
4222 This option substantially reduces the size of debugging information,
4223 but at significant potential loss in type information to the debugger.
4224 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4225 See @option{-femit-struct-debug-detailed} for more detailed control.
4227 This option works only with DWARF 2.
4229 @item -femit-struct-debug-reduced
4230 Emit debug information for struct-like types
4231 only when the base name of the compilation source file
4232 matches the base name of file in which the type was defined,
4233 unless the struct is a template or defined in a system header.
4235 This option significantly reduces the size of debugging information,
4236 with some potential loss in type information to the debugger.
4237 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4238 See @option{-femit-struct-debug-detailed} for more detailed control.
4240 This option works only with DWARF 2.
4242 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4243 Specify the struct-like types
4244 for which the compiler will generate debug information.
4245 The intent is to reduce duplicate struct debug information
4246 between different object files within the same program.
4248 This option is a detailed version of
4249 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4250 which will serve for most needs.
4252 A specification has the syntax
4253 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4255 The optional first word limits the specification to
4256 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4257 A struct type is used directly when it is the type of a variable, member.
4258 Indirect uses arise through pointers to structs.
4259 That is, when use of an incomplete struct would be legal, the use is indirect.
4261 @samp{struct one direct; struct two * indirect;}.
4263 The optional second word limits the specification to
4264 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4265 Generic structs are a bit complicated to explain.
4266 For C++, these are non-explicit specializations of template classes,
4267 or non-template classes within the above.
4268 Other programming languages have generics,
4269 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4271 The third word specifies the source files for those
4272 structs for which the compiler will emit debug information.
4273 The values @samp{none} and @samp{any} have the normal meaning.
4274 The value @samp{base} means that
4275 the base of name of the file in which the type declaration appears
4276 must match the base of the name of the main compilation file.
4277 In practice, this means that
4278 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4279 but types declared in other header will not.
4280 The value @samp{sys} means those types satisfying @samp{base}
4281 or declared in system or compiler headers.
4283 You may need to experiment to determine the best settings for your application.
4285 The default is @samp{-femit-struct-debug-detailed=all}.
4287 This option works only with DWARF 2.
4289 @item -fno-merge-debug-strings
4290 @opindex fmerge-debug-strings
4291 @opindex fno-merge-debug-strings
4292 Direct the linker to merge together strings which are identical in
4293 different object files. This is not supported by all assemblers or
4294 linker. This decreases the size of the debug information in the
4295 output file at the cost of increasing link processing time. This is
4298 @item -fdebug-prefix-map=@var{old}=@var{new}
4299 @opindex fdebug-prefix-map
4300 When compiling files in directory @file{@var{old}}, record debugging
4301 information describing them as in @file{@var{new}} instead.
4303 @cindex @command{prof}
4306 Generate extra code to write profile information suitable for the
4307 analysis program @command{prof}. You must use this option when compiling
4308 the source files you want data about, and you must also use it when
4311 @cindex @command{gprof}
4314 Generate extra code to write profile information suitable for the
4315 analysis program @command{gprof}. You must use this option when compiling
4316 the source files you want data about, and you must also use it when
4321 Makes the compiler print out each function name as it is compiled, and
4322 print some statistics about each pass when it finishes.
4325 @opindex ftime-report
4326 Makes the compiler print some statistics about the time consumed by each
4327 pass when it finishes.
4330 @opindex fmem-report
4331 Makes the compiler print some statistics about permanent memory
4332 allocation when it finishes.
4334 @item -fpre-ipa-mem-report
4335 @opindex fpre-ipa-mem-report
4336 @item -fpost-ipa-mem-report
4337 @opindex fpost-ipa-mem-report
4338 Makes the compiler print some statistics about permanent memory
4339 allocation before or after interprocedural optimization.
4341 @item -fprofile-arcs
4342 @opindex fprofile-arcs
4343 Add code so that program flow @dfn{arcs} are instrumented. During
4344 execution the program records how many times each branch and call is
4345 executed and how many times it is taken or returns. When the compiled
4346 program exits it saves this data to a file called
4347 @file{@var{auxname}.gcda} for each source file. The data may be used for
4348 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4349 test coverage analysis (@option{-ftest-coverage}). Each object file's
4350 @var{auxname} is generated from the name of the output file, if
4351 explicitly specified and it is not the final executable, otherwise it is
4352 the basename of the source file. In both cases any suffix is removed
4353 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4354 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4355 @xref{Cross-profiling}.
4357 @cindex @command{gcov}
4361 This option is used to compile and link code instrumented for coverage
4362 analysis. The option is a synonym for @option{-fprofile-arcs}
4363 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4364 linking). See the documentation for those options for more details.
4369 Compile the source files with @option{-fprofile-arcs} plus optimization
4370 and code generation options. For test coverage analysis, use the
4371 additional @option{-ftest-coverage} option. You do not need to profile
4372 every source file in a program.
4375 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4376 (the latter implies the former).
4379 Run the program on a representative workload to generate the arc profile
4380 information. This may be repeated any number of times. You can run
4381 concurrent instances of your program, and provided that the file system
4382 supports locking, the data files will be correctly updated. Also
4383 @code{fork} calls are detected and correctly handled (double counting
4387 For profile-directed optimizations, compile the source files again with
4388 the same optimization and code generation options plus
4389 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4390 Control Optimization}).
4393 For test coverage analysis, use @command{gcov} to produce human readable
4394 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4395 @command{gcov} documentation for further information.
4399 With @option{-fprofile-arcs}, for each function of your program GCC
4400 creates a program flow graph, then finds a spanning tree for the graph.
4401 Only arcs that are not on the spanning tree have to be instrumented: the
4402 compiler adds code to count the number of times that these arcs are
4403 executed. When an arc is the only exit or only entrance to a block, the
4404 instrumentation code can be added to the block; otherwise, a new basic
4405 block must be created to hold the instrumentation code.
4408 @item -ftest-coverage
4409 @opindex ftest-coverage
4410 Produce a notes file that the @command{gcov} code-coverage utility
4411 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4412 show program coverage. Each source file's note file is called
4413 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4414 above for a description of @var{auxname} and instructions on how to
4415 generate test coverage data. Coverage data will match the source files
4416 more closely, if you do not optimize.
4418 @item -fdbg-cnt-list
4419 @opindex fdbg-cnt-list
4420 Print the name and the counter upperbound for all debug counters.
4422 @item -fdbg-cnt=@var{counter-value-list}
4424 Set the internal debug counter upperbound. @var{counter-value-list}
4425 is a comma-separated list of @var{name}:@var{value} pairs
4426 which sets the upperbound of each debug counter @var{name} to @var{value}.
4427 All debug counters have the initial upperbound of @var{UINT_MAX},
4428 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4429 e.g. With -fdbg-cnt=dce:10,tail_call:0
4430 dbg_cnt(dce) will return true only for first 10 invocations
4431 and dbg_cnt(tail_call) will return false always.
4433 @item -d@var{letters}
4434 @item -fdump-rtl-@var{pass}
4436 Says to make debugging dumps during compilation at times specified by
4437 @var{letters}. This is used for debugging the RTL-based passes of the
4438 compiler. The file names for most of the dumps are made by appending a
4439 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4440 from the name of the output file, if explicitly specified and it is not
4441 an executable, otherwise it is the basename of the source file. These
4442 switches may have different effects when @option{-E} is used for
4445 Most debug dumps can be enabled either passing a letter to the @option{-d}
4446 option, or with a long @option{-fdump-rtl} switch; here are the possible
4447 letters for use in @var{letters} and @var{pass}, and their meanings:
4452 Annotate the assembler output with miscellaneous debugging information.
4454 @itemx -fdump-rtl-bbro
4455 @opindex fdump-rtl-bbro
4456 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4458 @itemx -fdump-rtl-combine
4459 @opindex fdump-rtl-combine
4460 Dump after the RTL instruction combination pass, to the file
4461 @file{@var{file}.129r.combine}.
4463 @itemx -fdump-rtl-ce1
4464 @itemx -fdump-rtl-ce2
4465 @opindex fdump-rtl-ce1
4466 @opindex fdump-rtl-ce2
4467 @option{-fdump-rtl-ce1} enable dumping after the
4468 first if conversion, to the file @file{@var{file}.117r.ce1}.
4469 @option{-fdump-rtl-ce2} enable dumping after the second if
4470 conversion, to the file @file{@var{file}.130r.ce2}.
4472 @itemx -fdump-rtl-btl
4473 @itemx -fdump-rtl-dbr
4474 @opindex fdump-rtl-btl
4475 @opindex fdump-rtl-dbr
4476 @option{-fdump-rtl-btl} enable dumping after branch
4477 target load optimization, to @file{@var{file}.31.btl}.
4478 @option{-fdump-rtl-dbr} enable dumping after delayed branch
4479 scheduling, to @file{@var{file}.36.dbr}.
4483 Dump all macro definitions, at the end of preprocessing, in addition to
4486 @itemx -fdump-rtl-ce3
4487 @opindex fdump-rtl-ce3
4488 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4490 @itemx -fdump-rtl-cfg
4491 @itemx -fdump-rtl-life
4492 @opindex fdump-rtl-cfg
4493 @opindex fdump-rtl-life
4494 @option{-fdump-rtl-cfg} enable dumping after control
4495 and data flow analysis, to @file{@var{file}.116r.cfg}.
4496 @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4497 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4499 @itemx -fdump-rtl-greg
4500 @opindex fdump-rtl-greg
4501 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4503 @itemx -fdump-rtl-gcse
4504 @itemx -fdump-rtl-bypass
4505 @opindex fdump-rtl-gcse
4506 @opindex fdump-rtl-bypass
4507 @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4508 @file{@var{file}.114r.gcse}. @option{-fdump-rtl-bypass}
4509 enable dumping after jump bypassing and control flow optimizations, to
4510 @file{@var{file}.115r.bypass}.
4512 @itemx -fdump-rtl-eh
4513 @opindex fdump-rtl-eh
4514 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4516 @itemx -fdump-rtl-sibling
4517 @opindex fdump-rtl-sibling
4518 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4520 @itemx -fdump-rtl-jump
4521 @opindex fdump-rtl-jump
4522 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4524 @itemx -fdump-rtl-stack
4525 @opindex fdump-rtl-stack
4526 Dump after conversion from GCC's "flat register file" registers to the
4527 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4529 @itemx -fdump-rtl-lreg
4530 @opindex fdump-rtl-lreg
4531 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4533 @itemx -fdump-rtl-loop2
4534 @opindex fdump-rtl-loop2
4535 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4536 loop optimization pass, to @file{@var{file}.119r.loop2},
4537 @file{@var{file}.120r.loop2_init},
4538 @file{@var{file}.121r.loop2_invariant}, and
4539 @file{@var{file}.125r.loop2_done}.
4541 @itemx -fdump-rtl-sms
4542 @opindex fdump-rtl-sms
4543 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4545 @itemx -fdump-rtl-mach
4546 @opindex fdump-rtl-mach
4547 Dump after performing the machine dependent reorganization pass, to
4548 @file{@var{file}.155r.mach} if that pass exists.
4550 @itemx -fdump-rtl-rnreg
4551 @opindex fdump-rtl-rnreg
4552 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4554 @itemx -fdump-rtl-regmove
4555 @opindex fdump-rtl-regmove
4556 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4558 @itemx -fdump-rtl-postreload
4559 @opindex fdump-rtl-postreload
4560 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4562 @itemx -fdump-rtl-expand
4563 @opindex fdump-rtl-expand
4564 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4566 @itemx -fdump-rtl-sched2
4567 @opindex fdump-rtl-sched2
4568 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4570 @itemx -fdump-rtl-cse
4571 @opindex fdump-rtl-cse
4572 Dump after CSE (including the jump optimization that sometimes follows
4573 CSE), to @file{@var{file}.113r.cse}.
4575 @itemx -fdump-rtl-sched1
4576 @opindex fdump-rtl-sched1
4577 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4579 @itemx -fdump-rtl-cse2
4580 @opindex fdump-rtl-cse2
4581 Dump after the second CSE pass (including the jump optimization that
4582 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4584 @itemx -fdump-rtl-tracer
4585 @opindex fdump-rtl-tracer
4586 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4588 @itemx -fdump-rtl-vpt
4589 @itemx -fdump-rtl-vartrack
4590 @opindex fdump-rtl-vpt
4591 @opindex fdump-rtl-vartrack
4592 @option{-fdump-rtl-vpt} enable dumping after the value
4593 profile transformations, to @file{@var{file}.10.vpt}.
4594 @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4595 to @file{@var{file}.154r.vartrack}.
4597 @itemx -fdump-rtl-flow2
4598 @opindex fdump-rtl-flow2
4599 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4601 @itemx -fdump-rtl-peephole2
4602 @opindex fdump-rtl-peephole2
4603 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4605 @itemx -fdump-rtl-web
4606 @opindex fdump-rtl-web
4607 Dump after live range splitting, to @file{@var{file}.126r.web}.
4609 @itemx -fdump-rtl-all
4610 @opindex fdump-rtl-all
4611 Produce all the dumps listed above.
4615 Produce a core dump whenever an error occurs.
4619 Print statistics on memory usage, at the end of the run, to
4624 Annotate the assembler output with a comment indicating which
4625 pattern and alternative was used. The length of each instruction is
4630 Dump the RTL in the assembler output as a comment before each instruction.
4631 Also turns on @option{-dp} annotation.
4635 For each of the other indicated dump files (either with @option{-d} or
4636 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4637 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4641 Just generate RTL for a function instead of compiling it. Usually used
4642 with @samp{r} (@option{-fdump-rtl-expand}).
4646 Dump debugging information during parsing, to standard error.
4650 @opindex fdump-noaddr
4651 When doing debugging dumps (see @option{-d} option above), suppress
4652 address output. This makes it more feasible to use diff on debugging
4653 dumps for compiler invocations with different compiler binaries and/or
4654 different text / bss / data / heap / stack / dso start locations.
4656 @item -fdump-unnumbered
4657 @opindex fdump-unnumbered
4658 When doing debugging dumps (see @option{-d} option above), suppress instruction
4659 numbers and address output. This makes it more feasible to
4660 use diff on debugging dumps for compiler invocations with different
4661 options, in particular with and without @option{-g}.
4663 @item -fdump-translation-unit @r{(C++ only)}
4664 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4665 @opindex fdump-translation-unit
4666 Dump a representation of the tree structure for the entire translation
4667 unit to a file. The file name is made by appending @file{.tu} to the
4668 source file name. If the @samp{-@var{options}} form is used, @var{options}
4669 controls the details of the dump as described for the
4670 @option{-fdump-tree} options.
4672 @item -fdump-class-hierarchy @r{(C++ only)}
4673 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4674 @opindex fdump-class-hierarchy
4675 Dump a representation of each class's hierarchy and virtual function
4676 table layout to a file. The file name is made by appending @file{.class}
4677 to the source file name. If the @samp{-@var{options}} form is used,
4678 @var{options} controls the details of the dump as described for the
4679 @option{-fdump-tree} options.
4681 @item -fdump-ipa-@var{switch}
4683 Control the dumping at various stages of inter-procedural analysis
4684 language tree to a file. The file name is generated by appending a switch
4685 specific suffix to the source file name. The following dumps are possible:
4689 Enables all inter-procedural analysis dumps.
4692 Dumps information about call-graph optimization, unused function removal,
4693 and inlining decisions.
4696 Dump after function inlining.
4700 @item -fdump-tree-@var{switch}
4701 @itemx -fdump-tree-@var{switch}-@var{options}
4703 Control the dumping at various stages of processing the intermediate
4704 language tree to a file. The file name is generated by appending a switch
4705 specific suffix to the source file name. If the @samp{-@var{options}}
4706 form is used, @var{options} is a list of @samp{-} separated options that
4707 control the details of the dump. Not all options are applicable to all
4708 dumps, those which are not meaningful will be ignored. The following
4709 options are available
4713 Print the address of each node. Usually this is not meaningful as it
4714 changes according to the environment and source file. Its primary use
4715 is for tying up a dump file with a debug environment.
4717 Inhibit dumping of members of a scope or body of a function merely
4718 because that scope has been reached. Only dump such items when they
4719 are directly reachable by some other path. When dumping pretty-printed
4720 trees, this option inhibits dumping the bodies of control structures.
4722 Print a raw representation of the tree. By default, trees are
4723 pretty-printed into a C-like representation.
4725 Enable more detailed dumps (not honored by every dump option).
4727 Enable dumping various statistics about the pass (not honored by every dump
4730 Enable showing basic block boundaries (disabled in raw dumps).
4732 Enable showing virtual operands for every statement.
4734 Enable showing line numbers for statements.
4736 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4738 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4741 The following tree dumps are possible:
4745 Dump before any tree based optimization, to @file{@var{file}.original}.
4748 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4751 @opindex fdump-tree-gimple
4752 Dump each function before and after the gimplification pass to a file. The
4753 file name is made by appending @file{.gimple} to the source file name.
4756 @opindex fdump-tree-cfg
4757 Dump the control flow graph of each function to a file. The file name is
4758 made by appending @file{.cfg} to the source file name.
4761 @opindex fdump-tree-vcg
4762 Dump the control flow graph of each function to a file in VCG format. The
4763 file name is made by appending @file{.vcg} to the source file name. Note
4764 that if the file contains more than one function, the generated file cannot
4765 be used directly by VCG@. You will need to cut and paste each function's
4766 graph into its own separate file first.
4769 @opindex fdump-tree-ch
4770 Dump each function after copying loop headers. The file name is made by
4771 appending @file{.ch} to the source file name.
4774 @opindex fdump-tree-ssa
4775 Dump SSA related information to a file. The file name is made by appending
4776 @file{.ssa} to the source file name.
4779 @opindex fdump-tree-salias
4780 Dump structure aliasing variable information to a file. This file name
4781 is made by appending @file{.salias} to the source file name.
4784 @opindex fdump-tree-alias
4785 Dump aliasing information for each function. The file name is made by
4786 appending @file{.alias} to the source file name.
4789 @opindex fdump-tree-ccp
4790 Dump each function after CCP@. The file name is made by appending
4791 @file{.ccp} to the source file name.
4794 @opindex fdump-tree-storeccp
4795 Dump each function after STORE-CCP@. The file name is made by appending
4796 @file{.storeccp} to the source file name.
4799 @opindex fdump-tree-pre
4800 Dump trees after partial redundancy elimination. The file name is made
4801 by appending @file{.pre} to the source file name.
4804 @opindex fdump-tree-fre
4805 Dump trees after full redundancy elimination. The file name is made
4806 by appending @file{.fre} to the source file name.
4809 @opindex fdump-tree-copyprop
4810 Dump trees after copy propagation. The file name is made
4811 by appending @file{.copyprop} to the source file name.
4813 @item store_copyprop
4814 @opindex fdump-tree-store_copyprop
4815 Dump trees after store copy-propagation. The file name is made
4816 by appending @file{.store_copyprop} to the source file name.
4819 @opindex fdump-tree-dce
4820 Dump each function after dead code elimination. The file name is made by
4821 appending @file{.dce} to the source file name.
4824 @opindex fdump-tree-mudflap
4825 Dump each function after adding mudflap instrumentation. The file name is
4826 made by appending @file{.mudflap} to the source file name.
4829 @opindex fdump-tree-sra
4830 Dump each function after performing scalar replacement of aggregates. The
4831 file name is made by appending @file{.sra} to the source file name.
4834 @opindex fdump-tree-sink
4835 Dump each function after performing code sinking. The file name is made
4836 by appending @file{.sink} to the source file name.
4839 @opindex fdump-tree-dom
4840 Dump each function after applying dominator tree optimizations. The file
4841 name is made by appending @file{.dom} to the source file name.
4844 @opindex fdump-tree-dse
4845 Dump each function after applying dead store elimination. The file
4846 name is made by appending @file{.dse} to the source file name.
4849 @opindex fdump-tree-phiopt
4850 Dump each function after optimizing PHI nodes into straightline code. The file
4851 name is made by appending @file{.phiopt} to the source file name.
4854 @opindex fdump-tree-forwprop
4855 Dump each function after forward propagating single use variables. The file
4856 name is made by appending @file{.forwprop} to the source file name.
4859 @opindex fdump-tree-copyrename
4860 Dump each function after applying the copy rename optimization. The file
4861 name is made by appending @file{.copyrename} to the source file name.
4864 @opindex fdump-tree-nrv
4865 Dump each function after applying the named return value optimization on
4866 generic trees. The file name is made by appending @file{.nrv} to the source
4870 @opindex fdump-tree-vect
4871 Dump each function after applying vectorization of loops. The file name is
4872 made by appending @file{.vect} to the source file name.
4875 @opindex fdump-tree-vrp
4876 Dump each function after Value Range Propagation (VRP). The file name
4877 is made by appending @file{.vrp} to the source file name.
4880 @opindex fdump-tree-all
4881 Enable all the available tree dumps with the flags provided in this option.
4884 @item -ftree-vectorizer-verbose=@var{n}
4885 @opindex ftree-vectorizer-verbose
4886 This option controls the amount of debugging output the vectorizer prints.
4887 This information is written to standard error, unless
4888 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4889 in which case it is output to the usual dump listing file, @file{.vect}.
4890 For @var{n}=0 no diagnostic information is reported.
4891 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4892 and the total number of loops that got vectorized.
4893 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4894 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
4895 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4896 level that @option{-fdump-tree-vect-stats} uses.
4897 Higher verbosity levels mean either more information dumped for each
4898 reported loop, or same amount of information reported for more loops:
4899 If @var{n}=3, alignment related information is added to the reports.
4900 If @var{n}=4, data-references related information (e.g.@: memory dependences,
4901 memory access-patterns) is added to the reports.
4902 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4903 that did not pass the first analysis phase (i.e., may not be countable, or
4904 may have complicated control-flow).
4905 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4906 For @var{n}=7, all the information the vectorizer generates during its
4907 analysis and transformation is reported. This is the same verbosity level
4908 that @option{-fdump-tree-vect-details} uses.
4910 @item -frandom-seed=@var{string}
4911 @opindex frandom-string
4912 This option provides a seed that GCC uses when it would otherwise use
4913 random numbers. It is used to generate certain symbol names
4914 that have to be different in every compiled file. It is also used to
4915 place unique stamps in coverage data files and the object files that
4916 produce them. You can use the @option{-frandom-seed} option to produce
4917 reproducibly identical object files.
4919 The @var{string} should be different for every file you compile.
4921 @item -fsched-verbose=@var{n}
4922 @opindex fsched-verbose
4923 On targets that use instruction scheduling, this option controls the
4924 amount of debugging output the scheduler prints. This information is
4925 written to standard error, unless @option{-dS} or @option{-dR} is
4926 specified, in which case it is output to the usual dump
4927 listing file, @file{.sched} or @file{.sched2} respectively. However
4928 for @var{n} greater than nine, the output is always printed to standard
4931 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4932 same information as @option{-dRS}. For @var{n} greater than one, it
4933 also output basic block probabilities, detailed ready list information
4934 and unit/insn info. For @var{n} greater than two, it includes RTL
4935 at abort point, control-flow and regions info. And for @var{n} over
4936 four, @option{-fsched-verbose} also includes dependence info.
4940 Store the usual ``temporary'' intermediate files permanently; place them
4941 in the current directory and name them based on the source file. Thus,
4942 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4943 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4944 preprocessed @file{foo.i} output file even though the compiler now
4945 normally uses an integrated preprocessor.
4947 When used in combination with the @option{-x} command line option,
4948 @option{-save-temps} is sensible enough to avoid over writing an
4949 input source file with the same extension as an intermediate file.
4950 The corresponding intermediate file may be obtained by renaming the
4951 source file before using @option{-save-temps}.
4955 Report the CPU time taken by each subprocess in the compilation
4956 sequence. For C source files, this is the compiler proper and assembler
4957 (plus the linker if linking is done). The output looks like this:
4964 The first number on each line is the ``user time'', that is time spent
4965 executing the program itself. The second number is ``system time'',
4966 time spent executing operating system routines on behalf of the program.
4967 Both numbers are in seconds.
4969 @item -fvar-tracking
4970 @opindex fvar-tracking
4971 Run variable tracking pass. It computes where variables are stored at each
4972 position in code. Better debugging information is then generated
4973 (if the debugging information format supports this information).
4975 It is enabled by default when compiling with optimization (@option{-Os},
4976 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
4977 the debug info format supports it.
4979 @item -print-file-name=@var{library}
4980 @opindex print-file-name
4981 Print the full absolute name of the library file @var{library} that
4982 would be used when linking---and don't do anything else. With this
4983 option, GCC does not compile or link anything; it just prints the
4986 @item -print-multi-directory
4987 @opindex print-multi-directory
4988 Print the directory name corresponding to the multilib selected by any
4989 other switches present in the command line. This directory is supposed
4990 to exist in @env{GCC_EXEC_PREFIX}.
4992 @item -print-multi-lib
4993 @opindex print-multi-lib
4994 Print the mapping from multilib directory names to compiler switches
4995 that enable them. The directory name is separated from the switches by
4996 @samp{;}, and each switch starts with an @samp{@@} instead of the
4997 @samp{-}, without spaces between multiple switches. This is supposed to
4998 ease shell-processing.
5000 @item -print-prog-name=@var{program}
5001 @opindex print-prog-name
5002 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5004 @item -print-libgcc-file-name
5005 @opindex print-libgcc-file-name
5006 Same as @option{-print-file-name=libgcc.a}.
5008 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5009 but you do want to link with @file{libgcc.a}. You can do
5012 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5015 @item -print-search-dirs
5016 @opindex print-search-dirs
5017 Print the name of the configured installation directory and a list of
5018 program and library directories @command{gcc} will search---and don't do anything else.
5020 This is useful when @command{gcc} prints the error message
5021 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5022 To resolve this you either need to put @file{cpp0} and the other compiler
5023 components where @command{gcc} expects to find them, or you can set the environment
5024 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5025 Don't forget the trailing @samp{/}.
5026 @xref{Environment Variables}.
5028 @item -print-sysroot-headers-suffix
5029 @opindex print-sysroot-headers-suffix
5030 Print the suffix added to the target sysroot when searching for
5031 headers, or give an error if the compiler is not configured with such
5032 a suffix---and don't do anything else.
5035 @opindex dumpmachine
5036 Print the compiler's target machine (for example,
5037 @samp{i686-pc-linux-gnu})---and don't do anything else.
5040 @opindex dumpversion
5041 Print the compiler version (for example, @samp{3.0})---and don't do
5046 Print the compiler's built-in specs---and don't do anything else. (This
5047 is used when GCC itself is being built.) @xref{Spec Files}.
5049 @item -feliminate-unused-debug-types
5050 @opindex feliminate-unused-debug-types
5051 Normally, when producing DWARF2 output, GCC will emit debugging
5052 information for all types declared in a compilation
5053 unit, regardless of whether or not they are actually used
5054 in that compilation unit. Sometimes this is useful, such as
5055 if, in the debugger, you want to cast a value to a type that is
5056 not actually used in your program (but is declared). More often,
5057 however, this results in a significant amount of wasted space.
5058 With this option, GCC will avoid producing debug symbol output
5059 for types that are nowhere used in the source file being compiled.
5062 @node Optimize Options
5063 @section Options That Control Optimization
5064 @cindex optimize options
5065 @cindex options, optimization
5067 These options control various sorts of optimizations.
5069 Without any optimization option, the compiler's goal is to reduce the
5070 cost of compilation and to make debugging produce the expected
5071 results. Statements are independent: if you stop the program with a
5072 breakpoint between statements, you can then assign a new value to any
5073 variable or change the program counter to any other statement in the
5074 function and get exactly the results you would expect from the source
5077 Turning on optimization flags makes the compiler attempt to improve
5078 the performance and/or code size at the expense of compilation time
5079 and possibly the ability to debug the program.
5081 The compiler performs optimization based on the knowledge it has of
5082 the program. Optimization levels @option{-O} and above, in
5083 particular, enable @emph{unit-at-a-time} mode, which allows the
5084 compiler to consider information gained from later functions in
5085 the file when compiling a function. Compiling multiple files at
5086 once to a single output file in @emph{unit-at-a-time} mode allows
5087 the compiler to use information gained from all of the files when
5088 compiling each of them.
5090 Not all optimizations are controlled directly by a flag. Only
5091 optimizations that have a flag are listed.
5098 Optimize. Optimizing compilation takes somewhat more time, and a lot
5099 more memory for a large function.
5101 With @option{-O}, the compiler tries to reduce code size and execution
5102 time, without performing any optimizations that take a great deal of
5105 @option{-O} turns on the following optimization flags:
5108 -fcprop-registers @gol
5111 -fdelayed-branch @gol
5113 -fguess-branch-probability @gol
5114 -fif-conversion2 @gol
5115 -fif-conversion @gol
5116 -finline-small-functions @gol
5117 -fipa-pure-const @gol
5118 -fipa-reference @gol
5120 -fsplit-wide-types @gol
5123 -ftree-copyrename @gol
5125 -ftree-dominator-opts @gol
5132 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5133 where doing so does not interfere with debugging.
5137 Optimize even more. GCC performs nearly all supported optimizations
5138 that do not involve a space-speed tradeoff. The compiler does not
5139 perform loop unrolling or function inlining when you specify @option{-O2}.
5140 As compared to @option{-O}, this option increases both compilation time
5141 and the performance of the generated code.
5143 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5144 also turns on the following optimization flags:
5145 @gccoptlist{-fthread-jumps @gol
5146 -falign-functions -falign-jumps @gol
5147 -falign-loops -falign-labels @gol
5150 -fcse-follow-jumps -fcse-skip-blocks @gol
5151 -fdelete-null-pointer-checks @gol
5152 -fexpensive-optimizations @gol
5153 -fgcse -fgcse-lm @gol
5154 -foptimize-sibling-calls @gol
5157 -freorder-blocks -freorder-functions @gol
5158 -frerun-cse-after-loop @gol
5159 -fsched-interblock -fsched-spec @gol
5160 -fschedule-insns -fschedule-insns2 @gol
5161 -fstrict-aliasing -fstrict-overflow @gol
5165 Please note the warning under @option{-fgcse} about
5166 invoking @option{-O2} on programs that use computed gotos.
5170 Optimize yet more. @option{-O3} turns on all optimizations specified by
5171 @option{-O2} and also turns on the @option{-finline-functions},
5172 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5173 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5177 Reduce compilation time and make debugging produce the expected
5178 results. This is the default.
5182 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5183 do not typically increase code size. It also performs further
5184 optimizations designed to reduce code size.
5186 @option{-Os} disables the following optimization flags:
5187 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5188 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5189 -fprefetch-loop-arrays -ftree-vect-loop-version}
5191 If you use multiple @option{-O} options, with or without level numbers,
5192 the last such option is the one that is effective.
5195 Options of the form @option{-f@var{flag}} specify machine-independent
5196 flags. Most flags have both positive and negative forms; the negative
5197 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5198 below, only one of the forms is listed---the one you typically will
5199 use. You can figure out the other form by either removing @samp{no-}
5202 The following options control specific optimizations. They are either
5203 activated by @option{-O} options or are related to ones that are. You
5204 can use the following flags in the rare cases when ``fine-tuning'' of
5205 optimizations to be performed is desired.
5208 @item -fno-default-inline
5209 @opindex fno-default-inline
5210 Do not make member functions inline by default merely because they are
5211 defined inside the class scope (C++ only). Otherwise, when you specify
5212 @w{@option{-O}}, member functions defined inside class scope are compiled
5213 inline by default; i.e., you don't need to add @samp{inline} in front of
5214 the member function name.
5216 @item -fno-defer-pop
5217 @opindex fno-defer-pop
5218 Always pop the arguments to each function call as soon as that function
5219 returns. For machines which must pop arguments after a function call,
5220 the compiler normally lets arguments accumulate on the stack for several
5221 function calls and pops them all at once.
5223 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5225 @item -fforward-propagate
5226 @opindex fforward-propagate
5227 Perform a forward propagation pass on RTL@. The pass tries to combine two
5228 instructions and checks if the result can be simplified. If loop unrolling
5229 is active, two passes are performed and the second is scheduled after
5232 This option is enabled by default at optimization levels @option{-O2},
5233 @option{-O3}, @option{-Os}.
5235 @item -fomit-frame-pointer
5236 @opindex fomit-frame-pointer
5237 Don't keep the frame pointer in a register for functions that
5238 don't need one. This avoids the instructions to save, set up and
5239 restore frame pointers; it also makes an extra register available
5240 in many functions. @strong{It also makes debugging impossible on
5243 On some machines, such as the VAX, this flag has no effect, because
5244 the standard calling sequence automatically handles the frame pointer
5245 and nothing is saved by pretending it doesn't exist. The
5246 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5247 whether a target machine supports this flag. @xref{Registers,,Register
5248 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5250 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5252 @item -foptimize-sibling-calls
5253 @opindex foptimize-sibling-calls
5254 Optimize sibling and tail recursive calls.
5256 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5260 Don't pay attention to the @code{inline} keyword. Normally this option
5261 is used to keep the compiler from expanding any functions inline.
5262 Note that if you are not optimizing, no functions can be expanded inline.
5264 @item -finline-small-functions
5265 @opindex finline-small-functions
5266 Integrate functions into their callers when their body is smaller than expected
5267 function call code (so overall size of program gets smaller). The compiler
5268 heuristically decides which functions are simple enough to be worth integrating
5271 Enabled at level @option{-O2}.
5273 @item -finline-functions
5274 @opindex finline-functions
5275 Integrate all simple functions into their callers. The compiler
5276 heuristically decides which functions are simple enough to be worth
5277 integrating in this way.
5279 If all calls to a given function are integrated, and the function is
5280 declared @code{static}, then the function is normally not output as
5281 assembler code in its own right.
5283 Enabled at level @option{-O3}.
5285 @item -finline-functions-called-once
5286 @opindex finline-functions-called-once
5287 Consider all @code{static} functions called once for inlining into their
5288 caller even if they are not marked @code{inline}. If a call to a given
5289 function is integrated, then the function is not output as assembler code
5292 Enabled if @option{-funit-at-a-time} is enabled.
5294 @item -fearly-inlining
5295 @opindex fearly-inlining
5296 Inline functions marked by @code{always_inline} and functions whose body seems
5297 smaller than the function call overhead early before doing
5298 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5299 makes profiling significantly cheaper and usually inlining faster on programs
5300 having large chains of nested wrapper functions.
5304 @item -finline-limit=@var{n}
5305 @opindex finline-limit
5306 By default, GCC limits the size of functions that can be inlined. This flag
5307 allows coarse control of this limit. @var{n} is the size of functions that
5308 can be inlined in number of pseudo instructions.
5310 Inlining is actually controlled by a number of parameters, which may be
5311 specified individually by using @option{--param @var{name}=@var{value}}.
5312 The @option{-finline-limit=@var{n}} option sets some of these parameters
5316 @item max-inline-insns-single
5317 is set to @var{n}/2.
5318 @item max-inline-insns-auto
5319 is set to @var{n}/2.
5322 See below for a documentation of the individual
5323 parameters controlling inlining and for the defaults of these parameters.
5325 @emph{Note:} there may be no value to @option{-finline-limit} that results
5326 in default behavior.
5328 @emph{Note:} pseudo instruction represents, in this particular context, an
5329 abstract measurement of function's size. In no way does it represent a count
5330 of assembly instructions and as such its exact meaning might change from one
5331 release to an another.
5333 @item -fkeep-inline-functions
5334 @opindex fkeep-inline-functions
5335 In C, emit @code{static} functions that are declared @code{inline}
5336 into the object file, even if the function has been inlined into all
5337 of its callers. This switch does not affect functions using the
5338 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5339 inline functions into the object file.
5341 @item -fkeep-static-consts
5342 @opindex fkeep-static-consts
5343 Emit variables declared @code{static const} when optimization isn't turned
5344 on, even if the variables aren't referenced.
5346 GCC enables this option by default. If you want to force the compiler to
5347 check if the variable was referenced, regardless of whether or not
5348 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5350 @item -fmerge-constants
5351 @opindex fmerge-constants
5352 Attempt to merge identical constants (string constants and floating point
5353 constants) across compilation units.
5355 This option is the default for optimized compilation if the assembler and
5356 linker support it. Use @option{-fno-merge-constants} to inhibit this
5359 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5361 @item -fmerge-all-constants
5362 @opindex fmerge-all-constants
5363 Attempt to merge identical constants and identical variables.
5365 This option implies @option{-fmerge-constants}. In addition to
5366 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5367 arrays or initialized constant variables with integral or floating point
5368 types. Languages like C or C++ require each non-automatic variable to
5369 have distinct location, so using this option will result in non-conforming
5372 @item -fmodulo-sched
5373 @opindex fmodulo-sched
5374 Perform swing modulo scheduling immediately before the first scheduling
5375 pass. This pass looks at innermost loops and reorders their
5376 instructions by overlapping different iterations.
5378 @item -fmodulo-sched-allow-regmoves
5379 @opindex fmodulo-sched-allow-regmoves
5380 Perform more aggressive SMS based modulo scheduling with register moves
5381 allowed. By setting this flag certain anti-dependences edges will be
5382 deleted which will trigger the generation of reg-moves based on the
5383 life-range analysis. This option is effective only with
5384 @option{-fmodulo-sched} enabled.
5386 @item -fno-branch-count-reg
5387 @opindex fno-branch-count-reg
5388 Do not use ``decrement and branch'' instructions on a count register,
5389 but instead generate a sequence of instructions that decrement a
5390 register, compare it against zero, then branch based upon the result.
5391 This option is only meaningful on architectures that support such
5392 instructions, which include x86, PowerPC, IA-64 and S/390.
5394 The default is @option{-fbranch-count-reg}.
5396 @item -fno-function-cse
5397 @opindex fno-function-cse
5398 Do not put function addresses in registers; make each instruction that
5399 calls a constant function contain the function's address explicitly.
5401 This option results in less efficient code, but some strange hacks
5402 that alter the assembler output may be confused by the optimizations
5403 performed when this option is not used.
5405 The default is @option{-ffunction-cse}
5407 @item -fno-zero-initialized-in-bss
5408 @opindex fno-zero-initialized-in-bss
5409 If the target supports a BSS section, GCC by default puts variables that
5410 are initialized to zero into BSS@. This can save space in the resulting
5413 This option turns off this behavior because some programs explicitly
5414 rely on variables going to the data section. E.g., so that the
5415 resulting executable can find the beginning of that section and/or make
5416 assumptions based on that.
5418 The default is @option{-fzero-initialized-in-bss}.
5420 @item -fmudflap -fmudflapth -fmudflapir
5424 @cindex bounds checking
5426 For front-ends that support it (C and C++), instrument all risky
5427 pointer/array dereferencing operations, some standard library
5428 string/heap functions, and some other associated constructs with
5429 range/validity tests. Modules so instrumented should be immune to
5430 buffer overflows, invalid heap use, and some other classes of C/C++
5431 programming errors. The instrumentation relies on a separate runtime
5432 library (@file{libmudflap}), which will be linked into a program if
5433 @option{-fmudflap} is given at link time. Run-time behavior of the
5434 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5435 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5438 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5439 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5440 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5441 instrumentation should ignore pointer reads. This produces less
5442 instrumentation (and therefore faster execution) and still provides
5443 some protection against outright memory corrupting writes, but allows
5444 erroneously read data to propagate within a program.
5446 @item -fthread-jumps
5447 @opindex fthread-jumps
5448 Perform optimizations where we check to see if a jump branches to a
5449 location where another comparison subsumed by the first is found. If
5450 so, the first branch is redirected to either the destination of the
5451 second branch or a point immediately following it, depending on whether
5452 the condition is known to be true or false.
5454 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5456 @item -fsplit-wide-types
5457 @opindex fsplit-wide-types
5458 When using a type that occupies multiple registers, such as @code{long
5459 long} on a 32-bit system, split the registers apart and allocate them
5460 independently. This normally generates better code for those types,
5461 but may make debugging more difficult.
5463 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5466 @item -fcse-follow-jumps
5467 @opindex fcse-follow-jumps
5468 In common subexpression elimination (CSE), scan through jump instructions
5469 when the target of the jump is not reached by any other path. For
5470 example, when CSE encounters an @code{if} statement with an
5471 @code{else} clause, CSE will follow the jump when the condition
5474 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5476 @item -fcse-skip-blocks
5477 @opindex fcse-skip-blocks
5478 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5479 follow jumps which conditionally skip over blocks. When CSE
5480 encounters a simple @code{if} statement with no else clause,
5481 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5482 body of the @code{if}.
5484 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5486 @item -frerun-cse-after-loop
5487 @opindex frerun-cse-after-loop
5488 Re-run common subexpression elimination after loop optimizations has been
5491 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5495 Perform a global common subexpression elimination pass.
5496 This pass also performs global constant and copy propagation.
5498 @emph{Note:} When compiling a program using computed gotos, a GCC
5499 extension, you may get better runtime performance if you disable
5500 the global common subexpression elimination pass by adding
5501 @option{-fno-gcse} to the command line.
5503 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5507 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5508 attempt to move loads which are only killed by stores into themselves. This
5509 allows a loop containing a load/store sequence to be changed to a load outside
5510 the loop, and a copy/store within the loop.
5512 Enabled by default when gcse is enabled.
5516 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5517 global common subexpression elimination. This pass will attempt to move
5518 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5519 loops containing a load/store sequence can be changed to a load before
5520 the loop and a store after the loop.
5522 Not enabled at any optimization level.
5526 When @option{-fgcse-las} is enabled, the global common subexpression
5527 elimination pass eliminates redundant loads that come after stores to the
5528 same memory location (both partial and full redundancies).
5530 Not enabled at any optimization level.
5532 @item -fgcse-after-reload
5533 @opindex fgcse-after-reload
5534 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5535 pass is performed after reload. The purpose of this pass is to cleanup
5538 @item -funsafe-loop-optimizations
5539 @opindex funsafe-loop-optimizations
5540 If given, the loop optimizer will assume that loop indices do not
5541 overflow, and that the loops with nontrivial exit condition are not
5542 infinite. This enables a wider range of loop optimizations even if
5543 the loop optimizer itself cannot prove that these assumptions are valid.
5544 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5545 if it finds this kind of loop.
5547 @item -fcrossjumping
5548 @opindex fcrossjumping
5549 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5550 resulting code may or may not perform better than without cross-jumping.
5552 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5554 @item -fauto-inc-dec
5555 @opindex fauto-inc-dec
5556 Combine increments or decrements of addresses with memory accesses.
5557 This pass is always skipped on architectures that do not have
5558 instructions to support this. Enabled by default at @option{-O} and
5559 higher on architectures that support this.
5563 Perform dead code elimination (DCE) on RTL@.
5564 Enabled by default at @option{-O} and higher.
5568 Perform dead store elimination (DSE) on RTL@.
5569 Enabled by default at @option{-O} and higher.
5571 @item -fif-conversion
5572 @opindex fif-conversion
5573 Attempt to transform conditional jumps into branch-less equivalents. This
5574 include use of conditional moves, min, max, set flags and abs instructions, and
5575 some tricks doable by standard arithmetics. The use of conditional execution
5576 on chips where it is available is controlled by @code{if-conversion2}.
5578 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5580 @item -fif-conversion2
5581 @opindex fif-conversion2
5582 Use conditional execution (where available) to transform conditional jumps into
5583 branch-less equivalents.
5585 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5587 @item -fdelete-null-pointer-checks
5588 @opindex fdelete-null-pointer-checks
5589 Use global dataflow analysis to identify and eliminate useless checks
5590 for null pointers. The compiler assumes that dereferencing a null
5591 pointer would have halted the program. If a pointer is checked after
5592 it has already been dereferenced, it cannot be null.
5594 In some environments, this assumption is not true, and programs can
5595 safely dereference null pointers. Use
5596 @option{-fno-delete-null-pointer-checks} to disable this optimization
5597 for programs which depend on that behavior.
