1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001 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 Free Software Foundation, Inc.
11 Permission is granted to make and distribute verbatim copies of this
12 manual provided the copyright notice and this permission notice are
13 preserved on all copies.
15 Permission is granted to copy and distribute modified versions of this
16 manual under the conditions for verbatim copying, provided also that the
17 entire resulting derived work is distributed under the terms of a
18 permission notice identical to this one.
20 Permission is granted to copy and distribute translations of this manual
21 into another language, under the above conditions for modified versions,
22 except that this permission notice may be included in translations
23 approved by the Free Software Foundation instead of in the original
26 @c Set file name and title for the man page.
28 @settitle GNU project C and C++ compiler
30 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
31 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
32 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
33 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
34 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
35 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
36 [@option{-o} @var{outfile}] @var{infile}@dots{}
38 Only the most useful options are listed here; see below for the
39 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
42 cpp(1), gcov(1), g77(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
43 and the Info entries for @file{gcc}, @file{cpp}, @file{g77}, @file{as},
44 @file{ld}, @file{binutils} and @file{gdb}.
47 For instructions on reporting bugs, see
48 @w{@uref{http://gcc.gnu.org/bugs.html}}. Use of the @command{gccbug}
49 script to report bugs is recommended.
52 See the Info entry for @file{gcc}, or
53 @w{@uref{http://gcc.gnu.org/thanks.html}}, for contributors to GCC@.
58 @chapter GCC Command Options
59 @cindex GCC command options
60 @cindex command options
61 @cindex options, GCC command
63 @c man begin DESCRIPTION
65 When you invoke GCC, it normally does preprocessing, compilation,
66 assembly and linking. The ``overall options'' allow you to stop this
67 process at an intermediate stage. For example, the @option{-c} option
68 says not to run the linker. Then the output consists of object files
69 output by the assembler.
71 Other options are passed on to one stage of processing. Some options
72 control the preprocessor and others the compiler itself. Yet other
73 options control the assembler and linker; most of these are not
74 documented here, since you rarely need to use any of them.
76 @cindex C compilation options
77 Most of the command line options that you can use with GCC are useful
78 for C programs; when an option is only useful with another language
79 (usually C++), the explanation says so explicitly. If the description
80 for a particular option does not mention a source language, you can use
81 that option with all supported languages.
83 @cindex C++ compilation options
84 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
85 options for compiling C++ programs.
87 @cindex grouping options
88 @cindex options, grouping
89 The @command{gcc} program accepts options and file names as operands. Many
90 options have multi-letter names; therefore multiple single-letter options
91 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
94 @cindex order of options
95 @cindex options, order
96 You can mix options and other arguments. For the most part, the order
97 you use doesn't matter. Order does matter when you use several options
98 of the same kind; for example, if you specify @option{-L} more than once,
99 the directories are searched in the order specified.
101 Many options have long names starting with @samp{-f} or with
102 @samp{-W}---for example, @option{-fforce-mem},
103 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
104 these have both positive and negative forms; the negative form of
105 @option{-ffoo} would be @option{-fno-foo}. This manual documents
106 only one of these two forms, whichever one is not the default.
110 @xref{Option Index}, for an index to GCC's options.
113 * Option Summary:: Brief list of all options, without explanations.
114 * Overall Options:: Controlling the kind of output:
115 an executable, object files, assembler files,
116 or preprocessed source.
117 * Invoking G++:: Compiling C++ programs.
118 * C Dialect Options:: Controlling the variant of C language compiled.
119 * C++ Dialect Options:: Variations on C++.
120 * Objective-C Dialect Options:: Variations on Objective-C.
121 * Language Independent Options:: Controlling how diagnostics should be
123 * Warning Options:: How picky should the compiler be?
124 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
125 * Optimize Options:: How much optimization?
126 * Preprocessor Options:: Controlling header files and macro definitions.
127 Also, getting dependency information for Make.
128 * Assembler Options:: Passing options to the assembler.
129 * Link Options:: Specifying libraries and so on.
130 * Directory Options:: Where to find header files and libraries.
131 Where to find the compiler executable files.
132 * Spec Files:: How to pass switches to sub-processes.
133 * Target Options:: Running a cross-compiler, or an old version of GCC.
134 * Submodel Options:: Specifying minor hardware or convention variations,
135 such as 68010 vs 68020.
136 * Code Gen Options:: Specifying conventions for function calls, data layout
138 * Environment Variables:: Env vars that affect GCC.
139 * Running Protoize:: Automatically adding or removing function prototypes.
145 @section Option Summary
147 Here is a summary of all the options, grouped by type. Explanations are
148 in the following sections.
151 @item Overall Options
152 @xref{Overall Options,,Options Controlling the Kind of Output}.
154 -c -S -E -o @var{file} -pipe -pass-exit-codes -x @var{language} @gol
155 -v --target-help --help}
157 @item C Language Options
158 @xref{C Dialect Options,,Options Controlling C Dialect}.
160 -ansi -std=@var{standard} -aux-info @var{filename} @gol
161 -fno-asm -fno-builtin @gol
162 -fhosted -ffreestanding @gol
163 -trigraphs -traditional -traditional-cpp @gol
164 -fallow-single-precision -fcond-mismatch @gol
165 -fsigned-bitfields -fsigned-char @gol
166 -funsigned-bitfields -funsigned-char @gol
167 -fwritable-strings -fshort-wchar}
169 @item C++ Language Options
170 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
172 -fno-access-control -fcheck-new -fconserve-space @gol
173 -fno-const-strings -fdollars-in-identifiers @gol
174 -fno-elide-constructors @gol
175 -fno-enforce-eh-specs -fexternal-templates @gol
176 -falt-external-templates @gol
177 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
178 -fno-implicit-templates @gol
179 -fno-implicit-inline-templates @gol
180 -fno-implement-inlines -fms-extensions @gol
181 -fno-nonansi-builtins -fno-operator-names @gol
182 -fno-optional-diags -fpermissive @gol
183 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
184 -fuse-cxa-atexit -fvtable-gc -fno-weak -nostdinc++ @gol
185 -fno-default-inline -Wctor-dtor-privacy @gol
186 -Wnon-virtual-dtor -Wreorder @gol
187 -Weffc++ -Wno-deprecated @gol
188 -Wno-non-template-friend -Wold-style-cast @gol
189 -Woverloaded-virtual -Wno-pmf-conversions @gol
190 -Wsign-promo -Wsynth}
192 @item Objective-C Language Options
193 @xref{Objective-C Dialect Options,,Options Controlling Objective-C Dialect}.
195 -fconstant-string-class=@var{class-name} @gol
196 -fgnu-runtime -fnext-runtime -gen-decls @gol
197 -Wno-protocol -Wselector}
199 @item Language Independent Options
200 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
202 -fmessage-length=@var{n} @gol
203 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
205 @item Warning Options
206 @xref{Warning Options,,Options to Request or Suppress Warnings}.
208 -fsyntax-only -pedantic -pedantic-errors @gol
209 -w -W -Wall -Waggregate-return @gol
210 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
211 -Wconversion -Wdisabled-optimization -Wdiv-by-zero -Werror @gol
212 -Wfloat-equal -Wformat -Wformat=2 @gol
213 -Wformat-nonliteral -Wformat-security @gol
214 -Wimplicit -Wimplicit-int @gol
215 -Wimplicit-function-declaration @gol
216 -Werror-implicit-function-declaration @gol
217 -Wimport -Winline @gol
218 -Wlarger-than-@var{len} -Wlong-long @gol
219 -Wmain -Wmissing-braces -Wmissing-declarations @gol
220 -Wmissing-format-attribute -Wmissing-noreturn @gol
221 -Wmultichar -Wno-format-extra-args -Wno-format-y2k @gol
222 -Wno-import -Wpacked -Wpadded @gol
223 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
224 -Wreturn-type -Wsequence-point -Wshadow @gol
225 -Wsign-compare -Wswitch -Wsystem-headers @gol
226 -Wtrigraphs -Wundef -Wuninitialized @gol
227 -Wunknown-pragmas -Wunreachable-code @gol
228 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
229 -Wunused-value -Wunused-variable -Wwrite-strings}
231 @item C-only Warning Options
233 -Wbad-function-cast -Wmissing-prototypes -Wnested-externs @gol
234 -Wstrict-prototypes -Wtraditional}
236 @item Debugging Options
237 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
239 -a -ax -d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
241 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
242 -fdump-tree-original@r{[}-@var{n}@r{]} -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
243 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
244 -fmem-report -fpretend-float @gol
245 -fprofile-arcs -ftest-coverage -ftime-report @gol
246 -g -g@var{level} -gcoff -gdwarf -gdwarf-1 -gdwarf-1+ -gdwarf-2 @gol
247 -ggdb -gstabs -gstabs+ -gxcoff -gxcoff+ @gol
248 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
249 -print-multi-directory -print-multi-lib @gol
250 -print-prog-name=@var{program} -print-search-dirs -Q @gol
253 @item Optimization Options
254 @xref{Optimize Options,,Options that Control Optimization}.
256 -falign-functions=@var{n} -falign-jumps=@var{n} @gol
257 -falign-labels=@var{n} -falign-loops=@var{n} @gol
258 -fbranch-probabilities -fcaller-saves @gol
259 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections @gol
260 -fdelayed-branch -fdelete-null-pointer-checks @gol
261 -fexpensive-optimizations -ffast-math -ffloat-store @gol
262 -fforce-addr -fforce-mem -ffunction-sections @gol
263 -fgcse -fgcse-lm -fgcse-sm @gol
264 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
265 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
266 -fmove-all-movables -fno-default-inline -fno-defer-pop @gol
267 -fno-function-cse -fno-guess-branch-probability @gol
268 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
269 -funsafe-math-optimizations -fno-trapping-math @gol
270 -fomit-frame-pointer -foptimize-register-move @gol
271 -foptimize-sibling-calls -freduce-all-givs @gol
272 -fregmove -frename-registers @gol
273 -frerun-cse-after-loop -frerun-loop-opt @gol
274 -fschedule-insns -fschedule-insns2 @gol
275 -fsingle-precision-constant -fssa -fssa-ccp -fssa-dce @gol
276 -fstrength-reduce -fstrict-aliasing -fthread-jumps -ftrapv @gol
277 -funroll-all-loops -funroll-loops @gol
278 --param @var{name}=@var{value}
279 -O -O0 -O1 -O2 -O3 -Os}
281 @item Preprocessor Options
282 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
284 -$ -A@var{question}=@var{answer} -A-@var{question}@r{[}=@var{answer}@r{]} @gol
285 -C -dD -dI -dM -dN @gol
286 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
287 -idirafter @var{dir} @gol
288 -include @var{file} -imacros @var{file} @gol
289 -iprefix @var{file} -iwithprefix @var{dir} @gol
290 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
291 -M -MM -MF -MG -MP -MQ -MT -nostdinc -P -remap @gol
292 -trigraphs -undef -U@var{macro} -Wp,@var{option}}
294 @item Assembler Option
295 @xref{Assembler Options,,Passing Options to the Assembler}.
300 @xref{Link Options,,Options for Linking}.
302 @var{object-file-name} -l@var{library} @gol
303 -nostartfiles -nodefaultlibs -nostdlib @gol
304 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
305 -Wl,@var{option} -Xlinker @var{option} @gol
308 @item Directory Options
309 @xref{Directory Options,,Options for Directory Search}.
311 -B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}}
314 @c I wrote this xref this way to avoid overfull hbox. -- rms
315 @xref{Target Options}.
317 -b @var{machine} -V @var{version}}
319 @item Machine Dependent Options
320 @xref{Submodel Options,,Hardware Models and Configurations}.
321 @emph{M680x0 Options}
323 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
324 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
325 -mfpa -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
326 -malign-int -mstrict-align}
328 @emph{M68hc1x Options}
330 -m6811 -m6812 -m68hc11 -m68hc12 @gol
331 -mauto-incdec -mshort -msoft-reg-count=@var{count}}
339 -mcpu=@var{cpu-type} @gol
340 -mtune=@var{cpu-type} @gol
341 -mcmodel=@var{code-model} @gol
343 -mapp-regs -mbroken-saverestore -mcypress @gol
344 -mepilogue -mfaster-structs -mflat @gol
345 -mfpu -mhard-float -mhard-quad-float @gol
346 -mimpure-text -mlive-g0 -mno-app-regs @gol
347 -mno-epilogue -mno-faster-structs -mno-flat -mno-fpu @gol
348 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles @gol
349 -msoft-float -msoft-quad-float -msparclite -mstack-bias @gol
350 -msupersparc -munaligned-doubles -mv8}
352 @emph{Convex Options}
354 -mc1 -mc2 -mc32 -mc34 -mc38 @gol
355 -margcount -mnoargcount @gol
356 -mlong32 -mlong64 @gol
357 -mvolatile-cache -mvolatile-nocache}
359 @emph{AMD29K Options}
361 -m29000 -m29050 -mbw -mnbw -mdw -mndw @gol
362 -mlarge -mnormal -msmall @gol
363 -mkernel-registers -mno-reuse-arg-regs @gol
364 -mno-stack-check -mno-storem-bug @gol
365 -mreuse-arg-regs -msoft-float -mstack-check @gol
366 -mstorem-bug -muser-registers}
370 -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 -mbsd -mxopen -mno-symrename @gol
383 -mabort-on-noreturn @gol
384 -mlong-calls -mno-long-calls @gol
385 -msingle-pic-base -mno-single-pic-base @gol
386 -mpic-register=@var{reg} @gol
387 -mnop-fun-dllimport @gol
388 -mpoke-function-name @gol
390 -mtpcs-frame -mtpcs-leaf-frame @gol
391 -mcaller-super-interworking -mcallee-super-interworking }
393 @emph{MN10200 Options}
397 @emph{MN10300 Options}
399 -mmult-bug -mno-mult-bug @gol
400 -mam33 -mno-am33 @gol
403 @emph{M32R/D Options}
405 -mcode-model=@var{model-type} -msdata=@var{sdata-type} @gol
410 -m88000 -m88100 -m88110 -mbig-pic @gol
411 -mcheck-zero-division -mhandle-large-shift @gol
412 -midentify-revision -mno-check-zero-division @gol
413 -mno-ocs-debug-info -mno-ocs-frame-position @gol
414 -mno-optimize-arg-area -mno-serialize-volatile @gol
415 -mno-underscores -mocs-debug-info @gol
416 -mocs-frame-position -moptimize-arg-area @gol
417 -mserialize-volatile -mshort-data-@var{num} -msvr3 @gol
418 -msvr4 -mtrap-large-shift -muse-div-instruction @gol
419 -mversion-03.00 -mwarn-passed-structs}
421 @emph{RS/6000 and PowerPC Options}
423 -mcpu=@var{cpu-type} @gol
424 -mtune=@var{cpu-type} @gol
425 -mpower -mno-power -mpower2 -mno-power2 @gol
426 -mpowerpc -mpowerpc64 -mno-powerpc @gol
427 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
428 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
429 -mnew-mnemonics -mold-mnemonics @gol
430 -mfull-toc -mminimal-toc -mno-fop-in-toc -mno-sum-in-toc @gol
431 -m64 -m32 -mxl-call -mno-xl-call -mthreads -mpe @gol
432 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
433 -mstring -mno-string -mupdate -mno-update @gol
434 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
435 -mstrict-align -mno-strict-align -mrelocatable @gol
436 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
437 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
438 -mcall-aix -mcall-sysv -mcall-netbsd @gol
439 -mprototype -mno-prototype @gol
440 -msim -mmvme -mads -myellowknife -memb -msdata @gol
441 -msdata=@var{opt} -mvxworks -G @var{num}}
445 -mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs @gol
446 -mfull-fp-blocks -mhc-struct-return -min-line-mul @gol
447 -mminimum-fp-blocks -mnohc-struct-return}
451 -mabicalls -march=@var{cpu-type} -mtune=@var{cpu=type} @gol
452 -mcpu=@var{cpu-type} -membedded-data -muninit-const-in-rodata @gol
453 -membedded-pic -mfp32 -mfp64 -mfused-madd -mno-fused-madd @gol
454 -mgas -mgp32 -mgp64 @gol
455 -mgpopt -mhalf-pic -mhard-float -mint64 -mips1 @gol
456 -mips2 -mips3 -mips4 -mlong64 -mlong32 -mlong-calls -mmemcpy @gol
457 -mmips-as -mmips-tfile -mno-abicalls @gol
458 -mno-embedded-data -mno-uninit-const-in-rodata @gol
459 -mno-embedded-pic -mno-gpopt -mno-long-calls @gol
460 -mno-memcpy -mno-mips-tfile -mno-rnames -mno-stats @gol
461 -mrnames -msoft-float @gol
462 -m4650 -msingle-float -mmad @gol
463 -mstats -EL -EB -G @var{num} -nocpp @gol
464 -mabi=32 -mabi=n32 -mabi=64 -mabi=eabi @gol
467 @emph{i386 and x86-64 Options}
469 -mcpu=@var{cpu-type} -march=@var{cpu-type} @gol
470 -mintel-syntax -mieee-fp -mno-fancy-math-387 @gol
471 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
472 -mno-wide-multiply -mrtd -malign-double @gol
473 -mpreferred-stack-boundary=@var{num} @gol
474 -mmmx -msse -m3dnow @gol
475 -mthreads -mno-align-stringops -minline-all-stringops @gol
476 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
477 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
483 -march=@var{architecture-type} @gol
484 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
485 -mfast-indirect-calls -mgas -mjump-in-delay @gol
486 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
487 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
488 -mno-jump-in-delay -mno-long-load-store @gol
489 -mno-portable-runtime -mno-soft-float @gol
490 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
491 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
492 -mschedule=@var{cpu-type} -mspace-regs}
494 @emph{Intel 960 Options}
496 -m@var{cpu-type} -masm-compat -mclean-linkage @gol
497 -mcode-align -mcomplex-addr -mleaf-procedures @gol
498 -mic-compat -mic2.0-compat -mic3.0-compat @gol
499 -mintel-asm -mno-clean-linkage -mno-code-align @gol
500 -mno-complex-addr -mno-leaf-procedures @gol
501 -mno-old-align -mno-strict-align -mno-tail-call @gol
502 -mnumerics -mold-align -msoft-float -mstrict-align @gol
505 @emph{DEC Alpha Options}
507 -mfp-regs -mno-fp-regs -mno-soft-float -msoft-float @gol
508 -malpha-as -mgas @gol
509 -mieee -mieee-with-inexact -mieee-conformant @gol
510 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
511 -mtrap-precision=@var{mode} -mbuild-constants @gol
512 -mcpu=@var{cpu-type} @gol
513 -mbwx -mno-bwx -mcix -mno-cix -mmax -mno-max @gol
514 -mmemory-latency=@var{time}}
516 @emph{Clipper Options}
520 @emph{H8/300 Options}
522 -mrelax -mh -ms -mint32 -malign-300}
526 -m1 -m2 -m3 -m3e @gol
527 -m4-nofpu -m4-single-only -m4-single -m4 @gol
528 -mb -ml -mdalign -mrelax @gol
529 -mbigtable -mfmovd -mhitachi -mnomacsave @gol
530 -mieee -misize -mpadstruct -mspace @gol
531 -mprefergot -musermode}
533 @emph{System V Options}
535 -Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
540 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
541 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
543 @emph{TMS320C3x/C4x Options}
545 -mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
546 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
547 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
548 -mparallel-insns -mparallel-mpy -mpreserve-float}
552 -mlong-calls -mno-long-calls -mep -mno-ep @gol
553 -mprolog-function -mno-prolog-function -mspace @gol
554 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
559 -m32032 -m32332 -m32532 -m32081 -m32381 @gol
560 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
561 -mregparam -mnoregparam -msb -mnosb @gol
562 -mbitfield -mnobitfield -mhimem -mnohimem}
566 -mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
567 -mcall-prologues -mno-tablejump -mtiny-stack}
571 -mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
572 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
573 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
574 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
575 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
579 -mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
580 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
581 -mconstant-gp -mauto-pic -minline-divide-min-latency @gol
582 -minline-divide-max-throughput -mno-dwarf2-asm @gol
583 -mfixed-range=@var{register-range}}
585 @emph{S/390 and zSeries Options}
587 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
588 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
589 -m64 -m31 -mdebug -mno-debug}
593 -mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
594 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
595 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
596 -mstack-align -mdata-align -mconst-align @gol
597 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
598 -melf -maout -melinux -mlinux -sim -sim2}
600 @item Code Generation Options
601 @xref{Code Gen Options,,Options for Code Generation Conventions}.
603 -fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
604 -ffixed-@var{reg} -fexceptions @gol
605 -fnon-call-exceptions -funwind-tables -fasynchronous-unwind-tables @gol
606 -finhibit-size-directive -finstrument-functions @gol
607 -fcheck-memory-usage -fprefix-function-name @gol
608 -fno-common -fno-ident -fno-gnu-linker @gol
609 -fpcc-struct-return -fpic -fPIC @gol
610 -freg-struct-return -fshared-data -fshort-enums @gol
611 -fshort-double -fvolatile @gol
612 -fvolatile-global -fvolatile-static @gol
613 -fverbose-asm -fpack-struct -fstack-check @gol
614 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
615 -fargument-alias -fargument-noalias @gol
616 -fargument-noalias-global -fleading-underscore}
620 * Overall Options:: Controlling the kind of output:
621 an executable, object files, assembler files,
622 or preprocessed source.
623 * C Dialect Options:: Controlling the variant of C language compiled.
624 * C++ Dialect Options:: Variations on C++.
625 * Objective-C Dialect Options:: Variations on Objective-C.
626 * Language Independent Options:: Controlling how diagnostics should be
628 * Warning Options:: How picky should the compiler be?
629 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
630 * Optimize Options:: How much optimization?
631 * Preprocessor Options:: Controlling header files and macro definitions.
632 Also, getting dependency information for Make.
633 * Assembler Options:: Passing options to the assembler.
634 * Link Options:: Specifying libraries and so on.
635 * Directory Options:: Where to find header files and libraries.
636 Where to find the compiler executable files.
637 * Spec Files:: How to pass switches to sub-processes.
638 * Target Options:: Running a cross-compiler, or an old version of GCC.
641 @node Overall Options
642 @section Options Controlling the Kind of Output
644 Compilation can involve up to four stages: preprocessing, compilation
645 proper, assembly and linking, always in that order. The first three
646 stages apply to an individual source file, and end by producing an
647 object file; linking combines all the object files (those newly
648 compiled, and those specified as input) into an executable file.
650 @cindex file name suffix
651 For any given input file, the file name suffix determines what kind of
656 C source code which must be preprocessed.
659 C source code which should not be preprocessed.
662 C++ source code which should not be preprocessed.
665 Objective-C source code. Note that you must link with the library
666 @file{libobjc.a} to make an Objective-C program work.
669 Objective-C source code which should not be preprocessed.
672 C header file (not to be compiled or linked).
676 @itemx @var{file}.cxx
677 @itemx @var{file}.cpp
678 @itemx @var{file}.c++
680 C++ source code which must be preprocessed. Note that in @samp{.cxx},
681 the last two letters must both be literally @samp{x}. Likewise,
682 @samp{.C} refers to a literal capital C@.
685 @itemx @var{file}.for
686 @itemx @var{file}.FOR
687 Fortran source code which should not be preprocessed.
690 @itemx @var{file}.fpp
691 @itemx @var{file}.FPP
692 Fortran source code which must be preprocessed (with the traditional
696 Fortran source code which must be preprocessed with a RATFOR
697 preprocessor (not included with GCC)@.
699 @xref{Overall Options,,Options Controlling the Kind of Output, g77,
700 Using and Porting GNU Fortran}, for more details of the handling of
703 @c FIXME: Descriptions of Java file types.
709 @c GCC also knows about some suffixes for languages not yet included:
719 @itemx @var{file}.chi
720 CHILL source code (preprocessed with the traditional preprocessor).
726 Assembler code which must be preprocessed.
729 An object file to be fed straight into linking.
730 Any file name with no recognized suffix is treated this way.
734 You can specify the input language explicitly with the @option{-x} option:
737 @item -x @var{language}
738 Specify explicitly the @var{language} for the following input files
739 (rather than letting the compiler choose a default based on the file
740 name suffix). This option applies to all following input files until
741 the next @option{-x} option. Possible values for @var{language} are:
743 c c-header cpp-output
745 objective-c objc-cpp-output
746 assembler assembler-with-cpp
747 f77 f77-cpp-input ratfor
750 @c Also f77-version, for internal use only.
753 Turn off any specification of a language, so that subsequent files are
754 handled according to their file name suffixes (as they are if @option{-x}
755 has not been used at all).
757 @item -pass-exit-codes
758 @opindex pass-exit-codes
759 Normally the @command{gcc} program will exit with the code of 1 if any
760 phase of the compiler returns a non-success return code. If you specify
761 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
762 numerically highest error produced by any phase that returned an error
766 If you only want some of the stages of compilation, you can use
767 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
768 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
769 @command{gcc} is to stop. Note that some combinations (for example,
770 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
775 Compile or assemble the source files, but do not link. The linking
776 stage simply is not done. The ultimate output is in the form of an
777 object file for each source file.
779 By default, the object file name for a source file is made by replacing
780 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
782 Unrecognized input files, not requiring compilation or assembly, are
787 Stop after the stage of compilation proper; do not assemble. The output
788 is in the form of an assembler code file for each non-assembler input
791 By default, the assembler file name for a source file is made by
792 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
794 Input files that don't require compilation are ignored.
798 Stop after the preprocessing stage; do not run the compiler proper. The
799 output is in the form of preprocessed source code, which is sent to the
802 Input files which don't require preprocessing are ignored.
804 @cindex output file option
807 Place output in file @var{file}. This applies regardless to whatever
808 sort of output is being produced, whether it be an executable file,
809 an object file, an assembler file or preprocessed C code.
811 Since only one output file can be specified, it does not make sense to
812 use @option{-o} when compiling more than one input file, unless you are
813 producing an executable file as output.
815 If @option{-o} is not specified, the default is to put an executable file
816 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
817 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
818 all preprocessed C source on standard output.
822 Print (on standard error output) the commands executed to run the stages
823 of compilation. Also print the version number of the compiler driver
824 program and of the preprocessor and the compiler proper.
828 Use pipes rather than temporary files for communication between the
829 various stages of compilation. This fails to work on some systems where
830 the assembler is unable to read from a pipe; but the GNU assembler has
835 Print (on the standard output) a description of the command line options
836 understood by @command{gcc}. If the @option{-v} option is also specified
837 then @option{--help} will also be passed on to the various processes
838 invoked by @command{gcc}, so that they can display the command line options
839 they accept. If the @option{-W} option is also specified then command
840 line options which have no documentation associated with them will also
845 Print (on the standard output) a description of target specific command
846 line options for each tool.
850 @section Compiling C++ Programs
852 @cindex suffixes for C++ source
853 @cindex C++ source file suffixes
854 C++ source files conventionally use one of the suffixes @samp{.C},
855 @samp{.cc}, @samp{.cpp}, @samp{.c++}, @samp{.cp}, or @samp{.cxx};
856 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
857 files with these names and compiles them as C++ programs even if you
858 call the compiler the same way as for compiling C programs (usually with
859 the name @command{gcc}).
863 However, C++ programs often require class libraries as well as a
864 compiler that understands the C++ language---and under some
865 circumstances, you might want to compile programs from standard input,
866 or otherwise without a suffix that flags them as C++ programs.
867 @command{g++} is a program that calls GCC with the default language
868 set to C++, and automatically specifies linking against the C++
869 library. On many systems, @command{g++} is also
870 installed with the name @command{c++}.
872 @cindex invoking @command{g++}
873 When you compile C++ programs, you may specify many of the same
874 command-line options that you use for compiling programs in any
875 language; or command-line options meaningful for C and related
876 languages; or options that are meaningful only for C++ programs.
877 @xref{C Dialect Options,,Options Controlling C Dialect}, for
878 explanations of options for languages related to C@.
879 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
880 explanations of options that are meaningful only for C++ programs.
882 @node C Dialect Options
883 @section Options Controlling C Dialect
884 @cindex dialect options
885 @cindex language dialect options
886 @cindex options, dialect
888 The following options control the dialect of C (or languages derived
889 from C, such as C++ and Objective-C) that the compiler accepts:
896 In C mode, support all ISO C89 programs. In C++ mode,
897 remove GNU extensions that conflict with ISO C++.
899 This turns off certain features of GCC that are incompatible with ISO
900 C89 (when compiling C code), or of standard C++ (when compiling C++ code),
901 such as the @code{asm} and @code{typeof} keywords, and
902 predefined macros such as @code{unix} and @code{vax} that identify the
903 type of system you are using. It also enables the undesirable and
904 rarely used ISO trigraph feature. For the C compiler,
905 it disables recognition of C++ style @samp{//} comments as well as
906 the @code{inline} keyword.
908 The alternate keywords @code{__asm__}, @code{__extension__},
909 @code{__inline__} and @code{__typeof__} continue to work despite
910 @option{-ansi}. You would not want to use them in an ISO C program, of
911 course, but it is useful to put them in header files that might be included
912 in compilations done with @option{-ansi}. Alternate predefined macros
913 such as @code{__unix__} and @code{__vax__} are also available, with or
914 without @option{-ansi}.
916 The @option{-ansi} option does not cause non-ISO programs to be
917 rejected gratuitously. For that, @option{-pedantic} is required in
918 addition to @option{-ansi}. @xref{Warning Options}.
920 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
921 option is used. Some header files may notice this macro and refrain
922 from declaring certain functions or defining certain macros that the
923 ISO standard doesn't call for; this is to avoid interfering with any
924 programs that might use these names for other things.
926 Functions which would normally be built in but do not have semantics
927 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
928 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
929 built-in functions provided by GCC}, for details of the functions
934 Determine the language standard. This option is currently only
935 supported when compiling C@. A value for this option must be provided;
941 ISO C89 (same as @option{-ansi}).
944 ISO C89 as modified in amendment 1.
950 ISO C99. Note that this standard is not yet fully supported; see
951 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
952 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
955 Default, ISO C89 plus GNU extensions (including some C99 features).
959 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
960 this will become the default. The name @samp{gnu9x} is deprecated.
964 Even when this option is not specified, you can still use some of the
965 features of newer standards in so far as they do not conflict with
966 previous C standards. For example, you may use @code{__restrict__} even
967 when @option{-std=c99} is not specified.
969 The @option{-std} options specifying some version of ISO C have the same
970 effects as @option{-ansi}, except that features that were not in ISO C89
971 but are in the specified version (for example, @samp{//} comments and
972 the @code{inline} keyword in ISO C99) are not disabled.
974 @xref{Standards,,Language Standards Supported by GCC}, for details of
975 these standard versions.
977 @item -aux-info @var{filename}
979 Output to the given filename prototyped declarations for all functions
980 declared and/or defined in a translation unit, including those in header
981 files. This option is silently ignored in any language other than C@.
