1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
8 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997,
9 1998, 1999, 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
11 Permission is granted to copy, distribute and/or modify this document
12 under the terms of the GNU Free Documentation License, Version 1.2 or
13 any later version published by the Free Software Foundation; with the
14 Invariant Sections being ``GNU General Public License'' and ``Funding
15 Free Software'', the Front-Cover texts being (a) (see below), and with
16 the Back-Cover Texts being (b) (see below). A copy of the license is
17 included in the gfdl(7) man page.
19 (a) The FSF's Front-Cover Text is:
23 (b) The FSF's Back-Cover Text is:
25 You have freedom to copy and modify this GNU Manual, like GNU
26 software. Copies published by the Free Software Foundation raise
27 funds for GNU development.
29 @c Set file name and title for the man page.
31 @settitle GNU project C and C++ compiler
33 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
34 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
35 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
36 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
37 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
38 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
39 [@option{-o} @var{outfile}] @var{infile}@dots{}
41 Only the most useful options are listed here; see below for the
42 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
45 gpl(7), gfdl(7), fsf-funding(7),
46 cpp(1), gcov(1), g77(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
47 and the Info entries for @file{gcc}, @file{cpp}, @file{g77}, @file{as},
48 @file{ld}, @file{binutils} and @file{gdb}.
51 For instructions on reporting bugs, see
52 @w{@uref{http://gcc.gnu.org/bugs.html}}. Use of the @command{gccbug}
53 script to report bugs is recommended.
56 See the Info entry for @command{gcc}, or
57 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
58 for contributors to GCC@.
63 @chapter GCC Command Options
64 @cindex GCC command options
65 @cindex command options
66 @cindex options, GCC command
68 @c man begin DESCRIPTION
69 When you invoke GCC, it normally does preprocessing, compilation,
70 assembly and linking. The ``overall options'' allow you to stop this
71 process at an intermediate stage. For example, the @option{-c} option
72 says not to run the linker. Then the output consists of object files
73 output by the assembler.
75 Other options are passed on to one stage of processing. Some options
76 control the preprocessor and others the compiler itself. Yet other
77 options control the assembler and linker; most of these are not
78 documented here, since you rarely need to use any of them.
80 @cindex C compilation options
81 Most of the command line options that you can use with GCC are useful
82 for C programs; when an option is only useful with another language
83 (usually C++), the explanation says so explicitly. If the description
84 for a particular option does not mention a source language, you can use
85 that option with all supported languages.
87 @cindex C++ compilation options
88 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
89 options for compiling C++ programs.
91 @cindex grouping options
92 @cindex options, grouping
93 The @command{gcc} program accepts options and file names as operands. Many
94 options have multi-letter names; therefore multiple single-letter options
95 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
98 @cindex order of options
99 @cindex options, order
100 You can mix options and other arguments. For the most part, the order
101 you use doesn't matter. Order does matter when you use several options
102 of the same kind; for example, if you specify @option{-L} more than once,
103 the directories are searched in the order specified.
105 Many options have long names starting with @samp{-f} or with
106 @samp{-W}---for example, @option{-fforce-mem},
107 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
108 these have both positive and negative forms; the negative form of
109 @option{-ffoo} would be @option{-fno-foo}. This manual documents
110 only one of these two forms, whichever one is not the default.
114 @xref{Option Index}, for an index to GCC's options.
117 * Option Summary:: Brief list of all options, without explanations.
118 * Overall Options:: Controlling the kind of output:
119 an executable, object files, assembler files,
120 or preprocessed source.
121 * Invoking G++:: Compiling C++ programs.
122 * C Dialect Options:: Controlling the variant of C language compiled.
123 * C++ Dialect Options:: Variations on C++.
124 * Objective-C Dialect Options:: Variations on Objective-C.
125 * Language Independent Options:: Controlling how diagnostics should be
127 * Warning Options:: How picky should the compiler be?
128 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
129 * Optimize Options:: How much optimization?
130 * Preprocessor Options:: Controlling header files and macro definitions.
131 Also, getting dependency information for Make.
132 * Assembler Options:: Passing options to the assembler.
133 * Link Options:: Specifying libraries and so on.
134 * Directory Options:: Where to find header files and libraries.
135 Where to find the compiler executable files.
136 * Spec Files:: How to pass switches to sub-processes.
137 * Target Options:: Running a cross-compiler, or an old version of GCC.
138 * Submodel Options:: Specifying minor hardware or convention variations,
139 such as 68010 vs 68020.
140 * Code Gen Options:: Specifying conventions for function calls, data layout
142 * Environment Variables:: Env vars that affect GCC.
143 * Precompiled Headers:: Compiling a header once, and using it many times.
144 * Running Protoize:: Automatically adding or removing function prototypes.
150 @section Option Summary
152 Here is a summary of all the options, grouped by type. Explanations are
153 in the following sections.
156 @item Overall Options
157 @xref{Overall Options,,Options Controlling the Kind of Output}.
158 @gccoptlist{-c -S -E -o @var{file} -pipe -pass-exit-codes @gol
159 -x @var{language} -v -### --help --target-help --version}
161 @item C Language Options
162 @xref{C Dialect Options,,Options Controlling C Dialect}.
163 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
164 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
165 -fhosted -ffreestanding -fms-extensions @gol
166 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
167 -fallow-single-precision -fcond-mismatch @gol
168 -fsigned-bitfields -fsigned-char @gol
169 -funsigned-bitfields -funsigned-char @gol
172 @item C++ Language Options
173 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
174 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
175 -fconserve-space -fno-const-strings @gol
176 -fno-elide-constructors @gol
177 -fno-enforce-eh-specs @gol
178 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
179 -fno-implicit-templates @gol
180 -fno-implicit-inline-templates @gol
181 -fno-implement-inlines -fms-extensions @gol
182 -fno-nonansi-builtins -fno-operator-names @gol
183 -fno-optional-diags -fpermissive @gol
184 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
185 -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
186 -fno-default-inline -Wabi -Wctor-dtor-privacy @gol
187 -Wnon-virtual-dtor -Wreorder @gol
188 -Weffc++ -Wno-deprecated @gol
189 -Wno-non-template-friend -Wold-style-cast @gol
190 -Woverloaded-virtual -Wno-pmf-conversions @gol
191 -Wsign-promo -Wsynth}
193 @item Objective-C Language Options
194 @xref{Objective-C Dialect Options,,Options Controlling Objective-C Dialect}.
196 -fconstant-string-class=@var{class-name} @gol
197 -fgnu-runtime -fnext-runtime @gol
198 -fno-nil-receivers @gol
199 -fobjc-exceptions @gol
200 -freplace-objc-classes @gol
203 -Wno-protocol -Wselector -Wundeclared-selector}
205 @item Language Independent Options
206 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
207 @gccoptlist{-fmessage-length=@var{n} @gol
208 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
210 @item Warning Options
211 @xref{Warning Options,,Options to Request or Suppress Warnings}.
212 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
213 -w -Wextra -Wall -Waggregate-return @gol
214 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
215 -Wconversion -Wno-deprecated-declarations @gol
216 -Wdisabled-optimization -Wno-div-by-zero -Wendif-labels @gol
217 -Werror -Werror-implicit-function-declaration @gol
218 -Wfloat-equal -Wformat -Wformat=2 @gol
219 -Wno-format-extra-args -Wformat-nonliteral @gol
220 -Wformat-security -Wformat-y2k @gol
221 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
222 -Wimport -Wno-import -Winit-self -Winline @gol
223 -Wno-invalid-offsetof -Winvalid-pch @gol
224 -Wlarger-than-@var{len} -Wlong-long @gol
225 -Wmain -Wmissing-braces @gol
226 -Wmissing-format-attribute -Wmissing-noreturn @gol
227 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
228 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
229 -Wreturn-type -Wsequence-point -Wshadow @gol
230 -Wsign-compare -Wstrict-aliasing @gol
231 -Wswitch -Wswitch-default -Wswitch-enum @gol
232 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
233 -Wunknown-pragmas -Wunreachable-code @gol
234 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
235 -Wunused-value -Wunused-variable -Wwrite-strings}
237 @item C-only Warning Options
238 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
239 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
240 -Wstrict-prototypes -Wtraditional @gol
241 -Wdeclaration-after-statement}
243 @item Debugging Options
244 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
245 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
246 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
247 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
248 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
249 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
250 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
251 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
252 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs @gol
253 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
254 -ftest-coverage -ftime-report @gol
255 -g -g@var{level} -gcoff -gdwarf-2 @gol
256 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
257 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
258 -print-multi-directory -print-multi-lib @gol
259 -print-prog-name=@var{program} -print-search-dirs -Q @gol
262 @item Optimization Options
263 @xref{Optimize Options,,Options that Control Optimization}.
264 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
265 -falign-labels=@var{n} -falign-loops=@var{n} @gol
266 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
267 -fbranch-target-load-optimize2 -fcaller-saves -fcprop-registers @gol
268 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections @gol
269 -fdelayed-branch -fdelete-null-pointer-checks @gol
270 -fexpensive-optimizations -ffast-math -ffloat-store @gol
271 -fforce-addr -fforce-mem -ffunction-sections @gol
272 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -floop-optimize @gol
273 -fcrossjumping -fif-conversion -fif-conversion2 @gol
274 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
275 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
276 -fmove-all-movables -fnew-ra -fno-branch-count-reg @gol
277 -fno-default-inline -fno-defer-pop @gol
278 -fno-function-cse -fno-guess-branch-probability @gol
279 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
280 -funsafe-math-optimizations -ffinite-math-only @gol
281 -fno-trapping-math -fno-zero-initialized-in-bss @gol
282 -fomit-frame-pointer -foptimize-register-move @gol
283 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
284 -fprofile-generate -fprofile-use @gol
285 -freduce-all-givs -fregmove -frename-registers @gol
286 -freorder-blocks -freorder-functions @gol
287 -frerun-cse-after-loop -frerun-loop-opt @gol
288 -frounding-math -fschedule-insns -fschedule-insns2 @gol
289 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
290 -fsched-spec-load-dangerous @gol
291 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
292 -fsched2-use-superblocks @gol
293 -fsched2-use-traces -fsignaling-nans @gol
294 -fsingle-precision-constant @gol
295 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
296 -funroll-all-loops -funroll-loops -fpeel-loops @gol
297 -funswitch-loops -fold-unroll-loops -fold-unroll-all-loops @gol
298 --param @var{name}=@var{value}
299 -O -O0 -O1 -O2 -O3 -Os}
301 @item Preprocessor Options
302 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
303 @gccoptlist{-A@var{question}=@var{answer} @gol
304 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
305 -C -dD -dI -dM -dN @gol
306 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
307 -idirafter @var{dir} @gol
308 -include @var{file} -imacros @var{file} @gol
309 -iprefix @var{file} -iwithprefix @var{dir} @gol
310 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
311 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
312 -P -fworking-directory -remap @gol
313 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
314 -Xpreprocessor @var{option}}
316 @item Assembler Option
317 @xref{Assembler Options,,Passing Options to the Assembler}.
318 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
321 @xref{Link Options,,Options for Linking}.
322 @gccoptlist{@var{object-file-name} -l@var{library} @gol
323 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
324 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
325 -Wl,@var{option} -Xlinker @var{option} @gol
328 @item Directory Options
329 @xref{Directory Options,,Options for Directory Search}.
330 @gccoptlist{-B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}}
333 @c I wrote this xref this way to avoid overfull hbox. -- rms
334 @xref{Target Options}.
335 @gccoptlist{-V @var{version} -b @var{machine}}
337 @item Machine Dependent Options
338 @xref{Submodel Options,,Hardware Models and Configurations}.
340 @emph{M680x0 Options}
341 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
342 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
343 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
344 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
345 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
347 @emph{M68hc1x Options}
348 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
349 -mauto-incdec -minmax -mlong-calls -mshort @gol
350 -msoft-reg-count=@var{count}}
353 @gccoptlist{-mg -mgnu -munix}
356 @gccoptlist{-mcpu=@var{cpu-type} @gol
357 -mtune=@var{cpu-type} @gol
358 -mcmodel=@var{code-model} @gol
360 -mapp-regs -mbroken-saverestore -mcypress @gol
361 -mfaster-structs -mflat @gol
362 -mfpu -mhard-float -mhard-quad-float @gol
363 -mimpure-text -mlittle-endian -mlive-g0 -mno-app-regs @gol
364 -mno-faster-structs -mno-flat -mno-fpu @gol
365 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles @gol
366 -msoft-float -msoft-quad-float -msparclite -mstack-bias @gol
367 -msupersparc -munaligned-doubles -mv8}
370 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
371 -mapcs-26 -mapcs-32 @gol
372 -mapcs-stack-check -mno-apcs-stack-check @gol
373 -mapcs-float -mno-apcs-float @gol
374 -mapcs-reentrant -mno-apcs-reentrant @gol
375 -msched-prolog -mno-sched-prolog @gol
376 -mlittle-endian -mbig-endian -mwords-little-endian @gol
377 -malignment-traps -mno-alignment-traps @gol
378 -msoft-float -mhard-float -mfpe @gol
379 -mthumb-interwork -mno-thumb-interwork @gol
380 -mcpu=@var{name} -march=@var{name} -mfpe=@var{name} @gol
381 -mstructure-size-boundary=@var{n} @gol
382 -mabort-on-noreturn @gol
383 -mlong-calls -mno-long-calls @gol
384 -msingle-pic-base -mno-single-pic-base @gol
385 -mpic-register=@var{reg} @gol
386 -mnop-fun-dllimport @gol
387 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
388 -mpoke-function-name @gol
390 -mtpcs-frame -mtpcs-leaf-frame @gol
391 -mcaller-super-interworking -mcallee-super-interworking}
393 @emph{MN10200 Options}
396 @emph{MN10300 Options}
397 @gccoptlist{-mmult-bug -mno-mult-bug @gol
398 -mam33 -mno-am33 @gol
399 -mam33-2 -mno-am33-2 @gol
402 @emph{M32R/D Options}
403 @gccoptlist{-m32r2 -m32rx -m32r @gol
405 -malign-loops -mno-align-loops @gol
406 -missue-rate=@var{number} @gol
407 -mbranch-cost=@var{number} @gol
408 -mmodel=@var{code-size-model-type} @gol
409 -msdata=@var{sdata-type} @gol
410 -mno-flush-func -mflush-func=@var{name} @gol
411 -mno-flush-trap -mflush-trap=@var{number} @gol
415 @gccoptlist{-m88000 -m88100 -m88110 -mbig-pic @gol
416 -mcheck-zero-division -mhandle-large-shift @gol
417 -midentify-revision -mno-check-zero-division @gol
418 -mno-ocs-debug-info -mno-ocs-frame-position @gol
419 -mno-optimize-arg-area -mno-serialize-volatile @gol
420 -mno-underscores -mocs-debug-info @gol
421 -mocs-frame-position -moptimize-arg-area @gol
422 -mserialize-volatile -mshort-data-@var{num} -msvr3 @gol
423 -msvr4 -mtrap-large-shift -muse-div-instruction @gol
424 -mversion-03.00 -mwarn-passed-structs}
426 @emph{RS/6000 and PowerPC Options}
427 @gccoptlist{-mcpu=@var{cpu-type} @gol
428 -mtune=@var{cpu-type} @gol
429 -mpower -mno-power -mpower2 -mno-power2 @gol
430 -mpowerpc -mpowerpc64 -mno-powerpc @gol
431 -maltivec -mno-altivec @gol
432 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
433 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
434 -mnew-mnemonics -mold-mnemonics @gol
435 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
436 -m64 -m32 -mxl-call -mno-xl-call -mpe @gol
437 -malign-power -malign-natural @gol
438 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
439 -mstring -mno-string -mupdate -mno-update @gol
440 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
441 -mstrict-align -mno-strict-align -mrelocatable @gol
442 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
443 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
444 -mdynamic-no-pic @gol
445 -mprioritize-restricted-insns=@var{priority} @gol
446 -msched-costly-dep=@var{dependence_type} @gol
447 -minsert-sched-nops=@var{scheme} @gol
448 -mcall-sysv -mcall-netbsd @gol
449 -maix-struct-return -msvr4-struct-return @gol
450 -mabi=altivec -mabi=no-altivec @gol
451 -mabi=spe -mabi=no-spe @gol
452 -misel=yes -misel=no @gol
453 -mspe=yes -mspe=no @gol
454 -mfloat-gprs=yes -mfloat-gprs=no @gol
455 -mprototype -mno-prototype @gol
456 -msim -mmvme -mads -myellowknife -memb -msdata @gol
457 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
459 @emph{Darwin Options}
460 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
461 -arch_only -bind_at_load -bundle -bundle_loader @gol
462 -client_name -compatibility_version -current_version @gol
463 -dependency-file -dylib_file -dylinker_install_name @gol
464 -dynamic -dynamiclib -exported_symbols_list @gol
465 -filelist -flat_namespace -force_cpusubtype_ALL @gol
466 -force_flat_namespace -headerpad_max_install_names @gol
467 -image_base -init -install_name -keep_private_externs @gol
468 -multi_module -multiply_defined -multiply_defined_unused @gol
469 -noall_load -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
470 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
471 -private_bundle -read_only_relocs -sectalign @gol
472 -sectobjectsymbols -whyload -seg1addr @gol
473 -sectcreate -sectobjectsymbols -sectorder @gol
474 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
475 -segprot -segs_read_only_addr -segs_read_write_addr @gol
476 -single_module -static -sub_library -sub_umbrella @gol
477 -twolevel_namespace -umbrella -undefined @gol
478 -unexported_symbols_list -weak_reference_mismatches @gol
482 @gccoptlist{-mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs @gol
483 -mfull-fp-blocks -mhc-struct-return -min-line-mul @gol
484 -mminimum-fp-blocks -mnohc-struct-return}
487 @gccoptlist{-mabicalls -march=@var{cpu-type} -mtune=@var{cpu=type} @gol
488 -mcpu=@var{cpu-type} -membedded-data -muninit-const-in-rodata @gol
489 -membedded-pic -mfp32 -mfp64 -mfused-madd -mno-fused-madd @gol
490 -mgas -mgp32 -mgp64 -mhard-float -mint64 -mips1 @gol
491 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
492 -mlong64 -mlong32 -mlong-calls -mmemcpy @gol
493 -mmips-as -mmips-tfile -mno-abicalls -mxgot @gol
494 -mno-embedded-data -mno-uninit-const-in-rodata @gol
495 -mno-embedded-pic -mno-long-calls @gol
496 -mno-memcpy -mno-mips-tfile -mno-rnames @gol
497 -mrnames -msoft-float @gol
498 -m4650 -msingle-float -mmad @gol
499 -EL -EB -G @var{num} -nocpp @gol
500 -mabi=32 -mabi=n32 -mabi=64 -mabi=eabi -mabi-fake-default @gol
501 -mfix7000 -mfix-sb1 -mno-fix-sb1 @gol
502 -mno-crt0 -mflush-func=@var{func} -mno-flush-func @gol
503 -mbranch-likely -mno-branch-likely}
505 @emph{i386 and x86-64 Options}
506 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
507 -mfpmath=@var{unit} @gol
508 -masm=@var{dialect} -mno-fancy-math-387 @gol
509 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
510 -mno-wide-multiply -mrtd -malign-double @gol
511 -mpreferred-stack-boundary=@var{num} @gol
512 -mmmx -msse -msse2 -mpni -m3dnow @gol
513 -mthreads -mno-align-stringops -minline-all-stringops @gol
514 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
515 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
516 -mno-red-zone -mno-tls-direct-seg-refs @gol
517 -mcmodel=@var{code-model} @gol
521 @gccoptlist{-march=@var{architecture-type} @gol
522 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
523 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
524 -mjump-in-delay -mlinker-opt -mlong-calls @gol
525 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
526 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
527 -mno-jump-in-delay -mno-long-load-store @gol
528 -mno-portable-runtime -mno-soft-float @gol
529 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
530 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
531 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
532 -nolibdld -static -threads}
534 @emph{Intel 960 Options}
535 @gccoptlist{-m@var{cpu-type} -masm-compat -mclean-linkage @gol
536 -mcode-align -mcomplex-addr -mleaf-procedures @gol
537 -mic-compat -mic2.0-compat -mic3.0-compat @gol
538 -mintel-asm -mno-clean-linkage -mno-code-align @gol
539 -mno-complex-addr -mno-leaf-procedures @gol
540 -mno-old-align -mno-strict-align -mno-tail-call @gol
541 -mnumerics -mold-align -msoft-float -mstrict-align @gol
544 @emph{DEC Alpha Options}
545 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
546 -mieee -mieee-with-inexact -mieee-conformant @gol
547 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
548 -mtrap-precision=@var{mode} -mbuild-constants @gol
549 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
550 -mbwx -mmax -mfix -mcix @gol
551 -mfloat-vax -mfloat-ieee @gol
552 -mexplicit-relocs -msmall-data -mlarge-data @gol
553 -msmall-text -mlarge-text @gol
554 -mmemory-latency=@var{time}}
556 @emph{DEC Alpha/VMS Options}
557 @gccoptlist{-mvms-return-codes}
559 @emph{H8/300 Options}
560 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
563 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
564 -m4-nofpu -m4-single-only -m4-single -m4 @gol
565 -m5-64media -m5-64media-nofpu @gol
566 -m5-32media -m5-32media-nofpu @gol
567 -m5-compact -m5-compact-nofpu @gol
568 -mb -ml -mdalign -mrelax @gol
569 -mbigtable -mfmovd -mhitachi -mnomacsave @gol
570 -mieee -misize -mpadstruct -mspace @gol
571 -mprefergot -musermode}
573 @emph{System V Options}
574 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
577 @gccoptlist{-EB -EL @gol
578 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
579 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
581 @emph{TMS320C3x/C4x Options}
582 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
583 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
584 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
585 -mparallel-insns -mparallel-mpy -mpreserve-float}
588 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
589 -mprolog-function -mno-prolog-function -mspace @gol
590 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
591 -mapp-regs -mno-app-regs @gol
592 -mdisable-callt -mno-disable-callt @gol
598 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
599 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
600 -mregparam -mnoregparam -msb -mnosb @gol
601 -mbitfield -mnobitfield -mhimem -mnohimem}
604 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
605 -mcall-prologues -mno-tablejump -mtiny-stack}
608 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
609 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
610 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
611 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
612 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
615 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
616 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
617 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
618 -mno-base-addresses -msingle-exit -mno-single-exit}
621 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
622 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
623 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
624 -minline-float-divide-max-throughput @gol
625 -minline-int-divide-min-latency @gol
626 -minline-int-divide-max-throughput -mno-dwarf2-asm @gol
627 -mfixed-range=@var{register-range}}
630 @gccoptlist{-mextmem -mextmemory -monchip -mno-asm-optimize @gol
631 -masm-optimize -mbranch-cost=@var{n} -mcond-exec=@var{n}}
633 @emph{S/390 and zSeries Options}
634 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
635 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
636 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
637 -m64 -m31 -mdebug -mno-debug -mesa -mzarch -mfused-madd -mno-fused-madd}
640 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
641 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
642 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
643 -mstack-align -mdata-align -mconst-align @gol
644 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
645 -melf -maout -melinux -mlinux -sim -sim2}
647 @emph{PDP-11 Options}
648 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
649 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
650 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
651 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
652 -mbranch-expensive -mbranch-cheap @gol
653 -msplit -mno-split -munix-asm -mdec-asm}
655 @emph{Xstormy16 Options}
658 @emph{Xtensa Options}
659 @gccoptlist{-mconst16 -mno-const16 @gol
660 -mfused-madd -mno-fused-madd @gol
661 -mtext-section-literals -mno-text-section-literals @gol
662 -mtarget-align -mno-target-align @gol
663 -mlongcalls -mno-longcalls}
666 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
667 -mhard-float -msoft-float @gol
668 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
669 -mdouble -mno-double @gol
670 -mmedia -mno-media -mmuladd -mno-muladd @gol
671 -mlibrary-pic -macc-4 -macc-8 @gol
672 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
673 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
674 -mvliw-branch -mno-vliw-branch @gol
675 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
676 -mno-nested-cond-exec -mtomcat-stats @gol
679 @item Code Generation Options
680 @xref{Code Gen Options,,Options for Code Generation Conventions}.
681 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
682 -ffixed-@var{reg} -fexceptions @gol
683 -fnon-call-exceptions -funwind-tables @gol
684 -fasynchronous-unwind-tables @gol
685 -finhibit-size-directive -finstrument-functions @gol
686 -fno-common -fno-ident @gol
687 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
688 -freg-struct-return -fshared-data -fshort-enums @gol
689 -fshort-double -fshort-wchar @gol
690 -fverbose-asm -fpack-struct -fstack-check @gol
691 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
692 -fargument-alias -fargument-noalias @gol
693 -fargument-noalias-global -fleading-underscore @gol
694 -ftls-model=@var{model} @gol
695 -ftrapv -fwrapv -fbounds-check}
699 * Overall Options:: Controlling the kind of output:
700 an executable, object files, assembler files,
701 or preprocessed source.
702 * C Dialect Options:: Controlling the variant of C language compiled.
703 * C++ Dialect Options:: Variations on C++.
704 * Objective-C Dialect Options:: Variations on Objective-C.
705 * Language Independent Options:: Controlling how diagnostics should be
707 * Warning Options:: How picky should the compiler be?
708 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
709 * Optimize Options:: How much optimization?
710 * Preprocessor Options:: Controlling header files and macro definitions.
711 Also, getting dependency information for Make.
712 * Assembler Options:: Passing options to the assembler.
713 * Link Options:: Specifying libraries and so on.
714 * Directory Options:: Where to find header files and libraries.
715 Where to find the compiler executable files.
716 * Spec Files:: How to pass switches to sub-processes.
717 * Target Options:: Running a cross-compiler, or an old version of GCC.
720 @node Overall Options
721 @section Options Controlling the Kind of Output
723 Compilation can involve up to four stages: preprocessing, compilation
724 proper, assembly and linking, always in that order. GCC is capable of
725 preprocessing and compiling several files either into several
726 assembler input files, or into one assembler input file; then each
727 assembler input file produces an object file, and linking combines all
728 the object files (those newly compiled, and those specified as input)
729 into an executable file.
731 @cindex file name suffix
732 For any given input file, the file name suffix determines what kind of
737 C source code which must be preprocessed.
740 C source code which should not be preprocessed.
743 C++ source code which should not be preprocessed.
746 Objective-C source code. Note that you must link with the library
747 @file{libobjc.a} to make an Objective-C program work.
750 Objective-C source code which should not be preprocessed.
753 C or C++ header file to be turned into a precompiled header.
757 @itemx @var{file}.cxx
758 @itemx @var{file}.cpp
759 @itemx @var{file}.CPP
760 @itemx @var{file}.c++
762 C++ source code which must be preprocessed. Note that in @samp{.cxx},
763 the last two letters must both be literally @samp{x}. Likewise,
764 @samp{.C} refers to a literal capital C@.
768 C++ header file to be turned into a precompiled header.
771 @itemx @var{file}.for
772 @itemx @var{file}.FOR
773 Fortran source code which should not be preprocessed.
776 @itemx @var{file}.fpp
777 @itemx @var{file}.FPP
778 Fortran source code which must be preprocessed (with the traditional
782 Fortran source code which must be preprocessed with a RATFOR
783 preprocessor (not included with GCC)@.
785 @xref{Overall Options,,Options Controlling the Kind of Output, g77,
786 Using and Porting GNU Fortran}, for more details of the handling of
789 @c FIXME: Descriptions of Java file types.
796 Ada source code file which contains a library unit declaration (a
797 declaration of a package, subprogram, or generic, or a generic
798 instantiation), or a library unit renaming declaration (a package,
799 generic, or subprogram renaming declaration). Such files are also
802 @itemx @var{file}.adb
803 Ada source code file containing a library unit body (a subprogram or
804 package body). Such files are also called @dfn{bodies}.
806 @c GCC also knows about some suffixes for languages not yet included:
815 Assembler code which must be preprocessed.
818 An object file to be fed straight into linking.
819 Any file name with no recognized suffix is treated this way.
823 You can specify the input language explicitly with the @option{-x} option:
826 @item -x @var{language}
827 Specify explicitly the @var{language} for the following input files
828 (rather than letting the compiler choose a default based on the file
829 name suffix). This option applies to all following input files until
830 the next @option{-x} option. Possible values for @var{language} are:
832 c c-header cpp-output
833 c++ c++-header c++-cpp-output
834 objective-c objective-c-header objc-cpp-output
835 assembler assembler-with-cpp
837 f77 f77-cpp-input ratfor
843 Turn off any specification of a language, so that subsequent files are
844 handled according to their file name suffixes (as they are if @option{-x}
845 has not been used at all).
847 @item -pass-exit-codes
848 @opindex pass-exit-codes
849 Normally the @command{gcc} program will exit with the code of 1 if any
850 phase of the compiler returns a non-success return code. If you specify
851 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
852 numerically highest error produced by any phase that returned an error
856 If you only want some of the stages of compilation, you can use
857 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
858 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
859 @command{gcc} is to stop. Note that some combinations (for example,
860 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
865 Compile or assemble the source files, but do not link. The linking
866 stage simply is not done. The ultimate output is in the form of an
867 object file for each source file.
869 By default, the object file name for a source file is made by replacing
870 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
872 Unrecognized input files, not requiring compilation or assembly, are
877 Stop after the stage of compilation proper; do not assemble. The output
878 is in the form of an assembler code file for each non-assembler input
881 By default, the assembler file name for a source file is made by
882 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
884 Input files that don't require compilation are ignored.
888 Stop after the preprocessing stage; do not run the compiler proper. The
889 output is in the form of preprocessed source code, which is sent to the
892 Input files which don't require preprocessing are ignored.
894 @cindex output file option
897 Place output in file @var{file}. This applies regardless to whatever
898 sort of output is being produced, whether it be an executable file,
899 an object file, an assembler file or preprocessed C code.
901 If you specify @option{-o} when compiling more than one input file, or
902 you are producing an executable file as output, all the source files
903 on the command line will be compiled at once.
905 If @option{-o} is not specified, the default is to put an executable file
906 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
907 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
908 all preprocessed C source on standard output.
912 Print (on standard error output) the commands executed to run the stages
913 of compilation. Also print the version number of the compiler driver
914 program and of the preprocessor and the compiler proper.
918 Like @option{-v} except the commands are not executed and all command
919 arguments are quoted. This is useful for shell scripts to capture the
920 driver-generated command lines.
924 Use pipes rather than temporary files for communication between the
925 various stages of compilation. This fails to work on some systems where
926 the assembler is unable to read from a pipe; but the GNU assembler has
931 Print (on the standard output) a description of the command line options
932 understood by @command{gcc}. If the @option{-v} option is also specified
933 then @option{--help} will also be passed on to the various processes
934 invoked by @command{gcc}, so that they can display the command line options
935 they accept. If the @option{-Wextra} option is also specified then command
936 line options which have no documentation associated with them will also
941 Print (on the standard output) a description of target specific command
942 line options for each tool.
946 Display the version number and copyrights of the invoked GCC.
950 @section Compiling C++ Programs
952 @cindex suffixes for C++ source
953 @cindex C++ source file suffixes
954 C++ source files conventionally use one of the suffixes @samp{.C},
955 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
956 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
957 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
958 files with these names and compiles them as C++ programs even if you
959 call the compiler the same way as for compiling C programs (usually
960 with the name @command{gcc}).
964 However, C++ programs often require class libraries as well as a
965 compiler that understands the C++ language---and under some
966 circumstances, you might want to compile programs or header files from
967 standard input, or otherwise without a suffix that flags them as C++
968 programs. You might also like to precompile a C header file with a
969 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
970 program that calls GCC with the default language set to C++, and
971 automatically specifies linking against the C++ library. On many
972 systems, @command{g++} is also installed with the name @command{c++}.
974 @cindex invoking @command{g++}
975 When you compile C++ programs, you may specify many of the same
976 command-line options that you use for compiling programs in any
977 language; or command-line options meaningful for C and related
978 languages; or options that are meaningful only for C++ programs.
979 @xref{C Dialect Options,,Options Controlling C Dialect}, for
980 explanations of options for languages related to C@.
981 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
982 explanations of options that are meaningful only for C++ programs.
984 @node C Dialect Options
985 @section Options Controlling C Dialect
986 @cindex dialect options
987 @cindex language dialect options
988 @cindex options, dialect
990 The following options control the dialect of C (or languages derived
991 from C, such as C++ and Objective-C) that the compiler accepts:
998 In C mode, support all ISO C90 programs. In C++ mode,
999 remove GNU extensions that conflict with ISO C++.
1001 This turns off certain features of GCC that are incompatible with ISO
1002 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1003 such as the @code{asm} and @code{typeof} keywords, and
1004 predefined macros such as @code{unix} and @code{vax} that identify the
1005 type of system you are using. It also enables the undesirable and
1006 rarely used ISO trigraph feature. For the C compiler,
1007 it disables recognition of C++ style @samp{//} comments as well as
1008 the @code{inline} keyword.
1010 The alternate keywords @code{__asm__}, @code{__extension__},
1011 @code{__inline__} and @code{__typeof__} continue to work despite
1012 @option{-ansi}. You would not want to use them in an ISO C program, of
1013 course, but it is useful to put them in header files that might be included
1014 in compilations done with @option{-ansi}. Alternate predefined macros
1015 such as @code{__unix__} and @code{__vax__} are also available, with or
1016 without @option{-ansi}.
1018 The @option{-ansi} option does not cause non-ISO programs to be
1019 rejected gratuitously. For that, @option{-pedantic} is required in
1020 addition to @option{-ansi}. @xref{Warning Options}.
1022 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1023 option is used. Some header files may notice this macro and refrain
1024 from declaring certain functions or defining certain macros that the
1025 ISO standard doesn't call for; this is to avoid interfering with any
1026 programs that might use these names for other things.
1028 Functions which would normally be built in but do not have semantics
1029 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1030 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1031 built-in functions provided by GCC}, for details of the functions
1036 Determine the language standard. This option is currently only
1037 supported when compiling C or C++. A value for this option must be
1038 provided; possible values are
1043 ISO C90 (same as @option{-ansi}).
1045 @item iso9899:199409
1046 ISO C90 as modified in amendment 1.
1052 ISO C99. Note that this standard is not yet fully supported; see
1053 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1054 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1057 Default, ISO C90 plus GNU extensions (including some C99 features).
1061 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1062 this will become the default. The name @samp{gnu9x} is deprecated.
1065 The 1998 ISO C++ standard plus amendments.
1068 The same as @option{-std=c++98} plus GNU extensions. This is the
1069 default for C++ code.
1072 Even when this option is not specified, you can still use some of the
1073 features of newer standards in so far as they do not conflict with
1074 previous C standards. For example, you may use @code{__restrict__} even
1075 when @option{-std=c99} is not specified.
1077 The @option{-std} options specifying some version of ISO C have the same
1078 effects as @option{-ansi}, except that features that were not in ISO C90
1079 but are in the specified version (for example, @samp{//} comments and
1080 the @code{inline} keyword in ISO C99) are not disabled.
1082 @xref{Standards,,Language Standards Supported by GCC}, for details of
1083 these standard versions.
1085 @item -aux-info @var{filename}
1087 Output to the given filename prototyped declarations for all functions
1088 declared and/or defined in a translation unit, including those in header
1089 files. This option is silently ignored in any language other than C@.
1091 Besides declarations, the file indicates, in comments, the origin of
1092 each declaration (source file and line), whether the declaration was
1093 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1094 @samp{O} for old, respectively, in the first character after the line
1095 number and the colon), and whether it came from a declaration or a
1096 definition (@samp{C} or @samp{F}, respectively, in the following
1097 character). In the case of function definitions, a K&R-style list of
1098 arguments followed by their declarations is also provided, inside
1099 comments, after the declaration.
1103 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1104 keyword, so that code can use these words as identifiers. You can use
1105 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1106 instead. @option{-ansi} implies @option{-fno-asm}.
1108 In C++, this switch only affects the @code{typeof} keyword, since
1109 @code{asm} and @code{inline} are standard keywords. You may want to
1110 use the @option{-fno-gnu-keywords} flag instead, which has the same
1111 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1112 switch only affects the @code{asm} and @code{typeof} keywords, since
1113 @code{inline} is a standard keyword in ISO C99.
1116 @itemx -fno-builtin-@var{function}
1117 @opindex fno-builtin
1118 @cindex built-in functions
1119 Don't recognize built-in functions that do not begin with
1120 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1121 functions provided by GCC}, for details of the functions affected,
1122 including those which are not built-in functions when @option{-ansi} or
1123 @option{-std} options for strict ISO C conformance are used because they
1124 do not have an ISO standard meaning.
1126 GCC normally generates special code to handle certain built-in functions
1127 more efficiently; for instance, calls to @code{alloca} may become single
1128 instructions that adjust the stack directly, and calls to @code{memcpy}
1129 may become inline copy loops. The resulting code is often both smaller
1130 and faster, but since the function calls no longer appear as such, you
1131 cannot set a breakpoint on those calls, nor can you change the behavior
1132 of the functions by linking with a different library.
1134 With the @option{-fno-builtin-@var{function}} option
1135 only the built-in function @var{function} is
1136 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1137 function is named this is not built-in in this version of GCC, this
1138 option is ignored. There is no corresponding
1139 @option{-fbuiltin-@var{function}} option; if you wish to enable
1140 built-in functions selectively when using @option{-fno-builtin} or
1141 @option{-ffreestanding}, you may define macros such as:
1144 #define abs(n) __builtin_abs ((n))
1145 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1150 @cindex hosted environment
1152 Assert that compilation takes place in a hosted environment. This implies
1153 @option{-fbuiltin}. A hosted environment is one in which the
1154 entire standard library is available, and in which @code{main} has a return
1155 type of @code{int}. Examples are nearly everything except a kernel.
1156 This is equivalent to @option{-fno-freestanding}.
1158 @item -ffreestanding
1159 @opindex ffreestanding
1160 @cindex hosted environment
1162 Assert that compilation takes place in a freestanding environment. This
1163 implies @option{-fno-builtin}. A freestanding environment
1164 is one in which the standard library may not exist, and program startup may
1165 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1166 This is equivalent to @option{-fno-hosted}.
1168 @xref{Standards,,Language Standards Supported by GCC}, for details of
1169 freestanding and hosted environments.
1171 @item -fms-extensions
1172 @opindex fms-extensions
1173 Accept some non-standard constructs used in Microsoft header files.
1177 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1178 options for strict ISO C conformance) implies @option{-trigraphs}.
1180 @item -no-integrated-cpp
1181 @opindex no-integrated-cpp
1182 Performs a compilation in two passes: preprocessing and compiling. This
1183 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1184 @option{-B} option. The user supplied compilation step can then add in
1185 an additional preprocessing step after normal preprocessing but before
1186 compiling. The default is to use the integrated cpp (internal cpp)
1188 The semantics of this option will change if "cc1", "cc1plus", and
1189 "cc1obj" are merged.
1191 @cindex traditional C language
1192 @cindex C language, traditional
1194 @itemx -traditional-cpp
1195 @opindex traditional-cpp
1196 @opindex traditional
1197 Formerly, these options caused GCC to attempt to emulate a pre-standard
1198 C compiler. They are now only supported with the @option{-E} switch.
1199 The preprocessor continues to support a pre-standard mode. See the GNU
1200 CPP manual for details.
1202 @item -fcond-mismatch
1203 @opindex fcond-mismatch
1204 Allow conditional expressions with mismatched types in the second and
1205 third arguments. The value of such an expression is void. This option
1206 is not supported for C++.
1208 @item -funsigned-char
1209 @opindex funsigned-char
1210 Let the type @code{char} be unsigned, like @code{unsigned char}.
1212 Each kind of machine has a default for what @code{char} should
1213 be. It is either like @code{unsigned char} by default or like
1214 @code{signed char} by default.
1216 Ideally, a portable program should always use @code{signed char} or
1217 @code{unsigned char} when it depends on the signedness of an object.
1218 But many programs have been written to use plain @code{char} and
1219 expect it to be signed, or expect it to be unsigned, depending on the
1220 machines they were written for. This option, and its inverse, let you
1221 make such a program work with the opposite default.
1223 The type @code{char} is always a distinct type from each of
1224 @code{signed char} or @code{unsigned char}, even though its behavior
1225 is always just like one of those two.
1228 @opindex fsigned-char
1229 Let the type @code{char} be signed, like @code{signed char}.
1231 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1232 the negative form of @option{-funsigned-char}. Likewise, the option
1233 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1235 @item -fsigned-bitfields
1236 @itemx -funsigned-bitfields
1237 @itemx -fno-signed-bitfields
1238 @itemx -fno-unsigned-bitfields
1239 @opindex fsigned-bitfields
1240 @opindex funsigned-bitfields
1241 @opindex fno-signed-bitfields
1242 @opindex fno-unsigned-bitfields
1243 These options control whether a bit-field is signed or unsigned, when the
1244 declaration does not use either @code{signed} or @code{unsigned}. By
1245 default, such a bit-field is signed, because this is consistent: the
1246 basic integer types such as @code{int} are signed types.
1248 @item -fwritable-strings
1249 @opindex fwritable-strings
1250 Store string constants in the writable data segment and don't uniquize
1251 them. This is for compatibility with old programs which assume they can
1252 write into string constants.
1254 Writing into string constants is a very bad idea; ``constants'' should
1257 This option is deprecated.
1260 @node C++ Dialect Options
1261 @section Options Controlling C++ Dialect
1263 @cindex compiler options, C++
1264 @cindex C++ options, command line
1265 @cindex options, C++
1266 This section describes the command-line options that are only meaningful
1267 for C++ programs; but you can also use most of the GNU compiler options
1268 regardless of what language your program is in. For example, you
1269 might compile a file @code{firstClass.C} like this:
1272 g++ -g -frepo -O -c firstClass.C
1276 In this example, only @option{-frepo} is an option meant
1277 only for C++ programs; you can use the other options with any
1278 language supported by GCC@.
1280 Here is a list of options that are @emph{only} for compiling C++ programs:
1284 @item -fabi-version=@var{n}
1285 @opindex fabi-version
1286 Use version @var{n} of the C++ ABI. Version 2 is the version of the
1287 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1288 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1289 the version that conforms most closely to the C++ ABI specification.
1290 Therefore, the ABI obtained using version 0 will change as ABI bugs
1293 The default is version 1.
1295 @item -fno-access-control
1296 @opindex fno-access-control
1297 Turn off all access checking. This switch is mainly useful for working
1298 around bugs in the access control code.
1302 Check that the pointer returned by @code{operator new} is non-null
1303 before attempting to modify the storage allocated. This check is
1304 normally unnecessary because the C++ standard specifies that
1305 @code{operator new} will only return @code{0} if it is declared
1306 @samp{throw()}, in which case the compiler will always check the
1307 return value even without this option. In all other cases, when
1308 @code{operator new} has a non-empty exception specification, memory
1309 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1310 @samp{new (nothrow)}.
1312 @item -fconserve-space
1313 @opindex fconserve-space
1314 Put uninitialized or runtime-initialized global variables into the
1315 common segment, as C does. This saves space in the executable at the
1316 cost of not diagnosing duplicate definitions. If you compile with this
1317 flag and your program mysteriously crashes after @code{main()} has
1318 completed, you may have an object that is being destroyed twice because
1319 two definitions were merged.
1321 This option is no longer useful on most targets, now that support has
1322 been added for putting variables into BSS without making them common.
1324 @item -fno-const-strings
1325 @opindex fno-const-strings
1326 Give string constants type @code{char *} instead of type @code{const
1327 char *}. By default, G++ uses type @code{const char *} as required by
1328 the standard. Even if you use @option{-fno-const-strings}, you cannot
1329 actually modify the value of a string constant, unless you also use
1330 @option{-fwritable-strings}.
1332 This option might be removed in a future release of G++. For maximum
1333 portability, you should structure your code so that it works with
1334 string constants that have type @code{const char *}.
1336 @item -fno-elide-constructors
1337 @opindex fno-elide-constructors
1338 The C++ standard allows an implementation to omit creating a temporary
1339 which is only used to initialize another object of the same type.
1340 Specifying this option disables that optimization, and forces G++ to
1341 call the copy constructor in all cases.
1343 @item -fno-enforce-eh-specs
1344 @opindex fno-enforce-eh-specs
1345 Don't check for violation of exception specifications at runtime. This
1346 option violates the C++ standard, but may be useful for reducing code
1347 size in production builds, much like defining @samp{NDEBUG}. The compiler
1348 will still optimize based on the exception specifications.
1351 @itemx -fno-for-scope
1353 @opindex fno-for-scope
1354 If @option{-ffor-scope} is specified, the scope of variables declared in
1355 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1356 as specified by the C++ standard.
1357 If @option{-fno-for-scope} is specified, the scope of variables declared in
1358 a @i{for-init-statement} extends to the end of the enclosing scope,
1359 as was the case in old versions of G++, and other (traditional)
1360 implementations of C++.
1362 The default if neither flag is given to follow the standard,
1363 but to allow and give a warning for old-style code that would
1364 otherwise be invalid, or have different behavior.
1366 @item -fno-gnu-keywords
1367 @opindex fno-gnu-keywords
1368 Do not recognize @code{typeof} as a keyword, so that code can use this
1369 word as an identifier. You can use the keyword @code{__typeof__} instead.
1370 @option{-ansi} implies @option{-fno-gnu-keywords}.
1372 @item -fno-implicit-templates
1373 @opindex fno-implicit-templates
1374 Never emit code for non-inline templates which are instantiated
1375 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1376 @xref{Template Instantiation}, for more information.
1378 @item -fno-implicit-inline-templates
1379 @opindex fno-implicit-inline-templates
1380 Don't emit code for implicit instantiations of inline templates, either.
1381 The default is to handle inlines differently so that compiles with and
1382 without optimization will need the same set of explicit instantiations.
1384 @item -fno-implement-inlines
1385 @opindex fno-implement-inlines
1386 To save space, do not emit out-of-line copies of inline functions
1387 controlled by @samp{#pragma implementation}. This will cause linker
1388 errors if these functions are not inlined everywhere they are called.
1390 @item -fms-extensions
1391 @opindex fms-extensions
1392 Disable pedantic warnings about constructs used in MFC, such as implicit
1393 int and getting a pointer to member function via non-standard syntax.