5599 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5601 @item -fexpensive-optimizations
5602 @opindex fexpensive-optimizations
5603 Perform a number of minor optimizations that are relatively expensive.
5605 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5607 @item -foptimize-register-move
5609 @opindex foptimize-register-move
5611 Attempt to reassign register numbers in move instructions and as
5612 operands of other simple instructions in order to maximize the amount of
5613 register tying. This is especially helpful on machines with two-operand
5616 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5619 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5621 @item -fdelayed-branch
5622 @opindex fdelayed-branch
5623 If supported for the target machine, attempt to reorder instructions
5624 to exploit instruction slots available after delayed branch
5627 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5629 @item -fschedule-insns
5630 @opindex fschedule-insns
5631 If supported for the target machine, attempt to reorder instructions to
5632 eliminate execution stalls due to required data being unavailable. This
5633 helps machines that have slow floating point or memory load instructions
5634 by allowing other instructions to be issued until the result of the load
5635 or floating point instruction is required.
5637 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5639 @item -fschedule-insns2
5640 @opindex fschedule-insns2
5641 Similar to @option{-fschedule-insns}, but requests an additional pass of
5642 instruction scheduling after register allocation has been done. This is
5643 especially useful on machines with a relatively small number of
5644 registers and where memory load instructions take more than one cycle.
5646 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5648 @item -fno-sched-interblock
5649 @opindex fno-sched-interblock
5650 Don't schedule instructions across basic blocks. This is normally
5651 enabled by default when scheduling before register allocation, i.e.@:
5652 with @option{-fschedule-insns} or at @option{-O2} or higher.
5654 @item -fno-sched-spec
5655 @opindex fno-sched-spec
5656 Don't allow speculative motion of non-load instructions. This is normally
5657 enabled by default when scheduling before register allocation, i.e.@:
5658 with @option{-fschedule-insns} or at @option{-O2} or higher.
5660 @item -fsched-spec-load
5661 @opindex fsched-spec-load
5662 Allow speculative motion of some load instructions. This only makes
5663 sense when scheduling before register allocation, i.e.@: with
5664 @option{-fschedule-insns} or at @option{-O2} or higher.
5666 @item -fsched-spec-load-dangerous
5667 @opindex fsched-spec-load-dangerous
5668 Allow speculative motion of more load instructions. This only makes
5669 sense when scheduling before register allocation, i.e.@: with
5670 @option{-fschedule-insns} or at @option{-O2} or higher.
5672 @item -fsched-stalled-insns
5673 @itemx -fsched-stalled-insns=@var{n}
5674 @opindex fsched-stalled-insns
5675 Define how many insns (if any) can be moved prematurely from the queue
5676 of stalled insns into the ready list, during the second scheduling pass.
5677 @option{-fno-sched-stalled-insns} means that no insns will be moved
5678 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
5679 on how many queued insns can be moved prematurely.
5680 @option{-fsched-stalled-insns} without a value is equivalent to
5681 @option{-fsched-stalled-insns=1}.
5683 @item -fsched-stalled-insns-dep
5684 @itemx -fsched-stalled-insns-dep=@var{n}
5685 @opindex fsched-stalled-insns-dep
5686 Define how many insn groups (cycles) will be examined for a dependency
5687 on a stalled insn that is candidate for premature removal from the queue
5688 of stalled insns. This has an effect only during the second scheduling pass,
5689 and only if @option{-fsched-stalled-insns} is used.
5690 @option{-fno-sched-stalled-insns-dep} is equivalent to
5691 @option{-fsched-stalled-insns-dep=0}.
5692 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5693 @option{-fsched-stalled-insns-dep=1}.
5695 @item -fsched2-use-superblocks
5696 @opindex fsched2-use-superblocks
5697 When scheduling after register allocation, do use superblock scheduling
5698 algorithm. Superblock scheduling allows motion across basic block boundaries
5699 resulting on faster schedules. This option is experimental, as not all machine
5700 descriptions used by GCC model the CPU closely enough to avoid unreliable
5701 results from the algorithm.
5703 This only makes sense when scheduling after register allocation, i.e.@: with
5704 @option{-fschedule-insns2} or at @option{-O2} or higher.
5706 @item -fsched2-use-traces
5707 @opindex fsched2-use-traces
5708 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5709 allocation and additionally perform code duplication in order to increase the
5710 size of superblocks using tracer pass. See @option{-ftracer} for details on
5713 This mode should produce faster but significantly longer programs. Also
5714 without @option{-fbranch-probabilities} the traces constructed may not
5715 match the reality and hurt the performance. This only makes
5716 sense when scheduling after register allocation, i.e.@: with
5717 @option{-fschedule-insns2} or at @option{-O2} or higher.
5721 Eliminate redundant sign extension instructions and move the non-redundant
5722 ones to optimal placement using lazy code motion (LCM).
5724 @item -freschedule-modulo-scheduled-loops
5725 @opindex freschedule-modulo-scheduled-loops
5726 The modulo scheduling comes before the traditional scheduling, if a loop
5727 was modulo scheduled we may want to prevent the later scheduling passes
5728 from changing its schedule, we use this option to control that.
5730 @item -fcaller-saves
5731 @opindex fcaller-saves
5732 Enable values to be allocated in registers that will be clobbered by
5733 function calls, by emitting extra instructions to save and restore the
5734 registers around such calls. Such allocation is done only when it
5735 seems to result in better code than would otherwise be produced.
5737 This option is always enabled by default on certain machines, usually
5738 those which have no call-preserved registers to use instead.
5740 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5742 @item -ftree-reassoc
5743 @opindex ftree-reassoc
5744 Perform reassociation on trees. This flag is enabled by default
5745 at @option{-O} and higher.
5749 Perform partial redundancy elimination (PRE) on trees. This flag is
5750 enabled by default at @option{-O2} and @option{-O3}.
5754 Perform full redundancy elimination (FRE) on trees. The difference
5755 between FRE and PRE is that FRE only considers expressions
5756 that are computed on all paths leading to the redundant computation.
5757 This analysis is faster than PRE, though it exposes fewer redundancies.
5758 This flag is enabled by default at @option{-O} and higher.
5760 @item -ftree-copy-prop
5761 @opindex ftree-copy-prop
5762 Perform copy propagation on trees. This pass eliminates unnecessary
5763 copy operations. This flag is enabled by default at @option{-O} and
5767 @opindex ftree-salias
5768 Perform structural alias analysis on trees. This flag
5769 is enabled by default at @option{-O} and higher.
5771 @item -fipa-pure-const
5772 @opindex fipa-pure-const
5773 Discover which functions are pure or constant.
5774 Enabled by default at @option{-O} and higher.
5776 @item -fipa-reference
5777 @opindex fipa-reference
5778 Discover which static variables do not escape cannot escape the
5780 Enabled by default at @option{-O} and higher.
5782 @item -fipa-struct-reorg
5783 @opindex fipa-struct-reorg
5784 Perform structure reorganization optimization, that change C-like structures
5785 layout in order to better utilize spatial locality. This transformation is
5786 affective for programs containing arrays of structures. Available in two
5787 compilation modes: profile-based (enabled with @option{-fprofile-generate})
5788 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
5789 to provide the safety of this transformation. It works only in whole program
5790 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
5791 enabled. Structures considered @samp{cold} by this transformation are not
5792 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
5794 With this flag, the program debug info reflects a new structure layout.
5798 Perform interprocedural pointer analysis.
5802 Perform interprocedural constant propagation.
5803 This optimization analyzes the program to determine when values passed
5804 to functions are constants and then optimizes accordingly.
5805 This optimization can substantially increase performance
5806 if the application has constants passed to functions, but
5807 because this optimization can create multiple copies of functions,
5808 it may significantly increase code size.
5810 @item -fipa-matrix-reorg
5811 @opindex fipa-matrix-reorg
5812 Perform matrix flattening and transposing.
5813 Matrix flattening tries to replace a m-dimensional matrix
5814 with its equivalent n-dimensional matrix, where n < m.
5815 This reduces the level of indirection needed for accessing the elements
5816 of the matrix. The second optimization is matrix transposing that
5817 attemps to change the order of the matrix's dimensions in order to
5818 improve cache locality.
5819 Both optimizations need fwhole-program flag.
5820 Transposing is enabled only if profiling information is avaliable.
5825 Perform forward store motion on trees. This flag is
5826 enabled by default at @option{-O} and higher.
5830 Perform sparse conditional constant propagation (CCP) on trees. This
5831 pass only operates on local scalar variables and is enabled by default
5832 at @option{-O} and higher.
5834 @item -ftree-store-ccp
5835 @opindex ftree-store-ccp
5836 Perform sparse conditional constant propagation (CCP) on trees. This
5837 pass operates on both local scalar variables and memory stores and
5838 loads (global variables, structures, arrays, etc). This flag is
5839 enabled by default at @option{-O2} and higher.
5843 Perform dead code elimination (DCE) on trees. This flag is enabled by
5844 default at @option{-O} and higher.
5846 @item -ftree-dominator-opts
5847 @opindex ftree-dominator-opts
5848 Perform a variety of simple scalar cleanups (constant/copy
5849 propagation, redundancy elimination, range propagation and expression
5850 simplification) based on a dominator tree traversal. This also
5851 performs jump threading (to reduce jumps to jumps). This flag is
5852 enabled by default at @option{-O} and higher.
5856 Perform dead store elimination (DSE) on trees. A dead store is a store into
5857 a memory location which will later be overwritten by another store without
5858 any intervening loads. In this case the earlier store can be deleted. This
5859 flag is enabled by default at @option{-O} and higher.
5863 Perform loop header copying on trees. This is beneficial since it increases
5864 effectiveness of code motion optimizations. It also saves one jump. This flag
5865 is enabled by default at @option{-O} and higher. It is not enabled
5866 for @option{-Os}, since it usually increases code size.
5868 @item -ftree-loop-optimize
5869 @opindex ftree-loop-optimize
5870 Perform loop optimizations on trees. This flag is enabled by default
5871 at @option{-O} and higher.
5873 @item -ftree-loop-linear
5874 @opindex ftree-loop-linear
5875 Perform linear loop transformations on tree. This flag can improve cache
5876 performance and allow further loop optimizations to take place.
5878 @item -fcheck-data-deps
5879 @opindex fcheck-data-deps
5880 Compare the results of several data dependence analyzers. This option
5881 is used for debugging the data dependence analyzers.
5883 @item -ftree-loop-distribution
5884 Perform loop distribution. This flag can improve cache performance on
5885 big loop bodies and allow further loop optimizations, like
5886 parallelization or vectorization, to take place. For example, the loop
5903 @item -ftree-loop-im
5904 @opindex ftree-loop-im
5905 Perform loop invariant motion on trees. This pass moves only invariants that
5906 would be hard to handle at RTL level (function calls, operations that expand to
5907 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5908 operands of conditions that are invariant out of the loop, so that we can use
5909 just trivial invariantness analysis in loop unswitching. The pass also includes
5912 @item -ftree-loop-ivcanon
5913 @opindex ftree-loop-ivcanon
5914 Create a canonical counter for number of iterations in the loop for that
5915 determining number of iterations requires complicated analysis. Later
5916 optimizations then may determine the number easily. Useful especially
5917 in connection with unrolling.
5921 Perform induction variable optimizations (strength reduction, induction
5922 variable merging and induction variable elimination) on trees.
5924 @item -ftree-parallelize-loops=n
5925 @opindex ftree-parallelize-loops
5926 Parallelize loops, i.e., split their iteration space to run in n threads.
5927 This is only possible for loops whose iterations are independent
5928 and can be arbitrarily reordered. The optimization is only
5929 profitable on multiprocessor machines, for loops that are CPU-intensive,
5930 rather than constrained e.g.@: by memory bandwidth. This option
5931 implies @option{-pthread}, and thus is only supported on targets
5932 that have support for @option{-pthread}.
5936 Perform scalar replacement of aggregates. This pass replaces structure
5937 references with scalars to prevent committing structures to memory too
5938 early. This flag is enabled by default at @option{-O} and higher.
5940 @item -ftree-copyrename
5941 @opindex ftree-copyrename
5942 Perform copy renaming on trees. This pass attempts to rename compiler
5943 temporaries to other variables at copy locations, usually resulting in
5944 variable names which more closely resemble the original variables. This flag
5945 is enabled by default at @option{-O} and higher.
5949 Perform temporary expression replacement during the SSA->normal phase. Single
5950 use/single def temporaries are replaced at their use location with their
5951 defining expression. This results in non-GIMPLE code, but gives the expanders
5952 much more complex trees to work on resulting in better RTL generation. This is
5953 enabled by default at @option{-O} and higher.
5955 @item -ftree-vectorize
5956 @opindex ftree-vectorize
5957 Perform loop vectorization on trees. This flag is enabled by default at
5960 @item -ftree-vect-loop-version
5961 @opindex ftree-vect-loop-version
5962 Perform loop versioning when doing loop vectorization on trees. When a loop
5963 appears to be vectorizable except that data alignment or data dependence cannot
5964 be determined at compile time then vectorized and non-vectorized versions of
5965 the loop are generated along with runtime checks for alignment or dependence
5966 to control which version is executed. This option is enabled by default
5967 except at level @option{-Os} where it is disabled.
5969 @item -fvect-cost-model
5970 @opindex fvect-cost-model
5971 Enable cost model for vectorization.
5975 Perform Value Range Propagation on trees. This is similar to the
5976 constant propagation pass, but instead of values, ranges of values are
5977 propagated. This allows the optimizers to remove unnecessary range
5978 checks like array bound checks and null pointer checks. This is
5979 enabled by default at @option{-O2} and higher. Null pointer check
5980 elimination is only done if @option{-fdelete-null-pointer-checks} is
5985 Perform tail duplication to enlarge superblock size. This transformation
5986 simplifies the control flow of the function allowing other optimizations to do
5989 @item -funroll-loops
5990 @opindex funroll-loops
5991 Unroll loops whose number of iterations can be determined at compile
5992 time or upon entry to the loop. @option{-funroll-loops} implies
5993 @option{-frerun-cse-after-loop}. This option makes code larger,
5994 and may or may not make it run faster.
5996 @item -funroll-all-loops
5997 @opindex funroll-all-loops
5998 Unroll all loops, even if their number of iterations is uncertain when
5999 the loop is entered. This usually makes programs run more slowly.
6000 @option{-funroll-all-loops} implies the same options as
6001 @option{-funroll-loops},
6003 @item -fsplit-ivs-in-unroller
6004 @opindex fsplit-ivs-in-unroller
6005 Enables expressing of values of induction variables in later iterations
6006 of the unrolled loop using the value in the first iteration. This breaks
6007 long dependency chains, thus improving efficiency of the scheduling passes.
6009 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6010 same effect. However in cases the loop body is more complicated than
6011 a single basic block, this is not reliable. It also does not work at all
6012 on some of the architectures due to restrictions in the CSE pass.
6014 This optimization is enabled by default.
6016 @item -fvariable-expansion-in-unroller
6017 @opindex fvariable-expansion-in-unroller
6018 With this option, the compiler will create multiple copies of some
6019 local variables when unrolling a loop which can result in superior code.
6021 @item -fpredictive-commoning
6022 @opindex fpredictive-commoning
6023 Perform predictive commoning optimization, i.e., reusing computations
6024 (especially memory loads and stores) performed in previous
6025 iterations of loops.
6027 This option is enabled at level @option{-O3}.
6029 @item -fprefetch-loop-arrays
6030 @opindex fprefetch-loop-arrays
6031 If supported by the target machine, generate instructions to prefetch
6032 memory to improve the performance of loops that access large arrays.
6034 This option may generate better or worse code; results are highly
6035 dependent on the structure of loops within the source code.
6037 Disabled at level @option{-Os}.
6040 @itemx -fno-peephole2
6041 @opindex fno-peephole
6042 @opindex fno-peephole2
6043 Disable any machine-specific peephole optimizations. The difference
6044 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6045 are implemented in the compiler; some targets use one, some use the
6046 other, a few use both.
6048 @option{-fpeephole} is enabled by default.
6049 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6051 @item -fno-guess-branch-probability
6052 @opindex fno-guess-branch-probability
6053 Do not guess branch probabilities using heuristics.
6055 GCC will use heuristics to guess branch probabilities if they are
6056 not provided by profiling feedback (@option{-fprofile-arcs}). These
6057 heuristics are based on the control flow graph. If some branch probabilities
6058 are specified by @samp{__builtin_expect}, then the heuristics will be
6059 used to guess branch probabilities for the rest of the control flow graph,
6060 taking the @samp{__builtin_expect} info into account. The interactions
6061 between the heuristics and @samp{__builtin_expect} can be complex, and in
6062 some cases, it may be useful to disable the heuristics so that the effects
6063 of @samp{__builtin_expect} are easier to understand.
6065 The default is @option{-fguess-branch-probability} at levels
6066 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6068 @item -freorder-blocks
6069 @opindex freorder-blocks
6070 Reorder basic blocks in the compiled function in order to reduce number of
6071 taken branches and improve code locality.
6073 Enabled at levels @option{-O2}, @option{-O3}.
6075 @item -freorder-blocks-and-partition
6076 @opindex freorder-blocks-and-partition
6077 In addition to reordering basic blocks in the compiled function, in order
6078 to reduce number of taken branches, partitions hot and cold basic blocks
6079 into separate sections of the assembly and .o files, to improve
6080 paging and cache locality performance.
6082 This optimization is automatically turned off in the presence of
6083 exception handling, for linkonce sections, for functions with a user-defined
6084 section attribute and on any architecture that does not support named
6087 @item -freorder-functions
6088 @opindex freorder-functions
6089 Reorder functions in the object file in order to
6090 improve code locality. This is implemented by using special
6091 subsections @code{.text.hot} for most frequently executed functions and
6092 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6093 the linker so object file format must support named sections and linker must
6094 place them in a reasonable way.
6096 Also profile feedback must be available in to make this option effective. See
6097 @option{-fprofile-arcs} for details.
6099 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6101 @item -fstrict-aliasing
6102 @opindex fstrict-aliasing
6103 Allows the compiler to assume the strictest aliasing rules applicable to
6104 the language being compiled. For C (and C++), this activates
6105 optimizations based on the type of expressions. In particular, an
6106 object of one type is assumed never to reside at the same address as an
6107 object of a different type, unless the types are almost the same. For
6108 example, an @code{unsigned int} can alias an @code{int}, but not a
6109 @code{void*} or a @code{double}. A character type may alias any other
6112 Pay special attention to code like this:
6125 The practice of reading from a different union member than the one most
6126 recently written to (called ``type-punning'') is common. Even with
6127 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6128 is accessed through the union type. So, the code above will work as
6129 expected. However, this code might not:
6140 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6142 @item -fstrict-overflow
6143 @opindex fstrict-overflow
6144 Allow the compiler to assume strict signed overflow rules, depending
6145 on the language being compiled. For C (and C++) this means that
6146 overflow when doing arithmetic with signed numbers is undefined, which
6147 means that the compiler may assume that it will not happen. This
6148 permits various optimizations. For example, the compiler will assume
6149 that an expression like @code{i + 10 > i} will always be true for
6150 signed @code{i}. This assumption is only valid if signed overflow is
6151 undefined, as the expression is false if @code{i + 10} overflows when
6152 using twos complement arithmetic. When this option is in effect any
6153 attempt to determine whether an operation on signed numbers will
6154 overflow must be written carefully to not actually involve overflow.
6156 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6157 that signed overflow is fully defined: it wraps. When
6158 @option{-fwrapv} is used, there is no difference between
6159 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
6160 @option{-fwrapv} certain types of overflow are permitted. For
6161 example, if the compiler gets an overflow when doing arithmetic on
6162 constants, the overflowed value can still be used with
6163 @option{-fwrapv}, but not otherwise.
6165 The @option{-fstrict-overflow} option is enabled at levels
6166 @option{-O2}, @option{-O3}, @option{-Os}.
6168 @item -falign-functions
6169 @itemx -falign-functions=@var{n}
6170 @opindex falign-functions
6171 Align the start of functions to the next power-of-two greater than
6172 @var{n}, skipping up to @var{n} bytes. For instance,
6173 @option{-falign-functions=32} aligns functions to the next 32-byte
6174 boundary, but @option{-falign-functions=24} would align to the next
6175 32-byte boundary only if this can be done by skipping 23 bytes or less.
6177 @option{-fno-align-functions} and @option{-falign-functions=1} are
6178 equivalent and mean that functions will not be aligned.
6180 Some assemblers only support this flag when @var{n} is a power of two;
6181 in that case, it is rounded up.
6183 If @var{n} is not specified or is zero, use a machine-dependent default.
6185 Enabled at levels @option{-O2}, @option{-O3}.
6187 @item -falign-labels
6188 @itemx -falign-labels=@var{n}
6189 @opindex falign-labels
6190 Align all branch targets to a power-of-two boundary, skipping up to
6191 @var{n} bytes like @option{-falign-functions}. This option can easily
6192 make code slower, because it must insert dummy operations for when the
6193 branch target is reached in the usual flow of the code.
6195 @option{-fno-align-labels} and @option{-falign-labels=1} are
6196 equivalent and mean that labels will not be aligned.
6198 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6199 are greater than this value, then their values are used instead.
6201 If @var{n} is not specified or is zero, use a machine-dependent default
6202 which is very likely to be @samp{1}, meaning no alignment.
6204 Enabled at levels @option{-O2}, @option{-O3}.
6207 @itemx -falign-loops=@var{n}
6208 @opindex falign-loops
6209 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6210 like @option{-falign-functions}. The hope is that the loop will be
6211 executed many times, which will make up for any execution of the dummy
6214 @option{-fno-align-loops} and @option{-falign-loops=1} are
6215 equivalent and mean that loops will not be aligned.
6217 If @var{n} is not specified or is zero, use a machine-dependent default.
6219 Enabled at levels @option{-O2}, @option{-O3}.
6222 @itemx -falign-jumps=@var{n}
6223 @opindex falign-jumps
6224 Align branch targets to a power-of-two boundary, for branch targets
6225 where the targets can only be reached by jumping, skipping up to @var{n}
6226 bytes like @option{-falign-functions}. In this case, no dummy operations
6229 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6230 equivalent and mean that loops will not be aligned.
6232 If @var{n} is not specified or is zero, use a machine-dependent default.
6234 Enabled at levels @option{-O2}, @option{-O3}.
6236 @item -funit-at-a-time
6237 @opindex funit-at-a-time
6238 Parse the whole compilation unit before starting to produce code.
6239 This allows some extra optimizations to take place but consumes
6240 more memory (in general). There are some compatibility issues
6241 with @emph{unit-at-a-time} mode:
6244 enabling @emph{unit-at-a-time} mode may change the order
6245 in which functions, variables, and top-level @code{asm} statements
6246 are emitted, and will likely break code relying on some particular
6247 ordering. The majority of such top-level @code{asm} statements,
6248 though, can be replaced by @code{section} attributes. The
6249 @option{fno-toplevel-reorder} option may be used to keep the ordering
6250 used in the input file, at the cost of some optimizations.
6253 @emph{unit-at-a-time} mode removes unreferenced static variables
6254 and functions. This may result in undefined references
6255 when an @code{asm} statement refers directly to variables or functions
6256 that are otherwise unused. In that case either the variable/function
6257 shall be listed as an operand of the @code{asm} statement operand or,
6258 in the case of top-level @code{asm} statements the attribute @code{used}
6259 shall be used on the declaration.
6262 Static functions now can use non-standard passing conventions that
6263 may break @code{asm} statements calling functions directly. Again,
6264 attribute @code{used} will prevent this behavior.
6267 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
6268 but this scheme may not be supported by future releases of GCC@.
6270 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6272 @item -fno-toplevel-reorder
6273 @opindex fno-toplevel-reorder
6274 Do not reorder top-level functions, variables, and @code{asm}
6275 statements. Output them in the same order that they appear in the
6276 input file. When this option is used, unreferenced static variables
6277 will not be removed. This option is intended to support existing code
6278 which relies on a particular ordering. For new code, it is better to
6283 Constructs webs as commonly used for register allocation purposes and assign
6284 each web individual pseudo register. This allows the register allocation pass
6285 to operate on pseudos directly, but also strengthens several other optimization
6286 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6287 however, make debugging impossible, since variables will no longer stay in a
6290 Enabled by default with @option{-funroll-loops}.
6292 @item -fwhole-program
6293 @opindex fwhole-program
6294 Assume that the current compilation unit represents whole program being
6295 compiled. All public functions and variables with the exception of @code{main}
6296 and those merged by attribute @code{externally_visible} become static functions
6297 and in a affect gets more aggressively optimized by interprocedural optimizers.
6298 While this option is equivalent to proper use of @code{static} keyword for
6299 programs consisting of single file, in combination with option
6300 @option{--combine} this flag can be used to compile most of smaller scale C
6301 programs since the functions and variables become local for the whole combined
6302 compilation unit, not for the single source file itself.
6304 This option is not supported for Fortran programs.
6306 @item -fcprop-registers
6307 @opindex fcprop-registers
6308 After register allocation and post-register allocation instruction splitting,
6309 we perform a copy-propagation pass to try to reduce scheduling dependencies
6310 and occasionally eliminate the copy.
6312 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6314 @item -fprofile-generate
6315 @opindex fprofile-generate
6317 Enable options usually used for instrumenting application to produce
6318 profile useful for later recompilation with profile feedback based
6319 optimization. You must use @option{-fprofile-generate} both when
6320 compiling and when linking your program.
6322 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6325 @opindex fprofile-use
6326 Enable profile feedback directed optimizations, and optimizations
6327 generally profitable only with profile feedback available.
6329 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6330 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6332 By default, GCC emits an error message if the feedback profiles do not
6333 match the source code. This error can be turned into a warning by using
6334 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6338 The following options control compiler behavior regarding floating
6339 point arithmetic. These options trade off between speed and
6340 correctness. All must be specifically enabled.
6344 @opindex ffloat-store
6345 Do not store floating point variables in registers, and inhibit other
6346 options that might change whether a floating point value is taken from a
6349 @cindex floating point precision
6350 This option prevents undesirable excess precision on machines such as
6351 the 68000 where the floating registers (of the 68881) keep more
6352 precision than a @code{double} is supposed to have. Similarly for the
6353 x86 architecture. For most programs, the excess precision does only
6354 good, but a few programs rely on the precise definition of IEEE floating
6355 point. Use @option{-ffloat-store} for such programs, after modifying
6356 them to store all pertinent intermediate computations into variables.
6360 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6361 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6362 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6364 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6366 This option is not turned on by any @option{-O} option since
6367 it can result in incorrect output for programs which depend on
6368 an exact implementation of IEEE or ISO rules/specifications for
6369 math functions. It may, however, yield faster code for programs
6370 that do not require the guarantees of these specifications.
6372 @item -fno-math-errno
6373 @opindex fno-math-errno
6374 Do not set ERRNO after calling math functions that are executed
6375 with a single instruction, e.g., sqrt. A program that relies on
6376 IEEE exceptions for math error handling may want to use this flag
6377 for speed while maintaining IEEE arithmetic compatibility.
6379 This option is not turned on by any @option{-O} option since
6380 it can result in incorrect output for programs which depend on
6381 an exact implementation of IEEE or ISO rules/specifications for
6382 math functions. It may, however, yield faster code for programs
6383 that do not require the guarantees of these specifications.
6385 The default is @option{-fmath-errno}.
6387 On Darwin systems, the math library never sets @code{errno}. There is
6388 therefore no reason for the compiler to consider the possibility that
6389 it might, and @option{-fno-math-errno} is the default.
6391 @item -funsafe-math-optimizations
6392 @opindex funsafe-math-optimizations
6394 Allow optimizations for floating-point arithmetic that (a) assume
6395 that arguments and results are valid and (b) may violate IEEE or
6396 ANSI standards. When used at link-time, it may include libraries
6397 or startup files that change the default FPU control word or other
6398 similar optimizations.
6400 This option is not turned on by any @option{-O} option since
6401 it can result in incorrect output for programs which depend on
6402 an exact implementation of IEEE or ISO rules/specifications for
6403 math functions. It may, however, yield faster code for programs
6404 that do not require the guarantees of these specifications.
6405 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6406 @option{-fassociative-math} and @option{-freciprocal-math}.
6408 The default is @option{-fno-unsafe-math-optimizations}.
6410 @item -fassociative-math
6411 @opindex fassociative-math
6413 Allow re-association of operands in series of floating-point operations.
6414 This violates the ISO C and C++ language standard by possibly changing
6415 computation result. NOTE: re-ordering may change the sign of zero as
6416 well as ignore NaNs and inhibit or create underflow or overflow (and
6417 thus cannot be used on a code which relies on rounding behavior like
6418 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6419 and thus may not be used when ordered comparisons are required.
6420 This option requires that both @option{-fno-signed-zeros} and
6421 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6422 much sense with @option{-frounding-math}.
6424 The default is @option{-fno-associative-math}.
6426 @item -freciprocal-math
6427 @opindex freciprocal-math
6429 Allow the reciprocal of a value to be used instead of dividing by
6430 the value if this enables optimizations. For example @code{x / y}
6431 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6432 is subject to common subexpression elimination. Note that this loses
6433 precision and increases the number of flops operating on the value.
6435 The default is @option{-fno-reciprocal-math}.
6437 @item -ffinite-math-only
6438 @opindex ffinite-math-only
6439 Allow optimizations for floating-point arithmetic that assume
6440 that arguments and results are not NaNs or +-Infs.
6442 This option is not turned on by any @option{-O} option since
6443 it can result in incorrect output for programs which depend on
6444 an exact implementation of IEEE or ISO rules/specifications for
6445 math functions. It may, however, yield faster code for programs
6446 that do not require the guarantees of these specifications.
6448 The default is @option{-fno-finite-math-only}.
6450 @item -fno-signed-zeros
6451 @opindex fno-signed-zeros
6452 Allow optimizations for floating point arithmetic that ignore the
6453 signedness of zero. IEEE arithmetic specifies the behavior of
6454 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6455 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6456 This option implies that the sign of a zero result isn't significant.
6458 The default is @option{-fsigned-zeros}.
6460 @item -fno-trapping-math
6461 @opindex fno-trapping-math
6462 Compile code assuming that floating-point operations cannot generate
6463 user-visible traps. These traps include division by zero, overflow,
6464 underflow, inexact result and invalid operation. This option requires
6465 that @option{-fno-signaling-nans} be in effect. Setting this option may
6466 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6468 This option should never be turned on by any @option{-O} option since
6469 it can result in incorrect output for programs which depend on
6470 an exact implementation of IEEE or ISO rules/specifications for
6473 The default is @option{-ftrapping-math}.
6475 @item -frounding-math
6476 @opindex frounding-math
6477 Disable transformations and optimizations that assume default floating
6478 point rounding behavior. This is round-to-zero for all floating point
6479 to integer conversions, and round-to-nearest for all other arithmetic
6480 truncations. This option should be specified for programs that change
6481 the FP rounding mode dynamically, or that may be executed with a
6482 non-default rounding mode. This option disables constant folding of
6483 floating point expressions at compile-time (which may be affected by
6484 rounding mode) and arithmetic transformations that are unsafe in the
6485 presence of sign-dependent rounding modes.
6487 The default is @option{-fno-rounding-math}.
6489 This option is experimental and does not currently guarantee to
6490 disable all GCC optimizations that are affected by rounding mode.
6491 Future versions of GCC may provide finer control of this setting
6492 using C99's @code{FENV_ACCESS} pragma. This command line option
6493 will be used to specify the default state for @code{FENV_ACCESS}.
6495 @item -frtl-abstract-sequences
6496 @opindex frtl-abstract-sequences
6497 It is a size optimization method. This option is to find identical
6498 sequences of code, which can be turned into pseudo-procedures and
6499 then replace all occurrences with calls to the newly created
6500 subroutine. It is kind of an opposite of @option{-finline-functions}.
6501 This optimization runs at RTL level.
6503 @item -fsignaling-nans
6504 @opindex fsignaling-nans
6505 Compile code assuming that IEEE signaling NaNs may generate user-visible
6506 traps during floating-point operations. Setting this option disables
6507 optimizations that may change the number of exceptions visible with
6508 signaling NaNs. This option implies @option{-ftrapping-math}.
6510 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6513 The default is @option{-fno-signaling-nans}.
6515 This option is experimental and does not currently guarantee to
6516 disable all GCC optimizations that affect signaling NaN behavior.
6518 @item -fsingle-precision-constant
6519 @opindex fsingle-precision-constant
6520 Treat floating point constant as single precision constant instead of
6521 implicitly converting it to double precision constant.
6523 @item -fcx-limited-range
6524 @opindex fcx-limited-range
6525 When enabled, this option states that a range reduction step is not
6526 needed when performing complex division. Also, there is no checking
6527 whether the result of a complex multiplication or division is @code{NaN
6528 + I*NaN}, with an attempt to rescue the situation in that case. The
6529 default is @option{-fno-cx-limited-range}, but is enabled by
6530 @option{-ffast-math}.
6532 This option controls the default setting of the ISO C99
6533 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6536 @item -fcx-fortran-rules
6537 @opindex fcx-fortran-rules
6538 Complex multiplication and division follow Fortran rules. Range
6539 reduction is done as part of complex division, but there is no checking
6540 whether the result of a complex multiplication or division is @code{NaN
6541 + I*NaN}, with an attempt to rescue the situation in that case.
6543 The default is @option{-fno-cx-fortran-rules}.
6547 The following options control optimizations that may improve
6548 performance, but are not enabled by any @option{-O} options. This
6549 section includes experimental options that may produce broken code.
6552 @item -fbranch-probabilities
6553 @opindex fbranch-probabilities
6554 After running a program compiled with @option{-fprofile-arcs}
6555 (@pxref{Debugging Options,, Options for Debugging Your Program or
6556 @command{gcc}}), you can compile it a second time using
6557 @option{-fbranch-probabilities}, to improve optimizations based on
6558 the number of times each branch was taken. When the program
6559 compiled with @option{-fprofile-arcs} exits it saves arc execution
6560 counts to a file called @file{@var{sourcename}.gcda} for each source
6561 file. The information in this data file is very dependent on the
6562 structure of the generated code, so you must use the same source code
6563 and the same optimization options for both compilations.
6565 With @option{-fbranch-probabilities}, GCC puts a
6566 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6567 These can be used to improve optimization. Currently, they are only
6568 used in one place: in @file{reorg.c}, instead of guessing which path a
6569 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6570 exactly determine which path is taken more often.
6572 @item -fprofile-values
6573 @opindex fprofile-values
6574 If combined with @option{-fprofile-arcs}, it adds code so that some
6575 data about values of expressions in the program is gathered.
6577 With @option{-fbranch-probabilities}, it reads back the data gathered
6578 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6579 notes to instructions for their later usage in optimizations.
6581 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6585 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6586 a code to gather information about values of expressions.
6588 With @option{-fbranch-probabilities}, it reads back the data gathered
6589 and actually performs the optimizations based on them.
6590 Currently the optimizations include specialization of division operation
6591 using the knowledge about the value of the denominator.
6593 @item -frename-registers
6594 @opindex frename-registers
6595 Attempt to avoid false dependencies in scheduled code by making use
6596 of registers left over after register allocation. This optimization
6597 will most benefit processors with lots of registers. Depending on the
6598 debug information format adopted by the target, however, it can
6599 make debugging impossible, since variables will no longer stay in
6600 a ``home register''.
6602 Enabled by default with @option{-funroll-loops}.
6606 Perform tail duplication to enlarge superblock size. This transformation
6607 simplifies the control flow of the function allowing other optimizations to do
6610 Enabled with @option{-fprofile-use}.
6612 @item -funroll-loops
6613 @opindex funroll-loops
6614 Unroll loops whose number of iterations can be determined at compile time or
6615 upon entry to the loop. @option{-funroll-loops} implies
6616 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6617 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6618 small constant number of iterations). This option makes code larger, and may
6619 or may not make it run faster.
6621 Enabled with @option{-fprofile-use}.
6623 @item -funroll-all-loops
6624 @opindex funroll-all-loops
6625 Unroll all loops, even if their number of iterations is uncertain when
6626 the loop is entered. This usually makes programs run more slowly.
6627 @option{-funroll-all-loops} implies the same options as
6628 @option{-funroll-loops}.
6631 @opindex fpeel-loops
6632 Peels the loops for that there is enough information that they do not
6633 roll much (from profile feedback). It also turns on complete loop peeling
6634 (i.e.@: complete removal of loops with small constant number of iterations).
6636 Enabled with @option{-fprofile-use}.
6638 @item -fmove-loop-invariants
6639 @opindex fmove-loop-invariants
6640 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6641 at level @option{-O1}
6643 @item -funswitch-loops
6644 @opindex funswitch-loops
6645 Move branches with loop invariant conditions out of the loop, with duplicates
6646 of the loop on both branches (modified according to result of the condition).
6648 @item -ffunction-sections
6649 @itemx -fdata-sections
6650 @opindex ffunction-sections
6651 @opindex fdata-sections
6652 Place each function or data item into its own section in the output
6653 file if the target supports arbitrary sections. The name of the
6654 function or the name of the data item determines the section's name
6657 Use these options on systems where the linker can perform optimizations
6658 to improve locality of reference in the instruction space. Most systems
6659 using the ELF object format and SPARC processors running Solaris 2 have
6660 linkers with such optimizations. AIX may have these optimizations in
6663 Only use these options when there are significant benefits from doing
6664 so. When you specify these options, the assembler and linker will
6665 create larger object and executable files and will also be slower.
6666 You will not be able to use @code{gprof} on all systems if you
6667 specify this option and you may have problems with debugging if
6668 you specify both this option and @option{-g}.
6670 @item -fbranch-target-load-optimize
6671 @opindex fbranch-target-load-optimize
6672 Perform branch target register load optimization before prologue / epilogue
6674 The use of target registers can typically be exposed only during reload,
6675 thus hoisting loads out of loops and doing inter-block scheduling needs
6676 a separate optimization pass.
6678 @item -fbranch-target-load-optimize2
6679 @opindex fbranch-target-load-optimize2
6680 Perform branch target register load optimization after prologue / epilogue
6683 @item -fbtr-bb-exclusive
6684 @opindex fbtr-bb-exclusive
6685 When performing branch target register load optimization, don't reuse
6686 branch target registers in within any basic block.
6688 @item -fstack-protector
6689 @opindex fstack-protector
6690 Emit extra code to check for buffer overflows, such as stack smashing
6691 attacks. This is done by adding a guard variable to functions with
6692 vulnerable objects. This includes functions that call alloca, and
6693 functions with buffers larger than 8 bytes. The guards are initialized
6694 when a function is entered and then checked when the function exits.
6695 If a guard check fails, an error message is printed and the program exits.
6697 @item -fstack-protector-all
6698 @opindex fstack-protector-all
6699 Like @option{-fstack-protector} except that all functions are protected.
6701 @item -fsection-anchors
6702 @opindex fsection-anchors
6703 Try to reduce the number of symbolic address calculations by using
6704 shared ``anchor'' symbols to address nearby objects. This transformation
6705 can help to reduce the number of GOT entries and GOT accesses on some
6708 For example, the implementation of the following function @code{foo}:
6712 int foo (void) @{ return a + b + c; @}
6715 would usually calculate the addresses of all three variables, but if you
6716 compile it with @option{-fsection-anchors}, it will access the variables
6717 from a common anchor point instead. The effect is similar to the
6718 following pseudocode (which isn't valid C):
6723 register int *xr = &x;
6724 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6728 Not all targets support this option.
6730 @item --param @var{name}=@var{value}
6732 In some places, GCC uses various constants to control the amount of
6733 optimization that is done. For example, GCC will not inline functions
6734 that contain more that a certain number of instructions. You can
6735 control some of these constants on the command-line using the
6736 @option{--param} option.
6738 The names of specific parameters, and the meaning of the values, are
6739 tied to the internals of the compiler, and are subject to change
6740 without notice in future releases.
6742 In each case, the @var{value} is an integer. The allowable choices for
6743 @var{name} are given in the following table:
6746 @item salias-max-implicit-fields
6747 The maximum number of fields in a variable without direct
6748 structure accesses for which structure aliasing will consider trying
6749 to track each field. The default is 5
6751 @item salias-max-array-elements
6752 The maximum number of elements an array can have and its elements
6753 still be tracked individually by structure aliasing. The default is 4
6755 @item sra-max-structure-size
6756 The maximum structure size, in bytes, at which the scalar replacement
6757 of aggregates (SRA) optimization will perform block copies. The
6758 default value, 0, implies that GCC will select the most appropriate
6761 @item sra-field-structure-ratio
6762 The threshold ratio (as a percentage) between instantiated fields and
6763 the complete structure size. We say that if the ratio of the number
6764 of bytes in instantiated fields to the number of bytes in the complete
6765 structure exceeds this parameter, then block copies are not used. The
6768 @item struct-reorg-cold-struct-ratio
6769 The threshold ratio (as a percentage) between a structure frequency
6770 and the frequency of the hottest structure in the program. This parameter
6771 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
6772 We say that if the ratio of a structure frequency, calculated by profiling,
6773 to the hottest structure frequency in the program is less than this
6774 parameter, then structure reorganization is not applied to this structure.
6777 @item max-crossjump-edges
6778 The maximum number of incoming edges to consider for crossjumping.
6779 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6780 the number of edges incoming to each block. Increasing values mean
6781 more aggressive optimization, making the compile time increase with
6782 probably small improvement in executable size.
6784 @item min-crossjump-insns
6785 The minimum number of instructions which must be matched at the end
6786 of two blocks before crossjumping will be performed on them. This
6787 value is ignored in the case where all instructions in the block being
6788 crossjumped from are matched. The default value is 5.
6790 @item max-grow-copy-bb-insns
6791 The maximum code size expansion factor when copying basic blocks
6792 instead of jumping. The expansion is relative to a jump instruction.
6793 The default value is 8.
6795 @item max-goto-duplication-insns
6796 The maximum number of instructions to duplicate to a block that jumps
6797 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6798 passes, GCC factors computed gotos early in the compilation process,
6799 and unfactors them as late as possible. Only computed jumps at the
6800 end of a basic blocks with no more than max-goto-duplication-insns are
6801 unfactored. The default value is 8.
6803 @item max-delay-slot-insn-search
6804 The maximum number of instructions to consider when looking for an
6805 instruction to fill a delay slot. If more than this arbitrary number of
6806 instructions is searched, the time savings from filling the delay slot
6807 will be minimal so stop searching. Increasing values mean more
6808 aggressive optimization, making the compile time increase with probably
6809 small improvement in executable run time.
6811 @item max-delay-slot-live-search
6812 When trying to fill delay slots, the maximum number of instructions to
6813 consider when searching for a block with valid live register
6814 information. Increasing this arbitrarily chosen value means more
6815 aggressive optimization, increasing the compile time. This parameter
6816 should be removed when the delay slot code is rewritten to maintain the
6819 @item max-gcse-memory
6820 The approximate maximum amount of memory that will be allocated in
6821 order to perform the global common subexpression elimination
6822 optimization. If more memory than specified is required, the
6823 optimization will not be done.
6825 @item max-gcse-passes
6826 The maximum number of passes of GCSE to run. The default is 1.
6828 @item max-pending-list-length
6829 The maximum number of pending dependencies scheduling will allow
6830 before flushing the current state and starting over. Large functions
6831 with few branches or calls can create excessively large lists which
6832 needlessly consume memory and resources.
6834 @item max-inline-insns-single
6835 Several parameters control the tree inliner used in gcc.
6836 This number sets the maximum number of instructions (counted in GCC's
6837 internal representation) in a single function that the tree inliner
6838 will consider for inlining. This only affects functions declared
6839 inline and methods implemented in a class declaration (C++).
6840 The default value is 450.
6842 @item max-inline-insns-auto
6843 When you use @option{-finline-functions} (included in @option{-O3}),
6844 a lot of functions that would otherwise not be considered for inlining
6845 by the compiler will be investigated. To those functions, a different
6846 (more restrictive) limit compared to functions declared inline can
6848 The default value is 90.
6850 @item large-function-insns
6851 The limit specifying really large functions. For functions larger than this
6852 limit after inlining inlining is constrained by
6853 @option{--param large-function-growth}. This parameter is useful primarily
6854 to avoid extreme compilation time caused by non-linear algorithms used by the
6856 This parameter is ignored when @option{-funit-at-a-time} is not used.
6857 The default value is 2700.
6859 @item large-function-growth
6860 Specifies maximal growth of large function caused by inlining in percents.
6861 This parameter is ignored when @option{-funit-at-a-time} is not used.
6862 The default value is 100 which limits large function growth to 2.0 times
6865 @item large-unit-insns
6866 The limit specifying large translation unit. Growth caused by inlining of
6867 units larger than this limit is limited by @option{--param inline-unit-growth}.