983 Besides declarations, the file indicates, in comments, the origin of
984 each declaration (source file and line), whether the declaration was
985 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
986 @samp{O} for old, respectively, in the first character after the line
987 number and the colon), and whether it came from a declaration or a
988 definition (@samp{C} or @samp{F}, respectively, in the following
989 character). In the case of function definitions, a K&R-style list of
990 arguments followed by their declarations is also provided, inside
991 comments, after the declaration.
995 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
996 keyword, so that code can use these words as identifiers. You can use
997 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
998 instead. @option{-ansi} implies @option{-fno-asm}.
1000 In C++, this switch only affects the @code{typeof} keyword, since
1001 @code{asm} and @code{inline} are standard keywords. You may want to
1002 use the @option{-fno-gnu-keywords} flag instead, which has the same
1003 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1004 switch only affects the @code{asm} and @code{typeof} keywords, since
1005 @code{inline} is a standard keyword in ISO C99.
1008 @opindex fno-builtin
1009 @cindex built-in functions
1010 Don't recognize built-in functions that do not begin with
1011 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1012 functions provided by GCC}, for details of the functions affected,
1013 including those which are not built-in functions when @option{-ansi} or
1014 @option{-std} options for strict ISO C conformance are used because they
1015 do not have an ISO standard meaning.
1017 GCC normally generates special code to handle certain built-in functions
1018 more efficiently; for instance, calls to @code{alloca} may become single
1019 instructions that adjust the stack directly, and calls to @code{memcpy}
1020 may become inline copy loops. The resulting code is often both smaller
1021 and faster, but since the function calls no longer appear as such, you
1022 cannot set a breakpoint on those calls, nor can you change the behavior
1023 of the functions by linking with a different library.
1025 In C++, @option{-fno-builtin} is always in effect. The @option{-fbuiltin}
1026 option has no effect. Therefore, in C++, the only way to get the
1027 optimization benefits of built-in functions is to call the function
1028 using the @samp{__builtin_} prefix. The GNU C++ Standard Library uses
1029 built-in functions to implement many functions (like
1030 @code{std::strchr}), so that you automatically get efficient code.
1034 @cindex hosted environment
1036 Assert that compilation takes place in a hosted environment. This implies
1037 @option{-fbuiltin}. A hosted environment is one in which the
1038 entire standard library is available, and in which @code{main} has a return
1039 type of @code{int}. Examples are nearly everything except a kernel.
1040 This is equivalent to @option{-fno-freestanding}.
1042 @item -ffreestanding
1043 @opindex ffreestanding
1044 @cindex hosted environment
1046 Assert that compilation takes place in a freestanding environment. This
1047 implies @option{-fno-builtin}. A freestanding environment
1048 is one in which the standard library may not exist, and program startup may
1049 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1050 This is equivalent to @option{-fno-hosted}.
1052 @xref{Standards,,Language Standards Supported by GCC}, for details of
1053 freestanding and hosted environments.
1057 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1058 options for strict ISO C conformance) implies @option{-trigraphs}.
1060 @cindex traditional C language
1061 @cindex C language, traditional
1063 @opindex traditional
1064 Attempt to support some aspects of traditional C compilers.
1069 All @code{extern} declarations take effect globally even if they
1070 are written inside of a function definition. This includes implicit
1071 declarations of functions.
1074 The newer keywords @code{typeof}, @code{inline}, @code{signed}, @code{const}
1075 and @code{volatile} are not recognized. (You can still use the
1076 alternative keywords such as @code{__typeof__}, @code{__inline__}, and
1080 Comparisons between pointers and integers are always allowed.
1083 Integer types @code{unsigned short} and @code{unsigned char} promote
1084 to @code{unsigned int}.
1087 Out-of-range floating point literals are not an error.
1090 Certain constructs which ISO regards as a single invalid preprocessing
1091 number, such as @samp{0xe-0xd}, are treated as expressions instead.
1094 String ``constants'' are not necessarily constant; they are stored in
1095 writable space, and identical looking constants are allocated
1096 separately. (This is the same as the effect of
1097 @option{-fwritable-strings}.)
1099 @cindex @code{longjmp} and automatic variables
1101 All automatic variables not declared @code{register} are preserved by
1102 @code{longjmp}. Ordinarily, GNU C follows ISO C: automatic variables
1103 not declared @code{volatile} may be clobbered.
1108 @cindex escape sequences, traditional
1109 The character escape sequences @samp{\x} and @samp{\a} evaluate as the
1110 literal characters @samp{x} and @samp{a} respectively. Without
1111 @w{@option{-traditional}}, @samp{\x} is a prefix for the hexadecimal
1112 representation of a character, and @samp{\a} produces a bell.
1115 You may wish to use @option{-fno-builtin} as well as @option{-traditional}
1116 if your program uses names that are normally GNU C built-in functions for
1117 other purposes of its own.
1119 You cannot use @option{-traditional} if you include any header files that
1120 rely on ISO C features. Some vendors are starting to ship systems with
1121 ISO C header files and you cannot use @option{-traditional} on such
1122 systems to compile files that include any system headers.
1124 The @option{-traditional} option also enables @option{-traditional-cpp}.
1126 @item -traditional-cpp
1127 @opindex traditional-cpp
1128 Attempt to support some aspects of traditional C preprocessors.
1129 See the GNU CPP manual for details.
1131 @item -fcond-mismatch
1132 @opindex fcond-mismatch
1133 Allow conditional expressions with mismatched types in the second and
1134 third arguments. The value of such an expression is void. This option
1135 is not supported for C++.
1137 @item -funsigned-char
1138 @opindex funsigned-char
1139 Let the type @code{char} be unsigned, like @code{unsigned char}.
1141 Each kind of machine has a default for what @code{char} should
1142 be. It is either like @code{unsigned char} by default or like
1143 @code{signed char} by default.
1145 Ideally, a portable program should always use @code{signed char} or
1146 @code{unsigned char} when it depends on the signedness of an object.
1147 But many programs have been written to use plain @code{char} and
1148 expect it to be signed, or expect it to be unsigned, depending on the
1149 machines they were written for. This option, and its inverse, let you
1150 make such a program work with the opposite default.
1152 The type @code{char} is always a distinct type from each of
1153 @code{signed char} or @code{unsigned char}, even though its behavior
1154 is always just like one of those two.
1157 @opindex fsigned-char
1158 Let the type @code{char} be signed, like @code{signed char}.
1160 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1161 the negative form of @option{-funsigned-char}. Likewise, the option
1162 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1164 @item -fsigned-bitfields
1165 @itemx -funsigned-bitfields
1166 @itemx -fno-signed-bitfields
1167 @itemx -fno-unsigned-bitfields
1168 @opindex fsigned-bitfields
1169 @opindex funsigned-bitfields
1170 @opindex fno-signed-bitfields
1171 @opindex fno-unsigned-bitfields
1172 These options control whether a bit-field is signed or unsigned, when the
1173 declaration does not use either @code{signed} or @code{unsigned}. By
1174 default, such a bit-field is signed, because this is consistent: the
1175 basic integer types such as @code{int} are signed types.
1177 However, when @option{-traditional} is used, bit-fields are all unsigned
1180 @item -fwritable-strings
1181 @opindex fwritable-strings
1182 Store string constants in the writable data segment and don't uniquize
1183 them. This is for compatibility with old programs which assume they can
1184 write into string constants. The option @option{-traditional} also has
1187 Writing into string constants is a very bad idea; ``constants'' should
1190 @item -fallow-single-precision
1191 @opindex fallow-single-precision
1192 Do not promote single precision math operations to double precision,
1193 even when compiling with @option{-traditional}.
1195 Traditional K&R C promotes all floating point operations to double
1196 precision, regardless of the sizes of the operands. On the
1197 architecture for which you are compiling, single precision may be faster
1198 than double precision. If you must use @option{-traditional}, but want
1199 to use single precision operations when the operands are single
1200 precision, use this option. This option has no effect when compiling
1201 with ISO or GNU C conventions (the default).
1204 @opindex fshort-wchar
1205 Override the underlying type for @samp{wchar_t} to be @samp{short
1206 unsigned int} instead of the default for the target. This option is
1207 useful for building programs to run under WINE@.
1210 @node C++ Dialect Options
1211 @section Options Controlling C++ Dialect
1213 @cindex compiler options, C++
1214 @cindex C++ options, command line
1215 @cindex options, C++
1216 This section describes the command-line options that are only meaningful
1217 for C++ programs; but you can also use most of the GNU compiler options
1218 regardless of what language your program is in. For example, you
1219 might compile a file @code{firstClass.C} like this:
1222 g++ -g -frepo -O -c firstClass.C
1226 In this example, only @option{-frepo} is an option meant
1227 only for C++ programs; you can use the other options with any
1228 language supported by GCC@.
1230 Here is a list of options that are @emph{only} for compiling C++ programs:
1233 @item -fno-access-control
1234 @opindex fno-access-control
1235 Turn off all access checking. This switch is mainly useful for working
1236 around bugs in the access control code.
1240 Check that the pointer returned by @code{operator new} is non-null
1241 before attempting to modify the storage allocated. The current Working
1242 Paper requires that @code{operator new} never return a null pointer, so
1243 this check is normally unnecessary.
1245 An alternative to using this option is to specify that your
1246 @code{operator new} does not throw any exceptions; if you declare it
1247 @samp{throw()}, G++ will check the return value. See also @samp{new
1250 @item -fconserve-space
1251 @opindex fconserve-space
1252 Put uninitialized or runtime-initialized global variables into the
1253 common segment, as C does. This saves space in the executable at the
1254 cost of not diagnosing duplicate definitions. If you compile with this
1255 flag and your program mysteriously crashes after @code{main()} has
1256 completed, you may have an object that is being destroyed twice because
1257 two definitions were merged.
1259 This option is no longer useful on most targets, now that support has
1260 been added for putting variables into BSS without making them common.
1262 @item -fno-const-strings
1263 @opindex fno-const-strings
1264 Give string constants type @code{char *} instead of type @code{const
1265 char *}. By default, G++ uses type @code{const char *} as required by
1266 the standard. Even if you use @option{-fno-const-strings}, you cannot
1267 actually modify the value of a string constant, unless you also use
1268 @option{-fwritable-strings}.
1270 This option might be removed in a future release of G++. For maximum
1271 portability, you should structure your code so that it works with
1272 string constants that have type @code{const char *}.
1274 @item -fdollars-in-identifiers
1275 @opindex fdollars-in-identifiers
1276 Accept @samp{$} in identifiers. You can also explicitly prohibit use of
1277 @samp{$} with the option @option{-fno-dollars-in-identifiers}. (GNU C allows
1278 @samp{$} by default on most target systems, but there are a few exceptions.)
1279 Traditional C allowed the character @samp{$} to form part of
1280 identifiers. However, ISO C and C++ forbid @samp{$} in identifiers.
1282 @item -fno-elide-constructors
1283 @opindex fno-elide-constructors
1284 The C++ standard allows an implementation to omit creating a temporary
1285 which is only used to initialize another object of the same type.
1286 Specifying this option disables that optimization, and forces G++ to
1287 call the copy constructor in all cases.
1289 @item -fno-enforce-eh-specs
1290 @opindex fno-enforce-eh-specs
1291 Don't check for violation of exception specifications at runtime. This
1292 option violates the C++ standard, but may be useful for reducing code
1293 size in production builds, much like defining @samp{NDEBUG}. The compiler
1294 will still optimize based on the exception specifications.
1296 @item -fexternal-templates
1297 @opindex fexternal-templates
1299 Cause @samp{#pragma interface} and @samp{implementation} to apply to
1300 template instantiation; template instances are emitted or not according
1301 to the location of the template definition. @xref{Template
1302 Instantiation}, for more information.
1304 This option is deprecated.
1306 @item -falt-external-templates
1307 @opindex falt-external-templates
1308 Similar to @option{-fexternal-templates}, but template instances are
1309 emitted or not according to the place where they are first instantiated.
1310 @xref{Template Instantiation}, for more information.
1312 This option is deprecated.
1315 @itemx -fno-for-scope
1317 @opindex fno-for-scope
1318 If @option{-ffor-scope} is specified, the scope of variables declared in
1319 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1320 as specified by the C++ standard.
1321 If @option{-fno-for-scope} is specified, the scope of variables declared in
1322 a @i{for-init-statement} extends to the end of the enclosing scope,
1323 as was the case in old versions of G++, and other (traditional)
1324 implementations of C++.
1326 The default if neither flag is given to follow the standard,
1327 but to allow and give a warning for old-style code that would
1328 otherwise be invalid, or have different behavior.
1330 @item -fno-gnu-keywords
1331 @opindex fno-gnu-keywords
1332 Do not recognize @code{typeof} as a keyword, so that code can use this
1333 word as an identifier. You can use the keyword @code{__typeof__} instead.
1334 @option{-ansi} implies @option{-fno-gnu-keywords}.
1336 @item -fno-implicit-templates
1337 @opindex fno-implicit-templates
1338 Never emit code for non-inline templates which are instantiated
1339 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1340 @xref{Template Instantiation}, for more information.
1342 @item -fno-implicit-inline-templates
1343 @opindex fno-implicit-inline-templates
1344 Don't emit code for implicit instantiations of inline templates, either.
1345 The default is to handle inlines differently so that compiles with and
1346 without optimization will need the same set of explicit instantiations.
1348 @item -fno-implement-inlines
1349 @opindex fno-implement-inlines
1350 To save space, do not emit out-of-line copies of inline functions
1351 controlled by @samp{#pragma implementation}. This will cause linker
1352 errors if these functions are not inlined everywhere they are called.
1354 @item -fms-extensions
1355 @opindex fms-extensions
1356 Disable pedantic warnings about constructs used in MFC, such as implicit
1357 int and getting a pointer to member function via non-standard syntax.
1359 @item -fno-nonansi-builtins
1360 @opindex fno-nonansi-builtins
1361 Disable built-in declarations of functions that are not mandated by
1362 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1363 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1365 @item -fno-operator-names
1366 @opindex fno-operator-names
1367 Do not treat the operator name keywords @code{and}, @code{bitand},
1368 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1369 synonyms as keywords.
1371 @item -fno-optional-diags
1372 @opindex fno-optional-diags
1373 Disable diagnostics that the standard says a compiler does not need to
1374 issue. Currently, the only such diagnostic issued by G++ is the one for
1375 a name having multiple meanings within a class.
1378 @opindex fpermissive
1379 Downgrade messages about nonconformant code from errors to warnings. By
1380 default, G++ effectively sets @option{-pedantic-errors} without
1381 @option{-pedantic}; this option reverses that. This behavior and this
1382 option are superseded by @option{-pedantic}, which works as it does for GNU C@.
1386 Enable automatic template instantiation at link time. This option also
1387 implies @option{-fno-implicit-templates}. @xref{Template
1388 Instantiation}, for more information.
1392 Disable generation of information about every class with virtual
1393 functions for use by the C++ runtime type identification features
1394 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1395 of the language, you can save some space by using this flag. Note that
1396 exception handling uses the same information, but it will generate it as
1401 Emit statistics about front-end processing at the end of the compilation.
1402 This information is generally only useful to the G++ development team.
1404 @item -ftemplate-depth-@var{n}
1405 @opindex ftemplate-depth
1406 Set the maximum instantiation depth for template classes to @var{n}.
1407 A limit on the template instantiation depth is needed to detect
1408 endless recursions during template class instantiation. ANSI/ISO C++
1409 conforming programs must not rely on a maximum depth greater than 17.
1411 @item -fuse-cxa-atexit
1412 @opindex fuse-cxa-atexit
1413 Register destructors for objects with static storage duration with the
1414 @code{__cxa_atexit} function rather than the @code{atexit} function.
1415 This option is required for fully standards-compliant handling of static
1416 destructors, but will only work if your C library supports
1417 @code{__cxa_atexit}.
1421 Emit special relocations for vtables and virtual function references
1422 so that the linker can identify unused virtual functions and zero out
1423 vtable slots that refer to them. This is most useful with
1424 @option{-ffunction-sections} and @option{-Wl,--gc-sections}, in order to
1425 also discard the functions themselves.
1427 This optimization requires GNU as and GNU ld. Not all systems support
1428 this option. @option{-Wl,--gc-sections} is ignored without @option{-static}.
1432 Do not use weak symbol support, even if it is provided by the linker.
1433 By default, G++ will use weak symbols if they are available. This
1434 option exists only for testing, and should not be used by end-users;
1435 it will result in inferior code and has no benefits. This option may
1436 be removed in a future release of G++.
1440 Do not search for header files in the standard directories specific to
1441 C++, but do still search the other standard directories. (This option
1442 is used when building the C++ library.)
1445 In addition, these optimization, warning, and code generation options
1446 have meanings only for C++ programs:
1449 @item -fno-default-inline
1450 @opindex fno-default-inline
1451 Do not assume @samp{inline} for functions defined inside a class scope.
1452 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1453 functions will have linkage like inline functions; they just won't be
1456 @item -Wctor-dtor-privacy @r{(C++ only)}
1457 @opindex Wctor-dtor-privacy
1458 Warn when a class seems unusable, because all the constructors or
1459 destructors in a class are private and the class has no friends or
1460 public static member functions.
1462 @item -Wnon-virtual-dtor @r{(C++ only)}
1463 @opindex Wnon-virtual-dtor
1464 Warn when a class declares a non-virtual destructor that should probably
1465 be virtual, because it looks like the class will be used polymorphically.
1467 @item -Wreorder @r{(C++ only)}
1469 @cindex reordering, warning
1470 @cindex warning for reordering of member initializers
1471 Warn when the order of member initializers given in the code does not
1472 match the order in which they must be executed. For instance:
1478 A(): j (0), i (1) @{ @}
1482 Here the compiler will warn that the member initializers for @samp{i}
1483 and @samp{j} will be rearranged to match the declaration order of the
1487 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1490 @item -Weffc++ @r{(C++ only)}
1492 Warn about violations of various style guidelines from Scott Meyers'
1493 @cite{Effective C++} books. If you use this option, you should be aware
1494 that the standard library headers do not obey all of these guidelines;
1495 you can use @samp{grep -v} to filter out those warnings.
1497 @item -Wno-deprecated @r{(C++ only)}
1498 @opindex Wno-deprecated
1499 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1501 @item -Wno-non-template-friend @r{(C++ only)}
1502 @opindex Wno-non-template-friend
1503 Disable warnings when non-templatized friend functions are declared
1504 within a template. With the advent of explicit template specification
1505 support in G++, if the name of the friend is an unqualified-id (i.e.,
1506 @samp{friend foo(int)}), the C++ language specification demands that the
1507 friend declare or define an ordinary, nontemplate function. (Section
1508 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1509 could be interpreted as a particular specialization of a templatized
1510 function. Because this non-conforming behavior is no longer the default
1511 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1512 check existing code for potential trouble spots, and is on by default.
1513 This new compiler behavior can be turned off with
1514 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1515 but disables the helpful warning.
1517 @item -Wold-style-cast @r{(C++ only)}
1518 @opindex Wold-style-cast
1519 Warn if an old-style (C-style) cast is used within a C++ program. The
1520 new-style casts (@samp{static_cast}, @samp{reinterpret_cast}, and
1521 @samp{const_cast}) are less vulnerable to unintended effects, and much
1524 @item -Woverloaded-virtual @r{(C++ only)}
1525 @opindex Woverloaded-virtual
1526 @cindex overloaded virtual fn, warning
1527 @cindex warning for overloaded virtual fn
1528 Warn when a function declaration hides virtual functions from a
1529 base class. For example, in:
1536 struct B: public A @{
1541 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1549 will fail to compile.
1551 @item -Wno-pmf-conversions @r{(C++ only)}
1552 @opindex Wno-pmf-conversions
1553 Disable the diagnostic for converting a bound pointer to member function
1556 @item -Wsign-promo @r{(C++ only)}
1557 @opindex Wsign-promo
1558 Warn when overload resolution chooses a promotion from unsigned or
1559 enumeral type to a signed type over a conversion to an unsigned type of
1560 the same size. Previous versions of G++ would try to preserve
1561 unsignedness, but the standard mandates the current behavior.
1563 @item -Wsynth @r{(C++ only)}
1565 @cindex warning for synthesized methods
1566 @cindex synthesized methods, warning
1567 Warn when G++'s synthesis behavior does not match that of cfront. For
1573 A& operator = (int);
1583 In this example, G++ will synthesize a default @samp{A& operator =
1584 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1587 @node Objective-C Dialect Options
1588 @section Options Controlling Objective-C Dialect
1590 @cindex compiler options, Objective-C
1591 @cindex Objective-C options, command line
1592 @cindex options, Objective-C
1593 This section describes the command-line options that are only meaningful
1594 for Objective-C programs; but you can also use most of the GNU compiler
1595 options regardless of what language your program is in. For example,
1596 you might compile a file @code{some_class.m} like this:
1599 gcc -g -fgnu-runtime -O -c some_class.m
1603 In this example, only @option{-fgnu-runtime} is an option meant only for
1604 Objective-C programs; you can use the other options with any language
1607 Here is a list of options that are @emph{only} for compiling Objective-C
1611 @item -fconstant-string-class=@var{class-name}
1612 @opindex fconstant-string-class
1613 Use @var{class-name} as the name of the class to instantiate for each
1614 literal string specified with the syntax @code{@@"@dots{}"}. The default
1615 class name is @code{NXConstantString}.
1618 @opindex fgnu-runtime
1619 Generate object code compatible with the standard GNU Objective-C
1620 runtime. This is the default for most types of systems.
1622 @item -fnext-runtime
1623 @opindex fnext-runtime
1624 Generate output compatible with the NeXT runtime. This is the default
1625 for NeXT-based systems, including Darwin and Mac OS X@.
1629 Dump interface declarations for all classes seen in the source file to a
1630 file named @file{@var{sourcename}.decl}.
1633 @opindex Wno-protocol
1634 Do not warn if methods required by a protocol are not implemented
1635 in the class adopting it.
1639 Warn if a selector has multiple methods of different types defined.
1641 @c not documented because only avail via -Wp
1642 @c @item -print-objc-runtime-info
1646 @node Language Independent Options
1647 @section Options to Control Diagnostic Messages Formatting
1648 @cindex options to control diagnostics formatting
1649 @cindex diagnostic messages
1650 @cindex message formatting
1652 Traditionally, diagnostic messages have been formatted irrespective of
1653 the output device's aspect (e.g.@: its width, @dots{}). The options described
1654 below can be used to control the diagnostic messages formatting
1655 algorithm, e.g.@: how many characters per line, how often source location
1656 information should be reported. Right now, only the C++ front end can
1657 honor these options. However it is expected, in the near future, that
1658 the remaining front ends would be able to digest them correctly.
1661 @item -fmessage-length=@var{n}
1662 @opindex fmessage-length
1663 Try to format error messages so that they fit on lines of about @var{n}
1664 characters. The default is 72 characters for @command{g++} and 0 for the rest of
1665 the front ends supported by GCC@. If @var{n} is zero, then no
1666 line-wrapping will be done; each error message will appear on a single
1669 @opindex fdiagnostics-show-location
1670 @item -fdiagnostics-show-location=once
1671 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1672 reporter to emit @emph{once} source location information; that is, in
1673 case the message is too long to fit on a single physical line and has to
1674 be wrapped, the source location won't be emitted (as prefix) again,
1675 over and over, in subsequent continuation lines. This is the default
1678 @item -fdiagnostics-show-location=every-line
1679 Only meaningful in line-wrapping mode. Instructs the diagnostic
1680 messages reporter to emit the same source location information (as
1681 prefix) for physical lines that result from the process of breaking
1682 a message which is too long to fit on a single line.
1686 @node Warning Options
1687 @section Options to Request or Suppress Warnings
1688 @cindex options to control warnings
1689 @cindex warning messages
1690 @cindex messages, warning
1691 @cindex suppressing warnings
1693 Warnings are diagnostic messages that report constructions which
1694 are not inherently erroneous but which are risky or suggest there
1695 may have been an error.
1697 You can request many specific warnings with options beginning @samp{-W},
1698 for example @option{-Wimplicit} to request warnings on implicit
1699 declarations. Each of these specific warning options also has a
1700 negative form beginning @samp{-Wno-} to turn off warnings;
1701 for example, @option{-Wno-implicit}. This manual lists only one of the
1702 two forms, whichever is not the default.
1704 These options control the amount and kinds of warnings produced by GCC:
1707 @cindex syntax checking
1709 @opindex fsyntax-only
1710 Check the code for syntax errors, but don't do anything beyond that.
1714 Issue all the warnings demanded by strict ISO C and ISO C++;
1715 reject all programs that use forbidden extensions, and some other
1716 programs that do not follow ISO C and ISO C++. For ISO C, follows the
1717 version of the ISO C standard specified by any @option{-std} option used.
1719 Valid ISO C and ISO C++ programs should compile properly with or without
1720 this option (though a rare few will require @option{-ansi} or a
1721 @option{-std} option specifying the required version of ISO C)@. However,
1722 without this option, certain GNU extensions and traditional C and C++
1723 features are supported as well. With this option, they are rejected.
1725 @option{-pedantic} does not cause warning messages for use of the
1726 alternate keywords whose names begin and end with @samp{__}. Pedantic
1727 warnings are also disabled in the expression that follows
1728 @code{__extension__}. However, only system header files should use
1729 these escape routes; application programs should avoid them.
1730 @xref{Alternate Keywords}.
1732 Some users try to use @option{-pedantic} to check programs for strict ISO
1733 C conformance. They soon find that it does not do quite what they want:
1734 it finds some non-ISO practices, but not all---only those for which
1735 ISO C @emph{requires} a diagnostic, and some others for which
1736 diagnostics have been added.
1738 A feature to report any failure to conform to ISO C might be useful in
1739 some instances, but would require considerable additional work and would
1740 be quite different from @option{-pedantic}. We don't have plans to
1741 support such a feature in the near future.
1743 Where the standard specified with @option{-std} represents a GNU
1744 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
1745 corresponding @dfn{base standard}, the version of ISO C on which the GNU
1746 extended dialect is based. Warnings from @option{-pedantic} are given
1747 where they are required by the base standard. (It would not make sense
1748 for such warnings to be given only for features not in the specified GNU
1749 C dialect, since by definition the GNU dialects of C include all
1750 features the compiler supports with the given option, and there would be
1751 nothing to warn about.)
1753 @item -pedantic-errors
1754 @opindex pedantic-errors
1755 Like @option{-pedantic}, except that errors are produced rather than
1760 Inhibit all warning messages.
1764 Inhibit warning messages about the use of @samp{#import}.
1766 @item -Wchar-subscripts
1767 @opindex Wchar-subscripts
1768 Warn if an array subscript has type @code{char}. This is a common cause
1769 of error, as programmers often forget that this type is signed on some
1774 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
1775 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
1779 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
1780 the arguments supplied have types appropriate to the format string
1781 specified, and that the conversions specified in the format string make
1782 sense. This includes standard functions, and others specified by format
1783 attributes (@pxref{Function Attributes}), in the @code{printf},
1784 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
1785 not in the C standard) families.
1787 The formats are checked against the format features supported by GNU
1788 libc version 2.2. These include all ISO C89 and C99 features, as well
1789 as features from the Single Unix Specification and some BSD and GNU
1790 extensions. Other library implementations may not support all these
1791 features; GCC does not support warning about features that go beyond a
1792 particular library's limitations. However, if @option{-pedantic} is used
1793 with @option{-Wformat}, warnings will be given about format features not
1794 in the selected standard version (but not for @code{strfmon} formats,
1795 since those are not in any version of the C standard). @xref{C Dialect
1796 Options,,Options Controlling C Dialect}.
1798 @option{-Wformat} is included in @option{-Wall}. For more control over some
1799 aspects of format checking, the options @option{-Wno-format-y2k},
1800 @option{-Wno-format-extra-args}, @option{-Wformat-nonliteral},
1801 @option{-Wformat-security} and @option{-Wformat=2} are available, but are
1802 not included in @option{-Wall}.
1804 @item -Wno-format-y2k
1805 @opindex Wno-format-y2k
1806 If @option{-Wformat} is specified, do not warn about @code{strftime}
1807 formats which may yield only a two-digit year.
1809 @item -Wno-format-extra-args
1810 @opindex Wno-format-extra-args
1811 If @option{-Wformat} is specified, do not warn about excess arguments to a
1812 @code{printf} or @code{scanf} format function. The C standard specifies
1813 that such arguments are ignored.
1815 @item -Wformat-nonliteral
1816 @opindex Wformat-nonliteral
1817 If @option{-Wformat} is specified, also warn if the format string is not a
1818 string literal and so cannot be checked, unless the format function
1819 takes its format arguments as a @code{va_list}.
1821 @item -Wformat-security
1822 @opindex Wformat-security
1823 If @option{-Wformat} is specified, also warn about uses of format
1824 functions that represent possible security problems. At present, this
1825 warns about calls to @code{printf} and @code{scanf} functions where the
1826 format string is not a string literal and there are no format arguments,
1827 as in @code{printf (foo);}. This may be a security hole if the format
1828 string came from untrusted input and contains @samp{%n}. (This is
1829 currently a subset of what @option{-Wformat-nonliteral} warns about, but
1830 in future warnings may be added to @option{-Wformat-security} that are not
1831 included in @option{-Wformat-nonliteral}.)
1835 Enable @option{-Wformat} plus format checks not included in
1836 @option{-Wformat}. Currently equivalent to @samp{-Wformat
1837 -Wformat-nonliteral -Wformat-security}.
1839 @item -Wimplicit-int
1840 @opindex Wimplicit-int
1841 Warn when a declaration does not specify a type.
1843 @item -Wimplicit-function-declaration
1844 @itemx -Werror-implicit-function-declaration
1845 @opindex Wimplicit-function-declaration
1846 @opindex Werror-implicit-function-declaration
1847 Give a warning (or error) whenever a function is used before being
1852 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
1856 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
1857 function with external linkage, returning int, taking either zero
1858 arguments, two, or three arguments of appropriate types.
1860 @item -Wmissing-braces
1861 @opindex Wmissing-braces
1862 Warn if an aggregate or union initializer is not fully bracketed. In
1863 the following example, the initializer for @samp{a} is not fully
1864 bracketed, but that for @samp{b} is fully bracketed.
1867 int a[2][2] = @{ 0, 1, 2, 3 @};
1868 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
1872 @opindex Wparentheses
1873 Warn if parentheses are omitted in certain contexts, such
1874 as when there is an assignment in a context where a truth value
1875 is expected, or when operators are nested whose precedence people
1876 often get confused about.
1878 Also warn about constructions where there may be confusion to which
1879 @code{if} statement an @code{else} branch belongs. Here is an example of
1894 In C, every @code{else} branch belongs to the innermost possible @code{if}
1895 statement, which in this example is @code{if (b)}. This is often not
1896 what the programmer expected, as illustrated in the above example by
1897 indentation the programmer chose. When there is the potential for this
1898 confusion, GCC will issue a warning when this flag is specified.