1395 @item -fno-nonansi-builtins
1396 @opindex fno-nonansi-builtins
1397 Disable built-in declarations of functions that are not mandated by
1398 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1399 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1401 @item -fno-operator-names
1402 @opindex fno-operator-names
1403 Do not treat the operator name keywords @code{and}, @code{bitand},
1404 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1405 synonyms as keywords.
1407 @item -fno-optional-diags
1408 @opindex fno-optional-diags
1409 Disable diagnostics that the standard says a compiler does not need to
1410 issue. Currently, the only such diagnostic issued by G++ is the one for
1411 a name having multiple meanings within a class.
1414 @opindex fpermissive
1415 Downgrade some diagnostics about nonconformant code from errors to
1416 warnings. Thus, using @option{-fpermissive} will allow some
1417 nonconforming code to compile.
1421 Enable automatic template instantiation at link time. This option also
1422 implies @option{-fno-implicit-templates}. @xref{Template
1423 Instantiation}, for more information.
1427 Disable generation of information about every class with virtual
1428 functions for use by the C++ runtime type identification features
1429 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1430 of the language, you can save some space by using this flag. Note that
1431 exception handling uses the same information, but it will generate it as
1436 Emit statistics about front-end processing at the end of the compilation.
1437 This information is generally only useful to the G++ development team.
1439 @item -ftemplate-depth-@var{n}
1440 @opindex ftemplate-depth
1441 Set the maximum instantiation depth for template classes to @var{n}.
1442 A limit on the template instantiation depth is needed to detect
1443 endless recursions during template class instantiation. ANSI/ISO C++
1444 conforming programs must not rely on a maximum depth greater than 17.
1446 @item -fuse-cxa-atexit
1447 @opindex fuse-cxa-atexit
1448 Register destructors for objects with static storage duration with the
1449 @code{__cxa_atexit} function rather than the @code{atexit} function.
1450 This option is required for fully standards-compliant handling of static
1451 destructors, but will only work if your C library supports
1452 @code{__cxa_atexit}.
1456 Do not use weak symbol support, even if it is provided by the linker.
1457 By default, G++ will use weak symbols if they are available. This
1458 option exists only for testing, and should not be used by end-users;
1459 it will result in inferior code and has no benefits. This option may
1460 be removed in a future release of G++.
1464 Do not search for header files in the standard directories specific to
1465 C++, but do still search the other standard directories. (This option
1466 is used when building the C++ library.)
1469 In addition, these optimization, warning, and code generation options
1470 have meanings only for C++ programs:
1473 @item -fno-default-inline
1474 @opindex fno-default-inline
1475 Do not assume @samp{inline} for functions defined inside a class scope.
1476 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1477 functions will have linkage like inline functions; they just won't be
1480 @item -Wabi @r{(C++ only)}
1482 Warn when G++ generates code that is probably not compatible with the
1483 vendor-neutral C++ ABI. Although an effort has been made to warn about
1484 all such cases, there are probably some cases that are not warned about,
1485 even though G++ is generating incompatible code. There may also be
1486 cases where warnings are emitted even though the code that is generated
1489 You should rewrite your code to avoid these warnings if you are
1490 concerned about the fact that code generated by G++ may not be binary
1491 compatible with code generated by other compilers.
1493 The known incompatibilities at this point include:
1498 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1499 pack data into the same byte as a base class. For example:
1502 struct A @{ virtual void f(); int f1 : 1; @};
1503 struct B : public A @{ int f2 : 1; @};
1507 In this case, G++ will place @code{B::f2} into the same byte
1508 as@code{A::f1}; other compilers will not. You can avoid this problem
1509 by explicitly padding @code{A} so that its size is a multiple of the
1510 byte size on your platform; that will cause G++ and other compilers to
1511 layout @code{B} identically.
1514 Incorrect handling of tail-padding for virtual bases. G++ does not use
1515 tail padding when laying out virtual bases. For example:
1518 struct A @{ virtual void f(); char c1; @};
1519 struct B @{ B(); char c2; @};
1520 struct C : public A, public virtual B @{@};
1524 In this case, G++ will not place @code{B} into the tail-padding for
1525 @code{A}; other compilers will. You can avoid this problem by
1526 explicitly padding @code{A} so that its size is a multiple of its
1527 alignment (ignoring virtual base classes); that will cause G++ and other
1528 compilers to layout @code{C} identically.
1531 Incorrect handling of bit-fields with declared widths greater than that
1532 of their underlying types, when the bit-fields appear in a union. For
1536 union U @{ int i : 4096; @};
1540 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1541 union too small by the number of bits in an @code{int}.
1544 Empty classes can be placed at incorrect offsets. For example:
1554 struct C : public B, public A @{@};
1558 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1559 it should be placed at offset zero. G++ mistakenly believes that the
1560 @code{A} data member of @code{B} is already at offset zero.
1563 Names of template functions whose types involve @code{typename} or
1564 template template parameters can be mangled incorrectly.
1567 template <typename Q>
1568 void f(typename Q::X) @{@}
1570 template <template <typename> class Q>
1571 void f(typename Q<int>::X) @{@}
1575 Instantiations of these templates may be mangled incorrectly.
1579 @item -Wctor-dtor-privacy @r{(C++ only)}
1580 @opindex Wctor-dtor-privacy
1581 Warn when a class seems unusable because all the constructors or
1582 destructors in that class are private, and it has neither friends nor
1583 public static member functions.
1585 @item -Wnon-virtual-dtor @r{(C++ only)}
1586 @opindex Wnon-virtual-dtor
1587 Warn when a class appears to be polymorphic, thereby requiring a virtual
1588 destructor, yet it declares a non-virtual one.
1589 This warning is enabled by @option{-Wall}.
1591 @item -Wreorder @r{(C++ only)}
1593 @cindex reordering, warning
1594 @cindex warning for reordering of member initializers
1595 Warn when the order of member initializers given in the code does not
1596 match the order in which they must be executed. For instance:
1602 A(): j (0), i (1) @{ @}
1606 The compiler will rearrange the member initializers for @samp{i}
1607 and @samp{j} to match the declaration order of the members, emitting
1608 a warning to that effect. This warning is enabled by @option{-Wall}.
1611 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1614 @item -Weffc++ @r{(C++ only)}
1616 Warn about violations of the following style guidelines from Scott Meyers'
1617 @cite{Effective C++} book:
1621 Item 11: Define a copy constructor and an assignment operator for classes
1622 with dynamically allocated memory.
1625 Item 12: Prefer initialization to assignment in constructors.
1628 Item 14: Make destructors virtual in base classes.
1631 Item 15: Have @code{operator=} return a reference to @code{*this}.
1634 Item 23: Don't try to return a reference when you must return an object.
1638 Also warn about violations of the following style guidelines from
1639 Scott Meyers' @cite{More Effective C++} book:
1643 Item 6: Distinguish between prefix and postfix forms of increment and
1644 decrement operators.
1647 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1651 When selecting this option, be aware that the standard library
1652 headers do not obey all of these guidelines; use @samp{grep -v}
1653 to filter out those warnings.
1655 @item -Wno-deprecated @r{(C++ only)}
1656 @opindex Wno-deprecated
1657 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1659 @item -Wno-non-template-friend @r{(C++ only)}
1660 @opindex Wno-non-template-friend
1661 Disable warnings when non-templatized friend functions are declared
1662 within a template. Since the advent of explicit template specification
1663 support in G++, if the name of the friend is an unqualified-id (i.e.,
1664 @samp{friend foo(int)}), the C++ language specification demands that the
1665 friend declare or define an ordinary, nontemplate function. (Section
1666 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1667 could be interpreted as a particular specialization of a templatized
1668 function. Because this non-conforming behavior is no longer the default
1669 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1670 check existing code for potential trouble spots and is on by default.
1671 This new compiler behavior can be turned off with
1672 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1673 but disables the helpful warning.
1675 @item -Wold-style-cast @r{(C++ only)}
1676 @opindex Wold-style-cast
1677 Warn if an old-style (C-style) cast to a non-void type is used within
1678 a C++ program. The new-style casts (@samp{static_cast},
1679 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1680 unintended effects and much easier to search for.
1682 @item -Woverloaded-virtual @r{(C++ only)}
1683 @opindex Woverloaded-virtual
1684 @cindex overloaded virtual fn, warning
1685 @cindex warning for overloaded virtual fn
1686 Warn when a function declaration hides virtual functions from a
1687 base class. For example, in:
1694 struct B: public A @{
1699 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1707 will fail to compile.
1709 @item -Wno-pmf-conversions @r{(C++ only)}
1710 @opindex Wno-pmf-conversions
1711 Disable the diagnostic for converting a bound pointer to member function
1714 @item -Wsign-promo @r{(C++ only)}
1715 @opindex Wsign-promo
1716 Warn when overload resolution chooses a promotion from unsigned or
1717 enumeral type to a signed type, over a conversion to an unsigned type of
1718 the same size. Previous versions of G++ would try to preserve
1719 unsignedness, but the standard mandates the current behavior.
1721 @item -Wsynth @r{(C++ only)}
1723 @cindex warning for synthesized methods
1724 @cindex synthesized methods, warning
1725 Warn when G++'s synthesis behavior does not match that of cfront. For
1731 A& operator = (int);
1741 In this example, G++ will synthesize a default @samp{A& operator =
1742 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1745 @node Objective-C Dialect Options
1746 @section Options Controlling Objective-C Dialect
1748 @cindex compiler options, Objective-C
1749 @cindex Objective-C options, command line
1750 @cindex options, Objective-C
1751 (NOTE: This manual does not describe the Objective-C language itself. See
1752 @w{@uref{http://gcc.gnu.org/readings.html}} for references.)
1754 This section describes the command-line options that are only meaningful
1755 for Objective-C programs, but you can also use most of the GNU compiler
1756 options regardless of what language your program is in. For example,
1757 you might compile a file @code{some_class.m} like this:
1760 gcc -g -fgnu-runtime -O -c some_class.m
1764 In this example, @option{-fgnu-runtime} is an option meant only for
1765 Objective-C programs; you can use the other options with any language
1768 Here is a list of options that are @emph{only} for compiling Objective-C
1772 @item -fconstant-string-class=@var{class-name}
1773 @opindex fconstant-string-class
1774 Use @var{class-name} as the name of the class to instantiate for each
1775 literal string specified with the syntax @code{@@"@dots{}"}. The default
1776 class name is @code{NXConstantString} if the GNU runtime is being used, and
1777 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1778 @option{-fconstant-cfstrings} option, if also present, will override the
1779 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1780 to be laid out as constant CoreFoundation strings.
1783 @opindex fgnu-runtime
1784 Generate object code compatible with the standard GNU Objective-C
1785 runtime. This is the default for most types of systems.
1787 @item -fnext-runtime
1788 @opindex fnext-runtime
1789 Generate output compatible with the NeXT runtime. This is the default
1790 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1791 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1794 @item -fno-nil-receivers
1795 @opindex -fno-nil-receivers
1796 Assume that all Objective-C message dispatches (e.g.,
1797 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1798 is not @code{nil}. This allows for more efficient entry points in the runtime to be
1799 used. Currently, this option is only available in conjunction with
1800 the NeXT runtime on Mac OS X 10.3 and later.
1802 @item -fobjc-exceptions
1803 @opindex -fobjc-exceptions
1804 Enable syntactic support for structured exception handling in Objective-C,
1805 similar to what is offered by C++ and Java. Currently, this option is only
1806 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1814 @@catch (AnObjCClass *exc) @{
1821 @@catch (AnotherClass *exc) @{
1824 @@catch (id allOthers) @{
1834 The @code{@@throw} statement may appear anywhere in an Objective-C or
1835 Objective-C++ program; when used inside of a @code{@@catch} block, the
1836 @code{@@throw} may appear without an argument (as shown above), in which case
1837 the object caught by the @code{@@catch} will be rethrown.
1839 Note that only (pointers to) Objective-C objects may be thrown and
1840 caught using this scheme. When an object is thrown, it will be caught
1841 by the nearest @code{@@catch} clause capable of handling objects of that type,
1842 analogously to how @code{catch} blocks work in C++ and Java. A
1843 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1844 any and all Objective-C exceptions not caught by previous @code{@@catch}
1847 The @code{@@finally} clause, if present, will be executed upon exit from the
1848 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1849 regardless of whether any exceptions are thrown, caught or rethrown
1850 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1851 of the @code{finally} clause in Java.
1853 There are several caveats to using the new exception mechanism:
1857 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1858 idioms provided by the @code{NSException} class, the new
1859 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1860 systems, due to additional functionality needed in the (NeXT) Objective-C
1864 As mentioned above, the new exceptions do not support handling
1865 types other than Objective-C objects. Furthermore, when used from
1866 Objective-C++, the Objective-C exception model does not interoperate with C++
1867 exceptions at this time. This means you cannot @code{@@throw} an exception
1868 from Objective-C and @code{catch} it in C++, or vice versa
1869 (i.e., @code{throw @dots{} @@catch}).
1872 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1873 blocks for thread-safe execution:
1876 @@synchronized (ObjCClass *guard) @{
1881 Upon entering the @code{@@synchronized} block, a thread of execution shall
1882 first check whether a lock has been placed on the corresponding @code{guard}
1883 object by another thread. If it has, the current thread shall wait until
1884 the other thread relinquishes its lock. Once @code{guard} becomes available,
1885 the current thread will place its own lock on it, execute the code contained in
1886 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1887 making @code{guard} available to other threads).
1889 Unlike Java, Objective-C does not allow for entire methods to be marked
1890 @code{@@synchronized}. Note that throwing exceptions out of
1891 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1892 to be unlocked properly.
1894 @item -freplace-objc-classes
1895 @opindex -freplace-objc-classes
1896 Emit a special marker instructing @command{ld(1)} not to statically link in
1897 the resulting object file, and allow @command{dyld(1)} to load it in at
1898 run time instead. This is used in conjunction with the Fix-and-Continue
1899 debugging mode, where the object file in question may be recompiled and
1900 dynamically reloaded in the course of program execution, without the need
1901 to restart the program itself. Currently, Fix-and-Continue functionality
1902 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1906 @opindex -fzero-link
1907 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1908 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1909 compile time) with static class references that get initialized at load time,
1910 which improves run-time performance. Specifying the @option{-fzero-link} flag
1911 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1912 to be retained. This is useful in Zero-Link debugging mode, since it allows
1913 for individual class implementations to be modified during program execution.
1917 Dump interface declarations for all classes seen in the source file to a
1918 file named @file{@var{sourcename}.decl}.
1921 @opindex Wno-protocol
1922 If a class is declared to implement a protocol, a warning is issued for
1923 every method in the protocol that is not implemented by the class. The
1924 default behavior is to issue a warning for every method not explicitly
1925 implemented in the class, even if a method implementation is inherited
1926 from the superclass. If you use the @code{-Wno-protocol} option, then
1927 methods inherited from the superclass are considered to be implemented,
1928 and no warning is issued for them.
1932 Warn if multiple methods of different types for the same selector are
1933 found during compilation. The check is performed on the list of methods
1934 in the final stage of compilation. Additionally, a check is performed
1935 for each selector appearing in a @code{@@selector(@dots{})}
1936 expression, and a corresponding method for that selector has been found
1937 during compilation. Because these checks scan the method table only at
1938 the end of compilation, these warnings are not produced if the final
1939 stage of compilation is not reached, for example because an error is
1940 found during compilation, or because the @code{-fsyntax-only} option is
1943 @item -Wundeclared-selector
1944 @opindex Wundeclared-selector
1945 Warn if a @code{@@selector(@dots{})} expression referring to an
1946 undeclared selector is found. A selector is considered undeclared if no
1947 method with that name has been declared before the
1948 @code{@@selector(@dots{})} expression, either explicitly in an
1949 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
1950 an @code{@@implementation} section. This option always performs its
1951 checks as soon as a @code{@@selector(@dots{})} expression is found,
1952 while @code{-Wselector} only performs its checks in the final stage of
1953 compilation. This also enforces the coding style convention
1954 that methods and selectors must be declared before being used.
1956 @item -print-objc-runtime-info
1957 @opindex -print-objc-runtime-info
1958 Generate C header describing the largest structure that is passed by
1963 @node Language Independent Options
1964 @section Options to Control Diagnostic Messages Formatting
1965 @cindex options to control diagnostics formatting
1966 @cindex diagnostic messages
1967 @cindex message formatting
1969 Traditionally, diagnostic messages have been formatted irrespective of
1970 the output device's aspect (e.g.@: its width, @dots{}). The options described
1971 below can be used to control the diagnostic messages formatting
1972 algorithm, e.g.@: how many characters per line, how often source location
1973 information should be reported. Right now, only the C++ front end can
1974 honor these options. However it is expected, in the near future, that
1975 the remaining front ends would be able to digest them correctly.
1978 @item -fmessage-length=@var{n}
1979 @opindex fmessage-length
1980 Try to format error messages so that they fit on lines of about @var{n}
1981 characters. The default is 72 characters for @command{g++} and 0 for the rest of
1982 the front ends supported by GCC@. If @var{n} is zero, then no
1983 line-wrapping will be done; each error message will appear on a single
1986 @opindex fdiagnostics-show-location
1987 @item -fdiagnostics-show-location=once
1988 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1989 reporter to emit @emph{once} source location information; that is, in
1990 case the message is too long to fit on a single physical line and has to
1991 be wrapped, the source location won't be emitted (as prefix) again,
1992 over and over, in subsequent continuation lines. This is the default
1995 @item -fdiagnostics-show-location=every-line
1996 Only meaningful in line-wrapping mode. Instructs the diagnostic
1997 messages reporter to emit the same source location information (as
1998 prefix) for physical lines that result from the process of breaking
1999 a message which is too long to fit on a single line.
2003 @node Warning Options
2004 @section Options to Request or Suppress Warnings
2005 @cindex options to control warnings
2006 @cindex warning messages
2007 @cindex messages, warning
2008 @cindex suppressing warnings
2010 Warnings are diagnostic messages that report constructions which
2011 are not inherently erroneous but which are risky or suggest there
2012 may have been an error.
2014 You can request many specific warnings with options beginning @samp{-W},
2015 for example @option{-Wimplicit} to request warnings on implicit
2016 declarations. Each of these specific warning options also has a
2017 negative form beginning @samp{-Wno-} to turn off warnings;
2018 for example, @option{-Wno-implicit}. This manual lists only one of the
2019 two forms, whichever is not the default.
2021 The following options control the amount and kinds of warnings produced
2022 by GCC; for further, language-specific options also refer to
2023 @ref{C++ Dialect Options} and @ref{Objective-C Dialect Options}.
2026 @cindex syntax checking
2028 @opindex fsyntax-only
2029 Check the code for syntax errors, but don't do anything beyond that.
2033 Issue all the warnings demanded by strict ISO C and ISO C++;
2034 reject all programs that use forbidden extensions, and some other
2035 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2036 version of the ISO C standard specified by any @option{-std} option used.
2038 Valid ISO C and ISO C++ programs should compile properly with or without
2039 this option (though a rare few will require @option{-ansi} or a
2040 @option{-std} option specifying the required version of ISO C)@. However,
2041 without this option, certain GNU extensions and traditional C and C++
2042 features are supported as well. With this option, they are rejected.
2044 @option{-pedantic} does not cause warning messages for use of the
2045 alternate keywords whose names begin and end with @samp{__}. Pedantic
2046 warnings are also disabled in the expression that follows
2047 @code{__extension__}. However, only system header files should use
2048 these escape routes; application programs should avoid them.
2049 @xref{Alternate Keywords}.
2051 Some users try to use @option{-pedantic} to check programs for strict ISO
2052 C conformance. They soon find that it does not do quite what they want:
2053 it finds some non-ISO practices, but not all---only those for which
2054 ISO C @emph{requires} a diagnostic, and some others for which
2055 diagnostics have been added.
2057 A feature to report any failure to conform to ISO C might be useful in
2058 some instances, but would require considerable additional work and would
2059 be quite different from @option{-pedantic}. We don't have plans to
2060 support such a feature in the near future.
2062 Where the standard specified with @option{-std} represents a GNU
2063 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2064 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2065 extended dialect is based. Warnings from @option{-pedantic} are given
2066 where they are required by the base standard. (It would not make sense
2067 for such warnings to be given only for features not in the specified GNU
2068 C dialect, since by definition the GNU dialects of C include all
2069 features the compiler supports with the given option, and there would be
2070 nothing to warn about.)
2072 @item -pedantic-errors
2073 @opindex pedantic-errors
2074 Like @option{-pedantic}, except that errors are produced rather than
2079 Inhibit all warning messages.
2083 Inhibit warning messages about the use of @samp{#import}.
2085 @item -Wchar-subscripts
2086 @opindex Wchar-subscripts
2087 Warn if an array subscript has type @code{char}. This is a common cause
2088 of error, as programmers often forget that this type is signed on some
2093 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2094 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2098 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2099 the arguments supplied have types appropriate to the format string
2100 specified, and that the conversions specified in the format string make
2101 sense. This includes standard functions, and others specified by format
2102 attributes (@pxref{Function Attributes}), in the @code{printf},
2103 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2104 not in the C standard) families.
2106 The formats are checked against the format features supported by GNU
2107 libc version 2.2. These include all ISO C90 and C99 features, as well
2108 as features from the Single Unix Specification and some BSD and GNU
2109 extensions. Other library implementations may not support all these
2110 features; GCC does not support warning about features that go beyond a
2111 particular library's limitations. However, if @option{-pedantic} is used
2112 with @option{-Wformat}, warnings will be given about format features not
2113 in the selected standard version (but not for @code{strfmon} formats,
2114 since those are not in any version of the C standard). @xref{C Dialect
2115 Options,,Options Controlling C Dialect}.
2117 Since @option{-Wformat} also checks for null format arguments for
2118 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2120 @option{-Wformat} is included in @option{-Wall}. For more control over some
2121 aspects of format checking, the options @option{-Wformat-y2k},
2122 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2123 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2124 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2127 @opindex Wformat-y2k
2128 If @option{-Wformat} is specified, also warn about @code{strftime}
2129 formats which may yield only a two-digit year.
2131 @item -Wno-format-extra-args
2132 @opindex Wno-format-extra-args
2133 If @option{-Wformat} is specified, do not warn about excess arguments to a
2134 @code{printf} or @code{scanf} format function. The C standard specifies
2135 that such arguments are ignored.
2137 Where the unused arguments lie between used arguments that are
2138 specified with @samp{$} operand number specifications, normally
2139 warnings are still given, since the implementation could not know what
2140 type to pass to @code{va_arg} to skip the unused arguments. However,
2141 in the case of @code{scanf} formats, this option will suppress the
2142 warning if the unused arguments are all pointers, since the Single
2143 Unix Specification says that such unused arguments are allowed.
2145 @item -Wno-format-zero-length
2146 @opindex Wno-format-zero-length
2147 If @option{-Wformat} is specified, do not warn about zero-length formats.
2148 The C standard specifies that zero-length formats are allowed.
2150 @item -Wformat-nonliteral
2151 @opindex Wformat-nonliteral
2152 If @option{-Wformat} is specified, also warn if the format string is not a
2153 string literal and so cannot be checked, unless the format function
2154 takes its format arguments as a @code{va_list}.
2156 @item -Wformat-security
2157 @opindex Wformat-security
2158 If @option{-Wformat} is specified, also warn about uses of format
2159 functions that represent possible security problems. At present, this
2160 warns about calls to @code{printf} and @code{scanf} functions where the
2161 format string is not a string literal and there are no format arguments,
2162 as in @code{printf (foo);}. This may be a security hole if the format
2163 string came from untrusted input and contains @samp{%n}. (This is
2164 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2165 in future warnings may be added to @option{-Wformat-security} that are not
2166 included in @option{-Wformat-nonliteral}.)
2170 Enable @option{-Wformat} plus format checks not included in
2171 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2172 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2176 Warn about passing a null pointer for arguments marked as
2177 requiring a non-null value by the @code{nonnull} function attribute.
2179 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2180 can be disabled with the @option{-Wno-nonnull} option.
2182 @item -Winit-self @r{(C, C++, and Objective-C only)}
2184 Warn about uninitialized variables which are initialized with themselves.
2185 Note this option can only be used with the @option{-Wuninitialized} option,
2186 which in turn only works with @option{-O1} and above.
2188 For example, GCC will warn about @code{i} being uninitialized in the
2189 following snippet only when @option{-Winit-self} has been specified:
2200 @item -Wimplicit-int
2201 @opindex Wimplicit-int
2202 Warn when a declaration does not specify a type.
2204 @item -Wimplicit-function-declaration
2205 @itemx -Werror-implicit-function-declaration
2206 @opindex Wimplicit-function-declaration
2207 @opindex Werror-implicit-function-declaration
2208 Give a warning (or error) whenever a function is used before being
2213 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2217 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2218 function with external linkage, returning int, taking either zero
2219 arguments, two, or three arguments of appropriate types.
2221 @item -Wmissing-braces
2222 @opindex Wmissing-braces
2223 Warn if an aggregate or union initializer is not fully bracketed. In
2224 the following example, the initializer for @samp{a} is not fully
2225 bracketed, but that for @samp{b} is fully bracketed.
2228 int a[2][2] = @{ 0, 1, 2, 3 @};
2229 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2233 @opindex Wparentheses
2234 Warn if parentheses are omitted in certain contexts, such
2235 as when there is an assignment in a context where a truth value
2236 is expected, or when operators are nested whose precedence people
2237 often get confused about.
2239 Also warn about constructions where there may be confusion to which
2240 @code{if} statement an @code{else} branch belongs. Here is an example of
2255 In C, every @code{else} branch belongs to the innermost possible @code{if}
2256 statement, which in this example is @code{if (b)}. This is often not
2257 what the programmer expected, as illustrated in the above example by
2258 indentation the programmer chose. When there is the potential for this
2259 confusion, GCC will issue a warning when this flag is specified.
2260 To eliminate the warning, add explicit braces around the innermost
2261 @code{if} statement so there is no way the @code{else} could belong to
2262 the enclosing @code{if}. The resulting code would look like this:
2278 @item -Wsequence-point
2279 @opindex Wsequence-point
2280 Warn about code that may have undefined semantics because of violations
2281 of sequence point rules in the C standard.
2283 The C standard defines the order in which expressions in a C program are
2284 evaluated in terms of @dfn{sequence points}, which represent a partial
2285 ordering between the execution of parts of the program: those executed
2286 before the sequence point, and those executed after it. These occur
2287 after the evaluation of a full expression (one which is not part of a
2288 larger expression), after the evaluation of the first operand of a
2289 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2290 function is called (but after the evaluation of its arguments and the
2291 expression denoting the called function), and in certain other places.
2292 Other than as expressed by the sequence point rules, the order of
2293 evaluation of subexpressions of an expression is not specified. All
2294 these rules describe only a partial order rather than a total order,
2295 since, for example, if two functions are called within one expression
2296 with no sequence point between them, the order in which the functions
2297 are called is not specified. However, the standards committee have
2298 ruled that function calls do not overlap.
2300 It is not specified when between sequence points modifications to the
2301 values of objects take effect. Programs whose behavior depends on this
2302 have undefined behavior; the C standard specifies that ``Between the
2303 previous and next sequence point an object shall have its stored value
2304 modified at most once by the evaluation of an expression. Furthermore,
2305 the prior value shall be read only to determine the value to be
2306 stored.''. If a program breaks these rules, the results on any
2307 particular implementation are entirely unpredictable.
2309 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2310 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2311 diagnosed by this option, and it may give an occasional false positive
2312 result, but in general it has been found fairly effective at detecting
2313 this sort of problem in programs.
2315 The present implementation of this option only works for C programs. A
2316 future implementation may also work for C++ programs.
2318 The C standard is worded confusingly, therefore there is some debate
2319 over the precise meaning of the sequence point rules in subtle cases.
2320 Links to discussions of the problem, including proposed formal
2321 definitions, may be found on our readings page, at
2322 @w{@uref{http://gcc.gnu.org/readings.html}}.
2325 @opindex Wreturn-type
2326 Warn whenever a function is defined with a return-type that defaults to
2327 @code{int}. Also warn about any @code{return} statement with no
2328 return-value in a function whose return-type is not @code{void}.
2330 For C++, a function without return type always produces a diagnostic
2331 message, even when @option{-Wno-return-type} is specified. The only
2332 exceptions are @samp{main} and functions defined in system headers.
2336 Warn whenever a @code{switch} statement has an index of enumeral type
2337 and lacks a @code{case} for one or more of the named codes of that
2338 enumeration. (The presence of a @code{default} label prevents this
2339 warning.) @code{case} labels outside the enumeration range also
2340 provoke warnings when this option is used.
2342 @item -Wswitch-default
2343 @opindex Wswitch-switch
2344 Warn whenever a @code{switch} statement does not have a @code{default}
2348 @opindex Wswitch-enum
2349 Warn whenever a @code{switch} statement has an index of enumeral type
2350 and lacks a @code{case} for one or more of the named codes of that
2351 enumeration. @code{case} labels outside the enumeration range also
2352 provoke warnings when this option is used.
2356 Warn if any trigraphs are encountered that might change the meaning of
2357 the program (trigraphs within comments are not warned about).
2359 @item -Wunused-function
2360 @opindex Wunused-function
2361 Warn whenever a static function is declared but not defined or a
2362 non\-inline static function is unused.
2364 @item -Wunused-label
2365 @opindex Wunused-label
2366 Warn whenever a label is declared but not used.
2368 To suppress this warning use the @samp{unused} attribute
2369 (@pxref{Variable Attributes}).
2371 @item -Wunused-parameter
2372 @opindex Wunused-parameter
2373 Warn whenever a function parameter is unused aside from its declaration.
2375 To suppress this warning use the @samp{unused} attribute
2376 (@pxref{Variable Attributes}).
2378 @item -Wunused-variable
2379 @opindex Wunused-variable
2380 Warn whenever a local variable or non-constant static variable is unused
2381 aside from its declaration
2383 To suppress this warning use the @samp{unused} attribute
2384 (@pxref{Variable Attributes}).
2386 @item -Wunused-value
2387 @opindex Wunused-value
2388 Warn whenever a statement computes a result that is explicitly not used.
2390 To suppress this warning cast the expression to @samp{void}.
2394 All the above @option{-Wunused} options combined.
2396 In order to get a warning about an unused function parameter, you must
2397 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2398 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2400 @item -Wuninitialized
2401 @opindex Wuninitialized
2402 Warn if an automatic variable is used without first being initialized or
2403 if a variable may be clobbered by a @code{setjmp} call.
2405 These warnings are possible only in optimizing compilation,
2406 because they require data flow information that is computed only
2407 when optimizing. If you don't specify @option{-O}, you simply won't
2410 If you want to warn about code which uses the uninitialized value of the
2411 variable in its own initializer, use the @option{-Winit-self} option.
2413 These warnings occur only for variables that are candidates for
2414 register allocation. Therefore, they do not occur for a variable that
2415 is declared @code{volatile}, or whose address is taken, or whose size
2416 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2417 structures, unions or arrays, even when they are in registers.
2419 Note that there may be no warning about a variable that is used only
2420 to compute a value that itself is never used, because such
2421 computations may be deleted by data flow analysis before the warnings
2424 These warnings are made optional because GCC is not smart
2425 enough to see all the reasons why the code might be correct
2426 despite appearing to have an error. Here is one example of how
2447 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2448 always initialized, but GCC doesn't know this. Here is
2449 another common case:
2454 if (change_y) save_y = y, y = new_y;
2456 if (change_y) y = save_y;
2461 This has no bug because @code{save_y} is used only if it is set.
2463 @cindex @code{longjmp} warnings
2464 This option also warns when a non-volatile automatic variable might be
2465 changed by a call to @code{longjmp}. These warnings as well are possible
2466 only in optimizing compilation.
2468 The compiler sees only the calls to @code{setjmp}. It cannot know
2469 where @code{longjmp} will be called; in fact, a signal handler could
2470 call it at any point in the code. As a result, you may get a warning
2471 even when there is in fact no problem because @code{longjmp} cannot
2472 in fact be called at the place which would cause a problem.
2474 Some spurious warnings can be avoided if you declare all the functions
2475 you use that never return as @code{noreturn}. @xref{Function
2478 @item -Wunknown-pragmas
2479 @opindex Wunknown-pragmas
2480 @cindex warning for unknown pragmas
2481 @cindex unknown pragmas, warning
2482 @cindex pragmas, warning of unknown
2483 Warn when a #pragma directive is encountered which is not understood by
2484 GCC@. If this command line option is used, warnings will even be issued
2485 for unknown pragmas in system header files. This is not the case if
2486 the warnings were only enabled by the @option{-Wall} command line option.
2488 @item -Wstrict-aliasing
2489 @opindex Wstrict-aliasing
2490 This option is only active when @option{-fstrict-aliasing} is active.
2491 It warns about code which might break the strict aliasing rules that the
2492 compiler is using for optimization. The warning does not catch all
2493 cases, but does attempt to catch the more common pitfalls. It is
2494 included in @option{-Wall}.
2498 All of the above @samp{-W} options combined. This enables all the
2499 warnings about constructions that some users consider questionable, and
2500 that are easy to avoid (or modify to prevent the warning), even in
2501 conjunction with macros. This also enables some language-specific
2502 warnings described in @ref{C++ Dialect Options} and
2503 @ref{Objective-C Dialect Options}.
2506 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2507 Some of them warn about constructions that users generally do not
2508 consider questionable, but which occasionally you might wish to check
2509 for; others warn about constructions that are necessary or hard to avoid
2510 in some cases, and there is no simple way to modify the code to suppress
2517 (This option used to be called @option{-W}. The older name is still
2518 supported, but the newer name is more descriptive.) Print extra warning
2519 messages for these events:
2523 A function can return either with or without a value. (Falling
2524 off the end of the function body is considered returning without
2525 a value.) For example, this function would evoke such a
2539 An expression-statement or the left-hand side of a comma expression
2540 contains no side effects.
2541 To suppress the warning, cast the unused expression to void.
2542 For example, an expression such as @samp{x[i,j]} will cause a warning,
2543 but @samp{x[(void)i,j]} will not.
2546 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2549 A comparison like @samp{x<=y<=z} appears; this is equivalent to
2550 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
2551 that of ordinary mathematical notation.
2554 Storage-class specifiers like @code{static} are not the first things in
2555 a declaration. According to the C Standard, this usage is obsolescent.
2558 The return type of a function has a type qualifier such as @code{const}.
2559 Such a type qualifier has no effect, since the value returned by a
2560 function is not an lvalue. (But don't warn about the GNU extension of
2561 @code{volatile void} return types. That extension will be warned about
2562 if @option{-pedantic} is specified.)
2565 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2569 A comparison between signed and unsigned values could produce an
2570 incorrect result when the signed value is converted to unsigned.
2571 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2574 An aggregate has an initializer which does not initialize all members.
2575 For example, the following code would cause such a warning, because
2576 @code{x.h} would be implicitly initialized to zero:
2579 struct s @{ int f, g, h; @};
2580 struct s x = @{ 3, 4 @};
2584 A function parameter is declared without a type specifier in K&R-style
2592 An empty body occurs in an @samp{if} or @samp{else} statement.
2595 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2596 @samp{>}, or @samp{>=}.
2599 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2602 Any of several floating-point events that often indicate errors, such as
2603 overflow, underflow, loss of precision, etc.
2605 @item @r{(C++ only)}
2606 An enumerator and a non-enumerator both appear in a conditional expression.
2608 @item @r{(C++ only)}
2609 A non-static reference or non-static @samp{const} member appears in a
2610 class without constructors.
2612 @item @r{(C++ only)}
2613 Ambiguous virtual bases.
2615 @item @r{(C++ only)}
2616 Subscripting an array which has been declared @samp{register}.
2618 @item @r{(C++ only)}
2619 Taking the address of a variable which has been declared @samp{register}.
2621 @item @r{(C++ only)}
2622 A base class is not initialized in a derived class' copy constructor.
2625 @item -Wno-div-by-zero
2626 @opindex Wno-div-by-zero
2627 @opindex Wdiv-by-zero
2628 Do not warn about compile-time integer division by zero. Floating point
2629 division by zero is not warned about, as it can be a legitimate way of
2630 obtaining infinities and NaNs.
2632 @item -Wsystem-headers
2633 @opindex Wsystem-headers
2634 @cindex warnings from system headers
2635 @cindex system headers, warnings from
2636 Print warning messages for constructs found in system header files.
2637 Warnings from system headers are normally suppressed, on the assumption
2638 that they usually do not indicate real problems and would only make the
2639 compiler output harder to read. Using this command line option tells
2640 GCC to emit warnings from system headers as if they occurred in user
2641 code. However, note that using @option{-Wall} in conjunction with this
2642 option will @emph{not} warn about unknown pragmas in system
2643 headers---for that, @option{-Wunknown-pragmas} must also be used.
2646 @opindex Wfloat-equal
2647 Warn if floating point values are used in equality comparisons.
2649 The idea behind this is that sometimes it is convenient (for the
2650 programmer) to consider floating-point values as approximations to
2651 infinitely precise real numbers. If you are doing this, then you need
2652 to compute (by analyzing the code, or in some other way) the maximum or
2653 likely maximum error that the computation introduces, and allow for it
2654 when performing comparisons (and when producing output, but that's a
2655 different problem). In particular, instead of testing for equality, you
2656 would check to see whether the two values have ranges that overlap; and
2657 this is done with the relational operators, so equality comparisons are
2660 @item -Wtraditional @r{(C only)}
2661 @opindex Wtraditional
2662 Warn about certain constructs that behave differently in traditional and
2663 ISO C@. Also warn about ISO C constructs that have no traditional C
2664 equivalent, and/or problematic constructs which should be avoided.
2668 Macro parameters that appear within string literals in the macro body.
2669 In traditional C macro replacement takes place within string literals,
2670 but does not in ISO C@.
2673 In traditional C, some preprocessor directives did not exist.
2674 Traditional preprocessors would only consider a line to be a directive
2675 if the @samp{#} appeared in column 1 on the line. Therefore
2676 @option{-Wtraditional} warns about directives that traditional C
2677 understands but would ignore because the @samp{#} does not appear as the
2678 first character on the line. It also suggests you hide directives like
2679 @samp{#pragma} not understood by traditional C by indenting them. Some
2680 traditional implementations would not recognize @samp{#elif}, so it
2681 suggests avoiding it altogether.
2684 A function-like macro that appears without arguments.
2687 The unary plus operator.
2690 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2691 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2692 constants.) Note, these suffixes appear in macros defined in the system
2693 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2694 Use of these macros in user code might normally lead to spurious
2695 warnings, however gcc's integrated preprocessor has enough context to
2696 avoid warning in these cases.
2699 A function declared external in one block and then used after the end of
2703 A @code{switch} statement has an operand of type @code{long}.
2706 A non-@code{static} function declaration follows a @code{static} one.
2707 This construct is not accepted by some traditional C compilers.
2710 The ISO type of an integer constant has a different width or
2711 signedness from its traditional type. This warning is only issued if
2712 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2713 typically represent bit patterns, are not warned about.
2716 Usage of ISO string concatenation is detected.
2719 Initialization of automatic aggregates.
2722 Identifier conflicts with labels. Traditional C lacks a separate
2723 namespace for labels.
2726 Initialization of unions. If the initializer is zero, the warning is
2727 omitted. This is done under the assumption that the zero initializer in
2728 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2729 initializer warnings and relies on default initialization to zero in the
2733 Conversions by prototypes between fixed/floating point values and vice
2734 versa. The absence of these prototypes when compiling with traditional
2735 C would cause serious problems. This is a subset of the possible
2736 conversion warnings, for the full set use @option{-Wconversion}.
2739 Use of ISO C style function definitions. This warning intentionally is
2740 @emph{not} issued for prototype declarations or variadic functions
2741 because these ISO C features will appear in your code when using
2742 libiberty's traditional C compatibility macros, @code{PARAMS} and
2743 @code{VPARAMS}. This warning is also bypassed for nested functions
2744 because that feature is already a gcc extension and thus not relevant to
2745 traditional C compatibility.
2748 @item -Wdeclaration-after-statement @r{(C only)}
2749 @opindex Wdeclaration-after-statement
2750 Warn when a declaration is found after a statement in a block. This
2751 construct, known from C++, was introduced with ISO C99 and is by default
2752 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2753 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2757 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2759 @item -Wendif-labels
2760 @opindex Wendif-labels
2761 Warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2765 Warn whenever a local variable shadows another local variable, parameter or
2766 global variable or whenever a built-in function is shadowed.
2768 @item -Wlarger-than-@var{len}
2769 @opindex Wlarger-than
2770 Warn whenever an object of larger than @var{len} bytes is defined.
2772 @item -Wpointer-arith
2773 @opindex Wpointer-arith
2774 Warn about anything that depends on the ``size of'' a function type or
2775 of @code{void}. GNU C assigns these types a size of 1, for
2776 convenience in calculations with @code{void *} pointers and pointers
2779 @item -Wbad-function-cast @r{(C only)}
2780 @opindex Wbad-function-cast
2781 Warn whenever a function call is cast to a non-matching type.
2782 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2786 Warn whenever a pointer is cast so as to remove a type qualifier from
2787 the target type. For example, warn if a @code{const char *} is cast
2788 to an ordinary @code{char *}.
2791 @opindex Wcast-align
2792 Warn whenever a pointer is cast such that the required alignment of the
2793 target is increased. For example, warn if a @code{char *} is cast to
2794 an @code{int *} on machines where integers can only be accessed at
2795 two- or four-byte boundaries.
2797 @item -Wwrite-strings
2798 @opindex Wwrite-strings
2799 When compiling C, give string constants the type @code{const
2800 char[@var{length}]} so that
2801 copying the address of one into a non-@code{const} @code{char *}
2802 pointer will get a warning; when compiling C++, warn about the
2803 deprecated conversion from string constants to @code{char *}.
2804 These warnings will help you find at
2805 compile time code that can try to write into a string constant, but
2806 only if you have been very careful about using @code{const} in
2807 declarations and prototypes. Otherwise, it will just be a nuisance;
2808 this is why we did not make @option{-Wall} request these warnings.
2811 @opindex Wconversion
2812 Warn if a prototype causes a type conversion that is different from what
2813 would happen to the same argument in the absence of a prototype. This
2814 includes conversions of fixed point to floating and vice versa, and
2815 conversions changing the width or signedness of a fixed point argument
2816 except when the same as the default promotion.
2818 Also, warn if a negative integer constant expression is implicitly
2819 converted to an unsigned type. For example, warn about the assignment
2820 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2821 casts like @code{(unsigned) -1}.
2823 @item -Wsign-compare
2824 @opindex Wsign-compare
2825 @cindex warning for comparison of signed and unsigned values
2826 @cindex comparison of signed and unsigned values, warning
2827 @cindex signed and unsigned values, comparison warning
2828 Warn when a comparison between signed and unsigned values could produce
2829 an incorrect result when the signed value is converted to unsigned.
2830 This warning is also enabled by @option{-Wextra}; to get the other warnings
2831 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2833 @item -Waggregate-return
2834 @opindex Waggregate-return
2835 Warn if any functions that return structures or unions are defined or
2836 called. (In languages where you can return an array, this also elicits
2839 @item -Wstrict-prototypes @r{(C only)}
2840 @opindex Wstrict-prototypes
2841 Warn if a function is declared or defined without specifying the
2842 argument types. (An old-style function definition is permitted without
2843 a warning if preceded by a declaration which specifies the argument
2846 @item -Wold-style-definition @r{(C only)}
2847 @opindex Wold-style-definition
2848 Warn if an old-style function definition is used. A warning is given
2849 even if there is a previous prototype.
2851 @item -Wmissing-prototypes @r{(C only)}
2852 @opindex Wmissing-prototypes
2853 Warn if a global function is defined without a previous prototype
2854 declaration. This warning is issued even if the definition itself
2855 provides a prototype. The aim is to detect global functions that fail
2856 to be declared in header files.
2858 @item -Wmissing-declarations @r{(C only)}
2859 @opindex Wmissing-declarations
2860 Warn if a global function is defined without a previous declaration.
2861 Do so even if the definition itself provides a prototype.
2862 Use this option to detect global functions that are not declared in
2865 @item -Wmissing-noreturn
2866 @opindex Wmissing-noreturn
2867 Warn about functions which might be candidates for attribute @code{noreturn}.
2868 Note these are only possible candidates, not absolute ones. Care should
2869 be taken to manually verify functions actually do not ever return before
2870 adding the @code{noreturn} attribute, otherwise subtle code generation
2871 bugs could be introduced. You will not get a warning for @code{main} in
2872 hosted C environments.
2874 @item -Wmissing-format-attribute
2875 @opindex Wmissing-format-attribute
2877 If @option{-Wformat} is enabled, also warn about functions which might be
2878 candidates for @code{format} attributes. Note these are only possible
2879 candidates, not absolute ones. GCC will guess that @code{format}
2880 attributes might be appropriate for any function that calls a function
2881 like @code{vprintf} or @code{vscanf}, but this might not always be the
2882 case, and some functions for which @code{format} attributes are
2883 appropriate may not be detected. This option has no effect unless
2884 @option{-Wformat} is enabled (possibly by @option{-Wall}).
2886 @item -Wno-multichar
2887 @opindex Wno-multichar
2889 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
2890 Usually they indicate a typo in the user's code, as they have
2891 implementation-defined values, and should not be used in portable code.
2893 @item -Wno-deprecated-declarations
2894 @opindex Wno-deprecated-declarations
2895 Do not warn about uses of functions, variables, and types marked as
2896 deprecated by using the @code{deprecated} attribute.
2897 (@pxref{Function Attributes}, @pxref{Variable Attributes},
2898 @pxref{Type Attributes}.)
2902 Warn if a structure is given the packed attribute, but the packed
2903 attribute has no effect on the layout or size of the structure.
2904 Such structures may be mis-aligned for little benefit. For
2905 instance, in this code, the variable @code{f.x} in @code{struct bar}
2906 will be misaligned even though @code{struct bar} does not itself
2907 have the packed attribute:
2914 @} __attribute__((packed));
2924 Warn if padding is included in a structure, either to align an element
2925 of the structure or to align the whole structure. Sometimes when this
2926 happens it is possible to rearrange the fields of the structure to
2927 reduce the padding and so make the structure smaller.