6868 For small units this might be too tight (consider unit consisting of function A
6869 that is inline and B that just calls A three time. If B is small relative to
6870 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6871 large units consisting of small inlineable functions however the overall unit
6872 growth limit is needed to avoid exponential explosion of code size. Thus for
6873 smaller units, the size is increased to @option{--param large-unit-insns}
6874 before applying @option{--param inline-unit-growth}. The default is 10000
6876 @item inline-unit-growth
6877 Specifies maximal overall growth of the compilation unit caused by inlining.
6878 This parameter is ignored when @option{-funit-at-a-time} is not used.
6879 The default value is 30 which limits unit growth to 1.3 times the original
6882 @item large-stack-frame
6883 The limit specifying large stack frames. While inlining the algorithm is trying
6884 to not grow past this limit too much. Default value is 256 bytes.
6886 @item large-stack-frame-growth
6887 Specifies maximal growth of large stack frames caused by inlining in percents.
6888 The default value is 1000 which limits large stack frame growth to 11 times
6891 @item max-inline-insns-recursive
6892 @itemx max-inline-insns-recursive-auto
6893 Specifies maximum number of instructions out-of-line copy of self recursive inline
6894 function can grow into by performing recursive inlining.
6896 For functions declared inline @option{--param max-inline-insns-recursive} is
6897 taken into account. For function not declared inline, recursive inlining
6898 happens only when @option{-finline-functions} (included in @option{-O3}) is
6899 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6900 default value is 450.
6902 @item max-inline-recursive-depth
6903 @itemx max-inline-recursive-depth-auto
6904 Specifies maximum recursion depth used by the recursive inlining.
6906 For functions declared inline @option{--param max-inline-recursive-depth} is
6907 taken into account. For function not declared inline, recursive inlining
6908 happens only when @option{-finline-functions} (included in @option{-O3}) is
6909 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6912 @item min-inline-recursive-probability
6913 Recursive inlining is profitable only for function having deep recursion
6914 in average and can hurt for function having little recursion depth by
6915 increasing the prologue size or complexity of function body to other
6918 When profile feedback is available (see @option{-fprofile-generate}) the actual
6919 recursion depth can be guessed from probability that function will recurse via
6920 given call expression. This parameter limits inlining only to call expression
6921 whose probability exceeds given threshold (in percents). The default value is
6924 @item inline-call-cost
6925 Specify cost of call instruction relative to simple arithmetics operations
6926 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6927 functions and at the same time increases size of leaf function that is believed to
6928 reduce function size by being inlined. In effect it increases amount of
6929 inlining for code having large abstraction penalty (many functions that just
6930 pass the arguments to other functions) and decrease inlining for code with low
6931 abstraction penalty. The default value is 12.
6933 @item min-vect-loop-bound
6934 The minimum number of iterations under which a loop will not get vectorized
6935 when @option{-ftree-vectorize} is used. The number of iterations after
6936 vectorization needs to be greater than the value specified by this option
6937 to allow vectorization. The default value is 0.
6939 @item max-unrolled-insns
6940 The maximum number of instructions that a loop should have if that loop
6941 is unrolled, and if the loop is unrolled, it determines how many times
6942 the loop code is unrolled.
6944 @item max-average-unrolled-insns
6945 The maximum number of instructions biased by probabilities of their execution
6946 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6947 it determines how many times the loop code is unrolled.
6949 @item max-unroll-times
6950 The maximum number of unrollings of a single loop.
6952 @item max-peeled-insns
6953 The maximum number of instructions that a loop should have if that loop
6954 is peeled, and if the loop is peeled, it determines how many times
6955 the loop code is peeled.
6957 @item max-peel-times
6958 The maximum number of peelings of a single loop.
6960 @item max-completely-peeled-insns
6961 The maximum number of insns of a completely peeled loop.
6963 @item max-completely-peel-times
6964 The maximum number of iterations of a loop to be suitable for complete peeling.
6966 @item max-unswitch-insns
6967 The maximum number of insns of an unswitched loop.
6969 @item max-unswitch-level
6970 The maximum number of branches unswitched in a single loop.
6973 The minimum cost of an expensive expression in the loop invariant motion.
6975 @item iv-consider-all-candidates-bound
6976 Bound on number of candidates for induction variables below that
6977 all candidates are considered for each use in induction variable
6978 optimizations. Only the most relevant candidates are considered
6979 if there are more candidates, to avoid quadratic time complexity.
6981 @item iv-max-considered-uses
6982 The induction variable optimizations give up on loops that contain more
6983 induction variable uses.
6985 @item iv-always-prune-cand-set-bound
6986 If number of candidates in the set is smaller than this value,
6987 we always try to remove unnecessary ivs from the set during its
6988 optimization when a new iv is added to the set.
6990 @item scev-max-expr-size
6991 Bound on size of expressions used in the scalar evolutions analyzer.
6992 Large expressions slow the analyzer.
6994 @item omega-max-vars
6995 The maximum number of variables in an Omega constraint system.
6996 The default value is 128.
6998 @item omega-max-geqs
6999 The maximum number of inequalities in an Omega constraint system.
7000 The default value is 256.
7003 The maximum number of equalities in an Omega constraint system.
7004 The default value is 128.
7006 @item omega-max-wild-cards
7007 The maximum number of wildcard variables that the Omega solver will
7008 be able to insert. The default value is 18.
7010 @item omega-hash-table-size
7011 The size of the hash table in the Omega solver. The default value is
7014 @item omega-max-keys
7015 The maximal number of keys used by the Omega solver. The default
7018 @item omega-eliminate-redundant-constraints
7019 When set to 1, use expensive methods to eliminate all redundant
7020 constraints. The default value is 0.
7022 @item vect-max-version-for-alignment-checks
7023 The maximum number of runtime checks that can be performed when
7024 doing loop versioning for alignment in the vectorizer. See option
7025 ftree-vect-loop-version for more information.
7027 @item vect-max-version-for-alias-checks
7028 The maximum number of runtime checks that can be performed when
7029 doing loop versioning for alias in the vectorizer. See option
7030 ftree-vect-loop-version for more information.
7032 @item max-iterations-to-track
7034 The maximum number of iterations of a loop the brute force algorithm
7035 for analysis of # of iterations of the loop tries to evaluate.
7037 @item hot-bb-count-fraction
7038 Select fraction of the maximal count of repetitions of basic block in program
7039 given basic block needs to have to be considered hot.
7041 @item hot-bb-frequency-fraction
7042 Select fraction of the maximal frequency of executions of basic block in
7043 function given basic block needs to have to be considered hot
7045 @item max-predicted-iterations
7046 The maximum number of loop iterations we predict statically. This is useful
7047 in cases where function contain single loop with known bound and other loop
7048 with unknown. We predict the known number of iterations correctly, while
7049 the unknown number of iterations average to roughly 10. This means that the
7050 loop without bounds would appear artificially cold relative to the other one.
7052 @item align-threshold
7054 Select fraction of the maximal frequency of executions of basic block in
7055 function given basic block will get aligned.
7057 @item align-loop-iterations
7059 A loop expected to iterate at lest the selected number of iterations will get
7062 @item tracer-dynamic-coverage
7063 @itemx tracer-dynamic-coverage-feedback
7065 This value is used to limit superblock formation once the given percentage of
7066 executed instructions is covered. This limits unnecessary code size
7069 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7070 feedback is available. The real profiles (as opposed to statically estimated
7071 ones) are much less balanced allowing the threshold to be larger value.
7073 @item tracer-max-code-growth
7074 Stop tail duplication once code growth has reached given percentage. This is
7075 rather hokey argument, as most of the duplicates will be eliminated later in
7076 cross jumping, so it may be set to much higher values than is the desired code
7079 @item tracer-min-branch-ratio
7081 Stop reverse growth when the reverse probability of best edge is less than this
7082 threshold (in percent).
7084 @item tracer-min-branch-ratio
7085 @itemx tracer-min-branch-ratio-feedback
7087 Stop forward growth if the best edge do have probability lower than this
7090 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7091 compilation for profile feedback and one for compilation without. The value
7092 for compilation with profile feedback needs to be more conservative (higher) in
7093 order to make tracer effective.
7095 @item max-cse-path-length
7097 Maximum number of basic blocks on path that cse considers. The default is 10.
7100 The maximum instructions CSE process before flushing. The default is 1000.
7102 @item max-aliased-vops
7104 Maximum number of virtual operands per function allowed to represent
7105 aliases before triggering the alias partitioning heuristic. Alias
7106 partitioning reduces compile times and memory consumption needed for
7107 aliasing at the expense of precision loss in alias information. The
7108 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
7111 Notice that if a function contains more memory statements than the
7112 value of this parameter, it is not really possible to achieve this
7113 reduction. In this case, the compiler will use the number of memory
7114 statements as the value for @option{max-aliased-vops}.
7116 @item avg-aliased-vops
7118 Average number of virtual operands per statement allowed to represent
7119 aliases before triggering the alias partitioning heuristic. This
7120 works in conjunction with @option{max-aliased-vops}. If a function
7121 contains more than @option{max-aliased-vops} virtual operators, then
7122 memory symbols will be grouped into memory partitions until either the
7123 total number of virtual operators is below @option{max-aliased-vops}
7124 or the average number of virtual operators per memory statement is
7125 below @option{avg-aliased-vops}. The default value for this parameter
7126 is 1 for -O1 and -O2, and 3 for -O3.
7128 @item ggc-min-expand
7130 GCC uses a garbage collector to manage its own memory allocation. This
7131 parameter specifies the minimum percentage by which the garbage
7132 collector's heap should be allowed to expand between collections.
7133 Tuning this may improve compilation speed; it has no effect on code
7136 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7137 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7138 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7139 GCC is not able to calculate RAM on a particular platform, the lower
7140 bound of 30% is used. Setting this parameter and
7141 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7142 every opportunity. This is extremely slow, but can be useful for
7145 @item ggc-min-heapsize
7147 Minimum size of the garbage collector's heap before it begins bothering
7148 to collect garbage. The first collection occurs after the heap expands
7149 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7150 tuning this may improve compilation speed, and has no effect on code
7153 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7154 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7155 with a lower bound of 4096 (four megabytes) and an upper bound of
7156 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7157 particular platform, the lower bound is used. Setting this parameter
7158 very large effectively disables garbage collection. Setting this
7159 parameter and @option{ggc-min-expand} to zero causes a full collection
7160 to occur at every opportunity.
7162 @item max-reload-search-insns
7163 The maximum number of instruction reload should look backward for equivalent
7164 register. Increasing values mean more aggressive optimization, making the
7165 compile time increase with probably slightly better performance. The default
7168 @item max-cselib-memory-locations
7169 The maximum number of memory locations cselib should take into account.
7170 Increasing values mean more aggressive optimization, making the compile time
7171 increase with probably slightly better performance. The default value is 500.
7173 @item max-flow-memory-locations
7174 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
7175 The default value is 100.
7177 @item reorder-blocks-duplicate
7178 @itemx reorder-blocks-duplicate-feedback
7180 Used by basic block reordering pass to decide whether to use unconditional
7181 branch or duplicate the code on its destination. Code is duplicated when its
7182 estimated size is smaller than this value multiplied by the estimated size of
7183 unconditional jump in the hot spots of the program.
7185 The @option{reorder-block-duplicate-feedback} is used only when profile
7186 feedback is available and may be set to higher values than
7187 @option{reorder-block-duplicate} since information about the hot spots is more
7190 @item max-sched-ready-insns
7191 The maximum number of instructions ready to be issued the scheduler should
7192 consider at any given time during the first scheduling pass. Increasing
7193 values mean more thorough searches, making the compilation time increase
7194 with probably little benefit. The default value is 100.
7196 @item max-sched-region-blocks
7197 The maximum number of blocks in a region to be considered for
7198 interblock scheduling. The default value is 10.
7200 @item max-sched-region-insns
7201 The maximum number of insns in a region to be considered for
7202 interblock scheduling. The default value is 100.
7205 The minimum probability (in percents) of reaching a source block
7206 for interblock speculative scheduling. The default value is 40.
7208 @item max-sched-extend-regions-iters
7209 The maximum number of iterations through CFG to extend regions.
7210 0 - disable region extension,
7211 N - do at most N iterations.
7212 The default value is 0.
7214 @item max-sched-insn-conflict-delay
7215 The maximum conflict delay for an insn to be considered for speculative motion.
7216 The default value is 3.
7218 @item sched-spec-prob-cutoff
7219 The minimal probability of speculation success (in percents), so that
7220 speculative insn will be scheduled.
7221 The default value is 40.
7223 @item max-last-value-rtl
7225 The maximum size measured as number of RTLs that can be recorded in an expression
7226 in combiner for a pseudo register as last known value of that register. The default
7229 @item integer-share-limit
7230 Small integer constants can use a shared data structure, reducing the
7231 compiler's memory usage and increasing its speed. This sets the maximum
7232 value of a shared integer constant. The default value is 256.
7234 @item min-virtual-mappings
7235 Specifies the minimum number of virtual mappings in the incremental
7236 SSA updater that should be registered to trigger the virtual mappings
7237 heuristic defined by virtual-mappings-ratio. The default value is
7240 @item virtual-mappings-ratio
7241 If the number of virtual mappings is virtual-mappings-ratio bigger
7242 than the number of virtual symbols to be updated, then the incremental
7243 SSA updater switches to a full update for those symbols. The default
7246 @item ssp-buffer-size
7247 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7248 protection when @option{-fstack-protection} is used.
7250 @item max-jump-thread-duplication-stmts
7251 Maximum number of statements allowed in a block that needs to be
7252 duplicated when threading jumps.
7254 @item max-fields-for-field-sensitive
7255 Maximum number of fields in a structure we will treat in
7256 a field sensitive manner during pointer analysis.
7258 @item prefetch-latency
7259 Estimate on average number of instructions that are executed before
7260 prefetch finishes. The distance we prefetch ahead is proportional
7261 to this constant. Increasing this number may also lead to less
7262 streams being prefetched (see @option{simultaneous-prefetches}).
7264 @item simultaneous-prefetches
7265 Maximum number of prefetches that can run at the same time.
7267 @item l1-cache-line-size
7268 The size of cache line in L1 cache, in bytes.
7271 The size of L1 cache, in kilobytes.
7274 The size of L2 cache, in kilobytes.
7276 @item use-canonical-types
7277 Whether the compiler should use the ``canonical'' type system. By
7278 default, this should always be 1, which uses a more efficient internal
7279 mechanism for comparing types in C++ and Objective-C++. However, if
7280 bugs in the canonical type system are causing compilation failures,
7281 set this value to 0 to disable canonical types.
7283 @item max-partial-antic-length
7284 Maximum length of the partial antic set computed during the tree
7285 partial redundancy elimination optimization (@option{-ftree-pre}) when
7286 optimizing at @option{-O3} and above. For some sorts of source code
7287 the enhanced partial redundancy elimination optimization can run away,
7288 consuming all of the memory available on the host machine. This
7289 parameter sets a limit on the length of the sets that are computed,
7290 which prevents the runaway behaviour. Setting a value of 0 for
7291 this paramter will allow an unlimited set length.
7293 @item sccvn-max-scc-size
7294 Maximum size of a strongly connected component (SCC) during SCCVN
7295 processing. If this limit is hit, SCCVN processing for the whole
7296 function will not be done and optimizations depending on it will
7297 be disabled. The default maximum SCC size is 10000.
7302 @node Preprocessor Options
7303 @section Options Controlling the Preprocessor
7304 @cindex preprocessor options
7305 @cindex options, preprocessor
7307 These options control the C preprocessor, which is run on each C source
7308 file before actual compilation.
7310 If you use the @option{-E} option, nothing is done except preprocessing.
7311 Some of these options make sense only together with @option{-E} because
7312 they cause the preprocessor output to be unsuitable for actual
7317 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7318 and pass @var{option} directly through to the preprocessor. If
7319 @var{option} contains commas, it is split into multiple options at the
7320 commas. However, many options are modified, translated or interpreted
7321 by the compiler driver before being passed to the preprocessor, and
7322 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7323 interface is undocumented and subject to change, so whenever possible
7324 you should avoid using @option{-Wp} and let the driver handle the
7327 @item -Xpreprocessor @var{option}
7328 @opindex preprocessor
7329 Pass @var{option} as an option to the preprocessor. You can use this to
7330 supply system-specific preprocessor options which GCC does not know how to
7333 If you want to pass an option that takes an argument, you must use
7334 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7337 @include cppopts.texi
7339 @node Assembler Options
7340 @section Passing Options to the Assembler
7342 @c prevent bad page break with this line
7343 You can pass options to the assembler.
7346 @item -Wa,@var{option}
7348 Pass @var{option} as an option to the assembler. If @var{option}
7349 contains commas, it is split into multiple options at the commas.
7351 @item -Xassembler @var{option}
7353 Pass @var{option} as an option to the assembler. You can use this to
7354 supply system-specific assembler options which GCC does not know how to
7357 If you want to pass an option that takes an argument, you must use
7358 @option{-Xassembler} twice, once for the option and once for the argument.
7363 @section Options for Linking
7364 @cindex link options
7365 @cindex options, linking
7367 These options come into play when the compiler links object files into
7368 an executable output file. They are meaningless if the compiler is
7369 not doing a link step.
7373 @item @var{object-file-name}
7374 A file name that does not end in a special recognized suffix is
7375 considered to name an object file or library. (Object files are
7376 distinguished from libraries by the linker according to the file
7377 contents.) If linking is done, these object files are used as input
7386 If any of these options is used, then the linker is not run, and
7387 object file names should not be used as arguments. @xref{Overall
7391 @item -l@var{library}
7392 @itemx -l @var{library}
7394 Search the library named @var{library} when linking. (The second
7395 alternative with the library as a separate argument is only for
7396 POSIX compliance and is not recommended.)
7398 It makes a difference where in the command you write this option; the
7399 linker searches and processes libraries and object files in the order they
7400 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7401 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7402 to functions in @samp{z}, those functions may not be loaded.
7404 The linker searches a standard list of directories for the library,
7405 which is actually a file named @file{lib@var{library}.a}. The linker
7406 then uses this file as if it had been specified precisely by name.
7408 The directories searched include several standard system directories
7409 plus any that you specify with @option{-L}.
7411 Normally the files found this way are library files---archive files
7412 whose members are object files. The linker handles an archive file by
7413 scanning through it for members which define symbols that have so far
7414 been referenced but not defined. But if the file that is found is an
7415 ordinary object file, it is linked in the usual fashion. The only
7416 difference between using an @option{-l} option and specifying a file name
7417 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7418 and searches several directories.
7422 You need this special case of the @option{-l} option in order to
7423 link an Objective-C or Objective-C++ program.
7426 @opindex nostartfiles
7427 Do not use the standard system startup files when linking.
7428 The standard system libraries are used normally, unless @option{-nostdlib}
7429 or @option{-nodefaultlibs} is used.
7431 @item -nodefaultlibs
7432 @opindex nodefaultlibs
7433 Do not use the standard system libraries when linking.
7434 Only the libraries you specify will be passed to the linker.
7435 The standard startup files are used normally, unless @option{-nostartfiles}
7436 is used. The compiler may generate calls to @code{memcmp},
7437 @code{memset}, @code{memcpy} and @code{memmove}.
7438 These entries are usually resolved by entries in
7439 libc. These entry points should be supplied through some other
7440 mechanism when this option is specified.
7444 Do not use the standard system startup files or libraries when linking.
7445 No startup files and only the libraries you specify will be passed to
7446 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7447 @code{memcpy} and @code{memmove}.
7448 These entries are usually resolved by entries in
7449 libc. These entry points should be supplied through some other
7450 mechanism when this option is specified.
7452 @cindex @option{-lgcc}, use with @option{-nostdlib}
7453 @cindex @option{-nostdlib} and unresolved references
7454 @cindex unresolved references and @option{-nostdlib}
7455 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7456 @cindex @option{-nodefaultlibs} and unresolved references
7457 @cindex unresolved references and @option{-nodefaultlibs}
7458 One of the standard libraries bypassed by @option{-nostdlib} and
7459 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7460 that GCC uses to overcome shortcomings of particular machines, or special
7461 needs for some languages.
7462 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7463 Collection (GCC) Internals},
7464 for more discussion of @file{libgcc.a}.)
7465 In most cases, you need @file{libgcc.a} even when you want to avoid
7466 other standard libraries. In other words, when you specify @option{-nostdlib}
7467 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7468 This ensures that you have no unresolved references to internal GCC
7469 library subroutines. (For example, @samp{__main}, used to ensure C++
7470 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7471 GNU Compiler Collection (GCC) Internals}.)
7475 Produce a position independent executable on targets which support it.
7476 For predictable results, you must also specify the same set of options
7477 that were used to generate code (@option{-fpie}, @option{-fPIE},
7478 or model suboptions) when you specify this option.
7482 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7483 that support it. This instructs the linker to add all symbols, not
7484 only used ones, to the dynamic symbol table. This option is needed
7485 for some uses of @code{dlopen} or to allow obtaining backtraces
7486 from within a program.
7490 Remove all symbol table and relocation information from the executable.
7494 On systems that support dynamic linking, this prevents linking with the shared
7495 libraries. On other systems, this option has no effect.
7499 Produce a shared object which can then be linked with other objects to
7500 form an executable. Not all systems support this option. For predictable
7501 results, you must also specify the same set of options that were used to
7502 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7503 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7504 needs to build supplementary stub code for constructors to work. On
7505 multi-libbed systems, @samp{gcc -shared} must select the correct support
7506 libraries to link against. Failing to supply the correct flags may lead
7507 to subtle defects. Supplying them in cases where they are not necessary
7510 @item -shared-libgcc
7511 @itemx -static-libgcc
7512 @opindex shared-libgcc
7513 @opindex static-libgcc
7514 On systems that provide @file{libgcc} as a shared library, these options
7515 force the use of either the shared or static version respectively.
7516 If no shared version of @file{libgcc} was built when the compiler was
7517 configured, these options have no effect.
7519 There are several situations in which an application should use the
7520 shared @file{libgcc} instead of the static version. The most common
7521 of these is when the application wishes to throw and catch exceptions
7522 across different shared libraries. In that case, each of the libraries
7523 as well as the application itself should use the shared @file{libgcc}.
7525 Therefore, the G++ and GCJ drivers automatically add
7526 @option{-shared-libgcc} whenever you build a shared library or a main
7527 executable, because C++ and Java programs typically use exceptions, so
7528 this is the right thing to do.
7530 If, instead, you use the GCC driver to create shared libraries, you may
7531 find that they will not always be linked with the shared @file{libgcc}.
7532 If GCC finds, at its configuration time, that you have a non-GNU linker
7533 or a GNU linker that does not support option @option{--eh-frame-hdr},
7534 it will link the shared version of @file{libgcc} into shared libraries
7535 by default. Otherwise, it will take advantage of the linker and optimize
7536 away the linking with the shared version of @file{libgcc}, linking with
7537 the static version of libgcc by default. This allows exceptions to
7538 propagate through such shared libraries, without incurring relocation
7539 costs at library load time.
7541 However, if a library or main executable is supposed to throw or catch
7542 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7543 for the languages used in the program, or using the option
7544 @option{-shared-libgcc}, such that it is linked with the shared
7549 Bind references to global symbols when building a shared object. Warn
7550 about any unresolved references (unless overridden by the link editor
7551 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7554 @item -Xlinker @var{option}
7556 Pass @var{option} as an option to the linker. You can use this to
7557 supply system-specific linker options which GCC does not know how to
7560 If you want to pass an option that takes an argument, you must use
7561 @option{-Xlinker} twice, once for the option and once for the argument.
7562 For example, to pass @option{-assert definitions}, you must write
7563 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7564 @option{-Xlinker "-assert definitions"}, because this passes the entire
7565 string as a single argument, which is not what the linker expects.
7567 @item -Wl,@var{option}
7569 Pass @var{option} as an option to the linker. If @var{option} contains
7570 commas, it is split into multiple options at the commas.
7572 @item -u @var{symbol}
7574 Pretend the symbol @var{symbol} is undefined, to force linking of
7575 library modules to define it. You can use @option{-u} multiple times with
7576 different symbols to force loading of additional library modules.
7579 @node Directory Options
7580 @section Options for Directory Search
7581 @cindex directory options
7582 @cindex options, directory search
7585 These options specify directories to search for header files, for
7586 libraries and for parts of the compiler:
7591 Add the directory @var{dir} to the head of the list of directories to be
7592 searched for header files. This can be used to override a system header
7593 file, substituting your own version, since these directories are
7594 searched before the system header file directories. However, you should
7595 not use this option to add directories that contain vendor-supplied
7596 system header files (use @option{-isystem} for that). If you use more than
7597 one @option{-I} option, the directories are scanned in left-to-right
7598 order; the standard system directories come after.
7600 If a standard system include directory, or a directory specified with
7601 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7602 option will be ignored. The directory will still be searched but as a
7603 system directory at its normal position in the system include chain.
7604 This is to ensure that GCC's procedure to fix buggy system headers and
7605 the ordering for the include_next directive are not inadvertently changed.
7606 If you really need to change the search order for system directories,
7607 use the @option{-nostdinc} and/or @option{-isystem} options.
7609 @item -iquote@var{dir}
7611 Add the directory @var{dir} to the head of the list of directories to
7612 be searched for header files only for the case of @samp{#include
7613 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7614 otherwise just like @option{-I}.
7618 Add directory @var{dir} to the list of directories to be searched
7621 @item -B@var{prefix}
7623 This option specifies where to find the executables, libraries,
7624 include files, and data files of the compiler itself.
7626 The compiler driver program runs one or more of the subprograms
7627 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7628 @var{prefix} as a prefix for each program it tries to run, both with and
7629 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7631 For each subprogram to be run, the compiler driver first tries the
7632 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7633 was not specified, the driver tries two standard prefixes, which are
7634 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7635 those results in a file name that is found, the unmodified program
7636 name is searched for using the directories specified in your
7637 @env{PATH} environment variable.
7639 The compiler will check to see if the path provided by the @option{-B}
7640 refers to a directory, and if necessary it will add a directory
7641 separator character at the end of the path.
7643 @option{-B} prefixes that effectively specify directory names also apply
7644 to libraries in the linker, because the compiler translates these
7645 options into @option{-L} options for the linker. They also apply to
7646 includes files in the preprocessor, because the compiler translates these
7647 options into @option{-isystem} options for the preprocessor. In this case,
7648 the compiler appends @samp{include} to the prefix.
7650 The run-time support file @file{libgcc.a} can also be searched for using
7651 the @option{-B} prefix, if needed. If it is not found there, the two
7652 standard prefixes above are tried, and that is all. The file is left
7653 out of the link if it is not found by those means.
7655 Another way to specify a prefix much like the @option{-B} prefix is to use
7656 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7659 As a special kludge, if the path provided by @option{-B} is
7660 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7661 9, then it will be replaced by @file{[dir/]include}. This is to help
7662 with boot-strapping the compiler.
7664 @item -specs=@var{file}
7666 Process @var{file} after the compiler reads in the standard @file{specs}
7667 file, in order to override the defaults that the @file{gcc} driver
7668 program uses when determining what switches to pass to @file{cc1},
7669 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7670 @option{-specs=@var{file}} can be specified on the command line, and they
7671 are processed in order, from left to right.
7673 @item --sysroot=@var{dir}
7675 Use @var{dir} as the logical root directory for headers and libraries.
7676 For example, if the compiler would normally search for headers in
7677 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7678 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7680 If you use both this option and the @option{-isysroot} option, then
7681 the @option{--sysroot} option will apply to libraries, but the
7682 @option{-isysroot} option will apply to header files.
7684 The GNU linker (beginning with version 2.16) has the necessary support
7685 for this option. If your linker does not support this option, the
7686 header file aspect of @option{--sysroot} will still work, but the
7687 library aspect will not.
7691 This option has been deprecated. Please use @option{-iquote} instead for
7692 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7693 Any directories you specify with @option{-I} options before the @option{-I-}
7694 option are searched only for the case of @samp{#include "@var{file}"};
7695 they are not searched for @samp{#include <@var{file}>}.
7697 If additional directories are specified with @option{-I} options after
7698 the @option{-I-}, these directories are searched for all @samp{#include}
7699 directives. (Ordinarily @emph{all} @option{-I} directories are used
7702 In addition, the @option{-I-} option inhibits the use of the current
7703 directory (where the current input file came from) as the first search
7704 directory for @samp{#include "@var{file}"}. There is no way to
7705 override this effect of @option{-I-}. With @option{-I.} you can specify
7706 searching the directory which was current when the compiler was
7707 invoked. That is not exactly the same as what the preprocessor does
7708 by default, but it is often satisfactory.
7710 @option{-I-} does not inhibit the use of the standard system directories
7711 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7718 @section Specifying subprocesses and the switches to pass to them
7721 @command{gcc} is a driver program. It performs its job by invoking a
7722 sequence of other programs to do the work of compiling, assembling and
7723 linking. GCC interprets its command-line parameters and uses these to
7724 deduce which programs it should invoke, and which command-line options
7725 it ought to place on their command lines. This behavior is controlled
7726 by @dfn{spec strings}. In most cases there is one spec string for each
7727 program that GCC can invoke, but a few programs have multiple spec
7728 strings to control their behavior. The spec strings built into GCC can
7729 be overridden by using the @option{-specs=} command-line switch to specify
7732 @dfn{Spec files} are plaintext files that are used to construct spec
7733 strings. They consist of a sequence of directives separated by blank
7734 lines. The type of directive is determined by the first non-whitespace
7735 character on the line and it can be one of the following:
7738 @item %@var{command}
7739 Issues a @var{command} to the spec file processor. The commands that can
7743 @item %include <@var{file}>
7745 Search for @var{file} and insert its text at the current point in the
7748 @item %include_noerr <@var{file}>
7749 @cindex %include_noerr
7750 Just like @samp{%include}, but do not generate an error message if the include
7751 file cannot be found.
7753 @item %rename @var{old_name} @var{new_name}
7755 Rename the spec string @var{old_name} to @var{new_name}.
7759 @item *[@var{spec_name}]:
7760 This tells the compiler to create, override or delete the named spec
7761 string. All lines after this directive up to the next directive or
7762 blank line are considered to be the text for the spec string. If this
7763 results in an empty string then the spec will be deleted. (Or, if the
7764 spec did not exist, then nothing will happened.) Otherwise, if the spec
7765 does not currently exist a new spec will be created. If the spec does
7766 exist then its contents will be overridden by the text of this
7767 directive, unless the first character of that text is the @samp{+}
7768 character, in which case the text will be appended to the spec.
7770 @item [@var{suffix}]:
7771 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7772 and up to the next directive or blank line are considered to make up the
7773 spec string for the indicated suffix. When the compiler encounters an
7774 input file with the named suffix, it will processes the spec string in
7775 order to work out how to compile that file. For example:
7782 This says that any input file whose name ends in @samp{.ZZ} should be
7783 passed to the program @samp{z-compile}, which should be invoked with the
7784 command-line switch @option{-input} and with the result of performing the
7785 @samp{%i} substitution. (See below.)
7787 As an alternative to providing a spec string, the text that follows a
7788 suffix directive can be one of the following:
7791 @item @@@var{language}
7792 This says that the suffix is an alias for a known @var{language}. This is
7793 similar to using the @option{-x} command-line switch to GCC to specify a
7794 language explicitly. For example:
7801 Says that .ZZ files are, in fact, C++ source files.
7804 This causes an error messages saying:
7807 @var{name} compiler not installed on this system.
7811 GCC already has an extensive list of suffixes built into it.
7812 This directive will add an entry to the end of the list of suffixes, but
7813 since the list is searched from the end backwards, it is effectively
7814 possible to override earlier entries using this technique.
7818 GCC has the following spec strings built into it. Spec files can
7819 override these strings or create their own. Note that individual
7820 targets can also add their own spec strings to this list.
7823 asm Options to pass to the assembler
7824 asm_final Options to pass to the assembler post-processor
7825 cpp Options to pass to the C preprocessor
7826 cc1 Options to pass to the C compiler
7827 cc1plus Options to pass to the C++ compiler
7828 endfile Object files to include at the end of the link
7829 link Options to pass to the linker
7830 lib Libraries to include on the command line to the linker
7831 libgcc Decides which GCC support library to pass to the linker
7832 linker Sets the name of the linker
7833 predefines Defines to be passed to the C preprocessor
7834 signed_char Defines to pass to CPP to say whether @code{char} is signed
7836 startfile Object files to include at the start of the link
7839 Here is a small example of a spec file:
7845 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7848 This example renames the spec called @samp{lib} to @samp{old_lib} and
7849 then overrides the previous definition of @samp{lib} with a new one.
7850 The new definition adds in some extra command-line options before
7851 including the text of the old definition.
7853 @dfn{Spec strings} are a list of command-line options to be passed to their
7854 corresponding program. In addition, the spec strings can contain
7855 @samp{%}-prefixed sequences to substitute variable text or to
7856 conditionally insert text into the command line. Using these constructs
7857 it is possible to generate quite complex command lines.
7859 Here is a table of all defined @samp{%}-sequences for spec
7860 strings. Note that spaces are not generated automatically around the
7861 results of expanding these sequences. Therefore you can concatenate them
7862 together or combine them with constant text in a single argument.
7866 Substitute one @samp{%} into the program name or argument.
7869 Substitute the name of the input file being processed.
7872 Substitute the basename of the input file being processed.
7873 This is the substring up to (and not including) the last period
7874 and not including the directory.
7877 This is the same as @samp{%b}, but include the file suffix (text after
7881 Marks the argument containing or following the @samp{%d} as a
7882 temporary file name, so that that file will be deleted if GCC exits
7883 successfully. Unlike @samp{%g}, this contributes no text to the
7886 @item %g@var{suffix}
7887 Substitute a file name that has suffix @var{suffix} and is chosen
7888 once per compilation, and mark the argument in the same way as
7889 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7890 name is now chosen in a way that is hard to predict even when previously
7891 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7892 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7893 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7894 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7895 was simply substituted with a file name chosen once per compilation,
7896 without regard to any appended suffix (which was therefore treated
7897 just like ordinary text), making such attacks more likely to succeed.
7899 @item %u@var{suffix}
7900 Like @samp{%g}, but generates a new temporary file name even if
7901 @samp{%u@var{suffix}} was already seen.
7903 @item %U@var{suffix}
7904 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7905 new one if there is no such last file name. In the absence of any
7906 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7907 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7908 would involve the generation of two distinct file names, one
7909 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7910 simply substituted with a file name chosen for the previous @samp{%u},
7911 without regard to any appended suffix.
7913 @item %j@var{suffix}
7914 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7915 writable, and if save-temps is off; otherwise, substitute the name
7916 of a temporary file, just like @samp{%u}. This temporary file is not
7917 meant for communication between processes, but rather as a junk
7920 @item %|@var{suffix}
7921 @itemx %m@var{suffix}
7922 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7923 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7924 all. These are the two most common ways to instruct a program that it
7925 should read from standard input or write to standard output. If you
7926 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7927 construct: see for example @file{f/lang-specs.h}.
7929 @item %.@var{SUFFIX}
7930 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7931 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7932 terminated by the next space or %.
7935 Marks the argument containing or following the @samp{%w} as the
7936 designated output file of this compilation. This puts the argument
7937 into the sequence of arguments that @samp{%o} will substitute later.
7940 Substitutes the names of all the output files, with spaces
7941 automatically placed around them. You should write spaces
7942 around the @samp{%o} as well or the results are undefined.
7943 @samp{%o} is for use in the specs for running the linker.
7944 Input files whose names have no recognized suffix are not compiled
7945 at all, but they are included among the output files, so they will
7949 Substitutes the suffix for object files. Note that this is
7950 handled specially when it immediately follows @samp{%g, %u, or %U},
7951 because of the need for those to form complete file names. The
7952 handling is such that @samp{%O} is treated exactly as if it had already
7953 been substituted, except that @samp{%g, %u, and %U} do not currently
7954 support additional @var{suffix} characters following @samp{%O} as they would
7955 following, for example, @samp{.o}.
7958 Substitutes the standard macro predefinitions for the
7959 current target machine. Use this when running @code{cpp}.
7962 Like @samp{%p}, but puts @samp{__} before and after the name of each
7963 predefined macro, except for macros that start with @samp{__} or with
7964 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
7968 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
7969 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
7970 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
7971 and @option{-imultilib} as necessary.
7974 Current argument is the name of a library or startup file of some sort.
7975 Search for that file in a standard list of directories and substitute
7976 the full name found.
7979 Print @var{str} as an error message. @var{str} is terminated by a newline.
7980 Use this when inconsistent options are detected.
7983 Substitute the contents of spec string @var{name} at this point.
7986 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
7988 @item %x@{@var{option}@}
7989 Accumulate an option for @samp{%X}.
7992 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
7996 Output the accumulated assembler options specified by @option{-Wa}.
7999 Output the accumulated preprocessor options specified by @option{-Wp}.
8002 Process the @code{asm} spec. This is used to compute the
8003 switches to be passed to the assembler.
8006 Process the @code{asm_final} spec. This is a spec string for
8007 passing switches to an assembler post-processor, if such a program is
8011 Process the @code{link} spec. This is the spec for computing the
8012 command line passed to the linker. Typically it will make use of the
8013 @samp{%L %G %S %D and %E} sequences.
8016 Dump out a @option{-L} option for each directory that GCC believes might
8017 contain startup files. If the target supports multilibs then the
8018 current multilib directory will be prepended to each of these paths.
8021 Process the @code{lib} spec. This is a spec string for deciding which
8022 libraries should be included on the command line to the linker.
8025 Process the @code{libgcc} spec. This is a spec string for deciding
8026 which GCC support library should be included on the command line to the linker.
8029 Process the @code{startfile} spec. This is a spec for deciding which
8030 object files should be the first ones passed to the linker. Typically
8031 this might be a file named @file{crt0.o}.
8034 Process the @code{endfile} spec. This is a spec string that specifies
8035 the last object files that will be passed to the linker.
8038 Process the @code{cpp} spec. This is used to construct the arguments
8039 to be passed to the C preprocessor.
8042 Process the @code{cc1} spec. This is used to construct the options to be
8043 passed to the actual C compiler (@samp{cc1}).
8046 Process the @code{cc1plus} spec. This is used to construct the options to be
8047 passed to the actual C++ compiler (@samp{cc1plus}).
8050 Substitute the variable part of a matched option. See below.
8051 Note that each comma in the substituted string is replaced by
8055 Remove all occurrences of @code{-S} from the command line. Note---this
8056 command is position dependent. @samp{%} commands in the spec string
8057 before this one will see @code{-S}, @samp{%} commands in the spec string
8058 after this one will not.
8060 @item %:@var{function}(@var{args})
8061 Call the named function @var{function}, passing it @var{args}.
8062 @var{args} is first processed as a nested spec string, then split
8063 into an argument vector in the usual fashion. The function returns
8064 a string which is processed as if it had appeared literally as part
8065 of the current spec.
8067 The following built-in spec functions are provided:
8071 The @code{getenv} spec function takes two arguments: an environment
8072 variable name and a string. If the environment variable is not
8073 defined, a fatal error is issued. Otherwise, the return value is the
8074 value of the environment variable concatenated with the string. For
8075 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8078 %:getenv(TOPDIR /include)
8081 expands to @file{/path/to/top/include}.
8083 @item @code{if-exists}
8084 The @code{if-exists} spec function takes one argument, an absolute
8085 pathname to a file. If the file exists, @code{if-exists} returns the
8086 pathname. Here is a small example of its usage:
8090 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8093 @item @code{if-exists-else}
8094 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8095 spec function, except that it takes two arguments. The first argument is
8096 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8097 returns the pathname. If it does not exist, it returns the second argument.
8098 This way, @code{if-exists-else} can be used to select one file or another,
8099 based on the existence of the first. Here is a small example of its usage:
8103 crt0%O%s %:if-exists(crti%O%s) \
8104 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8107 @item @code{replace-outfile}
8108 The @code{replace-outfile} spec function takes two arguments. It looks for the
8109 first argument in the outfiles array and replaces it with the second argument. Here
8110 is a small example of its usage:
8113 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8116 @item @code{print-asm-header}
8117 The @code{print-asm-header} function takes no arguments and simply
8118 prints a banner like:
8124 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8127 It is used to separate compiler options from assembler options
8128 in the @option{--target-help} output.
8132 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8133 If that switch was not specified, this substitutes nothing. Note that
8134 the leading dash is omitted when specifying this option, and it is
8135 automatically inserted if the substitution is performed. Thus the spec
8136 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8137 and would output the command line option @option{-foo}.
8139 @item %W@{@code{S}@}
8140 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8143 @item %@{@code{S}*@}
8144 Substitutes all the switches specified to GCC whose names start
8145 with @code{-S}, but which also take an argument. This is used for
8146 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8147 GCC considers @option{-o foo} as being
8148 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8149 text, including the space. Thus two arguments would be generated.
8151 @item %@{@code{S}*&@code{T}*@}
8152 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8153 (the order of @code{S} and @code{T} in the spec is not significant).
8154 There can be any number of ampersand-separated variables; for each the
8155 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8157 @item %@{@code{S}:@code{X}@}
8158 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8160 @item %@{!@code{S}:@code{X}@}
8161 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8163 @item %@{@code{S}*:@code{X}@}
8164 Substitutes @code{X} if one or more switches whose names start with
8165 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8166 once, no matter how many such switches appeared. However, if @code{%*}
8167 appears somewhere in @code{X}, then @code{X} will be substituted once
8168 for each matching switch, with the @code{%*} replaced by the part of
8169 that switch that matched the @code{*}.
8171 @item %@{.@code{S}:@code{X}@}
8172 Substitutes @code{X}, if processing a file with suffix @code{S}.
8174 @item %@{!.@code{S}:@code{X}@}
8175 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8177 @item %@{,@code{S}:@code{X}@}
8178 Substitutes @code{X}, if processing a file for language @code{S}.
8180 @item %@{!,@code{S}:@code{X}@}
8181 Substitutes @code{X}, if not processing a file for language @code{S}.
8183 @item %@{@code{S}|@code{P}:@code{X}@}
8184 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8185 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8186 @code{*} sequences as well, although they have a stronger binding than
8187 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8188 alternatives must be starred, and only the first matching alternative
8191 For example, a spec string like this:
8194 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8197 will output the following command-line options from the following input
8198 command-line options:
8203 -d fred.c -foo -baz -boggle
8204 -d jim.d -bar -baz -boggle
8207 @item %@{S:X; T:Y; :D@}
8209 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8210 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8211 be as many clauses as you need. This may be combined with @code{.},
8212 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8217 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8218 construct may contain other nested @samp{%} constructs or spaces, or
8219 even newlines. They are processed as usual, as described above.
8220 Trailing white space in @code{X} is ignored. White space may also
8221 appear anywhere on the left side of the colon in these constructs,
8222 except between @code{.} or @code{*} and the corresponding word.
8224 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8225 handled specifically in these constructs. If another value of
8226 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8227 @option{-W} switch is found later in the command line, the earlier
8228 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8229 just one letter, which passes all matching options.
8231 The character @samp{|} at the beginning of the predicate text is used to
8232 indicate that a command should be piped to the following command, but
8233 only if @option{-pipe} is specified.
8235 It is built into GCC which switches take arguments and which do not.
8236 (You might think it would be useful to generalize this to allow each
8237 compiler's spec to say which switches take arguments. But this cannot
8238 be done in a consistent fashion. GCC cannot even decide which input
8239 files have been specified without knowing which switches take arguments,
8240 and it must know which input files to compile in order to tell which
8243 GCC also knows implicitly that arguments starting in @option{-l} are to be
8244 treated as compiler output files, and passed to the linker in their
8245 proper position among the other output files.
8247 @c man begin OPTIONS
8249 @node Target Options
8250 @section Specifying Target Machine and Compiler Version
8251 @cindex target options
8252 @cindex cross compiling
8253 @cindex specifying machine version
8254 @cindex specifying compiler version and target machine
8255 @cindex compiler version, specifying
8256 @cindex target machine, specifying
8258 The usual way to run GCC is to run the executable called @file{gcc}, or
8259 @file{<machine>-gcc} when cross-compiling, or
8260 @file{<machine>-gcc-<version>} to run a version other than the one that
8261 was installed last. Sometimes this is inconvenient, so GCC provides
8262 options that will switch to another cross-compiler or version.