1899 To eliminate the warning, add explicit braces around the innermost
1900 @code{if} statement so there is no way the @code{else} could belong to
1901 the enclosing @code{if}. The resulting code would look like this:
1917 @item -Wsequence-point
1918 @opindex Wsequence-point
1919 Warn about code that may have undefined semantics because of violations
1920 of sequence point rules in the C standard.
1922 The C standard defines the order in which expressions in a C program are
1923 evaluated in terms of @dfn{sequence points}, which represent a partial
1924 ordering between the execution of parts of the program: those executed
1925 before the sequence point, and those executed after it. These occur
1926 after the evaluation of a full expression (one which is not part of a
1927 larger expression), after the evaluation of the first operand of a
1928 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
1929 function is called (but after the evaluation of its arguments and the
1930 expression denoting the called function), and in certain other places.
1931 Other than as expressed by the sequence point rules, the order of
1932 evaluation of subexpressions of an expression is not specified. All
1933 these rules describe only a partial order rather than a total order,
1934 since, for example, if two functions are called within one expression
1935 with no sequence point between them, the order in which the functions
1936 are called is not specified. However, the standards committee have
1937 ruled that function calls do not overlap.
1939 It is not specified when between sequence points modifications to the
1940 values of objects take effect. Programs whose behavior depends on this
1941 have undefined behavior; the C standard specifies that ``Between the
1942 previous and next sequence point an object shall have its stored value
1943 modified at most once by the evaluation of an expression. Furthermore,
1944 the prior value shall be read only to determine the value to be
1945 stored.''. If a program breaks these rules, the results on any
1946 particular implementation are entirely unpredictable.
1948 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
1949 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
1950 diagnosed by this option, and it may give an occasional false positive
1951 result, but in general it has been found fairly effective at detecting
1952 this sort of problem in programs.
1954 The present implementation of this option only works for C programs. A
1955 future implementation may also work for C++ programs.
1957 The C standard is worded confusingly, therefore there is some debate
1958 over the precise meaning of the sequence point rules in subtle cases.
1959 Links to discussions of the problem, including proposed formal
1960 definitions, may be found on our readings page, at
1961 @w{@uref{http://gcc.gnu.org/readings.html}}.
1964 @opindex Wreturn-type
1965 Warn whenever a function is defined with a return-type that defaults to
1966 @code{int}. Also warn about any @code{return} statement with no
1967 return-value in a function whose return-type is not @code{void}.
1969 For C++, a function without return type always produces a diagnostic
1970 message, even when @option{-Wno-return-type} is specified. The only
1971 exceptions are @samp{main} and functions defined in system headers.
1975 Warn whenever a @code{switch} statement has an index of enumeral type
1976 and lacks a @code{case} for one or more of the named codes of that
1977 enumeration. (The presence of a @code{default} label prevents this
1978 warning.) @code{case} labels outside the enumeration range also
1979 provoke warnings when this option is used.
1983 Warn if any trigraphs are encountered that might change the meaning of
1984 the program (trigraphs within comments are not warned about).
1986 @item -Wunused-function
1987 @opindex Wunused-function
1988 Warn whenever a static function is declared but not defined or a
1989 non\-inline static function is unused.
1991 @item -Wunused-label
1992 @opindex Wunused-label
1993 Warn whenever a label is declared but not used.
1995 To suppress this warning use the @samp{unused} attribute
1996 (@pxref{Variable Attributes}).
1998 @item -Wunused-parameter
1999 @opindex Wunused-parameter
2000 Warn whenever a function parameter is unused aside from its declaration.
2002 To suppress this warning use the @samp{unused} attribute
2003 (@pxref{Variable Attributes}).
2005 @item -Wunused-variable
2006 @opindex Wunused-variable
2007 Warn whenever a local variable or non-constant static variable is unused
2008 aside from its declaration
2010 To suppress this warning use the @samp{unused} attribute
2011 (@pxref{Variable Attributes}).
2013 @item -Wunused-value
2014 @opindex Wunused-value
2015 Warn whenever a statement computes a result that is explicitly not used.
2017 To suppress this warning cast the expression to @samp{void}.
2021 All all the above @option{-Wunused} options combined.
2023 In order to get a warning about an unused function parameter, you must
2024 either specify @samp{-W -Wunused} or separately specify
2025 @option{-Wunused-parameter}.
2027 @item -Wuninitialized
2028 @opindex Wuninitialized
2029 Warn if an automatic variable is used without first being initialized or
2030 if a variable may be clobbered by a @code{setjmp} call.
2032 These warnings are possible only in optimizing compilation,
2033 because they require data flow information that is computed only
2034 when optimizing. If you don't specify @option{-O}, you simply won't
2037 These warnings occur only for variables that are candidates for
2038 register allocation. Therefore, they do not occur for a variable that
2039 is declared @code{volatile}, or whose address is taken, or whose size
2040 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2041 structures, unions or arrays, even when they are in registers.
2043 Note that there may be no warning about a variable that is used only
2044 to compute a value that itself is never used, because such
2045 computations may be deleted by data flow analysis before the warnings
2048 These warnings are made optional because GCC is not smart
2049 enough to see all the reasons why the code might be correct
2050 despite appearing to have an error. Here is one example of how
2071 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2072 always initialized, but GCC doesn't know this. Here is
2073 another common case:
2078 if (change_y) save_y = y, y = new_y;
2080 if (change_y) y = save_y;
2085 This has no bug because @code{save_y} is used only if it is set.
2087 @cindex @code{longjmp} warnings
2088 This option also warns when a non-volatile automatic variable might be
2089 changed by a call to @code{longjmp}. These warnings as well are possible
2090 only in optimizing compilation.
2092 The compiler sees only the calls to @code{setjmp}. It cannot know
2093 where @code{longjmp} will be called; in fact, a signal handler could
2094 call it at any point in the code. As a result, you may get a warning
2095 even when there is in fact no problem because @code{longjmp} cannot
2096 in fact be called at the place which would cause a problem.
2098 Some spurious warnings can be avoided if you declare all the functions
2099 you use that never return as @code{noreturn}. @xref{Function
2102 @item -Wreorder @r{(C++ only)}
2104 @cindex reordering, warning
2105 @cindex warning for reordering of member initializers
2106 Warn when the order of member initializers given in the code does not
2107 match the order in which they must be executed. For instance:
2109 @item -Wunknown-pragmas
2110 @opindex Wunknown-pragmas
2111 @cindex warning for unknown pragmas
2112 @cindex unknown pragmas, warning
2113 @cindex pragmas, warning of unknown
2114 Warn when a #pragma directive is encountered which is not understood by
2115 GCC@. If this command line option is used, warnings will even be issued
2116 for unknown pragmas in system header files. This is not the case if
2117 the warnings were only enabled by the @option{-Wall} command line option.
2121 All of the above @samp{-W} options combined. This enables all the
2122 warnings about constructions that some users consider questionable, and
2123 that are easy to avoid (or modify to prevent the warning), even in
2124 conjunction with macros.
2127 @opindex Wno-div-by-zero
2128 @opindex Wdiv-by-zero
2129 Warn about compile-time integer division by zero. This is default. To
2130 inhibit the warning messages, use @option{-Wno-div-by-zero}. Floating
2131 point division by zero is not warned about, as it can be a legitimate
2132 way of obtaining infinities and NaNs.
2135 @opindex Wno-multichar
2137 Warn if a multicharacter constant (@samp{'FOOF'}) is used. This is
2138 default. To inhibit the warning messages, use @option{-Wno-multichar}.
2139 Usually they indicate a typo in the user's code, as they have
2140 implementation-defined values, and should not be used in portable code.
2142 @item -Wsystem-headers
2143 @opindex Wsystem-headers
2144 @cindex warnings from system headers
2145 @cindex system headers, warnings from
2146 Print warning messages for constructs found in system header files.
2147 Warnings from system headers are normally suppressed, on the assumption
2148 that they usually do not indicate real problems and would only make the
2149 compiler output harder to read. Using this command line option tells
2150 GCC to emit warnings from system headers as if they occurred in user
2151 code. However, note that using @option{-Wall} in conjunction with this
2152 option will @emph{not} warn about unknown pragmas in system
2153 headers---for that, @option{-Wunknown-pragmas} must also be used.
2156 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2157 Some of them warn about constructions that users generally do not
2158 consider questionable, but which occasionally you might wish to check
2159 for; others warn about constructions that are necessary or hard to avoid
2160 in some cases, and there is no simple way to modify the code to suppress
2166 Print extra warning messages for these events:
2170 A function can return either with or without a value. (Falling
2171 off the end of the function body is considered returning without
2172 a value.) For example, this function would evoke such a
2186 An expression-statement or the left-hand side of a comma expression
2187 contains no side effects.
2188 To suppress the warning, cast the unused expression to void.
2189 For example, an expression such as @samp{x[i,j]} will cause a warning,
2190 but @samp{x[(void)i,j]} will not.
2193 An unsigned value is compared against zero with @samp{<} or @samp{<=}.
2196 A comparison like @samp{x<=y<=z} appears; this is equivalent to
2197 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
2198 that of ordinary mathematical notation.
2201 Storage-class specifiers like @code{static} are not the first things in
2202 a declaration. According to the C Standard, this usage is obsolescent.
2205 The return type of a function has a type qualifier such as @code{const}.
2206 Such a type qualifier has no effect, since the value returned by a
2207 function is not an lvalue. (But don't warn about the GNU extension of
2208 @code{volatile void} return types. That extension will be warned about
2209 if @option{-pedantic} is specified.)
2212 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2216 A comparison between signed and unsigned values could produce an
2217 incorrect result when the signed value is converted to unsigned.
2218 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2221 An aggregate has a partly bracketed initializer.
2222 For example, the following code would evoke such a warning,
2223 because braces are missing around the initializer for @code{x.h}:
2226 struct s @{ int f, g; @};
2227 struct t @{ struct s h; int i; @};
2228 struct t x = @{ 1, 2, 3 @};
2232 An aggregate has an initializer which does not initialize all members.
2233 For example, the following code would cause such a warning, because
2234 @code{x.h} would be implicitly initialized to zero:
2237 struct s @{ int f, g, h; @};
2238 struct s x = @{ 3, 4 @};
2243 @opindex Wfloat-equal
2244 Warn if floating point values are used in equality comparisons.
2246 The idea behind this is that sometimes it is convenient (for the
2247 programmer) to consider floating-point values as approximations to
2248 infinitely precise real numbers. If you are doing this, then you need
2249 to compute (by analysing the code, or in some other way) the maximum or
2250 likely maximum error that the computation introduces, and allow for it
2251 when performing comparisons (and when producing output, but that's a
2252 different problem). In particular, instead of testing for equality, you
2253 would check to see whether the two values have ranges that overlap; and
2254 this is done with the relational operators, so equality comparisons are
2257 @item -Wtraditional @r{(C only)}
2258 @opindex Wtraditional
2259 Warn about certain constructs that behave differently in traditional and
2260 ISO C@. Also warn about ISO C constructs that have no traditional C
2261 equivalent, and/or problematic constructs which should be avoided.
2265 Macro parameters that appear within string literals in the macro body.
2266 In traditional C macro replacement takes place within string literals,
2267 but does not in ISO C@.
2270 In traditional C, some preprocessor directives did not exist.
2271 Traditional preprocessors would only consider a line to be a directive
2272 if the @samp{#} appeared in column 1 on the line. Therefore
2273 @option{-Wtraditional} warns about directives that traditional C
2274 understands but would ignore because the @samp{#} does not appear as the
2275 first character on the line. It also suggests you hide directives like
2276 @samp{#pragma} not understood by traditional C by indenting them. Some
2277 traditional implementations would not recognise @samp{#elif}, so it
2278 suggests avoiding it altogether.
2281 A function-like macro that appears without arguments.
2284 The unary plus operator.
2287 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2288 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2289 constants.) Note, these suffixes appear in macros defined in the system
2290 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2291 Use of these macros in user code might normally lead to spurious
2292 warnings, however gcc's integrated preprocessor has enough context to
2293 avoid warning in these cases.
2296 A function declared external in one block and then used after the end of
2300 A @code{switch} statement has an operand of type @code{long}.
2303 A non-@code{static} function declaration follows a @code{static} one.
2304 This construct is not accepted by some traditional C compilers.
2307 The ISO type of an integer constant has a different width or
2308 signedness from its traditional type. This warning is only issued if
2309 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2310 typically represent bit patterns, are not warned about.
2313 Usage of ISO string concatenation is detected.
2316 Initialization of automatic aggregates.
2319 Identifier conflicts with labels. Traditional C lacks a separate
2320 namespace for labels.
2323 Initialization of unions. If the initializer is zero, the warning is
2324 omitted. This is done under the assumption that the zero initializer in
2325 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2326 initializer warnings and relies on default initialization to zero in the
2330 Conversions by prototypes between fixed/floating point values and vice
2331 versa. The absence of these prototypes when compiling with traditional
2332 C would cause serious problems. This is a subset of the possible
2333 conversion warnings, for the full set use @option{-Wconversion}.
2338 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2342 Warn whenever a local variable shadows another local variable, parameter or
2343 global variable or whenever a built-in function is shadowed.
2345 @item -Wlarger-than-@var{len}
2346 @opindex Wlarger-than
2347 Warn whenever an object of larger than @var{len} bytes is defined.
2349 @item -Wpointer-arith
2350 @opindex Wpointer-arith
2351 Warn about anything that depends on the ``size of'' a function type or
2352 of @code{void}. GNU C assigns these types a size of 1, for
2353 convenience in calculations with @code{void *} pointers and pointers
2356 @item -Wbad-function-cast @r{(C only)}
2357 @opindex Wbad-function-cast
2358 Warn whenever a function call is cast to a non-matching type.
2359 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2363 Warn whenever a pointer is cast so as to remove a type qualifier from
2364 the target type. For example, warn if a @code{const char *} is cast
2365 to an ordinary @code{char *}.
2368 @opindex Wcast-align
2369 Warn whenever a pointer is cast such that the required alignment of the
2370 target is increased. For example, warn if a @code{char *} is cast to
2371 an @code{int *} on machines where integers can only be accessed at
2372 two- or four-byte boundaries.
2374 @item -Wwrite-strings
2375 @opindex Wwrite-strings
2376 When compiling C, give string constants the type @code{const
2377 char[@var{length}]} so that
2378 copying the address of one into a non-@code{const} @code{char *}
2379 pointer will get a warning; when compiling C++, warn about the
2380 deprecated conversion from string constants to @code{char *}.
2381 These warnings will help you find at
2382 compile time code that can try to write into a string constant, but
2383 only if you have been very careful about using @code{const} in
2384 declarations and prototypes. Otherwise, it will just be a nuisance;
2385 this is why we did not make @option{-Wall} request these warnings.
2388 @opindex Wconversion
2389 Warn if a prototype causes a type conversion that is different from what
2390 would happen to the same argument in the absence of a prototype. This
2391 includes conversions of fixed point to floating and vice versa, and
2392 conversions changing the width or signedness of a fixed point argument
2393 except when the same as the default promotion.
2395 Also, warn if a negative integer constant expression is implicitly
2396 converted to an unsigned type. For example, warn about the assignment
2397 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2398 casts like @code{(unsigned) -1}.
2400 @item -Wsign-compare
2401 @opindex Wsign-compare
2402 @cindex warning for comparison of signed and unsigned values
2403 @cindex comparison of signed and unsigned values, warning
2404 @cindex signed and unsigned values, comparison warning
2405 Warn when a comparison between signed and unsigned values could produce
2406 an incorrect result when the signed value is converted to unsigned.
2407 This warning is also enabled by @option{-W}; to get the other warnings
2408 of @option{-W} without this warning, use @samp{-W -Wno-sign-compare}.
2410 @item -Waggregate-return
2411 @opindex Waggregate-return
2412 Warn if any functions that return structures or unions are defined or
2413 called. (In languages where you can return an array, this also elicits
2416 @item -Wstrict-prototypes @r{(C only)}
2417 @opindex Wstrict-prototypes
2418 Warn if a function is declared or defined without specifying the
2419 argument types. (An old-style function definition is permitted without
2420 a warning if preceded by a declaration which specifies the argument
2423 @item -Wmissing-prototypes @r{(C only)}
2424 @opindex Wmissing-prototypes
2425 Warn if a global function is defined without a previous prototype
2426 declaration. This warning is issued even if the definition itself
2427 provides a prototype. The aim is to detect global functions that fail
2428 to be declared in header files.
2430 @item -Wmissing-declarations
2431 @opindex Wmissing-declarations
2432 Warn if a global function is defined without a previous declaration.
2433 Do so even if the definition itself provides a prototype.
2434 Use this option to detect global functions that are not declared in
2437 @item -Wmissing-noreturn
2438 @opindex Wmissing-noreturn
2439 Warn about functions which might be candidates for attribute @code{noreturn}.
2440 Note these are only possible candidates, not absolute ones. Care should
2441 be taken to manually verify functions actually do not ever return before
2442 adding the @code{noreturn} attribute, otherwise subtle code generation
2443 bugs could be introduced. You will not get a warning for @code{main} in
2444 hosted C environments.
2446 @item -Wmissing-format-attribute
2447 @opindex Wmissing-format-attribute
2449 If @option{-Wformat} is enabled, also warn about functions which might be
2450 candidates for @code{format} attributes. Note these are only possible
2451 candidates, not absolute ones. GCC will guess that @code{format}
2452 attributes might be appropriate for any function that calls a function
2453 like @code{vprintf} or @code{vscanf}, but this might not always be the
2454 case, and some functions for which @code{format} attributes are
2455 appropriate may not be detected. This option has no effect unless
2456 @option{-Wformat} is enabled (possibly by @option{-Wall}).
2460 Warn if a structure is given the packed attribute, but the packed
2461 attribute has no effect on the layout or size of the structure.
2462 Such structures may be mis-aligned for little benefit. For
2463 instance, in this code, the variable @code{f.x} in @code{struct bar}
2464 will be misaligned even though @code{struct bar} does not itself
2465 have the packed attribute:
2472 @} __attribute__((packed));
2482 Warn if padding is included in a structure, either to align an element
2483 of the structure or to align the whole structure. Sometimes when this
2484 happens it is possible to rearrange the fields of the structure to
2485 reduce the padding and so make the structure smaller.
2487 @item -Wredundant-decls
2488 @opindex Wredundant-decls
2489 Warn if anything is declared more than once in the same scope, even in
2490 cases where multiple declaration is valid and changes nothing.
2492 @item -Wnested-externs @r{(C only)}
2493 @opindex Wnested-externs
2494 Warn if an @code{extern} declaration is encountered within a function.
2496 @item -Wunreachable-code
2497 @opindex Wunreachable-code
2498 Warn if the compiler detects that code will never be executed.
2500 This option is intended to warn when the compiler detects that at
2501 least a whole line of source code will never be executed, because
2502 some condition is never satisfied or because it is after a
2503 procedure that never returns.
2505 It is possible for this option to produce a warning even though there
2506 are circumstances under which part of the affected line can be executed,
2507 so care should be taken when removing apparently-unreachable code.
2509 For instance, when a function is inlined, a warning may mean that the
2510 line is unreachable in only one inlined copy of the function.
2512 This option is not made part of @option{-Wall} because in a debugging
2513 version of a program there is often substantial code which checks
2514 correct functioning of the program and is, hopefully, unreachable
2515 because the program does work. Another common use of unreachable
2516 code is to provide behaviour which is selectable at compile-time.
2520 Warn if a function can not be inlined and it was declared as inline.
2524 @opindex Wno-long-long
2525 Warn if @samp{long long} type is used. This is default. To inhibit
2526 the warning messages, use @option{-Wno-long-long}. Flags
2527 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
2528 only when @option{-pedantic} flag is used.
2530 @item -Wdisabled-optimization
2531 @opindex Wdisabled-optimization
2532 Warn if a requested optimization pass is disabled. This warning does
2533 not generally indicate that there is anything wrong with your code; it
2534 merely indicates that GCC's optimizers were unable to handle the code
2535 effectively. Often, the problem is that your code is too big or too
2536 complex; GCC will refuse to optimize programs when the optimization
2537 itself is likely to take inordinate amounts of time.
2541 Make all warnings into errors.
2544 @node Debugging Options
2545 @section Options for Debugging Your Program or GCC
2546 @cindex options, debugging
2547 @cindex debugging information options
2549 GCC has various special options that are used for debugging
2550 either your program or GCC:
2555 Produce debugging information in the operating system's native format
2556 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
2559 On most systems that use stabs format, @option{-g} enables use of extra
2560 debugging information that only GDB can use; this extra information
2561 makes debugging work better in GDB but will probably make other debuggers
2563 refuse to read the program. If you want to control for certain whether
2564 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
2565 @option{-gxcoff+}, @option{-gxcoff}, @option{-gdwarf-1+}, or @option{-gdwarf-1}
2568 Unlike most other C compilers, GCC allows you to use @option{-g} with
2569 @option{-O}. The shortcuts taken by optimized code may occasionally
2570 produce surprising results: some variables you declared may not exist
2571 at all; flow of control may briefly move where you did not expect it;
2572 some statements may not be executed because they compute constant
2573 results or their values were already at hand; some statements may
2574 execute in different places because they were moved out of loops.
2576 Nevertheless it proves possible to debug optimized output. This makes
2577 it reasonable to use the optimizer for programs that might have bugs.
2579 The following options are useful when GCC is generated with the
2580 capability for more than one debugging format.
2584 Produce debugging information for use by GDB@. This means to use the
2585 most expressive format available (DWARF 2, stabs, or the native format
2586 if neither of those are supported), including GDB extensions if at all
2591 Produce debugging information in stabs format (if that is supported),
2592 without GDB extensions. This is the format used by DBX on most BSD
2593 systems. On MIPS, Alpha and System V Release 4 systems this option
2594 produces stabs debugging output which is not understood by DBX or SDB@.
2595 On System V Release 4 systems this option requires the GNU assembler.
2599 Produce debugging information in stabs format (if that is supported),
2600 using GNU extensions understood only by the GNU debugger (GDB)@. The
2601 use of these extensions is likely to make other debuggers crash or
2602 refuse to read the program.
2606 Produce debugging information in COFF format (if that is supported).
2607 This is the format used by SDB on most System V systems prior to
2612 Produce debugging information in XCOFF format (if that is supported).
2613 This is the format used by the DBX debugger on IBM RS/6000 systems.
2617 Produce debugging information in XCOFF format (if that is supported),
2618 using GNU extensions understood only by the GNU debugger (GDB)@. The
2619 use of these extensions is likely to make other debuggers crash or
2620 refuse to read the program, and may cause assemblers other than the GNU
2621 assembler (GAS) to fail with an error.
2625 Produce debugging information in DWARF version 1 format (if that is
2626 supported). This is the format used by SDB on most System V Release 4
2631 Produce debugging information in DWARF version 1 format (if that is
2632 supported), using GNU extensions understood only by the GNU debugger
2633 (GDB)@. The use of these extensions is likely to make other debuggers
2634 crash or refuse to read the program.
2638 Produce debugging information in DWARF version 2 format (if that is
2639 supported). This is the format used by DBX on IRIX 6.
2642 @itemx -ggdb@var{level}
2643 @itemx -gstabs@var{level}
2644 @itemx -gcoff@var{level}
2645 @itemx -gxcoff@var{level}
2646 @itemx -gdwarf@var{level}
2647 @itemx -gdwarf-2@var{level}
2648 Request debugging information and also use @var{level} to specify how
2649 much information. The default level is 2.
2651 Level 1 produces minimal information, enough for making backtraces in
2652 parts of the program that you don't plan to debug. This includes
2653 descriptions of functions and external variables, but no information
2654 about local variables and no line numbers.
2656 Level 3 includes extra information, such as all the macro definitions
2657 present in the program. Some debuggers support macro expansion when
2658 you use @option{-g3}.
2663 Generate extra code to write profile information suitable for the
2664 analysis program @code{prof}. You must use this option when compiling
2665 the source files you want data about, and you must also use it when
2668 @cindex @code{gprof}
2671 Generate extra code to write profile information suitable for the
2672 analysis program @code{gprof}. You must use this option when compiling
2673 the source files you want data about, and you must also use it when
2679 Generate extra code to write profile information for basic blocks, which will
2680 record the number of times each basic block is executed, the basic block start
2681 address, and the function name containing the basic block. If @option{-g} is
2682 used, the line number and filename of the start of the basic block will also be
2683 recorded. If not overridden by the machine description, the default action is
2684 to append to the text file @file{bb.out}.
2686 This data could be analyzed by a program like @code{tcov}. Note,
2687 however, that the format of the data is not what @code{tcov} expects.
2688 Eventually GNU @code{gprof} should be extended to process this data.
2692 Makes the compiler print out each function name as it is compiled, and
2693 print some statistics about each pass when it finishes.
2696 @opindex ftime-report
2697 Makes the compiler print some statistics about the time consumed by each
2698 pass when it finishes.
2701 @opindex fmem-report
2702 Makes the compiler print some statistics about permanent memory
2703 allocation when it finishes.
2707 Generate extra code to profile basic blocks. Your executable will
2708 produce output that is a superset of that produced when @option{-a} is
2709 used. Additional output is the source and target address of the basic
2710 blocks where a jump takes place, the number of times a jump is executed,
2711 and (optionally) the complete sequence of basic blocks being executed.
2712 The output is appended to file @file{bb.out}.
2714 You can examine different profiling aspects without recompilation. Your
2715 executable will read a list of function names from file @file{bb.in}.
2716 Profiling starts when a function on the list is entered and stops when
2717 that invocation is exited. To exclude a function from profiling, prefix
2718 its name with @samp{-}. If a function name is not unique, you can
2719 disambiguate it by writing it in the form
2720 @samp{/path/filename.d:functionname}. Your executable will write the
2721 available paths and filenames in file @file{bb.out}.
2723 Several function names have a special meaning:
2726 Write source, target and frequency of jumps to file @file{bb.out}.
2727 @item __bb_hidecall__
2728 Exclude function calls from frequency count.
2729 @item __bb_showret__
2730 Include function returns in frequency count.
2732 Write the sequence of basic blocks executed to file @file{bbtrace.gz}.
2733 The file will be compressed using the program @samp{gzip}, which must
2734 exist in your @env{PATH}. On systems without the @samp{popen}
2735 function, the file will be named @file{bbtrace} and will not be
2736 compressed. @strong{Profiling for even a few seconds on these systems
2737 will produce a very large file.} Note: @code{__bb_hidecall__} and
2738 @code{__bb_showret__} will not affect the sequence written to
2742 Here's a short example using different profiling parameters
2743 in file @file{bb.in}. Assume function @code{foo} consists of basic blocks
2744 1 and 2 and is called twice from block 3 of function @code{main}. After
2745 the calls, block 3 transfers control to block 4 of @code{main}.
2747 With @code{__bb_trace__} and @code{main} contained in file @file{bb.in},
2748 the following sequence of blocks is written to file @file{bbtrace.gz}:
2749 0 3 1 2 1 2 4. The return from block 2 to block 3 is not shown, because
2750 the return is to a point inside the block and not to the top. The
2751 block address 0 always indicates, that control is transferred
2752 to the trace from somewhere outside the observed functions. With
2753 @samp{-foo} added to @file{bb.in}, the blocks of function
2754 @code{foo} are removed from the trace, so only 0 3 4 remains.
2756 With @code{__bb_jumps__} and @code{main} contained in file @file{bb.in},
2757 jump frequencies will be written to file @file{bb.out}. The
2758 frequencies are obtained by constructing a trace of blocks
2759 and incrementing a counter for every neighbouring pair of blocks
2760 in the trace. The trace 0 3 1 2 1 2 4 displays the following
2764 Jump from block 0x0 to block 0x3 executed 1 time(s)
2765 Jump from block 0x3 to block 0x1 executed 1 time(s)
2766 Jump from block 0x1 to block 0x2 executed 2 time(s)
2767 Jump from block 0x2 to block 0x1 executed 1 time(s)
2768 Jump from block 0x2 to block 0x4 executed 1 time(s)
2771 With @code{__bb_hidecall__}, control transfer due to call instructions
2772 is removed from the trace, that is the trace is cut into three parts: 0
2773 3 4, 0 1 2 and 0 1 2. With @code{__bb_showret__}, control transfer due
2774 to return instructions is added to the trace. The trace becomes: 0 3 1
2775 2 3 1 2 3 4. Note, that this trace is not the same, as the sequence
2776 written to @file{bbtrace.gz}. It is solely used for counting jump
2779 @item -fprofile-arcs
2780 @opindex fprofile-arcs
2781 Instrument @dfn{arcs} during compilation to generate coverage data
2782 or for profile-directed block ordering. During execution the program
2783 records how many times each branch is executed and how many times it is
2784 taken. When the compiled program exits it saves this data to a file
2785 called @file{@var{sourcename}.da} for each source file.
2787 For profile-directed block ordering, compile the program with
2788 @option{-fprofile-arcs} plus optimization and code generation options,
2789 generate the arc profile information by running the program on a
2790 selected workload, and then compile the program again with the same
2791 optimization and code generation options plus
2792 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
2793 Control Optimization}).
2795 The other use of @option{-fprofile-arcs} is for use with @code{gcov},
2796 when it is used with the @option{-ftest-coverage} option. GCC
2797 supports two methods of determining code coverage: the options that
2798 support @code{gcov}, and options @option{-a} and @option{-ax}, which
2799 write information to text files. The options that support @code{gcov}
2800 do not need to instrument every arc in the program, so a program compiled
2801 with them runs faster than a program compiled with @option{-a}, which
2802 adds instrumentation code to every basic block in the program. The
2803 tradeoff: since @code{gcov} does not have execution counts for all
2804 branches, it must start with the execution counts for the instrumented
2805 branches, and then iterate over the program flow graph until the entire
2806 graph has been solved. Hence, @code{gcov} runs a little more slowly than
2807 a program which uses information from @option{-a} and @option{-ax}.
2809 With @option{-fprofile-arcs}, for each function of your program GCC
2810 creates a program flow graph, then finds a spanning tree for the graph.
2811 Only arcs that are not on the spanning tree have to be instrumented: the
2812 compiler adds code to count the number of times that these arcs are
2813 executed. When an arc is the only exit or only entrance to a block, the
2814 instrumentation code can be added to the block; otherwise, a new basic
2815 block must be created to hold the instrumentation code.
2817 This option makes it possible to estimate branch probabilities and to
2818 calculate basic block execution counts. In general, basic block
2819 execution counts as provided by @option{-a} do not give enough
2820 information to estimate all branch probabilities.
2823 @item -ftest-coverage
2824 @opindex ftest-coverage
2825 Create data files for the @code{gcov} code-coverage utility
2826 (@pxref{Gcov,, @code{gcov}: a GCC Test Coverage Program}).
2827 The data file names begin with the name of your source file:
2830 @item @var{sourcename}.bb
2831 A mapping from basic blocks to line numbers, which @code{gcov} uses to
2832 associate basic block execution counts with line numbers.
2834 @item @var{sourcename}.bbg
2835 A list of all arcs in the program flow graph. This allows @code{gcov}
2836 to reconstruct the program flow graph, so that it can compute all basic
2837 block and arc execution counts from the information in the
2838 @code{@var{sourcename}.da} file.