2929 @item -Wredundant-decls
2930 @opindex Wredundant-decls
2931 Warn if anything is declared more than once in the same scope, even in
2932 cases where multiple declaration is valid and changes nothing.
2934 @item -Wnested-externs @r{(C only)}
2935 @opindex Wnested-externs
2936 Warn if an @code{extern} declaration is encountered within a function.
2938 @item -Wunreachable-code
2939 @opindex Wunreachable-code
2940 Warn if the compiler detects that code will never be executed.
2942 This option is intended to warn when the compiler detects that at
2943 least a whole line of source code will never be executed, because
2944 some condition is never satisfied or because it is after a
2945 procedure that never returns.
2947 It is possible for this option to produce a warning even though there
2948 are circumstances under which part of the affected line can be executed,
2949 so care should be taken when removing apparently-unreachable code.
2951 For instance, when a function is inlined, a warning may mean that the
2952 line is unreachable in only one inlined copy of the function.
2954 This option is not made part of @option{-Wall} because in a debugging
2955 version of a program there is often substantial code which checks
2956 correct functioning of the program and is, hopefully, unreachable
2957 because the program does work. Another common use of unreachable
2958 code is to provide behavior which is selectable at compile-time.
2962 Warn if a function can not be inlined and it was declared as inline.
2963 Even with this option, the compiler will not warn about failures to
2964 inline functions declared in system headers.
2966 The compiler uses a variety of heuristics to determine whether or not
2967 to inline a function. For example, the compiler takes into account
2968 the size of the function being inlined and the the amount of inlining
2969 that has already been done in the current function. Therefore,
2970 seemingly insignificant changes in the source program can cause the
2971 warnings produced by @option{-Winline} to appear or disappear.
2973 @item -Wno-invalid-offsetof @r{(C++ only)}
2974 @opindex Wno-invalid-offsetof
2975 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
2976 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
2977 to a non-POD type is undefined. In existing C++ implementations,
2978 however, @samp{offsetof} typically gives meaningful results even when
2979 applied to certain kinds of non-POD types. (Such as a simple
2980 @samp{struct} that fails to be a POD type only by virtue of having a
2981 constructor.) This flag is for users who are aware that they are
2982 writing nonportable code and who have deliberately chosen to ignore the
2985 The restrictions on @samp{offsetof} may be relaxed in a future version
2986 of the C++ standard.
2989 @opindex Winvalid-pch
2990 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
2991 the search path but can't be used.
2995 @opindex Wno-long-long
2996 Warn if @samp{long long} type is used. This is default. To inhibit
2997 the warning messages, use @option{-Wno-long-long}. Flags
2998 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
2999 only when @option{-pedantic} flag is used.
3001 @item -Wdisabled-optimization
3002 @opindex Wdisabled-optimization
3003 Warn if a requested optimization pass is disabled. This warning does
3004 not generally indicate that there is anything wrong with your code; it
3005 merely indicates that GCC's optimizers were unable to handle the code
3006 effectively. Often, the problem is that your code is too big or too
3007 complex; GCC will refuse to optimize programs when the optimization
3008 itself is likely to take inordinate amounts of time.
3012 Make all warnings into errors.
3015 @node Debugging Options
3016 @section Options for Debugging Your Program or GCC
3017 @cindex options, debugging
3018 @cindex debugging information options
3020 GCC has various special options that are used for debugging
3021 either your program or GCC:
3026 Produce debugging information in the operating system's native format
3027 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
3030 On most systems that use stabs format, @option{-g} enables use of extra
3031 debugging information that only GDB can use; this extra information
3032 makes debugging work better in GDB but will probably make other debuggers
3034 refuse to read the program. If you want to control for certain whether
3035 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3036 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3038 Unlike most other C compilers, GCC allows you to use @option{-g} with
3039 @option{-O}. The shortcuts taken by optimized code may occasionally
3040 produce surprising results: some variables you declared may not exist
3041 at all; flow of control may briefly move where you did not expect it;
3042 some statements may not be executed because they compute constant
3043 results or their values were already at hand; some statements may
3044 execute in different places because they were moved out of loops.
3046 Nevertheless it proves possible to debug optimized output. This makes
3047 it reasonable to use the optimizer for programs that might have bugs.
3049 The following options are useful when GCC is generated with the
3050 capability for more than one debugging format.
3054 Produce debugging information for use by GDB@. This means to use the
3055 most expressive format available (DWARF 2, stabs, or the native format
3056 if neither of those are supported), including GDB extensions if at all
3061 Produce debugging information in stabs format (if that is supported),
3062 without GDB extensions. This is the format used by DBX on most BSD
3063 systems. On MIPS, Alpha and System V Release 4 systems this option
3064 produces stabs debugging output which is not understood by DBX or SDB@.
3065 On System V Release 4 systems this option requires the GNU assembler.
3067 @item -feliminate-unused-debug-symbols
3068 @opindex feliminate-unused-debug-symbols
3069 Produce debugging information in stabs format (if that is supported),
3070 for only symbols that are actually used.
3074 Produce debugging information in stabs format (if that is supported),
3075 using GNU extensions understood only by the GNU debugger (GDB)@. The
3076 use of these extensions is likely to make other debuggers crash or
3077 refuse to read the program.
3081 Produce debugging information in COFF format (if that is supported).
3082 This is the format used by SDB on most System V systems prior to
3087 Produce debugging information in XCOFF format (if that is supported).
3088 This is the format used by the DBX debugger on IBM RS/6000 systems.
3092 Produce debugging information in XCOFF format (if that is supported),
3093 using GNU extensions understood only by the GNU debugger (GDB)@. The
3094 use of these extensions is likely to make other debuggers crash or
3095 refuse to read the program, and may cause assemblers other than the GNU
3096 assembler (GAS) to fail with an error.
3100 Produce debugging information in DWARF version 2 format (if that is
3101 supported). This is the format used by DBX on IRIX 6.
3105 Produce debugging information in VMS debug format (if that is
3106 supported). This is the format used by DEBUG on VMS systems.
3109 @itemx -ggdb@var{level}
3110 @itemx -gstabs@var{level}
3111 @itemx -gcoff@var{level}
3112 @itemx -gxcoff@var{level}
3113 @itemx -gvms@var{level}
3114 Request debugging information and also use @var{level} to specify how
3115 much information. The default level is 2.
3117 Level 1 produces minimal information, enough for making backtraces in
3118 parts of the program that you don't plan to debug. This includes
3119 descriptions of functions and external variables, but no information
3120 about local variables and no line numbers.
3122 Level 3 includes extra information, such as all the macro definitions
3123 present in the program. Some debuggers support macro expansion when
3124 you use @option{-g3}.
3126 Note that in order to avoid confusion between DWARF1 debug level 2,
3127 and DWARF2 @option{-gdwarf-2} does not accept a concatenated debug
3128 level. Instead use an additional @option{-g@var{level}} option to
3129 change the debug level for DWARF2.
3131 @item -feliminate-dwarf2-dups
3132 @opindex feliminate-dwarf2-dups
3133 Compress DWARF2 debugging information by eliminating duplicated
3134 information about each symbol. This option only makes sense when
3135 generating DWARF2 debugging information with @option{-gdwarf-2}.
3137 @cindex @command{prof}
3140 Generate extra code to write profile information suitable for the
3141 analysis program @command{prof}. You must use this option when compiling
3142 the source files you want data about, and you must also use it when
3145 @cindex @command{gprof}
3148 Generate extra code to write profile information suitable for the
3149 analysis program @command{gprof}. You must use this option when compiling
3150 the source files you want data about, and you must also use it when
3155 Makes the compiler print out each function name as it is compiled, and
3156 print some statistics about each pass when it finishes.
3159 @opindex ftime-report
3160 Makes the compiler print some statistics about the time consumed by each
3161 pass when it finishes.
3164 @opindex fmem-report
3165 Makes the compiler print some statistics about permanent memory
3166 allocation when it finishes.
3168 @item -fprofile-arcs
3169 @opindex fprofile-arcs
3170 Add code so that program flow @dfn{arcs} are instrumented. During
3171 execution the program records how many times each branch and call is
3172 executed and how many times it is taken or returns. When the compiled
3173 program exits it saves this data to a file called
3174 @file{@var{auxname}.gcda} for each source file. The data may be used for
3175 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3176 test coverage analysis (@option{-ftest-coverage}). Each object file's
3177 @var{auxname} is generated from the name of the output file, if
3178 explicitly specified and it is not the final executable, otherwise it is
3179 the basename of the source file. In both cases any suffix is removed
3180 (e.g. @file{foo.gcda} for input file @file{dir/foo.c}, or
3181 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3186 Compile the source files with @option{-fprofile-arcs} plus optimization
3187 and code generation options. For test coverage analysis, use the
3188 additional @option{-ftest-coverage} option. You do not need to profile
3189 every source file in a program.
3192 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3193 (the latter implies the former).
3196 Run the program on a representative workload to generate the arc profile
3197 information. This may be repeated any number of times. You can run
3198 concurrent instances of your program, and provided that the file system
3199 supports locking, the data files will be correctly updated. Also
3200 @code{fork} calls are detected and correctly handled (double counting
3204 For profile-directed optimizations, compile the source files again with
3205 the same optimization and code generation options plus
3206 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3207 Control Optimization}).
3210 For test coverage analysis, use @command{gcov} to produce human readable
3211 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3212 @command{gcov} documentation for further information.
3216 With @option{-fprofile-arcs}, for each function of your program GCC
3217 creates a program flow graph, then finds a spanning tree for the graph.
3218 Only arcs that are not on the spanning tree have to be instrumented: the
3219 compiler adds code to count the number of times that these arcs are
3220 executed. When an arc is the only exit or only entrance to a block, the
3221 instrumentation code can be added to the block; otherwise, a new basic
3222 block must be created to hold the instrumentation code.
3225 @item -ftest-coverage
3226 @opindex ftest-coverage
3227 Produce a notes file that the @command{gcov} code-coverage utility
3228 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3229 show program coverage. Each source file's note file is called
3230 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3231 above for a description of @var{auxname} and instructions on how to
3232 generate test coverage data. Coverage data will match the source files
3233 more closely, if you do not optimize.
3235 @item -d@var{letters}
3237 Says to make debugging dumps during compilation at times specified by
3238 @var{letters}. This is used for debugging the compiler. The file names
3239 for most of the dumps are made by appending a pass number and a word to
3240 the @var{dumpname}. @var{dumpname} is generated from the name of the
3241 output file, if explicitly specified and it is not an executable,
3242 otherwise it is the basename of the source file. In both cases any
3243 suffix is removed (e.g. @file{foo.01.rtl} or @file{foo.02.sibling}).
3244 Here are the possible letters for use in @var{letters}, and their
3250 Annotate the assembler output with miscellaneous debugging information.
3253 Dump after computing branch probabilities, to @file{@var{file}.12.bp}.
3256 Dump after block reordering, to @file{@var{file}.30.bbro}.
3259 Dump after instruction combination, to the file @file{@var{file}.20.combine}.
3262 Dump after the first if conversion, to the file @file{@var{file}.14.ce1}.
3263 Also dump after the second if conversion, to the file @file{@var{file}.21.ce2}.
3266 Dump after branch target load optimization, to to @file{@var{file}.32.btl}.
3267 Also dump after delayed branch scheduling, to @file{@var{file}.36.dbr}.
3270 Dump all macro definitions, at the end of preprocessing, in addition to
3274 Dump after the second if conversion, to @file{@var{file}.31.ce3}.
3277 Dump after control and data flow analysis, to @file{@var{file}.11.cfg}.
3278 Also dump after life analysis, to @file{@var{file}.19.life}.
3281 Dump after purging @code{ADDRESSOF} codes, to @file{@var{file}.07.addressof}.
3284 Dump after global register allocation, to @file{@var{file}.25.greg}.
3287 Dump after GCSE, to @file{@var{file}.08.gcse}.
3288 Also dump after jump bypassing and control flow optimizations, to
3289 @file{@var{file}.10.bypass}.
3292 Dump after finalization of EH handling code, to @file{@var{file}.03.eh}.
3295 Dump after sibling call optimizations, to @file{@var{file}.02.sibling}.
3298 Dump after the first jump optimization, to @file{@var{file}.04.jump}.
3301 Dump after conversion from registers to stack, to @file{@var{file}.34.stack}.
3304 Dump after local register allocation, to @file{@var{file}.24.lreg}.
3307 Dump after loop optimization passes, to @file{@var{file}.09.loop} and
3308 @file{@var{file}.16.loop2}.
3311 Dump after performing the machine dependent reorganization pass, to
3312 @file{@var{file}.35.mach}.
3315 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3318 Dump after the register move pass, to @file{@var{file}.22.regmove}.
3321 Dump after post-reload optimizations, to @file{@var{file}.26.postreload}.
3324 Dump after RTL generation, to @file{@var{file}.01.rtl}.
3327 Dump after the second scheduling pass, to @file{@var{file}.33.sched2}.
3330 Dump after CSE (including the jump optimization that sometimes follows
3331 CSE), to @file{@var{file}.06.cse}.
3334 Dump after the first scheduling pass, to @file{@var{file}.23.sched}.
3337 Dump after the second CSE pass (including the jump optimization that
3338 sometimes follows CSE), to @file{@var{file}.18.cse2}.
3341 Dump after running tracer, to @file{@var{file}.15.tracer}.
3344 Dump after null pointer elimination pass to @file{@var{file}.05.null}.
3347 Dump callgraph and unit-at-a-time optimization @file{@var{file}.00.unit}.
3350 Dump after the value profile transformations, to @file{@var{file}.13.vpt}.
3353 Dump after the second flow pass, to @file{@var{file}.27.flow2}.
3356 Dump after the peephole pass, to @file{@var{file}.28.peephole2}.
3359 Dump after constructing the web, to @file{@var{file}.17.web}.
3362 Produce all the dumps listed above.
3365 Produce a core dump whenever an error occurs.
3368 Print statistics on memory usage, at the end of the run, to
3372 Annotate the assembler output with a comment indicating which
3373 pattern and alternative was used. The length of each instruction is
3377 Dump the RTL in the assembler output as a comment before each instruction.
3378 Also turns on @option{-dp} annotation.
3381 For each of the other indicated dump files (except for
3382 @file{@var{file}.01.rtl}), dump a representation of the control flow graph
3383 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3386 Just generate RTL for a function instead of compiling it. Usually used
3390 Dump debugging information during parsing, to standard error.
3393 @item -fdump-unnumbered
3394 @opindex fdump-unnumbered
3395 When doing debugging dumps (see @option{-d} option above), suppress instruction
3396 numbers and line number note output. This makes it more feasible to
3397 use diff on debugging dumps for compiler invocations with different
3398 options, in particular with and without @option{-g}.
3400 @item -fdump-translation-unit @r{(C and C++ only)}
3401 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3402 @opindex fdump-translation-unit
3403 Dump a representation of the tree structure for the entire translation
3404 unit to a file. The file name is made by appending @file{.tu} to the
3405 source file name. If the @samp{-@var{options}} form is used, @var{options}
3406 controls the details of the dump as described for the
3407 @option{-fdump-tree} options.
3409 @item -fdump-class-hierarchy @r{(C++ only)}
3410 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3411 @opindex fdump-class-hierarchy
3412 Dump a representation of each class's hierarchy and virtual function
3413 table layout to a file. The file name is made by appending @file{.class}
3414 to the source file name. If the @samp{-@var{options}} form is used,
3415 @var{options} controls the details of the dump as described for the
3416 @option{-fdump-tree} options.
3418 @item -fdump-tree-@var{switch} @r{(C++ only)}
3419 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C++ only)}
3421 Control the dumping at various stages of processing the intermediate
3422 language tree to a file. The file name is generated by appending a switch
3423 specific suffix to the source file name. If the @samp{-@var{options}}
3424 form is used, @var{options} is a list of @samp{-} separated options that
3425 control the details of the dump. Not all options are applicable to all
3426 dumps, those which are not meaningful will be ignored. The following
3427 options are available
3431 Print the address of each node. Usually this is not meaningful as it
3432 changes according to the environment and source file. Its primary use
3433 is for tying up a dump file with a debug environment.
3435 Inhibit dumping of members of a scope or body of a function merely
3436 because that scope has been reached. Only dump such items when they
3437 are directly reachable by some other path.
3439 Turn on all options.
3442 The following tree dumps are possible:
3445 Dump before any tree based optimization, to @file{@var{file}.original}.
3447 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3449 Dump after function inlining, to @file{@var{file}.inlined}.
3452 @item -frandom-seed=@var{string}
3453 @opindex frandom-string
3454 This option provides a seed that GCC uses when it would otherwise use
3455 random numbers. It is used to generate certain symbol names
3456 that have to be different in every compiled file. It is also used to
3457 place unique stamps in coverage data files and the object files that
3458 produce them. You can use the @option{-frandom-seed} option to produce
3459 reproducibly identical object files.
3461 The @var{string} should be different for every file you compile.
3463 @item -fsched-verbose=@var{n}
3464 @opindex fsched-verbose
3465 On targets that use instruction scheduling, this option controls the
3466 amount of debugging output the scheduler prints. This information is
3467 written to standard error, unless @option{-dS} or @option{-dR} is
3468 specified, in which case it is output to the usual dump
3469 listing file, @file{.sched} or @file{.sched2} respectively. However
3470 for @var{n} greater than nine, the output is always printed to standard
3473 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3474 same information as @option{-dRS}. For @var{n} greater than one, it
3475 also output basic block probabilities, detailed ready list information
3476 and unit/insn info. For @var{n} greater than two, it includes RTL
3477 at abort point, control-flow and regions info. And for @var{n} over
3478 four, @option{-fsched-verbose} also includes dependence info.
3482 Store the usual ``temporary'' intermediate files permanently; place them
3483 in the current directory and name them based on the source file. Thus,
3484 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3485 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3486 preprocessed @file{foo.i} output file even though the compiler now
3487 normally uses an integrated preprocessor.
3491 Report the CPU time taken by each subprocess in the compilation
3492 sequence. For C source files, this is the compiler proper and assembler
3493 (plus the linker if linking is done). The output looks like this:
3500 The first number on each line is the ``user time,'' that is time spent
3501 executing the program itself. The second number is ``system time,''
3502 time spent executing operating system routines on behalf of the program.
3503 Both numbers are in seconds.
3505 @item -print-file-name=@var{library}
3506 @opindex print-file-name
3507 Print the full absolute name of the library file @var{library} that
3508 would be used when linking---and don't do anything else. With this
3509 option, GCC does not compile or link anything; it just prints the
3512 @item -print-multi-directory
3513 @opindex print-multi-directory
3514 Print the directory name corresponding to the multilib selected by any
3515 other switches present in the command line. This directory is supposed
3516 to exist in @env{GCC_EXEC_PREFIX}.
3518 @item -print-multi-lib
3519 @opindex print-multi-lib
3520 Print the mapping from multilib directory names to compiler switches
3521 that enable them. The directory name is separated from the switches by
3522 @samp{;}, and each switch starts with an @samp{@@} instead of the
3523 @samp{-}, without spaces between multiple switches. This is supposed to
3524 ease shell-processing.
3526 @item -print-prog-name=@var{program}
3527 @opindex print-prog-name
3528 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3530 @item -print-libgcc-file-name
3531 @opindex print-libgcc-file-name
3532 Same as @option{-print-file-name=libgcc.a}.
3534 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3535 but you do want to link with @file{libgcc.a}. You can do
3538 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3541 @item -print-search-dirs
3542 @opindex print-search-dirs
3543 Print the name of the configured installation directory and a list of
3544 program and library directories gcc will search---and don't do anything else.
3546 This is useful when gcc prints the error message
3547 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3548 To resolve this you either need to put @file{cpp0} and the other compiler
3549 components where gcc expects to find them, or you can set the environment
3550 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3551 Don't forget the trailing '/'.
3552 @xref{Environment Variables}.
3555 @opindex dumpmachine
3556 Print the compiler's target machine (for example,
3557 @samp{i686-pc-linux-gnu})---and don't do anything else.
3560 @opindex dumpversion
3561 Print the compiler version (for example, @samp{3.0})---and don't do
3566 Print the compiler's built-in specs---and don't do anything else. (This
3567 is used when GCC itself is being built.) @xref{Spec Files}.
3569 @item -feliminate-unused-debug-types
3570 @opindex feliminate-unused-debug-types
3571 Normally, when producing DWARF2 output, GCC will emit debugging
3572 information for all types declared in a compilation
3573 unit, regardless of whether or not they are actually used
3574 in that compilation unit. Sometimes this is useful, such as
3575 if, in the debugger, you want to cast a value to a type that is
3576 not actually used in your program (but is declared). More often,
3577 however, this results in a significant amount of wasted space.
3578 With this option, GCC will avoid producing debug symbol output
3579 for types that are nowhere used in the source file being compiled.
3582 @node Optimize Options
3583 @section Options That Control Optimization
3584 @cindex optimize options
3585 @cindex options, optimization
3587 These options control various sorts of optimizations.
3589 Without any optimization option, the compiler's goal is to reduce the
3590 cost of compilation and to make debugging produce the expected
3591 results. Statements are independent: if you stop the program with a
3592 breakpoint between statements, you can then assign a new value to any
3593 variable or change the program counter to any other statement in the
3594 function and get exactly the results you would expect from the source
3597 Turning on optimization flags makes the compiler attempt to improve
3598 the performance and/or code size at the expense of compilation time
3599 and possibly the ability to debug the program.
3601 The compiler performs optimisation based on the knowledge it has of
3602 the program. Using the @option{-funit-at-a-time} flag will allow the
3603 compiler to consider information gained from later functions in the
3604 file when compiling a function. Compiling multiple files at once to a
3605 single output file (and using @option{-funit-at-a-time}) will allow
3606 the compiler to use information gained from all of the files when
3607 compiling each of them.
3609 Not all optimizations are controlled directly by a flag. Only
3610 optimizations that have a flag are listed.
3617 Optimize. Optimizing compilation takes somewhat more time, and a lot
3618 more memory for a large function.
3620 With @option{-O}, the compiler tries to reduce code size and execution
3621 time, without performing any optimizations that take a great deal of
3624 @option{-O} turns on the following optimization flags:
3625 @gccoptlist{-fdefer-pop @gol
3626 -fmerge-constants @gol
3628 -floop-optimize @gol
3629 -fif-conversion @gol
3630 -fif-conversion2 @gol
3631 -fdelayed-branch @gol
3632 -fguess-branch-probability @gol
3635 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
3636 where doing so does not interfere with debugging.
3640 Optimize even more. GCC performs nearly all supported optimizations
3641 that do not involve a space-speed tradeoff. The compiler does not
3642 perform loop unrolling or function inlining when you specify @option{-O2}.
3643 As compared to @option{-O}, this option increases both compilation time
3644 and the performance of the generated code.
3646 @option{-O2} turns on all optimization flags specified by @option{-O}. It
3647 also turns on the following optimization flags:
3648 @gccoptlist{-fforce-mem @gol
3649 -foptimize-sibling-calls @gol
3650 -fstrength-reduce @gol
3651 -fcse-follow-jumps -fcse-skip-blocks @gol
3652 -frerun-cse-after-loop -frerun-loop-opt @gol
3653 -fgcse -fgcse-lm -fgcse-sm -fgcse-las @gol
3654 -fdelete-null-pointer-checks @gol
3655 -fexpensive-optimizations @gol
3657 -fschedule-insns -fschedule-insns2 @gol
3658 -fsched-interblock -fsched-spec @gol
3661 -freorder-blocks -freorder-functions @gol
3662 -fstrict-aliasing @gol
3663 -funit-at-a-time @gol
3664 -falign-functions -falign-jumps @gol
3665 -falign-loops -falign-labels @gol
3668 Please note the warning under @option{-fgcse} about
3669 invoking @option{-O2} on programs that use computed gotos.
3673 Optimize yet more. @option{-O3} turns on all optimizations specified by
3674 @option{-O2} and also turns on the @option{-finline-functions},
3675 @option{-fweb} and @option{-frename-registers} options.
3679 Do not optimize. This is the default.
3683 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
3684 do not typically increase code size. It also performs further
3685 optimizations designed to reduce code size.
3687 @option{-Os} disables the following optimization flags:
3688 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
3689 -falign-labels -freorder-blocks -fprefetch-loop-arrays}
3691 If you use multiple @option{-O} options, with or without level numbers,
3692 the last such option is the one that is effective.
3695 Options of the form @option{-f@var{flag}} specify machine-independent
3696 flags. Most flags have both positive and negative forms; the negative
3697 form of @option{-ffoo} would be @option{-fno-foo}. In the table
3698 below, only one of the forms is listed---the one you typically will
3699 use. You can figure out the other form by either removing @samp{no-}
3702 The following options control specific optimizations. They are either
3703 activated by @option{-O} options or are related to ones that are. You
3704 can use the following flags in the rare cases when ``fine-tuning'' of
3705 optimizations to be performed is desired.
3708 @item -fno-default-inline
3709 @opindex fno-default-inline
3710 Do not make member functions inline by default merely because they are
3711 defined inside the class scope (C++ only). Otherwise, when you specify
3712 @w{@option{-O}}, member functions defined inside class scope are compiled
3713 inline by default; i.e., you don't need to add @samp{inline} in front of
3714 the member function name.
3716 @item -fno-defer-pop
3717 @opindex fno-defer-pop
3718 Always pop the arguments to each function call as soon as that function
3719 returns. For machines which must pop arguments after a function call,
3720 the compiler normally lets arguments accumulate on the stack for several
3721 function calls and pops them all at once.
3723 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3727 Force memory operands to be copied into registers before doing
3728 arithmetic on them. This produces better code by making all memory
3729 references potential common subexpressions. When they are not common
3730 subexpressions, instruction combination should eliminate the separate
3733 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3736 @opindex fforce-addr
3737 Force memory address constants to be copied into registers before
3738 doing arithmetic on them. This may produce better code just as
3739 @option{-fforce-mem} may.
3741 @item -fomit-frame-pointer
3742 @opindex fomit-frame-pointer
3743 Don't keep the frame pointer in a register for functions that
3744 don't need one. This avoids the instructions to save, set up and
3745 restore frame pointers; it also makes an extra register available
3746 in many functions. @strong{It also makes debugging impossible on
3749 On some machines, such as the VAX, this flag has no effect, because
3750 the standard calling sequence automatically handles the frame pointer
3751 and nothing is saved by pretending it doesn't exist. The
3752 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3753 whether a target machine supports this flag. @xref{Registers,,Register
3754 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
3756 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3758 @item -foptimize-sibling-calls
3759 @opindex foptimize-sibling-calls
3760 Optimize sibling and tail recursive calls.
3762 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3766 Don't pay attention to the @code{inline} keyword. Normally this option
3767 is used to keep the compiler from expanding any functions inline.
3768 Note that if you are not optimizing, no functions can be expanded inline.
3770 @item -finline-functions
3771 @opindex finline-functions
3772 Integrate all simple functions into their callers. The compiler
3773 heuristically decides which functions are simple enough to be worth
3774 integrating in this way.
3776 If all calls to a given function are integrated, and the function is
3777 declared @code{static}, then the function is normally not output as
3778 assembler code in its own right.
3780 Enabled at level @option{-O3}.
3782 @item -finline-limit=@var{n}
3783 @opindex finline-limit
3784 By default, gcc limits the size of functions that can be inlined. This flag
3785 allows the control of this limit for functions that are explicitly marked as
3786 inline (i.e., marked with the inline keyword or defined within the class
3787 definition in c++). @var{n} is the size of functions that can be inlined in
3788 number of pseudo instructions (not counting parameter handling). The default
3789 value of @var{n} is 600.
3790 Increasing this value can result in more inlined code at
3791 the cost of compilation time and memory consumption. Decreasing usually makes
3792 the compilation faster and less code will be inlined (which presumably
3793 means slower programs). This option is particularly useful for programs that
3794 use inlining heavily such as those based on recursive templates with C++.
3796 Inlining is actually controlled by a number of parameters, which may be
3797 specified individually by using @option{--param @var{name}=@var{value}}.
3798 The @option{-finline-limit=@var{n}} option sets some of these parameters
3802 @item max-inline-insns-single
3803 is set to @var{n}/2.
3804 @item max-inline-insns-auto
3805 is set to @var{n}/2.
3806 @item min-inline-insns
3807 is set to 130 or @var{n}/4, whichever is smaller.
3808 @item max-inline-insns-rtl
3812 See below for a documentation of the individual
3813 parameters controlling inlining.
3815 @emph{Note:} pseudo instruction represents, in this particular context, an
3816 abstract measurement of function's size. In no way, it represents a count
3817 of assembly instructions and as such its exact meaning might change from one
3818 release to an another.
3820 @item -fkeep-inline-functions
3821 @opindex fkeep-inline-functions
3822 Even if all calls to a given function are integrated, and the function
3823 is declared @code{static}, nevertheless output a separate run-time
3824 callable version of the function. This switch does not affect
3825 @code{extern inline} functions.
3827 @item -fkeep-static-consts
3828 @opindex fkeep-static-consts
3829 Emit variables declared @code{static const} when optimization isn't turned
3830 on, even if the variables aren't referenced.
3832 GCC enables this option by default. If you want to force the compiler to
3833 check if the variable was referenced, regardless of whether or not
3834 optimization is turned on, use the @option{-fno-keep-static-consts} option.
3836 @item -fmerge-constants
3837 Attempt to merge identical constants (string constants and floating point
3838 constants) across compilation units.
3840 This option is the default for optimized compilation if the assembler and
3841 linker support it. Use @option{-fno-merge-constants} to inhibit this
3844 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3846 @item -fmerge-all-constants
3847 Attempt to merge identical constants and identical variables.
3849 This option implies @option{-fmerge-constants}. In addition to
3850 @option{-fmerge-constants} this considers e.g. even constant initialized
3851 arrays or initialized constant variables with integral or floating point
3852 types. Languages like C or C++ require each non-automatic variable to
3853 have distinct location, so using this option will result in non-conforming
3858 Use a graph coloring register allocator. Currently this option is meant
3859 for testing, so we are interested to hear about miscompilations with
3862 @item -fno-branch-count-reg
3863 @opindex fno-branch-count-reg
3864 Do not use ``decrement and branch'' instructions on a count register,
3865 but instead generate a sequence of instructions that decrement a
3866 register, compare it against zero, then branch based upon the result.
3867 This option is only meaningful on architectures that support such
3868 instructions, which include x86, PowerPC, IA-64 and S/390.
3870 The default is @option{-fbranch-count-reg}, enabled when
3871 @option{-fstrength-reduce} is enabled.
3873 @item -fno-function-cse
3874 @opindex fno-function-cse
3875 Do not put function addresses in registers; make each instruction that
3876 calls a constant function contain the function's address explicitly.
3878 This option results in less efficient code, but some strange hacks
3879 that alter the assembler output may be confused by the optimizations
3880 performed when this option is not used.
3882 The default is @option{-ffunction-cse}
3884 @item -fno-zero-initialized-in-bss
3885 @opindex fno-zero-initialized-in-bss
3886 If the target supports a BSS section, GCC by default puts variables that
3887 are initialized to zero into BSS@. This can save space in the resulting
3890 This option turns off this behavior because some programs explicitly
3891 rely on variables going to the data section. E.g., so that the
3892 resulting executable can find the beginning of that section and/or make
3893 assumptions based on that.
3895 The default is @option{-fzero-initialized-in-bss}.
3897 @item -fstrength-reduce
3898 @opindex fstrength-reduce
3899 Perform the optimizations of loop strength reduction and
3900 elimination of iteration variables.
3902 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3904 @item -fthread-jumps
3905 @opindex fthread-jumps
3906 Perform optimizations where we check to see if a jump branches to a
3907 location where another comparison subsumed by the first is found. If
3908 so, the first branch is redirected to either the destination of the
3909 second branch or a point immediately following it, depending on whether
3910 the condition is known to be true or false.
3912 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3914 @item -fcse-follow-jumps
3915 @opindex fcse-follow-jumps
3916 In common subexpression elimination, scan through jump instructions
3917 when the target of the jump is not reached by any other path. For
3918 example, when CSE encounters an @code{if} statement with an
3919 @code{else} clause, CSE will follow the jump when the condition
3922 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3924 @item -fcse-skip-blocks
3925 @opindex fcse-skip-blocks
3926 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
3927 follow jumps which conditionally skip over blocks. When CSE
3928 encounters a simple @code{if} statement with no else clause,
3929 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
3930 body of the @code{if}.
3932 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3934 @item -frerun-cse-after-loop
3935 @opindex frerun-cse-after-loop
3936 Re-run common subexpression elimination after loop optimizations has been
3939 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3941 @item -frerun-loop-opt
3942 @opindex frerun-loop-opt
3943 Run the loop optimizer twice.
3945 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3949 Perform a global common subexpression elimination pass.
3950 This pass also performs global constant and copy propagation.
3952 @emph{Note:} When compiling a program using computed gotos, a GCC
3953 extension, you may get better runtime performance if you disable
3954 the global common subexpression elimination pass by adding
3955 @option{-fno-gcse} to the command line.
3957 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3961 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
3962 attempt to move loads which are only killed by stores into themselves. This
3963 allows a loop containing a load/store sequence to be changed to a load outside
3964 the loop, and a copy/store within the loop.
3966 Enabled by default when gcse is enabled.
3970 When @option{-fgcse-sm} is enabled, a store motion pass is run after
3971 global common subexpression elimination. This pass will attempt to move
3972 stores out of loops. When used in conjunction with @option{-fgcse-lm},
3973 loops containing a load/store sequence can be changed to a load before
3974 the loop and a store after the loop.
3976 Enabled by default when gcse is enabled.
3980 When @option{-fgcse-las} is enabled, the global common subexpression
3981 elimination pass eliminates redundant loads that come after stores to the
3982 same memory location (both partial and full redundancies).
3984 Enabled by default when gcse is enabled.
3986 @item -floop-optimize
3987 @opindex floop-optimize
3988 Perform loop optimizations: move constant expressions out of loops, simplify
3989 exit test conditions and optionally do strength-reduction and loop unrolling as
3992 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3994 @item -fcrossjumping
3995 @opindex crossjumping
3996 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
3997 resulting code may or may not perform better than without cross-jumping.
3999 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4001 @item -fif-conversion
4002 @opindex if-conversion
4003 Attempt to transform conditional jumps into branch-less equivalents. This
4004 include use of conditional moves, min, max, set flags and abs instructions, and
4005 some tricks doable by standard arithmetics. The use of conditional execution
4006 on chips where it is available is controlled by @code{if-conversion2}.
4008 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4010 @item -fif-conversion2
4011 @opindex if-conversion2
4012 Use conditional execution (where available) to transform conditional jumps into
4013 branch-less equivalents.
4015 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4017 @item -fdelete-null-pointer-checks
4018 @opindex fdelete-null-pointer-checks
4019 Use global dataflow analysis to identify and eliminate useless checks
4020 for null pointers. The compiler assumes that dereferencing a null
4021 pointer would have halted the program. If a pointer is checked after
4022 it has already been dereferenced, it cannot be null.
4024 In some environments, this assumption is not true, and programs can
4025 safely dereference null pointers. Use
4026 @option{-fno-delete-null-pointer-checks} to disable this optimization
4027 for programs which depend on that behavior.
4029 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4031 @item -fexpensive-optimizations
4032 @opindex fexpensive-optimizations
4033 Perform a number of minor optimizations that are relatively expensive.
4035 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4037 @item -foptimize-register-move
4039 @opindex foptimize-register-move
4041 Attempt to reassign register numbers in move instructions and as
4042 operands of other simple instructions in order to maximize the amount of
4043 register tying. This is especially helpful on machines with two-operand
4046 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4049 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4051 @item -fdelayed-branch
4052 @opindex fdelayed-branch
4053 If supported for the target machine, attempt to reorder instructions
4054 to exploit instruction slots available after delayed branch
4057 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4059 @item -fschedule-insns
4060 @opindex fschedule-insns
4061 If supported for the target machine, attempt to reorder instructions to
4062 eliminate execution stalls due to required data being unavailable. This
4063 helps machines that have slow floating point or memory load instructions
4064 by allowing other instructions to be issued until the result of the load
4065 or floating point instruction is required.
4067 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4069 @item -fschedule-insns2
4070 @opindex fschedule-insns2
4071 Similar to @option{-fschedule-insns}, but requests an additional pass of
4072 instruction scheduling after register allocation has been done. This is
4073 especially useful on machines with a relatively small number of
4074 registers and where memory load instructions take more than one cycle.
4076 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4078 @item -fno-sched-interblock
4079 @opindex fno-sched-interblock
4080 Don't schedule instructions across basic blocks. This is normally
4081 enabled by default when scheduling before register allocation, i.e.@:
4082 with @option{-fschedule-insns} or at @option{-O2} or higher.
4084 @item -fno-sched-spec
4085 @opindex fno-sched-spec
4086 Don't allow speculative motion of non-load instructions. This is normally
4087 enabled by default when scheduling before register allocation, i.e.@:
4088 with @option{-fschedule-insns} or at @option{-O2} or higher.
4090 @item -fsched-spec-load
4091 @opindex fsched-spec-load
4092 Allow speculative motion of some load instructions. This only makes
4093 sense when scheduling before register allocation, i.e.@: with
4094 @option{-fschedule-insns} or at @option{-O2} or higher.
4096 @item -fsched-spec-load-dangerous
4097 @opindex fsched-spec-load-dangerous
4098 Allow speculative motion of more load instructions. This only makes
4099 sense when scheduling before register allocation, i.e.@: with
4100 @option{-fschedule-insns} or at @option{-O2} or higher.
4102 @item -fsched-stalled-insns=@var{n}
4103 @opindex fsched-stalled-insns
4104 Define how many insns (if any) can be moved prematurely from the queue
4105 of stalled insns into the ready list, during the second scheduling pass.
4107 @item -fsched-stalled-insns-dep=@var{n}
4108 @opindex fsched-stalled-insns-dep
4109 Define how many insn groups (cycles) will be examined for a dependency
4110 on a stalled insn that is candidate for premature removal from the queue
4111 of stalled insns. Has an effect only during the second scheduling pass,
4112 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4114 @item -fsched2-use-superblocks
4115 @opindex fsched2-use-superblocks
4116 When scheduling after register allocation, do use superblock scheduling
4117 algorithm. Superblock scheduling allows motion across basic block boundaries
4118 resulting on faster schedules. This option is experimental, as not all machine
4119 descriptions used by GCC model the CPU closely enough to avoid unreliable
4120 results from the algorithm.
4122 This only makes sense when scheduling after register allocation, i.e.@: with
4123 @option{-fschedule-insns2} or at @option{-O2} or higher.
4125 @item -fsched2-use-traces
4126 @opindex fsched2-use-traces
4127 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4128 allocation and additionally perform code duplication in order to increase the
4129 size of superblocks using tracer pass. See @option{-ftracer} for details on
4132 This mode should produce faster but significantly longer programs. Also
4133 without @code{-fbranch-probabilities} the traces constructed may not match the
4134 reality and hurt the performance. This only makes
4135 sense when scheduling after register allocation, i.e.@: with
4136 @option{-fschedule-insns2} or at @option{-O2} or higher.
4138 @item -fcaller-saves
4139 @opindex fcaller-saves
4140 Enable values to be allocated in registers that will be clobbered by
4141 function calls, by emitting extra instructions to save and restore the
4142 registers around such calls. Such allocation is done only when it
4143 seems to result in better code than would otherwise be produced.
4145 This option is always enabled by default on certain machines, usually
4146 those which have no call-preserved registers to use instead.
4148 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4150 @item -fmove-all-movables
4151 @opindex fmove-all-movables
4152 Forces all invariant computations in loops to be moved
4155 @item -freduce-all-givs
4156 @opindex freduce-all-givs
4157 Forces all general-induction variables in loops to be
4160 @emph{Note:} When compiling programs written in Fortran,
4161 @option{-fmove-all-movables} and @option{-freduce-all-givs} are enabled
4162 by default when you use the optimizer.
4164 These options may generate better or worse code; results are highly
4165 dependent on the structure of loops within the source code.
4167 These two options are intended to be removed someday, once
4168 they have helped determine the efficacy of various
4169 approaches to improving loop optimizations.
4171 Please let us (@w{@email{gcc@@gcc.gnu.org}} and @w{@email{fortran@@gnu.org}})
4172 know how use of these options affects
4173 the performance of your production code.
4174 We're very interested in code that runs @emph{slower}
4175 when these options are @emph{enabled}.
4178 @itemx -fno-peephole2
4179 @opindex fno-peephole
4180 @opindex fno-peephole2
4181 Disable any machine-specific peephole optimizations. The difference
4182 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4183 are implemented in the compiler; some targets use one, some use the
4184 other, a few use both.
4186 @option{-fpeephole} is enabled by default.
4187 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4189 @item -fno-guess-branch-probability
4190 @opindex fno-guess-branch-probability
4191 Do not guess branch probabilities using a randomized model.
4193 Sometimes gcc will opt to use a randomized model to guess branch
4194 probabilities, when none are available from either profiling feedback
4195 (@option{-fprofile-arcs}) or @samp{__builtin_expect}. This means that
4196 different runs of the compiler on the same program may produce different
4199 In a hard real-time system, people don't want different runs of the
4200 compiler to produce code that has different behavior; minimizing
4201 non-determinism is of paramount import. This switch allows users to
4202 reduce non-determinism, possibly at the expense of inferior
4205 The default is @option{-fguess-branch-probability} at levels
4206 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4208 @item -freorder-blocks
4209 @opindex freorder-blocks
4210 Reorder basic blocks in the compiled function in order to reduce number of
4211 taken branches and improve code locality.
4213 Enabled at levels @option{-O2}, @option{-O3}.
4215 @item -freorder-functions
4216 @opindex freorder-functions
4217 Reorder basic blocks in the compiled function in order to reduce number of
4218 taken branches and improve code locality. This is implemented by using special
4219 subsections @code{text.hot} for most frequently executed functions and
4220 @code{text.unlikely} for unlikely executed functions. Reordering is done by
4221 the linker so object file format must support named sections and linker must
4222 place them in a reasonable way.
4224 Also profile feedback must be available in to make this option effective. See
4225 @option{-fprofile-arcs} for details.
4227 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4229 @item -fstrict-aliasing
4230 @opindex fstrict-aliasing
4231 Allows the compiler to assume the strictest aliasing rules applicable to
4232 the language being compiled. For C (and C++), this activates
4233 optimizations based on the type of expressions. In particular, an
4234 object of one type is assumed never to reside at the same address as an
4235 object of a different type, unless the types are almost the same. For
4236 example, an @code{unsigned int} can alias an @code{int}, but not a
4237 @code{void*} or a @code{double}. A character type may alias any other
4240 Pay special attention to code like this:
4253 The practice of reading from a different union member than the one most
4254 recently written to (called ``type-punning'') is common. Even with
4255 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4256 is accessed through the union type. So, the code above will work as
4257 expected. However, this code might not:
4268 Every language that wishes to perform language-specific alias analysis
4269 should define a function that computes, given an @code{tree}
4270 node, an alias set for the node. Nodes in different alias sets are not
4271 allowed to alias. For an example, see the C front-end function
4272 @code{c_get_alias_set}.
4274 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4276 @item -falign-functions
4277 @itemx -falign-functions=@var{n}
4278 @opindex falign-functions
4279 Align the start of functions to the next power-of-two greater than
4280 @var{n}, skipping up to @var{n} bytes. For instance,
4281 @option{-falign-functions=32} aligns functions to the next 32-byte
4282 boundary, but @option{-falign-functions=24} would align to the next
4283 32-byte boundary only if this can be done by skipping 23 bytes or less.
4285 @option{-fno-align-functions} and @option{-falign-functions=1} are
4286 equivalent and mean that functions will not be aligned.
4288 Some assemblers only support this flag when @var{n} is a power of two;
4289 in that case, it is rounded up.
4291 If @var{n} is not specified or is zero, use a machine-dependent default.
4293 Enabled at levels @option{-O2}, @option{-O3}.
4295 @item -falign-labels
4296 @itemx -falign-labels=@var{n}
4297 @opindex falign-labels
4298 Align all branch targets to a power-of-two boundary, skipping up to
4299 @var{n} bytes like @option{-falign-functions}. This option can easily
4300 make code slower, because it must insert dummy operations for when the
4301 branch target is reached in the usual flow of the code.
4303 @option{-fno-align-labels} and @option{-falign-labels=1} are
4304 equivalent and mean that labels will not be aligned.
4306 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
4307 are greater than this value, then their values are used instead.
4309 If @var{n} is not specified or is zero, use a machine-dependent default
4310 which is very likely to be @samp{1}, meaning no alignment.
4312 Enabled at levels @option{-O2}, @option{-O3}.
4315 @itemx -falign-loops=@var{n}
4316 @opindex falign-loops
4317 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
4318 like @option{-falign-functions}. The hope is that the loop will be
4319 executed many times, which will make up for any execution of the dummy
4322 @option{-fno-align-loops} and @option{-falign-loops=1} are
4323 equivalent and mean that loops will not be aligned.
4325 If @var{n} is not specified or is zero, use a machine-dependent default.
4327 Enabled at levels @option{-O2}, @option{-O3}.
4330 @itemx -falign-jumps=@var{n}
4331 @opindex falign-jumps
4332 Align branch targets to a power-of-two boundary, for branch targets
4333 where the targets can only be reached by jumping, skipping up to @var{n}
4334 bytes like @option{-falign-functions}. In this case, no dummy operations
4337 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
4338 equivalent and mean that loops will not be aligned.
4340 If @var{n} is not specified or is zero, use a machine-dependent default.
4342 Enabled at levels @option{-O2}, @option{-O3}.
4344 @item -frename-registers
4345 @opindex frename-registers
4346 Attempt to avoid false dependencies in scheduled code by making use
4347 of registers left over after register allocation. This optimization
4348 will most benefit processors with lots of registers. It can, however,
4349 make debugging impossible, since variables will no longer stay in
4350 a ``home register''.
4354 Constructs webs as commonly used for register allocation purposes and assign
4355 each web individual pseudo register. This allows our register allocation pass
4356 to operate on pseudos directly, but also strengthens several other optimization
4357 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
4358 however, make debugging impossible, since variables will no longer stay in a
4361 Enabled at levels @option{-O3}.
4363 @item -fno-cprop-registers
4364 @opindex fno-cprop-registers
4365 After register allocation and post-register allocation instruction splitting,
4366 we perform a copy-propagation pass to try to reduce scheduling dependencies
4367 and occasionally eliminate the copy.