8265 @item -b @var{machine}
8267 The argument @var{machine} specifies the target machine for compilation.
8269 The value to use for @var{machine} is the same as was specified as the
8270 machine type when configuring GCC as a cross-compiler. For
8271 example, if a cross-compiler was configured with @samp{configure
8272 arm-elf}, meaning to compile for an arm processor with elf binaries,
8273 then you would specify @option{-b arm-elf} to run that cross compiler.
8274 Because there are other options beginning with @option{-b}, the
8275 configuration must contain a hyphen.
8277 @item -V @var{version}
8279 The argument @var{version} specifies which version of GCC to run.
8280 This is useful when multiple versions are installed. For example,
8281 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8284 The @option{-V} and @option{-b} options work by running the
8285 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8286 use them if you can just run that directly.
8288 @node Submodel Options
8289 @section Hardware Models and Configurations
8290 @cindex submodel options
8291 @cindex specifying hardware config
8292 @cindex hardware models and configurations, specifying
8293 @cindex machine dependent options
8295 Earlier we discussed the standard option @option{-b} which chooses among
8296 different installed compilers for completely different target
8297 machines, such as VAX vs.@: 68000 vs.@: 80386.
8299 In addition, each of these target machine types can have its own
8300 special options, starting with @samp{-m}, to choose among various
8301 hardware models or configurations---for example, 68010 vs 68020,
8302 floating coprocessor or none. A single installed version of the
8303 compiler can compile for any model or configuration, according to the
8306 Some configurations of the compiler also support additional special
8307 options, usually for compatibility with other compilers on the same
8310 @c This list is ordered alphanumerically by subsection name.
8311 @c It should be the same order and spelling as these options are listed
8312 @c in Machine Dependent Options
8318 * Blackfin Options::
8322 * DEC Alpha Options::
8323 * DEC Alpha/VMS Options::
8325 * GNU/Linux Options::
8328 * i386 and x86-64 Options::
8341 * RS/6000 and PowerPC Options::
8342 * S/390 and zSeries Options::
8347 * System V Options::
8352 * Xstormy16 Options::
8358 @subsection ARC Options
8361 These options are defined for ARC implementations:
8366 Compile code for little endian mode. This is the default.
8370 Compile code for big endian mode.
8373 @opindex mmangle-cpu
8374 Prepend the name of the cpu to all public symbol names.
8375 In multiple-processor systems, there are many ARC variants with different
8376 instruction and register set characteristics. This flag prevents code
8377 compiled for one cpu to be linked with code compiled for another.
8378 No facility exists for handling variants that are ``almost identical''.
8379 This is an all or nothing option.
8381 @item -mcpu=@var{cpu}
8383 Compile code for ARC variant @var{cpu}.
8384 Which variants are supported depend on the configuration.
8385 All variants support @option{-mcpu=base}, this is the default.
8387 @item -mtext=@var{text-section}
8388 @itemx -mdata=@var{data-section}
8389 @itemx -mrodata=@var{readonly-data-section}
8393 Put functions, data, and readonly data in @var{text-section},
8394 @var{data-section}, and @var{readonly-data-section} respectively
8395 by default. This can be overridden with the @code{section} attribute.
8396 @xref{Variable Attributes}.
8401 @subsection ARM Options
8404 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8408 @item -mabi=@var{name}
8410 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8411 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8414 @opindex mapcs-frame
8415 Generate a stack frame that is compliant with the ARM Procedure Call
8416 Standard for all functions, even if this is not strictly necessary for
8417 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8418 with this option will cause the stack frames not to be generated for
8419 leaf functions. The default is @option{-mno-apcs-frame}.
8423 This is a synonym for @option{-mapcs-frame}.
8426 @c not currently implemented
8427 @item -mapcs-stack-check
8428 @opindex mapcs-stack-check
8429 Generate code to check the amount of stack space available upon entry to
8430 every function (that actually uses some stack space). If there is
8431 insufficient space available then either the function
8432 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8433 called, depending upon the amount of stack space required. The run time
8434 system is required to provide these functions. The default is
8435 @option{-mno-apcs-stack-check}, since this produces smaller code.
8437 @c not currently implemented
8439 @opindex mapcs-float
8440 Pass floating point arguments using the float point registers. This is
8441 one of the variants of the APCS@. This option is recommended if the
8442 target hardware has a floating point unit or if a lot of floating point
8443 arithmetic is going to be performed by the code. The default is
8444 @option{-mno-apcs-float}, since integer only code is slightly increased in
8445 size if @option{-mapcs-float} is used.
8447 @c not currently implemented
8448 @item -mapcs-reentrant
8449 @opindex mapcs-reentrant
8450 Generate reentrant, position independent code. The default is
8451 @option{-mno-apcs-reentrant}.
8454 @item -mthumb-interwork
8455 @opindex mthumb-interwork
8456 Generate code which supports calling between the ARM and Thumb
8457 instruction sets. Without this option the two instruction sets cannot
8458 be reliably used inside one program. The default is
8459 @option{-mno-thumb-interwork}, since slightly larger code is generated
8460 when @option{-mthumb-interwork} is specified.
8462 @item -mno-sched-prolog
8463 @opindex mno-sched-prolog
8464 Prevent the reordering of instructions in the function prolog, or the
8465 merging of those instruction with the instructions in the function's
8466 body. This means that all functions will start with a recognizable set
8467 of instructions (or in fact one of a choice from a small set of
8468 different function prologues), and this information can be used to
8469 locate the start if functions inside an executable piece of code. The
8470 default is @option{-msched-prolog}.
8473 @opindex mhard-float
8474 Generate output containing floating point instructions. This is the
8478 @opindex msoft-float
8479 Generate output containing library calls for floating point.
8480 @strong{Warning:} the requisite libraries are not available for all ARM
8481 targets. Normally the facilities of the machine's usual C compiler are
8482 used, but this cannot be done directly in cross-compilation. You must make
8483 your own arrangements to provide suitable library functions for
8486 @option{-msoft-float} changes the calling convention in the output file;
8487 therefore, it is only useful if you compile @emph{all} of a program with
8488 this option. In particular, you need to compile @file{libgcc.a}, the
8489 library that comes with GCC, with @option{-msoft-float} in order for
8492 @item -mfloat-abi=@var{name}
8494 Specifies which ABI to use for floating point values. Permissible values
8495 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8497 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
8498 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
8499 of floating point instructions, but still uses the soft-float calling
8502 @item -mlittle-endian
8503 @opindex mlittle-endian
8504 Generate code for a processor running in little-endian mode. This is
8505 the default for all standard configurations.
8508 @opindex mbig-endian
8509 Generate code for a processor running in big-endian mode; the default is
8510 to compile code for a little-endian processor.
8512 @item -mwords-little-endian
8513 @opindex mwords-little-endian
8514 This option only applies when generating code for big-endian processors.
8515 Generate code for a little-endian word order but a big-endian byte
8516 order. That is, a byte order of the form @samp{32107654}. Note: this
8517 option should only be used if you require compatibility with code for
8518 big-endian ARM processors generated by versions of the compiler prior to
8521 @item -mcpu=@var{name}
8523 This specifies the name of the target ARM processor. GCC uses this name
8524 to determine what kind of instructions it can emit when generating
8525 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8526 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8527 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8528 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8529 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8530 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8531 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8532 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8533 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8534 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8535 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8536 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8537 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8538 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8539 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3}, @samp{cortex-m1},
8540 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8542 @itemx -mtune=@var{name}
8544 This option is very similar to the @option{-mcpu=} option, except that
8545 instead of specifying the actual target processor type, and hence
8546 restricting which instructions can be used, it specifies that GCC should
8547 tune the performance of the code as if the target were of the type
8548 specified in this option, but still choosing the instructions that it
8549 will generate based on the cpu specified by a @option{-mcpu=} option.
8550 For some ARM implementations better performance can be obtained by using
8553 @item -march=@var{name}
8555 This specifies the name of the target ARM architecture. GCC uses this
8556 name to determine what kind of instructions it can emit when generating
8557 assembly code. This option can be used in conjunction with or instead
8558 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8559 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8560 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8561 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
8562 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
8563 @samp{iwmmxt}, @samp{ep9312}.
8565 @item -mfpu=@var{name}
8566 @itemx -mfpe=@var{number}
8567 @itemx -mfp=@var{number}
8571 This specifies what floating point hardware (or hardware emulation) is
8572 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8573 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
8574 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8575 with older versions of GCC@.
8577 If @option{-msoft-float} is specified this specifies the format of
8578 floating point values.
8580 @item -mstructure-size-boundary=@var{n}
8581 @opindex mstructure-size-boundary
8582 The size of all structures and unions will be rounded up to a multiple
8583 of the number of bits set by this option. Permissible values are 8, 32
8584 and 64. The default value varies for different toolchains. For the COFF
8585 targeted toolchain the default value is 8. A value of 64 is only allowed
8586 if the underlying ABI supports it.
8588 Specifying the larger number can produce faster, more efficient code, but
8589 can also increase the size of the program. Different values are potentially
8590 incompatible. Code compiled with one value cannot necessarily expect to
8591 work with code or libraries compiled with another value, if they exchange
8592 information using structures or unions.
8594 @item -mabort-on-noreturn
8595 @opindex mabort-on-noreturn
8596 Generate a call to the function @code{abort} at the end of a
8597 @code{noreturn} function. It will be executed if the function tries to
8601 @itemx -mno-long-calls
8602 @opindex mlong-calls
8603 @opindex mno-long-calls
8604 Tells the compiler to perform function calls by first loading the
8605 address of the function into a register and then performing a subroutine
8606 call on this register. This switch is needed if the target function
8607 will lie outside of the 64 megabyte addressing range of the offset based
8608 version of subroutine call instruction.
8610 Even if this switch is enabled, not all function calls will be turned
8611 into long calls. The heuristic is that static functions, functions
8612 which have the @samp{short-call} attribute, functions that are inside
8613 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8614 definitions have already been compiled within the current compilation
8615 unit, will not be turned into long calls. The exception to this rule is
8616 that weak function definitions, functions with the @samp{long-call}
8617 attribute or the @samp{section} attribute, and functions that are within
8618 the scope of a @samp{#pragma long_calls} directive, will always be
8619 turned into long calls.
8621 This feature is not enabled by default. Specifying
8622 @option{-mno-long-calls} will restore the default behavior, as will
8623 placing the function calls within the scope of a @samp{#pragma
8624 long_calls_off} directive. Note these switches have no effect on how
8625 the compiler generates code to handle function calls via function
8628 @item -mnop-fun-dllimport
8629 @opindex mnop-fun-dllimport
8630 Disable support for the @code{dllimport} attribute.
8632 @item -msingle-pic-base
8633 @opindex msingle-pic-base
8634 Treat the register used for PIC addressing as read-only, rather than
8635 loading it in the prologue for each function. The run-time system is
8636 responsible for initializing this register with an appropriate value
8637 before execution begins.
8639 @item -mpic-register=@var{reg}
8640 @opindex mpic-register
8641 Specify the register to be used for PIC addressing. The default is R10
8642 unless stack-checking is enabled, when R9 is used.
8644 @item -mcirrus-fix-invalid-insns
8645 @opindex mcirrus-fix-invalid-insns
8646 @opindex mno-cirrus-fix-invalid-insns
8647 Insert NOPs into the instruction stream to in order to work around
8648 problems with invalid Maverick instruction combinations. This option
8649 is only valid if the @option{-mcpu=ep9312} option has been used to
8650 enable generation of instructions for the Cirrus Maverick floating
8651 point co-processor. This option is not enabled by default, since the
8652 problem is only present in older Maverick implementations. The default
8653 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8656 @item -mpoke-function-name
8657 @opindex mpoke-function-name
8658 Write the name of each function into the text section, directly
8659 preceding the function prologue. The generated code is similar to this:
8663 .ascii "arm_poke_function_name", 0
8666 .word 0xff000000 + (t1 - t0)
8667 arm_poke_function_name
8669 stmfd sp!, @{fp, ip, lr, pc@}
8673 When performing a stack backtrace, code can inspect the value of
8674 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8675 location @code{pc - 12} and the top 8 bits are set, then we know that
8676 there is a function name embedded immediately preceding this location
8677 and has length @code{((pc[-3]) & 0xff000000)}.
8681 Generate code for the Thumb instruction set. The default is to
8682 use the 32-bit ARM instruction set.
8683 This option automatically enables either 16-bit Thumb-1 or
8684 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8685 and @option{-march=@var{name}} options.
8688 @opindex mtpcs-frame
8689 Generate a stack frame that is compliant with the Thumb Procedure Call
8690 Standard for all non-leaf functions. (A leaf function is one that does
8691 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8693 @item -mtpcs-leaf-frame
8694 @opindex mtpcs-leaf-frame
8695 Generate a stack frame that is compliant with the Thumb Procedure Call
8696 Standard for all leaf functions. (A leaf function is one that does
8697 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8699 @item -mcallee-super-interworking
8700 @opindex mcallee-super-interworking
8701 Gives all externally visible functions in the file being compiled an ARM
8702 instruction set header which switches to Thumb mode before executing the
8703 rest of the function. This allows these functions to be called from
8704 non-interworking code.
8706 @item -mcaller-super-interworking
8707 @opindex mcaller-super-interworking
8708 Allows calls via function pointers (including virtual functions) to
8709 execute correctly regardless of whether the target code has been
8710 compiled for interworking or not. There is a small overhead in the cost
8711 of executing a function pointer if this option is enabled.
8713 @item -mtp=@var{name}
8715 Specify the access model for the thread local storage pointer. The valid
8716 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8717 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8718 (supported in the arm6k architecture), and @option{auto}, which uses the
8719 best available method for the selected processor. The default setting is
8725 @subsection AVR Options
8728 These options are defined for AVR implementations:
8731 @item -mmcu=@var{mcu}
8733 Specify ATMEL AVR instruction set or MCU type.
8735 Instruction set avr1 is for the minimal AVR core, not supported by the C
8736 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8737 attiny11, attiny12, attiny15, attiny28).
8739 Instruction set avr2 (default) is for the classic AVR core with up to
8740 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8741 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8742 at90c8534, at90s8535).
8744 Instruction set avr3 is for the classic AVR core with up to 128K program
8745 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8747 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8748 memory space (MCU types: atmega8, atmega83, atmega85).
8750 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8751 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8752 atmega64, atmega128, at43usb355, at94k).
8756 Output instruction sizes to the asm file.
8758 @item -minit-stack=@var{N}
8759 @opindex minit-stack
8760 Specify the initial stack address, which may be a symbol or numeric value,
8761 @samp{__stack} is the default.
8763 @item -mno-interrupts
8764 @opindex mno-interrupts
8765 Generated code is not compatible with hardware interrupts.
8766 Code size will be smaller.
8768 @item -mcall-prologues
8769 @opindex mcall-prologues
8770 Functions prologues/epilogues expanded as call to appropriate
8771 subroutines. Code size will be smaller.
8773 @item -mno-tablejump
8774 @opindex mno-tablejump
8775 Do not generate tablejump insns which sometimes increase code size.
8778 @opindex mtiny-stack
8779 Change only the low 8 bits of the stack pointer.
8783 Assume int to be 8 bit integer. This affects the sizes of all types: A
8784 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8785 and long long will be 4 bytes. Please note that this option does not
8786 comply to the C standards, but it will provide you with smaller code
8790 @node Blackfin Options
8791 @subsection Blackfin Options
8792 @cindex Blackfin Options
8795 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
8797 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
8798 can be one of @samp{bf522}, @samp{bf523}, @samp{bf524},
8799 @samp{bf525}, @samp{bf526}, @samp{bf527},
8800 @samp{bf531}, @samp{bf532}, @samp{bf533}, @samp{bf534},
8801 @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
8802 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
8804 The optional @var{sirevision} specifies the silicon revision of the target
8805 Blackfin processor. Any workarounds available for the targeted silicon revision
8806 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
8807 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
8808 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
8809 hexadecimal digits representing the major and minor numbers in the silicon
8810 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
8811 is not defined. If @var{sirevision} is @samp{any}, the
8812 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
8813 If this optional @var{sirevision} is not used, GCC assumes the latest known
8814 silicon revision of the targeted Blackfin processor.
8816 Support for @samp{bf561} is incomplete. For @samp{bf561},
8817 Only the processor macro is defined.
8818 Without this option, @samp{bf532} is used as the processor by default.
8819 The corresponding predefined processor macros for @var{cpu} is to
8820 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
8821 provided by libgloss to be linked in if @option{-msim} is not given.
8825 Specifies that the program will be run on the simulator. This causes
8826 the simulator BSP provided by libgloss to be linked in. This option
8827 has effect only for @samp{bfin-elf} toolchain.
8828 Certain other options, such as @option{-mid-shared-library} and
8829 @option{-mfdpic}, imply @option{-msim}.
8831 @item -momit-leaf-frame-pointer
8832 @opindex momit-leaf-frame-pointer
8833 Don't keep the frame pointer in a register for leaf functions. This
8834 avoids the instructions to save, set up and restore frame pointers and
8835 makes an extra register available in leaf functions. The option
8836 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8837 which might make debugging harder.
8839 @item -mspecld-anomaly
8840 @opindex mspecld-anomaly
8841 When enabled, the compiler will ensure that the generated code does not
8842 contain speculative loads after jump instructions. If this option is used,
8843 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
8845 @item -mno-specld-anomaly
8846 @opindex mno-specld-anomaly
8847 Don't generate extra code to prevent speculative loads from occurring.
8849 @item -mcsync-anomaly
8850 @opindex mcsync-anomaly
8851 When enabled, the compiler will ensure that the generated code does not
8852 contain CSYNC or SSYNC instructions too soon after conditional branches.
8853 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
8855 @item -mno-csync-anomaly
8856 @opindex mno-csync-anomaly
8857 Don't generate extra code to prevent CSYNC or SSYNC instructions from
8858 occurring too soon after a conditional branch.
8862 When enabled, the compiler is free to take advantage of the knowledge that
8863 the entire program fits into the low 64k of memory.
8866 @opindex mno-low-64k
8867 Assume that the program is arbitrarily large. This is the default.
8869 @item -mstack-check-l1
8870 @opindex mstack-check-l1
8871 Do stack checking using information placed into L1 scratchpad memory by the
8874 @item -mid-shared-library
8875 @opindex mid-shared-library
8876 Generate code that supports shared libraries via the library ID method.
8877 This allows for execute in place and shared libraries in an environment
8878 without virtual memory management. This option implies @option{-fPIC}.
8879 With a @samp{bfin-elf} target, this option implies @option{-msim}.
8881 @item -mno-id-shared-library
8882 @opindex mno-id-shared-library
8883 Generate code that doesn't assume ID based shared libraries are being used.
8884 This is the default.
8886 @item -mleaf-id-shared-library
8887 @opindex mleaf-id-shared-library
8888 Generate code that supports shared libraries via the library ID method,
8889 but assumes that this library or executable won't link against any other
8890 ID shared libraries. That allows the compiler to use faster code for jumps
8893 @item -mno-leaf-id-shared-library
8894 @opindex mno-leaf-id-shared-library
8895 Do not assume that the code being compiled won't link against any ID shared
8896 libraries. Slower code will be generated for jump and call insns.
8898 @item -mshared-library-id=n
8899 @opindex mshared-library-id
8900 Specified the identification number of the ID based shared library being
8901 compiled. Specifying a value of 0 will generate more compact code, specifying
8902 other values will force the allocation of that number to the current
8903 library but is no more space or time efficient than omitting this option.
8907 Generate code that allows the data segment to be located in a different
8908 area of memory from the text segment. This allows for execute in place in
8909 an environment without virtual memory management by eliminating relocations
8910 against the text section.
8913 @opindex mno-sep-data
8914 Generate code that assumes that the data segment follows the text segment.
8915 This is the default.
8918 @itemx -mno-long-calls
8919 @opindex mlong-calls
8920 @opindex mno-long-calls
8921 Tells the compiler to perform function calls by first loading the
8922 address of the function into a register and then performing a subroutine
8923 call on this register. This switch is needed if the target function
8924 will lie outside of the 24 bit addressing range of the offset based
8925 version of subroutine call instruction.
8927 This feature is not enabled by default. Specifying
8928 @option{-mno-long-calls} will restore the default behavior. Note these
8929 switches have no effect on how the compiler generates code to handle
8930 function calls via function pointers.
8934 Link with the fast floating-point library. This library relaxes some of
8935 the IEEE floating-point standard's rules for checking inputs against
8936 Not-a-Number (NAN), in the interest of performance.
8939 @opindex minline-plt
8940 Enable inlining of PLT entries in function calls to functions that are
8941 not known to bind locally. It has no effect without @option{-mfdpic}.
8945 @subsection CRIS Options
8946 @cindex CRIS Options
8948 These options are defined specifically for the CRIS ports.
8951 @item -march=@var{architecture-type}
8952 @itemx -mcpu=@var{architecture-type}
8955 Generate code for the specified architecture. The choices for
8956 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
8957 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
8958 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
8961 @item -mtune=@var{architecture-type}
8963 Tune to @var{architecture-type} everything applicable about the generated
8964 code, except for the ABI and the set of available instructions. The
8965 choices for @var{architecture-type} are the same as for
8966 @option{-march=@var{architecture-type}}.
8968 @item -mmax-stack-frame=@var{n}
8969 @opindex mmax-stack-frame
8970 Warn when the stack frame of a function exceeds @var{n} bytes.
8972 @item -melinux-stacksize=@var{n}
8973 @opindex melinux-stacksize
8974 Only available with the @samp{cris-axis-aout} target. Arranges for
8975 indications in the program to the kernel loader that the stack of the
8976 program should be set to @var{n} bytes.
8982 The options @option{-metrax4} and @option{-metrax100} are synonyms for
8983 @option{-march=v3} and @option{-march=v8} respectively.
8985 @item -mmul-bug-workaround
8986 @itemx -mno-mul-bug-workaround
8987 @opindex mmul-bug-workaround
8988 @opindex mno-mul-bug-workaround
8989 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
8990 models where it applies. This option is active by default.
8994 Enable CRIS-specific verbose debug-related information in the assembly
8995 code. This option also has the effect to turn off the @samp{#NO_APP}
8996 formatted-code indicator to the assembler at the beginning of the
9001 Do not use condition-code results from previous instruction; always emit
9002 compare and test instructions before use of condition codes.
9004 @item -mno-side-effects
9005 @opindex mno-side-effects
9006 Do not emit instructions with side-effects in addressing modes other than
9010 @itemx -mno-stack-align
9012 @itemx -mno-data-align
9013 @itemx -mconst-align
9014 @itemx -mno-const-align
9015 @opindex mstack-align
9016 @opindex mno-stack-align
9017 @opindex mdata-align
9018 @opindex mno-data-align
9019 @opindex mconst-align
9020 @opindex mno-const-align
9021 These options (no-options) arranges (eliminate arrangements) for the
9022 stack-frame, individual data and constants to be aligned for the maximum
9023 single data access size for the chosen CPU model. The default is to
9024 arrange for 32-bit alignment. ABI details such as structure layout are
9025 not affected by these options.
9033 Similar to the stack- data- and const-align options above, these options
9034 arrange for stack-frame, writable data and constants to all be 32-bit,
9035 16-bit or 8-bit aligned. The default is 32-bit alignment.
9037 @item -mno-prologue-epilogue
9038 @itemx -mprologue-epilogue
9039 @opindex mno-prologue-epilogue
9040 @opindex mprologue-epilogue
9041 With @option{-mno-prologue-epilogue}, the normal function prologue and
9042 epilogue that sets up the stack-frame are omitted and no return
9043 instructions or return sequences are generated in the code. Use this
9044 option only together with visual inspection of the compiled code: no
9045 warnings or errors are generated when call-saved registers must be saved,
9046 or storage for local variable needs to be allocated.
9052 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9053 instruction sequences that load addresses for functions from the PLT part
9054 of the GOT rather than (traditional on other architectures) calls to the
9055 PLT@. The default is @option{-mgotplt}.
9059 Legacy no-op option only recognized with the cris-axis-aout target.
9063 Legacy no-op option only recognized with the cris-axis-elf and
9064 cris-axis-linux-gnu targets.
9068 Only recognized with the cris-axis-aout target, where it selects a
9069 GNU/linux-like multilib, include files and instruction set for
9074 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9078 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
9079 to link with input-output functions from a simulator library. Code,
9080 initialized data and zero-initialized data are allocated consecutively.
9084 Like @option{-sim}, but pass linker options to locate initialized data at
9085 0x40000000 and zero-initialized data at 0x80000000.
9089 @subsection CRX Options
9092 These options are defined specifically for the CRX ports.
9098 Enable the use of multiply-accumulate instructions. Disabled by default.
9102 Push instructions will be used to pass outgoing arguments when functions
9103 are called. Enabled by default.
9106 @node Darwin Options
9107 @subsection Darwin Options
9108 @cindex Darwin options
9110 These options are defined for all architectures running the Darwin operating
9113 FSF GCC on Darwin does not create ``fat'' object files; it will create
9114 an object file for the single architecture that it was built to
9115 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9116 @option{-arch} options are used; it does so by running the compiler or
9117 linker multiple times and joining the results together with
9120 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9121 @samp{i686}) is determined by the flags that specify the ISA
9122 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9123 @option{-force_cpusubtype_ALL} option can be used to override this.
9125 The Darwin tools vary in their behavior when presented with an ISA
9126 mismatch. The assembler, @file{as}, will only permit instructions to
9127 be used that are valid for the subtype of the file it is generating,
9128 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9129 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9130 and print an error if asked to create a shared library with a less
9131 restrictive subtype than its input files (for instance, trying to put
9132 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9133 for executables, @file{ld}, will quietly give the executable the most
9134 restrictive subtype of any of its input files.
9139 Add the framework directory @var{dir} to the head of the list of
9140 directories to be searched for header files. These directories are
9141 interleaved with those specified by @option{-I} options and are
9142 scanned in a left-to-right order.
9144 A framework directory is a directory with frameworks in it. A
9145 framework is a directory with a @samp{"Headers"} and/or
9146 @samp{"PrivateHeaders"} directory contained directly in it that ends
9147 in @samp{".framework"}. The name of a framework is the name of this
9148 directory excluding the @samp{".framework"}. Headers associated with
9149 the framework are found in one of those two directories, with
9150 @samp{"Headers"} being searched first. A subframework is a framework
9151 directory that is in a framework's @samp{"Frameworks"} directory.
9152 Includes of subframework headers can only appear in a header of a
9153 framework that contains the subframework, or in a sibling subframework
9154 header. Two subframeworks are siblings if they occur in the same
9155 framework. A subframework should not have the same name as a
9156 framework, a warning will be issued if this is violated. Currently a
9157 subframework cannot have subframeworks, in the future, the mechanism
9158 may be extended to support this. The standard frameworks can be found
9159 in @samp{"/System/Library/Frameworks"} and
9160 @samp{"/Library/Frameworks"}. An example include looks like
9161 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9162 the name of the framework and header.h is found in the
9163 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9165 @item -iframework@var{dir}
9167 Like @option{-F} except the directory is a treated as a system
9168 directory. The main difference between this @option{-iframework} and
9169 @option{-F} is that with @option{-iframework} the compiler does not
9170 warn about constructs contained within header files found via
9171 @var{dir}. This option is valid only for the C family of languages.
9175 Emit debugging information for symbols that are used. For STABS
9176 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9177 This is by default ON@.
9181 Emit debugging information for all symbols and types.
9183 @item -mmacosx-version-min=@var{version}
9184 The earliest version of MacOS X that this executable will run on
9185 is @var{version}. Typical values of @var{version} include @code{10.1},
9186 @code{10.2}, and @code{10.3.9}.
9188 If the compiler was built to use the system's headers by default,
9189 then the default for this option is the system version on which the
9190 compiler is running, otherwise the default is to make choices which
9191 are compatible with as many systems and code bases as possible.
9195 Enable kernel development mode. The @option{-mkernel} option sets
9196 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9197 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9198 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9199 applicable. This mode also sets @option{-mno-altivec},
9200 @option{-msoft-float}, @option{-fno-builtin} and
9201 @option{-mlong-branch} for PowerPC targets.
9203 @item -mone-byte-bool
9204 @opindex mone-byte-bool
9205 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9206 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9207 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9208 option has no effect on x86.
9210 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9211 to generate code that is not binary compatible with code generated
9212 without that switch. Using this switch may require recompiling all
9213 other modules in a program, including system libraries. Use this
9214 switch to conform to a non-default data model.
9216 @item -mfix-and-continue
9217 @itemx -ffix-and-continue
9218 @itemx -findirect-data
9219 @opindex mfix-and-continue
9220 @opindex ffix-and-continue
9221 @opindex findirect-data
9222 Generate code suitable for fast turn around development. Needed to
9223 enable gdb to dynamically load @code{.o} files into already running
9224 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9225 are provided for backwards compatibility.
9229 Loads all members of static archive libraries.
9230 See man ld(1) for more information.
9232 @item -arch_errors_fatal
9233 @opindex arch_errors_fatal
9234 Cause the errors having to do with files that have the wrong architecture
9238 @opindex bind_at_load
9239 Causes the output file to be marked such that the dynamic linker will
9240 bind all undefined references when the file is loaded or launched.
9244 Produce a Mach-o bundle format file.
9245 See man ld(1) for more information.
9247 @item -bundle_loader @var{executable}
9248 @opindex bundle_loader
9249 This option specifies the @var{executable} that will be loading the build
9250 output file being linked. See man ld(1) for more information.
9254 When passed this option, GCC will produce a dynamic library instead of
9255 an executable when linking, using the Darwin @file{libtool} command.
9257 @item -force_cpusubtype_ALL
9258 @opindex force_cpusubtype_ALL
9259 This causes GCC's output file to have the @var{ALL} subtype, instead of
9260 one controlled by the @option{-mcpu} or @option{-march} option.
9262 @item -allowable_client @var{client_name}
9264 @itemx -compatibility_version
9265 @itemx -current_version
9267 @itemx -dependency-file
9269 @itemx -dylinker_install_name
9271 @itemx -exported_symbols_list
9273 @itemx -flat_namespace
9274 @itemx -force_flat_namespace
9275 @itemx -headerpad_max_install_names
9278 @itemx -install_name
9279 @itemx -keep_private_externs
9280 @itemx -multi_module
9281 @itemx -multiply_defined
9282 @itemx -multiply_defined_unused
9284 @itemx -no_dead_strip_inits_and_terms
9285 @itemx -nofixprebinding
9288 @itemx -noseglinkedit
9289 @itemx -pagezero_size
9291 @itemx -prebind_all_twolevel_modules
9292 @itemx -private_bundle
9293 @itemx -read_only_relocs
9295 @itemx -sectobjectsymbols
9299 @itemx -sectobjectsymbols
9302 @itemx -segs_read_only_addr
9303 @itemx -segs_read_write_addr
9304 @itemx -seg_addr_table
9305 @itemx -seg_addr_table_filename
9308 @itemx -segs_read_only_addr
9309 @itemx -segs_read_write_addr
9310 @itemx -single_module
9313 @itemx -sub_umbrella
9314 @itemx -twolevel_namespace
9317 @itemx -unexported_symbols_list
9318 @itemx -weak_reference_mismatches
9321 @opindex allowable_client
9322 @opindex client_name
9323 @opindex compatibility_version
9324 @opindex current_version
9326 @opindex dependency-file
9328 @opindex dylinker_install_name
9330 @opindex exported_symbols_list
9332 @opindex flat_namespace
9333 @opindex force_flat_namespace
9334 @opindex headerpad_max_install_names
9337 @opindex install_name
9338 @opindex keep_private_externs
9339 @opindex multi_module
9340 @opindex multiply_defined
9341 @opindex multiply_defined_unused
9343 @opindex no_dead_strip_inits_and_terms
9344 @opindex nofixprebinding
9345 @opindex nomultidefs
9347 @opindex noseglinkedit
9348 @opindex pagezero_size
9350 @opindex prebind_all_twolevel_modules
9351 @opindex private_bundle
9352 @opindex read_only_relocs
9354 @opindex sectobjectsymbols
9358 @opindex sectobjectsymbols
9361 @opindex segs_read_only_addr
9362 @opindex segs_read_write_addr
9363 @opindex seg_addr_table
9364 @opindex seg_addr_table_filename
9365 @opindex seglinkedit
9367 @opindex segs_read_only_addr
9368 @opindex segs_read_write_addr
9369 @opindex single_module
9371 @opindex sub_library
9372 @opindex sub_umbrella
9373 @opindex twolevel_namespace
9376 @opindex unexported_symbols_list
9377 @opindex weak_reference_mismatches
9378 @opindex whatsloaded
9380 These options are passed to the Darwin linker. The Darwin linker man page
9381 describes them in detail.
9384 @node DEC Alpha Options
9385 @subsection DEC Alpha Options
9387 These @samp{-m} options are defined for the DEC Alpha implementations:
9390 @item -mno-soft-float
9392 @opindex mno-soft-float
9393 @opindex msoft-float
9394 Use (do not use) the hardware floating-point instructions for
9395 floating-point operations. When @option{-msoft-float} is specified,
9396 functions in @file{libgcc.a} will be used to perform floating-point
9397 operations. Unless they are replaced by routines that emulate the
9398 floating-point operations, or compiled in such a way as to call such
9399 emulations routines, these routines will issue floating-point
9400 operations. If you are compiling for an Alpha without floating-point
9401 operations, you must ensure that the library is built so as not to call
9404 Note that Alpha implementations without floating-point operations are
9405 required to have floating-point registers.
9410 @opindex mno-fp-regs
9411 Generate code that uses (does not use) the floating-point register set.
9412 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9413 register set is not used, floating point operands are passed in integer
9414 registers as if they were integers and floating-point results are passed
9415 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9416 so any function with a floating-point argument or return value called by code
9417 compiled with @option{-mno-fp-regs} must also be compiled with that
9420 A typical use of this option is building a kernel that does not use,
9421 and hence need not save and restore, any floating-point registers.
9425 The Alpha architecture implements floating-point hardware optimized for
9426 maximum performance. It is mostly compliant with the IEEE floating
9427 point standard. However, for full compliance, software assistance is
9428 required. This option generates code fully IEEE compliant code
9429 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9430 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9431 defined during compilation. The resulting code is less efficient but is
9432 able to correctly support denormalized numbers and exceptional IEEE
9433 values such as not-a-number and plus/minus infinity. Other Alpha
9434 compilers call this option @option{-ieee_with_no_inexact}.
9436 @item -mieee-with-inexact
9437 @opindex mieee-with-inexact
9438 This is like @option{-mieee} except the generated code also maintains
9439 the IEEE @var{inexact-flag}. Turning on this option causes the
9440 generated code to implement fully-compliant IEEE math. In addition to
9441 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9442 macro. On some Alpha implementations the resulting code may execute
9443 significantly slower than the code generated by default. Since there is
9444 very little code that depends on the @var{inexact-flag}, you should
9445 normally not specify this option. Other Alpha compilers call this
9446 option @option{-ieee_with_inexact}.
9448 @item -mfp-trap-mode=@var{trap-mode}
9449 @opindex mfp-trap-mode
9450 This option controls what floating-point related traps are enabled.
9451 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9452 The trap mode can be set to one of four values:
9456 This is the default (normal) setting. The only traps that are enabled
9457 are the ones that cannot be disabled in software (e.g., division by zero
9461 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9465 Like @samp{u}, but the instructions are marked to be safe for software
9466 completion (see Alpha architecture manual for details).
9469 Like @samp{su}, but inexact traps are enabled as well.
9472 @item -mfp-rounding-mode=@var{rounding-mode}
9473 @opindex mfp-rounding-mode
9474 Selects the IEEE rounding mode. Other Alpha compilers call this option
9475 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9480 Normal IEEE rounding mode. Floating point numbers are rounded towards
9481 the nearest machine number or towards the even machine number in case
9485 Round towards minus infinity.
9488 Chopped rounding mode. Floating point numbers are rounded towards zero.
9491 Dynamic rounding mode. A field in the floating point control register
9492 (@var{fpcr}, see Alpha architecture reference manual) controls the
9493 rounding mode in effect. The C library initializes this register for
9494 rounding towards plus infinity. Thus, unless your program modifies the
9495 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9498 @item -mtrap-precision=@var{trap-precision}
9499 @opindex mtrap-precision
9500 In the Alpha architecture, floating point traps are imprecise. This
9501 means without software assistance it is impossible to recover from a
9502 floating trap and program execution normally needs to be terminated.
9503 GCC can generate code that can assist operating system trap handlers
9504 in determining the exact location that caused a floating point trap.
9505 Depending on the requirements of an application, different levels of
9506 precisions can be selected:
9510 Program precision. This option is the default and means a trap handler
9511 can only identify which program caused a floating point exception.
9514 Function precision. The trap handler can determine the function that
9515 caused a floating point exception.
9518 Instruction precision. The trap handler can determine the exact
9519 instruction that caused a floating point exception.
9522 Other Alpha compilers provide the equivalent options called
9523 @option{-scope_safe} and @option{-resumption_safe}.
9525 @item -mieee-conformant
9526 @opindex mieee-conformant
9527 This option marks the generated code as IEEE conformant. You must not
9528 use this option unless you also specify @option{-mtrap-precision=i} and either
9529 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9530 is to emit the line @samp{.eflag 48} in the function prologue of the
9531 generated assembly file. Under DEC Unix, this has the effect that
9532 IEEE-conformant math library routines will be linked in.
9534 @item -mbuild-constants
9535 @opindex mbuild-constants
9536 Normally GCC examines a 32- or 64-bit integer constant to
9537 see if it can construct it from smaller constants in two or three
9538 instructions. If it cannot, it will output the constant as a literal and
9539 generate code to load it from the data segment at runtime.
9541 Use this option to require GCC to construct @emph{all} integer constants
9542 using code, even if it takes more instructions (the maximum is six).
9544 You would typically use this option to build a shared library dynamic
9545 loader. Itself a shared library, it must relocate itself in memory
9546 before it can find the variables and constants in its own data segment.
9552 Select whether to generate code to be assembled by the vendor-supplied
9553 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9571 Indicate whether GCC should generate code to use the optional BWX,
9572 CIX, FIX and MAX instruction sets. The default is to use the instruction
9573 sets supported by the CPU type specified via @option{-mcpu=} option or that
9574 of the CPU on which GCC was built if none was specified.
9579 @opindex mfloat-ieee
9580 Generate code that uses (does not use) VAX F and G floating point
9581 arithmetic instead of IEEE single and double precision.
9583 @item -mexplicit-relocs
9584 @itemx -mno-explicit-relocs
9585 @opindex mexplicit-relocs
9586 @opindex mno-explicit-relocs
9587 Older Alpha assemblers provided no way to generate symbol relocations
9588 except via assembler macros. Use of these macros does not allow
9589 optimal instruction scheduling. GNU binutils as of version 2.12
9590 supports a new syntax that allows the compiler to explicitly mark
9591 which relocations should apply to which instructions. This option
9592 is mostly useful for debugging, as GCC detects the capabilities of
9593 the assembler when it is built and sets the default accordingly.
9597 @opindex msmall-data
9598 @opindex mlarge-data
9599 When @option{-mexplicit-relocs} is in effect, static data is
9600 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9601 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9602 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9603 16-bit relocations off of the @code{$gp} register. This limits the
9604 size of the small data area to 64KB, but allows the variables to be
9605 directly accessed via a single instruction.
9607 The default is @option{-mlarge-data}. With this option the data area
9608 is limited to just below 2GB@. Programs that require more than 2GB of
9609 data must use @code{malloc} or @code{mmap} to allocate the data in the
9610 heap instead of in the program's data segment.
9612 When generating code for shared libraries, @option{-fpic} implies
9613 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9617 @opindex msmall-text
9618 @opindex mlarge-text
9619 When @option{-msmall-text} is used, the compiler assumes that the
9620 code of the entire program (or shared library) fits in 4MB, and is
9621 thus reachable with a branch instruction. When @option{-msmall-data}
9622 is used, the compiler can assume that all local symbols share the
9623 same @code{$gp} value, and thus reduce the number of instructions
9624 required for a function call from 4 to 1.
9626 The default is @option{-mlarge-text}.
9628 @item -mcpu=@var{cpu_type}
9630 Set the instruction set and instruction scheduling parameters for
9631 machine type @var{cpu_type}. You can specify either the @samp{EV}
9632 style name or the corresponding chip number. GCC supports scheduling
9633 parameters for the EV4, EV5 and EV6 family of processors and will
9634 choose the default values for the instruction set from the processor
9635 you specify. If you do not specify a processor type, GCC will default
9636 to the processor on which the compiler was built.
9638 Supported values for @var{cpu_type} are
9644 Schedules as an EV4 and has no instruction set extensions.
9648 Schedules as an EV5 and has no instruction set extensions.
9652 Schedules as an EV5 and supports the BWX extension.
9657 Schedules as an EV5 and supports the BWX and MAX extensions.
9661 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9665 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9668 @item -mtune=@var{cpu_type}
9670 Set only the instruction scheduling parameters for machine type
9671 @var{cpu_type}. The instruction set is not changed.
9673 @item -mmemory-latency=@var{time}
9674 @opindex mmemory-latency
9675 Sets the latency the scheduler should assume for typical memory
9676 references as seen by the application. This number is highly
9677 dependent on the memory access patterns used by the application
9678 and the size of the external cache on the machine.
9680 Valid options for @var{time} are
9684 A decimal number representing clock cycles.
9690 The compiler contains estimates of the number of clock cycles for
9691 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9692 (also called Dcache, Scache, and Bcache), as well as to main memory.
9693 Note that L3 is only valid for EV5.
9698 @node DEC Alpha/VMS Options
9699 @subsection DEC Alpha/VMS Options
9701 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9704 @item -mvms-return-codes
9705 @opindex mvms-return-codes
9706 Return VMS condition codes from main. The default is to return POSIX
9707 style condition (e.g.@: error) codes.
9711 @subsection FRV Options
9718 Only use the first 32 general purpose registers.
9723 Use all 64 general purpose registers.
9728 Use only the first 32 floating point registers.
9733 Use all 64 floating point registers
9736 @opindex mhard-float
9738 Use hardware instructions for floating point operations.
9741 @opindex msoft-float
9743 Use library routines for floating point operations.
9748 Dynamically allocate condition code registers.
9753 Do not try to dynamically allocate condition code registers, only
9754 use @code{icc0} and @code{fcc0}.
9759 Change ABI to use double word insns.
9764 Do not use double word instructions.
9769 Use floating point double instructions.
9774 Do not use floating point double instructions.
9779 Use media instructions.
9784 Do not use media instructions.
9789 Use multiply and add/subtract instructions.
9794 Do not use multiply and add/subtract instructions.
9799 Select the FDPIC ABI, that uses function descriptors to represent
9800 pointers to functions. Without any PIC/PIE-related options, it
9801 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9802 assumes GOT entries and small data are within a 12-bit range from the
9803 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9804 are computed with 32 bits.
9805 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9808 @opindex minline-plt
9810 Enable inlining of PLT entries in function calls to functions that are
9811 not known to bind locally. It has no effect without @option{-mfdpic}.
9812 It's enabled by default if optimizing for speed and compiling for
9813 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9814 optimization option such as @option{-O3} or above is present in the
9820 Assume a large TLS segment when generating thread-local code.
9825 Do not assume a large TLS segment when generating thread-local code.
9830 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9831 that is known to be in read-only sections. It's enabled by default,
9832 except for @option{-fpic} or @option{-fpie}: even though it may help
9833 make the global offset table smaller, it trades 1 instruction for 4.
9834 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9835 one of which may be shared by multiple symbols, and it avoids the need
9836 for a GOT entry for the referenced symbol, so it's more likely to be a
9837 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
9839 @item -multilib-library-pic
9840 @opindex multilib-library-pic
9842 Link with the (library, not FD) pic libraries. It's implied by
9843 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
9844 @option{-fpic} without @option{-mfdpic}. You should never have to use
9850 Follow the EABI requirement of always creating a frame pointer whenever
9851 a stack frame is allocated. This option is enabled by default and can
9852 be disabled with @option{-mno-linked-fp}.