2841 Use @option{-ftest-coverage} with @option{-fprofile-arcs}; the latter
2842 option adds instrumentation to the program, which then writes
2843 execution counts to another data file:
2846 @item @var{sourcename}.da
2847 Runtime arc execution counts, used in conjunction with the arc
2848 information in the file @code{@var{sourcename}.bbg}.
2851 Coverage data will map better to the source files if
2852 @option{-ftest-coverage} is used without optimization.
2854 @item -d@var{letters}
2856 Says to make debugging dumps during compilation at times specified by
2857 @var{letters}. This is used for debugging the compiler. The file names
2858 for most of the dumps are made by appending a pass number and a word to
2859 the source file name (e.g. @file{foo.c.00.rtl} or @file{foo.c.01.sibling}).
2860 Here are the possible letters for use in @var{letters}, and their meanings:
2865 Annotate the assembler output with miscellaneous debugging information.
2868 Dump after computing branch probabilities, to @file{@var{file}.14.bp}.
2871 Dump after block reordering, to @file{@var{file}.28.bbro}.
2874 Dump after instruction combination, to the file @file{@var{file}.16.combine}.
2877 Dump after the first if conversion, to the file @file{@var{file}.17.ce}.
2880 Dump after delayed branch scheduling, to @file{@var{file}.31.dbr}.
2883 Dump all macro definitions, at the end of preprocessing, in addition to
2887 Dump after SSA optimizations, to @file{@var{file}.04.ssa} and
2888 @file{@var{file}.07.ussa}.
2891 Dump after the second if conversion, to @file{@var{file}.26.ce2}.
2894 Dump after life analysis, to @file{@var{file}.15.life}.
2897 Dump after purging @code{ADDRESSOF} codes, to @file{@var{file}.09.addressof}.
2900 Dump after global register allocation, to @file{@var{file}.21.greg}.
2903 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
2906 Dump after post-reload optimizations, to @file{@var{file}.22.postreload}.
2909 Dump after GCSE, to @file{@var{file}.10.gcse}.
2912 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
2915 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
2918 Dump after conversion from registers to stack, to @file{@var{file}.32.stack}.
2921 Dump after local register allocation, to @file{@var{file}.20.lreg}.
2924 Dump after loop optimization, to @file{@var{file}.11.loop}.
2927 Dump after performing the machine dependent reorganisation pass, to
2928 @file{@var{file}.30.mach}.
2931 Dump after register renumbering, to @file{@var{file}.25.rnreg}.
2934 Dump after the register move pass, to @file{@var{file}.18.regmove}.
2937 Dump after RTL generation, to @file{@var{file}.00.rtl}.
2940 Dump after the second scheduling pass, to @file{@var{file}.27.sched2}.
2943 Dump after CSE (including the jump optimization that sometimes follows
2944 CSE), to @file{@var{file}.08.cse}.
2947 Dump after the first scheduling pass, to @file{@var{file}.19.sched}.
2950 Dump after the second CSE pass (including the jump optimization that
2951 sometimes follows CSE), to @file{@var{file}.12.cse2}.
2954 Dump after the second flow pass, to @file{@var{file}.23.flow2}.
2957 Dump after SSA dead code elimination, to @file{@var{file}.06.ssadce}.
2960 Dump after the peephole pass, to @file{@var{file}.24.peephole2}.
2963 Produce all the dumps listed above.
2966 Print statistics on memory usage, at the end of the run, to
2970 Annotate the assembler output with a comment indicating which
2971 pattern and alternative was used. The length of each instruction is
2975 Dump the RTL in the assembler output as a comment before each instruction.
2976 Also turns on @option{-dp} annotation.
2979 For each of the other indicated dump files (except for
2980 @file{@var{file}.00.rtl}), dump a representation of the control flow graph
2981 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
2984 Just generate RTL for a function instead of compiling it. Usually used
2988 Dump debugging information during parsing, to standard error.
2991 @item -fdump-unnumbered
2992 @opindex fdump-unnumbered
2993 When doing debugging dumps (see @option{-d} option above), suppress instruction
2994 numbers and line number note output. This makes it more feasible to
2995 use diff on debugging dumps for compiler invocations with different
2996 options, in particular with and without @option{-g}.
2998 @item -fdump-class-hierarchy @r{(C++ only)}
2999 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3000 @opindex fdump-class-hierarchy
3001 Dump a representation of each class's hierarchy and virtual function
3002 table layout to a file. The file name is made by appending @file{.class}
3003 to the source file name. If the @samp{-@var{options}} form is used,
3004 @var{options} controls the details of the dump as described for the
3005 @option{-fdump-tree} options.
3007 @item -fdump-tree-@var{switch} @r{(C++ only)}
3008 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C++ only)}
3010 Control the dumping at various stages of processing the intermediate
3011 language tree to a file. The file name is generated by appending a switch
3012 specific suffix to the source file name. If the @samp{-@var{options}}
3013 form is used, @var{options} is a list of @samp{-} separated options that
3014 control the details of the dump. Not all options are applicable to all
3015 dumps, those which are not meaningful will be ignored. The following
3016 options are available
3020 Print the address of each node. Usually this is not meaningful as it
3021 changes according to the environment and source file. Its primary use
3022 is for tying up a dump file with a debug environment.
3024 Inhibit dumping of members of a scope or body of a function merely
3025 because that scope has been reached. Only dump such items when they
3026 are directly reachable by some other path.
3028 Turn on all options.
3031 The following tree dumps are possible:
3034 Dump before any tree based optimization, to @file{@var{file}.original}.
3036 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3038 Dump after function inlining, to @file{@var{file}.inlined}.
3041 @item -fpretend-float
3042 @opindex fpretend-float
3043 When running a cross-compiler, pretend that the target machine uses the
3044 same floating point format as the host machine. This causes incorrect
3045 output of the actual floating constants, but the actual instruction
3046 sequence will probably be the same as GCC would make when running on
3051 Store the usual ``temporary'' intermediate files permanently; place them
3052 in the current directory and name them based on the source file. Thus,
3053 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3054 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3055 preprocessed @file{foo.i} output file even though the compiler now
3056 normally uses an integrated preprocessor.
3060 Report the CPU time taken by each subprocess in the compilation
3061 sequence. For C source files, this is the compiler proper and assembler
3062 (plus the linker if linking is done). The output looks like this:
3069 The first number on each line is the ``user time,'' that is time spent
3070 executing the program itself. The second number is ``system time,''
3071 time spent executing operating system routines on behalf of the program.
3072 Both numbers are in seconds.
3074 @item -print-file-name=@var{library}
3075 @opindex print-file-name
3076 Print the full absolute name of the library file @var{library} that
3077 would be used when linking---and don't do anything else. With this
3078 option, GCC does not compile or link anything; it just prints the
3081 @item -print-multi-directory
3082 @opindex print-multi-directory
3083 Print the directory name corresponding to the multilib selected by any
3084 other switches present in the command line. This directory is supposed
3085 to exist in @env{GCC_EXEC_PREFIX}.
3087 @item -print-multi-lib
3088 @opindex print-multi-lib
3089 Print the mapping from multilib directory names to compiler switches
3090 that enable them. The directory name is separated from the switches by
3091 @samp{;}, and each switch starts with an @samp{@@} instead of the
3092 @samp{-}, without spaces between multiple switches. This is supposed to
3093 ease shell-processing.
3095 @item -print-prog-name=@var{program}
3096 @opindex print-prog-name
3097 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3099 @item -print-libgcc-file-name
3100 @opindex print-libgcc-file-name
3101 Same as @option{-print-file-name=libgcc.a}.
3103 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3104 but you do want to link with @file{libgcc.a}. You can do
3107 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3110 @item -print-search-dirs
3111 @opindex print-search-dirs
3112 Print the name of the configured installation directory and a list of
3113 program and library directories gcc will search---and don't do anything else.
3115 This is useful when gcc prints the error message
3116 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3117 To resolve this you either need to put @file{cpp0} and the other compiler
3118 components where gcc expects to find them, or you can set the environment
3119 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3120 Don't forget the trailing '/'.
3121 @xref{Environment Variables}.
3124 @opindex dumpmachine
3125 Print the compiler's target machine (for example,
3126 @samp{i686-pc-linux-gnu})---and don't do anything else.
3129 @opindex dumpversion
3130 Print the compiler version (for example, @samp{3.0})---and don't do
3135 Print the compiler's built-in specs---and don't do anything else. (This
3136 is used when GCC itself is being built.) @xref{Spec Files}.
3139 @node Optimize Options
3140 @section Options That Control Optimization
3141 @cindex optimize options
3142 @cindex options, optimization
3144 These options control various sorts of optimizations:
3151 Optimize. Optimizing compilation takes somewhat more time, and a lot
3152 more memory for a large function.
3154 Without @option{-O}, the compiler's goal is to reduce the cost of
3155 compilation and to make debugging produce the expected results.
3156 Statements are independent: if you stop the program with a breakpoint
3157 between statements, you can then assign a new value to any variable or
3158 change the program counter to any other statement in the function and
3159 get exactly the results you would expect from the source code.
3161 With @option{-O}, the compiler tries to reduce code size and execution
3162 time, without performing any optimizations that take a great deal of
3167 Optimize even more. GCC performs nearly all supported optimizations
3168 that do not involve a space-speed tradeoff. The compiler does not
3169 perform loop unrolling or function inlining when you specify @option{-O2}.
3170 As compared to @option{-O}, this option increases both compilation time
3171 and the performance of the generated code.
3173 @option{-O2} turns on all optional optimizations except for loop unrolling,
3174 function inlining, and register renaming. It also turns on the
3175 @option{-fforce-mem} option on all machines and frame pointer elimination
3176 on machines where doing so does not interfere with debugging.
3178 Please note the warning under @option{-fgcse} about
3179 invoking @option{-O2} on programs that use computed gotos.
3183 Optimize yet more. @option{-O3} turns on all optimizations specified by
3184 @option{-O2} and also turns on the @option{-finline-functions} and
3185 @option{-frename-registers} options.
3193 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
3194 do not typically increase code size. It also performs further
3195 optimizations designed to reduce code size.
3197 If you use multiple @option{-O} options, with or without level numbers,
3198 the last such option is the one that is effective.
3201 Options of the form @option{-f@var{flag}} specify machine-independent
3202 flags. Most flags have both positive and negative forms; the negative
3203 form of @option{-ffoo} would be @option{-fno-foo}. In the table below,
3204 only one of the forms is listed---the one which is not the default.
3205 You can figure out the other form by either removing @samp{no-} or
3210 @opindex ffloat-store
3211 Do not store floating point variables in registers, and inhibit other
3212 options that might change whether a floating point value is taken from a
3215 @cindex floating point precision
3216 This option prevents undesirable excess precision on machines such as
3217 the 68000 where the floating registers (of the 68881) keep more
3218 precision than a @code{double} is supposed to have. Similarly for the
3219 x86 architecture. For most programs, the excess precision does only
3220 good, but a few programs rely on the precise definition of IEEE floating
3221 point. Use @option{-ffloat-store} for such programs, after modifying
3222 them to store all pertinent intermediate computations into variables.
3224 @item -fno-default-inline
3225 @opindex fno-default-inline
3226 Do not make member functions inline by default merely because they are
3227 defined inside the class scope (C++ only). Otherwise, when you specify
3228 @w{@option{-O}}, member functions defined inside class scope are compiled
3229 inline by default; i.e., you don't need to add @samp{inline} in front of
3230 the member function name.
3232 @item -fno-defer-pop
3233 @opindex fno-defer-pop
3234 Always pop the arguments to each function call as soon as that function
3235 returns. For machines which must pop arguments after a function call,
3236 the compiler normally lets arguments accumulate on the stack for several
3237 function calls and pops them all at once.
3241 Force memory operands to be copied into registers before doing
3242 arithmetic on them. This produces better code by making all memory
3243 references potential common subexpressions. When they are not common
3244 subexpressions, instruction combination should eliminate the separate
3245 register-load. The @option{-O2} option turns on this option.
3248 @opindex fforce-addr
3249 Force memory address constants to be copied into registers before
3250 doing arithmetic on them. This may produce better code just as
3251 @option{-fforce-mem} may.
3253 @item -fomit-frame-pointer
3254 @opindex fomit-frame-pointer
3255 Don't keep the frame pointer in a register for functions that
3256 don't need one. This avoids the instructions to save, set up and
3257 restore frame pointers; it also makes an extra register available
3258 in many functions. @strong{It also makes debugging impossible on
3262 On some machines, such as the VAX, this flag has no effect, because
3263 the standard calling sequence automatically handles the frame pointer
3264 and nothing is saved by pretending it doesn't exist. The
3265 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3266 whether a target machine supports this flag. @xref{Registers}.
3269 On some machines, such as the VAX, this flag has no effect, because
3270 the standard calling sequence automatically handles the frame pointer
3271 and nothing is saved by pretending it doesn't exist. The
3272 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3273 whether a target machine supports this flag. @xref{Registers,,Register
3274 Usage, gcc.info, Using and Porting GCC}.
3277 @item -foptimize-sibling-calls
3278 @opindex foptimize-sibling-calls
3279 Optimize sibling and tail recursive calls.
3283 This option generates traps for signed overflow on addition, subtraction,
3284 multiplication operations.
3288 Don't pay attention to the @code{inline} keyword. Normally this option
3289 is used to keep the compiler from expanding any functions inline.
3290 Note that if you are not optimizing, no functions can be expanded inline.
3292 @item -finline-functions
3293 @opindex finline-functions
3294 Integrate all simple functions into their callers. The compiler
3295 heuristically decides which functions are simple enough to be worth
3296 integrating in this way.
3298 If all calls to a given function are integrated, and the function is
3299 declared @code{static}, then the function is normally not output as
3300 assembler code in its own right.
3302 @item -finline-limit=@var{n}
3303 @opindex finline-limit
3304 By default, gcc limits the size of functions that can be inlined. This flag
3305 allows the control of this limit for functions that are explicitly marked as
3306 inline (ie marked with the inline keyword or defined within the class
3307 definition in c++). @var{n} is the size of functions that can be inlined in
3308 number of pseudo instructions (not counting parameter handling). The default
3309 value of @var{n} is 600.
3310 Increasing this value can result in more inlined code at
3311 the cost of compilation time and memory consumption. Decreasing usually makes
3312 the compilation faster and less code will be inlined (which presumably
3313 means slower programs). This option is particularly useful for programs that
3314 use inlining heavily such as those based on recursive templates with C++.
3316 @emph{Note:} pseudo instruction represents, in this particular context, an
3317 abstract measurement of function's size. In no way, it represents a count
3318 of assembly instructions and as such its exact meaning might change from one
3319 release to an another.
3321 @item -fkeep-inline-functions
3322 @opindex fkeep-inline-functions
3323 Even if all calls to a given function are integrated, and the function
3324 is declared @code{static}, nevertheless output a separate run-time
3325 callable version of the function. This switch does not affect
3326 @code{extern inline} functions.
3328 @item -fkeep-static-consts
3329 @opindex fkeep-static-consts
3330 Emit variables declared @code{static const} when optimization isn't turned
3331 on, even if the variables aren't referenced.
3333 GCC enables this option by default. If you want to force the compiler to
3334 check if the variable was referenced, regardless of whether or not
3335 optimization is turned on, use the @option{-fno-keep-static-consts} option.
3337 @item -fmerge-constants
3338 Attempt to merge identical constants (string constants and floating point
3339 constants) accross compilation units.
3341 This option is default for optimized compilation if assembler and linker
3342 support it. Use @option{-fno-merge-constants} to inhibit this behaviour.
3344 @item -fmerge-all-constants
3345 Attempt to merge identical constants and identical variables.
3347 This option implies @option{-fmerge-constants}. In addition to
3348 @option{-fmerge-constants} this considers e.g. even constant initialized
3349 arrays or initialized constant variables with integral or floating point
3350 types. Languages like C or C++ require each non-automatic variable to
3351 have distinct location, so using this option will result in non-conforming
3354 @item -fno-function-cse
3355 @opindex fno-function-cse
3356 Do not put function addresses in registers; make each instruction that
3357 calls a constant function contain the function's address explicitly.
3359 This option results in less efficient code, but some strange hacks
3360 that alter the assembler output may be confused by the optimizations
3361 performed when this option is not used.
3365 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, and @*
3366 @option{-fno-trapping-math}.
3368 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
3370 This option should never be turned on by any @option{-O} option since
3371 it can result in incorrect output for programs which depend on
3372 an exact implementation of IEEE or ISO rules/specifications for
3375 @item -fno-math-errno
3376 @opindex fno-math-errno
3377 Do not set ERRNO after calling math functions that are executed
3378 with a single instruction, e.g., sqrt. A program that relies on
3379 IEEE exceptions for math error handling may want to use this flag
3380 for speed while maintaining IEEE arithmetic compatibility.
3382 This option should never be turned on by any @option{-O} option since
3383 it can result in incorrect output for programs which depend on
3384 an exact implementation of IEEE or ISO rules/specifications for
3387 The default is @option{-fmath-errno}.
3389 @item -funsafe-math-optimizations
3390 @opindex funsafe-math-optimizations
3391 Allow optimizations for floating-point arithmetic that (a) assume
3392 that arguments and results are valid and (b) may violate IEEE or
3393 ANSI standards. When used at link-time, it may include libraries
3394 or startup files that change the default FPU control word or other
3395 similar optimizations.
3397 This option should never be turned on by any @option{-O} option since
3398 it can result in incorrect output for programs which depend on
3399 an exact implementation of IEEE or ISO rules/specifications for
3402 The default is @option{-fno-unsafe-math-optimizations}.
3404 @item -fno-trapping-math
3405 @opindex fno-trapping-math
3406 Compile code assuming that floating-point operations cannot generate
3407 user-visible traps. Setting this option may allow faster code
3408 if one relies on ``non-stop'' IEEE arithmetic, for example.
3410 This option should never be turned on by any @option{-O} option since
3411 it can result in incorrect output for programs which depend on
3412 an exact implementation of IEEE or ISO rules/specifications for
3415 The default is @option{-ftrapping-math}.
3418 The following options control specific optimizations. The @option{-O2}
3419 option turns on all of these optimizations except @option{-funroll-loops}
3420 and @option{-funroll-all-loops}. On most machines, the @option{-O} option
3421 turns on the @option{-fthread-jumps} and @option{-fdelayed-branch} options,
3422 but specific machines may handle it differently.
3424 You can use the following flags in the rare cases when ``fine-tuning''
3425 of optimizations to be performed is desired.
3427 Not all of the optimizations performed by GCC have @option{-f} options
3431 @item -fstrength-reduce
3432 @opindex fstrength-reduce
3433 Perform the optimizations of loop strength reduction and
3434 elimination of iteration variables.
3436 @item -fthread-jumps
3437 @opindex fthread-jumps
3438 Perform optimizations where we check to see if a jump branches to a
3439 location where another comparison subsumed by the first is found. If
3440 so, the first branch is redirected to either the destination of the
3441 second branch or a point immediately following it, depending on whether
3442 the condition is known to be true or false.
3444 @item -fcse-follow-jumps
3445 @opindex fcse-follow-jumps
3446 In common subexpression elimination, scan through jump instructions
3447 when the target of the jump is not reached by any other path. For
3448 example, when CSE encounters an @code{if} statement with an
3449 @code{else} clause, CSE will follow the jump when the condition
3452 @item -fcse-skip-blocks
3453 @opindex fcse-skip-blocks
3454 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
3455 follow jumps which conditionally skip over blocks. When CSE
3456 encounters a simple @code{if} statement with no else clause,
3457 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
3458 body of the @code{if}.
3460 @item -frerun-cse-after-loop
3461 @opindex frerun-cse-after-loop
3462 Re-run common subexpression elimination after loop optimizations has been
3465 @item -frerun-loop-opt
3466 @opindex frerun-loop-opt
3467 Run the loop optimizer twice.
3471 Perform a global common subexpression elimination pass.
3472 This pass also performs global constant and copy propagation.
3474 @emph{Note:} When compiling a program using computed gotos, a GCC
3475 extension, you may get better runtime performance if you disable
3476 the global common subexpression elmination pass by adding
3477 @option{-fno-gcse} to the command line.
3481 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
3482 attempt to move loads which are only killed by stores into themselves. This
3483 allows a loop containing a load/store sequence to be changed to a load outside
3484 the loop, and a copy/store within the loop.
3488 When @option{-fgcse-sm} is enabled, A store motion pass is run after global common
3489 subexpression elimination. This pass will attempt to move stores out of loops.
3490 When used in conjunction with @option{-fgcse-lm}, loops containing a load/store sequence
3491 can be changed to a load before the loop and a store after the loop.
3493 @item -fdelete-null-pointer-checks
3494 @opindex fdelete-null-pointer-checks
3495 Use global dataflow analysis to identify and eliminate useless checks
3496 for null pointers. The compiler assumes that dereferencing a null
3497 pointer would have halted the program. If a pointer is checked after
3498 it has already been dereferenced, it cannot be null.
3500 In some environments, this assumption is not true, and programs can
3501 safely dereference null pointers. Use
3502 @option{-fno-delete-null-pointer-checks} to disable this optimization
3503 for programs which depend on that behavior.
3505 @item -fexpensive-optimizations
3506 @opindex fexpensive-optimizations
3507 Perform a number of minor optimizations that are relatively expensive.
3509 @item -foptimize-register-move
3511 @opindex foptimize-register-move
3513 Attempt to reassign register numbers in move instructions and as
3514 operands of other simple instructions in order to maximize the amount of
3515 register tying. This is especially helpful on machines with two-operand
3516 instructions. GCC enables this optimization by default with @option{-O2}
3519 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
3522 @item -fdelayed-branch
3523 @opindex fdelayed-branch
3524 If supported for the target machine, attempt to reorder instructions
3525 to exploit instruction slots available after delayed branch
3528 @item -fschedule-insns
3529 @opindex fschedule-insns
3530 If supported for the target machine, attempt to reorder instructions to
3531 eliminate execution stalls due to required data being unavailable. This
3532 helps machines that have slow floating point or memory load instructions
3533 by allowing other instructions to be issued until the result of the load
3534 or floating point instruction is required.
3536 @item -fschedule-insns2
3537 @opindex fschedule-insns2
3538 Similar to @option{-fschedule-insns}, but requests an additional pass of
3539 instruction scheduling after register allocation has been done. This is
3540 especially useful on machines with a relatively small number of
3541 registers and where memory load instructions take more than one cycle.
3543 @item -ffunction-sections
3544 @itemx -fdata-sections
3545 @opindex ffunction-sections
3546 @opindex fdata-sections
3547 Place each function or data item into its own section in the output
3548 file if the target supports arbitrary sections. The name of the
3549 function or the name of the data item determines the section's name
3552 Use these options on systems where the linker can perform optimizations
3553 to improve locality of reference in the instruction space. HPPA
3554 processors running HP-UX and Sparc processors running Solaris 2 have
3555 linkers with such optimizations. Other systems using the ELF object format
3556 as well as AIX may have these optimizations in the future.
3558 Only use these options when there are significant benefits from doing
3559 so. When you specify these options, the assembler and linker will
3560 create larger object and executable files and will also be slower.
3561 You will not be able to use @code{gprof} on all systems if you
3562 specify this option and you may have problems with debugging if
3563 you specify both this option and @option{-g}.
3565 @item -fcaller-saves
3566 @opindex fcaller-saves
3567 Enable values to be allocated in registers that will be clobbered by
3568 function calls, by emitting extra instructions to save and restore the
3569 registers around such calls. Such allocation is done only when it
3570 seems to result in better code than would otherwise be produced.
3572 This option is always enabled by default on certain machines, usually
3573 those which have no call-preserved registers to use instead.
3575 For all machines, optimization level 2 and higher enables this flag by
3578 @item -funroll-loops
3579 @opindex funroll-loops
3580 Unroll loops whose number of iterations can be determined at compile
3581 time or upon entry to the loop. @option{-funroll-loops} implies both
3582 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
3583 option makes code larger, and may or may not make it run faster.
3585 @item -funroll-all-loops
3586 @opindex funroll-all-loops
3587 Unroll all loops, even if their number of iterations is uncertain when
3588 the loop is entered. This usually makes programs run more slowly.
3589 @option{-funroll-all-loops} implies the same options as
3590 @option{-funroll-loops},
3593 @item -fmove-all-movables
3594 @opindex fmove-all-movables
3595 Forces all invariant computations in loops to be moved
3598 @item -freduce-all-givs
3599 @opindex freduce-all-givs
3600 Forces all general-induction variables in loops to be
3603 @emph{Note:} When compiling programs written in Fortran,
3604 @option{-fmove-all-movables} and @option{-freduce-all-givs} are enabled
3605 by default when you use the optimizer.
3607 These options may generate better or worse code; results are highly
3608 dependent on the structure of loops within the source code.
3610 These two options are intended to be removed someday, once
3611 they have helped determine the efficacy of various
3612 approaches to improving loop optimizations.
3614 Please let us (@w{@email{gcc@@gcc.gnu.org}} and @w{@email{fortran@@gnu.org}})
3615 know how use of these options affects
3616 the performance of your production code.
3617 We're very interested in code that runs @emph{slower}
3618 when these options are @emph{enabled}.
3621 @itemx -fno-peephole2
3622 @opindex fno-peephole
3623 @opindex fno-peephole2
3624 Disable any machine-specific peephole optimizations. The difference
3625 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
3626 are implemented in the compiler; some targets use one, some use the
3627 other, a few use both.
3629 @item -fbranch-probabilities
3630 @opindex fbranch-probabilities
3631 After running a program compiled with @option{-fprofile-arcs}
3632 (@pxref{Debugging Options,, Options for Debugging Your Program or
3633 @command{gcc}}), you can compile it a second time using
3634 @option{-fbranch-probabilities}, to improve optimizations based on
3635 the number of times each branch was taken. When the program
3636 compiled with @option{-fprofile-arcs} exits it saves arc execution
3637 counts to a file called @file{@var{sourcename}.da} for each source
3638 file The information in this data file is very dependent on the
3639 structure of the generated code, so you must use the same source code
3640 and the same optimization options for both compilations.
3643 With @option{-fbranch-probabilities}, GCC puts a @samp{REG_EXEC_COUNT}
3644 note on the first instruction of each basic block, and a
3645 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
3646 These can be used to improve optimization. Currently, they are only
3647 used in one place: in @file{reorg.c}, instead of guessing which path a
3648 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
3649 exactly determine which path is taken more often.
3652 @item -fno-guess-branch-probability
3653 @opindex fno-guess-branch-probability
3654 Do not guess branch probabilities using a randomized model.
3656 Sometimes gcc will opt to use a randomized model to guess branch
3657 probabilities, when none are available from either profiling feedback
3658 (@option{-fprofile-arcs}) or @samp{__builtin_expect}. This means that
3659 different runs of the compiler on the same program may produce different
3662 In a hard real-time system, people don't want different runs of the
3663 compiler to produce code that has different behavior; minimizing
3664 non-determinism is of paramount import. This switch allows users to
3665 reduce non-determinism, possibly at the expense of inferior
3668 @item -fstrict-aliasing
3669 @opindex fstrict-aliasing
3670 Allows the compiler to assume the strictest aliasing rules applicable to
3671 the language being compiled. For C (and C++), this activates
3672 optimizations based on the type of expressions. In particular, an
3673 object of one type is assumed never to reside at the same address as an
3674 object of a different type, unless the types are almost the same. For
3675 example, an @code{unsigned int} can alias an @code{int}, but not a
3676 @code{void*} or a @code{double}. A character type may alias any other
3679 Pay special attention to code like this:
3692 The practice of reading from a different union member than the one most
3693 recently written to (called ``type-punning'') is common. Even with
3694 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
3695 is accessed through the union type. So, the code above will work as
3696 expected. However, this code might not:
3708 Every language that wishes to perform language-specific alias analysis
3709 should define a function that computes, given an @code{tree}
3710 node, an alias set for the node. Nodes in different alias sets are not
3711 allowed to alias. For an example, see the C front-end function
3712 @code{c_get_alias_set}.
3715 @item -falign-functions
3716 @itemx -falign-functions=@var{n}
3717 @opindex falign-functions
3718 Align the start of functions to the next power-of-two greater than
3719 @var{n}, skipping up to @var{n} bytes. For instance,
3720 @option{-falign-functions=32} aligns functions to the next 32-byte
3721 boundary, but @option{-falign-functions=24} would align to the next
3722 32-byte boundary only if this can be done by skipping 23 bytes or less.
3724 @option{-fno-align-functions} and @option{-falign-functions=1} are
3725 equivalent and mean that functions will not be aligned.
3727 Some assemblers only support this flag when @var{n} is a power of two;
3728 in that case, it is rounded up.
3730 If @var{n} is not specified, use a machine-dependent default.
3732 @item -falign-labels
3733 @itemx -falign-labels=@var{n}
3734 @opindex falign-labels
3735 Align all branch targets to a power-of-two boundary, skipping up to
3736 @var{n} bytes like @option{-falign-functions}. This option can easily
3737 make code slower, because it must insert dummy operations for when the
3738 branch target is reached in the usual flow of the code.
3740 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
3741 are greater than this value, then their values are used instead.
3743 If @var{n} is not specified, use a machine-dependent default which is
3744 very likely to be @samp{1}, meaning no alignment.
3747 @itemx -falign-loops=@var{n}
3748 @opindex falign-loops
3749 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
3750 like @option{-falign-functions}. The hope is that the loop will be
3751 executed many times, which will make up for any execution of the dummy
3754 If @var{n} is not specified, use a machine-dependent default.
3757 @itemx -falign-jumps=@var{n}
3758 @opindex falign-jumps
3759 Align branch targets to a power-of-two boundary, for branch targets
3760 where the targets can only be reached by jumping, skipping up to @var{n}
3761 bytes like @option{-falign-functions}. In this case, no dummy operations
3764 If @var{n} is not specified, use a machine-dependent default.
3768 Perform optimizations in static single assignment form. Each function's
3769 flow graph is translated into SSA form, optimizations are performed, and
3770 the flow graph is translated back from SSA form. Users should not
3771 specify this option, since it is not yet ready for production use.
3775 Perform Sparse Conditional Constant Propagation in SSA form. Requires
3776 @option{-fssa}. Like @option{-fssa}, this is an experimental feature.
3780 Perform aggressive dead-code elimination in SSA form. Requires @option{-fssa}.
3781 Like @option{-fssa}, this is an experimental feature.
3783 @item -fsingle-precision-constant
3784 @opindex fsingle-precision-constant
3785 Treat floating point constant as single precision constant instead of
3786 implicitly converting it to double precision constant.
3788 @item -frename-registers
3789 @opindex frename-registers
3790 Attempt to avoid false dependencies in scheduled code by making use
3791 of registers left over after register allocation. This optimization
3792 will most benefit processors with lots of registers. It can, however,
3793 make debugging impossible, since variables will no longer stay in
3794 a ``home register''.