4369 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4371 @item -fprofile-generate
4372 @opindex fprofile-generate
4373 Enable options usually used for instrumenting application to produce profile usefull
4374 for later recompilation profile feedback based optimization.
4376 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}
4379 @opindex fprofile-use
4380 Enable profile feedback directed optimizations, and optimizations
4381 generally profitable only with profile feedback available.
4383 The following options are enabled: @code{-fbranch-probabilities},
4384 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
4388 The following options control compiler behavior regarding floating
4389 point arithmetic. These options trade off between speed and
4390 correctness. All must be specifically enabled.
4394 @opindex ffloat-store
4395 Do not store floating point variables in registers, and inhibit other
4396 options that might change whether a floating point value is taken from a
4399 @cindex floating point precision
4400 This option prevents undesirable excess precision on machines such as
4401 the 68000 where the floating registers (of the 68881) keep more
4402 precision than a @code{double} is supposed to have. Similarly for the
4403 x86 architecture. For most programs, the excess precision does only
4404 good, but a few programs rely on the precise definition of IEEE floating
4405 point. Use @option{-ffloat-store} for such programs, after modifying
4406 them to store all pertinent intermediate computations into variables.
4410 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
4411 @option{-fno-trapping-math}, @option{-ffinite-math-only},
4412 @option{-fno-rounding-math} and @option{-fno-signaling-nans}.
4414 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
4416 This option should never be turned on by any @option{-O} option since
4417 it can result in incorrect output for programs which depend on
4418 an exact implementation of IEEE or ISO rules/specifications for
4421 @item -fno-math-errno
4422 @opindex fno-math-errno
4423 Do not set ERRNO after calling math functions that are executed
4424 with a single instruction, e.g., sqrt. A program that relies on
4425 IEEE exceptions for math error handling may want to use this flag
4426 for speed while maintaining IEEE arithmetic compatibility.
4428 This option should never be turned on by any @option{-O} option since
4429 it can result in incorrect output for programs which depend on
4430 an exact implementation of IEEE or ISO rules/specifications for
4433 The default is @option{-fmath-errno}.
4435 @item -funsafe-math-optimizations
4436 @opindex funsafe-math-optimizations
4437 Allow optimizations for floating-point arithmetic that (a) assume
4438 that arguments and results are valid and (b) may violate IEEE or
4439 ANSI standards. When used at link-time, it may include libraries
4440 or startup files that change the default FPU control word or other
4441 similar optimizations.
4443 This option should never be turned on by any @option{-O} option since
4444 it can result in incorrect output for programs which depend on
4445 an exact implementation of IEEE or ISO rules/specifications for
4448 The default is @option{-fno-unsafe-math-optimizations}.
4450 @item -ffinite-math-only
4451 @opindex ffinite-math-only
4452 Allow optimizations for floating-point arithmetic that assume
4453 that arguments and results are not NaNs or +-Infs.
4455 This option should never be turned on by any @option{-O} option since
4456 it can result in incorrect output for programs which depend on
4457 an exact implementation of IEEE or ISO rules/specifications.
4459 The default is @option{-fno-finite-math-only}.
4461 @item -fno-trapping-math
4462 @opindex fno-trapping-math
4463 Compile code assuming that floating-point operations cannot generate
4464 user-visible traps. These traps include division by zero, overflow,
4465 underflow, inexact result and invalid operation. This option implies
4466 @option{-fno-signaling-nans}. Setting this option may allow faster
4467 code if one relies on ``non-stop'' IEEE arithmetic, for example.
4469 This option should never be turned on by any @option{-O} option since
4470 it can result in incorrect output for programs which depend on
4471 an exact implementation of IEEE or ISO rules/specifications for
4474 The default is @option{-ftrapping-math}.
4476 @item -frounding-math
4477 @opindex frounding-math
4478 Disable transformations and optimizations that assume default floating
4479 point rounding behavior. This is round-to-zero for all floating point
4480 to integer conversions, and round-to-nearest for all other arithmetic
4481 truncations. This option should be specified for programs that change
4482 the FP rounding mode dynamically, or that may be executed with a
4483 non-default rounding mode. This option disables constant folding of
4484 floating point expressions at compile-time (which may be affected by
4485 rounding mode) and arithmetic transformations that are unsafe in the
4486 presence of sign-dependent rounding modes.
4488 The default is @option{-fno-rounding-math}.
4490 This option is experimental and does not currently guarantee to
4491 disable all GCC optimizations that are affected by rounding mode.
4492 Future versions of gcc may provide finer control of this setting
4493 using C99's @code{FENV_ACCESS} pragma. This command line option
4494 will be used to specify the default state for @code{FENV_ACCESS}.
4496 @item -fsignaling-nans
4497 @opindex fsignaling-nans
4498 Compile code assuming that IEEE signaling NaNs may generate user-visible
4499 traps during floating-point operations. Setting this option disables
4500 optimizations that may change the number of exceptions visible with
4501 signaling NaNs. This option implies @option{-ftrapping-math}.
4503 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
4506 The default is @option{-fno-signaling-nans}.
4508 This option is experimental and does not currently guarantee to
4509 disable all GCC optimizations that affect signaling NaN behavior.
4511 @item -fsingle-precision-constant
4512 @opindex fsingle-precision-constant
4513 Treat floating point constant as single precision constant instead of
4514 implicitly converting it to double precision constant.
4519 The following options control optimizations that may improve
4520 performance, but are not enabled by any @option{-O} options. This
4521 section includes experimental options that may produce broken code.
4524 @item -fbranch-probabilities
4525 @opindex fbranch-probabilities
4526 After running a program compiled with @option{-fprofile-arcs}
4527 (@pxref{Debugging Options,, Options for Debugging Your Program or
4528 @command{gcc}}), you can compile it a second time using
4529 @option{-fbranch-probabilities}, to improve optimizations based on
4530 the number of times each branch was taken. When the program
4531 compiled with @option{-fprofile-arcs} exits it saves arc execution
4532 counts to a file called @file{@var{sourcename}.gcda} for each source
4533 file The information in this data file is very dependent on the
4534 structure of the generated code, so you must use the same source code
4535 and the same optimization options for both compilations.
4537 With @option{-fbranch-probabilities}, GCC puts a
4538 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
4539 These can be used to improve optimization. Currently, they are only
4540 used in one place: in @file{reorg.c}, instead of guessing which path a
4541 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
4542 exactly determine which path is taken more often.
4544 @item -fprofile-values
4545 @opindex fprofile-values
4546 If combined with @option{-fprofile-arcs}, it adds code so that some
4547 data about values of expressions in the program is gathered.
4549 With @option{-fbranch-probabilities}, it reads back the data gathered
4550 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
4551 notes to instructions for their later usage in optimizations.
4555 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
4556 a code to gather information about values of expressions.
4558 With @option{-fbranch-probabilities}, it reads back the data gathered
4559 and actually performs the optimizations based on them.
4560 Currently the optimizations include specialization of division operation
4561 using the knowledge about the value of the denominator.
4565 Use a graph coloring register allocator. Currently this option is meant
4566 for testing, so we are interested to hear about miscompilations with
4571 Perform tail duplication to enlarge superblock size. This transformation
4572 simplifies the control flow of the function allowing other optimizations to do
4575 @item -funit-at-a-time
4576 @opindex funit-at-a-time
4577 Parse the whole compilation unit before starting to produce code.
4578 This allows some extra optimizations to take place but consumes more
4581 @item -funroll-loops
4582 @opindex funroll-loops
4583 Unroll loops whose number of iterations can be determined at compile time or
4584 upon entry to the loop. @option{-funroll-loops} implies
4585 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
4586 (i.e. complete removal of loops with small constant number of iterations).
4587 This option makes code larger, and may or may not make it run faster.
4589 @item -funroll-all-loops
4590 @opindex funroll-all-loops
4591 Unroll all loops, even if their number of iterations is uncertain when
4592 the loop is entered. This usually makes programs run more slowly.
4593 @option{-funroll-all-loops} implies the same options as
4594 @option{-funroll-loops}.
4597 @opindex fpeel-loops
4598 Peels the loops for that there is enough information that they do not
4599 roll much (from profile feedback). It also turns on complete loop peeling
4600 (i.e. complete removal of loops with small constant number of iterations).
4602 @item -funswitch-loops
4603 @opindex funswitch-loops
4604 Move branches with loop invariant conditions out of the loop, with duplicates
4605 of the loop on both branches (modified according to result of the condition).
4607 @item -fold-unroll-loops
4608 @opindex fold-unroll-loops
4609 Unroll loops whose number of iterations can be determined at compile
4610 time or upon entry to the loop, using the old loop unroller whose loop
4611 recognition is based on notes from frontend. @option{-fold-unroll-loops} implies
4612 both @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4613 option makes code larger, and may or may not make it run faster.
4615 @item -fold-unroll-all-loops
4616 @opindex fold-unroll-all-loops
4617 Unroll all loops, even if their number of iterations is uncertain when
4618 the loop is entered. This is done using the old loop unroller whose loop
4619 recognition is based on notes from frontend. This usually makes programs run more slowly.
4620 @option{-fold-unroll-all-loops} implies the same options as
4621 @option{-fold-unroll-loops}.
4623 @item -funswitch-loops
4624 @opindex funswitch-loops
4625 Move branches with loop invariant conditions out of the loop, with duplicates
4626 of the loop on both branches (modified according to result of the condition).
4628 @item -funswitch-loops
4629 @opindex funswitch-loops
4630 Move branches with loop invariant conditions out of the loop, with duplicates
4631 of the loop on both branches (modified according to result of the condition).
4633 @item -fprefetch-loop-arrays
4634 @opindex fprefetch-loop-arrays
4635 If supported by the target machine, generate instructions to prefetch
4636 memory to improve the performance of loops that access large arrays.
4638 Disabled at level @option{-Os}.
4640 @item -ffunction-sections
4641 @itemx -fdata-sections
4642 @opindex ffunction-sections
4643 @opindex fdata-sections
4644 Place each function or data item into its own section in the output
4645 file if the target supports arbitrary sections. The name of the
4646 function or the name of the data item determines the section's name
4649 Use these options on systems where the linker can perform optimizations
4650 to improve locality of reference in the instruction space. Most systems
4651 using the ELF object format and SPARC processors running Solaris 2 have
4652 linkers with such optimizations. AIX may have these optimizations in
4655 Only use these options when there are significant benefits from doing
4656 so. When you specify these options, the assembler and linker will
4657 create larger object and executable files and will also be slower.
4658 You will not be able to use @code{gprof} on all systems if you
4659 specify this option and you may have problems with debugging if
4660 you specify both this option and @option{-g}.
4662 @item -fbranch-target-load-optimize
4663 @opindex fbranch-target-load-optimize
4664 Perform branch target register load optimization before prologue / epilogue
4666 The use of target registers can typically be exposed only during reload,
4667 thus hoisting loads out of loops and doing inter-block scheduling needs
4668 a separate optimization pass.
4670 @item -fbranch-target-load-optimize2
4671 @opindex fbranch-target-load-optimize2
4672 Perform branch target register load optimization after prologue / epilogue
4675 @item --param @var{name}=@var{value}
4677 In some places, GCC uses various constants to control the amount of
4678 optimization that is done. For example, GCC will not inline functions
4679 that contain more that a certain number of instructions. You can
4680 control some of these constants on the command-line using the
4681 @option{--param} option.
4683 In each case, the @var{value} is an integer. The allowable choices for
4684 @var{name} are given in the following table:
4687 @item max-crossjump-edges
4688 The maximum number of incoming edges to consider for crossjumping.
4689 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
4690 the number of edges incoming to each block. Increasing values mean
4691 more aggressive optimization, making the compile time increase with
4692 probably small improvement in executable size.
4694 @item max-delay-slot-insn-search
4695 The maximum number of instructions to consider when looking for an
4696 instruction to fill a delay slot. If more than this arbitrary number of
4697 instructions is searched, the time savings from filling the delay slot
4698 will be minimal so stop searching. Increasing values mean more
4699 aggressive optimization, making the compile time increase with probably
4700 small improvement in executable run time.
4702 @item max-delay-slot-live-search
4703 When trying to fill delay slots, the maximum number of instructions to
4704 consider when searching for a block with valid live register
4705 information. Increasing this arbitrarily chosen value means more
4706 aggressive optimization, increasing the compile time. This parameter
4707 should be removed when the delay slot code is rewritten to maintain the
4710 @item max-gcse-memory
4711 The approximate maximum amount of memory that will be allocated in
4712 order to perform the global common subexpression elimination
4713 optimization. If more memory than specified is required, the
4714 optimization will not be done.
4716 @item max-gcse-passes
4717 The maximum number of passes of GCSE to run.
4719 @item max-pending-list-length
4720 The maximum number of pending dependencies scheduling will allow
4721 before flushing the current state and starting over. Large functions
4722 with few branches or calls can create excessively large lists which
4723 needlessly consume memory and resources.
4725 @item max-inline-insns-single
4726 Several parameters control the tree inliner used in gcc.
4727 This number sets the maximum number of instructions (counted in gcc's
4728 internal representation) in a single function that the tree inliner
4729 will consider for inlining. This only affects functions declared
4730 inline and methods implemented in a class declaration (C++).
4731 The default value is 500.
4733 @item max-inline-insns-auto
4734 When you use @option{-finline-functions} (included in @option{-O3}),
4735 a lot of functions that would otherwise not be considered for inlining
4736 by the compiler will be investigated. To those functions, a different
4737 (more restrictive) limit compared to functions declared inline can
4739 The default value is 150.
4741 @item large-function-insns
4742 The limit specifying really large functions. For functions greater than this
4743 limit inlining is constrained by @option{--param large-function-growth}.
4744 This parameter is useful primarily to avoid extreme compilation time caused by non-linear
4745 algorithms used by the backend.
4746 This parameter is ignored when @option{-funit-at-a-time} is not used.
4747 The default value is 30000.
4749 @item large-function-growth
4750 Specifies maximal growth of large functtion caused by inlining in percents.
4751 This parameter is ignored when @option{-funit-at-a-time} is not used.
4752 The default value is 200.
4754 @item inline-unit-growth
4755 Specifies maximal overall growth of the compilation unit caused by inlining.
4756 This parameter is ignored when @option{-funit-at-a-time} is not used.
4757 The default value is 150.
4759 @item max-inline-insns-rtl
4760 For languages that use the RTL inliner (this happens at a later stage
4761 than tree inlining), you can set the maximum allowable size (counted
4762 in RTL instructions) for the RTL inliner with this parameter.
4763 The default value is 600.
4765 @item max-unrolled-insns
4766 The maximum number of instructions that a loop should have if that loop
4767 is unrolled, and if the loop is unrolled, it determines how many times
4768 the loop code is unrolled.
4770 @item max-average-unrolled-insns
4771 The maximum number of instructions biased by probabilities of their execution
4772 that a loop should have if that loop is unrolled, and if the loop is unrolled,
4773 it determines how many times the loop code is unrolled.
4775 @item max-unroll-times
4776 The maximum number of unrollings of a single loop.
4778 @item max-peeled-insns
4779 The maximum number of instructions that a loop should have if that loop
4780 is peeled, and if the loop is peeled, it determines how many times
4781 the loop code is peeled.
4783 @item max-peel-times
4784 The maximum number of peelings of a single loop.
4786 @item max-completely-peeled-insns
4787 The maximum number of insns of a completely peeled loop.
4789 @item max-completely-peel-times
4790 The maximum number of iterations of a loop to be suitable for complete peeling.
4792 @item max-unswitch-insns
4793 The maximum number of insns of an unswitched loop.
4795 @item max-unswitch-level
4796 The maximum number of branches unswitched in a single loop.
4798 @item hot-bb-count-fraction
4799 Select fraction of the maximal count of repetitions of basic block in program
4800 given basic block needs to have to be considered hot.
4802 @item hot-bb-frequency-fraction
4803 Select fraction of the maximal frequency of executions of basic block in
4804 function given basic block needs to have to be considered hot
4806 @item tracer-dynamic-coverage
4807 @itemx tracer-dynamic-coverage-feedback
4809 This value is used to limit superblock formation once the given percentage of
4810 executed instructions is covered. This limits unnecessary code size
4813 The @option{tracer-dynamic-coverage-feedback} is used only when profile
4814 feedback is available. The real profiles (as opposed to statically estimated
4815 ones) are much less balanced allowing the threshold to be larger value.
4817 @item tracer-max-code-growth
4818 Stop tail duplication once code growth has reached given percentage. This is
4819 rather hokey argument, as most of the duplicates will be eliminated later in
4820 cross jumping, so it may be set to much higher values than is the desired code
4823 @item tracer-min-branch-ratio
4825 Stop reverse growth when the reverse probability of best edge is less than this
4826 threshold (in percent).
4828 @item tracer-min-branch-ratio
4829 @itemx tracer-min-branch-ratio-feedback
4831 Stop forward growth if the best edge do have probability lower than this
4834 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
4835 compilation for profile feedback and one for compilation without. The value
4836 for compilation with profile feedback needs to be more conservative (higher) in
4837 order to make tracer effective.
4839 @item max-cse-path-length
4841 Maximum number of basic blocks on path that cse considers.
4843 @item ggc-min-expand
4845 GCC uses a garbage collector to manage its own memory allocation. This
4846 parameter specifies the minimum percentage by which the garbage
4847 collector's heap should be allowed to expand between collections.
4848 Tuning this may improve compilation speed; it has no effect on code
4851 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
4852 RAM >= 1GB. If @code{getrlimit} is available, the notion of "RAM" is
4853 the smallest of actual RAM, RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If
4854 GCC is not able to calculate RAM on a particular platform, the lower
4855 bound of 30% is used. Setting this parameter and
4856 @option{ggc-min-heapsize} to zero causes a full collection to occur at
4857 every opportunity. This is extremely slow, but can be useful for
4860 @item ggc-min-heapsize
4862 Minimum size of the garbage collector's heap before it begins bothering
4863 to collect garbage. The first collection occurs after the heap expands
4864 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
4865 tuning this may improve compilation speed, and has no effect on code
4868 The default is RAM/8, with a lower bound of 4096 (four megabytes) and an
4869 upper bound of 131072 (128 megabytes). If @code{getrlimit} is
4870 available, the notion of "RAM" is the smallest of actual RAM,
4871 RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If GCC is not able to calculate
4872 RAM on a particular platform, the lower bound is used. Setting this
4873 parameter very large effectively disables garbage collection. Setting
4874 this parameter and @option{ggc-min-expand} to zero causes a full
4875 collection to occur at every opportunity.
4877 @item reorder-blocks-duplicate
4878 @itemx reorder-blocks-duplicate-feedback
4880 Used by basic block reordering pass to decide whether to use unconditional
4881 branch or duplicate the code on its destination. Code is duplicated when its
4882 estimated size is smaller than this value multiplied by the estimated size of
4883 unconditional jump in the hot spots of the program.
4885 The @option{reorder-block-duplicate-feedback} is used only when profile
4886 feedback is available and may be set to higher values than
4887 @option{reorder-block-duplicate} since information about the hot spots is more
4892 @node Preprocessor Options
4893 @section Options Controlling the Preprocessor
4894 @cindex preprocessor options
4895 @cindex options, preprocessor
4897 These options control the C preprocessor, which is run on each C source
4898 file before actual compilation.
4900 If you use the @option{-E} option, nothing is done except preprocessing.
4901 Some of these options make sense only together with @option{-E} because
4902 they cause the preprocessor output to be unsuitable for actual
4907 You can use @option{-Wp,@var{option}} to bypass the compiler driver
4908 and pass @var{option} directly through to the preprocessor. If
4909 @var{option} contains commas, it is split into multiple options at the
4910 commas. However, many options are modified, translated or interpreted
4911 by the compiler driver before being passed to the preprocessor, and
4912 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
4913 interface is undocumented and subject to change, so whenever possible
4914 you should avoid using @option{-Wp} and let the driver handle the
4917 @item -Xpreprocessor @var{option}
4918 @opindex preprocessor
4919 Pass @var{option} as an option to the preprocessor. You can use this to
4920 supply system-specific preprocessor options which GCC does not know how to
4923 If you want to pass an option that takes an argument, you must use
4924 @option{-Xpreprocessor} twice, once for the option and once for the argument.
4927 @include cppopts.texi
4929 @node Assembler Options
4930 @section Passing Options to the Assembler
4932 @c prevent bad page break with this line
4933 You can pass options to the assembler.
4936 @item -Wa,@var{option}
4938 Pass @var{option} as an option to the assembler. If @var{option}
4939 contains commas, it is split into multiple options at the commas.
4941 @item -Xassembler @var{option}
4943 Pass @var{option} as an option to the assembler. You can use this to
4944 supply system-specific assembler options which GCC does not know how to
4947 If you want to pass an option that takes an argument, you must use
4948 @option{-Xassembler} twice, once for the option and once for the argument.
4953 @section Options for Linking
4954 @cindex link options
4955 @cindex options, linking
4957 These options come into play when the compiler links object files into
4958 an executable output file. They are meaningless if the compiler is
4959 not doing a link step.
4963 @item @var{object-file-name}
4964 A file name that does not end in a special recognized suffix is
4965 considered to name an object file or library. (Object files are
4966 distinguished from libraries by the linker according to the file
4967 contents.) If linking is done, these object files are used as input
4976 If any of these options is used, then the linker is not run, and
4977 object file names should not be used as arguments. @xref{Overall
4981 @item -l@var{library}
4982 @itemx -l @var{library}
4984 Search the library named @var{library} when linking. (The second
4985 alternative with the library as a separate argument is only for
4986 POSIX compliance and is not recommended.)
4988 It makes a difference where in the command you write this option; the
4989 linker searches and processes libraries and object files in the order they
4990 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
4991 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
4992 to functions in @samp{z}, those functions may not be loaded.
4994 The linker searches a standard list of directories for the library,
4995 which is actually a file named @file{lib@var{library}.a}. The linker
4996 then uses this file as if it had been specified precisely by name.
4998 The directories searched include several standard system directories
4999 plus any that you specify with @option{-L}.
5001 Normally the files found this way are library files---archive files
5002 whose members are object files. The linker handles an archive file by
5003 scanning through it for members which define symbols that have so far
5004 been referenced but not defined. But if the file that is found is an
5005 ordinary object file, it is linked in the usual fashion. The only
5006 difference between using an @option{-l} option and specifying a file name
5007 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5008 and searches several directories.
5012 You need this special case of the @option{-l} option in order to
5013 link an Objective-C program.
5016 @opindex nostartfiles
5017 Do not use the standard system startup files when linking.
5018 The standard system libraries are used normally, unless @option{-nostdlib}
5019 or @option{-nodefaultlibs} is used.
5021 @item -nodefaultlibs
5022 @opindex nodefaultlibs
5023 Do not use the standard system libraries when linking.
5024 Only the libraries you specify will be passed to the linker.
5025 The standard startup files are used normally, unless @option{-nostartfiles}
5026 is used. The compiler may generate calls to memcmp, memset, and memcpy
5027 for System V (and ISO C) environments or to bcopy and bzero for
5028 BSD environments. These entries are usually resolved by entries in
5029 libc. These entry points should be supplied through some other
5030 mechanism when this option is specified.
5034 Do not use the standard system startup files or libraries when linking.
5035 No startup files and only the libraries you specify will be passed to
5036 the linker. The compiler may generate calls to memcmp, memset, and memcpy
5037 for System V (and ISO C) environments or to bcopy and bzero for
5038 BSD environments. These entries are usually resolved by entries in
5039 libc. These entry points should be supplied through some other
5040 mechanism when this option is specified.
5042 @cindex @option{-lgcc}, use with @option{-nostdlib}
5043 @cindex @option{-nostdlib} and unresolved references
5044 @cindex unresolved references and @option{-nostdlib}
5045 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5046 @cindex @option{-nodefaultlibs} and unresolved references
5047 @cindex unresolved references and @option{-nodefaultlibs}
5048 One of the standard libraries bypassed by @option{-nostdlib} and
5049 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5050 that GCC uses to overcome shortcomings of particular machines, or special
5051 needs for some languages.
5052 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5053 Collection (GCC) Internals},
5054 for more discussion of @file{libgcc.a}.)
5055 In most cases, you need @file{libgcc.a} even when you want to avoid
5056 other standard libraries. In other words, when you specify @option{-nostdlib}
5057 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5058 This ensures that you have no unresolved references to internal GCC
5059 library subroutines. (For example, @samp{__main}, used to ensure C++
5060 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5061 GNU Compiler Collection (GCC) Internals}.)
5065 Produce a position independent executable on targets which support it.
5066 For predictable results, you must also specify the same set of options
5067 that were used to generate code (@option{-fpie}, @option{-fPIE},
5068 or model suboptions) when you specify this option.
5072 Remove all symbol table and relocation information from the executable.
5076 On systems that support dynamic linking, this prevents linking with the shared
5077 libraries. On other systems, this option has no effect.
5081 Produce a shared object which can then be linked with other objects to
5082 form an executable. Not all systems support this option. For predictable
5083 results, you must also specify the same set of options that were used to
5084 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5085 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5086 needs to build supplementary stub code for constructors to work. On
5087 multi-libbed systems, @samp{gcc -shared} must select the correct support
5088 libraries to link against. Failing to supply the correct flags may lead
5089 to subtle defects. Supplying them in cases where they are not necessary
5092 @item -shared-libgcc
5093 @itemx -static-libgcc
5094 @opindex shared-libgcc
5095 @opindex static-libgcc
5096 On systems that provide @file{libgcc} as a shared library, these options
5097 force the use of either the shared or static version respectively.
5098 If no shared version of @file{libgcc} was built when the compiler was
5099 configured, these options have no effect.
5101 There are several situations in which an application should use the
5102 shared @file{libgcc} instead of the static version. The most common
5103 of these is when the application wishes to throw and catch exceptions
5104 across different shared libraries. In that case, each of the libraries
5105 as well as the application itself should use the shared @file{libgcc}.
5107 Therefore, the G++ and GCJ drivers automatically add
5108 @option{-shared-libgcc} whenever you build a shared library or a main
5109 executable, because C++ and Java programs typically use exceptions, so
5110 this is the right thing to do.
5112 If, instead, you use the GCC driver to create shared libraries, you may
5113 find that they will not always be linked with the shared @file{libgcc}.
5114 If GCC finds, at its configuration time, that you have a GNU linker that
5115 does not support option @option{--eh-frame-hdr}, it will link the shared
5116 version of @file{libgcc} into shared libraries by default. Otherwise,
5117 it will take advantage of the linker and optimize away the linking with
5118 the shared version of @file{libgcc}, linking with the static version of
5119 libgcc by default. This allows exceptions to propagate through such
5120 shared libraries, without incurring relocation costs at library load
5123 However, if a library or main executable is supposed to throw or catch
5124 exceptions, you must link it using the G++ or GCJ driver, as appropriate
5125 for the languages used in the program, or using the option
5126 @option{-shared-libgcc}, such that it is linked with the shared
5131 Bind references to global symbols when building a shared object. Warn
5132 about any unresolved references (unless overridden by the link editor
5133 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
5136 @item -Xlinker @var{option}
5138 Pass @var{option} as an option to the linker. You can use this to
5139 supply system-specific linker options which GCC does not know how to
5142 If you want to pass an option that takes an argument, you must use
5143 @option{-Xlinker} twice, once for the option and once for the argument.
5144 For example, to pass @option{-assert definitions}, you must write
5145 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
5146 @option{-Xlinker "-assert definitions"}, because this passes the entire
5147 string as a single argument, which is not what the linker expects.
5149 @item -Wl,@var{option}
5151 Pass @var{option} as an option to the linker. If @var{option} contains
5152 commas, it is split into multiple options at the commas.
5154 @item -u @var{symbol}
5156 Pretend the symbol @var{symbol} is undefined, to force linking of
5157 library modules to define it. You can use @option{-u} multiple times with
5158 different symbols to force loading of additional library modules.
5161 @node Directory Options
5162 @section Options for Directory Search
5163 @cindex directory options
5164 @cindex options, directory search
5167 These options specify directories to search for header files, for
5168 libraries and for parts of the compiler:
5173 Add the directory @var{dir} to the head of the list of directories to be
5174 searched for header files. This can be used to override a system header
5175 file, substituting your own version, since these directories are
5176 searched before the system header file directories. However, you should
5177 not use this option to add directories that contain vendor-supplied
5178 system header files (use @option{-isystem} for that). If you use more than
5179 one @option{-I} option, the directories are scanned in left-to-right
5180 order; the standard system directories come after.
5182 If a standard system include directory, or a directory specified with
5183 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
5184 option will be ignored. The directory will still be searched but as a
5185 system directory at its normal position in the system include chain.
5186 This is to ensure that GCC's procedure to fix buggy system headers and
5187 the ordering for the include_next directive are not inadvertently changed.
5188 If you really need to change the search order for system directories,
5189 use the @option{-nostdinc} and/or @option{-isystem} options.
5193 Any directories you specify with @option{-I} options before the @option{-I-}
5194 option are searched only for the case of @samp{#include "@var{file}"};
5195 they are not searched for @samp{#include <@var{file}>}.
5197 If additional directories are specified with @option{-I} options after
5198 the @option{-I-}, these directories are searched for all @samp{#include}
5199 directives. (Ordinarily @emph{all} @option{-I} directories are used
5202 In addition, the @option{-I-} option inhibits the use of the current
5203 directory (where the current input file came from) as the first search
5204 directory for @samp{#include "@var{file}"}. There is no way to
5205 override this effect of @option{-I-}. With @option{-I.} you can specify
5206 searching the directory which was current when the compiler was
5207 invoked. That is not exactly the same as what the preprocessor does
5208 by default, but it is often satisfactory.
5210 @option{-I-} does not inhibit the use of the standard system directories
5211 for header files. Thus, @option{-I-} and @option{-nostdinc} are
5216 Add directory @var{dir} to the list of directories to be searched
5219 @item -B@var{prefix}
5221 This option specifies where to find the executables, libraries,
5222 include files, and data files of the compiler itself.
5224 The compiler driver program runs one or more of the subprograms
5225 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
5226 @var{prefix} as a prefix for each program it tries to run, both with and
5227 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
5229 For each subprogram to be run, the compiler driver first tries the
5230 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
5231 was not specified, the driver tries two standard prefixes, which are
5232 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
5233 those results in a file name that is found, the unmodified program
5234 name is searched for using the directories specified in your
5235 @env{PATH} environment variable.
5237 The compiler will check to see if the path provided by the @option{-B}
5238 refers to a directory, and if necessary it will add a directory
5239 separator character at the end of the path.
5241 @option{-B} prefixes that effectively specify directory names also apply
5242 to libraries in the linker, because the compiler translates these
5243 options into @option{-L} options for the linker. They also apply to
5244 includes files in the preprocessor, because the compiler translates these
5245 options into @option{-isystem} options for the preprocessor. In this case,
5246 the compiler appends @samp{include} to the prefix.
5248 The run-time support file @file{libgcc.a} can also be searched for using
5249 the @option{-B} prefix, if needed. If it is not found there, the two
5250 standard prefixes above are tried, and that is all. The file is left
5251 out of the link if it is not found by those means.
5253 Another way to specify a prefix much like the @option{-B} prefix is to use
5254 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
5257 As a special kludge, if the path provided by @option{-B} is
5258 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
5259 9, then it will be replaced by @file{[dir/]include}. This is to help
5260 with boot-strapping the compiler.
5262 @item -specs=@var{file}
5264 Process @var{file} after the compiler reads in the standard @file{specs}
5265 file, in order to override the defaults that the @file{gcc} driver
5266 program uses when determining what switches to pass to @file{cc1},
5267 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
5268 @option{-specs=@var{file}} can be specified on the command line, and they
5269 are processed in order, from left to right.
5275 @section Specifying subprocesses and the switches to pass to them
5278 @command{gcc} is a driver program. It performs its job by invoking a
5279 sequence of other programs to do the work of compiling, assembling and
5280 linking. GCC interprets its command-line parameters and uses these to
5281 deduce which programs it should invoke, and which command-line options
5282 it ought to place on their command lines. This behavior is controlled
5283 by @dfn{spec strings}. In most cases there is one spec string for each
5284 program that GCC can invoke, but a few programs have multiple spec
5285 strings to control their behavior. The spec strings built into GCC can
5286 be overridden by using the @option{-specs=} command-line switch to specify
5289 @dfn{Spec files} are plaintext files that are used to construct spec
5290 strings. They consist of a sequence of directives separated by blank
5291 lines. The type of directive is determined by the first non-whitespace
5292 character on the line and it can be one of the following:
5295 @item %@var{command}
5296 Issues a @var{command} to the spec file processor. The commands that can
5300 @item %include <@var{file}>
5302 Search for @var{file} and insert its text at the current point in the
5305 @item %include_noerr <@var{file}>
5306 @cindex %include_noerr
5307 Just like @samp{%include}, but do not generate an error message if the include
5308 file cannot be found.
5310 @item %rename @var{old_name} @var{new_name}
5312 Rename the spec string @var{old_name} to @var{new_name}.
5316 @item *[@var{spec_name}]:
5317 This tells the compiler to create, override or delete the named spec
5318 string. All lines after this directive up to the next directive or
5319 blank line are considered to be the text for the spec string. If this
5320 results in an empty string then the spec will be deleted. (Or, if the
5321 spec did not exist, then nothing will happened.) Otherwise, if the spec
5322 does not currently exist a new spec will be created. If the spec does
5323 exist then its contents will be overridden by the text of this
5324 directive, unless the first character of that text is the @samp{+}
5325 character, in which case the text will be appended to the spec.
5327 @item [@var{suffix}]:
5328 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
5329 and up to the next directive or blank line are considered to make up the
5330 spec string for the indicated suffix. When the compiler encounters an
5331 input file with the named suffix, it will processes the spec string in
5332 order to work out how to compile that file. For example:
5339 This says that any input file whose name ends in @samp{.ZZ} should be
5340 passed to the program @samp{z-compile}, which should be invoked with the
5341 command-line switch @option{-input} and with the result of performing the
5342 @samp{%i} substitution. (See below.)
5344 As an alternative to providing a spec string, the text that follows a
5345 suffix directive can be one of the following:
5348 @item @@@var{language}
5349 This says that the suffix is an alias for a known @var{language}. This is
5350 similar to using the @option{-x} command-line switch to GCC to specify a
5351 language explicitly. For example:
5358 Says that .ZZ files are, in fact, C++ source files.
5361 This causes an error messages saying:
5364 @var{name} compiler not installed on this system.
5368 GCC already has an extensive list of suffixes built into it.
5369 This directive will add an entry to the end of the list of suffixes, but
5370 since the list is searched from the end backwards, it is effectively
5371 possible to override earlier entries using this technique.
5375 GCC has the following spec strings built into it. Spec files can
5376 override these strings or create their own. Note that individual
5377 targets can also add their own spec strings to this list.
5380 asm Options to pass to the assembler
5381 asm_final Options to pass to the assembler post-processor
5382 cpp Options to pass to the C preprocessor
5383 cc1 Options to pass to the C compiler
5384 cc1plus Options to pass to the C++ compiler
5385 endfile Object files to include at the end of the link
5386 link Options to pass to the linker
5387 lib Libraries to include on the command line to the linker
5388 libgcc Decides which GCC support library to pass to the linker
5389 linker Sets the name of the linker
5390 predefines Defines to be passed to the C preprocessor
5391 signed_char Defines to pass to CPP to say whether @code{char} is signed
5393 startfile Object files to include at the start of the link
5396 Here is a small example of a spec file:
5402 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
5405 This example renames the spec called @samp{lib} to @samp{old_lib} and
5406 then overrides the previous definition of @samp{lib} with a new one.
5407 The new definition adds in some extra command-line options before
5408 including the text of the old definition.
5410 @dfn{Spec strings} are a list of command-line options to be passed to their
5411 corresponding program. In addition, the spec strings can contain
5412 @samp{%}-prefixed sequences to substitute variable text or to
5413 conditionally insert text into the command line. Using these constructs
5414 it is possible to generate quite complex command lines.
5416 Here is a table of all defined @samp{%}-sequences for spec
5417 strings. Note that spaces are not generated automatically around the
5418 results of expanding these sequences. Therefore you can concatenate them
5419 together or combine them with constant text in a single argument.
5423 Substitute one @samp{%} into the program name or argument.
5426 Substitute the name of the input file being processed.
5429 Substitute the basename of the input file being processed.
5430 This is the substring up to (and not including) the last period
5431 and not including the directory.
5434 This is the same as @samp{%b}, but include the file suffix (text after
5438 Marks the argument containing or following the @samp{%d} as a
5439 temporary file name, so that that file will be deleted if GCC exits
5440 successfully. Unlike @samp{%g}, this contributes no text to the
5443 @item %g@var{suffix}
5444 Substitute a file name that has suffix @var{suffix} and is chosen
5445 once per compilation, and mark the argument in the same way as
5446 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
5447 name is now chosen in a way that is hard to predict even when previously
5448 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
5449 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
5450 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
5451 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
5452 was simply substituted with a file name chosen once per compilation,
5453 without regard to any appended suffix (which was therefore treated
5454 just like ordinary text), making such attacks more likely to succeed.
5456 @item %u@var{suffix}
5457 Like @samp{%g}, but generates a new temporary file name even if
5458 @samp{%u@var{suffix}} was already seen.
5460 @item %U@var{suffix}
5461 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
5462 new one if there is no such last file name. In the absence of any
5463 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
5464 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
5465 would involve the generation of two distinct file names, one
5466 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
5467 simply substituted with a file name chosen for the previous @samp{%u},
5468 without regard to any appended suffix.
5470 @item %j@var{suffix}
5471 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
5472 writable, and if save-temps is off; otherwise, substitute the name
5473 of a temporary file, just like @samp{%u}. This temporary file is not
5474 meant for communication between processes, but rather as a junk
5477 @item %|@var{suffix}
5478 @itemx %m@var{suffix}
5479 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
5480 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
5481 all. These are the two most common ways to instruct a program that it
5482 should read from standard input or write to standard output. If you
5483 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
5484 construct: see for example @file{f/lang-specs.h}.
5486 @item %.@var{SUFFIX}
5487 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
5488 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
5489 terminated by the next space or %.
5492 Marks the argument containing or following the @samp{%w} as the
5493 designated output file of this compilation. This puts the argument
5494 into the sequence of arguments that @samp{%o} will substitute later.
5497 Substitutes the names of all the output files, with spaces
5498 automatically placed around them. You should write spaces
5499 around the @samp{%o} as well or the results are undefined.
5500 @samp{%o} is for use in the specs for running the linker.
5501 Input files whose names have no recognized suffix are not compiled
5502 at all, but they are included among the output files, so they will
5506 Substitutes the suffix for object files. Note that this is
5507 handled specially when it immediately follows @samp{%g, %u, or %U},
5508 because of the need for those to form complete file names. The
5509 handling is such that @samp{%O} is treated exactly as if it had already
5510 been substituted, except that @samp{%g, %u, and %U} do not currently
5511 support additional @var{suffix} characters following @samp{%O} as they would
5512 following, for example, @samp{.o}.
5515 Substitutes the standard macro predefinitions for the
5516 current target machine. Use this when running @code{cpp}.
5519 Like @samp{%p}, but puts @samp{__} before and after the name of each
5520 predefined macro, except for macros that start with @samp{__} or with
5521 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
5525 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
5526 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
5527 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
5531 Current argument is the name of a library or startup file of some sort.
5532 Search for that file in a standard list of directories and substitute
5533 the full name found.
5536 Print @var{str} as an error message. @var{str} is terminated by a newline.
5537 Use this when inconsistent options are detected.
5540 Substitute the contents of spec string @var{name} at this point.
5543 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
5545 @item %x@{@var{option}@}
5546 Accumulate an option for @samp{%X}.
5549 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
5553 Output the accumulated assembler options specified by @option{-Wa}.
5556 Output the accumulated preprocessor options specified by @option{-Wp}.
5559 Process the @code{asm} spec. This is used to compute the
5560 switches to be passed to the assembler.
5563 Process the @code{asm_final} spec. This is a spec string for
5564 passing switches to an assembler post-processor, if such a program is
5568 Process the @code{link} spec. This is the spec for computing the
5569 command line passed to the linker. Typically it will make use of the
5570 @samp{%L %G %S %D and %E} sequences.
5573 Dump out a @option{-L} option for each directory that GCC believes might
5574 contain startup files. If the target supports multilibs then the
5575 current multilib directory will be prepended to each of these paths.
5578 Output the multilib directory with directory separators replaced with
5579 @samp{_}. If multilib directories are not set, or the multilib directory is
5580 @file{.} then this option emits nothing.
5583 Process the @code{lib} spec. This is a spec string for deciding which
5584 libraries should be included on the command line to the linker.
5587 Process the @code{libgcc} spec. This is a spec string for deciding
5588 which GCC support library should be included on the command line to the linker.
5591 Process the @code{startfile} spec. This is a spec for deciding which
5592 object files should be the first ones passed to the linker. Typically
5593 this might be a file named @file{crt0.o}.
5596 Process the @code{endfile} spec. This is a spec string that specifies
5597 the last object files that will be passed to the linker.
5600 Process the @code{cpp} spec. This is used to construct the arguments
5601 to be passed to the C preprocessor.
5604 Process the @code{signed_char} spec. This is intended to be used
5605 to tell cpp whether a char is signed. It typically has the definition:
5607 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
5611 Process the @code{cc1} spec. This is used to construct the options to be
5612 passed to the actual C compiler (@samp{cc1}).
5615 Process the @code{cc1plus} spec. This is used to construct the options to be
5616 passed to the actual C++ compiler (@samp{cc1plus}).
5619 Substitute the variable part of a matched option. See below.
5620 Note that each comma in the substituted string is replaced by
5624 Remove all occurrences of @code{-S} from the command line. Note---this
5625 command is position dependent. @samp{%} commands in the spec string
5626 before this one will see @code{-S}, @samp{%} commands in the spec string
5627 after this one will not.
5629 @item %:@var{function}(@var{args})
5630 Call the named function @var{function}, passing it @var{args}.
5631 @var{args} is first processed as a nested spec string, then split
5632 into an argument vector in the usual fashion. The function returns
5633 a string which is processed as if it had appeared literally as part
5634 of the current spec.
5636 The following built-in spec functions are provided:
5639 @item @code{if-exists}
5640 The @code{if-exists} spec function takes one argument, an absolute
5641 pathname to a file. If the file exists, @code{if-exists} returns the
5642 pathname. Here is a small example of its usage:
5646 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
5649 @item @code{if-exists-else}
5650 The @code{if-exists-else} spec function is similar to the @code{if-exists}
5651 spec function, except that it takes two arguments. The first argument is
5652 an absolute pathname to a file. If the file exists, @code{if-exists-else}
5653 returns the pathname. If it does not exist, it returns the second argument.
5654 This way, @code{if-exists-else} can be used to select one file or another,
5655 based on the existence of the first. Here is a small example of its usage:
5659 crt0%O%s %:if-exists(crti%O%s) \
5660 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
5665 Substitutes the @code{-S} switch, if that switch was given to GCC@.
5666 If that switch was not specified, this substitutes nothing. Note that
5667 the leading dash is omitted when specifying this option, and it is
5668 automatically inserted if the substitution is performed. Thus the spec
5669 string @samp{%@{foo@}} would match the command-line option @option{-foo}
5670 and would output the command line option @option{-foo}.
5672 @item %W@{@code{S}@}
5673 Like %@{@code{S}@} but mark last argument supplied within as a file to be
5676 @item %@{@code{S}*@}
5677 Substitutes all the switches specified to GCC whose names start
5678 with @code{-S}, but which also take an argument. This is used for
5679 switches like @option{-o}, @option{-D}, @option{-I}, etc.
5680 GCC considers @option{-o foo} as being
5681 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
5682 text, including the space. Thus two arguments would be generated.
5684 @item %@{@code{S}*&@code{T}*@}
5685 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
5686 (the order of @code{S} and @code{T} in the spec is not significant).
5687 There can be any number of ampersand-separated variables; for each the
5688 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
5690 @item %@{@code{S}:@code{X}@}
5691 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
5693 @item %@{!@code{S}:@code{X}@}
5694 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
5696 @item %@{@code{S}*:@code{X}@}
5697 Substitutes @code{X} if one or more switches whose names start with
5698 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
5699 once, no matter how many such switches appeared. However, if @code{%*}
5700 appears somewhere in @code{X}, then @code{X} will be substituted once
5701 for each matching switch, with the @code{%*} replaced by the part of
5702 that switch that matched the @code{*}.
5704 @item %@{.@code{S}:@code{X}@}
5705 Substitutes @code{X}, if processing a file with suffix @code{S}.
5707 @item %@{!.@code{S}:@code{X}@}
5708 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
5710 @item %@{@code{S}|@code{P}:@code{X}@}
5711 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
5712 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
5713 although they have a stronger binding than the @samp{|}. If @code{%*}
5714 appears in @code{X}, all of the alternatives must be starred, and only
5715 the first matching alternative is substituted.
5717 For example, a spec string like this:
5720 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
5723 will output the following command-line options from the following input
5724 command-line options:
5729 -d fred.c -foo -baz -boggle
5730 -d jim.d -bar -baz -boggle
5733 @item %@{S:X; T:Y; :D@}
5735 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
5736 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
5737 be as many clauses as you need. This may be combined with @code{.},
5738 @code{!}, @code{|}, and @code{*} as needed.
5743 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
5744 construct may contain other nested @samp{%} constructs or spaces, or
5745 even newlines. They are processed as usual, as described above.
5746 Trailing white space in @code{X} is ignored. White space may also
5747 appear anywhere on the left side of the colon in these constructs,
5748 except between @code{.} or @code{*} and the corresponding word.
5750 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
5751 handled specifically in these constructs. If another value of
5752 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
5753 @option{-W} switch is found later in the command line, the earlier
5754 switch value is ignored, except with @{@code{S}*@} where @code{S} is
5755 just one letter, which passes all matching options.
5757 The character @samp{|} at the beginning of the predicate text is used to
5758 indicate that a command should be piped to the following command, but
5759 only if @option{-pipe} is specified.