9855 @opindex mlong-calls
9857 Use indirect addressing to call functions outside the current
9858 compilation unit. This allows the functions to be placed anywhere
9859 within the 32-bit address space.
9861 @item -malign-labels
9862 @opindex malign-labels
9864 Try to align labels to an 8-byte boundary by inserting nops into the
9865 previous packet. This option only has an effect when VLIW packing
9866 is enabled. It doesn't create new packets; it merely adds nops to
9870 @opindex mlibrary-pic
9872 Generate position-independent EABI code.
9877 Use only the first four media accumulator registers.
9882 Use all eight media accumulator registers.
9887 Pack VLIW instructions.
9892 Do not pack VLIW instructions.
9897 Do not mark ABI switches in e_flags.
9902 Enable the use of conditional-move instructions (default).
9904 This switch is mainly for debugging the compiler and will likely be removed
9905 in a future version.
9907 @item -mno-cond-move
9908 @opindex mno-cond-move
9910 Disable the use of conditional-move instructions.
9912 This switch is mainly for debugging the compiler and will likely be removed
9913 in a future version.
9918 Enable the use of conditional set instructions (default).
9920 This switch is mainly for debugging the compiler and will likely be removed
9921 in a future version.
9926 Disable the use of conditional set instructions.
9928 This switch is mainly for debugging the compiler and will likely be removed
9929 in a future version.
9934 Enable the use of conditional execution (default).
9936 This switch is mainly for debugging the compiler and will likely be removed
9937 in a future version.
9939 @item -mno-cond-exec
9940 @opindex mno-cond-exec
9942 Disable the use of conditional execution.
9944 This switch is mainly for debugging the compiler and will likely be removed
9945 in a future version.
9948 @opindex mvliw-branch
9950 Run a pass to pack branches into VLIW instructions (default).
9952 This switch is mainly for debugging the compiler and will likely be removed
9953 in a future version.
9955 @item -mno-vliw-branch
9956 @opindex mno-vliw-branch
9958 Do not run a pass to pack branches into VLIW instructions.
9960 This switch is mainly for debugging the compiler and will likely be removed
9961 in a future version.
9963 @item -mmulti-cond-exec
9964 @opindex mmulti-cond-exec
9966 Enable optimization of @code{&&} and @code{||} in conditional execution
9969 This switch is mainly for debugging the compiler and will likely be removed
9970 in a future version.
9972 @item -mno-multi-cond-exec
9973 @opindex mno-multi-cond-exec
9975 Disable optimization of @code{&&} and @code{||} in conditional execution.
9977 This switch is mainly for debugging the compiler and will likely be removed
9978 in a future version.
9980 @item -mnested-cond-exec
9981 @opindex mnested-cond-exec
9983 Enable nested conditional execution optimizations (default).
9985 This switch is mainly for debugging the compiler and will likely be removed
9986 in a future version.
9988 @item -mno-nested-cond-exec
9989 @opindex mno-nested-cond-exec
9991 Disable nested conditional execution optimizations.
9993 This switch is mainly for debugging the compiler and will likely be removed
9994 in a future version.
9996 @item -moptimize-membar
9997 @opindex moptimize-membar
9999 This switch removes redundant @code{membar} instructions from the
10000 compiler generated code. It is enabled by default.
10002 @item -mno-optimize-membar
10003 @opindex mno-optimize-membar
10005 This switch disables the automatic removal of redundant @code{membar}
10006 instructions from the generated code.
10008 @item -mtomcat-stats
10009 @opindex mtomcat-stats
10011 Cause gas to print out tomcat statistics.
10013 @item -mcpu=@var{cpu}
10016 Select the processor type for which to generate code. Possible values are
10017 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10018 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10022 @node GNU/Linux Options
10023 @subsection GNU/Linux Options
10025 These @samp{-m} options are defined for GNU/Linux targets:
10030 Use the GNU C library instead of uClibc. This is the default except
10031 on @samp{*-*-linux-*uclibc*} targets.
10035 Use uClibc instead of the GNU C library. This is the default on
10036 @samp{*-*-linux-*uclibc*} targets.
10039 @node H8/300 Options
10040 @subsection H8/300 Options
10042 These @samp{-m} options are defined for the H8/300 implementations:
10047 Shorten some address references at link time, when possible; uses the
10048 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10049 ld, Using ld}, for a fuller description.
10053 Generate code for the H8/300H@.
10057 Generate code for the H8S@.
10061 Generate code for the H8S and H8/300H in the normal mode. This switch
10062 must be used either with @option{-mh} or @option{-ms}.
10066 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10070 Make @code{int} data 32 bits by default.
10073 @opindex malign-300
10074 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10075 The default for the H8/300H and H8S is to align longs and floats on 4
10077 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10078 This option has no effect on the H8/300.
10082 @subsection HPPA Options
10083 @cindex HPPA Options
10085 These @samp{-m} options are defined for the HPPA family of computers:
10088 @item -march=@var{architecture-type}
10090 Generate code for the specified architecture. The choices for
10091 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10092 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10093 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10094 architecture option for your machine. Code compiled for lower numbered
10095 architectures will run on higher numbered architectures, but not the
10098 @item -mpa-risc-1-0
10099 @itemx -mpa-risc-1-1
10100 @itemx -mpa-risc-2-0
10101 @opindex mpa-risc-1-0
10102 @opindex mpa-risc-1-1
10103 @opindex mpa-risc-2-0
10104 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10107 @opindex mbig-switch
10108 Generate code suitable for big switch tables. Use this option only if
10109 the assembler/linker complain about out of range branches within a switch
10112 @item -mjump-in-delay
10113 @opindex mjump-in-delay
10114 Fill delay slots of function calls with unconditional jump instructions
10115 by modifying the return pointer for the function call to be the target
10116 of the conditional jump.
10118 @item -mdisable-fpregs
10119 @opindex mdisable-fpregs
10120 Prevent floating point registers from being used in any manner. This is
10121 necessary for compiling kernels which perform lazy context switching of
10122 floating point registers. If you use this option and attempt to perform
10123 floating point operations, the compiler will abort.
10125 @item -mdisable-indexing
10126 @opindex mdisable-indexing
10127 Prevent the compiler from using indexing address modes. This avoids some
10128 rather obscure problems when compiling MIG generated code under MACH@.
10130 @item -mno-space-regs
10131 @opindex mno-space-regs
10132 Generate code that assumes the target has no space registers. This allows
10133 GCC to generate faster indirect calls and use unscaled index address modes.
10135 Such code is suitable for level 0 PA systems and kernels.
10137 @item -mfast-indirect-calls
10138 @opindex mfast-indirect-calls
10139 Generate code that assumes calls never cross space boundaries. This
10140 allows GCC to emit code which performs faster indirect calls.
10142 This option will not work in the presence of shared libraries or nested
10145 @item -mfixed-range=@var{register-range}
10146 @opindex mfixed-range
10147 Generate code treating the given register range as fixed registers.
10148 A fixed register is one that the register allocator can not use. This is
10149 useful when compiling kernel code. A register range is specified as
10150 two registers separated by a dash. Multiple register ranges can be
10151 specified separated by a comma.
10153 @item -mlong-load-store
10154 @opindex mlong-load-store
10155 Generate 3-instruction load and store sequences as sometimes required by
10156 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10159 @item -mportable-runtime
10160 @opindex mportable-runtime
10161 Use the portable calling conventions proposed by HP for ELF systems.
10165 Enable the use of assembler directives only GAS understands.
10167 @item -mschedule=@var{cpu-type}
10169 Schedule code according to the constraints for the machine type
10170 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10171 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10172 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10173 proper scheduling option for your machine. The default scheduling is
10177 @opindex mlinker-opt
10178 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10179 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10180 linkers in which they give bogus error messages when linking some programs.
10183 @opindex msoft-float
10184 Generate output containing library calls for floating point.
10185 @strong{Warning:} the requisite libraries are not available for all HPPA
10186 targets. Normally the facilities of the machine's usual C compiler are
10187 used, but this cannot be done directly in cross-compilation. You must make
10188 your own arrangements to provide suitable library functions for
10189 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
10190 does provide software floating point support.
10192 @option{-msoft-float} changes the calling convention in the output file;
10193 therefore, it is only useful if you compile @emph{all} of a program with
10194 this option. In particular, you need to compile @file{libgcc.a}, the
10195 library that comes with GCC, with @option{-msoft-float} in order for
10200 Generate the predefine, @code{_SIO}, for server IO@. The default is
10201 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10202 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10203 options are available under HP-UX and HI-UX@.
10207 Use GNU ld specific options. This passes @option{-shared} to ld when
10208 building a shared library. It is the default when GCC is configured,
10209 explicitly or implicitly, with the GNU linker. This option does not
10210 have any affect on which ld is called, it only changes what parameters
10211 are passed to that ld. The ld that is called is determined by the
10212 @option{--with-ld} configure option, GCC's program search path, and
10213 finally by the user's @env{PATH}. The linker used by GCC can be printed
10214 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10215 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10219 Use HP ld specific options. This passes @option{-b} to ld when building
10220 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10221 links. It is the default when GCC is configured, explicitly or
10222 implicitly, with the HP linker. This option does not have any affect on
10223 which ld is called, it only changes what parameters are passed to that
10224 ld. The ld that is called is determined by the @option{--with-ld}
10225 configure option, GCC's program search path, and finally by the user's
10226 @env{PATH}. The linker used by GCC can be printed using @samp{which
10227 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10228 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10231 @opindex mno-long-calls
10232 Generate code that uses long call sequences. This ensures that a call
10233 is always able to reach linker generated stubs. The default is to generate
10234 long calls only when the distance from the call site to the beginning
10235 of the function or translation unit, as the case may be, exceeds a
10236 predefined limit set by the branch type being used. The limits for
10237 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10238 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10241 Distances are measured from the beginning of functions when using the
10242 @option{-ffunction-sections} option, or when using the @option{-mgas}
10243 and @option{-mno-portable-runtime} options together under HP-UX with
10246 It is normally not desirable to use this option as it will degrade
10247 performance. However, it may be useful in large applications,
10248 particularly when partial linking is used to build the application.
10250 The types of long calls used depends on the capabilities of the
10251 assembler and linker, and the type of code being generated. The
10252 impact on systems that support long absolute calls, and long pic
10253 symbol-difference or pc-relative calls should be relatively small.
10254 However, an indirect call is used on 32-bit ELF systems in pic code
10255 and it is quite long.
10257 @item -munix=@var{unix-std}
10259 Generate compiler predefines and select a startfile for the specified
10260 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10261 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10262 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10263 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10264 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10267 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10268 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10269 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10270 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10271 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10272 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10274 It is @emph{important} to note that this option changes the interfaces
10275 for various library routines. It also affects the operational behavior
10276 of the C library. Thus, @emph{extreme} care is needed in using this
10279 Library code that is intended to operate with more than one UNIX
10280 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10281 as appropriate. Most GNU software doesn't provide this capability.
10285 Suppress the generation of link options to search libdld.sl when the
10286 @option{-static} option is specified on HP-UX 10 and later.
10290 The HP-UX implementation of setlocale in libc has a dependency on
10291 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10292 when the @option{-static} option is specified, special link options
10293 are needed to resolve this dependency.
10295 On HP-UX 10 and later, the GCC driver adds the necessary options to
10296 link with libdld.sl when the @option{-static} option is specified.
10297 This causes the resulting binary to be dynamic. On the 64-bit port,
10298 the linkers generate dynamic binaries by default in any case. The
10299 @option{-nolibdld} option can be used to prevent the GCC driver from
10300 adding these link options.
10304 Add support for multithreading with the @dfn{dce thread} library
10305 under HP-UX@. This option sets flags for both the preprocessor and
10309 @node i386 and x86-64 Options
10310 @subsection Intel 386 and AMD x86-64 Options
10311 @cindex i386 Options
10312 @cindex x86-64 Options
10313 @cindex Intel 386 Options
10314 @cindex AMD x86-64 Options
10316 These @samp{-m} options are defined for the i386 and x86-64 family of
10320 @item -mtune=@var{cpu-type}
10322 Tune to @var{cpu-type} everything applicable about the generated code, except
10323 for the ABI and the set of available instructions. The choices for
10324 @var{cpu-type} are:
10327 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10328 If you know the CPU on which your code will run, then you should use
10329 the corresponding @option{-mtune} option instead of
10330 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10331 of your application will have, then you should use this option.
10333 As new processors are deployed in the marketplace, the behavior of this
10334 option will change. Therefore, if you upgrade to a newer version of
10335 GCC, the code generated option will change to reflect the processors
10336 that were most common when that version of GCC was released.
10338 There is no @option{-march=generic} option because @option{-march}
10339 indicates the instruction set the compiler can use, and there is no
10340 generic instruction set applicable to all processors. In contrast,
10341 @option{-mtune} indicates the processor (or, in this case, collection of
10342 processors) for which the code is optimized.
10344 This selects the CPU to tune for at compilation time by determining
10345 the processor type of the compiling machine. Using @option{-mtune=native}
10346 will produce code optimized for the local machine under the constraints
10347 of the selected instruction set. Using @option{-march=native} will
10348 enable all instruction subsets supported by the local machine (hence
10349 the result might not run on different machines).
10351 Original Intel's i386 CPU@.
10353 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10354 @item i586, pentium
10355 Intel Pentium CPU with no MMX support.
10357 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10359 Intel PentiumPro CPU@.
10361 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10362 instruction set will be used, so the code will run on all i686 family chips.
10364 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10365 @item pentium3, pentium3m
10366 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10369 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10370 support. Used by Centrino notebooks.
10371 @item pentium4, pentium4m
10372 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10374 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10377 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10378 SSE2 and SSE3 instruction set support.
10380 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10381 instruction set support.
10383 AMD K6 CPU with MMX instruction set support.
10385 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10386 @item athlon, athlon-tbird
10387 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10389 @item athlon-4, athlon-xp, athlon-mp
10390 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10391 instruction set support.
10392 @item k8, opteron, athlon64, athlon-fx
10393 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10394 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10395 @item k8-sse3, opteron-sse3, athlon64-sse3
10396 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10397 @item amdfam10, barcelona
10398 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10399 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10400 instruction set extensions.)
10402 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10405 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
10406 instruction set support.
10408 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
10409 implemented for this chip.)
10411 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10412 implemented for this chip.)
10414 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10417 While picking a specific @var{cpu-type} will schedule things appropriately
10418 for that particular chip, the compiler will not generate any code that
10419 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10422 @item -march=@var{cpu-type}
10424 Generate instructions for the machine type @var{cpu-type}. The choices
10425 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10426 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10428 @item -mcpu=@var{cpu-type}
10430 A deprecated synonym for @option{-mtune}.
10432 @item -mfpmath=@var{unit}
10434 Generate floating point arithmetics for selected unit @var{unit}. The choices
10435 for @var{unit} are:
10439 Use the standard 387 floating point coprocessor present majority of chips and
10440 emulated otherwise. Code compiled with this option will run almost everywhere.
10441 The temporary results are computed in 80bit precision instead of precision
10442 specified by the type resulting in slightly different results compared to most
10443 of other chips. See @option{-ffloat-store} for more detailed description.
10445 This is the default choice for i386 compiler.
10448 Use scalar floating point instructions present in the SSE instruction set.
10449 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10450 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10451 instruction set supports only single precision arithmetics, thus the double and
10452 extended precision arithmetics is still done using 387. Later version, present
10453 only in Pentium4 and the future AMD x86-64 chips supports double precision
10456 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10457 or @option{-msse2} switches to enable SSE extensions and make this option
10458 effective. For the x86-64 compiler, these extensions are enabled by default.
10460 The resulting code should be considerably faster in the majority of cases and avoid
10461 the numerical instability problems of 387 code, but may break some existing
10462 code that expects temporaries to be 80bit.
10464 This is the default choice for the x86-64 compiler.
10467 Attempt to utilize both instruction sets at once. This effectively double the
10468 amount of available registers and on chips with separate execution units for
10469 387 and SSE the execution resources too. Use this option with care, as it is
10470 still experimental, because the GCC register allocator does not model separate
10471 functional units well resulting in instable performance.
10474 @item -masm=@var{dialect}
10475 @opindex masm=@var{dialect}
10476 Output asm instructions using selected @var{dialect}. Supported
10477 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10478 not support @samp{intel}.
10481 @itemx -mno-ieee-fp
10483 @opindex mno-ieee-fp
10484 Control whether or not the compiler uses IEEE floating point
10485 comparisons. These handle correctly the case where the result of a
10486 comparison is unordered.
10489 @opindex msoft-float
10490 Generate output containing library calls for floating point.
10491 @strong{Warning:} the requisite libraries are not part of GCC@.
10492 Normally the facilities of the machine's usual C compiler are used, but
10493 this can't be done directly in cross-compilation. You must make your
10494 own arrangements to provide suitable library functions for
10497 On machines where a function returns floating point results in the 80387
10498 register stack, some floating point opcodes may be emitted even if
10499 @option{-msoft-float} is used.
10501 @item -mno-fp-ret-in-387
10502 @opindex mno-fp-ret-in-387
10503 Do not use the FPU registers for return values of functions.
10505 The usual calling convention has functions return values of types
10506 @code{float} and @code{double} in an FPU register, even if there
10507 is no FPU@. The idea is that the operating system should emulate
10510 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10511 in ordinary CPU registers instead.
10513 @item -mno-fancy-math-387
10514 @opindex mno-fancy-math-387
10515 Some 387 emulators do not support the @code{sin}, @code{cos} and
10516 @code{sqrt} instructions for the 387. Specify this option to avoid
10517 generating those instructions. This option is the default on FreeBSD,
10518 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10519 indicates that the target cpu will always have an FPU and so the
10520 instruction will not need emulation. As of revision 2.6.1, these
10521 instructions are not generated unless you also use the
10522 @option{-funsafe-math-optimizations} switch.
10524 @item -malign-double
10525 @itemx -mno-align-double
10526 @opindex malign-double
10527 @opindex mno-align-double
10528 Control whether GCC aligns @code{double}, @code{long double}, and
10529 @code{long long} variables on a two word boundary or a one word
10530 boundary. Aligning @code{double} variables on a two word boundary will
10531 produce code that runs somewhat faster on a @samp{Pentium} at the
10532 expense of more memory.
10534 On x86-64, @option{-malign-double} is enabled by default.
10536 @strong{Warning:} if you use the @option{-malign-double} switch,
10537 structures containing the above types will be aligned differently than
10538 the published application binary interface specifications for the 386
10539 and will not be binary compatible with structures in code compiled
10540 without that switch.
10542 @item -m96bit-long-double
10543 @itemx -m128bit-long-double
10544 @opindex m96bit-long-double
10545 @opindex m128bit-long-double
10546 These switches control the size of @code{long double} type. The i386
10547 application binary interface specifies the size to be 96 bits,
10548 so @option{-m96bit-long-double} is the default in 32 bit mode.
10550 Modern architectures (Pentium and newer) would prefer @code{long double}
10551 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10552 conforming to the ABI, this would not be possible. So specifying a
10553 @option{-m128bit-long-double} will align @code{long double}
10554 to a 16 byte boundary by padding the @code{long double} with an additional
10557 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10558 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10560 Notice that neither of these options enable any extra precision over the x87
10561 standard of 80 bits for a @code{long double}.
10563 @strong{Warning:} if you override the default value for your target ABI, the
10564 structures and arrays containing @code{long double} variables will change
10565 their size as well as function calling convention for function taking
10566 @code{long double} will be modified. Hence they will not be binary
10567 compatible with arrays or structures in code compiled without that switch.
10569 @item -mmlarge-data-threshold=@var{number}
10570 @opindex mlarge-data-threshold=@var{number}
10571 When @option{-mcmodel=medium} is specified, the data greater than
10572 @var{threshold} are placed in large data section. This value must be the
10573 same across all object linked into the binary and defaults to 65535.
10577 Use a different function-calling convention, in which functions that
10578 take a fixed number of arguments return with the @code{ret} @var{num}
10579 instruction, which pops their arguments while returning. This saves one
10580 instruction in the caller since there is no need to pop the arguments
10583 You can specify that an individual function is called with this calling
10584 sequence with the function attribute @samp{stdcall}. You can also
10585 override the @option{-mrtd} option by using the function attribute
10586 @samp{cdecl}. @xref{Function Attributes}.
10588 @strong{Warning:} this calling convention is incompatible with the one
10589 normally used on Unix, so you cannot use it if you need to call
10590 libraries compiled with the Unix compiler.
10592 Also, you must provide function prototypes for all functions that
10593 take variable numbers of arguments (including @code{printf});
10594 otherwise incorrect code will be generated for calls to those
10597 In addition, seriously incorrect code will result if you call a
10598 function with too many arguments. (Normally, extra arguments are
10599 harmlessly ignored.)
10601 @item -mregparm=@var{num}
10603 Control how many registers are used to pass integer arguments. By
10604 default, no registers are used to pass arguments, and at most 3
10605 registers can be used. You can control this behavior for a specific
10606 function by using the function attribute @samp{regparm}.
10607 @xref{Function Attributes}.
10609 @strong{Warning:} if you use this switch, and
10610 @var{num} is nonzero, then you must build all modules with the same
10611 value, including any libraries. This includes the system libraries and
10615 @opindex msseregparm
10616 Use SSE register passing conventions for float and double arguments
10617 and return values. You can control this behavior for a specific
10618 function by using the function attribute @samp{sseregparm}.
10619 @xref{Function Attributes}.
10621 @strong{Warning:} if you use this switch then you must build all
10622 modules with the same value, including any libraries. This includes
10623 the system libraries and startup modules.
10632 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10633 is specified, the significands of results of floating-point operations are
10634 rounded to 24 bits (single precision); @option{-mpc64} rounds the the
10635 significands of results of floating-point operations to 53 bits (double
10636 precision) and @option{-mpc80} rounds the significands of results of
10637 floating-point operations to 64 bits (extended double precision), which is
10638 the default. When this option is used, floating-point operations in higher
10639 precisions are not available to the programmer without setting the FPU
10640 control word explicitly.
10642 Setting the rounding of floating-point operations to less than the default
10643 80 bits can speed some programs by 2% or more. Note that some mathematical
10644 libraries assume that extended precision (80 bit) floating-point operations
10645 are enabled by default; routines in such libraries could suffer significant
10646 loss of accuracy, typically through so-called "catastrophic cancellation",
10647 when this option is used to set the precision to less than extended precision.
10649 @item -mstackrealign
10650 @opindex mstackrealign
10651 Realign the stack at entry. On the Intel x86, the
10652 @option{-mstackrealign} option will generate an alternate prologue and
10653 epilogue that realigns the runtime stack. This supports mixing legacy
10654 codes that keep a 4-byte aligned stack with modern codes that keep a
10655 16-byte stack for SSE compatibility. The alternate prologue and
10656 epilogue are slower and bigger than the regular ones, and the
10657 alternate prologue requires an extra scratch register; this lowers the
10658 number of registers available if used in conjunction with the
10659 @code{regparm} attribute. The @option{-mstackrealign} option is
10660 incompatible with the nested function prologue; this is considered a
10661 hard error. See also the attribute @code{force_align_arg_pointer},
10662 applicable to individual functions.
10664 @item -mpreferred-stack-boundary=@var{num}
10665 @opindex mpreferred-stack-boundary
10666 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
10667 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
10668 the default is 4 (16 bytes or 128 bits).
10670 On Pentium and PentiumPro, @code{double} and @code{long double} values
10671 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
10672 suffer significant run time performance penalties. On Pentium III, the
10673 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
10674 properly if it is not 16 byte aligned.
10676 To ensure proper alignment of this values on the stack, the stack boundary
10677 must be as aligned as that required by any value stored on the stack.
10678 Further, every function must be generated such that it keeps the stack
10679 aligned. Thus calling a function compiled with a higher preferred
10680 stack boundary from a function compiled with a lower preferred stack
10681 boundary will most likely misalign the stack. It is recommended that
10682 libraries that use callbacks always use the default setting.
10684 This extra alignment does consume extra stack space, and generally
10685 increases code size. Code that is sensitive to stack space usage, such
10686 as embedded systems and operating system kernels, may want to reduce the
10687 preferred alignment to @option{-mpreferred-stack-boundary=2}.
10721 These switches enable or disable the use of instructions in the MMX,
10722 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4A, SSE5, ABM or 3DNow!@: extended
10724 These extensions are also available as built-in functions: see
10725 @ref{X86 Built-in Functions}, for details of the functions enabled and
10726 disabled by these switches.
10728 To have SSE/SSE2 instructions generated automatically from floating-point
10729 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
10731 These options will enable GCC to use these extended instructions in
10732 generated code, even without @option{-mfpmath=sse}. Applications which
10733 perform runtime CPU detection must compile separate files for each
10734 supported architecture, using the appropriate flags. In particular,
10735 the file containing the CPU detection code should be compiled without
10740 This option will enable GCC to use CMPXCHG16B instruction in generated code.
10741 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
10742 data types. This is useful for high resolution counters that could be updated
10743 by multiple processors (or cores). This instruction is generated as part of
10744 atomic built-in functions: see @ref{Atomic Builtins} for details.
10748 This option will enable GCC to use SAHF instruction in generated 64-bit code.
10749 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
10750 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
10751 SAHF are load and store instructions, respectively, for certain status flags.
10752 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
10753 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
10757 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
10758 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Rhapson step
10759 to increase precision instead of DIVSS and SQRTSS (and their vectorized
10760 variants) for single precision floating point arguments. These instructions
10761 are generated only when @option{-funsafe-math-optimizations} is enabled
10762 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
10763 Note that while the throughput of the sequence is higher than the throughput
10764 of the non-reciprocal instruction, the precision of the sequence can be
10765 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
10767 @item -mveclibabi=@var{type}
10768 @opindex mveclibabi
10769 Specifies the ABI type to use for vectorizing intrinsics using an
10770 external library. Supported types are @code{acml} for the AMD
10771 math core library style of interfacing. GCC will currently emit
10772 calls to @code{__vrd2_sin}, @code{__vrd2_cos}, @code{__vrd2_exp},
10773 @code{__vrd2_log}, @code{__vrd2_log2}, @code{__vrd2_log10},
10774 @code{__vrs4_sinf}, @code{__vrs4_cosf}, @code{__vrs4_expf},
10775 @code{__vrs4_logf}, @code{__vrs4_log2f}, @code{__vrs4_log10f}
10776 and @code{__vrs4_powf} when using this type and @option{-ftree-vectorize}
10777 is enabled. A ACML ABI compatible library will have to be specified
10781 @itemx -mno-push-args
10782 @opindex mpush-args
10783 @opindex mno-push-args
10784 Use PUSH operations to store outgoing parameters. This method is shorter
10785 and usually equally fast as method using SUB/MOV operations and is enabled
10786 by default. In some cases disabling it may improve performance because of
10787 improved scheduling and reduced dependencies.
10789 @item -maccumulate-outgoing-args
10790 @opindex maccumulate-outgoing-args
10791 If enabled, the maximum amount of space required for outgoing arguments will be
10792 computed in the function prologue. This is faster on most modern CPUs
10793 because of reduced dependencies, improved scheduling and reduced stack usage
10794 when preferred stack boundary is not equal to 2. The drawback is a notable
10795 increase in code size. This switch implies @option{-mno-push-args}.
10799 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
10800 on thread-safe exception handling must compile and link all code with the
10801 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10802 @option{-D_MT}; when linking, it links in a special thread helper library
10803 @option{-lmingwthrd} which cleans up per thread exception handling data.
10805 @item -mno-align-stringops
10806 @opindex mno-align-stringops
10807 Do not align destination of inlined string operations. This switch reduces
10808 code size and improves performance in case the destination is already aligned,
10809 but GCC doesn't know about it.
10811 @item -minline-all-stringops
10812 @opindex minline-all-stringops
10813 By default GCC inlines string operations only when destination is known to be
10814 aligned at least to 4 byte boundary. This enables more inlining, increase code
10815 size, but may improve performance of code that depends on fast memcpy, strlen
10816 and memset for short lengths.
10818 @item -minline-stringops-dynamically
10819 @opindex minline-stringops-dynamically
10820 For string operation of unknown size, inline runtime checks so for small
10821 blocks inline code is used, while for large blocks library call is used.
10823 @item -mstringop-strategy=@var{alg}
10824 @opindex mstringop-strategy=@var{alg}
10825 Overwrite internal decision heuristic about particular algorithm to inline
10826 string operation with. The allowed values are @code{rep_byte},
10827 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
10828 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
10829 expanding inline loop, @code{libcall} for always expanding library call.
10831 @item -momit-leaf-frame-pointer
10832 @opindex momit-leaf-frame-pointer
10833 Don't keep the frame pointer in a register for leaf functions. This
10834 avoids the instructions to save, set up and restore frame pointers and
10835 makes an extra register available in leaf functions. The option
10836 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10837 which might make debugging harder.
10839 @item -mtls-direct-seg-refs
10840 @itemx -mno-tls-direct-seg-refs
10841 @opindex mtls-direct-seg-refs
10842 Controls whether TLS variables may be accessed with offsets from the
10843 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
10844 or whether the thread base pointer must be added. Whether or not this
10845 is legal depends on the operating system, and whether it maps the
10846 segment to cover the entire TLS area.
10848 For systems that use GNU libc, the default is on.
10851 @itemx -mno-fused-madd
10852 @opindex mfused-madd
10853 Enable automatic generation of fused floating point multiply-add instructions
10854 if the ISA supports such instructions. The -mfused-madd option is on by
10855 default. The fused multiply-add instructions have a different
10856 rounding behavior compared to executing a multiply followed by an add.
10859 These @samp{-m} switches are supported in addition to the above
10860 on AMD x86-64 processors in 64-bit environments.
10867 Generate code for a 32-bit or 64-bit environment.
10868 The 32-bit environment sets int, long and pointer to 32 bits and
10869 generates code that runs on any i386 system.
10870 The 64-bit environment sets int to 32 bits and long and pointer
10871 to 64 bits and generates code for AMD's x86-64 architecture. For
10872 darwin only the -m64 option turns off the @option{-fno-pic} and
10873 @option{-mdynamic-no-pic} options.
10875 @item -mno-red-zone
10876 @opindex no-red-zone
10877 Do not use a so called red zone for x86-64 code. The red zone is mandated
10878 by the x86-64 ABI, it is a 128-byte area beyond the location of the
10879 stack pointer that will not be modified by signal or interrupt handlers
10880 and therefore can be used for temporary data without adjusting the stack
10881 pointer. The flag @option{-mno-red-zone} disables this red zone.
10883 @item -mcmodel=small
10884 @opindex mcmodel=small
10885 Generate code for the small code model: the program and its symbols must
10886 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
10887 Programs can be statically or dynamically linked. This is the default
10890 @item -mcmodel=kernel
10891 @opindex mcmodel=kernel
10892 Generate code for the kernel code model. The kernel runs in the
10893 negative 2 GB of the address space.
10894 This model has to be used for Linux kernel code.
10896 @item -mcmodel=medium
10897 @opindex mcmodel=medium
10898 Generate code for the medium model: The program is linked in the lower 2
10899 GB of the address space but symbols can be located anywhere in the
10900 address space. Programs can be statically or dynamically linked, but
10901 building of shared libraries are not supported with the medium model.
10903 @item -mcmodel=large
10904 @opindex mcmodel=large
10905 Generate code for the large model: This model makes no assumptions
10906 about addresses and sizes of sections.
10909 @node IA-64 Options
10910 @subsection IA-64 Options
10911 @cindex IA-64 Options
10913 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10917 @opindex mbig-endian
10918 Generate code for a big endian target. This is the default for HP-UX@.
10920 @item -mlittle-endian
10921 @opindex mlittle-endian
10922 Generate code for a little endian target. This is the default for AIX5
10928 @opindex mno-gnu-as
10929 Generate (or don't) code for the GNU assembler. This is the default.
10930 @c Also, this is the default if the configure option @option{--with-gnu-as}
10936 @opindex mno-gnu-ld
10937 Generate (or don't) code for the GNU linker. This is the default.
10938 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10943 Generate code that does not use a global pointer register. The result
10944 is not position independent code, and violates the IA-64 ABI@.
10946 @item -mvolatile-asm-stop
10947 @itemx -mno-volatile-asm-stop
10948 @opindex mvolatile-asm-stop
10949 @opindex mno-volatile-asm-stop
10950 Generate (or don't) a stop bit immediately before and after volatile asm
10953 @item -mregister-names
10954 @itemx -mno-register-names
10955 @opindex mregister-names
10956 @opindex mno-register-names
10957 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10958 the stacked registers. This may make assembler output more readable.
10964 Disable (or enable) optimizations that use the small data section. This may
10965 be useful for working around optimizer bugs.
10967 @item -mconstant-gp
10968 @opindex mconstant-gp
10969 Generate code that uses a single constant global pointer value. This is
10970 useful when compiling kernel code.
10974 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10975 This is useful when compiling firmware code.
10977 @item -minline-float-divide-min-latency
10978 @opindex minline-float-divide-min-latency
10979 Generate code for inline divides of floating point values
10980 using the minimum latency algorithm.
10982 @item -minline-float-divide-max-throughput
10983 @opindex minline-float-divide-max-throughput
10984 Generate code for inline divides of floating point values
10985 using the maximum throughput algorithm.
10987 @item -minline-int-divide-min-latency
10988 @opindex minline-int-divide-min-latency
10989 Generate code for inline divides of integer values
10990 using the minimum latency algorithm.
10992 @item -minline-int-divide-max-throughput
10993 @opindex minline-int-divide-max-throughput
10994 Generate code for inline divides of integer values
10995 using the maximum throughput algorithm.
10997 @item -minline-sqrt-min-latency
10998 @opindex minline-sqrt-min-latency
10999 Generate code for inline square roots
11000 using the minimum latency algorithm.
11002 @item -minline-sqrt-max-throughput
11003 @opindex minline-sqrt-max-throughput
11004 Generate code for inline square roots
11005 using the maximum throughput algorithm.
11007 @item -mno-dwarf2-asm
11008 @itemx -mdwarf2-asm
11009 @opindex mno-dwarf2-asm
11010 @opindex mdwarf2-asm
11011 Don't (or do) generate assembler code for the DWARF2 line number debugging
11012 info. This may be useful when not using the GNU assembler.
11014 @item -mearly-stop-bits
11015 @itemx -mno-early-stop-bits
11016 @opindex mearly-stop-bits
11017 @opindex mno-early-stop-bits
11018 Allow stop bits to be placed earlier than immediately preceding the
11019 instruction that triggered the stop bit. This can improve instruction
11020 scheduling, but does not always do so.
11022 @item -mfixed-range=@var{register-range}
11023 @opindex mfixed-range
11024 Generate code treating the given register range as fixed registers.
11025 A fixed register is one that the register allocator can not use. This is
11026 useful when compiling kernel code. A register range is specified as
11027 two registers separated by a dash. Multiple register ranges can be
11028 specified separated by a comma.
11030 @item -mtls-size=@var{tls-size}
11032 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11035 @item -mtune=@var{cpu-type}
11037 Tune the instruction scheduling for a particular CPU, Valid values are
11038 itanium, itanium1, merced, itanium2, and mckinley.
11044 Add support for multithreading using the POSIX threads library. This
11045 option sets flags for both the preprocessor and linker. It does
11046 not affect the thread safety of object code produced by the compiler or
11047 that of libraries supplied with it. These are HP-UX specific flags.
11053 Generate code for a 32-bit or 64-bit environment.
11054 The 32-bit environment sets int, long and pointer to 32 bits.
11055 The 64-bit environment sets int to 32 bits and long and pointer
11056 to 64 bits. These are HP-UX specific flags.
11058 @item -mno-sched-br-data-spec
11059 @itemx -msched-br-data-spec
11060 @opindex mno-sched-br-data-spec
11061 @opindex msched-br-data-spec
11062 (Dis/En)able data speculative scheduling before reload.
11063 This will result in generation of the ld.a instructions and
11064 the corresponding check instructions (ld.c / chk.a).
11065 The default is 'disable'.
11067 @item -msched-ar-data-spec
11068 @itemx -mno-sched-ar-data-spec
11069 @opindex msched-ar-data-spec
11070 @opindex mno-sched-ar-data-spec
11071 (En/Dis)able data speculative scheduling after reload.
11072 This will result in generation of the ld.a instructions and
11073 the corresponding check instructions (ld.c / chk.a).
11074 The default is 'enable'.
11076 @item -mno-sched-control-spec
11077 @itemx -msched-control-spec
11078 @opindex mno-sched-control-spec
11079 @opindex msched-control-spec
11080 (Dis/En)able control speculative scheduling. This feature is
11081 available only during region scheduling (i.e.@: before reload).
11082 This will result in generation of the ld.s instructions and
11083 the corresponding check instructions chk.s .
11084 The default is 'disable'.
11086 @item -msched-br-in-data-spec
11087 @itemx -mno-sched-br-in-data-spec
11088 @opindex msched-br-in-data-spec
11089 @opindex mno-sched-br-in-data-spec
11090 (En/Dis)able speculative scheduling of the instructions that
11091 are dependent on the data speculative loads before reload.
11092 This is effective only with @option{-msched-br-data-spec} enabled.
11093 The default is 'enable'.
11095 @item -msched-ar-in-data-spec
11096 @itemx -mno-sched-ar-in-data-spec
11097 @opindex msched-ar-in-data-spec
11098 @opindex mno-sched-ar-in-data-spec
11099 (En/Dis)able speculative scheduling of the instructions that
11100 are dependent on the data speculative loads after reload.
11101 This is effective only with @option{-msched-ar-data-spec} enabled.
11102 The default is 'enable'.
11104 @item -msched-in-control-spec
11105 @itemx -mno-sched-in-control-spec
11106 @opindex msched-in-control-spec
11107 @opindex mno-sched-in-control-spec
11108 (En/Dis)able speculative scheduling of the instructions that
11109 are dependent on the control speculative loads.
11110 This is effective only with @option{-msched-control-spec} enabled.
11111 The default is 'enable'.
11114 @itemx -mno-sched-ldc
11115 @opindex msched-ldc
11116 @opindex mno-sched-ldc
11117 (En/Dis)able use of simple data speculation checks ld.c .
11118 If disabled, only chk.a instructions will be emitted to check
11119 data speculative loads.
11120 The default is 'enable'.
11122 @item -mno-sched-control-ldc
11123 @itemx -msched-control-ldc
11124 @opindex mno-sched-control-ldc
11125 @opindex msched-control-ldc
11126 (Dis/En)able use of ld.c instructions to check control speculative loads.
11127 If enabled, in case of control speculative load with no speculatively
11128 scheduled dependent instructions this load will be emitted as ld.sa and
11129 ld.c will be used to check it.
11130 The default is 'disable'.
11132 @item -mno-sched-spec-verbose
11133 @itemx -msched-spec-verbose
11134 @opindex mno-sched-spec-verbose
11135 @opindex msched-spec-verbose
11136 (Dis/En)able printing of the information about speculative motions.
11138 @item -mno-sched-prefer-non-data-spec-insns
11139 @itemx -msched-prefer-non-data-spec-insns
11140 @opindex mno-sched-prefer-non-data-spec-insns
11141 @opindex msched-prefer-non-data-spec-insns
11142 If enabled, data speculative instructions will be chosen for schedule
11143 only if there are no other choices at the moment. This will make
11144 the use of the data speculation much more conservative.
11145 The default is 'disable'.
11147 @item -mno-sched-prefer-non-control-spec-insns
11148 @itemx -msched-prefer-non-control-spec-insns
11149 @opindex mno-sched-prefer-non-control-spec-insns
11150 @opindex msched-prefer-non-control-spec-insns
11151 If enabled, control speculative instructions will be chosen for schedule
11152 only if there are no other choices at the moment. This will make
11153 the use of the control speculation much more conservative.
11154 The default is 'disable'.
11156 @item -mno-sched-count-spec-in-critical-path
11157 @itemx -msched-count-spec-in-critical-path
11158 @opindex mno-sched-count-spec-in-critical-path
11159 @opindex msched-count-spec-in-critical-path
11160 If enabled, speculative dependencies will be considered during
11161 computation of the instructions priorities. This will make the use of the
11162 speculation a bit more conservative.
11163 The default is 'disable'.
11168 @subsection M32C Options
11169 @cindex M32C options
11172 @item -mcpu=@var{name}
11174 Select the CPU for which code is generated. @var{name} may be one of
11175 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11176 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11177 the M32C/80 series.
11181 Specifies that the program will be run on the simulator. This causes
11182 an alternate runtime library to be linked in which supports, for
11183 example, file I/O@. You must not use this option when generating
11184 programs that will run on real hardware; you must provide your own
11185 runtime library for whatever I/O functions are needed.
11187 @item -memregs=@var{number}
11189 Specifies the number of memory-based pseudo-registers GCC will use
11190 during code generation. These pseudo-registers will be used like real
11191 registers, so there is a tradeoff between GCC's ability to fit the
11192 code into available registers, and the performance penalty of using
11193 memory instead of registers. Note that all modules in a program must
11194 be compiled with the same value for this option. Because of that, you
11195 must not use this option with the default runtime libraries gcc
11200 @node M32R/D Options
11201 @subsection M32R/D Options
11202 @cindex M32R/D options
11204 These @option{-m} options are defined for Renesas M32R/D architectures:
11209 Generate code for the M32R/2@.
11213 Generate code for the M32R/X@.
11217 Generate code for the M32R@. This is the default.
11219 @item -mmodel=small
11220 @opindex mmodel=small
11221 Assume all objects live in the lower 16MB of memory (so that their addresses
11222 can be loaded with the @code{ld24} instruction), and assume all subroutines
11223 are reachable with the @code{bl} instruction.
11224 This is the default.
11226 The addressability of a particular object can be set with the
11227 @code{model} attribute.
11229 @item -mmodel=medium
11230 @opindex mmodel=medium
11231 Assume objects may be anywhere in the 32-bit address space (the compiler
11232 will generate @code{seth/add3} instructions to load their addresses), and
11233 assume all subroutines are reachable with the @code{bl} instruction.
11235 @item -mmodel=large
11236 @opindex mmodel=large
11237 Assume objects may be anywhere in the 32-bit address space (the compiler
11238 will generate @code{seth/add3} instructions to load their addresses), and
11239 assume subroutines may not be reachable with the @code{bl} instruction
11240 (the compiler will generate the much slower @code{seth/add3/jl}
11241 instruction sequence).
11244 @opindex msdata=none
11245 Disable use of the small data area. Variables will be put into
11246 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11247 @code{section} attribute has been specified).
11248 This is the default.
11250 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11251 Objects may be explicitly put in the small data area with the
11252 @code{section} attribute using one of these sections.
11254 @item -msdata=sdata
11255 @opindex msdata=sdata
11256 Put small global and static data in the small data area, but do not
11257 generate special code to reference them.
11260 @opindex msdata=use
11261 Put small global and static data in the small data area, and generate
11262 special instructions to reference them.
11266 @cindex smaller data references
11267 Put global and static objects less than or equal to @var{num} bytes
11268 into the small data or bss sections instead of the normal data or bss
11269 sections. The default value of @var{num} is 8.
11270 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11271 for this option to have any effect.
11273 All modules should be compiled with the same @option{-G @var{num}} value.
11274 Compiling with different values of @var{num} may or may not work; if it
11275 doesn't the linker will give an error message---incorrect code will not be
11280 Makes the M32R specific code in the compiler display some statistics
11281 that might help in debugging programs.
11283 @item -malign-loops
11284 @opindex malign-loops
11285 Align all loops to a 32-byte boundary.
11287 @item -mno-align-loops
11288 @opindex mno-align-loops
11289 Do not enforce a 32-byte alignment for loops. This is the default.
11291 @item -missue-rate=@var{number}
11292 @opindex missue-rate=@var{number}
11293 Issue @var{number} instructions per cycle. @var{number} can only be 1
11296 @item -mbranch-cost=@var{number}
11297 @opindex mbranch-cost=@var{number}
11298 @var{number} can only be 1 or 2. If it is 1 then branches will be
11299 preferred over conditional code, if it is 2, then the opposite will
11302 @item -mflush-trap=@var{number}
11303 @opindex mflush-trap=@var{number}
11304 Specifies the trap number to use to flush the cache. The default is
11305 12. Valid numbers are between 0 and 15 inclusive.