3796 @item --param @var{name}=@var{value}
3798 In some places, GCC uses various constants to control the amount of
3799 optimization that is done. For example, GCC will not inline functions
3800 that contain more that a certain number of instructions. You can
3801 control some of these constants on the command-line using the
3802 @option{--param} option.
3804 In each case, the @var{value} is an integer. The allowable choices for
3805 @var{name} are given in the following table:
3808 @item max-delay-slot-insn-search
3809 The maximum number of instructions to consider when looking for an
3810 instruction to fill a delay slot. If more than this arbitrary number of
3811 instructions is searched, the time savings from filling the delay slot
3812 will be minimal so stop searching. Increasing values mean more
3813 aggressive optimization, making the compile time increase with probably
3814 small improvement in executable run time.
3816 @item max-delay-slot-live-search
3817 When trying to fill delay slots, the maximum number of instructions to
3818 consider when searching for a block with valid live register
3819 information. Increasing this arbitrarily chosen value means more
3820 aggressive optimization, increasing the compile time. This parameter
3821 should be removed when the delay slot code is rewritten to maintain the
3824 @item max-gcse-memory
3825 The approximate maximum amount of memory that will be allocated in
3826 order to perform the global common subexpression elimination
3827 optimization. If more memory than specified is required, the
3828 optimization will not be done.
3830 @item max-gcse-passes
3831 The maximum number of passes of GCSE to run.
3833 @item max-pending-list-length
3834 The maximum number of pending dependancies scheduling will allow
3835 before flushing the current state and starting over. Large functions
3836 with few branches or calls can create excessively large lists which
3837 needlessly consume memory and resources.
3839 @item max-inline-insns
3840 If an function contains more than this many instructions, it
3841 will not be inlined. This option is precisely equivalent to
3842 @option{-finline-limit}.
3847 @node Preprocessor Options
3848 @section Options Controlling the Preprocessor
3849 @cindex preprocessor options
3850 @cindex options, preprocessor
3852 These options control the C preprocessor, which is run on each C source
3853 file before actual compilation.
3855 If you use the @option{-E} option, nothing is done except preprocessing.
3856 Some of these options make sense only together with @option{-E} because
3857 they cause the preprocessor output to be unsuitable for actual
3861 @item -include @var{file}
3863 Process @var{file} as input before processing the regular input file.
3864 In effect, the contents of @var{file} are compiled first. Any @option{-D}
3865 and @option{-U} options on the command line are always processed before
3866 @option{-include @var{file}}, regardless of the order in which they are
3867 written. All the @option{-include} and @option{-imacros} options are
3868 processed in the order in which they are written.
3870 @item -imacros @var{file}
3872 Process @var{file} as input, discarding the resulting output, before
3873 processing the regular input file. Because the output generated from
3874 @var{file} is discarded, the only effect of @option{-imacros @var{file}}
3875 is to make the macros defined in @var{file} available for use in the
3876 main input. All the @option{-include} and @option{-imacros} options are
3877 processed in the order in which they are written.
3879 @item -idirafter @var{dir}
3881 @cindex second include path
3882 Add the directory @var{dir} to the second include path. The directories
3883 on the second include path are searched when a header file is not found
3884 in any of the directories in the main include path (the one that
3885 @option{-I} adds to).
3887 @item -iprefix @var{prefix}
3889 Specify @var{prefix} as the prefix for subsequent @option{-iwithprefix}
3892 @item -iwithprefix @var{dir}
3893 @opindex iwithprefix
3894 Add a directory to the second include path. The directory's name is
3895 made by concatenating @var{prefix} and @var{dir}, where @var{prefix} was
3896 specified previously with @option{-iprefix}. If you have not specified a
3897 prefix yet, the directory containing the installed passes of the
3898 compiler is used as the default.
3900 @item -iwithprefixbefore @var{dir}
3901 @opindex iwithprefixbefore
3902 Add a directory to the main include path. The directory's name is made
3903 by concatenating @var{prefix} and @var{dir}, as in the case of
3904 @option{-iwithprefix}.
3906 @item -isystem @var{dir}
3908 Add a directory to the beginning of the second include path, marking it
3909 as a system directory, so that it gets the same special treatment as
3910 is applied to the standard system directories.
3914 Do not search the standard system directories for header files. Only
3915 the directories you have specified with @option{-I} options (and the
3916 current directory, if appropriate) are searched. @xref{Directory
3917 Options}, for information on @option{-I}.
3919 By using both @option{-nostdinc} and @option{-I-}, you can limit the include-file
3920 search path to only those directories you specify explicitly.
3924 When searching for a header file in a directory, remap file names if a
3925 file named @file{header.gcc} exists in that directory. This can be used
3926 to work around limitations of file systems with file name restrictions.
3927 The @file{header.gcc} file should contain a series of lines with two
3928 tokens on each line: the first token is the name to map, and the second
3929 token is the actual name to use.
3933 Do not predefine any nonstandard macros. (Including architecture flags).
3937 Run only the C preprocessor. Preprocess all the C source files
3938 specified and output the results to standard output or to the
3939 specified output file.
3943 Tell the preprocessor not to discard comments. Used with the
3948 Tell the preprocessor not to generate @samp{#line} directives.
3949 Used with the @option{-E} option.
3952 @cindex dependencies, make
3955 Instead of outputting the result of preprocessing, output a rule
3956 suitable for @code{make} describing the dependencies of the main source
3957 file. The preprocessor outputs one @code{make} rule containing the
3958 object file name for that source file, a colon, and the names of all the
3959 included files. Unless overridden explicitly, the object file name
3960 consists of the basename of the source file with any suffix replaced with
3961 object file suffix. If there are many included files then the
3962 rule is split into several lines using @samp{\}-newline.
3964 @option{-M} implies @option{-E}.
3968 Like @option{-M}, but mention only the files included with @samp{#include
3969 "@var{file}"}. System header files included with @samp{#include
3970 <@var{file}>} are omitted.
3974 Like @option{-M} but the dependency information is written to a file
3975 rather than stdout. @code{gcc} will use the same file name and
3976 directory as the object file, but with the suffix @file{.d} instead.
3978 This is in addition to compiling the main file as specified---@option{-MD}
3979 does not inhibit ordinary compilation the way @option{-M} does,
3980 unless you also specify @option{-MG}.
3982 With Mach, you can use the utility @code{md} to merge multiple
3983 dependency files into a single dependency file suitable for using with
3984 the @samp{make} command.
3988 Like @option{-MD} except mention only user header files, not system
3991 @item -MF @var{file}
3993 When used with @option{-M} or @option{-MM}, specifies a file to write the
3994 dependencies to. This allows the preprocessor to write the preprocessed
3995 file to stdout normally. If no @option{-MF} switch is given, CPP sends
3996 the rules to stdout and suppresses normal preprocessed output.
3998 Another way to specify output of a @code{make} rule is by setting
3999 the environment variable @env{DEPENDENCIES_OUTPUT} (@pxref{Environment
4004 When used with @option{-M} or @option{-MM}, @option{-MG} says to treat missing
4005 header files as generated files and assume they live in the same
4006 directory as the source file. It suppresses preprocessed output, as a
4007 missing header file is ordinarily an error.
4009 This feature is used in automatic updating of makefiles.
4013 This option instructs CPP to add a phony target for each dependency
4014 other than the main file, causing each to depend on nothing. These
4015 dummy rules work around errors @code{make} gives if you remove header
4016 files without updating the @code{Makefile} to match.
4018 This is typical output:-
4021 /tmp/test.o: /tmp/test.c /tmp/test.h
4026 @item -MQ @var{target}
4027 @item -MT @var{target}
4030 By default CPP uses the main file name, including any path, and appends
4031 the object suffix, normally ``.o'', to it to obtain the name of the
4032 target for dependency generation. With @option{-MT} you can specify a
4033 target yourself, overriding the default one.
4035 If you want multiple targets, you can specify them as a single argument
4036 to @option{-MT}, or use multiple @option{-MT} options.
4038 The targets you specify are output in the order they appear on the
4039 command line. @option{-MQ} is identical to @option{-MT}, except that the
4040 target name is quoted for Make, but with @option{-MT} it isn't. For
4041 example, @option{-MT '$(objpfx)foo.o'} gives
4044 $(objpfx)foo.o: /tmp/foo.c
4047 but @option{-MQ '$(objpfx)foo.o'} gives
4050 $$(objpfx)foo.o: /tmp/foo.c
4053 The default target is automatically quoted, as if it were given with
4058 Print the name of each header file used, in addition to other normal
4061 @item -A@var{question}(@var{answer})
4063 Assert the answer @var{answer} for @var{question}, in case it is tested
4064 with a preprocessing conditional such as @samp{#if
4065 #@var{question}(@var{answer})}. @option{-A-} disables the standard
4066 assertions that normally describe the target machine.
4070 Define macro @var{macro} with the string @samp{1} as its definition.
4072 @item -D@var{macro}=@var{defn}
4073 Define macro @var{macro} as @var{defn}. All instances of @option{-D} on
4074 the command line are processed before any @option{-U} options.
4076 Any @option{-D} and @option{-U} options on the command line are processed in
4077 order, and always before @option{-imacros @var{file}}, regardless of the
4078 order in which they are written.
4082 Undefine macro @var{macro}. @option{-U} options are evaluated after all
4083 @option{-D} options, but before any @option{-include} and @option{-imacros}
4086 Any @option{-D} and @option{-U} options on the command line are processed in
4087 order, and always before @option{-imacros @var{file}}, regardless of the
4088 order in which they are written.
4092 Tell the preprocessor to output only a list of the macro definitions
4093 that are in effect at the end of preprocessing. Used with the @option{-E}
4098 Tell the preprocessing to pass all macro definitions into the output, in
4099 their proper sequence in the rest of the output.
4103 Like @option{-dD} except that the macro arguments and contents are omitted.
4104 Only @samp{#define @var{name}} is included in the output.
4108 Output @samp{#include} directives in addition to the result of
4111 @item -fpreprocessed
4112 @opindex fpreprocessed
4113 Indicate to the preprocessor that the input file has already been
4114 preprocessed. This suppresses things like macro expansion, trigraph
4115 conversion, escaped newline splicing, and processing of most directives.
4116 The preprocessor still recognizes and removes comments, so that you can
4117 pass a file preprocessed with @option{-C} to the compiler without
4118 problems. In this mode the integrated preprocessor is little more than
4119 a tokenizer for the front ends.
4121 @option{-fpreprocessed} is implicit if the input file has one of the
4122 extensions @samp{i}, @samp{ii} or @samp{mi}. These are the extensions
4123 that GCC uses for preprocessed files created by @option{-save-temps}.
4127 Process ISO standard trigraph sequences. These are three-character
4128 sequences, all starting with @samp{??}, that are defined by ISO C to
4129 stand for single characters. For example, @samp{??/} stands for
4130 @samp{\}, so @samp{'??/n'} is a character constant for a newline. By
4131 default, GCC ignores trigraphs, but in standard-conforming modes it
4132 converts them. See the @option{-std} and @option{-ansi} options.
4134 The nine trigraph sequences are
4137 @expansion{} @samp{[}
4140 @expansion{} @samp{]}
4143 @expansion{} @samp{@{}
4146 @expansion{} @samp{@}}
4149 @expansion{} @samp{#}
4152 @expansion{} @samp{\}
4155 @expansion{} @samp{^}
4158 @expansion{} @samp{|}
4161 @expansion{} @samp{~}
4165 Trigraph support is not popular, so many compilers do not implement it
4166 properly. Portable code should not rely on trigraphs being either
4167 converted or ignored.
4169 @item -Wp,@var{option}
4171 Pass @var{option} as an option to the preprocessor. If @var{option}
4172 contains commas, it is split into multiple options at the commas.
4175 @node Assembler Options
4176 @section Passing Options to the Assembler
4178 @c prevent bad page break with this line
4179 You can pass options to the assembler.
4182 @item -Wa,@var{option}
4184 Pass @var{option} as an option to the assembler. If @var{option}
4185 contains commas, it is split into multiple options at the commas.
4189 @section Options for Linking
4190 @cindex link options
4191 @cindex options, linking
4193 These options come into play when the compiler links object files into
4194 an executable output file. They are meaningless if the compiler is
4195 not doing a link step.
4199 @item @var{object-file-name}
4200 A file name that does not end in a special recognized suffix is
4201 considered to name an object file or library. (Object files are
4202 distinguished from libraries by the linker according to the file
4203 contents.) If linking is done, these object files are used as input
4212 If any of these options is used, then the linker is not run, and
4213 object file names should not be used as arguments. @xref{Overall
4217 @item -l@var{library}
4218 @itemx -l @var{library}
4220 Search the library named @var{library} when linking. (The second
4221 alternative with the library as a separate argument is only for
4222 POSIX compliance and is not recommended.)
4224 It makes a difference where in the command you write this option; the
4225 linker searches and processes libraries and object files in the order they
4226 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
4227 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
4228 to functions in @samp{z}, those functions may not be loaded.
4230 The linker searches a standard list of directories for the library,
4231 which is actually a file named @file{lib@var{library}.a}. The linker
4232 then uses this file as if it had been specified precisely by name.
4234 The directories searched include several standard system directories
4235 plus any that you specify with @option{-L}.
4237 Normally the files found this way are library files---archive files
4238 whose members are object files. The linker handles an archive file by
4239 scanning through it for members which define symbols that have so far
4240 been referenced but not defined. But if the file that is found is an
4241 ordinary object file, it is linked in the usual fashion. The only
4242 difference between using an @option{-l} option and specifying a file name
4243 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
4244 and searches several directories.
4248 You need this special case of the @option{-l} option in order to
4249 link an Objective-C program.
4252 @opindex nostartfiles
4253 Do not use the standard system startup files when linking.
4254 The standard system libraries are used normally, unless @option{-nostdlib}
4255 or @option{-nodefaultlibs} is used.
4257 @item -nodefaultlibs
4258 @opindex nodefaultlibs
4259 Do not use the standard system libraries when linking.
4260 Only the libraries you specify will be passed to the linker.
4261 The standard startup files are used normally, unless @option{-nostartfiles}
4262 is used. The compiler may generate calls to memcmp, memset, and memcpy
4263 for System V (and ISO C) environments or to bcopy and bzero for
4264 BSD environments. These entries are usually resolved by entries in
4265 libc. These entry points should be supplied through some other
4266 mechanism when this option is specified.
4270 Do not use the standard system startup files or libraries when linking.
4271 No startup files and only the libraries you specify will be passed to
4272 the linker. The compiler may generate calls to memcmp, memset, and memcpy
4273 for System V (and ISO C) environments or to bcopy and bzero for
4274 BSD environments. These entries are usually resolved by entries in
4275 libc. These entry points should be supplied through some other
4276 mechanism when this option is specified.
4278 @cindex @option{-lgcc}, use with @option{-nostdlib}
4279 @cindex @option{-nostdlib} and unresolved references
4280 @cindex unresolved references and @option{-nostdlib}
4281 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
4282 @cindex @option{-nodefaultlibs} and unresolved references
4283 @cindex unresolved references and @option{-nodefaultlibs}
4284 One of the standard libraries bypassed by @option{-nostdlib} and
4285 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
4286 that GCC uses to overcome shortcomings of particular machines, or special
4287 needs for some languages.
4289 (@xref{Interface,,Interfacing to GCC Output}, for more discussion of
4293 (@xref{Interface,,Interfacing to GCC Output,gcc.info,Porting GCC},
4294 for more discussion of @file{libgcc.a}.)
4296 In most cases, you need @file{libgcc.a} even when you want to avoid
4297 other standard libraries. In other words, when you specify @option{-nostdlib}
4298 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
4299 This ensures that you have no unresolved references to internal GCC
4300 library subroutines. (For example, @samp{__main}, used to ensure C++
4301 constructors will be called; @pxref{Collect2,,@command{collect2}}.)
4305 Remove all symbol table and relocation information from the executable.
4309 On systems that support dynamic linking, this prevents linking with the shared
4310 libraries. On other systems, this option has no effect.
4314 Produce a shared object which can then be linked with other objects to
4315 form an executable. Not all systems support this option. For predictable
4316 results, you must also specify the same set of options that were used to
4317 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
4318 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
4319 needs to build supplementary stub code for constructors to work. On
4320 multi-libbed systems, @samp{gcc -shared} must select the correct support
4321 libraries to link against. Failing to supply the correct flags may lead
4322 to subtle defects. Supplying them in cases where they are not necessary
4325 @item -shared-libgcc
4326 @itemx -static-libgcc
4327 @opindex shared-libgcc
4328 @opindex static-libgcc
4329 On systems that provide @file{libgcc} as a shared library, these options
4330 force the use of either the shared or static version respectively.
4331 If no shared version of @file{libgcc} was built when the compiler was
4332 configured, these options have no effect.
4334 There are several situations in which an application should use the
4335 shared @file{libgcc} instead of the static version. The most common
4336 of these is when the application wishes to throw and catch exceptions
4337 across different shared libraries. In that case, each of the libraries
4338 as well as the application itself should use the shared @file{libgcc}.
4340 Therefore, whenever you specify the @option{-shared} option, the GCC
4341 driver automatically adds @option{-shared-libgcc}, unless you explicitly
4342 specify @option{-static-libgcc}. The G++ driver automatically adds
4343 @option{-shared-libgcc} when you build a main executable as well because
4344 for C++ programs that is typically the right thing to do.
4345 (Exception-handling will not work reliably otherwise.)
4347 However, when linking a main executable written in C, you must
4348 explicitly say @option{-shared-libgcc} if you want to use the shared
4353 Bind references to global symbols when building a shared object. Warn
4354 about any unresolved references (unless overridden by the link editor
4355 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
4358 @item -Xlinker @var{option}
4360 Pass @var{option} as an option to the linker. You can use this to
4361 supply system-specific linker options which GCC does not know how to
4364 If you want to pass an option that takes an argument, you must use
4365 @option{-Xlinker} twice, once for the option and once for the argument.
4366 For example, to pass @option{-assert definitions}, you must write
4367 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
4368 @option{-Xlinker "-assert definitions"}, because this passes the entire
4369 string as a single argument, which is not what the linker expects.
4371 @item -Wl,@var{option}
4373 Pass @var{option} as an option to the linker. If @var{option} contains
4374 commas, it is split into multiple options at the commas.
4376 @item -u @var{symbol}
4378 Pretend the symbol @var{symbol} is undefined, to force linking of
4379 library modules to define it. You can use @option{-u} multiple times with
4380 different symbols to force loading of additional library modules.
4383 @node Directory Options
4384 @section Options for Directory Search
4385 @cindex directory options
4386 @cindex options, directory search
4389 These options specify directories to search for header files, for
4390 libraries and for parts of the compiler:
4395 Add the directory @var{dir} to the head of the list of directories to be
4396 searched for header files. This can be used to override a system header
4397 file, substituting your own version, since these directories are
4398 searched before the system header file directories. However, you should
4399 not use this option to add directories that contain vendor-supplied
4400 system header files (use @option{-isystem} for that). If you use more than
4401 one @option{-I} option, the directories are scanned in left-to-right
4402 order; the standard system directories come after.
4404 If a standard system include directory, or a directory specified with
4405 @option{-isystem}, is also specified with @option{-I}, it will be
4406 searched only in the position requested by @option{-I}. Also, it will
4407 not be considered a system include directory. If that directory really
4408 does contain system headers, there is a good chance that they will
4409 break. For instance, if GCC's installation procedure edited the headers
4410 in @file{/usr/include} to fix bugs, @samp{-I/usr/include} will cause the
4411 original, buggy headers to be found instead of the corrected ones. GCC
4412 will issue a warning when a system include directory is hidden in this
4417 Any directories you specify with @option{-I} options before the @option{-I-}
4418 option are searched only for the case of @samp{#include "@var{file}"};
4419 they are not searched for @samp{#include <@var{file}>}.
4421 If additional directories are specified with @option{-I} options after
4422 the @option{-I-}, these directories are searched for all @samp{#include}
4423 directives. (Ordinarily @emph{all} @option{-I} directories are used
4426 In addition, the @option{-I-} option inhibits the use of the current
4427 directory (where the current input file came from) as the first search
4428 directory for @samp{#include "@var{file}"}. There is no way to
4429 override this effect of @option{-I-}. With @option{-I.} you can specify
4430 searching the directory which was current when the compiler was
4431 invoked. That is not exactly the same as what the preprocessor does
4432 by default, but it is often satisfactory.
4434 @option{-I-} does not inhibit the use of the standard system directories
4435 for header files. Thus, @option{-I-} and @option{-nostdinc} are
4440 Add directory @var{dir} to the list of directories to be searched
4443 @item -B@var{prefix}
4445 This option specifies where to find the executables, libraries,
4446 include files, and data files of the compiler itself.
4448 The compiler driver program runs one or more of the subprograms
4449 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
4450 @var{prefix} as a prefix for each program it tries to run, both with and
4451 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
4453 For each subprogram to be run, the compiler driver first tries the
4454 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
4455 was not specified, the driver tries two standard prefixes, which are
4456 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc-lib/}. If neither of
4457 those results in a file name that is found, the unmodified program
4458 name is searched for using the directories specified in your
4459 @env{PATH} environment variable.
4461 The compiler will check to see if the path provided by the @option{-B}
4462 refers to a directory, and if necessary it will add a directory
4463 separator character at the end of the path.
4465 @option{-B} prefixes that effectively specify directory names also apply
4466 to libraries in the linker, because the compiler translates these
4467 options into @option{-L} options for the linker. They also apply to
4468 includes files in the preprocessor, because the compiler translates these
4469 options into @option{-isystem} options for the preprocessor. In this case,
4470 the compiler appends @samp{include} to the prefix.
4472 The run-time support file @file{libgcc.a} can also be searched for using
4473 the @option{-B} prefix, if needed. If it is not found there, the two
4474 standard prefixes above are tried, and that is all. The file is left
4475 out of the link if it is not found by those means.
4477 Another way to specify a prefix much like the @option{-B} prefix is to use
4478 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
4481 As a special kludge, if the path provided by @option{-B} is
4482 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
4483 9, then it will be replaced by @file{[dir/]include}. This is to help
4484 with boot-strapping the compiler.
4486 @item -specs=@var{file}
4488 Process @var{file} after the compiler reads in the standard @file{specs}
4489 file, in order to override the defaults that the @file{gcc} driver
4490 program uses when determining what switches to pass to @file{cc1},
4491 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
4492 @option{-specs=@var{file}} can be specified on the command line, and they
4493 are processed in order, from left to right.
4499 @section Specifying subprocesses and the switches to pass to them
4501 @command{gcc} is a driver program. It performs its job by invoking a
4502 sequence of other programs to do the work of compiling, assembling and
4503 linking. GCC interprets its command-line parameters and uses these to
4504 deduce which programs it should invoke, and which command-line options
4505 it ought to place on their command lines. This behaviour is controlled
4506 by @dfn{spec strings}. In most cases there is one spec string for each
4507 program that GCC can invoke, but a few programs have multiple spec
4508 strings to control their behaviour. The spec strings built into GCC can
4509 be overridden by using the @option{-specs=} command-line switch to specify
4512 @dfn{Spec files} are plaintext files that are used to construct spec
4513 strings. They consist of a sequence of directives separated by blank
4514 lines. The type of directive is determined by the first non-whitespace
4515 character on the line and it can be one of the following:
4518 @item %@var{command}
4519 Issues a @var{command} to the spec file processor. The commands that can
4523 @item %include <@var{file}>
4525 Search for @var{file} and insert its text at the current point in the
4528 @item %include_noerr <@var{file}>
4529 @cindex %include_noerr
4530 Just like @samp{%include}, but do not generate an error message if the include
4531 file cannot be found.
4533 @item %rename @var{old_name} @var{new_name}
4535 Rename the spec string @var{old_name} to @var{new_name}.
4539 @item *[@var{spec_name}]:
4540 This tells the compiler to create, override or delete the named spec
4541 string. All lines after this directive up to the next directive or
4542 blank line are considered to be the text for the spec string. If this
4543 results in an empty string then the spec will be deleted. (Or, if the
4544 spec did not exist, then nothing will happened.) Otherwise, if the spec
4545 does not currently exist a new spec will be created. If the spec does
4546 exist then its contents will be overridden by the text of this
4547 directive, unless the first character of that text is the @samp{+}
4548 character, in which case the text will be appended to the spec.
4550 @item [@var{suffix}]:
4551 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
4552 and up to the next directive or blank line are considered to make up the
4553 spec string for the indicated suffix. When the compiler encounters an
4554 input file with the named suffix, it will processes the spec string in
4555 order to work out how to compile that file. For example:
4562 This says that any input file whose name ends in @samp{.ZZ} should be
4563 passed to the program @samp{z-compile}, which should be invoked with the
4564 command-line switch @option{-input} and with the result of performing the
4565 @samp{%i} substitution. (See below.)
4567 As an alternative to providing a spec string, the text that follows a
4568 suffix directive can be one of the following:
4571 @item @@@var{language}
4572 This says that the suffix is an alias for a known @var{language}. This is
4573 similar to using the @option{-x} command-line switch to GCC to specify a
4574 language explicitly. For example:
4581 Says that .ZZ files are, in fact, C++ source files.
4584 This causes an error messages saying:
4587 @var{name} compiler not installed on this system.
4591 GCC already has an extensive list of suffixes built into it.
4592 This directive will add an entry to the end of the list of suffixes, but
4593 since the list is searched from the end backwards, it is effectively
4594 possible to override earlier entries using this technique.
4598 GCC has the following spec strings built into it. Spec files can
4599 override these strings or create their own. Note that individual
4600 targets can also add their own spec strings to this list.
4603 asm Options to pass to the assembler
4604 asm_final Options to pass to the assembler post-processor
4605 cpp Options to pass to the C preprocessor
4606 cc1 Options to pass to the C compiler
4607 cc1plus Options to pass to the C++ compiler
4608 endfile Object files to include at the end of the link
4609 link Options to pass to the linker
4610 lib Libraries to include on the command line to the linker
4611 libgcc Decides which GCC support library to pass to the linker
4612 linker Sets the name of the linker
4613 predefines Defines to be passed to the C preprocessor
4614 signed_char Defines to pass to CPP to say whether @code{char} is signed
4616 startfile Object files to include at the start of the link
4619 Here is a small example of a spec file:
4625 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
4628 This example renames the spec called @samp{lib} to @samp{old_lib} and
4629 then overrides the previous definition of @samp{lib} with a new one.
4630 The new definition adds in some extra command-line options before
4631 including the text of the old definition.
4633 @dfn{Spec strings} are a list of command-line options to be passed to their
4634 corresponding program. In addition, the spec strings can contain
4635 @samp{%}-prefixed sequences to substitute variable text or to
4636 conditionally insert text into the command line. Using these constructs
4637 it is possible to generate quite complex command lines.
4639 Here is a table of all defined @samp{%}-sequences for spec
4640 strings. Note that spaces are not generated automatically around the
4641 results of expanding these sequences. Therefore you can concatenate them
4642 together or combine them with constant text in a single argument.
4646 Substitute one @samp{%} into the program name or argument.
4649 Substitute the name of the input file being processed.
4652 Substitute the basename of the input file being processed.
4653 This is the substring up to (and not including) the last period
4654 and not including the directory.
4657 This is the same as @samp{%b}, but include the file suffix (text after
4661 Marks the argument containing or following the @samp{%d} as a
4662 temporary file name, so that that file will be deleted if GCC exits
4663 successfully. Unlike @samp{%g}, this contributes no text to the
4666 @item %g@var{suffix}
4667 Substitute a file name that has suffix @var{suffix} and is chosen
4668 once per compilation, and mark the argument in the same way as
4669 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
4670 name is now chosen in a way that is hard to predict even when previously
4671 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
4672 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
4673 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
4674 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
4675 was simply substituted with a file name chosen once per compilation,
4676 without regard to any appended suffix (which was therefore treated
4677 just like ordinary text), making such attacks more likely to succeed.
4679 @item %u@var{suffix}
4680 Like @samp{%g}, but generates a new temporary file name even if
4681 @samp{%u@var{suffix}} was already seen.
4683 @item %U@var{suffix}
4684 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
4685 new one if there is no such last file name. In the absence of any
4686 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
4687 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
4688 would involve the generation of two distinct file names, one
4689 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
4690 simply substituted with a file name chosen for the previous @samp{%u},
4691 without regard to any appended suffix.
4693 @item %j@var{SUFFIX}
4694 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
4695 writable, and if save-temps is off; otherwise, substitute the name
4696 of a temporary file, just like @samp{%u}. This temporary file is not
4697 meant for communication between processes, but rather as a junk
4700 @item %.@var{SUFFIX}
4701 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
4702 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
4703 terminated by the next space or %.
4706 Marks the argument containing or following the @samp{%w} as the
4707 designated output file of this compilation. This puts the argument
4708 into the sequence of arguments that @samp{%o} will substitute later.
4711 Substitutes the names of all the output files, with spaces
4712 automatically placed around them. You should write spaces
4713 around the @samp{%o} as well or the results are undefined.
4714 @samp{%o} is for use in the specs for running the linker.
4715 Input files whose names have no recognized suffix are not compiled
4716 at all, but they are included among the output files, so they will
4720 Substitutes the suffix for object files. Note that this is
4721 handled specially when it immediately follows @samp{%g, %u, or %U},
4722 because of the need for those to form complete file names. The
4723 handling is such that @samp{%O} is treated exactly as if it had already
4724 been substituted, except that @samp{%g, %u, and %U} do not currently
4725 support additional @var{suffix} characters following @samp{%O} as they would
4726 following, for example, @samp{.o}.
4729 Substitutes the standard macro predefinitions for the
4730 current target machine. Use this when running @code{cpp}.
4733 Like @samp{%p}, but puts @samp{__} before and after the name of each
4734 predefined macro, except for macros that start with @samp{__} or with
4735 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
4739 Substitute a @option{-iprefix} option made from @env{GCC_EXEC_PREFIX}.
4742 Current argument is the name of a library or startup file of some sort.
4743 Search for that file in a standard list of directories and substitute
4744 the full name found.
4747 Print @var{str} as an error message. @var{str} is terminated by a newline.
4748 Use this when inconsistent options are detected.
4751 Output @samp{-} if the input for the current command is coming from a pipe.
4754 Substitute the contents of spec string @var{name} at this point.
4757 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
4759 @item %x@{@var{option}@}
4760 Accumulate an option for @samp{%X}.
4763 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
4767 Output the accumulated assembler options specified by @option{-Wa}.
4770 Output the accumulated preprocessor options specified by @option{-Wp}.
4773 Substitute the major version number of GCC@.
4774 (For version 2.9.5, this is 2.)
4777 Substitute the minor version number of GCC@.
4778 (For version 2.9.5, this is 9.)
4781 Substitute the patch level number of GCC@.
4782 (For version 2.9.5, this is 5.)
4785 Process the @code{asm} spec. This is used to compute the
4786 switches to be passed to the assembler.