5761 It is built into GCC which switches take arguments and which do not.
5762 (You might think it would be useful to generalize this to allow each
5763 compiler's spec to say which switches take arguments. But this cannot
5764 be done in a consistent fashion. GCC cannot even decide which input
5765 files have been specified without knowing which switches take arguments,
5766 and it must know which input files to compile in order to tell which
5769 GCC also knows implicitly that arguments starting in @option{-l} are to be
5770 treated as compiler output files, and passed to the linker in their
5771 proper position among the other output files.
5773 @c man begin OPTIONS
5775 @node Target Options
5776 @section Specifying Target Machine and Compiler Version
5777 @cindex target options
5778 @cindex cross compiling
5779 @cindex specifying machine version
5780 @cindex specifying compiler version and target machine
5781 @cindex compiler version, specifying
5782 @cindex target machine, specifying
5784 The usual way to run GCC is to run the executable called @file{gcc}, or
5785 @file{<machine>-gcc} when cross-compiling, or
5786 @file{<machine>-gcc-<version>} to run a version other than the one that
5787 was installed last. Sometimes this is inconvenient, so GCC provides
5788 options that will switch to another cross-compiler or version.
5791 @item -b @var{machine}
5793 The argument @var{machine} specifies the target machine for compilation.
5795 The value to use for @var{machine} is the same as was specified as the
5796 machine type when configuring GCC as a cross-compiler. For
5797 example, if a cross-compiler was configured with @samp{configure
5798 i386v}, meaning to compile for an 80386 running System V, then you
5799 would specify @option{-b i386v} to run that cross compiler.
5801 @item -V @var{version}
5803 The argument @var{version} specifies which version of GCC to run.
5804 This is useful when multiple versions are installed. For example,
5805 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
5808 The @option{-V} and @option{-b} options work by running the
5809 @file{<machine>-gcc-<version>} executable, so there's no real reason to
5810 use them if you can just run that directly.
5812 @node Submodel Options
5813 @section Hardware Models and Configurations
5814 @cindex submodel options
5815 @cindex specifying hardware config
5816 @cindex hardware models and configurations, specifying
5817 @cindex machine dependent options
5819 Earlier we discussed the standard option @option{-b} which chooses among
5820 different installed compilers for completely different target
5821 machines, such as VAX vs.@: 68000 vs.@: 80386.
5823 In addition, each of these target machine types can have its own
5824 special options, starting with @samp{-m}, to choose among various
5825 hardware models or configurations---for example, 68010 vs 68020,
5826 floating coprocessor or none. A single installed version of the
5827 compiler can compile for any model or configuration, according to the
5830 Some configurations of the compiler also support additional special
5831 options, usually for compatibility with other compilers on the same
5834 These options are defined by the macro @code{TARGET_SWITCHES} in the
5835 machine description. The default for the options is also defined by
5836 that macro, which enables you to change the defaults.
5848 * RS/6000 and PowerPC Options::
5852 * i386 and x86-64 Options::
5854 * Intel 960 Options::
5855 * DEC Alpha Options::
5856 * DEC Alpha/VMS Options::
5859 * System V Options::
5860 * TMS320C3x/C4x Options::
5868 * S/390 and zSeries Options::
5872 * Xstormy16 Options::
5877 @node M680x0 Options
5878 @subsection M680x0 Options
5879 @cindex M680x0 options
5881 These are the @samp{-m} options defined for the 68000 series. The default
5882 values for these options depends on which style of 68000 was selected when
5883 the compiler was configured; the defaults for the most common choices are
5891 Generate output for a 68000. This is the default
5892 when the compiler is configured for 68000-based systems.
5894 Use this option for microcontrollers with a 68000 or EC000 core,
5895 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
5901 Generate output for a 68020. This is the default
5902 when the compiler is configured for 68020-based systems.
5906 Generate output containing 68881 instructions for floating point.
5907 This is the default for most 68020 systems unless @option{--nfp} was
5908 specified when the compiler was configured.
5912 Generate output for a 68030. This is the default when the compiler is
5913 configured for 68030-based systems.
5917 Generate output for a 68040. This is the default when the compiler is
5918 configured for 68040-based systems.
5920 This option inhibits the use of 68881/68882 instructions that have to be
5921 emulated by software on the 68040. Use this option if your 68040 does not
5922 have code to emulate those instructions.
5926 Generate output for a 68060. This is the default when the compiler is
5927 configured for 68060-based systems.
5929 This option inhibits the use of 68020 and 68881/68882 instructions that
5930 have to be emulated by software on the 68060. Use this option if your 68060
5931 does not have code to emulate those instructions.
5935 Generate output for a CPU32. This is the default
5936 when the compiler is configured for CPU32-based systems.
5938 Use this option for microcontrollers with a
5939 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
5940 68336, 68340, 68341, 68349 and 68360.
5944 Generate output for a 520X ``coldfire'' family cpu. This is the default
5945 when the compiler is configured for 520X-based systems.
5947 Use this option for microcontroller with a 5200 core, including
5948 the MCF5202, MCF5203, MCF5204 and MCF5202.
5953 Generate output for a 68040, without using any of the new instructions.
5954 This results in code which can run relatively efficiently on either a
5955 68020/68881 or a 68030 or a 68040. The generated code does use the
5956 68881 instructions that are emulated on the 68040.
5960 Generate output for a 68060, without using any of the new instructions.
5961 This results in code which can run relatively efficiently on either a
5962 68020/68881 or a 68030 or a 68040. The generated code does use the
5963 68881 instructions that are emulated on the 68060.
5966 @opindex msoft-float
5967 Generate output containing library calls for floating point.
5968 @strong{Warning:} the requisite libraries are not available for all m68k
5969 targets. Normally the facilities of the machine's usual C compiler are
5970 used, but this can't be done directly in cross-compilation. You must
5971 make your own arrangements to provide suitable library functions for
5972 cross-compilation. The embedded targets @samp{m68k-*-aout} and
5973 @samp{m68k-*-coff} do provide software floating point support.
5977 Consider type @code{int} to be 16 bits wide, like @code{short int}.
5980 @opindex mnobitfield
5981 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
5982 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
5986 Do use the bit-field instructions. The @option{-m68020} option implies
5987 @option{-mbitfield}. This is the default if you use a configuration
5988 designed for a 68020.
5992 Use a different function-calling convention, in which functions
5993 that take a fixed number of arguments return with the @code{rtd}
5994 instruction, which pops their arguments while returning. This
5995 saves one instruction in the caller since there is no need to pop
5996 the arguments there.
5998 This calling convention is incompatible with the one normally
5999 used on Unix, so you cannot use it if you need to call libraries
6000 compiled with the Unix compiler.
6002 Also, you must provide function prototypes for all functions that
6003 take variable numbers of arguments (including @code{printf});
6004 otherwise incorrect code will be generated for calls to those
6007 In addition, seriously incorrect code will result if you call a
6008 function with too many arguments. (Normally, extra arguments are
6009 harmlessly ignored.)
6011 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
6012 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
6015 @itemx -mno-align-int
6017 @opindex mno-align-int
6018 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
6019 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
6020 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
6021 Aligning variables on 32-bit boundaries produces code that runs somewhat
6022 faster on processors with 32-bit busses at the expense of more memory.
6024 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
6025 align structures containing the above types differently than
6026 most published application binary interface specifications for the m68k.
6030 Use the pc-relative addressing mode of the 68000 directly, instead of
6031 using a global offset table. At present, this option implies @option{-fpic},
6032 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
6033 not presently supported with @option{-mpcrel}, though this could be supported for
6034 68020 and higher processors.
6036 @item -mno-strict-align
6037 @itemx -mstrict-align
6038 @opindex mno-strict-align
6039 @opindex mstrict-align
6040 Do not (do) assume that unaligned memory references will be handled by
6044 Generate code that allows the data segment to be located in a different
6045 area of memory from the text segment. This allows for execute in place in
6046 an environment without virtual memory management. This option implies -fPIC.
6049 Generate code that assumes that the data segment follows the text segment.
6050 This is the default.
6052 @item -mid-shared-library
6053 Generate code that supports shared libraries via the library ID method.
6054 This allows for execute in place and shared libraries in an environment
6055 without virtual memory management. This option implies -fPIC.
6057 @item -mno-id-shared-library
6058 Generate code that doesn't assume ID based shared libraries are being used.
6059 This is the default.
6061 @item -mshared-library-id=n
6062 Specified the identification number of the ID based shared library being
6063 compiled. Specifying a value of 0 will generate more compact code, specifying
6064 other values will force the allocation of that number to the current
6065 library but is no more space or time efficient than omitting this option.
6069 @node M68hc1x Options
6070 @subsection M68hc1x Options
6071 @cindex M68hc1x options
6073 These are the @samp{-m} options defined for the 68hc11 and 68hc12
6074 microcontrollers. The default values for these options depends on
6075 which style of microcontroller was selected when the compiler was configured;
6076 the defaults for the most common choices are given below.
6083 Generate output for a 68HC11. This is the default
6084 when the compiler is configured for 68HC11-based systems.
6090 Generate output for a 68HC12. This is the default
6091 when the compiler is configured for 68HC12-based systems.
6097 Generate output for a 68HCS12.
6100 @opindex mauto-incdec
6101 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
6108 Enable the use of 68HC12 min and max instructions.
6111 @itemx -mno-long-calls
6112 @opindex mlong-calls
6113 @opindex mno-long-calls
6114 Treat all calls as being far away (near). If calls are assumed to be
6115 far away, the compiler will use the @code{call} instruction to
6116 call a function and the @code{rtc} instruction for returning.
6120 Consider type @code{int} to be 16 bits wide, like @code{short int}.
6122 @item -msoft-reg-count=@var{count}
6123 @opindex msoft-reg-count
6124 Specify the number of pseudo-soft registers which are used for the
6125 code generation. The maximum number is 32. Using more pseudo-soft
6126 register may or may not result in better code depending on the program.
6127 The default is 4 for 68HC11 and 2 for 68HC12.
6132 @subsection VAX Options
6135 These @samp{-m} options are defined for the VAX:
6140 Do not output certain jump instructions (@code{aobleq} and so on)
6141 that the Unix assembler for the VAX cannot handle across long
6146 Do output those jump instructions, on the assumption that you
6147 will assemble with the GNU assembler.
6151 Output code for g-format floating point numbers instead of d-format.
6155 @subsection SPARC Options
6156 @cindex SPARC options
6158 These @samp{-m} switches are supported on the SPARC:
6163 @opindex mno-app-regs
6165 Specify @option{-mapp-regs} to generate output using the global registers
6166 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
6169 To be fully SVR4 ABI compliant at the cost of some performance loss,
6170 specify @option{-mno-app-regs}. You should compile libraries and system
6171 software with this option.
6176 @opindex mhard-float
6177 Generate output containing floating point instructions. This is the
6183 @opindex msoft-float
6184 Generate output containing library calls for floating point.
6185 @strong{Warning:} the requisite libraries are not available for all SPARC
6186 targets. Normally the facilities of the machine's usual C compiler are
6187 used, but this cannot be done directly in cross-compilation. You must make
6188 your own arrangements to provide suitable library functions for
6189 cross-compilation. The embedded targets @samp{sparc-*-aout} and
6190 @samp{sparclite-*-*} do provide software floating point support.
6192 @option{-msoft-float} changes the calling convention in the output file;
6193 therefore, it is only useful if you compile @emph{all} of a program with
6194 this option. In particular, you need to compile @file{libgcc.a}, the
6195 library that comes with GCC, with @option{-msoft-float} in order for
6198 @item -mhard-quad-float
6199 @opindex mhard-quad-float
6200 Generate output containing quad-word (long double) floating point
6203 @item -msoft-quad-float
6204 @opindex msoft-quad-float
6205 Generate output containing library calls for quad-word (long double)
6206 floating point instructions. The functions called are those specified
6207 in the SPARC ABI@. This is the default.
6209 As of this writing, there are no sparc implementations that have hardware
6210 support for the quad-word floating point instructions. They all invoke
6211 a trap handler for one of these instructions, and then the trap handler
6212 emulates the effect of the instruction. Because of the trap handler overhead,
6213 this is much slower than calling the ABI library routines. Thus the
6214 @option{-msoft-quad-float} option is the default.
6220 With @option{-mflat}, the compiler does not generate save/restore instructions
6221 and will use a ``flat'' or single register window calling convention.
6222 This model uses %i7 as the frame pointer and is compatible with the normal
6223 register window model. Code from either may be intermixed.
6224 The local registers and the input registers (0--5) are still treated as
6225 ``call saved'' registers and will be saved on the stack as necessary.
6227 With @option{-mno-flat} (the default), the compiler emits save/restore
6228 instructions (except for leaf functions) and is the normal mode of operation.
6230 @item -mno-unaligned-doubles
6231 @itemx -munaligned-doubles
6232 @opindex mno-unaligned-doubles
6233 @opindex munaligned-doubles
6234 Assume that doubles have 8 byte alignment. This is the default.
6236 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
6237 alignment only if they are contained in another type, or if they have an
6238 absolute address. Otherwise, it assumes they have 4 byte alignment.
6239 Specifying this option avoids some rare compatibility problems with code
6240 generated by other compilers. It is not the default because it results
6241 in a performance loss, especially for floating point code.
6243 @item -mno-faster-structs
6244 @itemx -mfaster-structs
6245 @opindex mno-faster-structs
6246 @opindex mfaster-structs
6247 With @option{-mfaster-structs}, the compiler assumes that structures
6248 should have 8 byte alignment. This enables the use of pairs of
6249 @code{ldd} and @code{std} instructions for copies in structure
6250 assignment, in place of twice as many @code{ld} and @code{st} pairs.
6251 However, the use of this changed alignment directly violates the SPARC
6252 ABI@. Thus, it's intended only for use on targets where the developer
6253 acknowledges that their resulting code will not be directly in line with
6254 the rules of the ABI@.
6257 @opindex mimpure-text
6258 @option{-mimpure-text}, used in addition to @option{-shared}, tells
6259 the compiler to not pass @option{-z text} to the linker when linking a
6260 shared object. Using this option, you can link position-dependent
6261 code into a shared object.
6263 @option{-mimpure-text} suppresses the ``relocations remain against
6264 allocatable but non-writable sections'' linker error message.
6265 However, the necessary relocations will trigger copy-on-write, and the
6266 shared object is not actually shared across processes. Instead of
6267 using @option{-mimpure-text}, you should compile all source code with
6268 @option{-fpic} or @option{-fPIC}.
6270 This option is only available on SunOS and Solaris.
6276 These two options select variations on the SPARC architecture.
6278 By default (unless specifically configured for the Fujitsu SPARClite),
6279 GCC generates code for the v7 variant of the SPARC architecture.
6281 @option{-mv8} will give you SPARC v8 code. The only difference from v7
6282 code is that the compiler emits the integer multiply and integer
6283 divide instructions which exist in SPARC v8 but not in SPARC v7.
6285 @option{-msparclite} will give you SPARClite code. This adds the integer
6286 multiply, integer divide step and scan (@code{ffs}) instructions which
6287 exist in SPARClite but not in SPARC v7.
6289 These options are deprecated and will be deleted in a future GCC release.
6290 They have been replaced with @option{-mcpu=xxx}.
6295 @opindex msupersparc
6296 These two options select the processor for which the code is optimized.
6298 With @option{-mcypress} (the default), the compiler optimizes code for the
6299 Cypress CY7C602 chip, as used in the SPARCStation/SPARCServer 3xx series.
6300 This is also appropriate for the older SPARCStation 1, 2, IPX etc.
6302 With @option{-msupersparc} the compiler optimizes code for the SuperSPARC cpu, as
6303 used in the SPARCStation 10, 1000 and 2000 series. This flag also enables use
6304 of the full SPARC v8 instruction set.
6306 These options are deprecated and will be deleted in a future GCC release.
6307 They have been replaced with @option{-mcpu=xxx}.
6309 @item -mcpu=@var{cpu_type}
6311 Set the instruction set, register set, and instruction scheduling parameters
6312 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
6313 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
6314 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
6315 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
6318 Default instruction scheduling parameters are used for values that select
6319 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
6320 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
6322 Here is a list of each supported architecture and their supported
6327 v8: supersparc, hypersparc
6328 sparclite: f930, f934, sparclite86x
6330 v9: ultrasparc, ultrasparc3
6333 @item -mtune=@var{cpu_type}
6335 Set the instruction scheduling parameters for machine type
6336 @var{cpu_type}, but do not set the instruction set or register set that the
6337 option @option{-mcpu=@var{cpu_type}} would.
6339 The same values for @option{-mcpu=@var{cpu_type}} can be used for
6340 @option{-mtune=@var{cpu_type}}, but the only useful values are those
6341 that select a particular cpu implementation. Those are @samp{cypress},
6342 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
6343 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
6348 These @samp{-m} switches are supported in addition to the above
6349 on the SPARCLET processor.
6352 @item -mlittle-endian
6353 @opindex mlittle-endian
6354 Generate code for a processor running in little-endian mode.
6358 Treat register @code{%g0} as a normal register.
6359 GCC will continue to clobber it as necessary but will not assume
6360 it always reads as 0.
6362 @item -mbroken-saverestore
6363 @opindex mbroken-saverestore
6364 Generate code that does not use non-trivial forms of the @code{save} and
6365 @code{restore} instructions. Early versions of the SPARCLET processor do
6366 not correctly handle @code{save} and @code{restore} instructions used with
6367 arguments. They correctly handle them used without arguments. A @code{save}
6368 instruction used without arguments increments the current window pointer
6369 but does not allocate a new stack frame. It is assumed that the window
6370 overflow trap handler will properly handle this case as will interrupt
6374 These @samp{-m} switches are supported in addition to the above
6375 on SPARC V9 processors in 64-bit environments.
6378 @item -mlittle-endian
6379 @opindex mlittle-endian
6380 Generate code for a processor running in little-endian mode. It is only
6381 available for a few configurations and most notably not on Solaris.
6387 Generate code for a 32-bit or 64-bit environment.
6388 The 32-bit environment sets int, long and pointer to 32 bits.
6389 The 64-bit environment sets int to 32 bits and long and pointer
6392 @item -mcmodel=medlow
6393 @opindex mcmodel=medlow
6394 Generate code for the Medium/Low code model: the program must be linked
6395 in the low 32 bits of the address space. Pointers are 64 bits.
6396 Programs can be statically or dynamically linked.
6398 @item -mcmodel=medmid
6399 @opindex mcmodel=medmid
6400 Generate code for the Medium/Middle code model: the program must be linked
6401 in the low 44 bits of the address space, the text segment must be less than
6402 2G bytes, and data segment must be within 2G of the text segment.
6403 Pointers are 64 bits.
6405 @item -mcmodel=medany
6406 @opindex mcmodel=medany
6407 Generate code for the Medium/Anywhere code model: the program may be linked
6408 anywhere in the address space, the text segment must be less than
6409 2G bytes, and data segment must be within 2G of the text segment.
6410 Pointers are 64 bits.
6412 @item -mcmodel=embmedany
6413 @opindex mcmodel=embmedany
6414 Generate code for the Medium/Anywhere code model for embedded systems:
6415 assume a 32-bit text and a 32-bit data segment, both starting anywhere
6416 (determined at link time). Register %g4 points to the base of the
6417 data segment. Pointers are still 64 bits.
6418 Programs are statically linked, PIC is not supported.
6421 @itemx -mno-stack-bias
6422 @opindex mstack-bias
6423 @opindex mno-stack-bias
6424 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
6425 frame pointer if present, are offset by @minus{}2047 which must be added back
6426 when making stack frame references.
6427 Otherwise, assume no such offset is present.
6431 @subsection ARM Options
6434 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6439 @opindex mapcs-frame
6440 Generate a stack frame that is compliant with the ARM Procedure Call
6441 Standard for all functions, even if this is not strictly necessary for
6442 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6443 with this option will cause the stack frames not to be generated for
6444 leaf functions. The default is @option{-mno-apcs-frame}.
6448 This is a synonym for @option{-mapcs-frame}.
6452 Generate code for a processor running with a 26-bit program counter,
6453 and conforming to the function calling standards for the APCS 26-bit
6454 option. This option replaces the @option{-m2} and @option{-m3} options
6455 of previous releases of the compiler.
6459 Generate code for a processor running with a 32-bit program counter,
6460 and conforming to the function calling standards for the APCS 32-bit
6461 option. This option replaces the @option{-m6} option of previous releases
6465 @c not currently implemented
6466 @item -mapcs-stack-check
6467 @opindex mapcs-stack-check
6468 Generate code to check the amount of stack space available upon entry to
6469 every function (that actually uses some stack space). If there is
6470 insufficient space available then either the function
6471 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6472 called, depending upon the amount of stack space required. The run time
6473 system is required to provide these functions. The default is
6474 @option{-mno-apcs-stack-check}, since this produces smaller code.
6476 @c not currently implemented
6478 @opindex mapcs-float
6479 Pass floating point arguments using the float point registers. This is
6480 one of the variants of the APCS@. This option is recommended if the
6481 target hardware has a floating point unit or if a lot of floating point
6482 arithmetic is going to be performed by the code. The default is
6483 @option{-mno-apcs-float}, since integer only code is slightly increased in
6484 size if @option{-mapcs-float} is used.
6486 @c not currently implemented
6487 @item -mapcs-reentrant
6488 @opindex mapcs-reentrant
6489 Generate reentrant, position independent code. The default is
6490 @option{-mno-apcs-reentrant}.
6493 @item -mthumb-interwork
6494 @opindex mthumb-interwork
6495 Generate code which supports calling between the ARM and Thumb
6496 instruction sets. Without this option the two instruction sets cannot
6497 be reliably used inside one program. The default is
6498 @option{-mno-thumb-interwork}, since slightly larger code is generated
6499 when @option{-mthumb-interwork} is specified.
6501 @item -mno-sched-prolog
6502 @opindex mno-sched-prolog
6503 Prevent the reordering of instructions in the function prolog, or the
6504 merging of those instruction with the instructions in the function's
6505 body. This means that all functions will start with a recognizable set
6506 of instructions (or in fact one of a choice from a small set of
6507 different function prologues), and this information can be used to
6508 locate the start if functions inside an executable piece of code. The
6509 default is @option{-msched-prolog}.
6512 @opindex mhard-float
6513 Generate output containing floating point instructions. This is the
6517 @opindex msoft-float
6518 Generate output containing library calls for floating point.
6519 @strong{Warning:} the requisite libraries are not available for all ARM
6520 targets. Normally the facilities of the machine's usual C compiler are
6521 used, but this cannot be done directly in cross-compilation. You must make
6522 your own arrangements to provide suitable library functions for
6525 @option{-msoft-float} changes the calling convention in the output file;
6526 therefore, it is only useful if you compile @emph{all} of a program with
6527 this option. In particular, you need to compile @file{libgcc.a}, the
6528 library that comes with GCC, with @option{-msoft-float} in order for
6531 @item -mlittle-endian
6532 @opindex mlittle-endian
6533 Generate code for a processor running in little-endian mode. This is
6534 the default for all standard configurations.
6537 @opindex mbig-endian
6538 Generate code for a processor running in big-endian mode; the default is
6539 to compile code for a little-endian processor.
6541 @item -mwords-little-endian
6542 @opindex mwords-little-endian
6543 This option only applies when generating code for big-endian processors.
6544 Generate code for a little-endian word order but a big-endian byte
6545 order. That is, a byte order of the form @samp{32107654}. Note: this
6546 option should only be used if you require compatibility with code for
6547 big-endian ARM processors generated by versions of the compiler prior to
6550 @item -malignment-traps
6551 @opindex malignment-traps
6552 Generate code that will not trap if the MMU has alignment traps enabled.
6553 On ARM architectures prior to ARMv4, there were no instructions to
6554 access half-word objects stored in memory. However, when reading from
6555 memory a feature of the ARM architecture allows a word load to be used,
6556 even if the address is unaligned, and the processor core will rotate the
6557 data as it is being loaded. This option tells the compiler that such
6558 misaligned accesses will cause a MMU trap and that it should instead
6559 synthesize the access as a series of byte accesses. The compiler can
6560 still use word accesses to load half-word data if it knows that the
6561 address is aligned to a word boundary.
6563 This option is ignored when compiling for ARM architecture 4 or later,
6564 since these processors have instructions to directly access half-word
6567 @item -mno-alignment-traps
6568 @opindex mno-alignment-traps
6569 Generate code that assumes that the MMU will not trap unaligned
6570 accesses. This produces better code when the target instruction set
6571 does not have half-word memory operations (i.e.@: implementations prior to
6574 Note that you cannot use this option to access unaligned word objects,
6575 since the processor will only fetch one 32-bit aligned object from
6578 The default setting for most targets is @option{-mno-alignment-traps}, since
6579 this produces better code when there are no half-word memory
6580 instructions available.
6582 @item -mshort-load-bytes
6583 @itemx -mno-short-load-words
6584 @opindex mshort-load-bytes
6585 @opindex mno-short-load-words
6586 These are deprecated aliases for @option{-malignment-traps}.
6588 @item -mno-short-load-bytes
6589 @itemx -mshort-load-words
6590 @opindex mno-short-load-bytes
6591 @opindex mshort-load-words
6592 This are deprecated aliases for @option{-mno-alignment-traps}.
6594 @item -mcpu=@var{name}
6596 This specifies the name of the target ARM processor. GCC uses this name
6597 to determine what kind of instructions it can emit when generating
6598 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6599 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6600 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6601 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6602 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6603 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm8},
6604 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6605 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6606 @samp{arm920t}, @samp{arm926ejs}, @samp{arm940t}, @samp{arm9tdmi},
6607 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ejs},
6608 @samp{arm1136js}, @samp{arm1136jfs} ,@samp{xscale}, @samp{iwmmxt},
6611 @itemx -mtune=@var{name}
6613 This option is very similar to the @option{-mcpu=} option, except that
6614 instead of specifying the actual target processor type, and hence
6615 restricting which instructions can be used, it specifies that GCC should
6616 tune the performance of the code as if the target were of the type
6617 specified in this option, but still choosing the instructions that it
6618 will generate based on the cpu specified by a @option{-mcpu=} option.
6619 For some ARM implementations better performance can be obtained by using
6622 @item -march=@var{name}
6624 This specifies the name of the target ARM architecture. GCC uses this
6625 name to determine what kind of instructions it can emit when generating
6626 assembly code. This option can be used in conjunction with or instead
6627 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6628 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6629 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6j},
6630 @samp{iwmmxt}, @samp{ep9312}.
6632 @item -mfpe=@var{number}
6633 @itemx -mfp=@var{number}
6636 This specifies the version of the floating point emulation available on
6637 the target. Permissible values are 2 and 3. @option{-mfp=} is a synonym
6638 for @option{-mfpe=}, for compatibility with older versions of GCC@.
6640 @item -mstructure-size-boundary=@var{n}
6641 @opindex mstructure-size-boundary
6642 The size of all structures and unions will be rounded up to a multiple
6643 of the number of bits set by this option. Permissible values are 8 and
6644 32. The default value varies for different toolchains. For the COFF
6645 targeted toolchain the default value is 8. Specifying the larger number
6646 can produce faster, more efficient code, but can also increase the size
6647 of the program. The two values are potentially incompatible. Code
6648 compiled with one value cannot necessarily expect to work with code or
6649 libraries compiled with the other value, if they exchange information
6650 using structures or unions.
6652 @item -mabort-on-noreturn
6653 @opindex mabort-on-noreturn
6654 Generate a call to the function @code{abort} at the end of a
6655 @code{noreturn} function. It will be executed if the function tries to
6659 @itemx -mno-long-calls
6660 @opindex mlong-calls
6661 @opindex mno-long-calls
6662 Tells the compiler to perform function calls by first loading the
6663 address of the function into a register and then performing a subroutine
6664 call on this register. This switch is needed if the target function
6665 will lie outside of the 64 megabyte addressing range of the offset based
6666 version of subroutine call instruction.
6668 Even if this switch is enabled, not all function calls will be turned
6669 into long calls. The heuristic is that static functions, functions
6670 which have the @samp{short-call} attribute, functions that are inside
6671 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6672 definitions have already been compiled within the current compilation
6673 unit, will not be turned into long calls. The exception to this rule is
6674 that weak function definitions, functions with the @samp{long-call}
6675 attribute or the @samp{section} attribute, and functions that are within
6676 the scope of a @samp{#pragma long_calls} directive, will always be
6677 turned into long calls.
6679 This feature is not enabled by default. Specifying
6680 @option{-mno-long-calls} will restore the default behavior, as will
6681 placing the function calls within the scope of a @samp{#pragma
6682 long_calls_off} directive. Note these switches have no effect on how
6683 the compiler generates code to handle function calls via function
6686 @item -mnop-fun-dllimport
6687 @opindex mnop-fun-dllimport
6688 Disable support for the @code{dllimport} attribute.
6690 @item -msingle-pic-base
6691 @opindex msingle-pic-base
6692 Treat the register used for PIC addressing as read-only, rather than
6693 loading it in the prologue for each function. The run-time system is
6694 responsible for initializing this register with an appropriate value
6695 before execution begins.
6697 @item -mpic-register=@var{reg}
6698 @opindex mpic-register
6699 Specify the register to be used for PIC addressing. The default is R10
6700 unless stack-checking is enabled, when R9 is used.
6702 @item -mcirrus-fix-invalid-insns
6703 @opindex mcirrus-fix-invalid-insns
6704 @opindex mno-cirrus-fix-invalid-insns
6705 Insert NOPs into the instruction stream to in order to work around
6706 problems with invalid Maverick instruction combinations. This option
6707 is only valid if the @option{-mcpu=ep9312} option has been used to
6708 enable generation of instructions for the Cirrus Maverick floating
6709 point co-processor. This option is not enabled by default, since the
6710 problem is only present in older Maverick implementations. The default
6711 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
6714 @item -mpoke-function-name
6715 @opindex mpoke-function-name
6716 Write the name of each function into the text section, directly
6717 preceding the function prologue. The generated code is similar to this:
6721 .ascii "arm_poke_function_name", 0
6724 .word 0xff000000 + (t1 - t0)
6725 arm_poke_function_name
6727 stmfd sp!, @{fp, ip, lr, pc@}
6731 When performing a stack backtrace, code can inspect the value of
6732 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6733 location @code{pc - 12} and the top 8 bits are set, then we know that
6734 there is a function name embedded immediately preceding this location
6735 and has length @code{((pc[-3]) & 0xff000000)}.
6739 Generate code for the 16-bit Thumb instruction set. The default is to
6740 use the 32-bit ARM instruction set.
6743 @opindex mtpcs-frame
6744 Generate a stack frame that is compliant with the Thumb Procedure Call
6745 Standard for all non-leaf functions. (A leaf function is one that does
6746 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6748 @item -mtpcs-leaf-frame
6749 @opindex mtpcs-leaf-frame
6750 Generate a stack frame that is compliant with the Thumb Procedure Call
6751 Standard for all leaf functions. (A leaf function is one that does
6752 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6754 @item -mcallee-super-interworking
6755 @opindex mcallee-super-interworking
6756 Gives all externally visible functions in the file being compiled an ARM
6757 instruction set header which switches to Thumb mode before executing the
6758 rest of the function. This allows these functions to be called from
6759 non-interworking code.
6761 @item -mcaller-super-interworking
6762 @opindex mcaller-super-interworking
6763 Allows calls via function pointers (including virtual functions) to
6764 execute correctly regardless of whether the target code has been
6765 compiled for interworking or not. There is a small overhead in the cost
6766 of executing a function pointer if this option is enabled.
6770 @node MN10200 Options
6771 @subsection MN10200 Options
6772 @cindex MN10200 options
6774 These @option{-m} options are defined for Matsushita MN10200 architectures:
6779 Indicate to the linker that it should perform a relaxation optimization pass
6780 to shorten branches, calls and absolute memory addresses. This option only
6781 has an effect when used on the command line for the final link step.
6783 This option makes symbolic debugging impossible.
6786 @node MN10300 Options
6787 @subsection MN10300 Options
6788 @cindex MN10300 options
6790 These @option{-m} options are defined for Matsushita MN10300 architectures:
6795 Generate code to avoid bugs in the multiply instructions for the MN10300
6796 processors. This is the default.
6799 @opindex mno-mult-bug
6800 Do not generate code to avoid bugs in the multiply instructions for the
6805 Generate code which uses features specific to the AM33 processor.
6809 Do not generate code which uses features specific to the AM33 processor. This
6814 Do not link in the C run-time initialization object file.
6818 Indicate to the linker that it should perform a relaxation optimization pass
6819 to shorten branches, calls and absolute memory addresses. This option only
6820 has an effect when used on the command line for the final link step.
6822 This option makes symbolic debugging impossible.
6826 @node M32R/D Options
6827 @subsection M32R/D Options
6828 @cindex M32R/D options
6830 These @option{-m} options are defined for Renesas M32R/D architectures:
6835 Generate code for the M32R/2@.
6839 Generate code for the M32R/X@.
6843 Generate code for the M32R@. This is the default.
6846 @opindex mmodel=small
6847 Assume all objects live in the lower 16MB of memory (so that their addresses
6848 can be loaded with the @code{ld24} instruction), and assume all subroutines
6849 are reachable with the @code{bl} instruction.
6850 This is the default.
6852 The addressability of a particular object can be set with the
6853 @code{model} attribute.
6855 @item -mmodel=medium
6856 @opindex mmodel=medium
6857 Assume objects may be anywhere in the 32-bit address space (the compiler
6858 will generate @code{seth/add3} instructions to load their addresses), and
6859 assume all subroutines are reachable with the @code{bl} instruction.
6862 @opindex mmodel=large
6863 Assume objects may be anywhere in the 32-bit address space (the compiler
6864 will generate @code{seth/add3} instructions to load their addresses), and
6865 assume subroutines may not be reachable with the @code{bl} instruction
6866 (the compiler will generate the much slower @code{seth/add3/jl}
6867 instruction sequence).
6870 @opindex msdata=none
6871 Disable use of the small data area. Variables will be put into
6872 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
6873 @code{section} attribute has been specified).
6874 This is the default.
6876 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
6877 Objects may be explicitly put in the small data area with the
6878 @code{section} attribute using one of these sections.
6881 @opindex msdata=sdata
6882 Put small global and static data in the small data area, but do not
6883 generate special code to reference them.
6887 Put small global and static data in the small data area, and generate
6888 special instructions to reference them.
6892 @cindex smaller data references
6893 Put global and static objects less than or equal to @var{num} bytes
6894 into the small data or bss sections instead of the normal data or bss
6895 sections. The default value of @var{num} is 8.
6896 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
6897 for this option to have any effect.
6899 All modules should be compiled with the same @option{-G @var{num}} value.
6900 Compiling with different values of @var{num} may or may not work; if it
6901 doesn't the linker will give an error message---incorrect code will not be
6906 Makes the M32R specific code in the compiler display some statistics
6907 that might help in debugging programs.
6910 @opindex malign-loops
6911 Align all loops to a 32-byte boundary.
6913 @item -mno-align-loops
6914 @opindex mno-align-loops
6915 Do not enforce a 32-byte alignment for loops. This is the default.
6917 @item -missue-rate=@var{number}
6918 @opindex missue-rate=@var{number}
6919 Issue @var{number} instructions per cycle. @var{number} can only be 1
6922 @item -mbranch-cost=@var{number}
6923 @opindex mbranch-cost=@var{number}
6924 @var{number} can only be 1 or 2. If it is 1 then branches will be
6925 prefered over conditional code, if it is 2, then the opposite will
6928 @item -mflush-trap=@var{number}
6929 @opindex mflush-trap=@var{number}
6930 Specifies the trap number to use to flush the cache. The default is
6931 12. Valid numbers are between 0 and 15 inclusive.
6933 @item -mno-flush-trap
6934 @opindex mno-flush-trap
6935 Specifies that the cache cannot be flushed by using a trap.
6937 @item -mflush-func=@var{name}
6938 @opindex mflush-func=@var{name}
6939 Specifies the name of the operating system function to call to flush
6940 the cache. The default is @emph{_flush_cache}, but a function call
6941 will only be used if a trap is not available.
6943 @item -mno-flush-func
6944 @opindex mno-flush-func
6945 Indicates that there is no OS function for flushing the cache.
6950 @subsection M88K Options
6951 @cindex M88k options
6953 These @samp{-m} options are defined for Motorola 88k architectures:
6958 Generate code that works well on both the m88100 and the
6963 Generate code that works best for the m88100, but that also
6968 Generate code that works best for the m88110, and may not run
6973 Obsolete option to be removed from the next revision.
6976 @item -midentify-revision
6977 @opindex midentify-revision
6978 @cindex identifying source, compiler (88k)
6979 Include an @code{ident} directive in the assembler output recording the
6980 source file name, compiler name and version, timestamp, and compilation
6983 @item -mno-underscores
6984 @opindex mno-underscores
6985 @cindex underscores, avoiding (88k)
6986 In assembler output, emit symbol names without adding an underscore
6987 character at the beginning of each name. The default is to use an
6988 underscore as prefix on each name.
6990 @item -mocs-debug-info
6991 @itemx -mno-ocs-debug-info
6992 @opindex mocs-debug-info
6993 @opindex mno-ocs-debug-info
6995 @cindex debugging, 88k OCS
6996 Include (or omit) additional debugging information (about registers used
6997 in each stack frame) as specified in the 88open Object Compatibility
6998 Standard, ``OCS''@. This extra information allows debugging of code that
6999 has had the frame pointer eliminated. The default for SVr4 and Delta 88
7000 SVr3.2 is to include this information; other 88k configurations omit this
7001 information by default.
7003 @item -mocs-frame-position
7004 @opindex mocs-frame-position
7005 @cindex register positions in frame (88k)
7006 When emitting COFF debugging information for automatic variables and
7007 parameters stored on the stack, use the offset from the canonical frame
7008 address, which is the stack pointer (register 31) on entry to the
7009 function. The SVr4 and Delta88 SVr3.2, and BCS configurations use
7010 @option{-mocs-frame-position}; other 88k configurations have the default
7011 @option{-mno-ocs-frame-position}.
7013 @item -mno-ocs-frame-position
7014 @opindex mno-ocs-frame-position
7015 @cindex register positions in frame (88k)
7016 When emitting COFF debugging information for automatic variables and
7017 parameters stored on the stack, use the offset from the frame pointer
7018 register (register 30). When this option is in effect, the frame
7019 pointer is not eliminated when debugging information is selected by the
7022 @item -moptimize-arg-area
7023 @opindex moptimize-arg-area
7024 @cindex arguments in frame (88k)
7025 Save space by reorganizing the stack frame. This option generates code
7026 that does not agree with the 88open specifications, but uses less
7029 @itemx -mno-optimize-arg-area
7030 @opindex mno-optimize-arg-area
7031 Do not reorganize the stack frame to save space. This is the default.
7032 The generated conforms to the specification, but uses more memory.
7034 @item -mshort-data-@var{num}
7035 @opindex mshort-data
7036 @cindex smaller data references (88k)
7037 @cindex r0-relative references (88k)
7038 Generate smaller data references by making them relative to @code{r0},
7039 which allows loading a value using a single instruction (rather than the
7040 usual two). You control which data references are affected by
7041 specifying @var{num} with this option. For example, if you specify
7042 @option{-mshort-data-512}, then the data references affected are those
7043 involving displacements of less than 512 bytes.
7044 @option{-mshort-data-@var{num}} is not effective for @var{num} greater
7047 @item -mserialize-volatile
7048 @opindex mserialize-volatile
7049 @itemx -mno-serialize-volatile
7050 @opindex mno-serialize-volatile
7051 @cindex sequential consistency on 88k
7052 Do, or don't, generate code to guarantee sequential consistency
7053 of volatile memory references. By default, consistency is
7056 The order of memory references made by the MC88110 processor does
7057 not always match the order of the instructions requesting those
7058 references. In particular, a load instruction may execute before
7059 a preceding store instruction. Such reordering violates
7060 sequential consistency of volatile memory references, when there
7061 are multiple processors. When consistency must be guaranteed,
7062 GCC generates special instructions, as needed, to force
7063 execution in the proper order.
7065 The MC88100 processor does not reorder memory references and so
7066 always provides sequential consistency. However, by default, GCC
7067 generates the special instructions to guarantee consistency
7068 even when you use @option{-m88100}, so that the code may be run on an
7069 MC88110 processor. If you intend to run your code only on the
7070 MC88100 processor, you may use @option{-mno-serialize-volatile}.
7072 The extra code generated to guarantee consistency may affect the
7073 performance of your application. If you know that you can safely
7074 forgo this guarantee, you may use @option{-mno-serialize-volatile}.
7080 @cindex assembler syntax, 88k
7082 Turn on (@option{-msvr4}) or off (@option{-msvr3}) compiler extensions
7083 related to System V release 4 (SVr4). This controls the following:
7087 Which variant of the assembler syntax to emit.
7089 @option{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
7090 that is used on System V release 4.
7092 @option{-msvr4} makes GCC issue additional declaration directives used in
7096 @option{-msvr4} is the default for the m88k-motorola-sysv4 configuration.
7097 @option{-msvr3} is the default for all other m88k configurations.
7099 @item -mversion-03.00
7100 @opindex mversion-03.00
7101 This option is obsolete, and is ignored.
7102 @c ??? which asm syntax better for GAS? option there too?
7104 @item -mno-check-zero-division
7105 @itemx -mcheck-zero-division
7106 @opindex mno-check-zero-division
7107 @opindex mcheck-zero-division
7108 @cindex zero division on 88k
7109 Do, or don't, generate code to guarantee that integer division by
7110 zero will be detected. By default, detection is guaranteed.
7112 Some models of the MC88100 processor fail to trap upon integer
7113 division by zero under certain conditions. By default, when
7114 compiling code that might be run on such a processor, GCC
7115 generates code that explicitly checks for zero-valued divisors
7116 and traps with exception number 503 when one is detected. Use of
7117 @option{-mno-check-zero-division} suppresses such checking for code
7118 generated to run on an MC88100 processor.
7120 GCC assumes that the MC88110 processor correctly detects all instances
7121 of integer division by zero. When @option{-m88110} is specified, no
7122 explicit checks for zero-valued divisors are generated, and both
7123 @option{-mcheck-zero-division} and @option{-mno-check-zero-division} are
7126 @item -muse-div-instruction
7127 @opindex muse-div-instruction
7128 @cindex divide instruction, 88k
7129 Use the div instruction for signed integer division on the
7130 MC88100 processor. By default, the div instruction is not used.
7132 On the MC88100 processor the signed integer division instruction
7133 div) traps to the operating system on a negative operand. The
7134 operating system transparently completes the operation, but at a
7135 large cost in execution time. By default, when compiling code
7136 that might be run on an MC88100 processor, GCC emulates signed
7137 integer division using the unsigned integer division instruction
7138 divu), thereby avoiding the large penalty of a trap to the
7139 operating system. Such emulation has its own, smaller, execution
7140 cost in both time and space. To the extent that your code's
7141 important signed integer division operations are performed on two
7142 nonnegative operands, it may be desirable to use the div
7143 instruction directly.
7145 On the MC88110 processor the div instruction (also known as the
7146 divs instruction) processes negative operands without trapping to
7147 the operating system. When @option{-m88110} is specified,
7148 @option{-muse-div-instruction} is ignored, and the div instruction is used
7149 for signed integer division.
7151 Note that the result of dividing @code{INT_MIN} by @minus{}1 is undefined. In
7152 particular, the behavior of such a division with and without
7153 @option{-muse-div-instruction} may differ.
7155 @item -mtrap-large-shift
7156 @itemx -mhandle-large-shift
7157 @opindex mtrap-large-shift
7158 @opindex mhandle-large-shift
7159 @cindex bit shift overflow (88k)
7160 @cindex large bit shifts (88k)
7161 Include code to detect bit-shifts of more than 31 bits; respectively,
7162 trap such shifts or emit code to handle them properly. By default GCC
7163 makes no special provision for large bit shifts.
7165 @item -mwarn-passed-structs
7166 @opindex mwarn-passed-structs
7167 @cindex structure passing (88k)
7168 Warn when a function passes a struct as an argument or result.
7169 Structure-passing conventions have changed during the evolution of the C
7170 language, and are often the source of portability problems. By default,
7171 GCC issues no such warning.
7174 @c break page here to avoid unsightly interparagraph stretch.
7178 @node RS/6000 and PowerPC Options
7179 @subsection IBM RS/6000 and PowerPC Options
7180 @cindex RS/6000 and PowerPC Options
7181 @cindex IBM RS/6000 and PowerPC Options
7183 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
7191 @itemx -mpowerpc-gpopt
7192 @itemx -mno-powerpc-gpopt
7193 @itemx -mpowerpc-gfxopt
7194 @itemx -mno-powerpc-gfxopt
7196 @itemx -mno-powerpc64
7202 @opindex mno-powerpc
7203 @opindex mpowerpc-gpopt
7204 @opindex mno-powerpc-gpopt
7205 @opindex mpowerpc-gfxopt
7206 @opindex mno-powerpc-gfxopt
7208 @opindex mno-powerpc64
7209 GCC supports two related instruction set architectures for the
7210 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
7211 instructions supported by the @samp{rios} chip set used in the original
7212 RS/6000 systems and the @dfn{PowerPC} instruction set is the
7213 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
7214 the IBM 4xx microprocessors.
7216 Neither architecture is a subset of the other. However there is a
7217 large common subset of instructions supported by both. An MQ
7218 register is included in processors supporting the POWER architecture.
7220 You use these options to specify which instructions are available on the
7221 processor you are using. The default value of these options is
7222 determined when configuring GCC@. Specifying the
7223 @option{-mcpu=@var{cpu_type}} overrides the specification of these
7224 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
7225 rather than the options listed above.
7227 The @option{-mpower} option allows GCC to generate instructions that
7228 are found only in the POWER architecture and to use the MQ register.
7229 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
7230 to generate instructions that are present in the POWER2 architecture but
7231 not the original POWER architecture.
7233 The @option{-mpowerpc} option allows GCC to generate instructions that
7234 are found only in the 32-bit subset of the PowerPC architecture.
7235 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
7236 GCC to use the optional PowerPC architecture instructions in the
7237 General Purpose group, including floating-point square root. Specifying
7238 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
7239 use the optional PowerPC architecture instructions in the Graphics
7240 group, including floating-point select.