11307 @item -mno-flush-trap
11308 @opindex mno-flush-trap
11309 Specifies that the cache cannot be flushed by using a trap.
11311 @item -mflush-func=@var{name}
11312 @opindex mflush-func=@var{name}
11313 Specifies the name of the operating system function to call to flush
11314 the cache. The default is @emph{_flush_cache}, but a function call
11315 will only be used if a trap is not available.
11317 @item -mno-flush-func
11318 @opindex mno-flush-func
11319 Indicates that there is no OS function for flushing the cache.
11323 @node M680x0 Options
11324 @subsection M680x0 Options
11325 @cindex M680x0 options
11327 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11328 The default settings depend on which architecture was selected when
11329 the compiler was configured; the defaults for the most common choices
11333 @item -march=@var{arch}
11335 Generate code for a specific M680x0 or ColdFire instruction set
11336 architecture. Permissible values of @var{arch} for M680x0
11337 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11338 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11339 architectures are selected according to Freescale's ISA classification
11340 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11341 @samp{isab} and @samp{isac}.
11343 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11344 code for a ColdFire target. The @var{arch} in this macro is one of the
11345 @option{-march} arguments given above.
11347 When used together, @option{-march} and @option{-mtune} select code
11348 that runs on a family of similar processors but that is optimized
11349 for a particular microarchitecture.
11351 @item -mcpu=@var{cpu}
11353 Generate code for a specific M680x0 or ColdFire processor.
11354 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11355 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11356 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11357 below, which also classifies the CPUs into families:
11359 @multitable @columnfractions 0.20 0.80
11360 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11361 @item @samp{51qe} @tab @samp{51qe}
11362 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11363 @item @samp{5206e} @tab @samp{5206e}
11364 @item @samp{5208} @tab @samp{5207} @samp{5208}
11365 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11366 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11367 @item @samp{5216} @tab @samp{5214} @samp{5216}
11368 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11369 @item @samp{5225} @tab @samp{5224} @samp{5225}
11370 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11371 @item @samp{5249} @tab @samp{5249}
11372 @item @samp{5250} @tab @samp{5250}
11373 @item @samp{5271} @tab @samp{5270} @samp{5271}
11374 @item @samp{5272} @tab @samp{5272}
11375 @item @samp{5275} @tab @samp{5274} @samp{5275}
11376 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11377 @item @samp{5307} @tab @samp{5307}
11378 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11379 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11380 @item @samp{5407} @tab @samp{5407}
11381 @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}
11384 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11385 @var{arch} is compatible with @var{cpu}. Other combinations of
11386 @option{-mcpu} and @option{-march} are rejected.
11388 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11389 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11390 where the value of @var{family} is given by the table above.
11392 @item -mtune=@var{tune}
11394 Tune the code for a particular microarchitecture, within the
11395 constraints set by @option{-march} and @option{-mcpu}.
11396 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11397 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11398 and @samp{cpu32}. The ColdFire microarchitectures
11399 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11401 You can also use @option{-mtune=68020-40} for code that needs
11402 to run relatively well on 68020, 68030 and 68040 targets.
11403 @option{-mtune=68020-60} is similar but includes 68060 targets
11404 as well. These two options select the same tuning decisions as
11405 @option{-m68020-40} and @option{-m68020-60} respectively.
11407 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11408 when tuning for 680x0 architecture @var{arch}. It also defines
11409 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11410 option is used. If gcc is tuning for a range of architectures,
11411 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11412 it defines the macros for every architecture in the range.
11414 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11415 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11416 of the arguments given above.
11422 Generate output for a 68000. This is the default
11423 when the compiler is configured for 68000-based systems.
11424 It is equivalent to @option{-march=68000}.
11426 Use this option for microcontrollers with a 68000 or EC000 core,
11427 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11431 Generate output for a 68010. This is the default
11432 when the compiler is configured for 68010-based systems.
11433 It is equivalent to @option{-march=68010}.
11439 Generate output for a 68020. This is the default
11440 when the compiler is configured for 68020-based systems.
11441 It is equivalent to @option{-march=68020}.
11445 Generate output for a 68030. This is the default when the compiler is
11446 configured for 68030-based systems. It is equivalent to
11447 @option{-march=68030}.
11451 Generate output for a 68040. This is the default when the compiler is
11452 configured for 68040-based systems. It is equivalent to
11453 @option{-march=68040}.
11455 This option inhibits the use of 68881/68882 instructions that have to be
11456 emulated by software on the 68040. Use this option if your 68040 does not
11457 have code to emulate those instructions.
11461 Generate output for a 68060. This is the default when the compiler is
11462 configured for 68060-based systems. It is equivalent to
11463 @option{-march=68060}.
11465 This option inhibits the use of 68020 and 68881/68882 instructions that
11466 have to be emulated by software on the 68060. Use this option if your 68060
11467 does not have code to emulate those instructions.
11471 Generate output for a CPU32. This is the default
11472 when the compiler is configured for CPU32-based systems.
11473 It is equivalent to @option{-march=cpu32}.
11475 Use this option for microcontrollers with a
11476 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11477 68336, 68340, 68341, 68349 and 68360.
11481 Generate output for a 520X ColdFire CPU@. This is the default
11482 when the compiler is configured for 520X-based systems.
11483 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11484 in favor of that option.
11486 Use this option for microcontroller with a 5200 core, including
11487 the MCF5202, MCF5203, MCF5204 and MCF5206.
11491 Generate output for a 5206e ColdFire CPU@. The option is now
11492 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11496 Generate output for a member of the ColdFire 528X family.
11497 The option is now deprecated in favor of the equivalent
11498 @option{-mcpu=528x}.
11502 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
11503 in favor of the equivalent @option{-mcpu=5307}.
11507 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
11508 in favor of the equivalent @option{-mcpu=5407}.
11512 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11513 This includes use of hardware floating point instructions.
11514 The option is equivalent to @option{-mcpu=547x}, and is now
11515 deprecated in favor of that option.
11519 Generate output for a 68040, without using any of the new instructions.
11520 This results in code which can run relatively efficiently on either a
11521 68020/68881 or a 68030 or a 68040. The generated code does use the
11522 68881 instructions that are emulated on the 68040.
11524 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11528 Generate output for a 68060, without using any of the new instructions.
11529 This results in code which can run relatively efficiently on either a
11530 68020/68881 or a 68030 or a 68040. The generated code does use the
11531 68881 instructions that are emulated on the 68060.
11533 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11537 @opindex mhard-float
11539 Generate floating-point instructions. This is the default for 68020
11540 and above, and for ColdFire devices that have an FPU@. It defines the
11541 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11542 on ColdFire targets.
11545 @opindex msoft-float
11546 Do not generate floating-point instructions; use library calls instead.
11547 This is the default for 68000, 68010, and 68832 targets. It is also
11548 the default for ColdFire devices that have no FPU.
11554 Generate (do not generate) ColdFire hardware divide and remainder
11555 instructions. If @option{-march} is used without @option{-mcpu},
11556 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11557 architectures. Otherwise, the default is taken from the target CPU
11558 (either the default CPU, or the one specified by @option{-mcpu}). For
11559 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11560 @option{-mcpu=5206e}.
11562 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11566 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11567 Additionally, parameters passed on the stack are also aligned to a
11568 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11572 Do not consider type @code{int} to be 16 bits wide. This is the default.
11575 @itemx -mno-bitfield
11576 @opindex mnobitfield
11577 @opindex mno-bitfield
11578 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11579 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11583 Do use the bit-field instructions. The @option{-m68020} option implies
11584 @option{-mbitfield}. This is the default if you use a configuration
11585 designed for a 68020.
11589 Use a different function-calling convention, in which functions
11590 that take a fixed number of arguments return with the @code{rtd}
11591 instruction, which pops their arguments while returning. This
11592 saves one instruction in the caller since there is no need to pop
11593 the arguments there.
11595 This calling convention is incompatible with the one normally
11596 used on Unix, so you cannot use it if you need to call libraries
11597 compiled with the Unix compiler.
11599 Also, you must provide function prototypes for all functions that
11600 take variable numbers of arguments (including @code{printf});
11601 otherwise incorrect code will be generated for calls to those
11604 In addition, seriously incorrect code will result if you call a
11605 function with too many arguments. (Normally, extra arguments are
11606 harmlessly ignored.)
11608 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
11609 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11613 Do not use the calling conventions selected by @option{-mrtd}.
11614 This is the default.
11617 @itemx -mno-align-int
11618 @opindex malign-int
11619 @opindex mno-align-int
11620 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
11621 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
11622 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
11623 Aligning variables on 32-bit boundaries produces code that runs somewhat
11624 faster on processors with 32-bit busses at the expense of more memory.
11626 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
11627 align structures containing the above types differently than
11628 most published application binary interface specifications for the m68k.
11632 Use the pc-relative addressing mode of the 68000 directly, instead of
11633 using a global offset table. At present, this option implies @option{-fpic},
11634 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
11635 not presently supported with @option{-mpcrel}, though this could be supported for
11636 68020 and higher processors.
11638 @item -mno-strict-align
11639 @itemx -mstrict-align
11640 @opindex mno-strict-align
11641 @opindex mstrict-align
11642 Do not (do) assume that unaligned memory references will be handled by
11646 Generate code that allows the data segment to be located in a different
11647 area of memory from the text segment. This allows for execute in place in
11648 an environment without virtual memory management. This option implies
11651 @item -mno-sep-data
11652 Generate code that assumes that the data segment follows the text segment.
11653 This is the default.
11655 @item -mid-shared-library
11656 Generate code that supports shared libraries via the library ID method.
11657 This allows for execute in place and shared libraries in an environment
11658 without virtual memory management. This option implies @option{-fPIC}.
11660 @item -mno-id-shared-library
11661 Generate code that doesn't assume ID based shared libraries are being used.
11662 This is the default.
11664 @item -mshared-library-id=n
11665 Specified the identification number of the ID based shared library being
11666 compiled. Specifying a value of 0 will generate more compact code, specifying
11667 other values will force the allocation of that number to the current
11668 library but is no more space or time efficient than omitting this option.
11672 @node M68hc1x Options
11673 @subsection M68hc1x Options
11674 @cindex M68hc1x options
11676 These are the @samp{-m} options defined for the 68hc11 and 68hc12
11677 microcontrollers. The default values for these options depends on
11678 which style of microcontroller was selected when the compiler was configured;
11679 the defaults for the most common choices are given below.
11686 Generate output for a 68HC11. This is the default
11687 when the compiler is configured for 68HC11-based systems.
11693 Generate output for a 68HC12. This is the default
11694 when the compiler is configured for 68HC12-based systems.
11700 Generate output for a 68HCS12.
11702 @item -mauto-incdec
11703 @opindex mauto-incdec
11704 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
11711 Enable the use of 68HC12 min and max instructions.
11714 @itemx -mno-long-calls
11715 @opindex mlong-calls
11716 @opindex mno-long-calls
11717 Treat all calls as being far away (near). If calls are assumed to be
11718 far away, the compiler will use the @code{call} instruction to
11719 call a function and the @code{rtc} instruction for returning.
11723 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11725 @item -msoft-reg-count=@var{count}
11726 @opindex msoft-reg-count
11727 Specify the number of pseudo-soft registers which are used for the
11728 code generation. The maximum number is 32. Using more pseudo-soft
11729 register may or may not result in better code depending on the program.
11730 The default is 4 for 68HC11 and 2 for 68HC12.
11734 @node MCore Options
11735 @subsection MCore Options
11736 @cindex MCore options
11738 These are the @samp{-m} options defined for the Motorola M*Core
11744 @itemx -mno-hardlit
11746 @opindex mno-hardlit
11747 Inline constants into the code stream if it can be done in two
11748 instructions or less.
11754 Use the divide instruction. (Enabled by default).
11756 @item -mrelax-immediate
11757 @itemx -mno-relax-immediate
11758 @opindex mrelax-immediate
11759 @opindex mno-relax-immediate
11760 Allow arbitrary sized immediates in bit operations.
11762 @item -mwide-bitfields
11763 @itemx -mno-wide-bitfields
11764 @opindex mwide-bitfields
11765 @opindex mno-wide-bitfields
11766 Always treat bit-fields as int-sized.
11768 @item -m4byte-functions
11769 @itemx -mno-4byte-functions
11770 @opindex m4byte-functions
11771 @opindex mno-4byte-functions
11772 Force all functions to be aligned to a four byte boundary.
11774 @item -mcallgraph-data
11775 @itemx -mno-callgraph-data
11776 @opindex mcallgraph-data
11777 @opindex mno-callgraph-data
11778 Emit callgraph information.
11781 @itemx -mno-slow-bytes
11782 @opindex mslow-bytes
11783 @opindex mno-slow-bytes
11784 Prefer word access when reading byte quantities.
11786 @item -mlittle-endian
11787 @itemx -mbig-endian
11788 @opindex mlittle-endian
11789 @opindex mbig-endian
11790 Generate code for a little endian target.
11796 Generate code for the 210 processor.
11800 @subsection MIPS Options
11801 @cindex MIPS options
11807 Generate big-endian code.
11811 Generate little-endian code. This is the default for @samp{mips*el-*-*}
11814 @item -march=@var{arch}
11816 Generate code that will run on @var{arch}, which can be the name of a
11817 generic MIPS ISA, or the name of a particular processor.
11819 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
11820 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
11821 The processor names are:
11822 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
11823 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
11824 @samp{5kc}, @samp{5kf},
11826 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
11827 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
11828 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
11829 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
11832 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
11833 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
11834 @samp{rm7000}, @samp{rm9000},
11837 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
11838 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
11839 The special value @samp{from-abi} selects the
11840 most compatible architecture for the selected ABI (that is,
11841 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
11843 In processor names, a final @samp{000} can be abbreviated as @samp{k}
11844 (for example, @samp{-march=r2k}). Prefixes are optional, and
11845 @samp{vr} may be written @samp{r}.
11847 Names of the form @samp{@var{n}f2_1} refer to processors with
11848 FPUs clocked at half the rate of the core, names of the form
11849 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
11850 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
11851 processors with FPUs clocked a ratio of 3:2 with respect to the core.
11852 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
11853 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
11854 accepted as synonyms for @samp{@var{n}f1_1}.
11856 GCC defines two macros based on the value of this option. The first
11857 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
11858 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
11859 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
11860 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
11861 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
11863 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
11864 above. In other words, it will have the full prefix and will not
11865 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
11866 the macro names the resolved architecture (either @samp{"mips1"} or
11867 @samp{"mips3"}). It names the default architecture when no
11868 @option{-march} option is given.
11870 @item -mtune=@var{arch}
11872 Optimize for @var{arch}. Among other things, this option controls
11873 the way instructions are scheduled, and the perceived cost of arithmetic
11874 operations. The list of @var{arch} values is the same as for
11877 When this option is not used, GCC will optimize for the processor
11878 specified by @option{-march}. By using @option{-march} and
11879 @option{-mtune} together, it is possible to generate code that will
11880 run on a family of processors, but optimize the code for one
11881 particular member of that family.
11883 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
11884 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
11885 @samp{-march} ones described above.
11889 Equivalent to @samp{-march=mips1}.
11893 Equivalent to @samp{-march=mips2}.
11897 Equivalent to @samp{-march=mips3}.
11901 Equivalent to @samp{-march=mips4}.
11905 Equivalent to @samp{-march=mips32}.
11909 Equivalent to @samp{-march=mips32r2}.
11913 Equivalent to @samp{-march=mips64}.
11918 @opindex mno-mips16
11919 Generate (do not generate) MIPS16 code. If GCC is targetting a
11920 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
11922 MIPS16 code generation can also be controlled on a per-function basis
11923 by means of @code{mips16} and @code{nomips16} attributes.
11924 @xref{Function Attributes}, for more information.
11926 @item -mflip-mips16
11927 @opindex mflip-mips16
11928 Generate MIPS16 code on alternating functions. This option is provided
11929 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
11930 not intended for ordinary use in compiling user code.
11932 @item -minterlink-mips16
11933 @itemx -mno-interlink-mips16
11934 @opindex minterlink-mips16
11935 @opindex mno-interlink-mips16
11936 Require (do not require) that non-MIPS16 code be link-compatible with
11939 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
11940 it must either use a call or an indirect jump. @option{-minterlink-mips16}
11941 therefore disables direct jumps unless GCC knows that the target of the
11942 jump is not MIPS16.
11954 Generate code for the given ABI@.
11956 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
11957 generates 64-bit code when you select a 64-bit architecture, but you
11958 can use @option{-mgp32} to get 32-bit code instead.
11960 For information about the O64 ABI, see
11961 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
11963 GCC supports a variant of the o32 ABI in which floating-point registers
11964 are 64 rather than 32 bits wide. You can select this combination with
11965 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
11966 and @samp{mfhc1} instructions and is therefore only supported for
11967 MIPS32R2 processors.
11969 The register assignments for arguments and return values remain the
11970 same, but each scalar value is passed in a single 64-bit register
11971 rather than a pair of 32-bit registers. For example, scalar
11972 floating-point values are returned in @samp{$f0} only, not a
11973 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
11974 remains the same, but all 64 bits are saved.
11977 @itemx -mno-abicalls
11979 @opindex mno-abicalls
11980 Generate (do not generate) code that is suitable for SVR4-style
11981 dynamic objects. @option{-mabicalls} is the default for SVR4-based
11986 Generate (do not generate) code that is fully position-independent,
11987 and that can therefore be linked into shared libraries. This option
11988 only affects @option{-mabicalls}.
11990 All @option{-mabicalls} code has traditionally been position-independent,
11991 regardless of options like @option{-fPIC} and @option{-fpic}. However,
11992 as an extension, the GNU toolchain allows executables to use absolute
11993 accesses for locally-binding symbols. It can also use shorter GP
11994 initialization sequences and generate direct calls to locally-defined
11995 functions. This mode is selected by @option{-mno-shared}.
11997 @option{-mno-shared} depends on binutils 2.16 or higher and generates
11998 objects that can only be linked by the GNU linker. However, the option
11999 does not affect the ABI of the final executable; it only affects the ABI
12000 of relocatable objects. Using @option{-mno-shared} will generally make
12001 executables both smaller and quicker.
12003 @option{-mshared} is the default.
12009 Lift (do not lift) the usual restrictions on the size of the global
12012 GCC normally uses a single instruction to load values from the GOT@.
12013 While this is relatively efficient, it will only work if the GOT
12014 is smaller than about 64k. Anything larger will cause the linker
12015 to report an error such as:
12017 @cindex relocation truncated to fit (MIPS)
12019 relocation truncated to fit: R_MIPS_GOT16 foobar
12022 If this happens, you should recompile your code with @option{-mxgot}.
12023 It should then work with very large GOTs, although it will also be
12024 less efficient, since it will take three instructions to fetch the
12025 value of a global symbol.
12027 Note that some linkers can create multiple GOTs. If you have such a
12028 linker, you should only need to use @option{-mxgot} when a single object
12029 file accesses more than 64k's worth of GOT entries. Very few do.
12031 These options have no effect unless GCC is generating position
12036 Assume that general-purpose registers are 32 bits wide.
12040 Assume that general-purpose registers are 64 bits wide.
12044 Assume that floating-point registers are 32 bits wide.
12048 Assume that floating-point registers are 64 bits wide.
12051 @opindex mhard-float
12052 Use floating-point coprocessor instructions.
12055 @opindex msoft-float
12056 Do not use floating-point coprocessor instructions. Implement
12057 floating-point calculations using library calls instead.
12059 @item -msingle-float
12060 @opindex msingle-float
12061 Assume that the floating-point coprocessor only supports single-precision
12064 @item -mdouble-float
12065 @opindex mdouble-float
12066 Assume that the floating-point coprocessor supports double-precision
12067 operations. This is the default.
12073 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12074 implement atomic memory built-in functions. When neither option is
12075 specified, GCC will use the instructions if the target architecture
12078 @option{-mllsc} is useful if the runtime environment can emulate the
12079 instructions and @option{-mno-llsc} can be useful when compiling for
12080 nonstandard ISAs. You can make either option the default by
12081 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12082 respectively. @option{--with-llsc} is the default for some
12083 configurations; see the installation documentation for details.
12089 Use (do not use) revision 1 of the MIPS DSP ASE@.
12090 @xref{MIPS DSP Built-in Functions}. This option defines the
12091 preprocessor macro @samp{__mips_dsp}. It also defines
12092 @samp{__mips_dsp_rev} to 1.
12098 Use (do not use) revision 2 of the MIPS DSP ASE@.
12099 @xref{MIPS DSP Built-in Functions}. This option defines the
12100 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12101 It also defines @samp{__mips_dsp_rev} to 2.
12104 @itemx -mno-smartmips
12105 @opindex msmartmips
12106 @opindex mno-smartmips
12107 Use (do not use) the MIPS SmartMIPS ASE.
12109 @item -mpaired-single
12110 @itemx -mno-paired-single
12111 @opindex mpaired-single
12112 @opindex mno-paired-single
12113 Use (do not use) paired-single floating-point instructions.
12114 @xref{MIPS Paired-Single Support}. This option requires
12115 hardware floating-point support to be enabled.
12121 Use (do not use) MIPS Digital Media Extension instructions.
12122 This option can only be used when generating 64-bit code and requires
12123 hardware floating-point support to be enabled.
12128 @opindex mno-mips3d
12129 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12130 The option @option{-mips3d} implies @option{-mpaired-single}.
12136 Use (do not use) MT Multithreading instructions.
12140 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12141 an explanation of the default and the way that the pointer size is
12146 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12148 The default size of @code{int}s, @code{long}s and pointers depends on
12149 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12150 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12151 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12152 or the same size as integer registers, whichever is smaller.
12158 Assume (do not assume) that all symbols have 32-bit values, regardless
12159 of the selected ABI@. This option is useful in combination with
12160 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12161 to generate shorter and faster references to symbolic addresses.
12165 Put definitions of externally-visible data in a small data section
12166 if that data is no bigger than @var{num} bytes. GCC can then access
12167 the data more efficiently; see @option{-mgpopt} for details.
12169 The default @option{-G} option depends on the configuration.
12171 @item -mlocal-sdata
12172 @itemx -mno-local-sdata
12173 @opindex mlocal-sdata
12174 @opindex mno-local-sdata
12175 Extend (do not extend) the @option{-G} behavior to local data too,
12176 such as to static variables in C@. @option{-mlocal-sdata} is the
12177 default for all configurations.
12179 If the linker complains that an application is using too much small data,
12180 you might want to try rebuilding the less performance-critical parts with
12181 @option{-mno-local-sdata}. You might also want to build large
12182 libraries with @option{-mno-local-sdata}, so that the libraries leave
12183 more room for the main program.
12185 @item -mextern-sdata
12186 @itemx -mno-extern-sdata
12187 @opindex mextern-sdata
12188 @opindex mno-extern-sdata
12189 Assume (do not assume) that externally-defined data will be in
12190 a small data section if that data is within the @option{-G} limit.
12191 @option{-mextern-sdata} is the default for all configurations.
12193 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12194 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12195 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12196 is placed in a small data section. If @var{Var} is defined by another
12197 module, you must either compile that module with a high-enough
12198 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12199 definition. If @var{Var} is common, you must link the application
12200 with a high-enough @option{-G} setting.
12202 The easiest way of satisfying these restrictions is to compile
12203 and link every module with the same @option{-G} option. However,
12204 you may wish to build a library that supports several different
12205 small data limits. You can do this by compiling the library with
12206 the highest supported @option{-G} setting and additionally using
12207 @option{-mno-extern-sdata} to stop the library from making assumptions
12208 about externally-defined data.
12214 Use (do not use) GP-relative accesses for symbols that are known to be
12215 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12216 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12219 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12220 might not hold the value of @code{_gp}. For example, if the code is
12221 part of a library that might be used in a boot monitor, programs that
12222 call boot monitor routines will pass an unknown value in @code{$gp}.
12223 (In such situations, the boot monitor itself would usually be compiled
12224 with @option{-G0}.)
12226 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12227 @option{-mno-extern-sdata}.
12229 @item -membedded-data
12230 @itemx -mno-embedded-data
12231 @opindex membedded-data
12232 @opindex mno-embedded-data
12233 Allocate variables to the read-only data section first if possible, then
12234 next in the small data section if possible, otherwise in data. This gives
12235 slightly slower code than the default, but reduces the amount of RAM required
12236 when executing, and thus may be preferred for some embedded systems.
12238 @item -muninit-const-in-rodata
12239 @itemx -mno-uninit-const-in-rodata
12240 @opindex muninit-const-in-rodata
12241 @opindex mno-uninit-const-in-rodata
12242 Put uninitialized @code{const} variables in the read-only data section.
12243 This option is only meaningful in conjunction with @option{-membedded-data}.
12245 @item -mcode-readable=@var{setting}
12246 @opindex mcode-readable
12247 Specify whether GCC may generate code that reads from executable sections.
12248 There are three possible settings:
12251 @item -mcode-readable=yes
12252 Instructions may freely access executable sections. This is the
12255 @item -mcode-readable=pcrel
12256 MIPS16 PC-relative load instructions can access executable sections,
12257 but other instructions must not do so. This option is useful on 4KSc
12258 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12259 It is also useful on processors that can be configured to have a dual
12260 instruction/data SRAM interface and that, like the M4K, automatically
12261 redirect PC-relative loads to the instruction RAM.
12263 @item -mcode-readable=no
12264 Instructions must not access executable sections. This option can be
12265 useful on targets that are configured to have a dual instruction/data
12266 SRAM interface but that (unlike the M4K) do not automatically redirect
12267 PC-relative loads to the instruction RAM.
12270 @item -msplit-addresses
12271 @itemx -mno-split-addresses
12272 @opindex msplit-addresses
12273 @opindex mno-split-addresses
12274 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12275 relocation operators. This option has been superseded by
12276 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12278 @item -mexplicit-relocs
12279 @itemx -mno-explicit-relocs
12280 @opindex mexplicit-relocs
12281 @opindex mno-explicit-relocs
12282 Use (do not use) assembler relocation operators when dealing with symbolic
12283 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12284 is to use assembler macros instead.
12286 @option{-mexplicit-relocs} is the default if GCC was configured
12287 to use an assembler that supports relocation operators.
12289 @item -mcheck-zero-division
12290 @itemx -mno-check-zero-division
12291 @opindex mcheck-zero-division
12292 @opindex mno-check-zero-division
12293 Trap (do not trap) on integer division by zero.
12295 The default is @option{-mcheck-zero-division}.
12297 @item -mdivide-traps
12298 @itemx -mdivide-breaks
12299 @opindex mdivide-traps
12300 @opindex mdivide-breaks
12301 MIPS systems check for division by zero by generating either a
12302 conditional trap or a break instruction. Using traps results in
12303 smaller code, but is only supported on MIPS II and later. Also, some
12304 versions of the Linux kernel have a bug that prevents trap from
12305 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12306 allow conditional traps on architectures that support them and
12307 @option{-mdivide-breaks} to force the use of breaks.
12309 The default is usually @option{-mdivide-traps}, but this can be
12310 overridden at configure time using @option{--with-divide=breaks}.
12311 Divide-by-zero checks can be completely disabled using
12312 @option{-mno-check-zero-division}.
12317 @opindex mno-memcpy
12318 Force (do not force) the use of @code{memcpy()} for non-trivial block
12319 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12320 most constant-sized copies.
12323 @itemx -mno-long-calls
12324 @opindex mlong-calls
12325 @opindex mno-long-calls
12326 Disable (do not disable) use of the @code{jal} instruction. Calling
12327 functions using @code{jal} is more efficient but requires the caller
12328 and callee to be in the same 256 megabyte segment.
12330 This option has no effect on abicalls code. The default is
12331 @option{-mno-long-calls}.
12337 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12338 instructions, as provided by the R4650 ISA@.
12341 @itemx -mno-fused-madd
12342 @opindex mfused-madd
12343 @opindex mno-fused-madd
12344 Enable (disable) use of the floating point multiply-accumulate
12345 instructions, when they are available. The default is
12346 @option{-mfused-madd}.
12348 When multiply-accumulate instructions are used, the intermediate
12349 product is calculated to infinite precision and is not subject to
12350 the FCSR Flush to Zero bit. This may be undesirable in some
12355 Tell the MIPS assembler to not run its preprocessor over user
12356 assembler files (with a @samp{.s} suffix) when assembling them.
12359 @itemx -mno-fix-r4000
12360 @opindex mfix-r4000
12361 @opindex mno-fix-r4000
12362 Work around certain R4000 CPU errata:
12365 A double-word or a variable shift may give an incorrect result if executed
12366 immediately after starting an integer division.
12368 A double-word or a variable shift may give an incorrect result if executed
12369 while an integer multiplication is in progress.
12371 An integer division may give an incorrect result if started in a delay slot
12372 of a taken branch or a jump.
12376 @itemx -mno-fix-r4400
12377 @opindex mfix-r4400
12378 @opindex mno-fix-r4400
12379 Work around certain R4400 CPU errata:
12382 A double-word or a variable shift may give an incorrect result if executed
12383 immediately after starting an integer division.
12387 @itemx -mno-fix-vr4120
12388 @opindex mfix-vr4120
12389 Work around certain VR4120 errata:
12392 @code{dmultu} does not always produce the correct result.
12394 @code{div} and @code{ddiv} do not always produce the correct result if one
12395 of the operands is negative.
12397 The workarounds for the division errata rely on special functions in
12398 @file{libgcc.a}. At present, these functions are only provided by
12399 the @code{mips64vr*-elf} configurations.
12401 Other VR4120 errata require a nop to be inserted between certain pairs of
12402 instructions. These errata are handled by the assembler, not by GCC itself.
12405 @opindex mfix-vr4130
12406 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
12407 workarounds are implemented by the assembler rather than by GCC,
12408 although GCC will avoid using @code{mflo} and @code{mfhi} if the
12409 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
12410 instructions are available instead.
12413 @itemx -mno-fix-sb1
12415 Work around certain SB-1 CPU core errata.
12416 (This flag currently works around the SB-1 revision 2
12417 ``F1'' and ``F2'' floating point errata.)
12419 @item -mflush-func=@var{func}
12420 @itemx -mno-flush-func
12421 @opindex mflush-func
12422 Specifies the function to call to flush the I and D caches, or to not
12423 call any such function. If called, the function must take the same
12424 arguments as the common @code{_flush_func()}, that is, the address of the
12425 memory range for which the cache is being flushed, the size of the
12426 memory range, and the number 3 (to flush both caches). The default
12427 depends on the target GCC was configured for, but commonly is either
12428 @samp{_flush_func} or @samp{__cpu_flush}.
12430 @item mbranch-cost=@var{num}
12431 @opindex mbranch-cost
12432 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12433 This cost is only a heuristic and is not guaranteed to produce
12434 consistent results across releases. A zero cost redundantly selects
12435 the default, which is based on the @option{-mtune} setting.
12437 @item -mbranch-likely
12438 @itemx -mno-branch-likely
12439 @opindex mbranch-likely
12440 @opindex mno-branch-likely
12441 Enable or disable use of Branch Likely instructions, regardless of the
12442 default for the selected architecture. By default, Branch Likely
12443 instructions may be generated if they are supported by the selected
12444 architecture. An exception is for the MIPS32 and MIPS64 architectures
12445 and processors which implement those architectures; for those, Branch
12446 Likely instructions will not be generated by default because the MIPS32
12447 and MIPS64 architectures specifically deprecate their use.
12449 @item -mfp-exceptions
12450 @itemx -mno-fp-exceptions
12451 @opindex mfp-exceptions
12452 Specifies whether FP exceptions are enabled. This affects how we schedule
12453 FP instructions for some processors. The default is that FP exceptions are
12456 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
12457 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
12460 @item -mvr4130-align
12461 @itemx -mno-vr4130-align
12462 @opindex mvr4130-align
12463 The VR4130 pipeline is two-way superscalar, but can only issue two
12464 instructions together if the first one is 8-byte aligned. When this
12465 option is enabled, GCC will align pairs of instructions that it
12466 thinks should execute in parallel.
12468 This option only has an effect when optimizing for the VR4130.
12469 It normally makes code faster, but at the expense of making it bigger.
12470 It is enabled by default at optimization level @option{-O3}.
12474 @subsection MMIX Options
12475 @cindex MMIX Options
12477 These options are defined for the MMIX:
12481 @itemx -mno-libfuncs
12483 @opindex mno-libfuncs
12484 Specify that intrinsic library functions are being compiled, passing all
12485 values in registers, no matter the size.
12488 @itemx -mno-epsilon
12490 @opindex mno-epsilon
12491 Generate floating-point comparison instructions that compare with respect
12492 to the @code{rE} epsilon register.
12494 @item -mabi=mmixware
12496 @opindex mabi-mmixware
12498 Generate code that passes function parameters and return values that (in
12499 the called function) are seen as registers @code{$0} and up, as opposed to
12500 the GNU ABI which uses global registers @code{$231} and up.
12502 @item -mzero-extend
12503 @itemx -mno-zero-extend
12504 @opindex mzero-extend
12505 @opindex mno-zero-extend
12506 When reading data from memory in sizes shorter than 64 bits, use (do not
12507 use) zero-extending load instructions by default, rather than
12508 sign-extending ones.
12511 @itemx -mno-knuthdiv
12513 @opindex mno-knuthdiv
12514 Make the result of a division yielding a remainder have the same sign as
12515 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
12516 remainder follows the sign of the dividend. Both methods are
12517 arithmetically valid, the latter being almost exclusively used.
12519 @item -mtoplevel-symbols
12520 @itemx -mno-toplevel-symbols
12521 @opindex mtoplevel-symbols
12522 @opindex mno-toplevel-symbols
12523 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
12524 code can be used with the @code{PREFIX} assembly directive.
12528 Generate an executable in the ELF format, rather than the default
12529 @samp{mmo} format used by the @command{mmix} simulator.
12531 @item -mbranch-predict
12532 @itemx -mno-branch-predict
12533 @opindex mbranch-predict
12534 @opindex mno-branch-predict
12535 Use (do not use) the probable-branch instructions, when static branch
12536 prediction indicates a probable branch.
12538 @item -mbase-addresses
12539 @itemx -mno-base-addresses
12540 @opindex mbase-addresses
12541 @opindex mno-base-addresses
12542 Generate (do not generate) code that uses @emph{base addresses}. Using a
12543 base address automatically generates a request (handled by the assembler
12544 and the linker) for a constant to be set up in a global register. The
12545 register is used for one or more base address requests within the range 0
12546 to 255 from the value held in the register. The generally leads to short
12547 and fast code, but the number of different data items that can be
12548 addressed is limited. This means that a program that uses lots of static
12549 data may require @option{-mno-base-addresses}.
12551 @item -msingle-exit
12552 @itemx -mno-single-exit
12553 @opindex msingle-exit
12554 @opindex mno-single-exit
12555 Force (do not force) generated code to have a single exit point in each
12559 @node MN10300 Options
12560 @subsection MN10300 Options
12561 @cindex MN10300 options
12563 These @option{-m} options are defined for Matsushita MN10300 architectures:
12568 Generate code to avoid bugs in the multiply instructions for the MN10300
12569 processors. This is the default.
12571 @item -mno-mult-bug
12572 @opindex mno-mult-bug
12573 Do not generate code to avoid bugs in the multiply instructions for the
12574 MN10300 processors.
12578 Generate code which uses features specific to the AM33 processor.
12582 Do not generate code which uses features specific to the AM33 processor. This
12585 @item -mreturn-pointer-on-d0
12586 @opindex mreturn-pointer-on-d0
12587 When generating a function which returns a pointer, return the pointer
12588 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
12589 only in a0, and attempts to call such functions without a prototype
12590 would result in errors. Note that this option is on by default; use
12591 @option{-mno-return-pointer-on-d0} to disable it.
12595 Do not link in the C run-time initialization object file.
12599 Indicate to the linker that it should perform a relaxation optimization pass
12600 to shorten branches, calls and absolute memory addresses. This option only
12601 has an effect when used on the command line for the final link step.
12603 This option makes symbolic debugging impossible.
12607 @subsection MT Options
12610 These @option{-m} options are defined for Morpho MT architectures:
12614 @item -march=@var{cpu-type}
12616 Generate code that will run on @var{cpu-type}, which is the name of a system
12617 representing a certain processor type. Possible values for
12618 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
12619 @samp{ms1-16-003} and @samp{ms2}.
12621 When this option is not used, the default is @option{-march=ms1-16-002}.
12625 Use byte loads and stores when generating code.
12629 Do not use byte loads and stores when generating code.
12633 Use simulator runtime
12637 Do not link in the C run-time initialization object file
12638 @file{crti.o}. Other run-time initialization and termination files
12639 such as @file{startup.o} and @file{exit.o} are still included on the
12640 linker command line.
12644 @node PDP-11 Options
12645 @subsection PDP-11 Options
12646 @cindex PDP-11 Options
12648 These options are defined for the PDP-11:
12653 Use hardware FPP floating point. This is the default. (FIS floating
12654 point on the PDP-11/40 is not supported.)
12657 @opindex msoft-float
12658 Do not use hardware floating point.
12662 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
12666 Return floating-point results in memory. This is the default.
12670 Generate code for a PDP-11/40.
12674 Generate code for a PDP-11/45. This is the default.
12678 Generate code for a PDP-11/10.
12680 @item -mbcopy-builtin
12681 @opindex bcopy-builtin
12682 Use inline @code{movmemhi} patterns for copying memory. This is the
12687 Do not use inline @code{movmemhi} patterns for copying memory.
12693 Use 16-bit @code{int}. This is the default.
12699 Use 32-bit @code{int}.
12702 @itemx -mno-float32
12704 @opindex mno-float32
12705 Use 64-bit @code{float}. This is the default.
12708 @itemx -mno-float64
12710 @opindex mno-float64
12711 Use 32-bit @code{float}.
12715 Use @code{abshi2} pattern. This is the default.
12719 Do not use @code{abshi2} pattern.
12721 @item -mbranch-expensive
12722 @opindex mbranch-expensive
12723 Pretend that branches are expensive. This is for experimenting with
12724 code generation only.
12726 @item -mbranch-cheap
12727 @opindex mbranch-cheap
12728 Do not pretend that branches are expensive. This is the default.
12732 Generate code for a system with split I&D@.
12736 Generate code for a system without split I&D@. This is the default.
12740 Use Unix assembler syntax. This is the default when configured for
12741 @samp{pdp11-*-bsd}.
12745 Use DEC assembler syntax. This is the default when configured for any
12746 PDP-11 target other than @samp{pdp11-*-bsd}.
12749 @node PowerPC Options
12750 @subsection PowerPC Options
12751 @cindex PowerPC options
12753 These are listed under @xref{RS/6000 and PowerPC Options}.
12755 @node RS/6000 and PowerPC Options
12756 @subsection IBM RS/6000 and PowerPC Options
12757 @cindex RS/6000 and PowerPC Options
12758 @cindex IBM RS/6000 and PowerPC Options
12760 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
12767 @itemx -mno-powerpc
12768 @itemx -mpowerpc-gpopt
12769 @itemx -mno-powerpc-gpopt
12770 @itemx -mpowerpc-gfxopt
12771 @itemx -mno-powerpc-gfxopt
12773 @itemx -mno-powerpc64
12777 @itemx -mno-popcntb
12785 @itemx -mno-hard-dfp
12789 @opindex mno-power2
12791 @opindex mno-powerpc
12792 @opindex mpowerpc-gpopt
12793 @opindex mno-powerpc-gpopt
12794 @opindex mpowerpc-gfxopt
12795 @opindex mno-powerpc-gfxopt
12796 @opindex mpowerpc64
12797 @opindex mno-powerpc64
12801 @opindex mno-popcntb
12807 @opindex mno-mfpgpr
12809 @opindex mno-hard-dfp
12810 GCC supports two related instruction set architectures for the
12811 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
12812 instructions supported by the @samp{rios} chip set used in the original
12813 RS/6000 systems and the @dfn{PowerPC} instruction set is the
12814 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
12815 the IBM 4xx, 6xx, and follow-on microprocessors.
12817 Neither architecture is a subset of the other. However there is a
12818 large common subset of instructions supported by both. An MQ
12819 register is included in processors supporting the POWER architecture.
12821 You use these options to specify which instructions are available on the
12822 processor you are using. The default value of these options is
12823 determined when configuring GCC@. Specifying the
12824 @option{-mcpu=@var{cpu_type}} overrides the specification of these
12825 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
12826 rather than the options listed above.
12828 The @option{-mpower} option allows GCC to generate instructions that
12829 are found only in the POWER architecture and to use the MQ register.
12830 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
12831 to generate instructions that are present in the POWER2 architecture but
12832 not the original POWER architecture.
12834 The @option{-mpowerpc} option allows GCC to generate instructions that
12835 are found only in the 32-bit subset of the PowerPC architecture.
12836 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
12837 GCC to use the optional PowerPC architecture instructions in the
12838 General Purpose group, including floating-point square root. Specifying
12839 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
12840 use the optional PowerPC architecture instructions in the Graphics
12841 group, including floating-point select.
12843 The @option{-mmfcrf} option allows GCC to generate the move from
12844 condition register field instruction implemented on the POWER4
12845 processor and other processors that support the PowerPC V2.01
12847 The @option{-mpopcntb} option allows GCC to generate the popcount and
12848 double precision FP reciprocal estimate instruction implemented on the
12849 POWER5 processor and other processors that support the PowerPC V2.02
12851 The @option{-mfprnd} option allows GCC to generate the FP round to
12852 integer instructions implemented on the POWER5+ processor and other
12853 processors that support the PowerPC V2.03 architecture.
12854 The @option{-mcmpb} option allows GCC to generate the compare bytes
12855 instruction implemented on the POWER6 processor and other processors
12856 that support the PowerPC V2.05 architecture.
12857 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
12858 general purpose register instructions implemented on the POWER6X
12859 processor and other processors that support the extended PowerPC V2.05
12861 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
12862 point instructions implemented on some POWER processors.
12864 The @option{-mpowerpc64} option allows GCC to generate the additional
12865 64-bit instructions that are found in the full PowerPC64 architecture
12866 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
12867 @option{-mno-powerpc64}.
12869 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
12870 will use only the instructions in the common subset of both
12871 architectures plus some special AIX common-mode calls, and will not use
12872 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
12873 permits GCC to use any instruction from either architecture and to
12874 allow use of the MQ register; specify this for the Motorola MPC601.
12876 @item -mnew-mnemonics
12877 @itemx -mold-mnemonics
12878 @opindex mnew-mnemonics
12879 @opindex mold-mnemonics
12880 Select which mnemonics to use in the generated assembler code. With
12881 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
12882 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
12883 assembler mnemonics defined for the POWER architecture. Instructions
12884 defined in only one architecture have only one mnemonic; GCC uses that
12885 mnemonic irrespective of which of these options is specified.
12887 GCC defaults to the mnemonics appropriate for the architecture in
12888 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
12889 value of these option. Unless you are building a cross-compiler, you
12890 should normally not specify either @option{-mnew-mnemonics} or
12891 @option{-mold-mnemonics}, but should instead accept the default.
12893 @item -mcpu=@var{cpu_type}
12895 Set architecture type, register usage, choice of mnemonics, and
12896 instruction scheduling parameters for machine type @var{cpu_type}.
12897 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
12898 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
12899 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
12900 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
12901 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
12902 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
12903 @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5}, @samp{power},
12904 @samp{power2}, @samp{power3}, @samp{power4}, @samp{power5},
12905 @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{common},
12906 @samp{powerpc}, @samp{powerpc64}, @samp{rios}, @samp{rios1},
12907 @samp{rios2}, @samp{rsc}, and @samp{rs64}.
12909 @option{-mcpu=common} selects a completely generic processor. Code
12910 generated under this option will run on any POWER or PowerPC processor.
12911 GCC will use only the instructions in the common subset of both
12912 architectures, and will not use the MQ register. GCC assumes a generic
12913 processor model for scheduling purposes.
12915 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
12916 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
12917 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
12918 types, with an appropriate, generic processor model assumed for
12919 scheduling purposes.