4789 Process the @code{asm_final} spec. This is a spec string for
4790 passing switches to an assembler post-processor, if such a program is
4794 Process the @code{link} spec. This is the spec for computing the
4795 command line passed to the linker. Typically it will make use of the
4796 @samp{%L %G %S %D and %E} sequences.
4799 Dump out a @option{-L} option for each directory that GCC believes might
4800 contain startup files. If the target supports multilibs then the
4801 current multilib directory will be prepended to each of these paths.
4804 Output the multilib directory with directory separators replaced with
4805 @samp{_}. If multilib directories are not set, or the multilib directory is
4806 @file{.} then this option emits nothing.
4809 Process the @code{lib} spec. This is a spec string for deciding which
4810 libraries should be included on the command line to the linker.
4813 Process the @code{libgcc} spec. This is a spec string for deciding
4814 which GCC support library should be included on the command line to the linker.
4817 Process the @code{startfile} spec. This is a spec for deciding which
4818 object files should be the first ones passed to the linker. Typically
4819 this might be a file named @file{crt0.o}.
4822 Process the @code{endfile} spec. This is a spec string that specifies
4823 the last object files that will be passed to the linker.
4826 Process the @code{cpp} spec. This is used to construct the arguments
4827 to be passed to the C preprocessor.
4830 Process the @code{signed_char} spec. This is intended to be used
4831 to tell cpp whether a char is signed. It typically has the definition:
4833 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
4837 Process the @code{cc1} spec. This is used to construct the options to be
4838 passed to the actual C compiler (@samp{cc1}).
4841 Process the @code{cc1plus} spec. This is used to construct the options to be
4842 passed to the actual C++ compiler (@samp{cc1plus}).
4845 Substitute the variable part of a matched option. See below.
4846 Note that each comma in the substituted string is replaced by
4850 Substitutes the @code{-S} switch, if that switch was given to GCC@.
4851 If that switch was not specified, this substitutes nothing. Note that
4852 the leading dash is omitted when specifying this option, and it is
4853 automatically inserted if the substitution is performed. Thus the spec
4854 string @samp{%@{foo@}} would match the command-line option @option{-foo}
4855 and would output the command line option @option{-foo}.
4857 @item %W@{@code{S}@}
4858 Like %@{@code{S}@} but mark last argument supplied within as a file to be
4861 @item %@{@code{S}*@}
4862 Substitutes all the switches specified to GCC whose names start
4863 with @code{-S}, but which also take an argument. This is used for
4864 switches like @option{-o}, @option{-D}, @option{-I}, etc.
4865 GCC considers @option{-o foo} as being
4866 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
4867 text, including the space. Thus two arguments would be generated.
4869 @item %@{^@code{S}*@}
4870 Like %@{@code{S}*@}, but don't put a blank between a switch and its
4871 argument. Thus %@{^o*@} would only generate one argument, not two.
4873 @item %@{@code{S}*&@code{T}*@}
4874 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
4875 (the order of @code{S} and @code{T} in the spec is not significant).
4876 There can be any number of ampersand-separated variables; for each the
4877 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
4879 @item %@{<@code{S}@}
4880 Remove all occurrences of @code{-S} from the command line. Note---this
4881 command is position dependent. @samp{%} commands in the spec string
4882 before this option will see @code{-S}, @samp{%} commands in the spec
4883 string after this option will not.
4885 @item %@{@code{S}*:@code{X}@}
4886 Substitutes @code{X} if one or more switches whose names start with
4887 @code{-S} are specified to GCC@. Note that the tail part of the
4888 @code{-S} option (i.e.@: the part matched by the @samp{*}) will be substituted
4889 for each occurrence of @samp{%*} within @code{X}.
4891 @item %@{@code{S}:@code{X}@}
4892 Substitutes @code{X}, but only if the @samp{-S} switch was given to GCC@.
4894 @item %@{!@code{S}:@code{X}@}
4895 Substitutes @code{X}, but only if the @samp{-S} switch was @emph{not} given to GCC@.
4897 @item %@{|@code{S}:@code{X}@}
4898 Like %@{@code{S}:@code{X}@}, but if no @code{S} switch, substitute @samp{-}.
4900 @item %@{|!@code{S}:@code{X}@}
4901 Like %@{!@code{S}:@code{X}@}, but if there is an @code{S} switch, substitute @samp{-}.
4903 @item %@{.@code{S}:@code{X}@}
4904 Substitutes @code{X}, but only if processing a file with suffix @code{S}.
4906 @item %@{!.@code{S}:@code{X}@}
4907 Substitutes @code{X}, but only if @emph{not} processing a file with suffix @code{S}.
4909 @item %@{@code{S}|@code{P}:@code{X}@}
4910 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@. This may be
4911 combined with @samp{!} and @samp{.} sequences as well, although they
4912 have a stronger binding than the @samp{|}. For example a spec string
4916 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
4919 will output the following command-line options from the following input
4920 command-line options:
4925 -d fred.c -foo -baz -boggle
4926 -d jim.d -bar -baz -boggle
4931 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or
4932 %@{!@code{S}:@code{X}@} construct may contain other nested @samp{%} constructs
4933 or spaces, or even newlines. They are processed as usual, as described
4936 The @option{-O}, @option{-f}, @option{-m}, and @option{-W}
4937 switches are handled specifically in these
4938 constructs. If another value of @option{-O} or the negated form of a @option{-f}, @option{-m}, or
4939 @option{-W} switch is found later in the command line, the earlier switch
4940 value is ignored, except with @{@code{S}*@} where @code{S} is just one
4941 letter, which passes all matching options.
4943 The character @samp{|} at the beginning of the predicate text is used to indicate
4944 that a command should be piped to the following command, but only if @option{-pipe}
4947 It is built into GCC which switches take arguments and which do not.
4948 (You might think it would be useful to generalize this to allow each
4949 compiler's spec to say which switches take arguments. But this cannot
4950 be done in a consistent fashion. GCC cannot even decide which input
4951 files have been specified without knowing which switches take arguments,
4952 and it must know which input files to compile in order to tell which
4955 GCC also knows implicitly that arguments starting in @option{-l} are to be
4956 treated as compiler output files, and passed to the linker in their
4957 proper position among the other output files.
4959 @c man begin OPTIONS
4961 @node Target Options
4962 @section Specifying Target Machine and Compiler Version
4963 @cindex target options
4964 @cindex cross compiling
4965 @cindex specifying machine version
4966 @cindex specifying compiler version and target machine
4967 @cindex compiler version, specifying
4968 @cindex target machine, specifying
4970 By default, GCC compiles code for the same type of machine that you
4971 are using. However, it can also be installed as a cross-compiler, to
4972 compile for some other type of machine. In fact, several different
4973 configurations of GCC, for different target machines, can be
4974 installed side by side. Then you specify which one to use with the
4977 In addition, older and newer versions of GCC can be installed side
4978 by side. One of them (probably the newest) will be the default, but
4979 you may sometimes wish to use another.
4982 @item -b @var{machine}
4984 The argument @var{machine} specifies the target machine for compilation.
4985 This is useful when you have installed GCC as a cross-compiler.
4987 The value to use for @var{machine} is the same as was specified as the
4988 machine type when configuring GCC as a cross-compiler. For
4989 example, if a cross-compiler was configured with @samp{configure
4990 i386v}, meaning to compile for an 80386 running System V, then you
4991 would specify @option{-b i386v} to run that cross compiler.
4993 When you do not specify @option{-b}, it normally means to compile for
4994 the same type of machine that you are using.
4996 @item -V @var{version}
4998 The argument @var{version} specifies which version of GCC to run.
4999 This is useful when multiple versions are installed. For example,
5000 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
5002 The default version, when you do not specify @option{-V}, is the last
5003 version of GCC that you installed.
5006 The @option{-b} and @option{-V} options actually work by controlling part of
5007 the file name used for the executable files and libraries used for
5008 compilation. A given version of GCC, for a given target machine, is
5009 normally kept in the directory @file{/usr/local/lib/gcc-lib/@var{machine}/@var{version}}.
5011 Thus, sites can customize the effect of @option{-b} or @option{-V} either by
5012 changing the names of these directories or adding alternate names (or
5013 symbolic links). If in directory @file{/usr/local/lib/gcc-lib/} the
5014 file @file{80386} is a link to the file @file{i386v}, then @option{-b
5015 80386} becomes an alias for @option{-b i386v}.
5017 In one respect, the @option{-b} or @option{-V} do not completely change
5018 to a different compiler: the top-level driver program @command{gcc}
5019 that you originally invoked continues to run and invoke the other
5020 executables (preprocessor, compiler per se, assembler and linker)
5021 that do the real work. However, since no real work is done in the
5022 driver program, it usually does not matter that the driver program
5023 in use is not the one for the specified target. It is common for the
5024 interface to the other executables to change incompatibly between
5025 compiler versions, so unless the version specified is very close to that
5026 of the driver (for example, @option{-V 3.0} with a driver program from GCC
5027 version 3.0.1), use of @option{-V} may not work; for example, using
5028 @option{-V 2.95.2} will not work with a driver program from GCC 3.0.
5030 The only way that the driver program depends on the target machine is
5031 in the parsing and handling of special machine-specific options.
5032 However, this is controlled by a file which is found, along with the
5033 other executables, in the directory for the specified version and
5034 target machine. As a result, a single installed driver program adapts
5035 to any specified target machine, and sufficiently similar compiler
5038 The driver program executable does control one significant thing,
5039 however: the default version and target machine. Therefore, you can
5040 install different instances of the driver program, compiled for
5041 different targets or versions, under different names.
5043 For example, if the driver for version 2.0 is installed as @command{ogcc}
5044 and that for version 2.1 is installed as @command{gcc}, then the command
5045 @command{gcc} will use version 2.1 by default, while @command{ogcc} will use
5046 2.0 by default. However, you can choose either version with either
5047 command with the @option{-V} option.
5049 @node Submodel Options
5050 @section Hardware Models and Configurations
5051 @cindex submodel options
5052 @cindex specifying hardware config
5053 @cindex hardware models and configurations, specifying
5054 @cindex machine dependent options
5056 Earlier we discussed the standard option @option{-b} which chooses among
5057 different installed compilers for completely different target
5058 machines, such as VAX vs.@: 68000 vs.@: 80386.
5060 In addition, each of these target machine types can have its own
5061 special options, starting with @samp{-m}, to choose among various
5062 hardware models or configurations---for example, 68010 vs 68020,
5063 floating coprocessor or none. A single installed version of the
5064 compiler can compile for any model or configuration, according to the
5067 Some configurations of the compiler also support additional special
5068 options, usually for compatibility with other compilers on the same
5072 These options are defined by the macro @code{TARGET_SWITCHES} in the
5073 machine description. The default for the options is also defined by
5074 that macro, which enables you to change the defaults.
5089 * RS/6000 and PowerPC Options::
5092 * i386 and x86-64 Options::
5094 * Intel 960 Options::
5095 * DEC Alpha Options::
5099 * System V Options::
5100 * TMS320C3x/C4x Options::
5108 * S/390 and zSeries Options::
5112 @node M680x0 Options
5113 @subsection M680x0 Options
5114 @cindex M680x0 options
5116 These are the @samp{-m} options defined for the 68000 series. The default
5117 values for these options depends on which style of 68000 was selected when
5118 the compiler was configured; the defaults for the most common choices are
5126 Generate output for a 68000. This is the default
5127 when the compiler is configured for 68000-based systems.
5129 Use this option for microcontrollers with a 68000 or EC000 core,
5130 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
5136 Generate output for a 68020. This is the default
5137 when the compiler is configured for 68020-based systems.
5141 Generate output containing 68881 instructions for floating point.
5142 This is the default for most 68020 systems unless @option{--nfp} was
5143 specified when the compiler was configured.
5147 Generate output for a 68030. This is the default when the compiler is
5148 configured for 68030-based systems.
5152 Generate output for a 68040. This is the default when the compiler is
5153 configured for 68040-based systems.
5155 This option inhibits the use of 68881/68882 instructions that have to be
5156 emulated by software on the 68040. Use this option if your 68040 does not
5157 have code to emulate those instructions.
5161 Generate output for a 68060. This is the default when the compiler is
5162 configured for 68060-based systems.
5164 This option inhibits the use of 68020 and 68881/68882 instructions that
5165 have to be emulated by software on the 68060. Use this option if your 68060
5166 does not have code to emulate those instructions.
5170 Generate output for a CPU32. This is the default
5171 when the compiler is configured for CPU32-based systems.
5173 Use this option for microcontrollers with a
5174 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
5175 68336, 68340, 68341, 68349 and 68360.
5179 Generate output for a 520X ``coldfire'' family cpu. This is the default
5180 when the compiler is configured for 520X-based systems.
5182 Use this option for microcontroller with a 5200 core, including
5183 the MCF5202, MCF5203, MCF5204 and MCF5202.
5188 Generate output for a 68040, without using any of the new instructions.
5189 This results in code which can run relatively efficiently on either a
5190 68020/68881 or a 68030 or a 68040. The generated code does use the
5191 68881 instructions that are emulated on the 68040.
5195 Generate output for a 68060, without using any of the new instructions.
5196 This results in code which can run relatively efficiently on either a
5197 68020/68881 or a 68030 or a 68040. The generated code does use the
5198 68881 instructions that are emulated on the 68060.
5202 Generate output containing Sun FPA instructions for floating point.
5205 @opindex msoft-float
5206 Generate output containing library calls for floating point.
5207 @strong{Warning:} the requisite libraries are not available for all m68k
5208 targets. Normally the facilities of the machine's usual C compiler are
5209 used, but this can't be done directly in cross-compilation. You must
5210 make your own arrangements to provide suitable library functions for
5211 cross-compilation. The embedded targets @samp{m68k-*-aout} and
5212 @samp{m68k-*-coff} do provide software floating point support.
5216 Consider type @code{int} to be 16 bits wide, like @code{short int}.
5219 @opindex mnobitfield
5220 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
5221 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
5225 Do use the bit-field instructions. The @option{-m68020} option implies
5226 @option{-mbitfield}. This is the default if you use a configuration
5227 designed for a 68020.
5231 Use a different function-calling convention, in which functions
5232 that take a fixed number of arguments return with the @code{rtd}
5233 instruction, which pops their arguments while returning. This
5234 saves one instruction in the caller since there is no need to pop
5235 the arguments there.
5237 This calling convention is incompatible with the one normally
5238 used on Unix, so you cannot use it if you need to call libraries
5239 compiled with the Unix compiler.
5241 Also, you must provide function prototypes for all functions that
5242 take variable numbers of arguments (including @code{printf});
5243 otherwise incorrect code will be generated for calls to those
5246 In addition, seriously incorrect code will result if you call a
5247 function with too many arguments. (Normally, extra arguments are
5248 harmlessly ignored.)
5250 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
5251 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
5254 @itemx -mno-align-int
5256 @opindex mno-align-int
5257 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
5258 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
5259 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
5260 Aligning variables on 32-bit boundaries produces code that runs somewhat
5261 faster on processors with 32-bit busses at the expense of more memory.
5263 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
5264 align structures containing the above types differently than
5265 most published application binary interface specifications for the m68k.
5269 Use the pc-relative addressing mode of the 68000 directly, instead of
5270 using a global offset table. At present, this option implies @option{-fpic},
5271 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
5272 not presently supported with @option{-mpcrel}, though this could be supported for
5273 68020 and higher processors.
5275 @item -mno-strict-align
5276 @itemx -mstrict-align
5277 @opindex mno-strict-align
5278 @opindex mstrict-align
5279 Do not (do) assume that unaligned memory references will be handled by
5284 @node M68hc1x Options
5285 @subsection M68hc1x Options
5286 @cindex M68hc1x options
5288 These are the @samp{-m} options defined for the 68hc11 and 68hc12
5289 microcontrollers. The default values for these options depends on
5290 which style of microcontroller was selected when the compiler was configured;
5291 the defaults for the most common choices are given below.
5298 Generate output for a 68HC11. This is the default
5299 when the compiler is configured for 68HC11-based systems.
5305 Generate output for a 68HC12. This is the default
5306 when the compiler is configured for 68HC12-based systems.
5309 @opindex mauto-incdec
5310 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
5315 Consider type @code{int} to be 16 bits wide, like @code{short int}.
5317 @item -msoft-reg-count=@var{count}
5318 @opindex msoft-reg-count
5319 Specify the number of pseudo-soft registers which are used for the
5320 code generation. The maximum number is 32. Using more pseudo-soft
5321 register may or may not result in better code depending on the program.
5322 The default is 4 for 68HC11 and 2 for 68HC12.
5327 @subsection VAX Options
5330 These @samp{-m} options are defined for the VAX:
5335 Do not output certain jump instructions (@code{aobleq} and so on)
5336 that the Unix assembler for the VAX cannot handle across long
5341 Do output those jump instructions, on the assumption that you
5342 will assemble with the GNU assembler.
5346 Output code for g-format floating point numbers instead of d-format.
5350 @subsection SPARC Options
5351 @cindex SPARC options
5353 These @samp{-m} switches are supported on the SPARC:
5358 @opindex mno-app-regs
5360 Specify @option{-mapp-regs} to generate output using the global registers
5361 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
5364 To be fully SVR4 ABI compliant at the cost of some performance loss,
5365 specify @option{-mno-app-regs}. You should compile libraries and system
5366 software with this option.
5371 @opindex mhard-float
5372 Generate output containing floating point instructions. This is the
5378 @opindex msoft-float
5379 Generate output containing library calls for floating point.
5380 @strong{Warning:} the requisite libraries are not available for all SPARC
5381 targets. Normally the facilities of the machine's usual C compiler are
5382 used, but this cannot be done directly in cross-compilation. You must make
5383 your own arrangements to provide suitable library functions for
5384 cross-compilation. The embedded targets @samp{sparc-*-aout} and
5385 @samp{sparclite-*-*} do provide software floating point support.
5387 @option{-msoft-float} changes the calling convention in the output file;
5388 therefore, it is only useful if you compile @emph{all} of a program with
5389 this option. In particular, you need to compile @file{libgcc.a}, the
5390 library that comes with GCC, with @option{-msoft-float} in order for
5393 @item -mhard-quad-float
5394 @opindex mhard-quad-float
5395 Generate output containing quad-word (long double) floating point
5398 @item -msoft-quad-float
5399 @opindex msoft-quad-float
5400 Generate output containing library calls for quad-word (long double)
5401 floating point instructions. The functions called are those specified
5402 in the SPARC ABI@. This is the default.
5404 As of this writing, there are no sparc implementations that have hardware
5405 support for the quad-word floating point instructions. They all invoke
5406 a trap handler for one of these instructions, and then the trap handler
5407 emulates the effect of the instruction. Because of the trap handler overhead,
5408 this is much slower than calling the ABI library routines. Thus the
5409 @option{-msoft-quad-float} option is the default.
5413 @opindex mno-epilogue
5415 With @option{-mepilogue} (the default), the compiler always emits code for
5416 function exit at the end of each function. Any function exit in
5417 the middle of the function (such as a return statement in C) will
5418 generate a jump to the exit code at the end of the function.
5420 With @option{-mno-epilogue}, the compiler tries to emit exit code inline
5421 at every function exit.
5427 With @option{-mflat}, the compiler does not generate save/restore instructions
5428 and will use a ``flat'' or single register window calling convention.
5429 This model uses %i7 as the frame pointer and is compatible with the normal
5430 register window model. Code from either may be intermixed.
5431 The local registers and the input registers (0--5) are still treated as
5432 ``call saved'' registers and will be saved on the stack as necessary.
5434 With @option{-mno-flat} (the default), the compiler emits save/restore
5435 instructions (except for leaf functions) and is the normal mode of operation.
5437 @item -mno-unaligned-doubles
5438 @itemx -munaligned-doubles
5439 @opindex mno-unaligned-doubles
5440 @opindex munaligned-doubles
5441 Assume that doubles have 8 byte alignment. This is the default.
5443 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
5444 alignment only if they are contained in another type, or if they have an
5445 absolute address. Otherwise, it assumes they have 4 byte alignment.
5446 Specifying this option avoids some rare compatibility problems with code
5447 generated by other compilers. It is not the default because it results
5448 in a performance loss, especially for floating point code.
5450 @item -mno-faster-structs
5451 @itemx -mfaster-structs
5452 @opindex mno-faster-structs
5453 @opindex mfaster-structs
5454 With @option{-mfaster-structs}, the compiler assumes that structures
5455 should have 8 byte alignment. This enables the use of pairs of
5456 @code{ldd} and @code{std} instructions for copies in structure
5457 assignment, in place of twice as many @code{ld} and @code{st} pairs.
5458 However, the use of this changed alignment directly violates the Sparc
5459 ABI@. Thus, it's intended only for use on targets where the developer
5460 acknowledges that their resulting code will not be directly in line with
5461 the rules of the ABI@.
5467 These two options select variations on the SPARC architecture.
5469 By default (unless specifically configured for the Fujitsu SPARClite),
5470 GCC generates code for the v7 variant of the SPARC architecture.
5472 @option{-mv8} will give you SPARC v8 code. The only difference from v7
5473 code is that the compiler emits the integer multiply and integer
5474 divide instructions which exist in SPARC v8 but not in SPARC v7.
5476 @option{-msparclite} will give you SPARClite code. This adds the integer
5477 multiply, integer divide step and scan (@code{ffs}) instructions which
5478 exist in SPARClite but not in SPARC v7.
5480 These options are deprecated and will be deleted in a future GCC release.
5481 They have been replaced with @option{-mcpu=xxx}.
5486 @opindex msupersparc
5487 These two options select the processor for which the code is optimised.
5489 With @option{-mcypress} (the default), the compiler optimizes code for the
5490 Cypress CY7C602 chip, as used in the SparcStation/SparcServer 3xx series.
5491 This is also appropriate for the older SparcStation 1, 2, IPX etc.
5493 With @option{-msupersparc} the compiler optimizes code for the SuperSparc cpu, as
5494 used in the SparcStation 10, 1000 and 2000 series. This flag also enables use
5495 of the full SPARC v8 instruction set.
5497 These options are deprecated and will be deleted in a future GCC release.
5498 They have been replaced with @option{-mcpu=xxx}.
5500 @item -mcpu=@var{cpu_type}
5502 Set the instruction set, register set, and instruction scheduling parameters
5503 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
5504 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
5505 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
5506 @samp{sparclet}, @samp{tsc701}, @samp{v9}, and @samp{ultrasparc}.
5508 Default instruction scheduling parameters are used for values that select
5509 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
5510 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
5512 Here is a list of each supported architecture and their supported
5517 v8: supersparc, hypersparc
5518 sparclite: f930, f934, sparclite86x
5523 @item -mtune=@var{cpu_type}
5525 Set the instruction scheduling parameters for machine type
5526 @var{cpu_type}, but do not set the instruction set or register set that the
5527 option @option{-mcpu=@var{cpu_type}} would.
5529 The same values for @option{-mcpu=@var{cpu_type}} can be used for
5530 @option{-mtune=@var{cpu_type}}, but the only useful values are those
5531 that select a particular cpu implementation. Those are @samp{cypress},
5532 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
5533 @samp{sparclite86x}, @samp{tsc701}, and @samp{ultrasparc}.
5537 These @samp{-m} switches are supported in addition to the above
5538 on the SPARCLET processor.
5541 @item -mlittle-endian
5542 @opindex mlittle-endian
5543 Generate code for a processor running in little-endian mode.
5547 Treat register @code{%g0} as a normal register.
5548 GCC will continue to clobber it as necessary but will not assume
5549 it always reads as 0.
5551 @item -mbroken-saverestore
5552 @opindex mbroken-saverestore
5553 Generate code that does not use non-trivial forms of the @code{save} and
5554 @code{restore} instructions. Early versions of the SPARCLET processor do
5555 not correctly handle @code{save} and @code{restore} instructions used with
5556 arguments. They correctly handle them used without arguments. A @code{save}
5557 instruction used without arguments increments the current window pointer
5558 but does not allocate a new stack frame. It is assumed that the window
5559 overflow trap handler will properly handle this case as will interrupt
5563 These @samp{-m} switches are supported in addition to the above
5564 on SPARC V9 processors in 64-bit environments.
5567 @item -mlittle-endian
5568 @opindex mlittle-endian
5569 Generate code for a processor running in little-endian mode.
5575 Generate code for a 32-bit or 64-bit environment.
5576 The 32-bit environment sets int, long and pointer to 32 bits.
5577 The 64-bit environment sets int to 32 bits and long and pointer
5580 @item -mcmodel=medlow
5581 @opindex mcmodel=medlow
5582 Generate code for the Medium/Low code model: the program must be linked
5583 in the low 32 bits of the address space. Pointers are 64 bits.
5584 Programs can be statically or dynamically linked.
5586 @item -mcmodel=medmid
5587 @opindex mcmodel=medmid
5588 Generate code for the Medium/Middle code model: the program must be linked
5589 in the low 44 bits of the address space, the text segment must be less than
5590 2G bytes, and data segment must be within 2G of the text segment.
5591 Pointers are 64 bits.
5593 @item -mcmodel=medany
5594 @opindex mcmodel=medany
5595 Generate code for the Medium/Anywhere code model: the program may be linked
5596 anywhere in the address space, the text segment must be less than
5597 2G bytes, and data segment must be within 2G of the text segment.
5598 Pointers are 64 bits.
5600 @item -mcmodel=embmedany
5601 @opindex mcmodel=embmedany
5602 Generate code for the Medium/Anywhere code model for embedded systems:
5603 assume a 32-bit text and a 32-bit data segment, both starting anywhere
5604 (determined at link time). Register %g4 points to the base of the
5605 data segment. Pointers are still 64 bits.
5606 Programs are statically linked, PIC is not supported.
5609 @itemx -mno-stack-bias
5610 @opindex mstack-bias
5611 @opindex mno-stack-bias
5612 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
5613 frame pointer if present, are offset by @minus{}2047 which must be added back
5614 when making stack frame references.
5615 Otherwise, assume no such offset is present.
5618 @node Convex Options
5619 @subsection Convex Options
5620 @cindex Convex options
5622 These @samp{-m} options are defined for Convex:
5627 Generate output for C1. The code will run on any Convex machine.
5628 The preprocessor symbol @code{__convex__c1__} is defined.
5632 Generate output for C2. Uses instructions not available on C1.
5633 Scheduling and other optimizations are chosen for max performance on C2.
5634 The preprocessor symbol @code{__convex_c2__} is defined.
5638 Generate output for C32xx. Uses instructions not available on C1.
5639 Scheduling and other optimizations are chosen for max performance on C32.
5640 The preprocessor symbol @code{__convex_c32__} is defined.
5644 Generate output for C34xx. Uses instructions not available on C1.
5645 Scheduling and other optimizations are chosen for max performance on C34.
5646 The preprocessor symbol @code{__convex_c34__} is defined.
5650 Generate output for C38xx. Uses instructions not available on C1.
5651 Scheduling and other optimizations are chosen for max performance on C38.
5652 The preprocessor symbol @code{__convex_c38__} is defined.
5656 Generate code which puts an argument count in the word preceding each
5657 argument list. This is compatible with regular CC, and a few programs
5658 may need the argument count word. GDB and other source-level debuggers
5659 do not need it; this info is in the symbol table.
5662 @opindex mnoargcount
5663 Omit the argument count word. This is the default.
5665 @item -mvolatile-cache
5666 @opindex mvolatile-cache
5667 Allow volatile references to be cached. This is the default.
5669 @item -mvolatile-nocache
5670 @opindex mvolatile-nocache
5671 Volatile references bypass the data cache, going all the way to memory.
5672 This is only needed for multi-processor code that does not use standard
5673 synchronization instructions. Making non-volatile references to volatile
5674 locations will not necessarily work.
5678 Type long is 32 bits, the same as type int. This is the default.
5682 Type long is 64 bits, the same as type long long. This option is useless,
5683 because no library support exists for it.
5686 @node AMD29K Options
5687 @subsection AMD29K Options
5688 @cindex AMD29K options
5690 These @samp{-m} options are defined for the AMD Am29000:
5695 @cindex DW bit (29k)
5696 Generate code that assumes the @code{DW} bit is set, i.e., that byte and
5697 halfword operations are directly supported by the hardware. This is the
5702 Generate code that assumes the @code{DW} bit is not set.
5706 @cindex byte writes (29k)
5707 Generate code that assumes the system supports byte and halfword write
5708 operations. This is the default.
5712 Generate code that assumes the systems does not support byte and
5713 halfword write operations. @option{-mnbw} implies @option{-mndw}.
5717 @cindex memory model (29k)
5718 Use a small memory model that assumes that all function addresses are
5719 either within a single 256 KB segment or at an absolute address of less
5720 than 256k. This allows the @code{call} instruction to be used instead
5721 of a @code{const}, @code{consth}, @code{calli} sequence.
5725 Use the normal memory model: Generate @code{call} instructions only when
5726 calling functions in the same file and @code{calli} instructions
5727 otherwise. This works if each file occupies less than 256 KB but allows
5728 the entire executable to be larger than 256 KB@. This is the default.
5732 Always use @code{calli} instructions. Specify this option if you expect
5733 a single file to compile into more than 256 KB of code.
5737 @cindex processor selection (29k)
5738 Generate code for the Am29050.
5742 Generate code for the Am29000. This is the default.
5744 @item -mkernel-registers
5745 @opindex mkernel-registers
5746 @cindex kernel and user registers (29k)
5747 Generate references to registers @code{gr64-gr95} instead of to
5748 registers @code{gr96-gr127}. This option can be used when compiling
5749 kernel code that wants a set of global registers disjoint from that used
5752 Note that when this option is used, register names in @samp{-f} flags
5753 must use the normal, user-mode, names.
5755 @item -muser-registers
5756 @opindex muser-registers
5757 Use the normal set of global registers, @code{gr96-gr127}. This is the
5761 @itemx -mno-stack-check
5762 @opindex mstack-check
5763 @opindex mno-stack-check
5764 @cindex stack checks (29k)
5765 Insert (or do not insert) a call to @code{__msp_check} after each stack
5766 adjustment. This is often used for kernel code.
5769 @itemx -mno-storem-bug
5770 @opindex mstorem-bug
5771 @opindex mno-storem-bug
5772 @cindex storem bug (29k)
5773 @option{-mstorem-bug} handles 29k processors which cannot handle the
5774 separation of a mtsrim insn and a storem instruction (most 29000 chips
5775 to date, but not the 29050).
5777 @item -mno-reuse-arg-regs
5778 @itemx -mreuse-arg-regs
5779 @opindex mno-reuse-arg-regs
5780 @opindex mreuse-arg-regs
5781 @option{-mno-reuse-arg-regs} tells the compiler to only use incoming argument
5782 registers for copying out arguments. This helps detect calling a function
5783 with fewer arguments than it was declared with.
5785 @item -mno-impure-text
5786 @itemx -mimpure-text
5787 @opindex mno-impure-text
5788 @opindex mimpure-text
5789 @option{-mimpure-text}, used in addition to @option{-shared}, tells the compiler to
5790 not pass @option{-assert pure-text} to the linker when linking a shared object.