7242 The @option{-mpowerpc64} option allows GCC to generate the additional
7243 64-bit instructions that are found in the full PowerPC64 architecture
7244 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
7245 @option{-mno-powerpc64}.
7247 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
7248 will use only the instructions in the common subset of both
7249 architectures plus some special AIX common-mode calls, and will not use
7250 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
7251 permits GCC to use any instruction from either architecture and to
7252 allow use of the MQ register; specify this for the Motorola MPC601.
7254 @item -mnew-mnemonics
7255 @itemx -mold-mnemonics
7256 @opindex mnew-mnemonics
7257 @opindex mold-mnemonics
7258 Select which mnemonics to use in the generated assembler code. With
7259 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
7260 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
7261 assembler mnemonics defined for the POWER architecture. Instructions
7262 defined in only one architecture have only one mnemonic; GCC uses that
7263 mnemonic irrespective of which of these options is specified.
7265 GCC defaults to the mnemonics appropriate for the architecture in
7266 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
7267 value of these option. Unless you are building a cross-compiler, you
7268 should normally not specify either @option{-mnew-mnemonics} or
7269 @option{-mold-mnemonics}, but should instead accept the default.
7271 @item -mcpu=@var{cpu_type}
7273 Set architecture type, register usage, choice of mnemonics, and
7274 instruction scheduling parameters for machine type @var{cpu_type}.
7275 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
7276 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
7277 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
7278 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
7279 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
7280 @samp{860}, @samp{970}, @samp{common}, @samp{ec603e}, @samp{G3},
7281 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
7282 @samp{power4}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
7283 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
7285 @option{-mcpu=common} selects a completely generic processor. Code
7286 generated under this option will run on any POWER or PowerPC processor.
7287 GCC will use only the instructions in the common subset of both
7288 architectures, and will not use the MQ register. GCC assumes a generic
7289 processor model for scheduling purposes.
7291 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
7292 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
7293 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
7294 types, with an appropriate, generic processor model assumed for
7295 scheduling purposes.
7297 The other options specify a specific processor. Code generated under
7298 those options will run best on that processor, and may not run at all on
7301 The @option{-mcpu} options automatically enable or disable the
7302 following options: @option{-maltivec}, @option{-mhard-float},
7303 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
7304 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
7305 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
7306 @option{-mstring}. The particular options set for any particular CPU
7307 will vary between compiler versions, depending on what setting seems
7308 to produce optimal code for that CPU; it doesn't necessarily reflect
7309 the actual hardware's capabilities. If you wish to set an individual
7310 option to a particular value, you may specify it after the
7311 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
7313 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
7314 not enabled or disabled by the @option{-mcpu} option at present, since
7315 AIX does not have full support for these options. You may still
7316 enable or disable them individually if you're sure it'll work in your
7319 @item -mtune=@var{cpu_type}
7321 Set the instruction scheduling parameters for machine type
7322 @var{cpu_type}, but do not set the architecture type, register usage, or
7323 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
7324 values for @var{cpu_type} are used for @option{-mtune} as for
7325 @option{-mcpu}. If both are specified, the code generated will use the
7326 architecture, registers, and mnemonics set by @option{-mcpu}, but the
7327 scheduling parameters set by @option{-mtune}.
7332 @opindex mno-altivec
7333 These switches enable or disable the use of built-in functions that
7334 allow access to the AltiVec instruction set. You may also need to set
7335 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
7340 Extend the current ABI with SPE ABI extensions. This does not change
7341 the default ABI, instead it adds the SPE ABI extensions to the current
7345 @opindex mabi=no-spe
7346 Disable Booke SPE ABI extensions for the current ABI.
7348 @item -misel=@var{yes/no}
7351 This switch enables or disables the generation of ISEL instructions.
7353 @item -mspe=@var{yes/no}
7356 This switch enables or disables the generation of SPE simd
7359 @item -mfloat-gprs=@var{yes/no}
7361 @opindex mfloat-gprs
7362 This switch enables or disables the generation of floating point
7363 operations on the general purpose registers for architectures that
7364 support it. This option is currently only available on the MPC8540.
7367 @itemx -mno-fp-in-toc
7368 @itemx -mno-sum-in-toc
7369 @itemx -mminimal-toc
7371 @opindex mno-fp-in-toc
7372 @opindex mno-sum-in-toc
7373 @opindex mminimal-toc
7374 Modify generation of the TOC (Table Of Contents), which is created for
7375 every executable file. The @option{-mfull-toc} option is selected by
7376 default. In that case, GCC will allocate at least one TOC entry for
7377 each unique non-automatic variable reference in your program. GCC
7378 will also place floating-point constants in the TOC@. However, only
7379 16,384 entries are available in the TOC@.
7381 If you receive a linker error message that saying you have overflowed
7382 the available TOC space, you can reduce the amount of TOC space used
7383 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
7384 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
7385 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
7386 generate code to calculate the sum of an address and a constant at
7387 run-time instead of putting that sum into the TOC@. You may specify one
7388 or both of these options. Each causes GCC to produce very slightly
7389 slower and larger code at the expense of conserving TOC space.
7391 If you still run out of space in the TOC even when you specify both of
7392 these options, specify @option{-mminimal-toc} instead. This option causes
7393 GCC to make only one TOC entry for every file. When you specify this
7394 option, GCC will produce code that is slower and larger but which
7395 uses extremely little TOC space. You may wish to use this option
7396 only on files that contain less frequently executed code.
7402 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
7403 @code{long} type, and the infrastructure needed to support them.
7404 Specifying @option{-maix64} implies @option{-mpowerpc64} and
7405 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
7406 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
7411 @opindex mno-xl-call
7412 On AIX, pass floating-point arguments to prototyped functions beyond the
7413 register save area (RSA) on the stack in addition to argument FPRs. The
7414 AIX calling convention was extended but not initially documented to
7415 handle an obscure K&R C case of calling a function that takes the
7416 address of its arguments with fewer arguments than declared. AIX XL
7417 compilers access floating point arguments which do not fit in the
7418 RSA from the stack when a subroutine is compiled without
7419 optimization. Because always storing floating-point arguments on the
7420 stack is inefficient and rarely needed, this option is not enabled by
7421 default and only is necessary when calling subroutines compiled by AIX
7422 XL compilers without optimization.
7426 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
7427 application written to use message passing with special startup code to
7428 enable the application to run. The system must have PE installed in the
7429 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
7430 must be overridden with the @option{-specs=} option to specify the
7431 appropriate directory location. The Parallel Environment does not
7432 support threads, so the @option{-mpe} option and the @option{-pthread}
7433 option are incompatible.
7435 @item -malign-natural
7436 @itemx -malign-power
7437 @opindex malign-natural
7438 @opindex malign-power
7439 On AIX, Darwin, and 64-bit PowerPC Linux, the option
7440 @option{-malign-natural} overrides the ABI-defined alignment of larger
7441 types, such as floating-point doubles, on their natural size-based boundary.
7442 The option @option{-malign-power} instructs GCC to follow the ABI-specified
7443 alignment rules. GCC defaults to the standard alignment defined in the ABI.
7447 @opindex msoft-float
7448 @opindex mhard-float
7449 Generate code that does not use (uses) the floating-point register set.
7450 Software floating point emulation is provided if you use the
7451 @option{-msoft-float} option, and pass the option to GCC when linking.
7454 @itemx -mno-multiple
7456 @opindex mno-multiple
7457 Generate code that uses (does not use) the load multiple word
7458 instructions and the store multiple word instructions. These
7459 instructions are generated by default on POWER systems, and not
7460 generated on PowerPC systems. Do not use @option{-mmultiple} on little
7461 endian PowerPC systems, since those instructions do not work when the
7462 processor is in little endian mode. The exceptions are PPC740 and
7463 PPC750 which permit the instructions usage in little endian mode.
7469 Generate code that uses (does not use) the load string instructions
7470 and the store string word instructions to save multiple registers and
7471 do small block moves. These instructions are generated by default on
7472 POWER systems, and not generated on PowerPC systems. Do not use
7473 @option{-mstring} on little endian PowerPC systems, since those
7474 instructions do not work when the processor is in little endian mode.
7475 The exceptions are PPC740 and PPC750 which permit the instructions
7476 usage in little endian mode.
7482 Generate code that uses (does not use) the load or store instructions
7483 that update the base register to the address of the calculated memory
7484 location. These instructions are generated by default. If you use
7485 @option{-mno-update}, there is a small window between the time that the
7486 stack pointer is updated and the address of the previous frame is
7487 stored, which means code that walks the stack frame across interrupts or
7488 signals may get corrupted data.
7491 @itemx -mno-fused-madd
7492 @opindex mfused-madd
7493 @opindex mno-fused-madd
7494 Generate code that uses (does not use) the floating point multiply and
7495 accumulate instructions. These instructions are generated by default if
7496 hardware floating is used.
7498 @item -mno-bit-align
7500 @opindex mno-bit-align
7502 On System V.4 and embedded PowerPC systems do not (do) force structures
7503 and unions that contain bit-fields to be aligned to the base type of the
7506 For example, by default a structure containing nothing but 8
7507 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
7508 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
7509 the structure would be aligned to a 1 byte boundary and be one byte in
7512 @item -mno-strict-align
7513 @itemx -mstrict-align
7514 @opindex mno-strict-align
7515 @opindex mstrict-align
7516 On System V.4 and embedded PowerPC systems do not (do) assume that
7517 unaligned memory references will be handled by the system.
7520 @itemx -mno-relocatable
7521 @opindex mrelocatable
7522 @opindex mno-relocatable
7523 On embedded PowerPC systems generate code that allows (does not allow)
7524 the program to be relocated to a different address at runtime. If you
7525 use @option{-mrelocatable} on any module, all objects linked together must
7526 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
7528 @item -mrelocatable-lib
7529 @itemx -mno-relocatable-lib
7530 @opindex mrelocatable-lib
7531 @opindex mno-relocatable-lib
7532 On embedded PowerPC systems generate code that allows (does not allow)
7533 the program to be relocated to a different address at runtime. Modules
7534 compiled with @option{-mrelocatable-lib} can be linked with either modules
7535 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
7536 with modules compiled with the @option{-mrelocatable} options.
7542 On System V.4 and embedded PowerPC systems do not (do) assume that
7543 register 2 contains a pointer to a global area pointing to the addresses
7544 used in the program.
7547 @itemx -mlittle-endian
7549 @opindex mlittle-endian
7550 On System V.4 and embedded PowerPC systems compile code for the
7551 processor in little endian mode. The @option{-mlittle-endian} option is
7552 the same as @option{-mlittle}.
7557 @opindex mbig-endian
7558 On System V.4 and embedded PowerPC systems compile code for the
7559 processor in big endian mode. The @option{-mbig-endian} option is
7560 the same as @option{-mbig}.
7562 @item -mdynamic-no-pic
7563 @opindex mdynamic-no-pic
7564 On Darwin and Mac OS X systems, compile code so that it is not
7565 relocatable, but that its external references are relocatable. The
7566 resulting code is suitable for applications, but not shared
7569 @item -mprioritize-restricted-insns=@var{priority}
7570 @opindex mprioritize-restricted-insns
7571 This option controls the priority that is assigned to
7572 dispatch-slot restricted instructions during the second scheduling
7573 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
7574 @var{no/highest/second-highest} priority to dispatch slot restricted
7577 @item -msched-costly-dep=@var{dependence_type}
7578 @opindex msched-costly-dep
7579 This option controls which dependences are considered costly
7580 by the target during instruction scheduling. The argument
7581 @var{dependence_type} takes one of the following values:
7582 @var{no}: no dependence is costly,
7583 @var{all}: all dependences are costly,
7584 @var{true_store_to_load}: a true dependence from store to load is costly,
7585 @var{store_to_load}: any dependence from store to load is costly,
7586 @var{number}: any dependence which latency >= @var{number} is costly.
7588 @item -minsert-sched-nops=@var{scheme}
7589 @opindex minsert-sched-nops
7590 This option controls which nop insertion scheme will be used during
7591 the second scheduling pass. The argument @var{scheme} takes one of the
7593 @var{no}: Don't insert nops.
7594 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
7595 according to the scheduler's grouping.
7596 @var{regroup_exact}: Insert nops to force costly dependent insns into
7597 separate groups. Insert exactly as many nops as needed to force an insn
7598 to a new group, according to the estimatied processor grouping.
7599 @var{number}: Insert nops to force costly dependent insns into
7600 separate groups. Insert @var{number} nops to force an insn to a new group.
7604 On System V.4 and embedded PowerPC systems compile code using calling
7605 conventions that adheres to the March 1995 draft of the System V
7606 Application Binary Interface, PowerPC processor supplement. This is the
7607 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
7609 @item -mcall-sysv-eabi
7610 @opindex mcall-sysv-eabi
7611 Specify both @option{-mcall-sysv} and @option{-meabi} options.
7613 @item -mcall-sysv-noeabi
7614 @opindex mcall-sysv-noeabi
7615 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
7617 @item -mcall-solaris
7618 @opindex mcall-solaris
7619 On System V.4 and embedded PowerPC systems compile code for the Solaris
7623 @opindex mcall-linux
7624 On System V.4 and embedded PowerPC systems compile code for the
7625 Linux-based GNU system.
7629 On System V.4 and embedded PowerPC systems compile code for the
7630 Hurd-based GNU system.
7633 @opindex mcall-netbsd
7634 On System V.4 and embedded PowerPC systems compile code for the
7635 NetBSD operating system.
7637 @item -maix-struct-return
7638 @opindex maix-struct-return
7639 Return all structures in memory (as specified by the AIX ABI)@.
7641 @item -msvr4-struct-return
7642 @opindex msvr4-struct-return
7643 Return structures smaller than 8 bytes in registers (as specified by the
7647 @opindex mabi=altivec
7648 Extend the current ABI with AltiVec ABI extensions. This does not
7649 change the default ABI, instead it adds the AltiVec ABI extensions to
7652 @item -mabi=no-altivec
7653 @opindex mabi=no-altivec
7654 Disable AltiVec ABI extensions for the current ABI.
7657 @itemx -mno-prototype
7659 @opindex mno-prototype
7660 On System V.4 and embedded PowerPC systems assume that all calls to
7661 variable argument functions are properly prototyped. Otherwise, the
7662 compiler must insert an instruction before every non prototyped call to
7663 set or clear bit 6 of the condition code register (@var{CR}) to
7664 indicate whether floating point values were passed in the floating point
7665 registers in case the function takes a variable arguments. With
7666 @option{-mprototype}, only calls to prototyped variable argument functions
7667 will set or clear the bit.
7671 On embedded PowerPC systems, assume that the startup module is called
7672 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
7673 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
7678 On embedded PowerPC systems, assume that the startup module is called
7679 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
7684 On embedded PowerPC systems, assume that the startup module is called
7685 @file{crt0.o} and the standard C libraries are @file{libads.a} and
7689 @opindex myellowknife
7690 On embedded PowerPC systems, assume that the startup module is called
7691 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
7696 On System V.4 and embedded PowerPC systems, specify that you are
7697 compiling for a VxWorks system.
7701 Specify that you are compiling for the WindISS simulation environment.
7705 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
7706 header to indicate that @samp{eabi} extended relocations are used.
7712 On System V.4 and embedded PowerPC systems do (do not) adhere to the
7713 Embedded Applications Binary Interface (eabi) which is a set of
7714 modifications to the System V.4 specifications. Selecting @option{-meabi}
7715 means that the stack is aligned to an 8 byte boundary, a function
7716 @code{__eabi} is called to from @code{main} to set up the eabi
7717 environment, and the @option{-msdata} option can use both @code{r2} and
7718 @code{r13} to point to two separate small data areas. Selecting
7719 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
7720 do not call an initialization function from @code{main}, and the
7721 @option{-msdata} option will only use @code{r13} to point to a single
7722 small data area. The @option{-meabi} option is on by default if you
7723 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
7726 @opindex msdata=eabi
7727 On System V.4 and embedded PowerPC systems, put small initialized
7728 @code{const} global and static data in the @samp{.sdata2} section, which
7729 is pointed to by register @code{r2}. Put small initialized
7730 non-@code{const} global and static data in the @samp{.sdata} section,
7731 which is pointed to by register @code{r13}. Put small uninitialized
7732 global and static data in the @samp{.sbss} section, which is adjacent to
7733 the @samp{.sdata} section. The @option{-msdata=eabi} option is
7734 incompatible with the @option{-mrelocatable} option. The
7735 @option{-msdata=eabi} option also sets the @option{-memb} option.
7738 @opindex msdata=sysv
7739 On System V.4 and embedded PowerPC systems, put small global and static
7740 data in the @samp{.sdata} section, which is pointed to by register
7741 @code{r13}. Put small uninitialized global and static data in the
7742 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
7743 The @option{-msdata=sysv} option is incompatible with the
7744 @option{-mrelocatable} option.
7746 @item -msdata=default
7748 @opindex msdata=default
7750 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
7751 compile code the same as @option{-msdata=eabi}, otherwise compile code the
7752 same as @option{-msdata=sysv}.
7755 @opindex msdata-data
7756 On System V.4 and embedded PowerPC systems, put small global and static
7757 data in the @samp{.sdata} section. Put small uninitialized global and
7758 static data in the @samp{.sbss} section. Do not use register @code{r13}
7759 to address small data however. This is the default behavior unless
7760 other @option{-msdata} options are used.
7764 @opindex msdata=none
7766 On embedded PowerPC systems, put all initialized global and static data
7767 in the @samp{.data} section, and all uninitialized data in the
7768 @samp{.bss} section.
7772 @cindex smaller data references (PowerPC)
7773 @cindex .sdata/.sdata2 references (PowerPC)
7774 On embedded PowerPC systems, put global and static items less than or
7775 equal to @var{num} bytes into the small data or bss sections instead of
7776 the normal data or bss section. By default, @var{num} is 8. The
7777 @option{-G @var{num}} switch is also passed to the linker.
7778 All modules should be compiled with the same @option{-G @var{num}} value.
7781 @itemx -mno-regnames
7783 @opindex mno-regnames
7784 On System V.4 and embedded PowerPC systems do (do not) emit register
7785 names in the assembly language output using symbolic forms.
7788 @itemx -mno-longcall
7790 @opindex mno-longcall
7791 Default to making all function calls via pointers, so that functions
7792 which reside further than 64 megabytes (67,108,864 bytes) from the
7793 current location can be called. This setting can be overridden by the
7794 @code{shortcall} function attribute, or by @code{#pragma longcall(0)}.
7796 Some linkers are capable of detecting out-of-range calls and generating
7797 glue code on the fly. On these systems, long calls are unnecessary and
7798 generate slower code. As of this writing, the AIX linker can do this,
7799 as can the GNU linker for PowerPC/64. It is planned to add this feature
7800 to the GNU linker for 32-bit PowerPC systems as well.
7802 On Mach-O (Darwin) systems, this option directs the compiler emit to
7803 the glue for every direct call, and the Darwin linker decides whether
7804 to use or discard it.
7806 In the future, we may cause GCC to ignore all longcall specifications
7807 when the linker is known to generate glue.
7811 Adds support for multithreading with the @dfn{pthreads} library.
7812 This option sets flags for both the preprocessor and linker.
7816 @node Darwin Options
7817 @subsection Darwin Options
7818 @cindex Darwin options
7820 These options are defined for all architectures running the Darwin operating
7821 system. They are useful for compatibility with other Mac OS compilers.
7826 Loads all members of static archive libraries.
7827 See man ld(1) for more information.
7829 @item -arch_errors_fatal
7830 @opindex arch_errors_fatal
7831 Cause the errors having to do with files that have the wrong architecture
7835 @opindex bind_at_load
7836 Causes the output file to be marked such that the dynamic linker will
7837 bind all undefined references when the file is loaded or launched.
7841 Produce a Mach-o bundle format file.
7842 See man ld(1) for more information.
7844 @item -bundle_loader @var{executable}
7845 @opindex bundle_loader
7846 This specifies the @var{executable} that will be loading the build
7847 output file being linked. See man ld(1) for more information.
7849 @item -allowable_client @var{client_name}
7853 @itemx -compatibility_version
7854 @itemx -current_version
7855 @itemx -dependency-file
7857 @itemx -dylinker_install_name
7860 @itemx -exported_symbols_list
7862 @itemx -flat_namespace
7863 @itemx -force_cpusubtype_ALL
7864 @itemx -force_flat_namespace
7865 @itemx -headerpad_max_install_names
7868 @itemx -install_name
7869 @itemx -keep_private_externs
7870 @itemx -multi_module
7871 @itemx -multiply_defined
7872 @itemx -multiply_defined_unused
7874 @itemx -nofixprebinding
7877 @itemx -noseglinkedit
7878 @itemx -pagezero_size
7880 @itemx -prebind_all_twolevel_modules
7881 @itemx -private_bundle
7882 @itemx -read_only_relocs
7884 @itemx -sectobjectsymbols
7888 @itemx -sectobjectsymbols
7890 @itemx -seg_addr_table
7891 @itemx -seg_addr_table_filename
7894 @itemx -segs_read_only_addr
7895 @itemx -segs_read_write_addr
7896 @itemx -single_module
7899 @itemx -sub_umbrella
7900 @itemx -twolevel_namespace
7903 @itemx -unexported_symbols_list
7904 @itemx -weak_reference_mismatches
7907 @opindex allowable_client
7909 @opindex client_name
7910 @opindex compatibility_version
7911 @opindex current_version
7912 @opindex dependency-file
7914 @opindex dylinker_install_name
7917 @opindex exported_symbols_list
7919 @opindex flat_namespace
7920 @opindex force_cpusubtype_ALL
7921 @opindex force_flat_namespace
7922 @opindex headerpad_max_install_names
7925 @opindex install_name
7926 @opindex keep_private_externs
7927 @opindex multi_module
7928 @opindex multiply_defined
7929 @opindex multiply_defined_unused
7931 @opindex nofixprebinding
7932 @opindex nomultidefs
7934 @opindex noseglinkedit
7935 @opindex pagezero_size
7937 @opindex prebind_all_twolevel_modules
7938 @opindex private_bundle
7939 @opindex read_only_relocs
7941 @opindex sectobjectsymbols
7945 @opindex sectobjectsymbols
7947 @opindex seg_addr_table
7948 @opindex seg_addr_table_filename
7949 @opindex seglinkedit
7951 @opindex segs_read_only_addr
7952 @opindex segs_read_write_addr
7953 @opindex single_module
7955 @opindex sub_library
7956 @opindex sub_umbrella
7957 @opindex twolevel_namespace
7960 @opindex unexported_symbols_list
7961 @opindex weak_reference_mismatches
7962 @opindex whatsloaded
7964 These options are available for Darwin linker. Darwin linker man page
7965 describes them in detail.
7970 @subsection IBM RT Options
7972 @cindex IBM RT options
7974 These @samp{-m} options are defined for the IBM RT PC:
7978 @opindex min-line-mul
7979 Use an in-line code sequence for integer multiplies. This is the
7982 @item -mcall-lib-mul
7983 @opindex mcall-lib-mul
7984 Call @code{lmul$$} for integer multiples.
7986 @item -mfull-fp-blocks
7987 @opindex mfull-fp-blocks
7988 Generate full-size floating point data blocks, including the minimum
7989 amount of scratch space recommended by IBM@. This is the default.
7991 @item -mminimum-fp-blocks
7992 @opindex mminimum-fp-blocks
7993 Do not include extra scratch space in floating point data blocks. This
7994 results in smaller code, but slower execution, since scratch space must
7995 be allocated dynamically.
7997 @cindex @file{stdarg.h} and RT PC
7998 @item -mfp-arg-in-fpregs
7999 @opindex mfp-arg-in-fpregs
8000 Use a calling sequence incompatible with the IBM calling convention in
8001 which floating point arguments are passed in floating point registers.
8002 Note that @code{stdarg.h} will not work with floating point operands
8003 if this option is specified.
8005 @item -mfp-arg-in-gregs
8006 @opindex mfp-arg-in-gregs
8007 Use the normal calling convention for floating point arguments. This is
8010 @item -mhc-struct-return
8011 @opindex mhc-struct-return
8012 Return structures of more than one word in memory, rather than in a
8013 register. This provides compatibility with the MetaWare HighC (hc)
8014 compiler. Use the option @option{-fpcc-struct-return} for compatibility
8015 with the Portable C Compiler (pcc).
8017 @item -mnohc-struct-return
8018 @opindex mnohc-struct-return
8019 Return some structures of more than one word in registers, when
8020 convenient. This is the default. For compatibility with the
8021 IBM-supplied compilers, use the option @option{-fpcc-struct-return} or the
8022 option @option{-mhc-struct-return}.
8026 @subsection MIPS Options
8027 @cindex MIPS options
8029 These @samp{-m} options are defined for the MIPS family of computers:
8033 @item -march=@var{arch}
8035 Generate code that will run on @var{arch}, which can be the name of a
8036 generic MIPS ISA, or the name of a particular processor.
8038 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
8039 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
8040 The processor names are:
8041 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
8043 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
8044 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
8048 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4300},
8049 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
8050 The special value @samp{from-abi} selects the
8051 most compatible architecture for the selected ABI (that is,
8052 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
8054 In processor names, a final @samp{000} can be abbreviated as @samp{k}
8055 (for example, @samp{-march=r2k}). Prefixes are optional, and
8056 @samp{vr} may be written @samp{r}.
8058 GCC defines two macros based on the value of this option. The first
8059 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
8060 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
8061 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
8062 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
8063 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
8065 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
8066 above. In other words, it will have the full prefix and will not
8067 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
8068 the macro names the resolved architecture (either @samp{"mips1"} or
8069 @samp{"mips3"}). It names the default architecture when no
8070 @option{-march} option is given.
8072 @item -mtune=@var{arch}
8074 Optimize for @var{arch}. Among other things, this option controls
8075 the way instructions are scheduled, and the perceived cost of arithmetic
8076 operations. The list of @var{arch} values is the same as for
8079 When this option is not used, GCC will optimize for the processor
8080 specified by @option{-march}. By using @option{-march} and
8081 @option{-mtune} together, it is possible to generate code that will
8082 run on a family of processors, but optimize the code for one
8083 particular member of that family.
8085 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
8086 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
8087 @samp{-march} ones described above.
8091 Equivalent to @samp{-march=mips1}.
8095 Equivalent to @samp{-march=mips2}.
8099 Equivalent to @samp{-march=mips3}.
8103 Equivalent to @samp{-march=mips4}.
8107 Equivalent to @samp{-march=mips32}.
8111 Equivalent to @samp{-march=mips32r2}.
8115 Equivalent to @samp{-march=mips64}.
8118 @itemx -mno-fused-madd
8119 @opindex mfused-madd
8120 @opindex mno-fused-madd
8121 Generate code that uses (does not use) the floating point multiply and
8122 accumulate instructions, when they are available. These instructions
8123 are generated by default if they are available, but this may be
8124 undesirable if the extra precision causes problems or on certain chips
8125 in the mode where denormals are rounded to zero where denormals
8126 generated by multiply and accumulate instructions cause exceptions
8131 Assume that floating point registers are 32 bits wide.
8135 Assume that floating point registers are 64 bits wide.
8139 Assume that general purpose registers are 32 bits wide.
8143 Assume that general purpose registers are 64 bits wide.
8147 Force int and long types to be 64 bits wide. See @option{-mlong32} for an
8148 explanation of the default, and the width of pointers.
8152 Force long types to be 64 bits wide. See @option{-mlong32} for an
8153 explanation of the default, and the width of pointers.
8157 Force long, int, and pointer types to be 32 bits wide.
8159 The default size of ints, longs and pointers depends on the ABI@. All
8160 the supported ABIs use 32-bit ints. The n64 ABI uses 64-bit longs, as
8161 does the 64-bit Cygnus EABI; the others use 32-bit longs. Pointers
8162 are the same size as longs, or the same size as integer registers,
8163 whichever is smaller.
8177 Generate code for the given ABI@.
8179 Note that there are two embedded ABIs: @option{-mabi=eabi}
8180 selects the one defined by Cygnus while @option{-meabi=meabi}
8181 selects the one defined by MIPS@. Both these ABIs have
8182 32-bit and 64-bit variants. Normally, GCC will generate
8183 64-bit code when you select a 64-bit architecture, but you
8184 can use @option{-mgp32} to get 32-bit code instead.
8186 @item -mabi-fake-default
8187 @opindex mabi-fake-default
8188 You don't want to know what this option does. No, really. I mean
8189 it. Move on to the next option.
8191 What? You're still here? Oh, well@enddots{} Ok, here's the deal. GCC
8192 wants the default set of options to get the root of the multilib tree,
8193 and the shared library SONAMEs without any multilib-indicating
8194 suffixes. This is not convenience for @samp{mips64-linux-gnu}, since
8195 we want to default to the N32 ABI, while still being binary-compatible
8196 with @samp{mips-linux-gnu} if you stick to the O32 ABI@. Being
8197 binary-compatible means shared libraries should have the same SONAMEs,
8198 and libraries should live in the same location. Having O32 libraries
8199 in a sub-directory named say @file{o32} is not acceptable.
8201 So we trick GCC into believing that O32 is the default ABI, except
8202 that we override the default with some internal command-line
8203 processing magic. Problem is, if we stopped at that, and you then
8204 created a multilib-aware package that used the output of @command{gcc
8205 -print-multi-lib} to decide which multilibs to build, and how, and
8206 you'd find yourself in an awkward situation when you found out that
8207 some of the options listed ended up mapping to the same multilib, and
8208 none of your libraries was actually built for the multilib that
8209 @option{-print-multi-lib} claims to be the default. So we added this
8210 option that disables the default switcher, falling back to GCC's
8211 original notion of the default library. Confused yet?
8213 For short: don't ever use this option, unless you find it in the list
8214 of additional options to be used when building for multilibs, in the
8215 output of @option{gcc -print-multi-lib}.
8219 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
8220 add normal debug information. This is the default for all
8221 platforms except for the OSF/1 reference platform, using the OSF/rose
8222 object format. If the either of the @option{-gstabs} or @option{-gstabs+}
8223 switches are used, the @file{mips-tfile} program will encapsulate the
8224 stabs within MIPS ECOFF@.
8228 Generate code for the GNU assembler. This is the default on the OSF/1
8229 reference platform, using the OSF/rose object format. Also, this is
8230 the default if the configure option @option{--with-gnu-as} is used.
8232 @item -msplit-addresses
8233 @itemx -mno-split-addresses
8234 @opindex msplit-addresses
8235 @opindex mno-split-addresses
8236 Generate code to load the high and low parts of address constants separately.
8237 This allows GCC to optimize away redundant loads of the high order
8238 bits of addresses. This optimization requires GNU as and GNU ld.
8239 This optimization is enabled by default for some embedded targets where
8240 GNU as and GNU ld are standard.
8246 The @option{-mrnames} switch says to output code using the MIPS software
8247 names for the registers, instead of the hardware names (ie, @var{a0}
8248 instead of @var{$4}). The only known assembler that supports this option
8249 is the Algorithmics assembler.
8255 The @option{-mmemcpy} switch makes all block moves call the appropriate
8256 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
8257 generating inline code.
8260 @itemx -mno-mips-tfile
8261 @opindex mmips-tfile
8262 @opindex mno-mips-tfile
8263 The @option{-mno-mips-tfile} switch causes the compiler not
8264 postprocess the object file with the @file{mips-tfile} program,
8265 after the MIPS assembler has generated it to add debug support. If
8266 @file{mips-tfile} is not run, then no local variables will be
8267 available to the debugger. In addition, @file{stage2} and
8268 @file{stage3} objects will have the temporary file names passed to the
8269 assembler embedded in the object file, which means the objects will
8270 not compare the same. The @option{-mno-mips-tfile} switch should only
8271 be used when there are bugs in the @file{mips-tfile} program that
8272 prevents compilation.
8275 @opindex msoft-float
8276 Generate output containing library calls for floating point.
8277 @strong{Warning:} the requisite libraries are not part of GCC@.
8278 Normally the facilities of the machine's usual C compiler are used, but
8279 this can't be done directly in cross-compilation. You must make your
8280 own arrangements to provide suitable library functions for
8284 @opindex mhard-float
8285 Generate output containing floating point instructions. This is the
8286 default if you use the unmodified sources.
8289 @itemx -mno-abicalls
8291 @opindex mno-abicalls
8292 Emit (or do not emit) the pseudo operations @samp{.abicalls},
8293 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
8294 position independent code.
8300 Lift (or do not lift) the usual restrictions on the size of the global
8303 GCC normally uses a single instruction to load values from the GOT.
8304 While this is relatively efficient, it will only work if the GOT
8305 is smaller than about 64k. Anything larger will cause the linker
8306 to report an error such as:
8308 @cindex relocation truncated to fit (MIPS)
8310 relocation truncated to fit: R_MIPS_GOT16 foobar
8313 If this happens, you should recompile your code with @option{-mxgot}.
8314 It should then work with very large GOTs, although it will also be
8315 less efficient, since it will take three instructions to fetch the
8316 value of a global symbol.
8318 Note that some linkers can create multiple GOTs. If you have such a
8319 linker, you should only need to use @option{-mxgot} when a single object
8320 file accesses more than 64k's worth of GOT entries. Very few do.
8322 These options have no effect unless GCC is generating position
8326 @itemx -mno-long-calls
8327 @opindex mlong-calls
8328 @opindex mno-long-calls
8329 Do all calls with the @samp{JALR} instruction, which requires
8330 loading up a function's address into a register before the call.
8331 You need to use this switch, if you call outside of the current
8332 512 megabyte segment to functions that are not through pointers.
8334 @item -membedded-pic
8335 @itemx -mno-embedded-pic
8336 @opindex membedded-pic
8337 @opindex mno-embedded-pic
8338 Generate PIC code suitable for some embedded systems. All calls are
8339 made using PC relative address, and all data is addressed using the $gp
8340 register. No more than 65536 bytes of global data may be used. This
8341 requires GNU as and GNU ld which do most of the work. This currently
8342 only works on targets which use ECOFF; it does not work with ELF@.
8344 @item -membedded-data
8345 @itemx -mno-embedded-data
8346 @opindex membedded-data
8347 @opindex mno-embedded-data
8348 Allocate variables to the read-only data section first if possible, then
8349 next in the small data section if possible, otherwise in data. This gives
8350 slightly slower code than the default, but reduces the amount of RAM required
8351 when executing, and thus may be preferred for some embedded systems.
8353 @item -muninit-const-in-rodata
8354 @itemx -mno-uninit-const-in-rodata
8355 @opindex muninit-const-in-rodata
8356 @opindex mno-uninit-const-in-rodata
8357 When used together with @option{-membedded-data}, it will always store uninitialized
8358 const variables in the read-only data section.
8360 @item -msingle-float
8361 @itemx -mdouble-float
8362 @opindex msingle-float
8363 @opindex mdouble-float
8364 The @option{-msingle-float} switch tells gcc to assume that the floating
8365 point coprocessor only supports single precision operations, as on the
8366 @samp{r4650} chip. The @option{-mdouble-float} switch permits gcc to use
8367 double precision operations. This is the default.
8373 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
8374 as on the @samp{r4650} chip.
8378 Turns on @option{-msingle-float}, @option{-mmad}, and, at least for now,
8379 @option{-mcpu=r4650}.
8385 Enable 16-bit instructions.
8389 Compile code for the processor in little endian mode.
8390 The requisite libraries are assumed to exist.
8394 Compile code for the processor in big endian mode.
8395 The requisite libraries are assumed to exist.
8399 @cindex smaller data references (MIPS)
8400 @cindex gp-relative references (MIPS)
8401 Put global and static items less than or equal to @var{num} bytes into
8402 the small data or bss sections instead of the normal data or bss
8403 section. This allows the assembler to emit one word memory reference
8404 instructions based on the global pointer (@var{gp} or @var{$28}),
8405 instead of the normal two words used. By default, @var{num} is 8 when
8406 the MIPS assembler is used, and 0 when the GNU assembler is used. The
8407 @option{-G @var{num}} switch is also passed to the assembler and linker.
8408 All modules should be compiled with the same @option{-G @var{num}}
8413 Tell the MIPS assembler to not run its preprocessor over user
8414 assembler files (with a @samp{.s} suffix) when assembling them.
8418 Pass an option to gas which will cause nops to be inserted if
8419 the read of the destination register of an mfhi or mflo instruction
8420 occurs in the following two instructions.
8425 Work around certain SB-1 CPU core errata.
8426 (This flag currently works around the SB-1 revision 2
8427 ``F1'' and ``F2'' floating point errata.)
8431 Do not include the default crt0.
8433 @item -mflush-func=@var{func}
8434 @itemx -mno-flush-func
8435 @opindex mflush-func
8436 Specifies the function to call to flush the I and D caches, or to not
8437 call any such function. If called, the function must take the same
8438 arguments as the common @code{_flush_func()}, that is, the address of the
8439 memory range for which the cache is being flushed, the size of the
8440 memory range, and the number 3 (to flush both caches). The default
8441 depends on the target gcc was configured for, but commonly is either
8442 @samp{_flush_func} or @samp{__cpu_flush}.
8444 @item -mbranch-likely
8445 @itemx -mno-branch-likely
8446 @opindex mbranch-likely
8447 @opindex mno-branch-likely
8448 Enable or disable use of Branch Likely instructions, regardless of the
8449 default for the selected architecture. By default, Branch Likely
8450 instructions may be generated if they are supported by the selected
8451 architecture. An exception is for the MIPS32 and MIPS64 architectures
8452 and processors which implement those architectures; for those, Branch
8453 Likely instructions will not be generated by default because the MIPS32
8454 and MIPS64 architectures specifically deprecate their use.
8457 @node i386 and x86-64 Options
8458 @subsection Intel 386 and AMD x86-64 Options
8459 @cindex i386 Options
8460 @cindex x86-64 Options
8461 @cindex Intel 386 Options
8462 @cindex AMD x86-64 Options
8464 These @samp{-m} options are defined for the i386 and x86-64 family of
8468 @item -mtune=@var{cpu-type}
8470 Tune to @var{cpu-type} everything applicable about the generated code, except
8471 for the ABI and the set of available instructions. The choices for
8472 @var{cpu-type} are @samp{i386}, @samp{i486}, @samp{i586}, @samp{i686},
8473 @samp{pentium}, @samp{pentium-mmx}, @samp{pentiumpro}, @samp{pentium2},
8474 @samp{pentium3}, @samp{pentium4}, @samp{k6}, @samp{k6-2}, @samp{k6-3},
8475 @samp{athlon}, @samp{athlon-tbird}, @samp{athlon-4}, @samp{athlon-xp},
8476 @samp{athlon-mp}, @samp{winchip-c6}, @samp{winchip2}, @samp{k8}, @samp{c3}
8479 While picking a specific @var{cpu-type} will schedule things appropriately
8480 for that particular chip, the compiler will not generate any code that
8481 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8482 being used. @samp{i586} is equivalent to @samp{pentium} and @samp{i686}
8483 is equivalent to @samp{pentiumpro}. @samp{k6} and @samp{athlon} are the
8484 AMD chips as opposed to the Intel ones.
8486 @item -march=@var{cpu-type}
8488 Generate instructions for the machine type @var{cpu-type}. The choices
8489 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8490 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8492 @item -mcpu=@var{cpu-type}
8494 A deprecated synonym for @option{-mtune}.
8503 @opindex mpentiumpro
8504 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8505 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8506 These synonyms are deprecated.
8508 @item -mfpmath=@var{unit}
8510 generate floating point arithmetics for selected unit @var{unit}. the choices
8515 Use the standard 387 floating point coprocessor present majority of chips and
8516 emulated otherwise. Code compiled with this option will run almost everywhere.
8517 The temporary results are computed in 80bit precision instead of precision
8518 specified by the type resulting in slightly different results compared to most
8519 of other chips. See @option{-ffloat-store} for more detailed description.
8521 This is the default choice for i386 compiler.
8524 Use scalar floating point instructions present in the SSE instruction set.
8525 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8526 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8527 instruction set supports only single precision arithmetics, thus the double and
8528 extended precision arithmetics is still done using 387. Later version, present
8529 only in Pentium4 and the future AMD x86-64 chips supports double precision
8532 For i387 you need to use @option{-march=@var{cpu-type}}, @option{-msse} or
8533 @option{-msse2} switches to enable SSE extensions and make this option
8534 effective. For x86-64 compiler, these extensions are enabled by default.
8536 The resulting code should be considerably faster in majority of cases and avoid
8537 the numerical instability problems of 387 code, but may break some existing
8538 code that expects temporaries to be 80bit.
8540 This is the default choice for x86-64 compiler.
8543 Use all SSE extensions enabled by @option{-msse2} as well as the new
8544 SSE extensions in Prescott New Instructions. @option{-mpni} also
8545 enables 2 builtin functions, @code{__builtin_ia32_monitor} and
8546 @code{__builtin_ia32_mwait}, for new instructions @code{monitor} and
8550 Attempt to utilize both instruction sets at once. This effectively double the
8551 amount of available registers and on chips with separate execution units for
8552 387 and SSE the execution resources too. Use this option with care, as it is
8553 still experimental, because gcc register allocator does not model separate
8554 functional units well resulting in instable performance.
8557 @item -masm=@var{dialect}
8558 @opindex masm=@var{dialect}
8559 Output asm instructions using selected @var{dialect}. Supported choices are
8560 @samp{intel} or @samp{att} (the default one).
8565 @opindex mno-ieee-fp
8566 Control whether or not the compiler uses IEEE floating point
8567 comparisons. These handle correctly the case where the result of a
8568 comparison is unordered.
8571 @opindex msoft-float
8572 Generate output containing library calls for floating point.
8573 @strong{Warning:} the requisite libraries are not part of GCC@.
8574 Normally the facilities of the machine's usual C compiler are used, but
8575 this can't be done directly in cross-compilation. You must make your
8576 own arrangements to provide suitable library functions for
8579 On machines where a function returns floating point results in the 80387
8580 register stack, some floating point opcodes may be emitted even if
8581 @option{-msoft-float} is used.
8583 @item -mno-fp-ret-in-387
8584 @opindex mno-fp-ret-in-387
8585 Do not use the FPU registers for return values of functions.
8587 The usual calling convention has functions return values of types
8588 @code{float} and @code{double} in an FPU register, even if there
8589 is no FPU@. The idea is that the operating system should emulate
8592 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8593 in ordinary CPU registers instead.
8595 @item -mno-fancy-math-387
8596 @opindex mno-fancy-math-387
8597 Some 387 emulators do not support the @code{sin}, @code{cos} and
8598 @code{sqrt} instructions for the 387. Specify this option to avoid
8599 generating those instructions. This option is the default on FreeBSD,
8600 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8601 indicates that the target cpu will always have an FPU and so the
8602 instruction will not need emulation. As of revision 2.6.1, these
8603 instructions are not generated unless you also use the
8604 @option{-funsafe-math-optimizations} switch.
8606 @item -malign-double
8607 @itemx -mno-align-double
8608 @opindex malign-double
8609 @opindex mno-align-double
8610 Control whether GCC aligns @code{double}, @code{long double}, and
8611 @code{long long} variables on a two word boundary or a one word
8612 boundary. Aligning @code{double} variables on a two word boundary will
8613 produce code that runs somewhat faster on a @samp{Pentium} at the
8614 expense of more memory.
8616 @strong{Warning:} if you use the @option{-malign-double} switch,
8617 structures containing the above types will be aligned differently than
8618 the published application binary interface specifications for the 386
8619 and will not be binary compatible with structures in code compiled
8620 without that switch.
8622 @item -m96bit-long-double
8623 @itemx -m128bit-long-double
8624 @opindex m96bit-long-double
8625 @opindex m128bit-long-double
8626 These switches control the size of @code{long double} type. The i386
8627 application binary interface specifies the size to be 96 bits,
8628 so @option{-m96bit-long-double} is the default in 32 bit mode.
8630 Modern architectures (Pentium and newer) would prefer @code{long double}
8631 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8632 conforming to the ABI, this would not be possible. So specifying a
8633 @option{-m128bit-long-double} will align @code{long double}
8634 to a 16 byte boundary by padding the @code{long double} with an additional
8637 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8638 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8640 Notice that neither of these options enable any extra precision over the x87
8641 standard of 80 bits for a @code{long double}.
8643 @strong{Warning:} if you override the default value for your target ABI, the
8644 structures and arrays containing @code{long double} will change their size as
8645 well as function calling convention for function taking @code{long double}
8646 will be modified. Hence they will not be binary compatible with arrays or
8647 structures in code compiled without that switch.
8651 @itemx -mno-svr3-shlib
8652 @opindex msvr3-shlib
8653 @opindex mno-svr3-shlib
8654 Control whether GCC places uninitialized local variables into the
8655 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8656 into @code{bss}. These options are meaningful only on System V Release 3.
8660 Use a different function-calling convention, in which functions that
8661 take a fixed number of arguments return with the @code{ret} @var{num}
8662 instruction, which pops their arguments while returning. This saves one
8663 instruction in the caller since there is no need to pop the arguments
8666 You can specify that an individual function is called with this calling
8667 sequence with the function attribute @samp{stdcall}. You can also
8668 override the @option{-mrtd} option by using the function attribute
8669 @samp{cdecl}. @xref{Function Attributes}.
8671 @strong{Warning:} this calling convention is incompatible with the one
8672 normally used on Unix, so you cannot use it if you need to call
8673 libraries compiled with the Unix compiler.
8675 Also, you must provide function prototypes for all functions that
8676 take variable numbers of arguments (including @code{printf});
8677 otherwise incorrect code will be generated for calls to those
8680 In addition, seriously incorrect code will result if you call a
8681 function with too many arguments. (Normally, extra arguments are
8682 harmlessly ignored.)
8684 @item -mregparm=@var{num}
8686 Control how many registers are used to pass integer arguments. By
8687 default, no registers are used to pass arguments, and at most 3
8688 registers can be used. You can control this behavior for a specific
8689 function by using the function attribute @samp{regparm}.
8690 @xref{Function Attributes}.
8692 @strong{Warning:} if you use this switch, and
8693 @var{num} is nonzero, then you must build all modules with the same
8694 value, including any libraries. This includes the system libraries and
8697 @item -mpreferred-stack-boundary=@var{num}
8698 @opindex mpreferred-stack-boundary
8699 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8700 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8701 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8702 size (@option{-Os}), in which case the default is the minimum correct
8703 alignment (4 bytes for x86, and 8 bytes for x86-64).