12921 The other options specify a specific processor. Code generated under
12922 those options will run best on that processor, and may not run at all on
12925 The @option{-mcpu} options automatically enable or disable the
12928 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
12929 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
12930 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
12932 The particular options set for any particular CPU will vary between
12933 compiler versions, depending on what setting seems to produce optimal
12934 code for that CPU; it doesn't necessarily reflect the actual hardware's
12935 capabilities. If you wish to set an individual option to a particular
12936 value, you may specify it after the @option{-mcpu} option, like
12937 @samp{-mcpu=970 -mno-altivec}.
12939 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
12940 not enabled or disabled by the @option{-mcpu} option at present because
12941 AIX does not have full support for these options. You may still
12942 enable or disable them individually if you're sure it'll work in your
12945 @item -mtune=@var{cpu_type}
12947 Set the instruction scheduling parameters for machine type
12948 @var{cpu_type}, but do not set the architecture type, register usage, or
12949 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
12950 values for @var{cpu_type} are used for @option{-mtune} as for
12951 @option{-mcpu}. If both are specified, the code generated will use the
12952 architecture, registers, and mnemonics set by @option{-mcpu}, but the
12953 scheduling parameters set by @option{-mtune}.
12959 Generate code to compute division as reciprocal estimate and iterative
12960 refinement, creating opportunities for increased throughput. This
12961 feature requires: optional PowerPC Graphics instruction set for single
12962 precision and FRE instruction for double precision, assuming divides
12963 cannot generate user-visible traps, and the domain values not include
12964 Infinities, denormals or zero denominator.
12967 @itemx -mno-altivec
12969 @opindex mno-altivec
12970 Generate code that uses (does not use) AltiVec instructions, and also
12971 enable the use of built-in functions that allow more direct access to
12972 the AltiVec instruction set. You may also need to set
12973 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
12979 @opindex mno-vrsave
12980 Generate VRSAVE instructions when generating AltiVec code.
12983 @opindex msecure-plt
12984 Generate code that allows ld and ld.so to build executables and shared
12985 libraries with non-exec .plt and .got sections. This is a PowerPC
12986 32-bit SYSV ABI option.
12990 Generate code that uses a BSS .plt section that ld.so fills in, and
12991 requires .plt and .got sections that are both writable and executable.
12992 This is a PowerPC 32-bit SYSV ABI option.
12998 This switch enables or disables the generation of ISEL instructions.
13000 @item -misel=@var{yes/no}
13001 This switch has been deprecated. Use @option{-misel} and
13002 @option{-mno-isel} instead.
13008 This switch enables or disables the generation of SPE simd
13014 @opindex mno-paired
13015 This switch enables or disables the generation of PAIRED simd
13018 @item -mspe=@var{yes/no}
13019 This option has been deprecated. Use @option{-mspe} and
13020 @option{-mno-spe} instead.
13022 @item -mfloat-gprs=@var{yes/single/double/no}
13023 @itemx -mfloat-gprs
13024 @opindex mfloat-gprs
13025 This switch enables or disables the generation of floating point
13026 operations on the general purpose registers for architectures that
13029 The argument @var{yes} or @var{single} enables the use of
13030 single-precision floating point operations.
13032 The argument @var{double} enables the use of single and
13033 double-precision floating point operations.
13035 The argument @var{no} disables floating point operations on the
13036 general purpose registers.
13038 This option is currently only available on the MPC854x.
13044 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13045 targets (including GNU/Linux). The 32-bit environment sets int, long
13046 and pointer to 32 bits and generates code that runs on any PowerPC
13047 variant. The 64-bit environment sets int to 32 bits and long and
13048 pointer to 64 bits, and generates code for PowerPC64, as for
13049 @option{-mpowerpc64}.
13052 @itemx -mno-fp-in-toc
13053 @itemx -mno-sum-in-toc
13054 @itemx -mminimal-toc
13056 @opindex mno-fp-in-toc
13057 @opindex mno-sum-in-toc
13058 @opindex mminimal-toc
13059 Modify generation of the TOC (Table Of Contents), which is created for
13060 every executable file. The @option{-mfull-toc} option is selected by
13061 default. In that case, GCC will allocate at least one TOC entry for
13062 each unique non-automatic variable reference in your program. GCC
13063 will also place floating-point constants in the TOC@. However, only
13064 16,384 entries are available in the TOC@.
13066 If you receive a linker error message that saying you have overflowed
13067 the available TOC space, you can reduce the amount of TOC space used
13068 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13069 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13070 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13071 generate code to calculate the sum of an address and a constant at
13072 run-time instead of putting that sum into the TOC@. You may specify one
13073 or both of these options. Each causes GCC to produce very slightly
13074 slower and larger code at the expense of conserving TOC space.
13076 If you still run out of space in the TOC even when you specify both of
13077 these options, specify @option{-mminimal-toc} instead. This option causes
13078 GCC to make only one TOC entry for every file. When you specify this
13079 option, GCC will produce code that is slower and larger but which
13080 uses extremely little TOC space. You may wish to use this option
13081 only on files that contain less frequently executed code.
13087 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13088 @code{long} type, and the infrastructure needed to support them.
13089 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13090 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13091 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13094 @itemx -mno-xl-compat
13095 @opindex mxl-compat
13096 @opindex mno-xl-compat
13097 Produce code that conforms more closely to IBM XL compiler semantics
13098 when using AIX-compatible ABI@. Pass floating-point arguments to
13099 prototyped functions beyond the register save area (RSA) on the stack
13100 in addition to argument FPRs. Do not assume that most significant
13101 double in 128-bit long double value is properly rounded when comparing
13102 values and converting to double. Use XL symbol names for long double
13105 The AIX calling convention was extended but not initially documented to
13106 handle an obscure K&R C case of calling a function that takes the
13107 address of its arguments with fewer arguments than declared. IBM XL
13108 compilers access floating point arguments which do not fit in the
13109 RSA from the stack when a subroutine is compiled without
13110 optimization. Because always storing floating-point arguments on the
13111 stack is inefficient and rarely needed, this option is not enabled by
13112 default and only is necessary when calling subroutines compiled by IBM
13113 XL compilers without optimization.
13117 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13118 application written to use message passing with special startup code to
13119 enable the application to run. The system must have PE installed in the
13120 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13121 must be overridden with the @option{-specs=} option to specify the
13122 appropriate directory location. The Parallel Environment does not
13123 support threads, so the @option{-mpe} option and the @option{-pthread}
13124 option are incompatible.
13126 @item -malign-natural
13127 @itemx -malign-power
13128 @opindex malign-natural
13129 @opindex malign-power
13130 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13131 @option{-malign-natural} overrides the ABI-defined alignment of larger
13132 types, such as floating-point doubles, on their natural size-based boundary.
13133 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13134 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13136 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13140 @itemx -mhard-float
13141 @opindex msoft-float
13142 @opindex mhard-float
13143 Generate code that does not use (uses) the floating-point register set.
13144 Software floating point emulation is provided if you use the
13145 @option{-msoft-float} option, and pass the option to GCC when linking.
13148 @itemx -mno-multiple
13150 @opindex mno-multiple
13151 Generate code that uses (does not use) the load multiple word
13152 instructions and the store multiple word instructions. These
13153 instructions are generated by default on POWER systems, and not
13154 generated on PowerPC systems. Do not use @option{-mmultiple} on little
13155 endian PowerPC systems, since those instructions do not work when the
13156 processor is in little endian mode. The exceptions are PPC740 and
13157 PPC750 which permit the instructions usage in little endian mode.
13162 @opindex mno-string
13163 Generate code that uses (does not use) the load string instructions
13164 and the store string word instructions to save multiple registers and
13165 do small block moves. These instructions are generated by default on
13166 POWER systems, and not generated on PowerPC systems. Do not use
13167 @option{-mstring} on little endian PowerPC systems, since those
13168 instructions do not work when the processor is in little endian mode.
13169 The exceptions are PPC740 and PPC750 which permit the instructions
13170 usage in little endian mode.
13175 @opindex mno-update
13176 Generate code that uses (does not use) the load or store instructions
13177 that update the base register to the address of the calculated memory
13178 location. These instructions are generated by default. If you use
13179 @option{-mno-update}, there is a small window between the time that the
13180 stack pointer is updated and the address of the previous frame is
13181 stored, which means code that walks the stack frame across interrupts or
13182 signals may get corrupted data.
13185 @itemx -mno-fused-madd
13186 @opindex mfused-madd
13187 @opindex mno-fused-madd
13188 Generate code that uses (does not use) the floating point multiply and
13189 accumulate instructions. These instructions are generated by default if
13190 hardware floating is used.
13196 Generate code that uses (does not use) the half-word multiply and
13197 multiply-accumulate instructions on the IBM 405 and 440 processors.
13198 These instructions are generated by default when targetting those
13205 Generate code that uses (does not use) the string-search @samp{dlmzb}
13206 instruction on the IBM 405 and 440 processors. This instruction is
13207 generated by default when targetting those processors.
13209 @item -mno-bit-align
13211 @opindex mno-bit-align
13212 @opindex mbit-align
13213 On System V.4 and embedded PowerPC systems do not (do) force structures
13214 and unions that contain bit-fields to be aligned to the base type of the
13217 For example, by default a structure containing nothing but 8
13218 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13219 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13220 the structure would be aligned to a 1 byte boundary and be one byte in
13223 @item -mno-strict-align
13224 @itemx -mstrict-align
13225 @opindex mno-strict-align
13226 @opindex mstrict-align
13227 On System V.4 and embedded PowerPC systems do not (do) assume that
13228 unaligned memory references will be handled by the system.
13230 @item -mrelocatable
13231 @itemx -mno-relocatable
13232 @opindex mrelocatable
13233 @opindex mno-relocatable
13234 On embedded PowerPC systems generate code that allows (does not allow)
13235 the program to be relocated to a different address at runtime. If you
13236 use @option{-mrelocatable} on any module, all objects linked together must
13237 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
13239 @item -mrelocatable-lib
13240 @itemx -mno-relocatable-lib
13241 @opindex mrelocatable-lib
13242 @opindex mno-relocatable-lib
13243 On embedded PowerPC systems generate code that allows (does not allow)
13244 the program to be relocated to a different address at runtime. Modules
13245 compiled with @option{-mrelocatable-lib} can be linked with either modules
13246 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
13247 with modules compiled with the @option{-mrelocatable} options.
13253 On System V.4 and embedded PowerPC systems do not (do) assume that
13254 register 2 contains a pointer to a global area pointing to the addresses
13255 used in the program.
13258 @itemx -mlittle-endian
13260 @opindex mlittle-endian
13261 On System V.4 and embedded PowerPC systems compile code for the
13262 processor in little endian mode. The @option{-mlittle-endian} option is
13263 the same as @option{-mlittle}.
13266 @itemx -mbig-endian
13268 @opindex mbig-endian
13269 On System V.4 and embedded PowerPC systems compile code for the
13270 processor in big endian mode. The @option{-mbig-endian} option is
13271 the same as @option{-mbig}.
13273 @item -mdynamic-no-pic
13274 @opindex mdynamic-no-pic
13275 On Darwin and Mac OS X systems, compile code so that it is not
13276 relocatable, but that its external references are relocatable. The
13277 resulting code is suitable for applications, but not shared
13280 @item -mprioritize-restricted-insns=@var{priority}
13281 @opindex mprioritize-restricted-insns
13282 This option controls the priority that is assigned to
13283 dispatch-slot restricted instructions during the second scheduling
13284 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
13285 @var{no/highest/second-highest} priority to dispatch slot restricted
13288 @item -msched-costly-dep=@var{dependence_type}
13289 @opindex msched-costly-dep
13290 This option controls which dependences are considered costly
13291 by the target during instruction scheduling. The argument
13292 @var{dependence_type} takes one of the following values:
13293 @var{no}: no dependence is costly,
13294 @var{all}: all dependences are costly,
13295 @var{true_store_to_load}: a true dependence from store to load is costly,
13296 @var{store_to_load}: any dependence from store to load is costly,
13297 @var{number}: any dependence which latency >= @var{number} is costly.
13299 @item -minsert-sched-nops=@var{scheme}
13300 @opindex minsert-sched-nops
13301 This option controls which nop insertion scheme will be used during
13302 the second scheduling pass. The argument @var{scheme} takes one of the
13304 @var{no}: Don't insert nops.
13305 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
13306 according to the scheduler's grouping.
13307 @var{regroup_exact}: Insert nops to force costly dependent insns into
13308 separate groups. Insert exactly as many nops as needed to force an insn
13309 to a new group, according to the estimated processor grouping.
13310 @var{number}: Insert nops to force costly dependent insns into
13311 separate groups. Insert @var{number} nops to force an insn to a new group.
13314 @opindex mcall-sysv
13315 On System V.4 and embedded PowerPC systems compile code using calling
13316 conventions that adheres to the March 1995 draft of the System V
13317 Application Binary Interface, PowerPC processor supplement. This is the
13318 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
13320 @item -mcall-sysv-eabi
13321 @opindex mcall-sysv-eabi
13322 Specify both @option{-mcall-sysv} and @option{-meabi} options.
13324 @item -mcall-sysv-noeabi
13325 @opindex mcall-sysv-noeabi
13326 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
13328 @item -mcall-solaris
13329 @opindex mcall-solaris
13330 On System V.4 and embedded PowerPC systems compile code for the Solaris
13334 @opindex mcall-linux
13335 On System V.4 and embedded PowerPC systems compile code for the
13336 Linux-based GNU system.
13340 On System V.4 and embedded PowerPC systems compile code for the
13341 Hurd-based GNU system.
13343 @item -mcall-netbsd
13344 @opindex mcall-netbsd
13345 On System V.4 and embedded PowerPC systems compile code for the
13346 NetBSD operating system.
13348 @item -maix-struct-return
13349 @opindex maix-struct-return
13350 Return all structures in memory (as specified by the AIX ABI)@.
13352 @item -msvr4-struct-return
13353 @opindex msvr4-struct-return
13354 Return structures smaller than 8 bytes in registers (as specified by the
13357 @item -mabi=@var{abi-type}
13359 Extend the current ABI with a particular extension, or remove such extension.
13360 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
13361 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
13365 Extend the current ABI with SPE ABI extensions. This does not change
13366 the default ABI, instead it adds the SPE ABI extensions to the current
13370 @opindex mabi=no-spe
13371 Disable Booke SPE ABI extensions for the current ABI@.
13373 @item -mabi=ibmlongdouble
13374 @opindex mabi=ibmlongdouble
13375 Change the current ABI to use IBM extended precision long double.
13376 This is a PowerPC 32-bit SYSV ABI option.
13378 @item -mabi=ieeelongdouble
13379 @opindex mabi=ieeelongdouble
13380 Change the current ABI to use IEEE extended precision long double.
13381 This is a PowerPC 32-bit Linux ABI option.
13384 @itemx -mno-prototype
13385 @opindex mprototype
13386 @opindex mno-prototype
13387 On System V.4 and embedded PowerPC systems assume that all calls to
13388 variable argument functions are properly prototyped. Otherwise, the
13389 compiler must insert an instruction before every non prototyped call to
13390 set or clear bit 6 of the condition code register (@var{CR}) to
13391 indicate whether floating point values were passed in the floating point
13392 registers in case the function takes a variable arguments. With
13393 @option{-mprototype}, only calls to prototyped variable argument functions
13394 will set or clear the bit.
13398 On embedded PowerPC systems, assume that the startup module is called
13399 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
13400 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
13405 On embedded PowerPC systems, assume that the startup module is called
13406 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
13411 On embedded PowerPC systems, assume that the startup module is called
13412 @file{crt0.o} and the standard C libraries are @file{libads.a} and
13415 @item -myellowknife
13416 @opindex myellowknife
13417 On embedded PowerPC systems, assume that the startup module is called
13418 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
13423 On System V.4 and embedded PowerPC systems, specify that you are
13424 compiling for a VxWorks system.
13428 Specify that you are compiling for the WindISS simulation environment.
13432 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
13433 header to indicate that @samp{eabi} extended relocations are used.
13439 On System V.4 and embedded PowerPC systems do (do not) adhere to the
13440 Embedded Applications Binary Interface (eabi) which is a set of
13441 modifications to the System V.4 specifications. Selecting @option{-meabi}
13442 means that the stack is aligned to an 8 byte boundary, a function
13443 @code{__eabi} is called to from @code{main} to set up the eabi
13444 environment, and the @option{-msdata} option can use both @code{r2} and
13445 @code{r13} to point to two separate small data areas. Selecting
13446 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
13447 do not call an initialization function from @code{main}, and the
13448 @option{-msdata} option will only use @code{r13} to point to a single
13449 small data area. The @option{-meabi} option is on by default if you
13450 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
13453 @opindex msdata=eabi
13454 On System V.4 and embedded PowerPC systems, put small initialized
13455 @code{const} global and static data in the @samp{.sdata2} section, which
13456 is pointed to by register @code{r2}. Put small initialized
13457 non-@code{const} global and static data in the @samp{.sdata} section,
13458 which is pointed to by register @code{r13}. Put small uninitialized
13459 global and static data in the @samp{.sbss} section, which is adjacent to
13460 the @samp{.sdata} section. The @option{-msdata=eabi} option is
13461 incompatible with the @option{-mrelocatable} option. The
13462 @option{-msdata=eabi} option also sets the @option{-memb} option.
13465 @opindex msdata=sysv
13466 On System V.4 and embedded PowerPC systems, put small global and static
13467 data in the @samp{.sdata} section, which is pointed to by register
13468 @code{r13}. Put small uninitialized global and static data in the
13469 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
13470 The @option{-msdata=sysv} option is incompatible with the
13471 @option{-mrelocatable} option.
13473 @item -msdata=default
13475 @opindex msdata=default
13477 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
13478 compile code the same as @option{-msdata=eabi}, otherwise compile code the
13479 same as @option{-msdata=sysv}.
13482 @opindex msdata-data
13483 On System V.4 and embedded PowerPC systems, put small global
13484 data in the @samp{.sdata} section. Put small uninitialized global
13485 data in the @samp{.sbss} section. Do not use register @code{r13}
13486 to address small data however. This is the default behavior unless
13487 other @option{-msdata} options are used.
13491 @opindex msdata=none
13493 On embedded PowerPC systems, put all initialized global and static data
13494 in the @samp{.data} section, and all uninitialized data in the
13495 @samp{.bss} section.
13499 @cindex smaller data references (PowerPC)
13500 @cindex .sdata/.sdata2 references (PowerPC)
13501 On embedded PowerPC systems, put global and static items less than or
13502 equal to @var{num} bytes into the small data or bss sections instead of
13503 the normal data or bss section. By default, @var{num} is 8. The
13504 @option{-G @var{num}} switch is also passed to the linker.
13505 All modules should be compiled with the same @option{-G @var{num}} value.
13508 @itemx -mno-regnames
13510 @opindex mno-regnames
13511 On System V.4 and embedded PowerPC systems do (do not) emit register
13512 names in the assembly language output using symbolic forms.
13515 @itemx -mno-longcall
13517 @opindex mno-longcall
13518 By default assume that all calls are far away so that a longer more
13519 expensive calling sequence is required. This is required for calls
13520 further than 32 megabytes (33,554,432 bytes) from the current location.
13521 A short call will be generated if the compiler knows
13522 the call cannot be that far away. This setting can be overridden by
13523 the @code{shortcall} function attribute, or by @code{#pragma
13526 Some linkers are capable of detecting out-of-range calls and generating
13527 glue code on the fly. On these systems, long calls are unnecessary and
13528 generate slower code. As of this writing, the AIX linker can do this,
13529 as can the GNU linker for PowerPC/64. It is planned to add this feature
13530 to the GNU linker for 32-bit PowerPC systems as well.
13532 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
13533 callee, L42'', plus a ``branch island'' (glue code). The two target
13534 addresses represent the callee and the ``branch island''. The
13535 Darwin/PPC linker will prefer the first address and generate a ``bl
13536 callee'' if the PPC ``bl'' instruction will reach the callee directly;
13537 otherwise, the linker will generate ``bl L42'' to call the ``branch
13538 island''. The ``branch island'' is appended to the body of the
13539 calling function; it computes the full 32-bit address of the callee
13542 On Mach-O (Darwin) systems, this option directs the compiler emit to
13543 the glue for every direct call, and the Darwin linker decides whether
13544 to use or discard it.
13546 In the future, we may cause GCC to ignore all longcall specifications
13547 when the linker is known to generate glue.
13551 Adds support for multithreading with the @dfn{pthreads} library.
13552 This option sets flags for both the preprocessor and linker.
13556 @node S/390 and zSeries Options
13557 @subsection S/390 and zSeries Options
13558 @cindex S/390 and zSeries Options
13560 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
13564 @itemx -msoft-float
13565 @opindex mhard-float
13566 @opindex msoft-float
13567 Use (do not use) the hardware floating-point instructions and registers
13568 for floating-point operations. When @option{-msoft-float} is specified,
13569 functions in @file{libgcc.a} will be used to perform floating-point
13570 operations. When @option{-mhard-float} is specified, the compiler
13571 generates IEEE floating-point instructions. This is the default.
13573 @item -mlong-double-64
13574 @itemx -mlong-double-128
13575 @opindex mlong-double-64
13576 @opindex mlong-double-128
13577 These switches control the size of @code{long double} type. A size
13578 of 64bit makes the @code{long double} type equivalent to the @code{double}
13579 type. This is the default.
13582 @itemx -mno-backchain
13583 @opindex mbackchain
13584 @opindex mno-backchain
13585 Store (do not store) the address of the caller's frame as backchain pointer
13586 into the callee's stack frame.
13587 A backchain may be needed to allow debugging using tools that do not understand
13588 DWARF-2 call frame information.
13589 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
13590 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
13591 the backchain is placed into the topmost word of the 96/160 byte register
13594 In general, code compiled with @option{-mbackchain} is call-compatible with
13595 code compiled with @option{-mmo-backchain}; however, use of the backchain
13596 for debugging purposes usually requires that the whole binary is built with
13597 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
13598 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13599 to build a linux kernel use @option{-msoft-float}.
13601 The default is to not maintain the backchain.
13603 @item -mpacked-stack
13604 @item -mno-packed-stack
13605 @opindex mpacked-stack
13606 @opindex mno-packed-stack
13607 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
13608 specified, the compiler uses the all fields of the 96/160 byte register save
13609 area only for their default purpose; unused fields still take up stack space.
13610 When @option{-mpacked-stack} is specified, register save slots are densely
13611 packed at the top of the register save area; unused space is reused for other
13612 purposes, allowing for more efficient use of the available stack space.
13613 However, when @option{-mbackchain} is also in effect, the topmost word of
13614 the save area is always used to store the backchain, and the return address
13615 register is always saved two words below the backchain.
13617 As long as the stack frame backchain is not used, code generated with
13618 @option{-mpacked-stack} is call-compatible with code generated with
13619 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
13620 S/390 or zSeries generated code that uses the stack frame backchain at run
13621 time, not just for debugging purposes. Such code is not call-compatible
13622 with code compiled with @option{-mpacked-stack}. Also, note that the
13623 combination of @option{-mbackchain},
13624 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13625 to build a linux kernel use @option{-msoft-float}.
13627 The default is to not use the packed stack layout.
13630 @itemx -mno-small-exec
13631 @opindex msmall-exec
13632 @opindex mno-small-exec
13633 Generate (or do not generate) code using the @code{bras} instruction
13634 to do subroutine calls.
13635 This only works reliably if the total executable size does not
13636 exceed 64k. The default is to use the @code{basr} instruction instead,
13637 which does not have this limitation.
13643 When @option{-m31} is specified, generate code compliant to the
13644 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
13645 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
13646 particular to generate 64-bit instructions. For the @samp{s390}
13647 targets, the default is @option{-m31}, while the @samp{s390x}
13648 targets default to @option{-m64}.
13654 When @option{-mzarch} is specified, generate code using the
13655 instructions available on z/Architecture.
13656 When @option{-mesa} is specified, generate code using the
13657 instructions available on ESA/390. Note that @option{-mesa} is
13658 not possible with @option{-m64}.
13659 When generating code compliant to the GNU/Linux for S/390 ABI,
13660 the default is @option{-mesa}. When generating code compliant
13661 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
13667 Generate (or do not generate) code using the @code{mvcle} instruction
13668 to perform block moves. When @option{-mno-mvcle} is specified,
13669 use a @code{mvc} loop instead. This is the default unless optimizing for
13676 Print (or do not print) additional debug information when compiling.
13677 The default is to not print debug information.
13679 @item -march=@var{cpu-type}
13681 Generate code that will run on @var{cpu-type}, which is the name of a system
13682 representing a certain processor type. Possible values for
13683 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
13684 When generating code using the instructions available on z/Architecture,
13685 the default is @option{-march=z900}. Otherwise, the default is
13686 @option{-march=g5}.
13688 @item -mtune=@var{cpu-type}
13690 Tune to @var{cpu-type} everything applicable about the generated code,
13691 except for the ABI and the set of available instructions.
13692 The list of @var{cpu-type} values is the same as for @option{-march}.
13693 The default is the value used for @option{-march}.
13696 @itemx -mno-tpf-trace
13697 @opindex mtpf-trace
13698 @opindex mno-tpf-trace
13699 Generate code that adds (does not add) in TPF OS specific branches to trace
13700 routines in the operating system. This option is off by default, even
13701 when compiling for the TPF OS@.
13704 @itemx -mno-fused-madd
13705 @opindex mfused-madd
13706 @opindex mno-fused-madd
13707 Generate code that uses (does not use) the floating point multiply and
13708 accumulate instructions. These instructions are generated by default if
13709 hardware floating point is used.
13711 @item -mwarn-framesize=@var{framesize}
13712 @opindex mwarn-framesize
13713 Emit a warning if the current function exceeds the given frame size. Because
13714 this is a compile time check it doesn't need to be a real problem when the program
13715 runs. It is intended to identify functions which most probably cause
13716 a stack overflow. It is useful to be used in an environment with limited stack
13717 size e.g.@: the linux kernel.
13719 @item -mwarn-dynamicstack
13720 @opindex mwarn-dynamicstack
13721 Emit a warning if the function calls alloca or uses dynamically
13722 sized arrays. This is generally a bad idea with a limited stack size.
13724 @item -mstack-guard=@var{stack-guard}
13725 @item -mstack-size=@var{stack-size}
13726 @opindex mstack-guard
13727 @opindex mstack-size
13728 If these options are provided the s390 back end emits additional instructions in
13729 the function prologue which trigger a trap if the stack size is @var{stack-guard}
13730 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
13731 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
13732 the frame size of the compiled function is chosen.
13733 These options are intended to be used to help debugging stack overflow problems.
13734 The additionally emitted code causes only little overhead and hence can also be
13735 used in production like systems without greater performance degradation. The given
13736 values have to be exact powers of 2 and @var{stack-size} has to be greater than
13737 @var{stack-guard} without exceeding 64k.
13738 In order to be efficient the extra code makes the assumption that the stack starts
13739 at an address aligned to the value given by @var{stack-size}.
13740 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
13743 @node Score Options
13744 @subsection Score Options
13745 @cindex Score Options
13747 These options are defined for Score implementations:
13752 Compile code for big endian mode. This is the default.
13756 Compile code for little endian mode.
13760 Disable generate bcnz instruction.
13764 Enable generate unaligned load and store instruction.
13768 Enable the use of multiply-accumulate instructions. Disabled by default.
13772 Specify the SCORE5 as the target architecture.
13776 Specify the SCORE5U of the target architecture.
13780 Specify the SCORE7 as the target architecture. This is the default.
13784 Specify the SCORE7D as the target architecture.
13788 @subsection SH Options
13790 These @samp{-m} options are defined for the SH implementations:
13795 Generate code for the SH1.
13799 Generate code for the SH2.
13802 Generate code for the SH2e.
13806 Generate code for the SH3.
13810 Generate code for the SH3e.
13814 Generate code for the SH4 without a floating-point unit.
13816 @item -m4-single-only
13817 @opindex m4-single-only
13818 Generate code for the SH4 with a floating-point unit that only
13819 supports single-precision arithmetic.
13823 Generate code for the SH4 assuming the floating-point unit is in
13824 single-precision mode by default.
13828 Generate code for the SH4.
13832 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
13833 floating-point unit is not used.
13835 @item -m4a-single-only
13836 @opindex m4a-single-only
13837 Generate code for the SH4a, in such a way that no double-precision
13838 floating point operations are used.
13841 @opindex m4a-single
13842 Generate code for the SH4a assuming the floating-point unit is in
13843 single-precision mode by default.
13847 Generate code for the SH4a.
13851 Same as @option{-m4a-nofpu}, except that it implicitly passes
13852 @option{-dsp} to the assembler. GCC doesn't generate any DSP
13853 instructions at the moment.
13857 Compile code for the processor in big endian mode.
13861 Compile code for the processor in little endian mode.
13865 Align doubles at 64-bit boundaries. Note that this changes the calling
13866 conventions, and thus some functions from the standard C library will
13867 not work unless you recompile it first with @option{-mdalign}.
13871 Shorten some address references at link time, when possible; uses the
13872 linker option @option{-relax}.
13876 Use 32-bit offsets in @code{switch} tables. The default is to use
13881 Enable the use of the instruction @code{fmovd}.
13885 Comply with the calling conventions defined by Renesas.
13889 Comply with the calling conventions defined by Renesas.
13893 Comply with the calling conventions defined for GCC before the Renesas
13894 conventions were available. This option is the default for all
13895 targets of the SH toolchain except for @samp{sh-symbianelf}.
13898 @opindex mnomacsave
13899 Mark the @code{MAC} register as call-clobbered, even if
13900 @option{-mhitachi} is given.
13904 Increase IEEE-compliance of floating-point code.
13905 At the moment, this is equivalent to @option{-fno-finite-math-only}.
13906 When generating 16 bit SH opcodes, getting IEEE-conforming results for
13907 comparisons of NANs / infinities incurs extra overhead in every
13908 floating point comparison, therefore the default is set to
13909 @option{-ffinite-math-only}.
13911 @item -minline-ic_invalidate
13912 @opindex minline-ic_invalidate
13913 Inline code to invalidate instruction cache entries after setting up
13914 nested function trampolines.
13915 This option has no effect if -musermode is in effect and the selected
13916 code generation option (e.g. -m4) does not allow the use of the icbi
13918 If the selected code generation option does not allow the use of the icbi
13919 instruction, and -musermode is not in effect, the inlined code will
13920 manipulate the instruction cache address array directly with an associative
13921 write. This not only requires privileged mode, but it will also
13922 fail if the cache line had been mapped via the TLB and has become unmapped.
13926 Dump instruction size and location in the assembly code.
13929 @opindex mpadstruct
13930 This option is deprecated. It pads structures to multiple of 4 bytes,
13931 which is incompatible with the SH ABI@.
13935 Optimize for space instead of speed. Implied by @option{-Os}.
13938 @opindex mprefergot
13939 When generating position-independent code, emit function calls using
13940 the Global Offset Table instead of the Procedure Linkage Table.
13944 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
13945 if the inlined code would not work in user mode.
13946 This is the default when the target is @code{sh-*-linux*}.
13948 @item -multcost=@var{number}
13949 @opindex multcost=@var{number}
13950 Set the cost to assume for a multiply insn.
13952 @item -mdiv=@var{strategy}
13953 @opindex mdiv=@var{strategy}
13954 Set the division strategy to use for SHmedia code. @var{strategy} must be
13955 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
13956 inv:call2, inv:fp .
13957 "fp" performs the operation in floating point. This has a very high latency,
13958 but needs only a few instructions, so it might be a good choice if
13959 your code has enough easily exploitable ILP to allow the compiler to
13960 schedule the floating point instructions together with other instructions.
13961 Division by zero causes a floating point exception.
13962 "inv" uses integer operations to calculate the inverse of the divisor,
13963 and then multiplies the dividend with the inverse. This strategy allows
13964 cse and hoisting of the inverse calculation. Division by zero calculates
13965 an unspecified result, but does not trap.
13966 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
13967 have been found, or if the entire operation has been hoisted to the same
13968 place, the last stages of the inverse calculation are intertwined with the
13969 final multiply to reduce the overall latency, at the expense of using a few
13970 more instructions, and thus offering fewer scheduling opportunities with
13972 "call" calls a library function that usually implements the inv:minlat
13974 This gives high code density for m5-*media-nofpu compilations.
13975 "call2" uses a different entry point of the same library function, where it
13976 assumes that a pointer to a lookup table has already been set up, which
13977 exposes the pointer load to cse / code hoisting optimizations.
13978 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
13979 code generation, but if the code stays unoptimized, revert to the "call",
13980 "call2", or "fp" strategies, respectively. Note that the
13981 potentially-trapping side effect of division by zero is carried by a
13982 separate instruction, so it is possible that all the integer instructions
13983 are hoisted out, but the marker for the side effect stays where it is.
13984 A recombination to fp operations or a call is not possible in that case.
13985 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
13986 that the inverse calculation was nor separated from the multiply, they speed
13987 up division where the dividend fits into 20 bits (plus sign where applicable),
13988 by inserting a test to skip a number of operations in this case; this test
13989 slows down the case of larger dividends. inv20u assumes the case of a such
13990 a small dividend to be unlikely, and inv20l assumes it to be likely.
13992 @item -mdivsi3_libfunc=@var{name}
13993 @opindex mdivsi3_libfunc=@var{name}
13994 Set the name of the library function used for 32 bit signed division to
13995 @var{name}. This only affect the name used in the call and inv:call
13996 division strategies, and the compiler will still expect the same
13997 sets of input/output/clobbered registers as if this option was not present.
13999 @item -madjust-unroll
14000 @opindex madjust-unroll
14001 Throttle unrolling to avoid thrashing target registers.
14002 This option only has an effect if the gcc code base supports the
14003 TARGET_ADJUST_UNROLL_MAX target hook.
14005 @item -mindexed-addressing
14006 @opindex mindexed-addressing
14007 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14008 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14009 semantics for the indexed addressing mode. The architecture allows the
14010 implementation of processors with 64 bit MMU, which the OS could use to
14011 get 32 bit addressing, but since no current hardware implementation supports
14012 this or any other way to make the indexed addressing mode safe to use in
14013 the 32 bit ABI, the default is -mno-indexed-addressing.
14015 @item -mgettrcost=@var{number}
14016 @opindex mgettrcost=@var{number}
14017 Set the cost assumed for the gettr instruction to @var{number}.
14018 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14022 Assume pt* instructions won't trap. This will generally generate better
14023 scheduled code, but is unsafe on current hardware. The current architecture
14024 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14025 This has the unintentional effect of making it unsafe to schedule ptabs /
14026 ptrel before a branch, or hoist it out of a loop. For example,
14027 __do_global_ctors, a part of libgcc that runs constructors at program
14028 startup, calls functions in a list which is delimited by @minus{}1. With the
14029 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14030 That means that all the constructors will be run a bit quicker, but when
14031 the loop comes to the end of the list, the program crashes because ptabs
14032 loads @minus{}1 into a target register. Since this option is unsafe for any
14033 hardware implementing the current architecture specification, the default
14034 is -mno-pt-fixed. Unless the user specifies a specific cost with
14035 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14036 this deters register allocation using target registers for storing
14039 @item -minvalid-symbols
14040 @opindex minvalid-symbols
14041 Assume symbols might be invalid. Ordinary function symbols generated by
14042 the compiler will always be valid to load with movi/shori/ptabs or
14043 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14044 to generate symbols that will cause ptabs / ptrel to trap.
14045 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14046 It will then prevent cross-basic-block cse, hoisting and most scheduling
14047 of symbol loads. The default is @option{-mno-invalid-symbols}.
14050 @node SPARC Options
14051 @subsection SPARC Options
14052 @cindex SPARC options
14054 These @samp{-m} options are supported on the SPARC:
14057 @item -mno-app-regs
14059 @opindex mno-app-regs
14061 Specify @option{-mapp-regs} to generate output using the global registers
14062 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14065 To be fully SVR4 ABI compliant at the cost of some performance loss,
14066 specify @option{-mno-app-regs}. You should compile libraries and system
14067 software with this option.
14070 @itemx -mhard-float
14072 @opindex mhard-float
14073 Generate output containing floating point instructions. This is the
14077 @itemx -msoft-float
14079 @opindex msoft-float
14080 Generate output containing library calls for floating point.
14081 @strong{Warning:} the requisite libraries are not available for all SPARC
14082 targets. Normally the facilities of the machine's usual C compiler are
14083 used, but this cannot be done directly in cross-compilation. You must make
14084 your own arrangements to provide suitable library functions for
14085 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14086 @samp{sparclite-*-*} do provide software floating point support.
14088 @option{-msoft-float} changes the calling convention in the output file;
14089 therefore, it is only useful if you compile @emph{all} of a program with
14090 this option. In particular, you need to compile @file{libgcc.a}, the
14091 library that comes with GCC, with @option{-msoft-float} in order for
14094 @item -mhard-quad-float
14095 @opindex mhard-quad-float
14096 Generate output containing quad-word (long double) floating point
14099 @item -msoft-quad-float
14100 @opindex msoft-quad-float
14101 Generate output containing library calls for quad-word (long double)
14102 floating point instructions. The functions called are those specified
14103 in the SPARC ABI@. This is the default.
14105 As of this writing, there are no SPARC implementations that have hardware
14106 support for the quad-word floating point instructions. They all invoke
14107 a trap handler for one of these instructions, and then the trap handler
14108 emulates the effect of the instruction. Because of the trap handler overhead,
14109 this is much slower than calling the ABI library routines. Thus the
14110 @option{-msoft-quad-float} option is the default.
14112 @item -mno-unaligned-doubles
14113 @itemx -munaligned-doubles
14114 @opindex mno-unaligned-doubles
14115 @opindex munaligned-doubles
14116 Assume that doubles have 8 byte alignment. This is the default.
14118 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
14119 alignment only if they are contained in another type, or if they have an
14120 absolute address. Otherwise, it assumes they have 4 byte alignment.
14121 Specifying this option avoids some rare compatibility problems with code
14122 generated by other compilers. It is not the default because it results
14123 in a performance loss, especially for floating point code.
14125 @item -mno-faster-structs
14126 @itemx -mfaster-structs
14127 @opindex mno-faster-structs
14128 @opindex mfaster-structs
14129 With @option{-mfaster-structs}, the compiler assumes that structures
14130 should have 8 byte alignment. This enables the use of pairs of
14131 @code{ldd} and @code{std} instructions for copies in structure
14132 assignment, in place of twice as many @code{ld} and @code{st} pairs.
14133 However, the use of this changed alignment directly violates the SPARC
14134 ABI@. Thus, it's intended only for use on targets where the developer
14135 acknowledges that their resulting code will not be directly in line with
14136 the rules of the ABI@.
14138 @item -mimpure-text
14139 @opindex mimpure-text
14140 @option{-mimpure-text}, used in addition to @option{-shared}, tells
14141 the compiler to not pass @option{-z text} to the linker when linking a
14142 shared object. Using this option, you can link position-dependent
14143 code into a shared object.
14145 @option{-mimpure-text} suppresses the ``relocations remain against
14146 allocatable but non-writable sections'' linker error message.
14147 However, the necessary relocations will trigger copy-on-write, and the
14148 shared object is not actually shared across processes. Instead of
14149 using @option{-mimpure-text}, you should compile all source code with
14150 @option{-fpic} or @option{-fPIC}.
14152 This option is only available on SunOS and Solaris.
14154 @item -mcpu=@var{cpu_type}
14156 Set the instruction set, register set, and instruction scheduling parameters
14157 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
14158 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
14159 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
14160 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
14161 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
14163 Default instruction scheduling parameters are used for values that select
14164 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
14165 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
14167 Here is a list of each supported architecture and their supported
14172 v8: supersparc, hypersparc
14173 sparclite: f930, f934, sparclite86x
14175 v9: ultrasparc, ultrasparc3, niagara, niagara2
14178 By default (unless configured otherwise), GCC generates code for the V7
14179 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
14180 additionally optimizes it for the Cypress CY7C602 chip, as used in the
14181 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
14182 SPARCStation 1, 2, IPX etc.
14184 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
14185 architecture. The only difference from V7 code is that the compiler emits
14186 the integer multiply and integer divide instructions which exist in SPARC-V8
14187 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
14188 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
14191 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
14192 the SPARC architecture. This adds the integer multiply, integer divide step
14193 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
14194 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14195 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14196 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14197 MB86934 chip, which is the more recent SPARClite with FPU@.
14199 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14200 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
14201 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
14202 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
14203 optimizes it for the TEMIC SPARClet chip.
14205 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
14206 architecture. This adds 64-bit integer and floating-point move instructions,
14207 3 additional floating-point condition code registers and conditional move
14208 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
14209 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
14210 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
14211 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
14212 @option{-mcpu=niagara}, the compiler additionally optimizes it for
14213 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
14214 additionally optimizes it for Sun UltraSPARC T2 chips.
14216 @item -mtune=@var{cpu_type}
14218 Set the instruction scheduling parameters for machine type
14219 @var{cpu_type}, but do not set the instruction set or register set that the
14220 option @option{-mcpu=@var{cpu_type}} would.
14222 The same values for @option{-mcpu=@var{cpu_type}} can be used for
14223 @option{-mtune=@var{cpu_type}}, but the only useful values are those
14224 that select a particular cpu implementation. Those are @samp{cypress},
14225 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
14226 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
14227 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
14232 @opindex mno-v8plus
14233 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
14234 difference from the V8 ABI is that the global and out registers are
14235 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
14236 mode for all SPARC-V9 processors.
14242 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
14243 Visual Instruction Set extensions. The default is @option{-mno-vis}.
14246 These @samp{-m} options are supported in addition to the above
14247 on SPARC-V9 processors in 64-bit environments:
14250 @item -mlittle-endian
14251 @opindex mlittle-endian
14252 Generate code for a processor running in little-endian mode. It is only
14253 available for a few configurations and most notably not on Solaris and Linux.
14259 Generate code for a 32-bit or 64-bit environment.
14260 The 32-bit environment sets int, long and pointer to 32 bits.
14261 The 64-bit environment sets int to 32 bits and long and pointer
14264 @item -mcmodel=medlow
14265 @opindex mcmodel=medlow
14266 Generate code for the Medium/Low code model: 64-bit addresses, programs
14267 must be linked in the low 32 bits of memory. Programs can be statically
14268 or dynamically linked.
14270 @item -mcmodel=medmid
14271 @opindex mcmodel=medmid
14272 Generate code for the Medium/Middle code model: 64-bit addresses, programs
14273 must be linked in the low 44 bits of memory, the text and data segments must
14274 be less than 2GB in size and the data segment must be located within 2GB of
14277 @item -mcmodel=medany
14278 @opindex mcmodel=medany
14279 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
14280 may be linked anywhere in memory, the text and data segments must be less
14281 than 2GB in size and the data segment must be located within 2GB of the
14284 @item -mcmodel=embmedany
14285 @opindex mcmodel=embmedany
14286 Generate code for the Medium/Anywhere code model for embedded systems:
14287 64-bit addresses, the text and data segments must be less than 2GB in
14288 size, both starting anywhere in memory (determined at link time). The
14289 global register %g4 points to the base of the data segment. Programs
14290 are statically linked and PIC is not supported.
14293 @itemx -mno-stack-bias
14294 @opindex mstack-bias
14295 @opindex mno-stack-bias
14296 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
14297 frame pointer if present, are offset by @minus{}2047 which must be added back
14298 when making stack frame references. This is the default in 64-bit mode.
14299 Otherwise, assume no such offset is present.
14302 These switches are supported in addition to the above on Solaris:
14307 Add support for multithreading using the Solaris threads library. This
14308 option sets flags for both the preprocessor and linker. This option does
14309 not affect the thread safety of object code produced by the compiler or
14310 that of libraries supplied with it.
14314 Add support for multithreading using the POSIX threads library. This
14315 option sets flags for both the preprocessor and linker. This option does
14316 not affect the thread safety of object code produced by the compiler or
14317 that of libraries supplied with it.
14321 This is a synonym for @option{-pthreads}.
14325 @subsection SPU Options
14326 @cindex SPU options
14328 These @samp{-m} options are supported on the SPU:
14332 @itemx -merror-reloc
14333 @opindex mwarn-reloc
14334 @opindex merror-reloc
14336 The loader for SPU does not handle dynamic relocations. By default, GCC
14337 will give an error when it generates code that requires a dynamic
14338 relocation. @option{-mno-error-reloc} disables the error,
14339 @option{-mwarn-reloc} will generate a warning instead.