5793 @opindex msoft-float
5794 Generate output containing library calls for floating point.
5795 @strong{Warning:} the requisite libraries are not part of GCC@.
5796 Normally the facilities of the machine's usual C compiler are used, but
5797 this can't be done directly in cross-compilation. You must make your
5798 own arrangements to provide suitable library functions for
5803 Do not generate multm or multmu instructions. This is useful for some embedded
5804 systems which do not have trap handlers for these instructions.
5808 @subsection ARM Options
5811 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
5816 @opindex mapcs-frame
5817 Generate a stack frame that is compliant with the ARM Procedure Call
5818 Standard for all functions, even if this is not strictly necessary for
5819 correct execution of the code. Specifying @option{-fomit-frame-pointer}
5820 with this option will cause the stack frames not to be generated for
5821 leaf functions. The default is @option{-mno-apcs-frame}.
5825 This is a synonym for @option{-mapcs-frame}.
5829 Generate code for a processor running with a 26-bit program counter,
5830 and conforming to the function calling standards for the APCS 26-bit
5831 option. This option replaces the @option{-m2} and @option{-m3} options
5832 of previous releases of the compiler.
5836 Generate code for a processor running with a 32-bit program counter,
5837 and conforming to the function calling standards for the APCS 32-bit
5838 option. This option replaces the @option{-m6} option of previous releases
5842 @c not currently implemented
5843 @item -mapcs-stack-check
5844 @opindex mapcs-stack-check
5845 Generate code to check the amount of stack space available upon entry to
5846 every function (that actually uses some stack space). If there is
5847 insufficient space available then either the function
5848 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
5849 called, depending upon the amount of stack space required. The run time
5850 system is required to provide these functions. The default is
5851 @option{-mno-apcs-stack-check}, since this produces smaller code.
5853 @c not currently implemented
5855 @opindex mapcs-float
5856 Pass floating point arguments using the float point registers. This is
5857 one of the variants of the APCS@. This option is recommended if the
5858 target hardware has a floating point unit or if a lot of floating point
5859 arithmetic is going to be performed by the code. The default is
5860 @option{-mno-apcs-float}, since integer only code is slightly increased in
5861 size if @option{-mapcs-float} is used.
5863 @c not currently implemented
5864 @item -mapcs-reentrant
5865 @opindex mapcs-reentrant
5866 Generate reentrant, position independent code. The default is
5867 @option{-mno-apcs-reentrant}.
5870 @item -mthumb-interwork
5871 @opindex mthumb-interwork
5872 Generate code which supports calling between the ARM and Thumb
5873 instruction sets. Without this option the two instruction sets cannot
5874 be reliably used inside one program. The default is
5875 @option{-mno-thumb-interwork}, since slightly larger code is generated
5876 when @option{-mthumb-interwork} is specified.
5878 @item -mno-sched-prolog
5879 @opindex mno-sched-prolog
5880 Prevent the reordering of instructions in the function prolog, or the
5881 merging of those instruction with the instructions in the function's
5882 body. This means that all functions will start with a recognizable set
5883 of instructions (or in fact one of a choice from a small set of
5884 different function prologues), and this information can be used to
5885 locate the start if functions inside an executable piece of code. The
5886 default is @option{-msched-prolog}.
5889 @opindex mhard-float
5890 Generate output containing floating point instructions. This is the
5894 @opindex msoft-float
5895 Generate output containing library calls for floating point.
5896 @strong{Warning:} the requisite libraries are not available for all ARM
5897 targets. Normally the facilities of the machine's usual C compiler are
5898 used, but this cannot be done directly in cross-compilation. You must make
5899 your own arrangements to provide suitable library functions for
5902 @option{-msoft-float} changes the calling convention in the output file;
5903 therefore, it is only useful if you compile @emph{all} of a program with
5904 this option. In particular, you need to compile @file{libgcc.a}, the
5905 library that comes with GCC, with @option{-msoft-float} in order for
5908 @item -mlittle-endian
5909 @opindex mlittle-endian
5910 Generate code for a processor running in little-endian mode. This is
5911 the default for all standard configurations.
5914 @opindex mbig-endian
5915 Generate code for a processor running in big-endian mode; the default is
5916 to compile code for a little-endian processor.
5918 @item -mwords-little-endian
5919 @opindex mwords-little-endian
5920 This option only applies when generating code for big-endian processors.
5921 Generate code for a little-endian word order but a big-endian byte
5922 order. That is, a byte order of the form @samp{32107654}. Note: this
5923 option should only be used if you require compatibility with code for
5924 big-endian ARM processors generated by versions of the compiler prior to
5927 @item -malignment-traps
5928 @opindex malignment-traps
5929 Generate code that will not trap if the MMU has alignment traps enabled.
5930 On ARM architectures prior to ARMv4, there were no instructions to
5931 access half-word objects stored in memory. However, when reading from
5932 memory a feature of the ARM architecture allows a word load to be used,
5933 even if the address is unaligned, and the processor core will rotate the
5934 data as it is being loaded. This option tells the compiler that such
5935 misaligned accesses will cause a MMU trap and that it should instead
5936 synthesise the access as a series of byte accesses. The compiler can
5937 still use word accesses to load half-word data if it knows that the
5938 address is aligned to a word boundary.
5940 This option is ignored when compiling for ARM architecture 4 or later,
5941 since these processors have instructions to directly access half-word
5944 @item -mno-alignment-traps
5945 @opindex mno-alignment-traps
5946 Generate code that assumes that the MMU will not trap unaligned
5947 accesses. This produces better code when the target instruction set
5948 does not have half-word memory operations (i.e.@: implementations prior to
5951 Note that you cannot use this option to access unaligned word objects,
5952 since the processor will only fetch one 32-bit aligned object from
5955 The default setting for most targets is @option{-mno-alignment-traps}, since
5956 this produces better code when there are no half-word memory
5957 instructions available.
5959 @item -mshort-load-bytes
5960 @itemx -mno-short-load-words
5961 @opindex mshort-load-bytes
5962 @opindex mno-short-load-words
5963 These are deprecated aliases for @option{-malignment-traps}.
5965 @item -mno-short-load-bytes
5966 @itemx -mshort-load-words
5967 @opindex mno-short-load-bytes
5968 @opindex mshort-load-words
5969 This are deprecated aliases for @option{-mno-alignment-traps}.
5973 This option only applies to RISC iX@. Emulate the native BSD-mode
5974 compiler. This is the default if @option{-ansi} is not specified.
5978 This option only applies to RISC iX@. Emulate the native X/Open-mode
5981 @item -mno-symrename
5982 @opindex mno-symrename
5983 This option only applies to RISC iX@. Do not run the assembler
5984 post-processor, @samp{symrename}, after code has been assembled.
5985 Normally it is necessary to modify some of the standard symbols in
5986 preparation for linking with the RISC iX C library; this option
5987 suppresses this pass. The post-processor is never run when the
5988 compiler is built for cross-compilation.
5990 @item -mcpu=@var{name}
5992 This specifies the name of the target ARM processor. GCC uses this name
5993 to determine what kind of instructions it can emit when generating
5994 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
5995 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
5996 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
5997 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
5998 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
5999 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm8},
6000 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6001 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6002 @samp{arm920t}, @samp{arm940t}, @samp{arm9tdmi}, @samp{arm10tdmi},
6003 @samp{arm1020t}, @samp{xscale}.
6005 @itemx -mtune=@var{name}
6007 This option is very similar to the @option{-mcpu=} option, except that
6008 instead of specifying the actual target processor type, and hence
6009 restricting which instructions can be used, it specifies that GCC should
6010 tune the performance of the code as if the target were of the type
6011 specified in this option, but still choosing the instructions that it
6012 will generate based on the cpu specified by a @option{-mcpu=} option.
6013 For some ARM implementations better performance can be obtained by using
6016 @item -march=@var{name}
6018 This specifies the name of the target ARM architecture. GCC uses this
6019 name to determine what kind of instructions it can emit when generating
6020 assembly code. This option can be used in conjunction with or instead
6021 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6022 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6023 @samp{armv5}, @samp{armv5t}, @samp{armv5te}.
6025 @item -mfpe=@var{number}
6026 @itemx -mfp=@var{number}
6029 This specifies the version of the floating point emulation available on
6030 the target. Permissible values are 2 and 3. @option{-mfp=} is a synonym
6031 for @option{-mfpe=}, for compatibility with older versions of GCC@.
6033 @item -mstructure-size-boundary=@var{n}
6034 @opindex mstructure-size-boundary
6035 The size of all structures and unions will be rounded up to a multiple
6036 of the number of bits set by this option. Permissible values are 8 and
6037 32. The default value varies for different toolchains. For the COFF
6038 targeted toolchain the default value is 8. Specifying the larger number
6039 can produce faster, more efficient code, but can also increase the size
6040 of the program. The two values are potentially incompatible. Code
6041 compiled with one value cannot necessarily expect to work with code or
6042 libraries compiled with the other value, if they exchange information
6043 using structures or unions.
6045 @item -mabort-on-noreturn
6046 @opindex mabort-on-noreturn
6047 Generate a call to the function @code{abort} at the end of a
6048 @code{noreturn} function. It will be executed if the function tries to
6052 @itemx -mno-long-calls
6053 @opindex mlong-calls
6054 @opindex mno-long-calls
6055 Tells the compiler to perform function calls by first loading the
6056 address of the function into a register and then performing a subroutine
6057 call on this register. This switch is needed if the target function
6058 will lie outside of the 64 megabyte addressing range of the offset based
6059 version of subroutine call instruction.
6061 Even if this switch is enabled, not all function calls will be turned
6062 into long calls. The heuristic is that static functions, functions
6063 which have the @samp{short-call} attribute, functions that are inside
6064 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6065 definitions have already been compiled within the current compilation
6066 unit, will not be turned into long calls. The exception to this rule is
6067 that weak function definitions, functions with the @samp{long-call}
6068 attribute or the @samp{section} attribute, and functions that are within
6069 the scope of a @samp{#pragma long_calls} directive, will always be
6070 turned into long calls.
6072 This feature is not enabled by default. Specifying
6073 @option{-mno-long-calls} will restore the default behaviour, as will
6074 placing the function calls within the scope of a @samp{#pragma
6075 long_calls_off} directive. Note these switches have no effect on how
6076 the compiler generates code to handle function calls via function
6079 @item -mnop-fun-dllimport
6080 @opindex mnop-fun-dllimport
6081 Disable support for the @code{dllimport} attribute.
6083 @item -msingle-pic-base
6084 @opindex msingle-pic-base
6085 Treat the register used for PIC addressing as read-only, rather than
6086 loading it in the prologue for each function. The run-time system is
6087 responsible for initialising this register with an appropriate value
6088 before execution begins.
6090 @item -mpic-register=@var{reg}
6091 @opindex mpic-register
6092 Specify the register to be used for PIC addressing. The default is R10
6093 unless stack-checking is enabled, when R9 is used.
6095 @item -mpoke-function-name
6096 @opindex mpoke-function-name
6097 Write the name of each function into the text section, directly
6098 preceding the function prologue. The generated code is similar to this:
6102 .ascii "arm_poke_function_name", 0
6105 .word 0xff000000 + (t1 - t0)
6106 arm_poke_function_name
6108 stmfd sp!, @{fp, ip, lr, pc@}
6112 When performing a stack backtrace, code can inspect the value of
6113 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6114 location @code{pc - 12} and the top 8 bits are set, then we know that
6115 there is a function name embedded immediately preceding this location
6116 and has length @code{((pc[-3]) & 0xff000000)}.
6120 Generate code for the 16-bit Thumb instruction set. The default is to
6121 use the 32-bit ARM instruction set.
6124 @opindex mtpcs-frame
6125 Generate a stack frame that is compliant with the Thumb Procedure Call
6126 Standard for all non-leaf functions. (A leaf function is one that does
6127 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6129 @item -mtpcs-leaf-frame
6130 @opindex mtpcs-leaf-frame
6131 Generate a stack frame that is compliant with the Thumb Procedure Call
6132 Standard for all leaf functions. (A leaf function is one that does
6133 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6135 @item -mcallee-super-interworking
6136 @opindex mcallee-super-interworking
6137 Gives all externally visible functions in the file being compiled an ARM
6138 instruction set header which switches to Thumb mode before executing the
6139 rest of the function. This allows these functions to be called from
6140 non-interworking code.
6142 @item -mcaller-super-interworking
6143 @opindex mcaller-super-interworking
6144 Allows calls via function pointers (including virtual functions) to
6145 execute correctly regardless of whether the target code has been
6146 compiled for interworking or not. There is a small overhead in the cost
6147 of executing a function pointer if this option is enabled.
6151 @node MN10200 Options
6152 @subsection MN10200 Options
6153 @cindex MN10200 options
6154 These @option{-m} options are defined for Matsushita MN10200 architectures:
6159 Indicate to the linker that it should perform a relaxation optimization pass
6160 to shorten branches, calls and absolute memory addresses. This option only
6161 has an effect when used on the command line for the final link step.
6163 This option makes symbolic debugging impossible.
6166 @node MN10300 Options
6167 @subsection MN10300 Options
6168 @cindex MN10300 options
6169 These @option{-m} options are defined for Matsushita MN10300 architectures:
6174 Generate code to avoid bugs in the multiply instructions for the MN10300
6175 processors. This is the default.
6178 @opindex mno-mult-bug
6179 Do not generate code to avoid bugs in the multiply instructions for the
6184 Generate code which uses features specific to the AM33 processor.
6188 Do not generate code which uses features specific to the AM33 processor. This
6193 Do not link in the C run-time initialization object file.
6197 Indicate to the linker that it should perform a relaxation optimization pass
6198 to shorten branches, calls and absolute memory addresses. This option only
6199 has an effect when used on the command line for the final link step.
6201 This option makes symbolic debugging impossible.
6205 @node M32R/D Options
6206 @subsection M32R/D Options
6207 @cindex M32R/D options
6209 These @option{-m} options are defined for Mitsubishi M32R/D architectures:
6212 @item -mcode-model=small
6213 @opindex mcode-model=small
6214 Assume all objects live in the lower 16MB of memory (so that their addresses
6215 can be loaded with the @code{ld24} instruction), and assume all subroutines
6216 are reachable with the @code{bl} instruction.
6217 This is the default.
6219 The addressability of a particular object can be set with the
6220 @code{model} attribute.
6222 @item -mcode-model=medium
6223 @opindex mcode-model=medium
6224 Assume objects may be anywhere in the 32-bit address space (the compiler
6225 will generate @code{seth/add3} instructions to load their addresses), and
6226 assume all subroutines are reachable with the @code{bl} instruction.
6228 @item -mcode-model=large
6229 @opindex mcode-model=large
6230 Assume objects may be anywhere in the 32-bit address space (the compiler
6231 will generate @code{seth/add3} instructions to load their addresses), and
6232 assume subroutines may not be reachable with the @code{bl} instruction
6233 (the compiler will generate the much slower @code{seth/add3/jl}
6234 instruction sequence).
6237 @opindex msdata=none
6238 Disable use of the small data area. Variables will be put into
6239 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
6240 @code{section} attribute has been specified).
6241 This is the default.
6243 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
6244 Objects may be explicitly put in the small data area with the
6245 @code{section} attribute using one of these sections.
6248 @opindex msdata=sdata
6249 Put small global and static data in the small data area, but do not
6250 generate special code to reference them.
6254 Put small global and static data in the small data area, and generate
6255 special instructions to reference them.
6259 @cindex smaller data references
6260 Put global and static objects less than or equal to @var{num} bytes
6261 into the small data or bss sections instead of the normal data or bss
6262 sections. The default value of @var{num} is 8.
6263 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
6264 for this option to have any effect.
6266 All modules should be compiled with the same @option{-G @var{num}} value.
6267 Compiling with different values of @var{num} may or may not work; if it
6268 doesn't the linker will give an error message---incorrect code will not be
6274 @subsection M88K Options
6275 @cindex M88k options
6277 These @samp{-m} options are defined for Motorola 88k architectures:
6282 Generate code that works well on both the m88100 and the
6287 Generate code that works best for the m88100, but that also
6292 Generate code that works best for the m88110, and may not run
6297 Obsolete option to be removed from the next revision.
6300 @item -midentify-revision
6301 @opindex midentify-revision
6302 @cindex identifying source, compiler (88k)
6303 Include an @code{ident} directive in the assembler output recording the
6304 source file name, compiler name and version, timestamp, and compilation
6307 @item -mno-underscores
6308 @opindex mno-underscores
6309 @cindex underscores, avoiding (88k)
6310 In assembler output, emit symbol names without adding an underscore
6311 character at the beginning of each name. The default is to use an
6312 underscore as prefix on each name.
6314 @item -mocs-debug-info
6315 @itemx -mno-ocs-debug-info
6316 @opindex mocs-debug-info
6317 @opindex mno-ocs-debug-info
6319 @cindex debugging, 88k OCS
6320 Include (or omit) additional debugging information (about registers used
6321 in each stack frame) as specified in the 88open Object Compatibility
6322 Standard, ``OCS''@. This extra information allows debugging of code that
6323 has had the frame pointer eliminated. The default for DG/UX, SVr4, and
6324 Delta 88 SVr3.2 is to include this information; other 88k configurations
6325 omit this information by default.
6327 @item -mocs-frame-position
6328 @opindex mocs-frame-position
6329 @cindex register positions in frame (88k)
6330 When emitting COFF debugging information for automatic variables and
6331 parameters stored on the stack, use the offset from the canonical frame
6332 address, which is the stack pointer (register 31) on entry to the
6333 function. The DG/UX, SVr4, Delta88 SVr3.2, and BCS configurations use
6334 @option{-mocs-frame-position}; other 88k configurations have the default
6335 @option{-mno-ocs-frame-position}.
6337 @item -mno-ocs-frame-position
6338 @opindex mno-ocs-frame-position
6339 @cindex register positions in frame (88k)
6340 When emitting COFF debugging information for automatic variables and
6341 parameters stored on the stack, use the offset from the frame pointer
6342 register (register 30). When this option is in effect, the frame
6343 pointer is not eliminated when debugging information is selected by the
6346 @item -moptimize-arg-area
6347 @opindex moptimize-arg-area
6348 @cindex arguments in frame (88k)
6349 Save space by reorganizing the stack frame. This option generates code
6350 that does not agree with the 88open specifications, but uses less
6353 @itemx -mno-optimize-arg-area
6354 @opindex mno-optimize-arg-area
6355 Do not reorganize the stack frame to save space. This is the default.
6356 The generated conforms to the specification, but uses more memory.
6358 @item -mshort-data-@var{num}
6359 @opindex mshort-data
6360 @cindex smaller data references (88k)
6361 @cindex r0-relative references (88k)
6362 Generate smaller data references by making them relative to @code{r0},
6363 which allows loading a value using a single instruction (rather than the
6364 usual two). You control which data references are affected by
6365 specifying @var{num} with this option. For example, if you specify
6366 @option{-mshort-data-512}, then the data references affected are those
6367 involving displacements of less than 512 bytes.
6368 @option{-mshort-data-@var{num}} is not effective for @var{num} greater
6371 @item -mserialize-volatile
6372 @opindex mserialize-volatile
6373 @itemx -mno-serialize-volatile
6374 @opindex mno-serialize-volatile
6375 @cindex sequential consistency on 88k
6376 Do, or don't, generate code to guarantee sequential consistency
6377 of volatile memory references. By default, consistency is
6380 The order of memory references made by the MC88110 processor does
6381 not always match the order of the instructions requesting those
6382 references. In particular, a load instruction may execute before
6383 a preceding store instruction. Such reordering violates
6384 sequential consistency of volatile memory references, when there
6385 are multiple processors. When consistency must be guaranteed,
6386 GCC generates special instructions, as needed, to force
6387 execution in the proper order.
6389 The MC88100 processor does not reorder memory references and so
6390 always provides sequential consistency. However, by default, GCC
6391 generates the special instructions to guarantee consistency
6392 even when you use @option{-m88100}, so that the code may be run on an
6393 MC88110 processor. If you intend to run your code only on the
6394 MC88100 processor, you may use @option{-mno-serialize-volatile}.
6396 The extra code generated to guarantee consistency may affect the
6397 performance of your application. If you know that you can safely
6398 forgo this guarantee, you may use @option{-mno-serialize-volatile}.
6404 @cindex assembler syntax, 88k
6406 Turn on (@option{-msvr4}) or off (@option{-msvr3}) compiler extensions
6407 related to System V release 4 (SVr4). This controls the following:
6411 Which variant of the assembler syntax to emit.
6413 @option{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
6414 that is used on System V release 4.
6416 @option{-msvr4} makes GCC issue additional declaration directives used in
6420 @option{-msvr4} is the default for the m88k-motorola-sysv4 and
6421 m88k-dg-dgux m88k configurations. @option{-msvr3} is the default for all
6422 other m88k configurations.
6424 @item -mversion-03.00
6425 @opindex mversion-03.00
6426 This option is obsolete, and is ignored.
6427 @c ??? which asm syntax better for GAS? option there too?
6429 @item -mno-check-zero-division
6430 @itemx -mcheck-zero-division
6431 @opindex mno-check-zero-division
6432 @opindex mcheck-zero-division
6433 @cindex zero division on 88k
6434 Do, or don't, generate code to guarantee that integer division by
6435 zero will be detected. By default, detection is guaranteed.
6437 Some models of the MC88100 processor fail to trap upon integer
6438 division by zero under certain conditions. By default, when
6439 compiling code that might be run on such a processor, GCC
6440 generates code that explicitly checks for zero-valued divisors
6441 and traps with exception number 503 when one is detected. Use of
6442 @option{-mno-check-zero-division} suppresses such checking for code
6443 generated to run on an MC88100 processor.
6445 GCC assumes that the MC88110 processor correctly detects all instances
6446 of integer division by zero. When @option{-m88110} is specified, no
6447 explicit checks for zero-valued divisors are generated, and both
6448 @option{-mcheck-zero-division} and @option{-mno-check-zero-division} are
6451 @item -muse-div-instruction
6452 @opindex muse-div-instruction
6453 @cindex divide instruction, 88k
6454 Use the div instruction for signed integer division on the
6455 MC88100 processor. By default, the div instruction is not used.
6457 On the MC88100 processor the signed integer division instruction
6458 div) traps to the operating system on a negative operand. The
6459 operating system transparently completes the operation, but at a
6460 large cost in execution time. By default, when compiling code
6461 that might be run on an MC88100 processor, GCC emulates signed
6462 integer division using the unsigned integer division instruction
6463 divu), thereby avoiding the large penalty of a trap to the
6464 operating system. Such emulation has its own, smaller, execution
6465 cost in both time and space. To the extent that your code's
6466 important signed integer division operations are performed on two
6467 nonnegative operands, it may be desirable to use the div
6468 instruction directly.
6470 On the MC88110 processor the div instruction (also known as the
6471 divs instruction) processes negative operands without trapping to
6472 the operating system. When @option{-m88110} is specified,
6473 @option{-muse-div-instruction} is ignored, and the div instruction is used
6474 for signed integer division.
6476 Note that the result of dividing @code{INT_MIN} by @minus{}1 is undefined. In
6477 particular, the behavior of such a division with and without
6478 @option{-muse-div-instruction} may differ.
6480 @item -mtrap-large-shift
6481 @itemx -mhandle-large-shift
6482 @opindex mtrap-large-shift
6483 @opindex mhandle-large-shift
6484 @cindex bit shift overflow (88k)
6485 @cindex large bit shifts (88k)
6486 Include code to detect bit-shifts of more than 31 bits; respectively,
6487 trap such shifts or emit code to handle them properly. By default GCC
6488 makes no special provision for large bit shifts.
6490 @item -mwarn-passed-structs
6491 @opindex mwarn-passed-structs
6492 @cindex structure passing (88k)
6493 Warn when a function passes a struct as an argument or result.
6494 Structure-passing conventions have changed during the evolution of the C
6495 language, and are often the source of portability problems. By default,
6496 GCC issues no such warning.
6499 @c break page here to avoid unsightly interparagraph stretch.
6503 @node RS/6000 and PowerPC Options
6504 @subsection IBM RS/6000 and PowerPC Options
6505 @cindex RS/6000 and PowerPC Options
6506 @cindex IBM RS/6000 and PowerPC Options
6508 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
6516 @itemx -mpowerpc-gpopt
6517 @itemx -mno-powerpc-gpopt
6518 @itemx -mpowerpc-gfxopt
6519 @itemx -mno-powerpc-gfxopt
6521 @itemx -mno-powerpc64
6527 @opindex mno-powerpc
6528 @opindex mpowerpc-gpopt
6529 @opindex mno-powerpc-gpopt
6530 @opindex mpowerpc-gfxopt
6531 @opindex mno-powerpc-gfxopt
6533 @opindex mno-powerpc64
6534 GCC supports two related instruction set architectures for the
6535 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
6536 instructions supported by the @samp{rios} chip set used in the original
6537 RS/6000 systems and the @dfn{PowerPC} instruction set is the
6538 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
6539 the IBM 4xx microprocessors.
6541 Neither architecture is a subset of the other. However there is a
6542 large common subset of instructions supported by both. An MQ
6543 register is included in processors supporting the POWER architecture.
6545 You use these options to specify which instructions are available on the
6546 processor you are using. The default value of these options is
6547 determined when configuring GCC@. Specifying the
6548 @option{-mcpu=@var{cpu_type}} overrides the specification of these
6549 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
6550 rather than the options listed above.
6552 The @option{-mpower} option allows GCC to generate instructions that
6553 are found only in the POWER architecture and to use the MQ register.
6554 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
6555 to generate instructions that are present in the POWER2 architecture but
6556 not the original POWER architecture.
6558 The @option{-mpowerpc} option allows GCC to generate instructions that
6559 are found only in the 32-bit subset of the PowerPC architecture.
6560 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
6561 GCC to use the optional PowerPC architecture instructions in the
6562 General Purpose group, including floating-point square root. Specifying
6563 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
6564 use the optional PowerPC architecture instructions in the Graphics
6565 group, including floating-point select.
6567 The @option{-mpowerpc64} option allows GCC to generate the additional
6568 64-bit instructions that are found in the full PowerPC64 architecture
6569 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
6570 @option{-mno-powerpc64}.
6572 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
6573 will use only the instructions in the common subset of both
6574 architectures plus some special AIX common-mode calls, and will not use
6575 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
6576 permits GCC to use any instruction from either architecture and to
6577 allow use of the MQ register; specify this for the Motorola MPC601.
6579 @item -mnew-mnemonics
6580 @itemx -mold-mnemonics
6581 @opindex mnew-mnemonics
6582 @opindex mold-mnemonics
6583 Select which mnemonics to use in the generated assembler code. With
6584 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
6585 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
6586 assembler mnemonics defined for the POWER architecture. Instructions
6587 defined in only one architecture have only one mnemonic; GCC uses that
6588 mnemonic irrespective of which of these options is specified.
6590 GCC defaults to the mnemonics appropriate for the architecture in
6591 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
6592 value of these option. Unless you are building a cross-compiler, you
6593 should normally not specify either @option{-mnew-mnemonics} or
6594 @option{-mold-mnemonics}, but should instead accept the default.
6596 @item -mcpu=@var{cpu_type}
6598 Set architecture type, register usage, choice of mnemonics, and
6599 instruction scheduling parameters for machine type @var{cpu_type}.
6600 Supported values for @var{cpu_type} are @samp{rios}, @samp{rios1},
6601 @samp{rsc}, @samp{rios2}, @samp{rs64a}, @samp{601}, @samp{602},
6602 @samp{603}, @samp{603e}, @samp{604}, @samp{604e}, @samp{620},
6603 @samp{630}, @samp{740}, @samp{7400}, @samp{7450}, @samp{750},
6604 @samp{power}, @samp{power2}, @samp{powerpc}, @samp{403}, @samp{505},
6605 @samp{801}, @samp{821}, @samp{823}, and @samp{860} and @samp{common}.
6607 @option{-mcpu=common} selects a completely generic processor. Code
6608 generated under this option will run on any POWER or PowerPC processor.
6609 GCC will use only the instructions in the common subset of both
6610 architectures, and will not use the MQ register. GCC assumes a generic
6611 processor model for scheduling purposes.
6613 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
6614 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
6615 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
6616 types, with an appropriate, generic processor model assumed for
6617 scheduling purposes.
6619 The other options specify a specific processor. Code generated under
6620 those options will run best on that processor, and may not run at all on
6623 The @option{-mcpu} options automatically enable or disable other
6624 @option{-m} options as follows:
6628 @option{-mno-power}, @option{-mno-powerc}
6635 @option{-mpower}, @option{-mno-powerpc}, @option{-mno-new-mnemonics}
6650 @option{-mno-power}, @option{-mpowerpc}, @option{-mnew-mnemonics}
6653 @option{-mpower}, @option{-mpowerpc}, @option{-mnew-mnemonics}
6658 @option{-mno-power}, @option{-mpowerpc}, @option{-mnew-mnemonics}, @option{-msoft-float}
6661 @item -mtune=@var{cpu_type}
6663 Set the instruction scheduling parameters for machine type
6664 @var{cpu_type}, but do not set the architecture type, register usage, or
6665 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
6666 values for @var{cpu_type} are used for @option{-mtune} as for
6667 @option{-mcpu}. If both are specified, the code generated will use the
6668 architecture, registers, and mnemonics set by @option{-mcpu}, but the
6669 scheduling parameters set by @option{-mtune}.
6672 @itemx -mno-fp-in-toc
6673 @itemx -mno-sum-in-toc
6674 @itemx -mminimal-toc
6676 @opindex mno-fp-in-toc
6677 @opindex mno-sum-in-toc
6678 @opindex mminimal-toc
6679 Modify generation of the TOC (Table Of Contents), which is created for
6680 every executable file. The @option{-mfull-toc} option is selected by
6681 default. In that case, GCC will allocate at least one TOC entry for
6682 each unique non-automatic variable reference in your program. GCC
6683 will also place floating-point constants in the TOC@. However, only
6684 16,384 entries are available in the TOC@.
6686 If you receive a linker error message that saying you have overflowed
6687 the available TOC space, you can reduce the amount of TOC space used
6688 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
6689 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
6690 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
6691 generate code to calculate the sum of an address and a constant at
6692 run-time instead of putting that sum into the TOC@. You may specify one
6693 or both of these options. Each causes GCC to produce very slightly
6694 slower and larger code at the expense of conserving TOC space.
6696 If you still run out of space in the TOC even when you specify both of
6697 these options, specify @option{-mminimal-toc} instead. This option causes
6698 GCC to make only one TOC entry for every file. When you specify this
6699 option, GCC will produce code that is slower and larger but which
6700 uses extremely little TOC space. You may wish to use this option
6701 only on files that contain less frequently executed code.
6707 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
6708 @code{long} type, and the infrastructure needed to support them.