8705 On Pentium and PentiumPro, @code{double} and @code{long double} values
8706 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8707 suffer significant run time performance penalties. On Pentium III, the
8708 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8709 penalties if it is not 16 byte aligned.
8711 To ensure proper alignment of this values on the stack, the stack boundary
8712 must be as aligned as that required by any value stored on the stack.
8713 Further, every function must be generated such that it keeps the stack
8714 aligned. Thus calling a function compiled with a higher preferred
8715 stack boundary from a function compiled with a lower preferred stack
8716 boundary will most likely misalign the stack. It is recommended that
8717 libraries that use callbacks always use the default setting.
8719 This extra alignment does consume extra stack space, and generally
8720 increases code size. Code that is sensitive to stack space usage, such
8721 as embedded systems and operating system kernels, may want to reduce the
8722 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8740 These switches enable or disable the use of built-in functions that allow
8741 direct access to the MMX, SSE and 3Dnow extensions of the instruction set.
8743 @xref{X86 Built-in Functions}, for details of the functions enabled
8744 and disabled by these switches.
8746 To have SSE/SSE2 instructions generated automatically from floating-point
8747 code, see @option{-mfpmath=sse}.
8750 @itemx -mno-push-args
8752 @opindex mno-push-args
8753 Use PUSH operations to store outgoing parameters. This method is shorter
8754 and usually equally fast as method using SUB/MOV operations and is enabled
8755 by default. In some cases disabling it may improve performance because of
8756 improved scheduling and reduced dependencies.
8758 @item -maccumulate-outgoing-args
8759 @opindex maccumulate-outgoing-args
8760 If enabled, the maximum amount of space required for outgoing arguments will be
8761 computed in the function prologue. This is faster on most modern CPUs
8762 because of reduced dependencies, improved scheduling and reduced stack usage
8763 when preferred stack boundary is not equal to 2. The drawback is a notable
8764 increase in code size. This switch implies @option{-mno-push-args}.
8768 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8769 on thread-safe exception handling must compile and link all code with the
8770 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8771 @option{-D_MT}; when linking, it links in a special thread helper library
8772 @option{-lmingwthrd} which cleans up per thread exception handling data.
8774 @item -mno-align-stringops
8775 @opindex mno-align-stringops
8776 Do not align destination of inlined string operations. This switch reduces
8777 code size and improves performance in case the destination is already aligned,
8778 but gcc don't know about it.
8780 @item -minline-all-stringops
8781 @opindex minline-all-stringops
8782 By default GCC inlines string operations only when destination is known to be
8783 aligned at least to 4 byte boundary. This enables more inlining, increase code
8784 size, but may improve performance of code that depends on fast memcpy, strlen
8785 and memset for short lengths.
8787 @item -momit-leaf-frame-pointer
8788 @opindex momit-leaf-frame-pointer
8789 Don't keep the frame pointer in a register for leaf functions. This
8790 avoids the instructions to save, set up and restore frame pointers and
8791 makes an extra register available in leaf functions. The option
8792 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8793 which might make debugging harder.
8795 @item -mtls-direct-seg-refs
8796 @itemx -mno-tls-direct-seg-refs
8797 @opindex mtls-direct-seg-refs
8798 Controls whether TLS variables may be accessed with offsets from the
8799 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8800 or whether the thread base pointer must be added. Whether or not this
8801 is legal depends on the operating system, and whether it maps the
8802 segment to cover the entire TLS area.
8804 For systems that use GNU libc, the default is on.
8807 These @samp{-m} switches are supported in addition to the above
8808 on AMD x86-64 processors in 64-bit environments.
8815 Generate code for a 32-bit or 64-bit environment.
8816 The 32-bit environment sets int, long and pointer to 32 bits and
8817 generates code that runs on any i386 system.
8818 The 64-bit environment sets int to 32 bits and long and pointer
8819 to 64 bits and generates code for AMD's x86-64 architecture.
8822 @opindex no-red-zone
8823 Do not use a so called red zone for x86-64 code. The red zone is mandated
8824 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8825 stack pointer that will not be modified by signal or interrupt handlers
8826 and therefore can be used for temporary data without adjusting the stack
8827 pointer. The flag @option{-mno-red-zone} disables this red zone.
8829 @item -mcmodel=small
8830 @opindex mcmodel=small
8831 Generate code for the small code model: the program and its symbols must
8832 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8833 Programs can be statically or dynamically linked. This is the default
8836 @item -mcmodel=kernel
8837 @opindex mcmodel=kernel
8838 Generate code for the kernel code model. The kernel runs in the
8839 negative 2 GB of the address space.
8840 This model has to be used for Linux kernel code.
8842 @item -mcmodel=medium
8843 @opindex mcmodel=medium
8844 Generate code for the medium model: The program is linked in the lower 2
8845 GB of the address space but symbols can be located anywhere in the
8846 address space. Programs can be statically or dynamically linked, but
8847 building of shared libraries are not supported with the medium model.
8849 @item -mcmodel=large
8850 @opindex mcmodel=large
8851 Generate code for the large model: This model makes no assumptions
8852 about addresses and sizes of sections. Currently GCC does not implement
8857 @subsection HPPA Options
8858 @cindex HPPA Options
8860 These @samp{-m} options are defined for the HPPA family of computers:
8863 @item -march=@var{architecture-type}
8865 Generate code for the specified architecture. The choices for
8866 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8867 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8868 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8869 architecture option for your machine. Code compiled for lower numbered
8870 architectures will run on higher numbered architectures, but not the
8873 PA 2.0 support currently requires gas snapshot 19990413 or later. The
8874 next release of binutils (current is 2.9.1) will probably contain PA 2.0
8878 @itemx -mpa-risc-1-1
8879 @itemx -mpa-risc-2-0
8880 @opindex mpa-risc-1-0
8881 @opindex mpa-risc-1-1
8882 @opindex mpa-risc-2-0
8883 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8886 @opindex mbig-switch
8887 Generate code suitable for big switch tables. Use this option only if
8888 the assembler/linker complain about out of range branches within a switch
8891 @item -mjump-in-delay
8892 @opindex mjump-in-delay
8893 Fill delay slots of function calls with unconditional jump instructions
8894 by modifying the return pointer for the function call to be the target
8895 of the conditional jump.
8897 @item -mdisable-fpregs
8898 @opindex mdisable-fpregs
8899 Prevent floating point registers from being used in any manner. This is
8900 necessary for compiling kernels which perform lazy context switching of
8901 floating point registers. If you use this option and attempt to perform
8902 floating point operations, the compiler will abort.
8904 @item -mdisable-indexing
8905 @opindex mdisable-indexing
8906 Prevent the compiler from using indexing address modes. This avoids some
8907 rather obscure problems when compiling MIG generated code under MACH@.
8909 @item -mno-space-regs
8910 @opindex mno-space-regs
8911 Generate code that assumes the target has no space registers. This allows
8912 GCC to generate faster indirect calls and use unscaled index address modes.
8914 Such code is suitable for level 0 PA systems and kernels.
8916 @item -mfast-indirect-calls
8917 @opindex mfast-indirect-calls
8918 Generate code that assumes calls never cross space boundaries. This
8919 allows GCC to emit code which performs faster indirect calls.
8921 This option will not work in the presence of shared libraries or nested
8924 @item -mlong-load-store
8925 @opindex mlong-load-store
8926 Generate 3-instruction load and store sequences as sometimes required by
8927 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8930 @item -mportable-runtime
8931 @opindex mportable-runtime
8932 Use the portable calling conventions proposed by HP for ELF systems.
8936 Enable the use of assembler directives only GAS understands.
8938 @item -mschedule=@var{cpu-type}
8940 Schedule code according to the constraints for the machine type
8941 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8942 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8943 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8944 proper scheduling option for your machine. The default scheduling is
8948 @opindex mlinker-opt
8949 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8950 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8951 linkers in which they give bogus error messages when linking some programs.
8954 @opindex msoft-float
8955 Generate output containing library calls for floating point.
8956 @strong{Warning:} the requisite libraries are not available for all HPPA
8957 targets. Normally the facilities of the machine's usual C compiler are
8958 used, but this cannot be done directly in cross-compilation. You must make
8959 your own arrangements to provide suitable library functions for
8960 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8961 does provide software floating point support.
8963 @option{-msoft-float} changes the calling convention in the output file;
8964 therefore, it is only useful if you compile @emph{all} of a program with
8965 this option. In particular, you need to compile @file{libgcc.a}, the
8966 library that comes with GCC, with @option{-msoft-float} in order for
8971 Generate the predefine, @code{_SIO}, for server IO. The default is
8972 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8973 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO. These
8974 options are available under HP-UX and HI-UX.
8978 Use GNU ld specific options. This passes @option{-shared} to ld when
8979 building a shared library. It is the default when GCC is configured,
8980 explicitly or implicitly, with the GNU linker. This option does not
8981 have any affect on which ld is called, it only changes what parameters
8982 are passed to that ld. The ld that is called is determined by the
8983 @option{--with-ld} configure option, gcc's program search path, and
8984 finally by the user's @env{PATH}. The linker used by GCC can be printed
8985 using @samp{which `gcc -print-prog-name=ld`}.
8989 Use HP ld specific options. This passes @option{-b} to ld when building
8990 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8991 links. It is the default when GCC is configured, explicitly or
8992 implicitly, with the HP linker. This option does not have any affect on
8993 which ld is called, it only changes what parameters are passed to that
8994 ld. The ld that is called is determined by the @option{--with-ld}
8995 configure option, gcc's program search path, and finally by the user's
8996 @env{PATH}. The linker used by GCC can be printed using @samp{which
8997 `gcc -print-prog-name=ld`}.
9000 @opindex mno-long-calls
9001 Generate code that uses long call sequences. This ensures that a call
9002 is always able to reach linker generated stubs. The default is to generate
9003 long calls only when the distance from the call site to the beginning
9004 of the function or translation unit, as the case may be, exceeds a
9005 predefined limit set by the branch type being used. The limits for
9006 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9007 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9010 Distances are measured from the beginning of functions when using the
9011 @option{-ffunction-sections} option, or when using the @option{-mgas}
9012 and @option{-mno-portable-runtime} options together under HP-UX with
9015 It is normally not desirable to use this option as it will degrade
9016 performance. However, it may be useful in large applications,
9017 particularly when partial linking is used to build the application.
9019 The types of long calls used depends on the capabilities of the
9020 assembler and linker, and the type of code being generated. The
9021 impact on systems that support long absolute calls, and long pic
9022 symbol-difference or pc-relative calls should be relatively small.
9023 However, an indirect call is used on 32-bit ELF systems in pic code
9024 and it is quite long.
9028 Suppress the generation of link options to search libdld.sl when the
9029 @option{-static} option is specified on HP-UX 10 and later.
9033 The HP-UX implementation of setlocale in libc has a dependency on
9034 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9035 when the @option{-static} option is specified, special link options
9036 are needed to resolve this dependency.
9038 On HP-UX 10 and later, the GCC driver adds the necessary options to
9039 link with libdld.sl when the @option{-static} option is specified.
9040 This causes the resulting binary to be dynamic. On the 64-bit port,
9041 the linkers generate dynamic binaries by default in any case. The
9042 @option{-nolibdld} option can be used to prevent the GCC driver from
9043 adding these link options.
9047 Add support for multithreading with the @dfn{dce thread} library
9048 under HP-UX. This option sets flags for both the preprocessor and
9052 @node Intel 960 Options
9053 @subsection Intel 960 Options
9055 These @samp{-m} options are defined for the Intel 960 implementations:
9058 @item -m@var{cpu-type}
9066 Assume the defaults for the machine type @var{cpu-type} for some of
9067 the other options, including instruction scheduling, floating point
9068 support, and addressing modes. The choices for @var{cpu-type} are
9069 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
9070 @samp{sa}, and @samp{sb}.
9077 @opindex msoft-float
9078 The @option{-mnumerics} option indicates that the processor does support
9079 floating-point instructions. The @option{-msoft-float} option indicates
9080 that floating-point support should not be assumed.
9082 @item -mleaf-procedures
9083 @itemx -mno-leaf-procedures
9084 @opindex mleaf-procedures
9085 @opindex mno-leaf-procedures
9086 Do (or do not) attempt to alter leaf procedures to be callable with the
9087 @code{bal} instruction as well as @code{call}. This will result in more
9088 efficient code for explicit calls when the @code{bal} instruction can be
9089 substituted by the assembler or linker, but less efficient code in other
9090 cases, such as calls via function pointers, or using a linker that doesn't
9091 support this optimization.
9094 @itemx -mno-tail-call
9096 @opindex mno-tail-call
9097 Do (or do not) make additional attempts (beyond those of the
9098 machine-independent portions of the compiler) to optimize tail-recursive
9099 calls into branches. You may not want to do this because the detection of
9100 cases where this is not valid is not totally complete. The default is
9101 @option{-mno-tail-call}.
9103 @item -mcomplex-addr
9104 @itemx -mno-complex-addr
9105 @opindex mcomplex-addr
9106 @opindex mno-complex-addr
9107 Assume (or do not assume) that the use of a complex addressing mode is a
9108 win on this implementation of the i960. Complex addressing modes may not
9109 be worthwhile on the K-series, but they definitely are on the C-series.
9110 The default is currently @option{-mcomplex-addr} for all processors except
9114 @itemx -mno-code-align
9115 @opindex mcode-align
9116 @opindex mno-code-align
9117 Align code to 8-byte boundaries for faster fetching (or don't bother).
9118 Currently turned on by default for C-series implementations only.
9121 @item -mclean-linkage
9122 @itemx -mno-clean-linkage
9123 @opindex mclean-linkage
9124 @opindex mno-clean-linkage
9125 These options are not fully implemented.
9129 @itemx -mic2.0-compat
9130 @itemx -mic3.0-compat
9132 @opindex mic2.0-compat
9133 @opindex mic3.0-compat
9134 Enable compatibility with iC960 v2.0 or v3.0.
9138 @opindex masm-compat
9140 Enable compatibility with the iC960 assembler.
9142 @item -mstrict-align
9143 @itemx -mno-strict-align
9144 @opindex mstrict-align
9145 @opindex mno-strict-align
9146 Do not permit (do permit) unaligned accesses.
9150 Enable structure-alignment compatibility with Intel's gcc release version
9151 1.3 (based on gcc 1.37). This option implies @option{-mstrict-align}.
9153 @item -mlong-double-64
9154 @opindex mlong-double-64
9155 Implement type @samp{long double} as 64-bit floating point numbers.
9156 Without the option @samp{long double} is implemented by 80-bit
9157 floating point numbers. The only reason we have it because there is
9158 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
9159 is only useful for people using soft-float targets. Otherwise, we
9160 should recommend against use of it.
9164 @node DEC Alpha Options
9165 @subsection DEC Alpha Options
9167 These @samp{-m} options are defined for the DEC Alpha implementations:
9170 @item -mno-soft-float
9172 @opindex mno-soft-float
9173 @opindex msoft-float
9174 Use (do not use) the hardware floating-point instructions for
9175 floating-point operations. When @option{-msoft-float} is specified,
9176 functions in @file{libgcc.a} will be used to perform floating-point
9177 operations. Unless they are replaced by routines that emulate the
9178 floating-point operations, or compiled in such a way as to call such
9179 emulations routines, these routines will issue floating-point
9180 operations. If you are compiling for an Alpha without floating-point
9181 operations, you must ensure that the library is built so as not to call
9184 Note that Alpha implementations without floating-point operations are
9185 required to have floating-point registers.
9190 @opindex mno-fp-regs
9191 Generate code that uses (does not use) the floating-point register set.
9192 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9193 register set is not used, floating point operands are passed in integer
9194 registers as if they were integers and floating-point results are passed
9195 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9196 so any function with a floating-point argument or return value called by code
9197 compiled with @option{-mno-fp-regs} must also be compiled with that
9200 A typical use of this option is building a kernel that does not use,
9201 and hence need not save and restore, any floating-point registers.
9205 The Alpha architecture implements floating-point hardware optimized for
9206 maximum performance. It is mostly compliant with the IEEE floating
9207 point standard. However, for full compliance, software assistance is
9208 required. This option generates code fully IEEE compliant code
9209 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9210 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9211 defined during compilation. The resulting code is less efficient but is
9212 able to correctly support denormalized numbers and exceptional IEEE
9213 values such as not-a-number and plus/minus infinity. Other Alpha
9214 compilers call this option @option{-ieee_with_no_inexact}.
9216 @item -mieee-with-inexact
9217 @opindex mieee-with-inexact
9218 This is like @option{-mieee} except the generated code also maintains
9219 the IEEE @var{inexact-flag}. Turning on this option causes the
9220 generated code to implement fully-compliant IEEE math. In addition to
9221 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9222 macro. On some Alpha implementations the resulting code may execute
9223 significantly slower than the code generated by default. Since there is
9224 very little code that depends on the @var{inexact-flag}, you should
9225 normally not specify this option. Other Alpha compilers call this
9226 option @option{-ieee_with_inexact}.
9228 @item -mfp-trap-mode=@var{trap-mode}
9229 @opindex mfp-trap-mode
9230 This option controls what floating-point related traps are enabled.
9231 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9232 The trap mode can be set to one of four values:
9236 This is the default (normal) setting. The only traps that are enabled
9237 are the ones that cannot be disabled in software (e.g., division by zero
9241 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9245 Like @samp{su}, but the instructions are marked to be safe for software
9246 completion (see Alpha architecture manual for details).
9249 Like @samp{su}, but inexact traps are enabled as well.
9252 @item -mfp-rounding-mode=@var{rounding-mode}
9253 @opindex mfp-rounding-mode
9254 Selects the IEEE rounding mode. Other Alpha compilers call this option
9255 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9260 Normal IEEE rounding mode. Floating point numbers are rounded towards
9261 the nearest machine number or towards the even machine number in case
9265 Round towards minus infinity.
9268 Chopped rounding mode. Floating point numbers are rounded towards zero.
9271 Dynamic rounding mode. A field in the floating point control register
9272 (@var{fpcr}, see Alpha architecture reference manual) controls the
9273 rounding mode in effect. The C library initializes this register for
9274 rounding towards plus infinity. Thus, unless your program modifies the
9275 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9278 @item -mtrap-precision=@var{trap-precision}
9279 @opindex mtrap-precision
9280 In the Alpha architecture, floating point traps are imprecise. This
9281 means without software assistance it is impossible to recover from a
9282 floating trap and program execution normally needs to be terminated.
9283 GCC can generate code that can assist operating system trap handlers
9284 in determining the exact location that caused a floating point trap.
9285 Depending on the requirements of an application, different levels of
9286 precisions can be selected:
9290 Program precision. This option is the default and means a trap handler
9291 can only identify which program caused a floating point exception.
9294 Function precision. The trap handler can determine the function that
9295 caused a floating point exception.
9298 Instruction precision. The trap handler can determine the exact
9299 instruction that caused a floating point exception.
9302 Other Alpha compilers provide the equivalent options called
9303 @option{-scope_safe} and @option{-resumption_safe}.
9305 @item -mieee-conformant
9306 @opindex mieee-conformant
9307 This option marks the generated code as IEEE conformant. You must not
9308 use this option unless you also specify @option{-mtrap-precision=i} and either
9309 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9310 is to emit the line @samp{.eflag 48} in the function prologue of the
9311 generated assembly file. Under DEC Unix, this has the effect that
9312 IEEE-conformant math library routines will be linked in.
9314 @item -mbuild-constants
9315 @opindex mbuild-constants
9316 Normally GCC examines a 32- or 64-bit integer constant to
9317 see if it can construct it from smaller constants in two or three
9318 instructions. If it cannot, it will output the constant as a literal and
9319 generate code to load it from the data segment at runtime.
9321 Use this option to require GCC to construct @emph{all} integer constants
9322 using code, even if it takes more instructions (the maximum is six).
9324 You would typically use this option to build a shared library dynamic
9325 loader. Itself a shared library, it must relocate itself in memory
9326 before it can find the variables and constants in its own data segment.
9332 Select whether to generate code to be assembled by the vendor-supplied
9333 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9351 Indicate whether GCC should generate code to use the optional BWX,
9352 CIX, FIX and MAX instruction sets. The default is to use the instruction
9353 sets supported by the CPU type specified via @option{-mcpu=} option or that
9354 of the CPU on which GCC was built if none was specified.
9359 @opindex mfloat-ieee
9360 Generate code that uses (does not use) VAX F and G floating point
9361 arithmetic instead of IEEE single and double precision.
9363 @item -mexplicit-relocs
9364 @itemx -mno-explicit-relocs
9365 @opindex mexplicit-relocs
9366 @opindex mno-explicit-relocs
9367 Older Alpha assemblers provided no way to generate symbol relocations
9368 except via assembler macros. Use of these macros does not allow
9369 optimal instruction scheduling. GNU binutils as of version 2.12
9370 supports a new syntax that allows the compiler to explicitly mark
9371 which relocations should apply to which instructions. This option
9372 is mostly useful for debugging, as GCC detects the capabilities of
9373 the assembler when it is built and sets the default accordingly.
9377 @opindex msmall-data
9378 @opindex mlarge-data
9379 When @option{-mexplicit-relocs} is in effect, static data is
9380 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9381 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9382 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9383 16-bit relocations off of the @code{$gp} register. This limits the
9384 size of the small data area to 64KB, but allows the variables to be
9385 directly accessed via a single instruction.
9387 The default is @option{-mlarge-data}. With this option the data area
9388 is limited to just below 2GB. Programs that require more than 2GB of
9389 data must use @code{malloc} or @code{mmap} to allocate the data in the
9390 heap instead of in the program's data segment.
9392 When generating code for shared libraries, @option{-fpic} implies
9393 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9397 @opindex msmall-text
9398 @opindex mlarge-text
9399 When @option{-msmall-text} is used, the compiler assumes that the
9400 code of the entire program (or shared library) fits in 4MB, and is
9401 thus reachable with a branch instruction. When @option{-msmall-data}
9402 is used, the compiler can assume that all local symbols share the
9403 same @code{$gp} value, and thus reduce the number of instructions
9404 required for a function call from 4 to 1.
9406 The default is @option{-mlarge-text}.
9408 @item -mcpu=@var{cpu_type}
9410 Set the instruction set and instruction scheduling parameters for
9411 machine type @var{cpu_type}. You can specify either the @samp{EV}
9412 style name or the corresponding chip number. GCC supports scheduling
9413 parameters for the EV4, EV5 and EV6 family of processors and will
9414 choose the default values for the instruction set from the processor
9415 you specify. If you do not specify a processor type, GCC will default
9416 to the processor on which the compiler was built.
9418 Supported values for @var{cpu_type} are
9424 Schedules as an EV4 and has no instruction set extensions.
9428 Schedules as an EV5 and has no instruction set extensions.
9432 Schedules as an EV5 and supports the BWX extension.
9437 Schedules as an EV5 and supports the BWX and MAX extensions.
9441 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9445 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9448 @item -mtune=@var{cpu_type}
9450 Set only the instruction scheduling parameters for machine type
9451 @var{cpu_type}. The instruction set is not changed.
9453 @item -mmemory-latency=@var{time}
9454 @opindex mmemory-latency
9455 Sets the latency the scheduler should assume for typical memory
9456 references as seen by the application. This number is highly
9457 dependent on the memory access patterns used by the application
9458 and the size of the external cache on the machine.
9460 Valid options for @var{time} are
9464 A decimal number representing clock cycles.
9470 The compiler contains estimates of the number of clock cycles for
9471 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9472 (also called Dcache, Scache, and Bcache), as well as to main memory.
9473 Note that L3 is only valid for EV5.
9478 @node DEC Alpha/VMS Options
9479 @subsection DEC Alpha/VMS Options
9481 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9484 @item -mvms-return-codes
9485 @opindex mvms-return-codes
9486 Return VMS condition codes from main. The default is to return POSIX
9487 style condition (e.g.@ error) codes.
9490 @node H8/300 Options
9491 @subsection H8/300 Options
9493 These @samp{-m} options are defined for the H8/300 implementations:
9498 Shorten some address references at link time, when possible; uses the
9499 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9500 ld, Using ld}, for a fuller description.
9504 Generate code for the H8/300H@.
9508 Generate code for the H8S@.
9512 Generate code for the H8S and H8/300H in the normal mode. This switch
9513 must be used either with -mh or -ms.
9517 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9521 Make @code{int} data 32 bits by default.
9525 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9526 The default for the H8/300H and H8S is to align longs and floats on 4
9528 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9529 This option has no effect on the H8/300.
9533 @subsection SH Options
9535 These @samp{-m} options are defined for the SH implementations:
9540 Generate code for the SH1.
9544 Generate code for the SH2.
9547 Generate code for the SH2e.
9551 Generate code for the SH3.
9555 Generate code for the SH3e.
9559 Generate code for the SH4 without a floating-point unit.
9561 @item -m4-single-only
9562 @opindex m4-single-only
9563 Generate code for the SH4 with a floating-point unit that only
9564 supports single-precision arithmetic.
9568 Generate code for the SH4 assuming the floating-point unit is in
9569 single-precision mode by default.
9573 Generate code for the SH4.
9577 Compile code for the processor in big endian mode.
9581 Compile code for the processor in little endian mode.
9585 Align doubles at 64-bit boundaries. Note that this changes the calling
9586 conventions, and thus some functions from the standard C library will
9587 not work unless you recompile it first with @option{-mdalign}.
9591 Shorten some address references at link time, when possible; uses the
9592 linker option @option{-relax}.
9596 Use 32-bit offsets in @code{switch} tables. The default is to use
9601 Enable the use of the instruction @code{fmovd}.
9605 Comply with the calling conventions defined by Renesas.
9609 Mark the @code{MAC} register as call-clobbered, even if
9610 @option{-mhitachi} is given.
9614 Increase IEEE-compliance of floating-point code.
9618 Dump instruction size and location in the assembly code.
9622 This option is deprecated. It pads structures to multiple of 4 bytes,
9623 which is incompatible with the SH ABI@.
9627 Optimize for space instead of speed. Implied by @option{-Os}.
9631 When generating position-independent code, emit function calls using
9632 the Global Offset Table instead of the Procedure Linkage Table.
9636 Generate a library function call to invalidate instruction cache
9637 entries, after fixing up a trampoline. This library function call
9638 doesn't assume it can write to the whole memory address space. This
9639 is the default when the target is @code{sh-*-linux*}.
9642 @node System V Options
9643 @subsection Options for System V
9645 These additional options are available on System V Release 4 for
9646 compatibility with other compilers on those systems:
9651 Create a shared object.
9652 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
9656 Identify the versions of each tool used by the compiler, in a
9657 @code{.ident} assembler directive in the output.
9661 Refrain from adding @code{.ident} directives to the output file (this is
9664 @item -YP,@var{dirs}
9666 Search the directories @var{dirs}, and no others, for libraries
9667 specified with @option{-l}.
9671 Look in the directory @var{dir} to find the M4 preprocessor.
9672 The assembler uses this option.
9673 @c This is supposed to go with a -Yd for predefined M4 macro files, but
9674 @c the generic assembler that comes with Solaris takes just -Ym.
9677 @node TMS320C3x/C4x Options
9678 @subsection TMS320C3x/C4x Options
9679 @cindex TMS320C3x/C4x Options
9681 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
9685 @item -mcpu=@var{cpu_type}
9687 Set the instruction set, register set, and instruction scheduling
9688 parameters for machine type @var{cpu_type}. Supported values for
9689 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
9690 @samp{c44}. The default is @samp{c40} to generate code for the
9695 @itemx -msmall-memory
9697 @opindex mbig-memory
9699 @opindex msmall-memory
9701 Generates code for the big or small memory model. The small memory
9702 model assumed that all data fits into one 64K word page. At run-time
9703 the data page (DP) register must be set to point to the 64K page
9704 containing the .bss and .data program sections. The big memory model is
9705 the default and requires reloading of the DP register for every direct
9712 Allow (disallow) allocation of general integer operands into the block
9719 Enable (disable) generation of code using decrement and branch,
9720 DBcond(D), instructions. This is enabled by default for the C4x. To be
9721 on the safe side, this is disabled for the C3x, since the maximum
9722 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
9723 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
9724 that it can utilize the decrement and branch instruction, but will give
9725 up if there is more than one memory reference in the loop. Thus a loop
9726 where the loop counter is decremented can generate slightly more
9727 efficient code, in cases where the RPTB instruction cannot be utilized.
9729 @item -mdp-isr-reload
9731 @opindex mdp-isr-reload
9733 Force the DP register to be saved on entry to an interrupt service
9734 routine (ISR), reloaded to point to the data section, and restored on
9735 exit from the ISR@. This should not be required unless someone has
9736 violated the small memory model by modifying the DP register, say within
9743 For the C3x use the 24-bit MPYI instruction for integer multiplies
9744 instead of a library call to guarantee 32-bit results. Note that if one
9745 of the operands is a constant, then the multiplication will be performed
9746 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
9747 then squaring operations are performed inline instead of a library call.
9750 @itemx -mno-fast-fix
9752 @opindex mno-fast-fix
9753 The C3x/C4x FIX instruction to convert a floating point value to an
9754 integer value chooses the nearest integer less than or equal to the
9755 floating point value rather than to the nearest integer. Thus if the
9756 floating point number is negative, the result will be incorrectly
9757 truncated an additional code is necessary to detect and correct this
9758 case. This option can be used to disable generation of the additional
9759 code required to correct the result.
9765 Enable (disable) generation of repeat block sequences using the RPTB
9766 instruction for zero overhead looping. The RPTB construct is only used
9767 for innermost loops that do not call functions or jump across the loop
9768 boundaries. There is no advantage having nested RPTB loops due to the
9769 overhead required to save and restore the RC, RS, and RE registers.
9770 This is enabled by default with @option{-O2}.
9772 @item -mrpts=@var{count}
9776 Enable (disable) the use of the single instruction repeat instruction
9777 RPTS@. If a repeat block contains a single instruction, and the loop
9778 count can be guaranteed to be less than the value @var{count}, GCC will
9779 emit a RPTS instruction instead of a RPTB@. If no value is specified,
9780 then a RPTS will be emitted even if the loop count cannot be determined
9781 at compile time. Note that the repeated instruction following RPTS does
9782 not have to be reloaded from memory each iteration, thus freeing up the
9783 CPU buses for operands. However, since interrupts are blocked by this
9784 instruction, it is disabled by default.
9786 @item -mloop-unsigned
9787 @itemx -mno-loop-unsigned
9788 @opindex mloop-unsigned
9789 @opindex mno-loop-unsigned
9790 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
9791 is @math{2^{31} + 1} since these instructions test if the iteration count is
9792 negative to terminate the loop. If the iteration count is unsigned
9793 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
9794 exceeded. This switch allows an unsigned iteration count.
9798 Try to emit an assembler syntax that the TI assembler (asm30) is happy
9799 with. This also enforces compatibility with the API employed by the TI
9800 C3x C compiler. For example, long doubles are passed as structures
9801 rather than in floating point registers.
9807 Generate code that uses registers (stack) for passing arguments to functions.
9808 By default, arguments are passed in registers where possible rather
9809 than by pushing arguments on to the stack.
9811 @item -mparallel-insns
9812 @itemx -mno-parallel-insns
9813 @opindex mparallel-insns
9814 @opindex mno-parallel-insns
9815 Allow the generation of parallel instructions. This is enabled by
9816 default with @option{-O2}.
9818 @item -mparallel-mpy
9819 @itemx -mno-parallel-mpy
9820 @opindex mparallel-mpy
9821 @opindex mno-parallel-mpy
9822 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
9823 provided @option{-mparallel-insns} is also specified. These instructions have
9824 tight register constraints which can pessimize the code generation
9830 @subsection V850 Options
9831 @cindex V850 Options
9833 These @samp{-m} options are defined for V850 implementations:
9837 @itemx -mno-long-calls
9838 @opindex mlong-calls
9839 @opindex mno-long-calls
9840 Treat all calls as being far away (near). If calls are assumed to be
9841 far away, the compiler will always load the functions address up into a
9842 register, and call indirect through the pointer.
9848 Do not optimize (do optimize) basic blocks that use the same index
9849 pointer 4 or more times to copy pointer into the @code{ep} register, and
9850 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
9851 option is on by default if you optimize.
9853 @item -mno-prolog-function
9854 @itemx -mprolog-function
9855 @opindex mno-prolog-function
9856 @opindex mprolog-function
9857 Do not use (do use) external functions to save and restore registers
9858 at the prologue and epilogue of a function. The external functions
9859 are slower, but use less code space if more than one function saves
9860 the same number of registers. The @option{-mprolog-function} option
9861 is on by default if you optimize.
9865 Try to make the code as small as possible. At present, this just turns
9866 on the @option{-mep} and @option{-mprolog-function} options.
9870 Put static or global variables whose size is @var{n} bytes or less into
9871 the tiny data area that register @code{ep} points to. The tiny data
9872 area can hold up to 256 bytes in total (128 bytes for byte references).
9876 Put static or global variables whose size is @var{n} bytes or less into
9877 the small data area that register @code{gp} points to. The small data
9878 area can hold up to 64 kilobytes.
9882 Put static or global variables whose size is @var{n} bytes or less into
9883 the first 32 kilobytes of memory.
9887 Specify that the target processor is the V850.
9890 @opindex mbig-switch
9891 Generate code suitable for big switch tables. Use this option only if
9892 the assembler/linker complain about out of range branches within a switch
9897 This option will cause r2 and r5 to be used in the code generated by
9898 the compiler. This setting is the default.
9901 @opindex mno-app-regs
9902 This option will cause r2 and r5 to be treated as fixed registers.
9906 Specify that the target processor is the V850E1. The preprocessor
9907 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
9908 this option is used.
9912 Specify that the target processor is the V850E. The preprocessor
9913 constant @samp{__v850e__} will be defined if this option is used.
9915 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
9916 are defined then a default target processor will be chosen and the
9917 relevant @samp{__v850*__} preprocessor constant will be defined.
9919 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
9920 defined, regardless of which processor variant is the target.
9922 @item -mdisable-callt
9923 @opindex mdisable-callt
9924 This option will suppress generation of the CALLT instruction for the
9925 v850e and v850e1 flavors of the v850 architecture. The default is
9926 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
9931 @subsection ARC Options
9934 These options are defined for ARC implementations:
9939 Compile code for little endian mode. This is the default.
9943 Compile code for big endian mode.
9946 @opindex mmangle-cpu
9947 Prepend the name of the cpu to all public symbol names.
9948 In multiple-processor systems, there are many ARC variants with different
9949 instruction and register set characteristics. This flag prevents code
9950 compiled for one cpu to be linked with code compiled for another.
9951 No facility exists for handling variants that are ``almost identical''.
9952 This is an all or nothing option.
9954 @item -mcpu=@var{cpu}
9956 Compile code for ARC variant @var{cpu}.
9957 Which variants are supported depend on the configuration.
9958 All variants support @option{-mcpu=base}, this is the default.
9960 @item -mtext=@var{text-section}
9961 @itemx -mdata=@var{data-section}
9962 @itemx -mrodata=@var{readonly-data-section}
9966 Put functions, data, and readonly data in @var{text-section},
9967 @var{data-section}, and @var{readonly-data-section} respectively
9968 by default. This can be overridden with the @code{section} attribute.
9969 @xref{Variable Attributes}.
9974 @subsection NS32K Options
9975 @cindex NS32K options
9977 These are the @samp{-m} options defined for the 32000 series. The default
9978 values for these options depends on which style of 32000 was selected when
9979 the compiler was configured; the defaults for the most common choices are
9987 Generate output for a 32032. This is the default
9988 when the compiler is configured for 32032 and 32016 based systems.
9994 Generate output for a 32332. This is the default
9995 when the compiler is configured for 32332-based systems.
10001 Generate output for a 32532. This is the default
10002 when the compiler is configured for 32532-based systems.
10006 Generate output containing 32081 instructions for floating point.
10007 This is the default for all systems.
10011 Generate output containing 32381 instructions for floating point. This
10012 also implies @option{-m32081}. The 32381 is only compatible with the 32332
10013 and 32532 cpus. This is the default for the pc532-netbsd configuration.
10016 @opindex mmulti-add
10017 Try and generate multiply-add floating point instructions @code{polyF}
10018 and @code{dotF}. This option is only available if the @option{-m32381}
10019 option is in effect. Using these instructions requires changes to
10020 register allocation which generally has a negative impact on
10021 performance. This option should only be enabled when compiling code
10022 particularly likely to make heavy use of multiply-add instructions.
10024 @item -mnomulti-add
10025 @opindex mnomulti-add
10026 Do not try and generate multiply-add floating point instructions
10027 @code{polyF} and @code{dotF}. This is the default on all platforms.
10030 @opindex msoft-float
10031 Generate output containing library calls for floating point.
10032 @strong{Warning:} the requisite libraries may not be available.
10034 @item -mieee-compare
10035 @itemx -mno-ieee-compare
10036 @opindex mieee-compare
10037 @opindex mno-ieee-compare
10038 Control whether or not the compiler uses IEEE floating point
10039 comparisons. These handle correctly the case where the result of a
10040 comparison is unordered.
10041 @strong{Warning:} the requisite kernel support may not be available.
10044 @opindex mnobitfield
10045 Do not use the bit-field instructions. On some machines it is faster to
10046 use shifting and masking operations. This is the default for the pc532.
10050 Do use the bit-field instructions. This is the default for all platforms
10055 Use a different function-calling convention, in which functions
10056 that take a fixed number of arguments return pop their
10057 arguments on return with the @code{ret} instruction.
10059 This calling convention is incompatible with the one normally
10060 used on Unix, so you cannot use it if you need to call libraries
10061 compiled with the Unix compiler.
10063 Also, you must provide function prototypes for all functions that
10064 take variable numbers of arguments (including @code{printf});
10065 otherwise incorrect code will be generated for calls to those
10068 In addition, seriously incorrect code will result if you call a
10069 function with too many arguments. (Normally, extra arguments are
10070 harmlessly ignored.)
10072 This option takes its name from the 680x0 @code{rtd} instruction.
10077 Use a different function-calling convention where the first two arguments
10078 are passed in registers.
10080 This calling convention is incompatible with the one normally
10081 used on Unix, so you cannot use it if you need to call libraries
10082 compiled with the Unix compiler.
10085 @opindex mnoregparam
10086 Do not pass any arguments in registers. This is the default for all
10091 It is OK to use the sb as an index register which is always loaded with
10092 zero. This is the default for the pc532-netbsd target.
10096 The sb register is not available for use or has not been initialized to
10097 zero by the run time system. This is the default for all targets except
10098 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10099 @option{-fpic} is set.
10103 Many ns32000 series addressing modes use displacements of up to 512MB@.
10104 If an address is above 512MB then displacements from zero can not be used.
10105 This option causes code to be generated which can be loaded above 512MB@.
10106 This may be useful for operating systems or ROM code.
10110 Assume code will be loaded in the first 512MB of virtual address space.
10111 This is the default for all platforms.
10117 @subsection AVR Options
10118 @cindex AVR Options
10120 These options are defined for AVR implementations:
10123 @item -mmcu=@var{mcu}
10125 Specify ATMEL AVR instruction set or MCU type.
10127 Instruction set avr1 is for the minimal AVR core, not supported by the C
10128 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10129 attiny11, attiny12, attiny15, attiny28).
10131 Instruction set avr2 (default) is for the classic AVR core with up to
10132 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10133 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10134 at90c8534, at90s8535).
10136 Instruction set avr3 is for the classic AVR core with up to 128K program
10137 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10139 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10140 memory space (MCU types: atmega8, atmega83, atmega85).
10142 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10143 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10144 atmega64, atmega128, at43usb355, at94k).
10148 Output instruction sizes to the asm file.
10150 @item -minit-stack=@var{N}
10151 @opindex minit-stack
10152 Specify the initial stack address, which may be a symbol or numeric value,
10153 @samp{__stack} is the default.
10155 @item -mno-interrupts
10156 @opindex mno-interrupts
10157 Generated code is not compatible with hardware interrupts.
10158 Code size will be smaller.
10160 @item -mcall-prologues
10161 @opindex mcall-prologues
10162 Functions prologues/epilogues expanded as call to appropriate
10163 subroutines. Code size will be smaller.
10165 @item -mno-tablejump
10166 @opindex mno-tablejump
10167 Do not generate tablejump insns which sometimes increase code size.
10170 @opindex mtiny-stack
10171 Change only the low 8 bits of the stack pointer.
10174 @node MCore Options
10175 @subsection MCore Options
10176 @cindex MCore options
10178 These are the @samp{-m} options defined for the Motorola M*Core
10184 @itemx -mno-hardlit
10186 @opindex mno-hardlit
10187 Inline constants into the code stream if it can be done in two
10188 instructions or less.
10194 Use the divide instruction. (Enabled by default).
10196 @item -mrelax-immediate
10197 @itemx -mno-relax-immediate
10198 @opindex mrelax-immediate
10199 @opindex mno-relax-immediate
10200 Allow arbitrary sized immediates in bit operations.
10202 @item -mwide-bitfields
10203 @itemx -mno-wide-bitfields
10204 @opindex mwide-bitfields
10205 @opindex mno-wide-bitfields
10206 Always treat bit-fields as int-sized.
10208 @item -m4byte-functions
10209 @itemx -mno-4byte-functions
10210 @opindex m4byte-functions
10211 @opindex mno-4byte-functions
10212 Force all functions to be aligned to a four byte boundary.
10214 @item -mcallgraph-data
10215 @itemx -mno-callgraph-data
10216 @opindex mcallgraph-data
10217 @opindex mno-callgraph-data
10218 Emit callgraph information.
10221 @itemx -mno-slow-bytes
10222 @opindex mslow-bytes
10223 @opindex mno-slow-bytes
10224 Prefer word access when reading byte quantities.
10226 @item -mlittle-endian
10227 @itemx -mbig-endian
10228 @opindex mlittle-endian
10229 @opindex mbig-endian
10230 Generate code for a little endian target.
10236 Generate code for the 210 processor.
10239 @node IA-64 Options
10240 @subsection IA-64 Options
10241 @cindex IA-64 Options
10243 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10247 @opindex mbig-endian
10248 Generate code for a big endian target. This is the default for HP-UX@.
10250 @item -mlittle-endian
10251 @opindex mlittle-endian
10252 Generate code for a little endian target. This is the default for AIX5
10258 @opindex mno-gnu-as
10259 Generate (or don't) code for the GNU assembler. This is the default.
10260 @c Also, this is the default if the configure option @option{--with-gnu-as}
10266 @opindex mno-gnu-ld
10267 Generate (or don't) code for the GNU linker. This is the default.
10268 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10273 Generate code that does not use a global pointer register. The result
10274 is not position independent code, and violates the IA-64 ABI@.
10276 @item -mvolatile-asm-stop
10277 @itemx -mno-volatile-asm-stop
10278 @opindex mvolatile-asm-stop
10279 @opindex mno-volatile-asm-stop
10280 Generate (or don't) a stop bit immediately before and after volatile asm
10285 Generate code that works around Itanium B step errata.
10287 @item -mregister-names
10288 @itemx -mno-register-names
10289 @opindex mregister-names
10290 @opindex mno-register-names
10291 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10292 the stacked registers. This may make assembler output more readable.
10298 Disable (or enable) optimizations that use the small data section. This may
10299 be useful for working around optimizer bugs.
10301 @item -mconstant-gp
10302 @opindex mconstant-gp
10303 Generate code that uses a single constant global pointer value. This is
10304 useful when compiling kernel code.
10308 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10309 This is useful when compiling firmware code.
10311 @item -minline-float-divide-min-latency
10312 @opindex minline-float-divide-min-latency
10313 Generate code for inline divides of floating point values
10314 using the minimum latency algorithm.
10316 @item -minline-float-divide-max-throughput
10317 @opindex minline-float-divide-max-throughput
10318 Generate code for inline divides of floating point values
10319 using the maximum throughput algorithm.
10321 @item -minline-int-divide-min-latency
10322 @opindex minline-int-divide-min-latency
10323 Generate code for inline divides of integer values
10324 using the minimum latency algorithm.
10326 @item -minline-int-divide-max-throughput
10327 @opindex minline-int-divide-max-throughput
10328 Generate code for inline divides of integer values
10329 using the maximum throughput algorithm.
10331 @item -mno-dwarf2-asm
10332 @itemx -mdwarf2-asm
10333 @opindex mno-dwarf2-asm
10334 @opindex mdwarf2-asm
10335 Don't (or do) generate assembler code for the DWARF2 line number debugging
10336 info. This may be useful when not using the GNU assembler.
10338 @item -mfixed-range=@var{register-range}
10339 @opindex mfixed-range
10340 Generate code treating the given register range as fixed registers.
10341 A fixed register is one that the register allocator can not use. This is
10342 useful when compiling kernel code. A register range is specified as
10343 two registers separated by a dash. Multiple register ranges can be
10344 specified separated by a comma.
10346 @item -mearly-stop-bits
10347 @itemx -mno-early-stop-bits
10348 @opindex mearly-stop-bits
10349 @opindex mno-early-stop-bits
10350 Allow stop bits to be placed earlier than immediately preceding the
10351 instruction that triggered the stop bit. This can improve instruction
10352 scheduling, but does not always do so.
10356 @subsection D30V Options
10357 @cindex D30V Options
10359 These @samp{-m} options are defined for D30V implementations:
10364 Link the @samp{.text}, @samp{.data}, @samp{.bss}, @samp{.strings},
10365 @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections into external
10366 memory, which starts at location @code{0x80000000}.
10369 @opindex mextmemory
10370 Same as the @option{-mextmem} switch.
10374 Link the @samp{.text} section into onchip text memory, which starts at
10375 location @code{0x0}. Also link @samp{.data}, @samp{.bss},
10376 @samp{.strings}, @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections
10377 into onchip data memory, which starts at location @code{0x20000000}.
10379 @item -mno-asm-optimize
10380 @itemx -masm-optimize
10381 @opindex mno-asm-optimize
10382 @opindex masm-optimize
10383 Disable (enable) passing @option{-O} to the assembler when optimizing.
10384 The assembler uses the @option{-O} option to automatically parallelize
10385 adjacent short instructions where possible.