14342 @itemx -munsafe-dma
14344 @opindex munsafe-dma
14346 Instructions which initiate or test completion of DMA must not be
14347 reordered with respect to loads and stores of the memory which is being
14348 accessed. Users typically address this problem using the volatile
14349 keyword, but that can lead to inefficient code in places where the
14350 memory is known to not change. Rather than mark the memory as volatile
14351 we treat the DMA instructions as potentially effecting all memory. With
14352 @option{-munsafe-dma} users must use the volatile keyword to protect
14355 @item -mbranch-hints
14356 @opindex mbranch-hints
14358 By default, GCC will generate a branch hint instruction to avoid
14359 pipeline stalls for always taken or probably taken branches. A hint
14360 will not be generated closer than 8 instructions away from its branch.
14361 There is little reason to disable them, except for debugging purposes,
14362 or to make an object a little bit smaller.
14366 @opindex msmall-mem
14367 @opindex mlarge-mem
14369 By default, GCC generates code assuming that addresses are never larger
14370 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
14371 a full 32 bit address.
14376 By default, GCC links against startup code that assumes the SPU-style
14377 main function interface (which has an unconventional parameter list).
14378 With @option{-mstdmain}, GCC will link your program against startup
14379 code that assumes a C99-style interface to @code{main}, including a
14380 local copy of @code{argv} strings.
14382 @item -mfixed-range=@var{register-range}
14383 @opindex mfixed-range
14384 Generate code treating the given register range as fixed registers.
14385 A fixed register is one that the register allocator can not use. This is
14386 useful when compiling kernel code. A register range is specified as
14387 two registers separated by a dash. Multiple register ranges can be
14388 specified separated by a comma.
14392 @node System V Options
14393 @subsection Options for System V
14395 These additional options are available on System V Release 4 for
14396 compatibility with other compilers on those systems:
14401 Create a shared object.
14402 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
14406 Identify the versions of each tool used by the compiler, in a
14407 @code{.ident} assembler directive in the output.
14411 Refrain from adding @code{.ident} directives to the output file (this is
14414 @item -YP,@var{dirs}
14416 Search the directories @var{dirs}, and no others, for libraries
14417 specified with @option{-l}.
14419 @item -Ym,@var{dir}
14421 Look in the directory @var{dir} to find the M4 preprocessor.
14422 The assembler uses this option.
14423 @c This is supposed to go with a -Yd for predefined M4 macro files, but
14424 @c the generic assembler that comes with Solaris takes just -Ym.
14428 @subsection V850 Options
14429 @cindex V850 Options
14431 These @samp{-m} options are defined for V850 implementations:
14435 @itemx -mno-long-calls
14436 @opindex mlong-calls
14437 @opindex mno-long-calls
14438 Treat all calls as being far away (near). If calls are assumed to be
14439 far away, the compiler will always load the functions address up into a
14440 register, and call indirect through the pointer.
14446 Do not optimize (do optimize) basic blocks that use the same index
14447 pointer 4 or more times to copy pointer into the @code{ep} register, and
14448 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
14449 option is on by default if you optimize.
14451 @item -mno-prolog-function
14452 @itemx -mprolog-function
14453 @opindex mno-prolog-function
14454 @opindex mprolog-function
14455 Do not use (do use) external functions to save and restore registers
14456 at the prologue and epilogue of a function. The external functions
14457 are slower, but use less code space if more than one function saves
14458 the same number of registers. The @option{-mprolog-function} option
14459 is on by default if you optimize.
14463 Try to make the code as small as possible. At present, this just turns
14464 on the @option{-mep} and @option{-mprolog-function} options.
14466 @item -mtda=@var{n}
14468 Put static or global variables whose size is @var{n} bytes or less into
14469 the tiny data area that register @code{ep} points to. The tiny data
14470 area can hold up to 256 bytes in total (128 bytes for byte references).
14472 @item -msda=@var{n}
14474 Put static or global variables whose size is @var{n} bytes or less into
14475 the small data area that register @code{gp} points to. The small data
14476 area can hold up to 64 kilobytes.
14478 @item -mzda=@var{n}
14480 Put static or global variables whose size is @var{n} bytes or less into
14481 the first 32 kilobytes of memory.
14485 Specify that the target processor is the V850.
14488 @opindex mbig-switch
14489 Generate code suitable for big switch tables. Use this option only if
14490 the assembler/linker complain about out of range branches within a switch
14495 This option will cause r2 and r5 to be used in the code generated by
14496 the compiler. This setting is the default.
14498 @item -mno-app-regs
14499 @opindex mno-app-regs
14500 This option will cause r2 and r5 to be treated as fixed registers.
14504 Specify that the target processor is the V850E1. The preprocessor
14505 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
14506 this option is used.
14510 Specify that the target processor is the V850E@. The preprocessor
14511 constant @samp{__v850e__} will be defined if this option is used.
14513 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
14514 are defined then a default target processor will be chosen and the
14515 relevant @samp{__v850*__} preprocessor constant will be defined.
14517 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
14518 defined, regardless of which processor variant is the target.
14520 @item -mdisable-callt
14521 @opindex mdisable-callt
14522 This option will suppress generation of the CALLT instruction for the
14523 v850e and v850e1 flavors of the v850 architecture. The default is
14524 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
14529 @subsection VAX Options
14530 @cindex VAX options
14532 These @samp{-m} options are defined for the VAX:
14537 Do not output certain jump instructions (@code{aobleq} and so on)
14538 that the Unix assembler for the VAX cannot handle across long
14543 Do output those jump instructions, on the assumption that you
14544 will assemble with the GNU assembler.
14548 Output code for g-format floating point numbers instead of d-format.
14551 @node VxWorks Options
14552 @subsection VxWorks Options
14553 @cindex VxWorks Options
14555 The options in this section are defined for all VxWorks targets.
14556 Options specific to the target hardware are listed with the other
14557 options for that target.
14562 GCC can generate code for both VxWorks kernels and real time processes
14563 (RTPs). This option switches from the former to the latter. It also
14564 defines the preprocessor macro @code{__RTP__}.
14567 @opindex non-static
14568 Link an RTP executable against shared libraries rather than static
14569 libraries. The options @option{-static} and @option{-shared} can
14570 also be used for RTPs (@pxref{Link Options}); @option{-static}
14577 These options are passed down to the linker. They are defined for
14578 compatibility with Diab.
14581 @opindex Xbind-lazy
14582 Enable lazy binding of function calls. This option is equivalent to
14583 @option{-Wl,-z,now} and is defined for compatibility with Diab.
14587 Disable lazy binding of function calls. This option is the default and
14588 is defined for compatibility with Diab.
14591 @node x86-64 Options
14592 @subsection x86-64 Options
14593 @cindex x86-64 options
14595 These are listed under @xref{i386 and x86-64 Options}.
14597 @node Xstormy16 Options
14598 @subsection Xstormy16 Options
14599 @cindex Xstormy16 Options
14601 These options are defined for Xstormy16:
14606 Choose startup files and linker script suitable for the simulator.
14609 @node Xtensa Options
14610 @subsection Xtensa Options
14611 @cindex Xtensa Options
14613 These options are supported for Xtensa targets:
14617 @itemx -mno-const16
14619 @opindex mno-const16
14620 Enable or disable use of @code{CONST16} instructions for loading
14621 constant values. The @code{CONST16} instruction is currently not a
14622 standard option from Tensilica. When enabled, @code{CONST16}
14623 instructions are always used in place of the standard @code{L32R}
14624 instructions. The use of @code{CONST16} is enabled by default only if
14625 the @code{L32R} instruction is not available.
14628 @itemx -mno-fused-madd
14629 @opindex mfused-madd
14630 @opindex mno-fused-madd
14631 Enable or disable use of fused multiply/add and multiply/subtract
14632 instructions in the floating-point option. This has no effect if the
14633 floating-point option is not also enabled. Disabling fused multiply/add
14634 and multiply/subtract instructions forces the compiler to use separate
14635 instructions for the multiply and add/subtract operations. This may be
14636 desirable in some cases where strict IEEE 754-compliant results are
14637 required: the fused multiply add/subtract instructions do not round the
14638 intermediate result, thereby producing results with @emph{more} bits of
14639 precision than specified by the IEEE standard. Disabling fused multiply
14640 add/subtract instructions also ensures that the program output is not
14641 sensitive to the compiler's ability to combine multiply and add/subtract
14644 @item -mserialize-volatile
14645 @itemx -mno-serialize-volatile
14646 @opindex mserialize-volatile
14647 @opindex mno-serialize-volatile
14648 When this option is enabled, GCC inserts @code{MEMW} instructions before
14649 @code{volatile} memory references to guarantee sequential consistency.
14650 The default is @option{-mserialize-volatile}. Use
14651 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
14653 @item -mtext-section-literals
14654 @itemx -mno-text-section-literals
14655 @opindex mtext-section-literals
14656 @opindex mno-text-section-literals
14657 Control the treatment of literal pools. The default is
14658 @option{-mno-text-section-literals}, which places literals in a separate
14659 section in the output file. This allows the literal pool to be placed
14660 in a data RAM/ROM, and it also allows the linker to combine literal
14661 pools from separate object files to remove redundant literals and
14662 improve code size. With @option{-mtext-section-literals}, the literals
14663 are interspersed in the text section in order to keep them as close as
14664 possible to their references. This may be necessary for large assembly
14667 @item -mtarget-align
14668 @itemx -mno-target-align
14669 @opindex mtarget-align
14670 @opindex mno-target-align
14671 When this option is enabled, GCC instructs the assembler to
14672 automatically align instructions to reduce branch penalties at the
14673 expense of some code density. The assembler attempts to widen density
14674 instructions to align branch targets and the instructions following call
14675 instructions. If there are not enough preceding safe density
14676 instructions to align a target, no widening will be performed. The
14677 default is @option{-mtarget-align}. These options do not affect the
14678 treatment of auto-aligned instructions like @code{LOOP}, which the
14679 assembler will always align, either by widening density instructions or
14680 by inserting no-op instructions.
14683 @itemx -mno-longcalls
14684 @opindex mlongcalls
14685 @opindex mno-longcalls
14686 When this option is enabled, GCC instructs the assembler to translate
14687 direct calls to indirect calls unless it can determine that the target
14688 of a direct call is in the range allowed by the call instruction. This
14689 translation typically occurs for calls to functions in other source
14690 files. Specifically, the assembler translates a direct @code{CALL}
14691 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
14692 The default is @option{-mno-longcalls}. This option should be used in
14693 programs where the call target can potentially be out of range. This
14694 option is implemented in the assembler, not the compiler, so the
14695 assembly code generated by GCC will still show direct call
14696 instructions---look at the disassembled object code to see the actual
14697 instructions. Note that the assembler will use an indirect call for
14698 every cross-file call, not just those that really will be out of range.
14701 @node zSeries Options
14702 @subsection zSeries Options
14703 @cindex zSeries options
14705 These are listed under @xref{S/390 and zSeries Options}.
14707 @node Code Gen Options
14708 @section Options for Code Generation Conventions
14709 @cindex code generation conventions
14710 @cindex options, code generation
14711 @cindex run-time options
14713 These machine-independent options control the interface conventions
14714 used in code generation.
14716 Most of them have both positive and negative forms; the negative form
14717 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
14718 one of the forms is listed---the one which is not the default. You
14719 can figure out the other form by either removing @samp{no-} or adding
14723 @item -fbounds-check
14724 @opindex fbounds-check
14725 For front-ends that support it, generate additional code to check that
14726 indices used to access arrays are within the declared range. This is
14727 currently only supported by the Java and Fortran front-ends, where
14728 this option defaults to true and false respectively.
14732 This option generates traps for signed overflow on addition, subtraction,
14733 multiplication operations.
14737 This option instructs the compiler to assume that signed arithmetic
14738 overflow of addition, subtraction and multiplication wraps around
14739 using twos-complement representation. This flag enables some optimizations
14740 and disables others. This option is enabled by default for the Java
14741 front-end, as required by the Java language specification.
14744 @opindex fexceptions
14745 Enable exception handling. Generates extra code needed to propagate
14746 exceptions. For some targets, this implies GCC will generate frame
14747 unwind information for all functions, which can produce significant data
14748 size overhead, although it does not affect execution. If you do not
14749 specify this option, GCC will enable it by default for languages like
14750 C++ which normally require exception handling, and disable it for
14751 languages like C that do not normally require it. However, you may need
14752 to enable this option when compiling C code that needs to interoperate
14753 properly with exception handlers written in C++. You may also wish to
14754 disable this option if you are compiling older C++ programs that don't
14755 use exception handling.
14757 @item -fnon-call-exceptions
14758 @opindex fnon-call-exceptions
14759 Generate code that allows trapping instructions to throw exceptions.
14760 Note that this requires platform-specific runtime support that does
14761 not exist everywhere. Moreover, it only allows @emph{trapping}
14762 instructions to throw exceptions, i.e.@: memory references or floating
14763 point instructions. It does not allow exceptions to be thrown from
14764 arbitrary signal handlers such as @code{SIGALRM}.
14766 @item -funwind-tables
14767 @opindex funwind-tables
14768 Similar to @option{-fexceptions}, except that it will just generate any needed
14769 static data, but will not affect the generated code in any other way.
14770 You will normally not enable this option; instead, a language processor
14771 that needs this handling would enable it on your behalf.
14773 @item -fasynchronous-unwind-tables
14774 @opindex fasynchronous-unwind-tables
14775 Generate unwind table in dwarf2 format, if supported by target machine. The
14776 table is exact at each instruction boundary, so it can be used for stack
14777 unwinding from asynchronous events (such as debugger or garbage collector).
14779 @item -fpcc-struct-return
14780 @opindex fpcc-struct-return
14781 Return ``short'' @code{struct} and @code{union} values in memory like
14782 longer ones, rather than in registers. This convention is less
14783 efficient, but it has the advantage of allowing intercallability between
14784 GCC-compiled files and files compiled with other compilers, particularly
14785 the Portable C Compiler (pcc).
14787 The precise convention for returning structures in memory depends
14788 on the target configuration macros.
14790 Short structures and unions are those whose size and alignment match
14791 that of some integer type.
14793 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
14794 switch is not binary compatible with code compiled with the
14795 @option{-freg-struct-return} switch.
14796 Use it to conform to a non-default application binary interface.
14798 @item -freg-struct-return
14799 @opindex freg-struct-return
14800 Return @code{struct} and @code{union} values in registers when possible.
14801 This is more efficient for small structures than
14802 @option{-fpcc-struct-return}.
14804 If you specify neither @option{-fpcc-struct-return} nor
14805 @option{-freg-struct-return}, GCC defaults to whichever convention is
14806 standard for the target. If there is no standard convention, GCC
14807 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
14808 the principal compiler. In those cases, we can choose the standard, and
14809 we chose the more efficient register return alternative.
14811 @strong{Warning:} code compiled with the @option{-freg-struct-return}
14812 switch is not binary compatible with code compiled with the
14813 @option{-fpcc-struct-return} switch.
14814 Use it to conform to a non-default application binary interface.
14816 @item -fshort-enums
14817 @opindex fshort-enums
14818 Allocate to an @code{enum} type only as many bytes as it needs for the
14819 declared range of possible values. Specifically, the @code{enum} type
14820 will be equivalent to the smallest integer type which has enough room.
14822 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
14823 code that is not binary compatible with code generated without that switch.
14824 Use it to conform to a non-default application binary interface.
14826 @item -fshort-double
14827 @opindex fshort-double
14828 Use the same size for @code{double} as for @code{float}.
14830 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
14831 code that is not binary compatible with code generated without that switch.
14832 Use it to conform to a non-default application binary interface.
14834 @item -fshort-wchar
14835 @opindex fshort-wchar
14836 Override the underlying type for @samp{wchar_t} to be @samp{short
14837 unsigned int} instead of the default for the target. This option is
14838 useful for building programs to run under WINE@.
14840 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
14841 code that is not binary compatible with code generated without that switch.
14842 Use it to conform to a non-default application binary interface.
14845 @opindex fno-common
14846 In C, allocate even uninitialized global variables in the data section of the
14847 object file, rather than generating them as common blocks. This has the
14848 effect that if the same variable is declared (without @code{extern}) in
14849 two different compilations, you will get an error when you link them.
14850 The only reason this might be useful is if you wish to verify that the
14851 program will work on other systems which always work this way.
14855 Ignore the @samp{#ident} directive.
14857 @item -finhibit-size-directive
14858 @opindex finhibit-size-directive
14859 Don't output a @code{.size} assembler directive, or anything else that
14860 would cause trouble if the function is split in the middle, and the
14861 two halves are placed at locations far apart in memory. This option is
14862 used when compiling @file{crtstuff.c}; you should not need to use it
14865 @item -fverbose-asm
14866 @opindex fverbose-asm
14867 Put extra commentary information in the generated assembly code to
14868 make it more readable. This option is generally only of use to those
14869 who actually need to read the generated assembly code (perhaps while
14870 debugging the compiler itself).
14872 @option{-fno-verbose-asm}, the default, causes the
14873 extra information to be omitted and is useful when comparing two assembler
14876 @item -frecord-gcc-switches
14877 @opindex frecord-gcc-switches
14878 This switch causes the command line that was used to invoke the
14879 compiler to be recorded into the object file that is being created.
14880 This switch is only implemented on some targets and the exact format
14881 of the recording is target and binary file format dependent, but it
14882 usually takes the form of a section containing ASCII text. This
14883 switch is related to the @option{-fverbose-asm} switch, but that
14884 switch only records information in the assembler output file as
14885 comments, so it never reaches the object file.
14889 @cindex global offset table
14891 Generate position-independent code (PIC) suitable for use in a shared
14892 library, if supported for the target machine. Such code accesses all
14893 constant addresses through a global offset table (GOT)@. The dynamic
14894 loader resolves the GOT entries when the program starts (the dynamic
14895 loader is not part of GCC; it is part of the operating system). If
14896 the GOT size for the linked executable exceeds a machine-specific
14897 maximum size, you get an error message from the linker indicating that
14898 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14899 instead. (These maximums are 8k on the SPARC and 32k
14900 on the m68k and RS/6000. The 386 has no such limit.)
14902 Position-independent code requires special support, and therefore works
14903 only on certain machines. For the 386, GCC supports PIC for System V
14904 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14905 position-independent.
14907 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14912 If supported for the target machine, emit position-independent code,
14913 suitable for dynamic linking and avoiding any limit on the size of the
14914 global offset table. This option makes a difference on the m68k,
14915 PowerPC and SPARC@.
14917 Position-independent code requires special support, and therefore works
14918 only on certain machines.
14920 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14927 These options are similar to @option{-fpic} and @option{-fPIC}, but
14928 generated position independent code can be only linked into executables.
14929 Usually these options are used when @option{-pie} GCC option will be
14930 used during linking.
14932 @option{-fpie} and @option{-fPIE} both define the macros
14933 @code{__pie__} and @code{__PIE__}. The macros have the value 1
14934 for @option{-fpie} and 2 for @option{-fPIE}.
14936 @item -fno-jump-tables
14937 @opindex fno-jump-tables
14938 Do not use jump tables for switch statements even where it would be
14939 more efficient than other code generation strategies. This option is
14940 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14941 building code which forms part of a dynamic linker and cannot
14942 reference the address of a jump table. On some targets, jump tables
14943 do not require a GOT and this option is not needed.
14945 @item -ffixed-@var{reg}
14947 Treat the register named @var{reg} as a fixed register; generated code
14948 should never refer to it (except perhaps as a stack pointer, frame
14949 pointer or in some other fixed role).
14951 @var{reg} must be the name of a register. The register names accepted
14952 are machine-specific and are defined in the @code{REGISTER_NAMES}
14953 macro in the machine description macro file.
14955 This flag does not have a negative form, because it specifies a
14958 @item -fcall-used-@var{reg}
14959 @opindex fcall-used
14960 Treat the register named @var{reg} as an allocable register that is
14961 clobbered by function calls. It may be allocated for temporaries or
14962 variables that do not live across a call. Functions compiled this way
14963 will not save and restore the register @var{reg}.
14965 It is an error to used this flag with the frame pointer or stack pointer.
14966 Use of this flag for other registers that have fixed pervasive roles in
14967 the machine's execution model will produce disastrous results.
14969 This flag does not have a negative form, because it specifies a
14972 @item -fcall-saved-@var{reg}
14973 @opindex fcall-saved
14974 Treat the register named @var{reg} as an allocable register saved by
14975 functions. It may be allocated even for temporaries or variables that
14976 live across a call. Functions compiled this way will save and restore
14977 the register @var{reg} if they use it.
14979 It is an error to used this flag with the frame pointer or stack pointer.
14980 Use of this flag for other registers that have fixed pervasive roles in
14981 the machine's execution model will produce disastrous results.
14983 A different sort of disaster will result from the use of this flag for
14984 a register in which function values may be returned.
14986 This flag does not have a negative form, because it specifies a
14989 @item -fpack-struct[=@var{n}]
14990 @opindex fpack-struct
14991 Without a value specified, pack all structure members together without
14992 holes. When a value is specified (which must be a small power of two), pack
14993 structure members according to this value, representing the maximum
14994 alignment (that is, objects with default alignment requirements larger than
14995 this will be output potentially unaligned at the next fitting location.
14997 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14998 code that is not binary compatible with code generated without that switch.
14999 Additionally, it makes the code suboptimal.
15000 Use it to conform to a non-default application binary interface.
15002 @item -finstrument-functions
15003 @opindex finstrument-functions
15004 Generate instrumentation calls for entry and exit to functions. Just
15005 after function entry and just before function exit, the following
15006 profiling functions will be called with the address of the current
15007 function and its call site. (On some platforms,
15008 @code{__builtin_return_address} does not work beyond the current
15009 function, so the call site information may not be available to the
15010 profiling functions otherwise.)
15013 void __cyg_profile_func_enter (void *this_fn,
15015 void __cyg_profile_func_exit (void *this_fn,
15019 The first argument is the address of the start of the current function,
15020 which may be looked up exactly in the symbol table.
15022 This instrumentation is also done for functions expanded inline in other
15023 functions. The profiling calls will indicate where, conceptually, the
15024 inline function is entered and exited. This means that addressable
15025 versions of such functions must be available. If all your uses of a
15026 function are expanded inline, this may mean an additional expansion of
15027 code size. If you use @samp{extern inline} in your C code, an
15028 addressable version of such functions must be provided. (This is
15029 normally the case anyways, but if you get lucky and the optimizer always
15030 expands the functions inline, you might have gotten away without
15031 providing static copies.)
15033 A function may be given the attribute @code{no_instrument_function}, in
15034 which case this instrumentation will not be done. This can be used, for
15035 example, for the profiling functions listed above, high-priority
15036 interrupt routines, and any functions from which the profiling functions
15037 cannot safely be called (perhaps signal handlers, if the profiling
15038 routines generate output or allocate memory).
15040 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
15041 @opindex finstrument-functions-exclude-file-list
15043 Set the list of functions that are excluded from instrumentation (see
15044 the description of @code{-finstrument-functions}). If the file that
15045 contains a function definition matches with one of @var{file}, then
15046 that function is not instrumented. The match is done on substrings:
15047 if the @var{file} parameter is a substring of the file name, it is
15048 considered to be a match.
15051 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15052 will exclude any inline function defined in files whose pathnames
15053 contain @code{/bits/stl} or @code{include/sys}.
15055 If, for some reason, you want to include letter @code{','} in one of
15056 @var{sym}, write @code{'\,'}. For example,
15057 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15058 (note the single quote surrounding the option).
15060 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15061 @opindex finstrument-functions-exclude-function-list
15063 This is similar to @code{-finstrument-functions-exclude-file-list},
15064 but this option sets the list of function names to be excluded from
15065 instrumentation. The function name to be matched is its user-visible
15066 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15067 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15068 match is done on substrings: if the @var{sym} parameter is a substring
15069 of the function name, it is considered to be a match.
15071 @item -fstack-check
15072 @opindex fstack-check
15073 Generate code to verify that you do not go beyond the boundary of the
15074 stack. You should specify this flag if you are running in an
15075 environment with multiple threads, but only rarely need to specify it in
15076 a single-threaded environment since stack overflow is automatically
15077 detected on nearly all systems if there is only one stack.
15079 Note that this switch does not actually cause checking to be done; the
15080 operating system must do that. The switch causes generation of code
15081 to ensure that the operating system sees the stack being extended.
15083 @item -fstack-limit-register=@var{reg}
15084 @itemx -fstack-limit-symbol=@var{sym}
15085 @itemx -fno-stack-limit
15086 @opindex fstack-limit-register
15087 @opindex fstack-limit-symbol
15088 @opindex fno-stack-limit
15089 Generate code to ensure that the stack does not grow beyond a certain value,
15090 either the value of a register or the address of a symbol. If the stack
15091 would grow beyond the value, a signal is raised. For most targets,
15092 the signal is raised before the stack overruns the boundary, so
15093 it is possible to catch the signal without taking special precautions.
15095 For instance, if the stack starts at absolute address @samp{0x80000000}
15096 and grows downwards, you can use the flags
15097 @option{-fstack-limit-symbol=__stack_limit} and
15098 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15099 of 128KB@. Note that this may only work with the GNU linker.
15101 @cindex aliasing of parameters
15102 @cindex parameters, aliased
15103 @item -fargument-alias
15104 @itemx -fargument-noalias
15105 @itemx -fargument-noalias-global
15106 @itemx -fargument-noalias-anything
15107 @opindex fargument-alias
15108 @opindex fargument-noalias
15109 @opindex fargument-noalias-global
15110 @opindex fargument-noalias-anything
15111 Specify the possible relationships among parameters and between
15112 parameters and global data.
15114 @option{-fargument-alias} specifies that arguments (parameters) may
15115 alias each other and may alias global storage.@*
15116 @option{-fargument-noalias} specifies that arguments do not alias
15117 each other, but may alias global storage.@*
15118 @option{-fargument-noalias-global} specifies that arguments do not
15119 alias each other and do not alias global storage.
15120 @option{-fargument-noalias-anything} specifies that arguments do not
15121 alias any other storage.
15123 Each language will automatically use whatever option is required by
15124 the language standard. You should not need to use these options yourself.
15126 @item -fleading-underscore
15127 @opindex fleading-underscore
15128 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15129 change the way C symbols are represented in the object file. One use
15130 is to help link with legacy assembly code.
15132 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15133 generate code that is not binary compatible with code generated without that
15134 switch. Use it to conform to a non-default application binary interface.
15135 Not all targets provide complete support for this switch.
15137 @item -ftls-model=@var{model}
15138 @opindex ftls-model
15139 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15140 The @var{model} argument should be one of @code{global-dynamic},
15141 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15143 The default without @option{-fpic} is @code{initial-exec}; with
15144 @option{-fpic} the default is @code{global-dynamic}.
15146 @item -fvisibility=@var{default|internal|hidden|protected}
15147 @opindex fvisibility
15148 Set the default ELF image symbol visibility to the specified option---all
15149 symbols will be marked with this unless overridden within the code.
15150 Using this feature can very substantially improve linking and
15151 load times of shared object libraries, produce more optimized
15152 code, provide near-perfect API export and prevent symbol clashes.
15153 It is @strong{strongly} recommended that you use this in any shared objects
15156 Despite the nomenclature, @code{default} always means public ie;
15157 available to be linked against from outside the shared object.
15158 @code{protected} and @code{internal} are pretty useless in real-world
15159 usage so the only other commonly used option will be @code{hidden}.
15160 The default if @option{-fvisibility} isn't specified is
15161 @code{default}, i.e., make every
15162 symbol public---this causes the same behavior as previous versions of
15165 A good explanation of the benefits offered by ensuring ELF
15166 symbols have the correct visibility is given by ``How To Write
15167 Shared Libraries'' by Ulrich Drepper (which can be found at
15168 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
15169 solution made possible by this option to marking things hidden when
15170 the default is public is to make the default hidden and mark things
15171 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
15172 and @code{__attribute__ ((visibility("default")))} instead of
15173 @code{__declspec(dllexport)} you get almost identical semantics with
15174 identical syntax. This is a great boon to those working with
15175 cross-platform projects.
15177 For those adding visibility support to existing code, you may find
15178 @samp{#pragma GCC visibility} of use. This works by you enclosing
15179 the declarations you wish to set visibility for with (for example)
15180 @samp{#pragma GCC visibility push(hidden)} and
15181 @samp{#pragma GCC visibility pop}.
15182 Bear in mind that symbol visibility should be viewed @strong{as
15183 part of the API interface contract} and thus all new code should
15184 always specify visibility when it is not the default ie; declarations
15185 only for use within the local DSO should @strong{always} be marked explicitly
15186 as hidden as so to avoid PLT indirection overheads---making this
15187 abundantly clear also aids readability and self-documentation of the code.
15188 Note that due to ISO C++ specification requirements, operator new and
15189 operator delete must always be of default visibility.
15191 Be aware that headers from outside your project, in particular system
15192 headers and headers from any other library you use, may not be
15193 expecting to be compiled with visibility other than the default. You
15194 may need to explicitly say @samp{#pragma GCC visibility push(default)}
15195 before including any such headers.
15197 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
15198 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
15199 no modifications. However, this means that calls to @samp{extern}
15200 functions with no explicit visibility will use the PLT, so it is more
15201 effective to use @samp{__attribute ((visibility))} and/or
15202 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
15203 declarations should be treated as hidden.
15205 Note that @samp{-fvisibility} does affect C++ vague linkage
15206 entities. This means that, for instance, an exception class that will
15207 be thrown between DSOs must be explicitly marked with default
15208 visibility so that the @samp{type_info} nodes will be unified between
15211 An overview of these techniques, their benefits and how to use them
15212 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
15218 @node Environment Variables
15219 @section Environment Variables Affecting GCC
15220 @cindex environment variables
15222 @c man begin ENVIRONMENT
15223 This section describes several environment variables that affect how GCC
15224 operates. Some of them work by specifying directories or prefixes to use
15225 when searching for various kinds of files. Some are used to specify other
15226 aspects of the compilation environment.
15228 Note that you can also specify places to search using options such as
15229 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
15230 take precedence over places specified using environment variables, which
15231 in turn take precedence over those specified by the configuration of GCC@.
15232 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
15233 GNU Compiler Collection (GCC) Internals}.
15238 @c @itemx LC_COLLATE
15240 @c @itemx LC_MONETARY
15241 @c @itemx LC_NUMERIC
15246 @c @findex LC_COLLATE
15247 @findex LC_MESSAGES
15248 @c @findex LC_MONETARY
15249 @c @findex LC_NUMERIC
15253 These environment variables control the way that GCC uses
15254 localization information that allow GCC to work with different
15255 national conventions. GCC inspects the locale categories
15256 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
15257 so. These locale categories can be set to any value supported by your
15258 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
15259 Kingdom encoded in UTF-8.
15261 The @env{LC_CTYPE} environment variable specifies character
15262 classification. GCC uses it to determine the character boundaries in
15263 a string; this is needed for some multibyte encodings that contain quote
15264 and escape characters that would otherwise be interpreted as a string
15267 The @env{LC_MESSAGES} environment variable specifies the language to
15268 use in diagnostic messages.
15270 If the @env{LC_ALL} environment variable is set, it overrides the value
15271 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
15272 and @env{LC_MESSAGES} default to the value of the @env{LANG}
15273 environment variable. If none of these variables are set, GCC
15274 defaults to traditional C English behavior.
15278 If @env{TMPDIR} is set, it specifies the directory to use for temporary
15279 files. GCC uses temporary files to hold the output of one stage of
15280 compilation which is to be used as input to the next stage: for example,
15281 the output of the preprocessor, which is the input to the compiler
15284 @item GCC_EXEC_PREFIX
15285 @findex GCC_EXEC_PREFIX
15286 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
15287 names of the subprograms executed by the compiler. No slash is added
15288 when this prefix is combined with the name of a subprogram, but you can
15289 specify a prefix that ends with a slash if you wish.
15291 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
15292 an appropriate prefix to use based on the pathname it was invoked with.
15294 If GCC cannot find the subprogram using the specified prefix, it
15295 tries looking in the usual places for the subprogram.
15297 The default value of @env{GCC_EXEC_PREFIX} is
15298 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
15299 the installed compiler. In many cases @var{prefix} is the value
15300 of @code{prefix} when you ran the @file{configure} script.
15302 Other prefixes specified with @option{-B} take precedence over this prefix.
15304 This prefix is also used for finding files such as @file{crt0.o} that are
15307 In addition, the prefix is used in an unusual way in finding the
15308 directories to search for header files. For each of the standard
15309 directories whose name normally begins with @samp{/usr/local/lib/gcc}
15310 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
15311 replacing that beginning with the specified prefix to produce an
15312 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
15313 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
15314 These alternate directories are searched first; the standard directories
15315 come next. If a standard directory begins with the configured
15316 @var{prefix} then the value of @var{prefix} is replaced by
15317 @env{GCC_EXEC_PREFIX} when looking for header files.
15319 @item COMPILER_PATH
15320 @findex COMPILER_PATH
15321 The value of @env{COMPILER_PATH} is a colon-separated list of
15322 directories, much like @env{PATH}. GCC tries the directories thus
15323 specified when searching for subprograms, if it can't find the
15324 subprograms using @env{GCC_EXEC_PREFIX}.
15327 @findex LIBRARY_PATH
15328 The value of @env{LIBRARY_PATH} is a colon-separated list of
15329 directories, much like @env{PATH}. When configured as a native compiler,
15330 GCC tries the directories thus specified when searching for special
15331 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
15332 using GCC also uses these directories when searching for ordinary
15333 libraries for the @option{-l} option (but directories specified with
15334 @option{-L} come first).
15338 @cindex locale definition
15339 This variable is used to pass locale information to the compiler. One way in
15340 which this information is used is to determine the character set to be used
15341 when character literals, string literals and comments are parsed in C and C++.
15342 When the compiler is configured to allow multibyte characters,
15343 the following values for @env{LANG} are recognized:
15347 Recognize JIS characters.
15349 Recognize SJIS characters.
15351 Recognize EUCJP characters.
15354 If @env{LANG} is not defined, or if it has some other value, then the
15355 compiler will use mblen and mbtowc as defined by the default locale to
15356 recognize and translate multibyte characters.
15360 Some additional environments variables affect the behavior of the
15363 @include cppenv.texi
15367 @node Precompiled Headers
15368 @section Using Precompiled Headers
15369 @cindex precompiled headers
15370 @cindex speed of compilation
15372 Often large projects have many header files that are included in every
15373 source file. The time the compiler takes to process these header files
15374 over and over again can account for nearly all of the time required to
15375 build the project. To make builds faster, GCC allows users to
15376 `precompile' a header file; then, if builds can use the precompiled
15377 header file they will be much faster.
15379 To create a precompiled header file, simply compile it as you would any
15380 other file, if necessary using the @option{-x} option to make the driver
15381 treat it as a C or C++ header file. You will probably want to use a
15382 tool like @command{make} to keep the precompiled header up-to-date when
15383 the headers it contains change.
15385 A precompiled header file will be searched for when @code{#include} is
15386 seen in the compilation. As it searches for the included file
15387 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
15388 compiler looks for a precompiled header in each directory just before it
15389 looks for the include file in that directory. The name searched for is
15390 the name specified in the @code{#include} with @samp{.gch} appended. If
15391 the precompiled header file can't be used, it is ignored.
15393 For instance, if you have @code{#include "all.h"}, and you have
15394 @file{all.h.gch} in the same directory as @file{all.h}, then the
15395 precompiled header file will be used if possible, and the original
15396 header will be used otherwise.
15398 Alternatively, you might decide to put the precompiled header file in a
15399 directory and use @option{-I} to ensure that directory is searched
15400 before (or instead of) the directory containing the original header.
15401 Then, if you want to check that the precompiled header file is always
15402 used, you can put a file of the same name as the original header in this
15403 directory containing an @code{#error} command.
15405 This also works with @option{-include}. So yet another way to use
15406 precompiled headers, good for projects not designed with precompiled
15407 header files in mind, is to simply take most of the header files used by
15408 a project, include them from another header file, precompile that header
15409 file, and @option{-include} the precompiled header. If the header files
15410 have guards against multiple inclusion, they will be skipped because
15411 they've already been included (in the precompiled header).
15413 If you need to precompile the same header file for different
15414 languages, targets, or compiler options, you can instead make a
15415 @emph{directory} named like @file{all.h.gch}, and put each precompiled
15416 header in the directory, perhaps using @option{-o}. It doesn't matter
15417 what you call the files in the directory, every precompiled header in
15418 the directory will be considered. The first precompiled header
15419 encountered in the directory that is valid for this compilation will
15420 be used; they're searched in no particular order.
15422 There are many other possibilities, limited only by your imagination,
15423 good sense, and the constraints of your build system.
15425 A precompiled header file can be used only when these conditions apply:
15429 Only one precompiled header can be used in a particular compilation.
15432 A precompiled header can't be used once the first C token is seen. You
15433 can have preprocessor directives before a precompiled header; you can
15434 even include a precompiled header from inside another header, so long as
15435 there are no C tokens before the @code{#include}.
15438 The precompiled header file must be produced for the same language as
15439 the current compilation. You can't use a C precompiled header for a C++
15443 The precompiled header file must have been produced by the same compiler
15444 binary as the current compilation is using.
15447 Any macros defined before the precompiled header is included must
15448 either be defined in the same way as when the precompiled header was
15449 generated, or must not affect the precompiled header, which usually
15450 means that they don't appear in the precompiled header at all.
15452 The @option{-D} option is one way to define a macro before a
15453 precompiled header is included; using a @code{#define} can also do it.
15454 There are also some options that define macros implicitly, like
15455 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
15458 @item If debugging information is output when using the precompiled
15459 header, using @option{-g} or similar, the same kind of debugging information
15460 must have been output when building the precompiled header. However,
15461 a precompiled header built using @option{-g} can be used in a compilation
15462 when no debugging information is being output.
15464 @item The same @option{-m} options must generally be used when building
15465 and using the precompiled header. @xref{Submodel Options},
15466 for any cases where this rule is relaxed.
15468 @item Each of the following options must be the same when building and using
15469 the precompiled header:
15471 @gccoptlist{-fexceptions -funit-at-a-time}
15474 Some other command-line options starting with @option{-f},
15475 @option{-p}, or @option{-O} must be defined in the same way as when
15476 the precompiled header was generated. At present, it's not clear
15477 which options are safe to change and which are not; the safest choice
15478 is to use exactly the same options when generating and using the
15479 precompiled header. The following are known to be safe:
15481 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
15482 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
15483 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
15488 For all of these except the last, the compiler will automatically
15489 ignore the precompiled header if the conditions aren't met. If you
15490 find an option combination that doesn't work and doesn't cause the
15491 precompiled header to be ignored, please consider filing a bug report,
15494 If you do use differing options when generating and using the
15495 precompiled header, the actual behavior will be a mixture of the
15496 behavior for the options. For instance, if you use @option{-g} to
15497 generate the precompiled header but not when using it, you may or may
15498 not get debugging information for routines in the precompiled header.
15500 @node Running Protoize
15501 @section Running Protoize
15503 The program @code{protoize} is an optional part of GCC@. You can use
15504 it to add prototypes to a program, thus converting the program to ISO
15505 C in one respect. The companion program @code{unprotoize} does the
15506 reverse: it removes argument types from any prototypes that are found.
15508 When you run these programs, you must specify a set of source files as
15509 command line arguments. The conversion programs start out by compiling
15510 these files to see what functions they define. The information gathered
15511 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
15513 After scanning comes actual conversion. The specified files are all
15514 eligible to be converted; any files they include (whether sources or
15515 just headers) are eligible as well.
15517 But not all the eligible files are converted. By default,
15518 @code{protoize} and @code{unprotoize} convert only source and header
15519 files in the current directory. You can specify additional directories
15520 whose files should be converted with the @option{-d @var{directory}}
15521 option. You can also specify particular files to exclude with the
15522 @option{-x @var{file}} option. A file is converted if it is eligible, its
15523 directory name matches one of the specified directory names, and its
15524 name within the directory has not been excluded.
15526 Basic conversion with @code{protoize} consists of rewriting most
15527 function definitions and function declarations to specify the types of
15528 the arguments. The only ones not rewritten are those for varargs
15531 @code{protoize} optionally inserts prototype declarations at the
15532 beginning of the source file, to make them available for any calls that
15533 precede the function's definition. Or it can insert prototype
15534 declarations with block scope in the blocks where undeclared functions
15537 Basic conversion with @code{unprotoize} consists of rewriting most
15538 function declarations to remove any argument types, and rewriting
15539 function definitions to the old-style pre-ISO form.
15541 Both conversion programs print a warning for any function declaration or
15542 definition that they can't convert. You can suppress these warnings
15545 The output from @code{protoize} or @code{unprotoize} replaces the
15546 original source file. The original file is renamed to a name ending
15547 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
15548 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
15549 for DOS) file already exists, then the source file is simply discarded.
15551 @code{protoize} and @code{unprotoize} both depend on GCC itself to
15552 scan the program and collect information about the functions it uses.
15553 So neither of these programs will work until GCC is installed.
15555 Here is a table of the options you can use with @code{protoize} and
15556 @code{unprotoize}. Each option works with both programs unless
15560 @item -B @var{directory}
15561 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
15562 usual directory (normally @file{/usr/local/lib}). This file contains
15563 prototype information about standard system functions. This option
15564 applies only to @code{protoize}.
15566 @item -c @var{compilation-options}
15567 Use @var{compilation-options} as the options when running @command{gcc} to
15568 produce the @samp{.X} files. The special option @option{-aux-info} is
15569 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
15571 Note that the compilation options must be given as a single argument to
15572 @code{protoize} or @code{unprotoize}. If you want to specify several
15573 @command{gcc} options, you must quote the entire set of compilation options
15574 to make them a single word in the shell.
15576 There are certain @command{gcc} arguments that you cannot use, because they
15577 would produce the wrong kind of output. These include @option{-g},
15578 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
15579 the @var{compilation-options}, they are ignored.
15582 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
15583 systems) instead of @samp{.c}. This is convenient if you are converting
15584 a C program to C++. This option applies only to @code{protoize}.
15587 Add explicit global declarations. This means inserting explicit
15588 declarations at the beginning of each source file for each function
15589 that is called in the file and was not declared. These declarations
15590 precede the first function definition that contains a call to an
15591 undeclared function. This option applies only to @code{protoize}.
15593 @item -i @var{string}
15594 Indent old-style parameter declarations with the string @var{string}.
15595 This option applies only to @code{protoize}.
15597 @code{unprotoize} converts prototyped function definitions to old-style
15598 function definitions, where the arguments are declared between the
15599 argument list and the initial @samp{@{}. By default, @code{unprotoize}
15600 uses five spaces as the indentation. If you want to indent with just
15601 one space instead, use @option{-i " "}.
15604 Keep the @samp{.X} files. Normally, they are deleted after conversion
15608 Add explicit local declarations. @code{protoize} with @option{-l} inserts
15609 a prototype declaration for each function in each block which calls the
15610 function without any declaration. This option applies only to
15614 Make no real changes. This mode just prints information about the conversions
15615 that would have been done without @option{-n}.
15618 Make no @samp{.save} files. The original files are simply deleted.
15619 Use this option with caution.
15621 @item -p @var{program}
15622 Use the program @var{program} as the compiler. Normally, the name
15623 @file{gcc} is used.
15626 Work quietly. Most warnings are suppressed.
15629 Print the version number, just like @option{-v} for @command{gcc}.
15632 If you need special compiler options to compile one of your program's
15633 source files, then you should generate that file's @samp{.X} file
15634 specially, by running @command{gcc} on that source file with the
15635 appropriate options and the option @option{-aux-info}. Then run
15636 @code{protoize} on the entire set of files. @code{protoize} will use
15637 the existing @samp{.X} file because it is newer than the source file.
15641 gcc -Dfoo=bar file1.c -aux-info file1.X
15646 You need to include the special files along with the rest in the
15647 @code{protoize} command, even though their @samp{.X} files already
15648 exist, because otherwise they won't get converted.
15650 @xref{Protoize Caveats}, for more information on how to use
15651 @code{protoize} successfully.