10387 @item -mbranch-cost=@var{n}
10388 @opindex mbranch-cost
10389 Increase the internal costs of branches to @var{n}. Higher costs means
10390 that the compiler will issue more instructions to avoid doing a branch.
10393 @item -mcond-exec=@var{n}
10394 @opindex mcond-exec
10395 Specify the maximum number of conditionally executed instructions that
10396 replace a branch. The default is 4.
10399 @node S/390 and zSeries Options
10400 @subsection S/390 and zSeries Options
10401 @cindex S/390 and zSeries Options
10403 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
10407 @itemx -msoft-float
10408 @opindex mhard-float
10409 @opindex msoft-float
10410 Use (do not use) the hardware floating-point instructions and registers
10411 for floating-point operations. When @option{-msoft-float} is specified,
10412 functions in @file{libgcc.a} will be used to perform floating-point
10413 operations. When @option{-mhard-float} is specified, the compiler
10414 generates IEEE floating-point instructions. This is the default.
10417 @itemx -mno-backchain
10418 @opindex mbackchain
10419 @opindex mno-backchain
10420 Generate (or do not generate) code which maintains an explicit
10421 backchain within the stack frame that points to the caller's frame.
10422 This is currently needed to allow debugging. The default is to
10423 generate the backchain.
10426 @itemx -mno-small-exec
10427 @opindex msmall-exec
10428 @opindex mno-small-exec
10429 Generate (or do not generate) code using the @code{bras} instruction
10430 to do subroutine calls.
10431 This only works reliably if the total executable size does not
10432 exceed 64k. The default is to use the @code{basr} instruction instead,
10433 which does not have this limitation.
10439 When @option{-m31} is specified, generate code compliant to the
10440 Linux for S/390 ABI@. When @option{-m64} is specified, generate
10441 code compliant to the Linux for zSeries ABI@. This allows GCC in
10442 particular to generate 64-bit instructions. For the @samp{s390}
10443 targets, the default is @option{-m31}, while the @samp{s390x}
10444 targets default to @option{-m64}.
10450 When @option{-mzarch} is specified, generate code using the
10451 instructions available on z/Architecture.
10452 When @option{-mesa} is specified, generate code using the
10453 instructions available on ESA/390. Note that @option{-mesa} is
10454 not possible with @option{-m64}.
10455 When generating code compliant to the Linux for S/390 ABI,
10456 the default is @option{-mesa}. When generating code compliant
10457 to the Linux for zSeries ABI, the default is @option{-mzarch}.
10463 Generate (or do not generate) code using the @code{mvcle} instruction
10464 to perform block moves. When @option{-mno-mvcle} is specified,
10465 use a @code{mvc} loop instead. This is the default.
10471 Print (or do not print) additional debug information when compiling.
10472 The default is to not print debug information.
10474 @item -march=@var{cpu-type}
10476 Generate code that will run on @var{cpu-type}, which is the name of a system
10477 representing a certain processor type. Possible values for
10478 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
10479 When generating code using the instructions available on z/Architecture,
10480 the default is @option{-march=z900}. Otherwise, the default is
10481 @option{-march=g5}.
10483 @item -mtune=@var{cpu-type}
10485 Tune to @var{cpu-type} everything applicable about the generated code,
10486 except for the ABI and the set of available instructions.
10487 The list of @var{cpu-type} values is the same as for @option{-march}.
10488 The default is the value used for @option{-march}.
10491 @itemx -mno-fused-madd
10492 @opindex mfused-madd
10493 @opindex mno-fused-madd
10494 Generate code that uses (does not use) the floating point multiply and
10495 accumulate instructions. These instructions are generated by default if
10496 hardware floating point is used.
10500 @subsection CRIS Options
10501 @cindex CRIS Options
10503 These options are defined specifically for the CRIS ports.
10506 @item -march=@var{architecture-type}
10507 @itemx -mcpu=@var{architecture-type}
10510 Generate code for the specified architecture. The choices for
10511 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10512 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX.
10513 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10516 @item -mtune=@var{architecture-type}
10518 Tune to @var{architecture-type} everything applicable about the generated
10519 code, except for the ABI and the set of available instructions. The
10520 choices for @var{architecture-type} are the same as for
10521 @option{-march=@var{architecture-type}}.
10523 @item -mmax-stack-frame=@var{n}
10524 @opindex mmax-stack-frame
10525 Warn when the stack frame of a function exceeds @var{n} bytes.
10527 @item -melinux-stacksize=@var{n}
10528 @opindex melinux-stacksize
10529 Only available with the @samp{cris-axis-aout} target. Arranges for
10530 indications in the program to the kernel loader that the stack of the
10531 program should be set to @var{n} bytes.
10537 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10538 @option{-march=v3} and @option{-march=v8} respectively.
10542 Enable CRIS-specific verbose debug-related information in the assembly
10543 code. This option also has the effect to turn off the @samp{#NO_APP}
10544 formatted-code indicator to the assembler at the beginning of the
10549 Do not use condition-code results from previous instruction; always emit
10550 compare and test instructions before use of condition codes.
10552 @item -mno-side-effects
10553 @opindex mno-side-effects
10554 Do not emit instructions with side-effects in addressing modes other than
10557 @item -mstack-align
10558 @itemx -mno-stack-align
10559 @itemx -mdata-align
10560 @itemx -mno-data-align
10561 @itemx -mconst-align
10562 @itemx -mno-const-align
10563 @opindex mstack-align
10564 @opindex mno-stack-align
10565 @opindex mdata-align
10566 @opindex mno-data-align
10567 @opindex mconst-align
10568 @opindex mno-const-align
10569 These options (no-options) arranges (eliminate arrangements) for the
10570 stack-frame, individual data and constants to be aligned for the maximum
10571 single data access size for the chosen CPU model. The default is to
10572 arrange for 32-bit alignment. ABI details such as structure layout are
10573 not affected by these options.
10581 Similar to the stack- data- and const-align options above, these options
10582 arrange for stack-frame, writable data and constants to all be 32-bit,
10583 16-bit or 8-bit aligned. The default is 32-bit alignment.
10585 @item -mno-prologue-epilogue
10586 @itemx -mprologue-epilogue
10587 @opindex mno-prologue-epilogue
10588 @opindex mprologue-epilogue
10589 With @option{-mno-prologue-epilogue}, the normal function prologue and
10590 epilogue that sets up the stack-frame are omitted and no return
10591 instructions or return sequences are generated in the code. Use this
10592 option only together with visual inspection of the compiled code: no
10593 warnings or errors are generated when call-saved registers must be saved,
10594 or storage for local variable needs to be allocated.
10598 @opindex mno-gotplt
10600 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10601 instruction sequences that load addresses for functions from the PLT part
10602 of the GOT rather than (traditional on other architectures) calls to the
10603 PLT. The default is @option{-mgotplt}.
10607 Legacy no-op option only recognized with the cris-axis-aout target.
10611 Legacy no-op option only recognized with the cris-axis-elf and
10612 cris-axis-linux-gnu targets.
10616 Only recognized with the cris-axis-aout target, where it selects a
10617 GNU/linux-like multilib, include files and instruction set for
10618 @option{-march=v8}.
10622 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10626 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
10627 to link with input-output functions from a simulator library. Code,
10628 initialized data and zero-initialized data are allocated consecutively.
10632 Like @option{-sim}, but pass linker options to locate initialized data at
10633 0x40000000 and zero-initialized data at 0x80000000.
10637 @subsection MMIX Options
10638 @cindex MMIX Options
10640 These options are defined for the MMIX:
10644 @itemx -mno-libfuncs
10646 @opindex mno-libfuncs
10647 Specify that intrinsic library functions are being compiled, passing all
10648 values in registers, no matter the size.
10651 @itemx -mno-epsilon
10653 @opindex mno-epsilon
10654 Generate floating-point comparison instructions that compare with respect
10655 to the @code{rE} epsilon register.
10657 @item -mabi=mmixware
10659 @opindex mabi-mmixware
10661 Generate code that passes function parameters and return values that (in
10662 the called function) are seen as registers @code{$0} and up, as opposed to
10663 the GNU ABI which uses global registers @code{$231} and up.
10665 @item -mzero-extend
10666 @itemx -mno-zero-extend
10667 @opindex mzero-extend
10668 @opindex mno-zero-extend
10669 When reading data from memory in sizes shorter than 64 bits, use (do not
10670 use) zero-extending load instructions by default, rather than
10671 sign-extending ones.
10674 @itemx -mno-knuthdiv
10676 @opindex mno-knuthdiv
10677 Make the result of a division yielding a remainder have the same sign as
10678 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10679 remainder follows the sign of the dividend. Both methods are
10680 arithmetically valid, the latter being almost exclusively used.
10682 @item -mtoplevel-symbols
10683 @itemx -mno-toplevel-symbols
10684 @opindex mtoplevel-symbols
10685 @opindex mno-toplevel-symbols
10686 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10687 code can be used with the @code{PREFIX} assembly directive.
10691 Generate an executable in the ELF format, rather than the default
10692 @samp{mmo} format used by the @command{mmix} simulator.
10694 @item -mbranch-predict
10695 @itemx -mno-branch-predict
10696 @opindex mbranch-predict
10697 @opindex mno-branch-predict
10698 Use (do not use) the probable-branch instructions, when static branch
10699 prediction indicates a probable branch.
10701 @item -mbase-addresses
10702 @itemx -mno-base-addresses
10703 @opindex mbase-addresses
10704 @opindex mno-base-addresses
10705 Generate (do not generate) code that uses @emph{base addresses}. Using a
10706 base address automatically generates a request (handled by the assembler
10707 and the linker) for a constant to be set up in a global register. The
10708 register is used for one or more base address requests within the range 0
10709 to 255 from the value held in the register. The generally leads to short
10710 and fast code, but the number of different data items that can be
10711 addressed is limited. This means that a program that uses lots of static
10712 data may require @option{-mno-base-addresses}.
10714 @item -msingle-exit
10715 @itemx -mno-single-exit
10716 @opindex msingle-exit
10717 @opindex mno-single-exit
10718 Force (do not force) generated code to have a single exit point in each
10722 @node PDP-11 Options
10723 @subsection PDP-11 Options
10724 @cindex PDP-11 Options
10726 These options are defined for the PDP-11:
10731 Use hardware FPP floating point. This is the default. (FIS floating
10732 point on the PDP-11/40 is not supported.)
10735 @opindex msoft-float
10736 Do not use hardware floating point.
10740 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10744 Return floating-point results in memory. This is the default.
10748 Generate code for a PDP-11/40.
10752 Generate code for a PDP-11/45. This is the default.
10756 Generate code for a PDP-11/10.
10758 @item -mbcopy-builtin
10759 @opindex bcopy-builtin
10760 Use inline @code{movstrhi} patterns for copying memory. This is the
10765 Do not use inline @code{movstrhi} patterns for copying memory.
10771 Use 16-bit @code{int}. This is the default.
10777 Use 32-bit @code{int}.
10780 @itemx -mno-float32
10782 @opindex mno-float32
10783 Use 64-bit @code{float}. This is the default.
10786 @itemx -mno-float64
10788 @opindex mno-float64
10789 Use 32-bit @code{float}.
10793 Use @code{abshi2} pattern. This is the default.
10797 Do not use @code{abshi2} pattern.
10799 @item -mbranch-expensive
10800 @opindex mbranch-expensive
10801 Pretend that branches are expensive. This is for experimenting with
10802 code generation only.
10804 @item -mbranch-cheap
10805 @opindex mbranch-cheap
10806 Do not pretend that branches are expensive. This is the default.
10810 Generate code for a system with split I&D.
10814 Generate code for a system without split I&D. This is the default.
10818 Use Unix assembler syntax. This is the default when configured for
10819 @samp{pdp11-*-bsd}.
10823 Use DEC assembler syntax. This is the default when configured for any
10824 PDP-11 target other than @samp{pdp11-*-bsd}.
10827 @node Xstormy16 Options
10828 @subsection Xstormy16 Options
10829 @cindex Xstormy16 Options
10831 These options are defined for Xstormy16:
10836 Choose startup files and linker script suitable for the simulator.
10840 @subsection FRV Options
10841 @cindex FRV Options
10847 Only use the first 32 general purpose registers.
10852 Use all 64 general purpose registers.
10857 Use only the first 32 floating point registers.
10862 Use all 64 floating point registers
10865 @opindex mhard-float
10867 Use hardware instructions for floating point operations.
10870 @opindex msoft-float
10872 Use library routines for floating point operations.
10877 Dynamically allocate condition code registers.
10882 Do not try to dynamically allocate condition code registers, only
10883 use @code{icc0} and @code{fcc0}.
10888 Change ABI to use double word insns.
10893 Do not use double word instructions.
10898 Use floating point double instructions.
10901 @opindex mno-double
10903 Do not use floating point double instructions.
10908 Use media instructions.
10913 Do not use media instructions.
10918 Use multiply and add/subtract instructions.
10921 @opindex mno-muladd
10923 Do not use multiply and add/subtract instructions.
10925 @item -mlibrary-pic
10926 @opindex mlibrary-pic
10928 Enable PIC support for building libraries
10933 Use only the first four media accumulator registers.
10938 Use all eight media accumulator registers.
10943 Pack VLIW instructions.
10948 Do not pack VLIW instructions.
10951 @opindex mno-eflags
10953 Do not mark ABI switches in e_flags.
10956 @opindex mcond-move
10958 Enable the use of conditional-move instructions (default).
10960 This switch is mainly for debugging the compiler and will likely be removed
10961 in a future version.
10963 @item -mno-cond-move
10964 @opindex mno-cond-move
10966 Disable the use of conditional-move instructions.
10968 This switch is mainly for debugging the compiler and will likely be removed
10969 in a future version.
10974 Enable the use of conditional set instructions (default).
10976 This switch is mainly for debugging the compiler and will likely be removed
10977 in a future version.
10982 Disable the use of conditional set instructions.
10984 This switch is mainly for debugging the compiler and will likely be removed
10985 in a future version.
10988 @opindex mcond-exec
10990 Enable the use of conditional execution (default).
10992 This switch is mainly for debugging the compiler and will likely be removed
10993 in a future version.
10995 @item -mno-cond-exec
10996 @opindex mno-cond-exec
10998 Disable the use of conditional execution.
11000 This switch is mainly for debugging the compiler and will likely be removed
11001 in a future version.
11003 @item -mvliw-branch
11004 @opindex mvliw-branch
11006 Run a pass to pack branches into VLIW instructions (default).
11008 This switch is mainly for debugging the compiler and will likely be removed
11009 in a future version.
11011 @item -mno-vliw-branch
11012 @opindex mno-vliw-branch
11014 Do not run a pass to pack branches into VLIW instructions.
11016 This switch is mainly for debugging the compiler and will likely be removed
11017 in a future version.
11019 @item -mmulti-cond-exec
11020 @opindex mmulti-cond-exec
11022 Enable optimization of @code{&&} and @code{||} in conditional execution
11025 This switch is mainly for debugging the compiler and will likely be removed
11026 in a future version.
11028 @item -mno-multi-cond-exec
11029 @opindex mno-multi-cond-exec
11031 Disable optimization of @code{&&} and @code{||} in conditional execution.
11033 This switch is mainly for debugging the compiler and will likely be removed
11034 in a future version.
11036 @item -mnested-cond-exec
11037 @opindex mnested-cond-exec
11039 Enable nested conditional execution optimizations (default).
11041 This switch is mainly for debugging the compiler and will likely be removed
11042 in a future version.
11044 @item -mno-nested-cond-exec
11045 @opindex mno-nested-cond-exec
11047 Disable nested conditional execution optimizations.
11049 This switch is mainly for debugging the compiler and will likely be removed
11050 in a future version.
11052 @item -mtomcat-stats
11053 @opindex mtomcat-stats
11055 Cause gas to print out tomcat statistics.
11057 @item -mcpu=@var{cpu}
11060 Select the processor type for which to generate code. Possible values are
11061 @samp{simple}, @samp{tomcat}, @samp{fr500}, @samp{fr400}, @samp{fr300},
11066 @node Xtensa Options
11067 @subsection Xtensa Options
11068 @cindex Xtensa Options
11070 These options are supported for Xtensa targets:
11074 @itemx -mno-const16
11076 @opindex mno-const16
11077 Enable or disable use of @code{CONST16} instructions for loading
11078 constant values. The @code{CONST16} instruction is currently not a
11079 standard option from Tensilica. When enabled, @code{CONST16}
11080 instructions are always used in place of the standard @code{L32R}
11081 instructions. The use of @code{CONST16} is enabled by default only if
11082 the @code{L32R} instruction is not available.
11085 @itemx -mno-fused-madd
11086 @opindex mfused-madd
11087 @opindex mno-fused-madd
11088 Enable or disable use of fused multiply/add and multiply/subtract
11089 instructions in the floating-point option. This has no effect if the
11090 floating-point option is not also enabled. Disabling fused multiply/add
11091 and multiply/subtract instructions forces the compiler to use separate
11092 instructions for the multiply and add/subtract operations. This may be
11093 desirable in some cases where strict IEEE 754-compliant results are
11094 required: the fused multiply add/subtract instructions do not round the
11095 intermediate result, thereby producing results with @emph{more} bits of
11096 precision than specified by the IEEE standard. Disabling fused multiply
11097 add/subtract instructions also ensures that the program output is not
11098 sensitive to the compiler's ability to combine multiply and add/subtract
11101 @item -mtext-section-literals
11102 @itemx -mno-text-section-literals
11103 @opindex mtext-section-literals
11104 @opindex mno-text-section-literals
11105 Control the treatment of literal pools. The default is
11106 @option{-mno-text-section-literals}, which places literals in a separate
11107 section in the output file. This allows the literal pool to be placed
11108 in a data RAM/ROM, and it also allows the linker to combine literal
11109 pools from separate object files to remove redundant literals and
11110 improve code size. With @option{-mtext-section-literals}, the literals
11111 are interspersed in the text section in order to keep them as close as
11112 possible to their references. This may be necessary for large assembly
11115 @item -mtarget-align
11116 @itemx -mno-target-align
11117 @opindex mtarget-align
11118 @opindex mno-target-align
11119 When this option is enabled, GCC instructs the assembler to
11120 automatically align instructions to reduce branch penalties at the
11121 expense of some code density. The assembler attempts to widen density
11122 instructions to align branch targets and the instructions following call
11123 instructions. If there are not enough preceding safe density
11124 instructions to align a target, no widening will be performed. The
11125 default is @option{-mtarget-align}. These options do not affect the
11126 treatment of auto-aligned instructions like @code{LOOP}, which the
11127 assembler will always align, either by widening density instructions or
11128 by inserting no-op instructions.
11131 @itemx -mno-longcalls
11132 @opindex mlongcalls
11133 @opindex mno-longcalls
11134 When this option is enabled, GCC instructs the assembler to translate
11135 direct calls to indirect calls unless it can determine that the target
11136 of a direct call is in the range allowed by the call instruction. This
11137 translation typically occurs for calls to functions in other source
11138 files. Specifically, the assembler translates a direct @code{CALL}
11139 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
11140 The default is @option{-mno-longcalls}. This option should be used in
11141 programs where the call target can potentially be out of range. This
11142 option is implemented in the assembler, not the compiler, so the
11143 assembly code generated by GCC will still show direct call
11144 instructions---look at the disassembled object code to see the actual
11145 instructions. Note that the assembler will use an indirect call for
11146 every cross-file call, not just those that really will be out of range.
11149 @node Code Gen Options
11150 @section Options for Code Generation Conventions
11151 @cindex code generation conventions
11152 @cindex options, code generation
11153 @cindex run-time options
11155 These machine-independent options control the interface conventions
11156 used in code generation.
11158 Most of them have both positive and negative forms; the negative form
11159 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
11160 one of the forms is listed---the one which is not the default. You
11161 can figure out the other form by either removing @samp{no-} or adding
11165 @item -fbounds-check
11166 @opindex fbounds-check
11167 For front-ends that support it, generate additional code to check that
11168 indices used to access arrays are within the declared range. This is
11169 currently only supported by the Java and Fortran 77 front-ends, where
11170 this option defaults to true and false respectively.
11174 This option generates traps for signed overflow on addition, subtraction,
11175 multiplication operations.
11179 This option instructs the compiler to assume that signed arithmetic
11180 overflow of addition, subtraction and multiplication wraps around
11181 using twos-complement representation. This flag enables some optimizations
11182 and disables other. This option is enabled by default for the Java
11183 front-end, as required by the Java language specification.
11186 @opindex fexceptions
11187 Enable exception handling. Generates extra code needed to propagate
11188 exceptions. For some targets, this implies GCC will generate frame
11189 unwind information for all functions, which can produce significant data
11190 size overhead, although it does not affect execution. If you do not
11191 specify this option, GCC will enable it by default for languages like
11192 C++ which normally require exception handling, and disable it for
11193 languages like C that do not normally require it. However, you may need
11194 to enable this option when compiling C code that needs to interoperate
11195 properly with exception handlers written in C++. You may also wish to
11196 disable this option if you are compiling older C++ programs that don't
11197 use exception handling.
11199 @item -fnon-call-exceptions
11200 @opindex fnon-call-exceptions
11201 Generate code that allows trapping instructions to throw exceptions.
11202 Note that this requires platform-specific runtime support that does
11203 not exist everywhere. Moreover, it only allows @emph{trapping}
11204 instructions to throw exceptions, i.e.@: memory references or floating
11205 point instructions. It does not allow exceptions to be thrown from
11206 arbitrary signal handlers such as @code{SIGALRM}.
11208 @item -funwind-tables
11209 @opindex funwind-tables
11210 Similar to @option{-fexceptions}, except that it will just generate any needed
11211 static data, but will not affect the generated code in any other way.
11212 You will normally not enable this option; instead, a language processor
11213 that needs this handling would enable it on your behalf.
11215 @item -fasynchronous-unwind-tables
11216 @opindex funwind-tables
11217 Generate unwind table in dwarf2 format, if supported by target machine. The
11218 table is exact at each instruction boundary, so it can be used for stack
11219 unwinding from asynchronous events (such as debugger or garbage collector).
11221 @item -fpcc-struct-return
11222 @opindex fpcc-struct-return
11223 Return ``short'' @code{struct} and @code{union} values in memory like
11224 longer ones, rather than in registers. This convention is less
11225 efficient, but it has the advantage of allowing intercallability between
11226 GCC-compiled files and files compiled with other compilers, particularly
11227 the Portable C Compiler (pcc).
11229 The precise convention for returning structures in memory depends
11230 on the target configuration macros.
11232 Short structures and unions are those whose size and alignment match
11233 that of some integer type.
11235 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11236 switch is not binary compatible with code compiled with the
11237 @option{-freg-struct-return} switch.
11238 Use it to conform to a non-default application binary interface.
11240 @item -freg-struct-return
11241 @opindex freg-struct-return
11242 Return @code{struct} and @code{union} values in registers when possible.
11243 This is more efficient for small structures than
11244 @option{-fpcc-struct-return}.
11246 If you specify neither @option{-fpcc-struct-return} nor
11247 @option{-freg-struct-return}, GCC defaults to whichever convention is
11248 standard for the target. If there is no standard convention, GCC
11249 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11250 the principal compiler. In those cases, we can choose the standard, and
11251 we chose the more efficient register return alternative.
11253 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11254 switch is not binary compatible with code compiled with the
11255 @option{-fpcc-struct-return} switch.
11256 Use it to conform to a non-default application binary interface.
11258 @item -fshort-enums
11259 @opindex fshort-enums
11260 Allocate to an @code{enum} type only as many bytes as it needs for the
11261 declared range of possible values. Specifically, the @code{enum} type
11262 will be equivalent to the smallest integer type which has enough room.
11264 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11265 code that is not binary compatible with code generated without that switch.
11266 Use it to conform to a non-default application binary interface.
11268 @item -fshort-double
11269 @opindex fshort-double
11270 Use the same size for @code{double} as for @code{float}.
11272 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
11273 code that is not binary compatible with code generated without that switch.
11274 Use it to conform to a non-default application binary interface.
11276 @item -fshort-wchar
11277 @opindex fshort-wchar
11278 Override the underlying type for @samp{wchar_t} to be @samp{short
11279 unsigned int} instead of the default for the target. This option is
11280 useful for building programs to run under WINE@.
11282 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11283 code that is not binary compatible with code generated without that switch.
11284 Use it to conform to a non-default application binary interface.
11286 @item -fshared-data
11287 @opindex fshared-data
11288 Requests that the data and non-@code{const} variables of this
11289 compilation be shared data rather than private data. The distinction
11290 makes sense only on certain operating systems, where shared data is
11291 shared between processes running the same program, while private data
11292 exists in one copy per process.
11295 @opindex fno-common
11296 In C, allocate even uninitialized global variables in the data section of the
11297 object file, rather than generating them as common blocks. This has the
11298 effect that if the same variable is declared (without @code{extern}) in
11299 two different compilations, you will get an error when you link them.
11300 The only reason this might be useful is if you wish to verify that the
11301 program will work on other systems which always work this way.
11305 Ignore the @samp{#ident} directive.
11307 @item -finhibit-size-directive
11308 @opindex finhibit-size-directive
11309 Don't output a @code{.size} assembler directive, or anything else that
11310 would cause trouble if the function is split in the middle, and the
11311 two halves are placed at locations far apart in memory. This option is
11312 used when compiling @file{crtstuff.c}; you should not need to use it
11315 @item -fverbose-asm
11316 @opindex fverbose-asm
11317 Put extra commentary information in the generated assembly code to
11318 make it more readable. This option is generally only of use to those
11319 who actually need to read the generated assembly code (perhaps while
11320 debugging the compiler itself).
11322 @option{-fno-verbose-asm}, the default, causes the
11323 extra information to be omitted and is useful when comparing two assembler
11328 @cindex global offset table
11330 Generate position-independent code (PIC) suitable for use in a shared
11331 library, if supported for the target machine. Such code accesses all
11332 constant addresses through a global offset table (GOT)@. The dynamic
11333 loader resolves the GOT entries when the program starts (the dynamic
11334 loader is not part of GCC; it is part of the operating system). If
11335 the GOT size for the linked executable exceeds a machine-specific
11336 maximum size, you get an error message from the linker indicating that
11337 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11338 instead. (These maximums are 16k on the m88k, 8k on the SPARC, and 32k
11339 on the m68k and RS/6000. The 386 has no such limit.)
11341 Position-independent code requires special support, and therefore works
11342 only on certain machines. For the 386, GCC supports PIC for System V
11343 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11344 position-independent.
11348 If supported for the target machine, emit position-independent code,
11349 suitable for dynamic linking and avoiding any limit on the size of the
11350 global offset table. This option makes a difference on the m68k, m88k,
11353 Position-independent code requires special support, and therefore works
11354 only on certain machines.
11360 These options are similar to @option{-fpic} and @option{-fPIC}, but
11361 generated position independent code can be only linked into executables.
11362 Usually these options are used when @option{-pie} GCC option will be
11363 used during linking.
11365 @item -ffixed-@var{reg}
11367 Treat the register named @var{reg} as a fixed register; generated code
11368 should never refer to it (except perhaps as a stack pointer, frame
11369 pointer or in some other fixed role).
11371 @var{reg} must be the name of a register. The register names accepted
11372 are machine-specific and are defined in the @code{REGISTER_NAMES}
11373 macro in the machine description macro file.
11375 This flag does not have a negative form, because it specifies a
11378 @item -fcall-used-@var{reg}
11379 @opindex fcall-used
11380 Treat the register named @var{reg} as an allocable register that is
11381 clobbered by function calls. It may be allocated for temporaries or
11382 variables that do not live across a call. Functions compiled this way
11383 will not save and restore the register @var{reg}.
11385 It is an error to used this flag with the frame pointer or stack pointer.
11386 Use of this flag for other registers that have fixed pervasive roles in
11387 the machine's execution model will produce disastrous results.
11389 This flag does not have a negative form, because it specifies a
11392 @item -fcall-saved-@var{reg}
11393 @opindex fcall-saved
11394 Treat the register named @var{reg} as an allocable register saved by
11395 functions. It may be allocated even for temporaries or variables that
11396 live across a call. Functions compiled this way will save and restore
11397 the register @var{reg} if they use it.
11399 It is an error to used this flag with the frame pointer or stack pointer.
11400 Use of this flag for other registers that have fixed pervasive roles in
11401 the machine's execution model will produce disastrous results.
11403 A different sort of disaster will result from the use of this flag for
11404 a register in which function values may be returned.
11406 This flag does not have a negative form, because it specifies a
11409 @item -fpack-struct
11410 @opindex fpack-struct
11411 Pack all structure members together without holes.
11413 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11414 code that is not binary compatible with code generated without that switch.
11415 Additionally, it makes the code suboptimal.
11416 Use it to conform to a non-default application binary interface.
11418 @item -finstrument-functions
11419 @opindex finstrument-functions
11420 Generate instrumentation calls for entry and exit to functions. Just
11421 after function entry and just before function exit, the following
11422 profiling functions will be called with the address of the current
11423 function and its call site. (On some platforms,
11424 @code{__builtin_return_address} does not work beyond the current
11425 function, so the call site information may not be available to the
11426 profiling functions otherwise.)
11429 void __cyg_profile_func_enter (void *this_fn,
11431 void __cyg_profile_func_exit (void *this_fn,
11435 The first argument is the address of the start of the current function,
11436 which may be looked up exactly in the symbol table.
11438 This instrumentation is also done for functions expanded inline in other
11439 functions. The profiling calls will indicate where, conceptually, the
11440 inline function is entered and exited. This means that addressable
11441 versions of such functions must be available. If all your uses of a
11442 function are expanded inline, this may mean an additional expansion of
11443 code size. If you use @samp{extern inline} in your C code, an
11444 addressable version of such functions must be provided. (This is
11445 normally the case anyways, but if you get lucky and the optimizer always
11446 expands the functions inline, you might have gotten away without
11447 providing static copies.)
11449 A function may be given the attribute @code{no_instrument_function}, in
11450 which case this instrumentation will not be done. This can be used, for
11451 example, for the profiling functions listed above, high-priority
11452 interrupt routines, and any functions from which the profiling functions
11453 cannot safely be called (perhaps signal handlers, if the profiling
11454 routines generate output or allocate memory).
11456 @item -fstack-check
11457 @opindex fstack-check
11458 Generate code to verify that you do not go beyond the boundary of the
11459 stack. You should specify this flag if you are running in an
11460 environment with multiple threads, but only rarely need to specify it in
11461 a single-threaded environment since stack overflow is automatically
11462 detected on nearly all systems if there is only one stack.
11464 Note that this switch does not actually cause checking to be done; the
11465 operating system must do that. The switch causes generation of code
11466 to ensure that the operating system sees the stack being extended.
11468 @item -fstack-limit-register=@var{reg}
11469 @itemx -fstack-limit-symbol=@var{sym}
11470 @itemx -fno-stack-limit
11471 @opindex fstack-limit-register
11472 @opindex fstack-limit-symbol
11473 @opindex fno-stack-limit
11474 Generate code to ensure that the stack does not grow beyond a certain value,
11475 either the value of a register or the address of a symbol. If the stack
11476 would grow beyond the value, a signal is raised. For most targets,
11477 the signal is raised before the stack overruns the boundary, so
11478 it is possible to catch the signal without taking special precautions.
11480 For instance, if the stack starts at absolute address @samp{0x80000000}
11481 and grows downwards, you can use the flags
11482 @option{-fstack-limit-symbol=__stack_limit} and
11483 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11484 of 128KB@. Note that this may only work with the GNU linker.
11486 @cindex aliasing of parameters
11487 @cindex parameters, aliased
11488 @item -fargument-alias
11489 @itemx -fargument-noalias
11490 @itemx -fargument-noalias-global
11491 @opindex fargument-alias
11492 @opindex fargument-noalias
11493 @opindex fargument-noalias-global
11494 Specify the possible relationships among parameters and between
11495 parameters and global data.
11497 @option{-fargument-alias} specifies that arguments (parameters) may
11498 alias each other and may alias global storage.@*
11499 @option{-fargument-noalias} specifies that arguments do not alias
11500 each other, but may alias global storage.@*
11501 @option{-fargument-noalias-global} specifies that arguments do not
11502 alias each other and do not alias global storage.
11504 Each language will automatically use whatever option is required by
11505 the language standard. You should not need to use these options yourself.
11507 @item -fleading-underscore
11508 @opindex fleading-underscore
11509 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11510 change the way C symbols are represented in the object file. One use
11511 is to help link with legacy assembly code.
11513 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11514 generate code that is not binary compatible with code generated without that
11515 switch. Use it to conform to a non-default application binary interface.
11516 Not all targets provide complete support for this switch.
11518 @item -ftls-model=@var{model}
11519 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11520 The @var{model} argument should be one of @code{global-dynamic},
11521 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
11523 The default without @option{-fpic} is @code{initial-exec}; with
11524 @option{-fpic} the default is @code{global-dynamic}.
11529 @node Environment Variables
11530 @section Environment Variables Affecting GCC
11531 @cindex environment variables
11533 @c man begin ENVIRONMENT
11534 This section describes several environment variables that affect how GCC
11535 operates. Some of them work by specifying directories or prefixes to use
11536 when searching for various kinds of files. Some are used to specify other
11537 aspects of the compilation environment.
11539 Note that you can also specify places to search using options such as
11540 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
11541 take precedence over places specified using environment variables, which
11542 in turn take precedence over those specified by the configuration of GCC@.
11543 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
11544 GNU Compiler Collection (GCC) Internals}.
11549 @c @itemx LC_COLLATE
11551 @c @itemx LC_MONETARY
11552 @c @itemx LC_NUMERIC
11557 @c @findex LC_COLLATE
11558 @findex LC_MESSAGES
11559 @c @findex LC_MONETARY
11560 @c @findex LC_NUMERIC
11564 These environment variables control the way that GCC uses
11565 localization information that allow GCC to work with different
11566 national conventions. GCC inspects the locale categories
11567 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
11568 so. These locale categories can be set to any value supported by your
11569 installation. A typical value is @samp{en_UK} for English in the United
11572 The @env{LC_CTYPE} environment variable specifies character
11573 classification. GCC uses it to determine the character boundaries in
11574 a string; this is needed for some multibyte encodings that contain quote
11575 and escape characters that would otherwise be interpreted as a string
11578 The @env{LC_MESSAGES} environment variable specifies the language to
11579 use in diagnostic messages.
11581 If the @env{LC_ALL} environment variable is set, it overrides the value
11582 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
11583 and @env{LC_MESSAGES} default to the value of the @env{LANG}
11584 environment variable. If none of these variables are set, GCC
11585 defaults to traditional C English behavior.
11589 If @env{TMPDIR} is set, it specifies the directory to use for temporary
11590 files. GCC uses temporary files to hold the output of one stage of
11591 compilation which is to be used as input to the next stage: for example,
11592 the output of the preprocessor, which is the input to the compiler
11595 @item GCC_EXEC_PREFIX
11596 @findex GCC_EXEC_PREFIX
11597 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
11598 names of the subprograms executed by the compiler. No slash is added
11599 when this prefix is combined with the name of a subprogram, but you can
11600 specify a prefix that ends with a slash if you wish.
11602 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
11603 an appropriate prefix to use based on the pathname it was invoked with.
11605 If GCC cannot find the subprogram using the specified prefix, it
11606 tries looking in the usual places for the subprogram.
11608 The default value of @env{GCC_EXEC_PREFIX} is
11609 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
11610 of @code{prefix} when you ran the @file{configure} script.
11612 Other prefixes specified with @option{-B} take precedence over this prefix.
11614 This prefix is also used for finding files such as @file{crt0.o} that are
11617 In addition, the prefix is used in an unusual way in finding the
11618 directories to search for header files. For each of the standard
11619 directories whose name normally begins with @samp{/usr/local/lib/gcc}
11620 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
11621 replacing that beginning with the specified prefix to produce an
11622 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
11623 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
11624 These alternate directories are searched first; the standard directories
11627 @item COMPILER_PATH
11628 @findex COMPILER_PATH
11629 The value of @env{COMPILER_PATH} is a colon-separated list of
11630 directories, much like @env{PATH}. GCC tries the directories thus
11631 specified when searching for subprograms, if it can't find the
11632 subprograms using @env{GCC_EXEC_PREFIX}.
11635 @findex LIBRARY_PATH
11636 The value of @env{LIBRARY_PATH} is a colon-separated list of
11637 directories, much like @env{PATH}. When configured as a native compiler,
11638 GCC tries the directories thus specified when searching for special
11639 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
11640 using GCC also uses these directories when searching for ordinary
11641 libraries for the @option{-l} option (but directories specified with
11642 @option{-L} come first).
11646 @cindex locale definition
11647 This variable is used to pass locale information to the compiler. One way in
11648 which this information is used is to determine the character set to be used
11649 when character literals, string literals and comments are parsed in C and C++.
11650 When the compiler is configured to allow multibyte characters,
11651 the following values for @env{LANG} are recognized:
11655 Recognize JIS characters.
11657 Recognize SJIS characters.
11659 Recognize EUCJP characters.
11662 If @env{LANG} is not defined, or if it has some other value, then the
11663 compiler will use mblen and mbtowc as defined by the default locale to
11664 recognize and translate multibyte characters.
11668 Some additional environments variables affect the behavior of the
11671 @include cppenv.texi
11675 @node Precompiled Headers
11676 @section Using Precompiled Headers
11677 @cindex precompiled headers
11678 @cindex speed of compilation
11680 Often large projects have many header files that are included in every
11681 source file. The time the compiler takes to process these header files
11682 over and over again can account for nearly all of the time required to
11683 build the project. To make builds faster, GCC allows users to
11684 `precompile' a header file; then, if builds can use the precompiled
11685 header file they will be much faster.
11687 To create a precompiled header file, simply compile it as you would any
11688 other file, if necessary using the @option{-x} option to make the driver
11689 treat it as a C or C++ header file. You will probably want to use a
11690 tool like @command{make} to keep the precompiled header up-to-date when
11691 the headers it contains change.
11693 A precompiled header file will be searched for when @code{#include} is
11694 seen in the compilation. As it searches for the included file
11695 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
11696 compiler looks for a precompiled header in each directory just before it
11697 looks for the include file in that directory. The name searched for is
11698 the name specified in the @code{#include} with @samp{.gch} appended. If
11699 the precompiled header file can't be used, it is ignored.
11701 For instance, if you have @code{#include "all.h"}, and you have
11702 @file{all.h.gch} in the same directory as @file{all.h}, then the
11703 precompiled header file will be used if possible, and the original
11704 header will be used otherwise.
11706 Alternatively, you might decide to put the precompiled header file in a
11707 directory and use @option{-I} to ensure that directory is searched
11708 before (or instead of) the directory containing the original header.
11709 Then, if you want to check that the precompiled header file is always
11710 used, you can put a file of the same name as the original header in this
11711 directory containing an @code{#error} command.
11713 This also works with @option{-include}. So yet another way to use
11714 precompiled headers, good for projects not designed with precompiled
11715 header files in mind, is to simply take most of the header files used by
11716 a project, include them from another header file, precompile that header
11717 file, and @option{-include} the precompiled header. If the header files
11718 have guards against multiple inclusion, they will be skipped because
11719 they've already been included (in the precompiled header).
11721 If you need to precompile the same header file for different
11722 languages, targets, or compiler options, you can instead make a
11723 @emph{directory} named like @file{all.h.gch}, and put each precompiled
11724 header in the directory. (It doesn't matter what you call the files
11725 in the directory, every precompiled header in the directory will be
11726 considered.) The first precompiled header encountered in the
11727 directory that is valid for this compilation will be used; they're
11728 searched in no particular order.
11730 There are many other possibilities, limited only by your imagination,
11731 good sense, and the constraints of your build system.
11733 A precompiled header file can be used only when these conditions apply:
11737 Only one precompiled header can be used in a particular compilation.
11739 A precompiled header can't be used once the first C token is seen. You
11740 can have preprocessor directives before a precompiled header; you can
11741 even include a precompiled header from inside another header, so long as
11742 there are no C tokens before the @code{#include}.
11744 The precompiled header file must be produced for the same language as
11745 the current compilation. You can't use a C precompiled header for a C++
11748 The precompiled header file must be produced by the same compiler
11749 version and configuration as the current compilation is using.
11750 The easiest way to guarantee this is to use the same compiler binary
11751 for creating and using precompiled headers.
11753 Any macros defined before the precompiled header (including with
11754 @option{-D}) must either be defined in the same way as when the
11755 precompiled header was generated, or must not affect the precompiled
11756 header, which usually means that the they don't appear in the
11757 precompiled header at all.
11759 Certain command-line options must be defined in the same way as when the
11760 precompiled header was generated. At present, it's not clear which
11761 options are safe to change and which are not; the safest choice is to
11762 use exactly the same options when generating and using the precompiled
11766 For all of these but the last, the compiler will automatically ignore
11767 the precompiled header if the conditions aren't met. For the last item,
11768 some option changes will cause the precompiled header to be rejected,
11769 but not all incompatible option combinations have yet been found. If
11770 you find a new incompatible combination, please consider filing a bug
11771 report, see @ref{Bugs}.
11773 @node Running Protoize
11774 @section Running Protoize
11776 The program @code{protoize} is an optional part of GCC@. You can use
11777 it to add prototypes to a program, thus converting the program to ISO
11778 C in one respect. The companion program @code{unprotoize} does the
11779 reverse: it removes argument types from any prototypes that are found.
11781 When you run these programs, you must specify a set of source files as
11782 command line arguments. The conversion programs start out by compiling
11783 these files to see what functions they define. The information gathered
11784 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
11786 After scanning comes actual conversion. The specified files are all
11787 eligible to be converted; any files they include (whether sources or
11788 just headers) are eligible as well.
11790 But not all the eligible files are converted. By default,
11791 @code{protoize} and @code{unprotoize} convert only source and header
11792 files in the current directory. You can specify additional directories
11793 whose files should be converted with the @option{-d @var{directory}}
11794 option. You can also specify particular files to exclude with the
11795 @option{-x @var{file}} option. A file is converted if it is eligible, its
11796 directory name matches one of the specified directory names, and its
11797 name within the directory has not been excluded.
11799 Basic conversion with @code{protoize} consists of rewriting most
11800 function definitions and function declarations to specify the types of
11801 the arguments. The only ones not rewritten are those for varargs
11804 @code{protoize} optionally inserts prototype declarations at the
11805 beginning of the source file, to make them available for any calls that
11806 precede the function's definition. Or it can insert prototype
11807 declarations with block scope in the blocks where undeclared functions
11810 Basic conversion with @code{unprotoize} consists of rewriting most
11811 function declarations to remove any argument types, and rewriting
11812 function definitions to the old-style pre-ISO form.
11814 Both conversion programs print a warning for any function declaration or
11815 definition that they can't convert. You can suppress these warnings
11818 The output from @code{protoize} or @code{unprotoize} replaces the
11819 original source file. The original file is renamed to a name ending
11820 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
11821 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
11822 for DOS) file already exists, then the source file is simply discarded.
11824 @code{protoize} and @code{unprotoize} both depend on GCC itself to
11825 scan the program and collect information about the functions it uses.
11826 So neither of these programs will work until GCC is installed.
11828 Here is a table of the options you can use with @code{protoize} and
11829 @code{unprotoize}. Each option works with both programs unless
11833 @item -B @var{directory}
11834 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
11835 usual directory (normally @file{/usr/local/lib}). This file contains
11836 prototype information about standard system functions. This option
11837 applies only to @code{protoize}.
11839 @item -c @var{compilation-options}
11840 Use @var{compilation-options} as the options when running @command{gcc} to
11841 produce the @samp{.X} files. The special option @option{-aux-info} is
11842 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
11844 Note that the compilation options must be given as a single argument to
11845 @code{protoize} or @code{unprotoize}. If you want to specify several
11846 @command{gcc} options, you must quote the entire set of compilation options
11847 to make them a single word in the shell.
11849 There are certain @command{gcc} arguments that you cannot use, because they
11850 would produce the wrong kind of output. These include @option{-g},
11851 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
11852 the @var{compilation-options}, they are ignored.
11855 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
11856 systems) instead of @samp{.c}. This is convenient if you are converting
11857 a C program to C++. This option applies only to @code{protoize}.
11860 Add explicit global declarations. This means inserting explicit
11861 declarations at the beginning of each source file for each function
11862 that is called in the file and was not declared. These declarations
11863 precede the first function definition that contains a call to an
11864 undeclared function. This option applies only to @code{protoize}.
11866 @item -i @var{string}
11867 Indent old-style parameter declarations with the string @var{string}.
11868 This option applies only to @code{protoize}.
11870 @code{unprotoize} converts prototyped function definitions to old-style
11871 function definitions, where the arguments are declared between the
11872 argument list and the initial @samp{@{}. By default, @code{unprotoize}
11873 uses five spaces as the indentation. If you want to indent with just
11874 one space instead, use @option{-i " "}.
11877 Keep the @samp{.X} files. Normally, they are deleted after conversion
11881 Add explicit local declarations. @code{protoize} with @option{-l} inserts
11882 a prototype declaration for each function in each block which calls the
11883 function without any declaration. This option applies only to
11887 Make no real changes. This mode just prints information about the conversions
11888 that would have been done without @option{-n}.
11891 Make no @samp{.save} files. The original files are simply deleted.
11892 Use this option with caution.
11894 @item -p @var{program}
11895 Use the program @var{program} as the compiler. Normally, the name
11896 @file{gcc} is used.
11899 Work quietly. Most warnings are suppressed.
11902 Print the version number, just like @option{-v} for @command{gcc}.
11905 If you need special compiler options to compile one of your program's
11906 source files, then you should generate that file's @samp{.X} file
11907 specially, by running @command{gcc} on that source file with the
11908 appropriate options and the option @option{-aux-info}. Then run
11909 @code{protoize} on the entire set of files. @code{protoize} will use
11910 the existing @samp{.X} file because it is newer than the source file.
11914 gcc -Dfoo=bar file1.c -aux-info file1.X
11919 You need to include the special files along with the rest in the
11920 @code{protoize} command, even though their @samp{.X} files already
11921 exist, because otherwise they won't get converted.
11923 @xref{Protoize Caveats}, for more information on how to use
11924 @code{protoize} successfully.