6709 Specifying @option{-maix64} implies @option{-mpowerpc64} and
6710 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
6711 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
6716 @opindex mno-xl-call
6717 On AIX, pass floating-point arguments to prototyped functions beyond the
6718 register save area (RSA) on the stack in addition to argument FPRs. The
6719 AIX calling convention was extended but not initially documented to
6720 handle an obscure K&R C case of calling a function that takes the
6721 address of its arguments with fewer arguments than declared. AIX XL
6722 compilers access floating point arguments which do not fit in the
6723 RSA from the stack when a subroutine is compiled without
6724 optimization. Because always storing floating-point arguments on the
6725 stack is inefficient and rarely needed, this option is not enabled by
6726 default and only is necessary when calling subroutines compiled by AIX
6727 XL compilers without optimization.
6731 Support @dfn{AIX Threads}. Link an application written to use
6732 @dfn{pthreads} with special libraries and startup code to enable the
6737 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
6738 application written to use message passing with special startup code to
6739 enable the application to run. The system must have PE installed in the
6740 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
6741 must be overridden with the @option{-specs=} option to specify the
6742 appropriate directory location. The Parallel Environment does not
6743 support threads, so the @option{-mpe} option and the @option{-mthreads}
6744 option are incompatible.
6748 @opindex msoft-float
6749 @opindex mhard-float
6750 Generate code that does not use (uses) the floating-point register set.
6751 Software floating point emulation is provided if you use the
6752 @option{-msoft-float} option, and pass the option to GCC when linking.
6755 @itemx -mno-multiple
6757 @opindex mno-multiple
6758 Generate code that uses (does not use) the load multiple word
6759 instructions and the store multiple word instructions. These
6760 instructions are generated by default on POWER systems, and not
6761 generated on PowerPC systems. Do not use @option{-mmultiple} on little
6762 endian PowerPC systems, since those instructions do not work when the
6763 processor is in little endian mode. The exceptions are PPC740 and
6764 PPC750 which permit the instructions usage in little endian mode.
6770 Generate code that uses (does not use) the load string instructions
6771 and the store string word instructions to save multiple registers and
6772 do small block moves. These instructions are generated by default on
6773 POWER systems, and not generated on PowerPC systems. Do not use
6774 @option{-mstring} on little endian PowerPC systems, since those
6775 instructions do not work when the processor is in little endian mode.
6776 The exceptions are PPC740 and PPC750 which permit the instructions
6777 usage in little endian mode.
6783 Generate code that uses (does not use) the load or store instructions
6784 that update the base register to the address of the calculated memory
6785 location. These instructions are generated by default. If you use
6786 @option{-mno-update}, there is a small window between the time that the
6787 stack pointer is updated and the address of the previous frame is
6788 stored, which means code that walks the stack frame across interrupts or
6789 signals may get corrupted data.
6792 @itemx -mno-fused-madd
6793 @opindex mfused-madd
6794 @opindex mno-fused-madd
6795 Generate code that uses (does not use) the floating point multiply and
6796 accumulate instructions. These instructions are generated by default if
6797 hardware floating is used.
6799 @item -mno-bit-align
6801 @opindex mno-bit-align
6803 On System V.4 and embedded PowerPC systems do not (do) force structures
6804 and unions that contain bit-fields to be aligned to the base type of the
6807 For example, by default a structure containing nothing but 8
6808 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
6809 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
6810 the structure would be aligned to a 1 byte boundary and be one byte in
6813 @item -mno-strict-align
6814 @itemx -mstrict-align
6815 @opindex mno-strict-align
6816 @opindex mstrict-align
6817 On System V.4 and embedded PowerPC systems do not (do) assume that
6818 unaligned memory references will be handled by the system.
6821 @itemx -mno-relocatable
6822 @opindex mrelocatable
6823 @opindex mno-relocatable
6824 On embedded PowerPC systems generate code that allows (does not allow)
6825 the program to be relocated to a different address at runtime. If you
6826 use @option{-mrelocatable} on any module, all objects linked together must
6827 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
6829 @item -mrelocatable-lib
6830 @itemx -mno-relocatable-lib
6831 @opindex mrelocatable-lib
6832 @opindex mno-relocatable-lib
6833 On embedded PowerPC systems generate code that allows (does not allow)
6834 the program to be relocated to a different address at runtime. Modules
6835 compiled with @option{-mrelocatable-lib} can be linked with either modules
6836 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
6837 with modules compiled with the @option{-mrelocatable} options.
6843 On System V.4 and embedded PowerPC systems do not (do) assume that
6844 register 2 contains a pointer to a global area pointing to the addresses
6845 used in the program.
6848 @itemx -mlittle-endian
6850 @opindex mlittle-endian
6851 On System V.4 and embedded PowerPC systems compile code for the
6852 processor in little endian mode. The @option{-mlittle-endian} option is
6853 the same as @option{-mlittle}.
6858 @opindex mbig-endian
6859 On System V.4 and embedded PowerPC systems compile code for the
6860 processor in big endian mode. The @option{-mbig-endian} option is
6861 the same as @option{-mbig}.
6865 On System V.4 and embedded PowerPC systems compile code using calling
6866 conventions that adheres to the March 1995 draft of the System V
6867 Application Binary Interface, PowerPC processor supplement. This is the
6868 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
6870 @item -mcall-sysv-eabi
6871 @opindex mcall-sysv-eabi
6872 Specify both @option{-mcall-sysv} and @option{-meabi} options.
6874 @item -mcall-sysv-noeabi
6875 @opindex mcall-sysv-noeabi
6876 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
6880 On System V.4 and embedded PowerPC systems compile code using calling
6881 conventions that are similar to those used on AIX@. This is the
6882 default if you configured GCC using @samp{powerpc-*-eabiaix}.
6884 @item -mcall-solaris
6885 @opindex mcall-solaris
6886 On System V.4 and embedded PowerPC systems compile code for the Solaris
6890 @opindex mcall-linux
6891 On System V.4 and embedded PowerPC systems compile code for the
6892 Linux-based GNU system.
6895 @opindex mcall-netbsd
6896 On System V.4 and embedded PowerPC systems compile code for the
6897 NetBSD operating system.
6900 @itemx -mno-prototype
6902 @opindex mno-prototype
6903 On System V.4 and embedded PowerPC systems assume that all calls to
6904 variable argument functions are properly prototyped. Otherwise, the
6905 compiler must insert an instruction before every non prototyped call to
6906 set or clear bit 6 of the condition code register (@var{CR}) to
6907 indicate whether floating point values were passed in the floating point
6908 registers in case the function takes a variable arguments. With
6909 @option{-mprototype}, only calls to prototyped variable argument functions
6910 will set or clear the bit.
6914 On embedded PowerPC systems, assume that the startup module is called
6915 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
6916 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
6921 On embedded PowerPC systems, assume that the startup module is called
6922 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
6927 On embedded PowerPC systems, assume that the startup module is called
6928 @file{crt0.o} and the standard C libraries are @file{libads.a} and
6932 @opindex myellowknife
6933 On embedded PowerPC systems, assume that the startup module is called
6934 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
6939 On System V.4 and embedded PowerPC systems, specify that you are
6940 compiling for a VxWorks system.
6944 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
6945 header to indicate that @samp{eabi} extended relocations are used.
6951 On System V.4 and embedded PowerPC systems do (do not) adhere to the
6952 Embedded Applications Binary Interface (eabi) which is a set of
6953 modifications to the System V.4 specifications. Selecting @option{-meabi}
6954 means that the stack is aligned to an 8 byte boundary, a function
6955 @code{__eabi} is called to from @code{main} to set up the eabi
6956 environment, and the @option{-msdata} option can use both @code{r2} and
6957 @code{r13} to point to two separate small data areas. Selecting
6958 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
6959 do not call an initialization function from @code{main}, and the
6960 @option{-msdata} option will only use @code{r13} to point to a single
6961 small data area. The @option{-meabi} option is on by default if you
6962 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
6965 @opindex msdata=eabi
6966 On System V.4 and embedded PowerPC systems, put small initialized
6967 @code{const} global and static data in the @samp{.sdata2} section, which
6968 is pointed to by register @code{r2}. Put small initialized
6969 non-@code{const} global and static data in the @samp{.sdata} section,
6970 which is pointed to by register @code{r13}. Put small uninitialized
6971 global and static data in the @samp{.sbss} section, which is adjacent to
6972 the @samp{.sdata} section. The @option{-msdata=eabi} option is
6973 incompatible with the @option{-mrelocatable} option. The
6974 @option{-msdata=eabi} option also sets the @option{-memb} option.
6977 @opindex msdata=sysv
6978 On System V.4 and embedded PowerPC systems, put small global and static
6979 data in the @samp{.sdata} section, which is pointed to by register
6980 @code{r13}. Put small uninitialized global and static data in the
6981 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
6982 The @option{-msdata=sysv} option is incompatible with the
6983 @option{-mrelocatable} option.
6985 @item -msdata=default
6987 @opindex msdata=default
6989 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
6990 compile code the same as @option{-msdata=eabi}, otherwise compile code the
6991 same as @option{-msdata=sysv}.
6994 @opindex msdata-data
6995 On System V.4 and embedded PowerPC systems, put small global and static
6996 data in the @samp{.sdata} section. Put small uninitialized global and
6997 static data in the @samp{.sbss} section. Do not use register @code{r13}
6998 to address small data however. This is the default behavior unless
6999 other @option{-msdata} options are used.
7003 @opindex msdata=none
7005 On embedded PowerPC systems, put all initialized global and static data
7006 in the @samp{.data} section, and all uninitialized data in the
7007 @samp{.bss} section.
7011 @cindex smaller data references (PowerPC)
7012 @cindex .sdata/.sdata2 references (PowerPC)
7013 On embedded PowerPC systems, put global and static items less than or
7014 equal to @var{num} bytes into the small data or bss sections instead of
7015 the normal data or bss section. By default, @var{num} is 8. The
7016 @option{-G @var{num}} switch is also passed to the linker.
7017 All modules should be compiled with the same @option{-G @var{num}} value.
7020 @itemx -mno-regnames
7022 @opindex mno-regnames
7023 On System V.4 and embedded PowerPC systems do (do not) emit register
7024 names in the assembly language output using symbolic forms.
7029 @subsection IBM RT Options
7031 @cindex IBM RT options
7033 These @samp{-m} options are defined for the IBM RT PC:
7037 @opindex min-line-mul
7038 Use an in-line code sequence for integer multiplies. This is the
7041 @item -mcall-lib-mul
7042 @opindex mcall-lib-mul
7043 Call @code{lmul$$} for integer multiples.
7045 @item -mfull-fp-blocks
7046 @opindex mfull-fp-blocks
7047 Generate full-size floating point data blocks, including the minimum
7048 amount of scratch space recommended by IBM@. This is the default.
7050 @item -mminimum-fp-blocks
7051 @opindex mminimum-fp-blocks
7052 Do not include extra scratch space in floating point data blocks. This
7053 results in smaller code, but slower execution, since scratch space must
7054 be allocated dynamically.
7056 @cindex @file{varargs.h} and RT PC
7057 @cindex @file{stdarg.h} and RT PC
7058 @item -mfp-arg-in-fpregs
7059 @opindex mfp-arg-in-fpregs
7060 Use a calling sequence incompatible with the IBM calling convention in
7061 which floating point arguments are passed in floating point registers.
7062 Note that @code{varargs.h} and @code{stdarg.h} will not work with
7063 floating point operands if this option is specified.
7065 @item -mfp-arg-in-gregs
7066 @opindex mfp-arg-in-gregs
7067 Use the normal calling convention for floating point arguments. This is
7070 @item -mhc-struct-return
7071 @opindex mhc-struct-return
7072 Return structures of more than one word in memory, rather than in a
7073 register. This provides compatibility with the MetaWare HighC (hc)
7074 compiler. Use the option @option{-fpcc-struct-return} for compatibility
7075 with the Portable C Compiler (pcc).
7077 @item -mnohc-struct-return
7078 @opindex mnohc-struct-return
7079 Return some structures of more than one word in registers, when
7080 convenient. This is the default. For compatibility with the
7081 IBM-supplied compilers, use the option @option{-fpcc-struct-return} or the
7082 option @option{-mhc-struct-return}.
7086 @subsection MIPS Options
7087 @cindex MIPS options
7089 These @samp{-m} options are defined for the MIPS family of computers:
7093 @item -march=@var{cpu-type}
7095 Assume the defaults for the machine type @var{cpu-type} when generating
7096 instructions. The choices for @var{cpu-type} are @samp{r2000}, @samp{r3000},
7097 @samp{r3900}, @samp{r4000}, @samp{r4100}, @samp{r4300}, @samp{r4400},
7098 @samp{r4600}, @samp{r4650}, @samp{r5000}, @samp{r6000}, @samp{r8000},
7099 and @samp{orion}. Additionally, the @samp{r2000}, @samp{r3000},
7100 @samp{r4000}, @samp{r5000}, and @samp{r6000} can be abbreviated as
7101 @samp{r2k} (or @samp{r2K}), @samp{r3k}, etc.
7103 @item -mtune=@var{cpu-type}
7105 Assume the defaults for the machine type @var{cpu-type} when scheduling
7106 instructions. The choices for @var{cpu-type} are @samp{r2000}, @samp{r3000},
7107 @samp{r3900}, @samp{r4000}, @samp{r4100}, @samp{r4300}, @samp{r4400},
7108 @samp{r4600}, @samp{r4650}, @samp{r5000}, @samp{r6000}, @samp{r8000},
7109 and @samp{orion}. Additionally, the @samp{r2000}, @samp{r3000},
7110 @samp{r4000}, @samp{r5000}, and @samp{r6000} can be abbreviated as
7111 @samp{r2k} (or @samp{r2K}), @samp{r3k}, etc. While picking a specific
7112 @var{cpu-type} will schedule things appropriately for that particular
7113 chip, the compiler will not generate any code that does not meet level 1
7114 of the MIPS ISA (instruction set architecture) without a @option{-mipsX}
7115 or @option{-mabi} switch being used.
7117 @item -mcpu=@var{cpu-type}
7119 This is identical to specifying both @option{-march} and @option{-mtune}.
7123 Issue instructions from level 1 of the MIPS ISA@. This is the default.
7124 @samp{r3000} is the default @var{cpu-type} at this ISA level.
7128 Issue instructions from level 2 of the MIPS ISA (branch likely, square
7129 root instructions). @samp{r6000} is the default @var{cpu-type} at this
7134 Issue instructions from level 3 of the MIPS ISA (64-bit instructions).
7135 @samp{r4000} is the default @var{cpu-type} at this ISA level.
7139 Issue instructions from level 4 of the MIPS ISA (conditional move,
7140 prefetch, enhanced FPU instructions). @samp{r8000} is the default
7141 @var{cpu-type} at this ISA level.
7145 Assume that 32 32-bit floating point registers are available. This is
7150 Assume that 32 64-bit floating point registers are available. This is
7151 the default when the @option{-mips3} option is used.
7154 @itemx -mno-fused-madd
7155 @opindex mfused-madd
7156 @opindex mno-fused-madd
7157 Generate code that uses (does not use) the floating point multiply and
7158 accumulate instructions, when they are available. These instructions
7159 are generated by default if they are available, but this may be
7160 undesirable if the extra precision causes problems or on certain chips
7161 in the mode where denormals are rounded to zero where denormals
7162 generated by multiply and accumulate instructions cause exceptions
7167 Assume that 32 32-bit general purpose registers are available. This is
7172 Assume that 32 64-bit general purpose registers are available. This is
7173 the default when the @option{-mips3} option is used.
7177 Force int and long types to be 64 bits wide. See @option{-mlong32} for an
7178 explanation of the default, and the width of pointers.
7182 Force long types to be 64 bits wide. See @option{-mlong32} for an
7183 explanation of the default, and the width of pointers.
7187 Force long, int, and pointer types to be 32 bits wide.
7189 If none of @option{-mlong32}, @option{-mlong64}, or @option{-mint64} are set,
7190 the size of ints, longs, and pointers depends on the ABI and ISA chosen.
7191 For @option{-mabi=32}, and @option{-mabi=n32}, ints and longs are 32 bits
7192 wide. For @option{-mabi=64}, ints are 32 bits, and longs are 64 bits wide.
7193 For @option{-mabi=eabi} and either @option{-mips1} or @option{-mips2}, ints
7194 and longs are 32 bits wide. For @option{-mabi=eabi} and higher ISAs, ints
7195 are 32 bits, and longs are 64 bits wide. The width of pointer types is
7196 the smaller of the width of longs or the width of general purpose
7197 registers (which in turn depends on the ISA)@.
7209 Generate code for the indicated ABI@. The default instruction level is
7210 @option{-mips1} for @samp{32}, @option{-mips3} for @samp{n32}, and
7211 @option{-mips4} otherwise. Conversely, with @option{-mips1} or
7212 @option{-mips2}, the default ABI is @samp{32}; otherwise, the default ABI
7217 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
7218 add normal debug information. This is the default for all
7219 platforms except for the OSF/1 reference platform, using the OSF/rose
7220 object format. If the either of the @option{-gstabs} or @option{-gstabs+}
7221 switches are used, the @file{mips-tfile} program will encapsulate the
7222 stabs within MIPS ECOFF@.
7226 Generate code for the GNU assembler. This is the default on the OSF/1
7227 reference platform, using the OSF/rose object format. Also, this is
7228 the default if the configure option @option{--with-gnu-as} is used.
7230 @item -msplit-addresses
7231 @itemx -mno-split-addresses
7232 @opindex msplit-addresses
7233 @opindex mno-split-addresses
7234 Generate code to load the high and low parts of address constants separately.
7235 This allows GCC to optimize away redundant loads of the high order
7236 bits of addresses. This optimization requires GNU as and GNU ld.
7237 This optimization is enabled by default for some embedded targets where
7238 GNU as and GNU ld are standard.
7244 The @option{-mrnames} switch says to output code using the MIPS software
7245 names for the registers, instead of the hardware names (ie, @var{a0}
7246 instead of @var{$4}). The only known assembler that supports this option
7247 is the Algorithmics assembler.
7253 The @option{-mgpopt} switch says to write all of the data declarations
7254 before the instructions in the text section, this allows the MIPS
7255 assembler to generate one word memory references instead of using two
7256 words for short global or static data items. This is on by default if
7257 optimization is selected.
7263 For each non-inline function processed, the @option{-mstats} switch
7264 causes the compiler to emit one line to the standard error file to
7265 print statistics about the program (number of registers saved, stack
7272 The @option{-mmemcpy} switch makes all block moves call the appropriate
7273 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
7274 generating inline code.
7277 @itemx -mno-mips-tfile
7278 @opindex mmips-tfile
7279 @opindex mno-mips-tfile
7280 The @option{-mno-mips-tfile} switch causes the compiler not
7281 postprocess the object file with the @file{mips-tfile} program,
7282 after the MIPS assembler has generated it to add debug support. If
7283 @file{mips-tfile} is not run, then no local variables will be
7284 available to the debugger. In addition, @file{stage2} and
7285 @file{stage3} objects will have the temporary file names passed to the
7286 assembler embedded in the object file, which means the objects will
7287 not compare the same. The @option{-mno-mips-tfile} switch should only
7288 be used when there are bugs in the @file{mips-tfile} program that
7289 prevents compilation.
7292 @opindex msoft-float
7293 Generate output containing library calls for floating point.
7294 @strong{Warning:} the requisite libraries are not part of GCC@.
7295 Normally the facilities of the machine's usual C compiler are used, but
7296 this can't be done directly in cross-compilation. You must make your
7297 own arrangements to provide suitable library functions for
7301 @opindex mhard-float
7302 Generate output containing floating point instructions. This is the
7303 default if you use the unmodified sources.
7306 @itemx -mno-abicalls
7308 @opindex mno-abicalls
7309 Emit (or do not emit) the pseudo operations @samp{.abicalls},
7310 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
7311 position independent code.
7314 @itemx -mno-long-calls
7315 @opindex mlong-calls
7316 @opindex mno-long-calls
7317 Do all calls with the @samp{JALR} instruction, which requires
7318 loading up a function's address into a register before the call.
7319 You need to use this switch, if you call outside of the current
7320 512 megabyte segment to functions that are not through pointers.
7323 @itemx -mno-half-pic
7325 @opindex mno-half-pic
7326 Put pointers to extern references into the data section and load them
7327 up, rather than put the references in the text section.
7329 @item -membedded-pic
7330 @itemx -mno-embedded-pic
7331 @opindex membedded-pic
7332 @opindex mno-embedded-pic
7333 Generate PIC code suitable for some embedded systems. All calls are
7334 made using PC relative address, and all data is addressed using the $gp
7335 register. No more than 65536 bytes of global data may be used. This
7336 requires GNU as and GNU ld which do most of the work. This currently
7337 only works on targets which use ECOFF; it does not work with ELF@.
7339 @item -membedded-data
7340 @itemx -mno-embedded-data
7341 @opindex membedded-data
7342 @opindex mno-embedded-data
7343 Allocate variables to the read-only data section first if possible, then
7344 next in the small data section if possible, otherwise in data. This gives
7345 slightly slower code than the default, but reduces the amount of RAM required
7346 when executing, and thus may be preferred for some embedded systems.
7348 @item -muninit-const-in-rodata
7349 @itemx -mno-uninit-const-in-rodata
7350 @opindex muninit-const-in-rodata
7351 @opindex mno-uninit-const-in-rodata
7352 When used together with @option{-membedded-data}, it will always store uninitialized
7353 const variables in the read-only data section.
7355 @item -msingle-float
7356 @itemx -mdouble-float
7357 @opindex msingle-float
7358 @opindex mdouble-float
7359 The @option{-msingle-float} switch tells gcc to assume that the floating
7360 point coprocessor only supports single precision operations, as on the
7361 @samp{r4650} chip. The @option{-mdouble-float} switch permits gcc to use
7362 double precision operations. This is the default.
7368 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
7369 as on the @samp{r4650} chip.
7373 Turns on @option{-msingle-float}, @option{-mmad}, and, at least for now,
7374 @option{-mcpu=r4650}.
7380 Enable 16-bit instructions.
7384 Use the entry and exit pseudo ops. This option can only be used with
7389 Compile code for the processor in little endian mode.
7390 The requisite libraries are assumed to exist.
7394 Compile code for the processor in big endian mode.
7395 The requisite libraries are assumed to exist.
7399 @cindex smaller data references (MIPS)
7400 @cindex gp-relative references (MIPS)
7401 Put global and static items less than or equal to @var{num} bytes into
7402 the small data or bss sections instead of the normal data or bss
7403 section. This allows the assembler to emit one word memory reference
7404 instructions based on the global pointer (@var{gp} or @var{$28}),
7405 instead of the normal two words used. By default, @var{num} is 8 when
7406 the MIPS assembler is used, and 0 when the GNU assembler is used. The
7407 @option{-G @var{num}} switch is also passed to the assembler and linker.
7408 All modules should be compiled with the same @option{-G @var{num}}
7413 Tell the MIPS assembler to not run its preprocessor over user
7414 assembler files (with a @samp{.s} suffix) when assembling them.
7418 Pass an option to gas which will cause nops to be inserted if
7419 the read of the destination register of an mfhi or mflo instruction
7420 occurs in the following two instructions.
7424 Do not include the default crt0.
7428 These options are defined by the macro
7429 @code{TARGET_SWITCHES} in the machine description. The default for the
7430 options is also defined by that macro, which enables you to change the
7434 @node i386 and x86-64 Options
7435 @subsection Intel 386 and AMD x86-64 Options
7436 @cindex i386 Options
7437 @cindex x86-64 Options
7438 @cindex Intel 386 Options
7439 @cindex AMD x86-64 Options
7441 These @samp{-m} options are defined for the i386 and x86-64 family of
7445 @item -mcpu=@var{cpu-type}
7447 Assume the defaults for the machine type @var{cpu-type} when scheduling
7448 instructions. The choices for @var{cpu-type} are @samp{i386},
7449 @samp{i486}, @samp{i586}, @samp{i686}, @samp{pentium},
7450 @samp{pentiumpro}, @samp{pentium4}, @samp{k6}, and @samp{athlon}
7452 While picking a specific @var{cpu-type} will schedule things appropriately
7453 for that particular chip, the compiler will not generate any code that
7454 does not run on the i386 without the @option{-march=@var{cpu-type}} option
7455 being used. @samp{i586} is equivalent to @samp{pentium} and @samp{i686}
7456 is equivalent to @samp{pentiumpro}. @samp{k6} and @samp{athlon} are the
7457 AMD chips as opposed to the Intel ones.
7459 @item -march=@var{cpu-type}
7461 Generate instructions for the machine type @var{cpu-type}. The choices
7462 for @var{cpu-type} are the same as for @option{-mcpu}. Moreover,
7463 specifying @option{-march=@var{cpu-type}} implies @option{-mcpu=@var{cpu-type}}.
7472 @opindex mpentiumpro
7473 These options are synonyms for @option{-mcpu=i386}, @option{-mcpu=i486},
7474 @option{-mcpu=pentium}, and @option{-mcpu=pentiumpro} respectively.
7475 These synonyms are deprecated.
7477 @item -mintel-syntax
7478 @opindex mintel-syntax
7479 Emit assembly using Intel syntax opcodes instead of AT&T syntax.
7484 @opindex mno-ieee-fp
7485 Control whether or not the compiler uses IEEE floating point
7486 comparisons. These handle correctly the case where the result of a
7487 comparison is unordered.
7490 @opindex msoft-float
7491 Generate output containing library calls for floating point.
7492 @strong{Warning:} the requisite libraries are not part of GCC@.
7493 Normally the facilities of the machine's usual C compiler are used, but
7494 this can't be done directly in cross-compilation. You must make your
7495 own arrangements to provide suitable library functions for
7498 On machines where a function returns floating point results in the 80387
7499 register stack, some floating point opcodes may be emitted even if
7500 @option{-msoft-float} is used.
7502 @item -mno-fp-ret-in-387
7503 @opindex mno-fp-ret-in-387
7504 Do not use the FPU registers for return values of functions.
7506 The usual calling convention has functions return values of types
7507 @code{float} and @code{double} in an FPU register, even if there
7508 is no FPU@. The idea is that the operating system should emulate
7511 The option @option{-mno-fp-ret-in-387} causes such values to be returned
7512 in ordinary CPU registers instead.
7514 @item -mno-fancy-math-387
7515 @opindex mno-fancy-math-387
7516 Some 387 emulators do not support the @code{sin}, @code{cos} and
7517 @code{sqrt} instructions for the 387. Specify this option to avoid
7518 generating those instructions. This option is the default on FreeBSD@.
7519 As of revision 2.6.1, these instructions are not generated unless you
7520 also use the @option{-funsafe-math-optimizations} switch.
7522 @item -malign-double
7523 @itemx -mno-align-double
7524 @opindex malign-double
7525 @opindex mno-align-double
7526 Control whether GCC aligns @code{double}, @code{long double}, and
7527 @code{long long} variables on a two word boundary or a one word
7528 boundary. Aligning @code{double} variables on a two word boundary will
7529 produce code that runs somewhat faster on a @samp{Pentium} at the
7530 expense of more memory.
7532 @item -m128bit-long-double
7533 @opindex m128bit-long-double
7534 Control the size of @code{long double} type. i386 application binary interface
7535 specify the size to be 12 bytes, while modern architectures (Pentium and newer)
7536 prefer @code{long double} aligned to 8 or 16 byte boundary. This is
7537 impossible to reach with 12 byte long doubles in the array accesses.
7539 @strong{Warning:} if you use the @option{-m128bit-long-double} switch, the
7540 structures and arrays containing @code{long double} will change their size as
7541 well as function calling convention for function taking @code{long double}
7544 @item -m96bit-long-double
7545 @opindex m96bit-long-double
7546 Set the size of @code{long double} to 96 bits as required by the i386
7547 application binary interface. This is the default.
7550 @itemx -mno-svr3-shlib
7551 @opindex msvr3-shlib
7552 @opindex mno-svr3-shlib
7553 Control whether GCC places uninitialized local variables into the
7554 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
7555 into @code{bss}. These options are meaningful only on System V Release 3.
7559 Use a different function-calling convention, in which functions that
7560 take a fixed number of arguments return with the @code{ret} @var{num}
7561 instruction, which pops their arguments while returning. This saves one
7562 instruction in the caller since there is no need to pop the arguments
7565 You can specify that an individual function is called with this calling
7566 sequence with the function attribute @samp{stdcall}. You can also
7567 override the @option{-mrtd} option by using the function attribute
7568 @samp{cdecl}. @xref{Function Attributes}.
7570 @strong{Warning:} this calling convention is incompatible with the one
7571 normally used on Unix, so you cannot use it if you need to call
7572 libraries compiled with the Unix compiler.
7574 Also, you must provide function prototypes for all functions that
7575 take variable numbers of arguments (including @code{printf});
7576 otherwise incorrect code will be generated for calls to those
7579 In addition, seriously incorrect code will result if you call a
7580 function with too many arguments. (Normally, extra arguments are
7581 harmlessly ignored.)
7583 @item -mregparm=@var{num}
7585 Control how many registers are used to pass integer arguments. By
7586 default, no registers are used to pass arguments, and at most 3
7587 registers can be used. You can control this behavior for a specific
7588 function by using the function attribute @samp{regparm}.
7589 @xref{Function Attributes}.
7591 @strong{Warning:} if you use this switch, and
7592 @var{num} is nonzero, then you must build all modules with the same
7593 value, including any libraries. This includes the system libraries and
7596 @item -mpreferred-stack-boundary=@var{num}
7597 @opindex mpreferred-stack-boundary
7598 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
7599 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
7600 the default is 4 (16 bytes or 128 bits), except when optimizing for code
7601 size (@option{-Os}), in which case the default is the minimum correct
7602 alignment (4 bytes for x86, and 8 bytes for x86-64).
7604 On Pentium and PentiumPro, @code{double} and @code{long double} values
7605 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
7606 suffer significant run time performance penalties. On Pentium III, the
7607 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
7608 penalties if it is not 16 byte aligned.
7610 To ensure proper alignment of this values on the stack, the stack boundary
7611 must be as aligned as that required by any value stored on the stack.
7612 Further, every function must be generated such that it keeps the stack
7613 aligned. Thus calling a function compiled with a higher preferred
7614 stack boundary from a function compiled with a lower preferred stack
7615 boundary will most likely misalign the stack. It is recommended that
7616 libraries that use callbacks always use the default setting.
7618 This extra alignment does consume extra stack space, and generally
7619 increases code size. Code that is sensitive to stack space usage, such
7620 as embedded systems and operating system kernels, may want to reduce the
7621 preferred alignment to @option{-mpreferred-stack-boundary=2}.
7635 These switches enable or disable the use of built-in functions that allow
7636 direct access to the MMX, SSE and 3Dnow extensions of the instruction set.
7638 The following machine modes are available for use with MMX built-in functions
7639 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32 bit integers,
7640 @code{V4HI} for a vector of four 16 bit integers, and @code{V8QI} for a
7641 vector of eight 8 bit integers. Some of the built-in functions operate on
7642 MMX registers as a whole 64 bit entity, these use @code{DI} as their mode.
7644 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
7645 of two 32 bit floating point values.
7647 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32 bit
7648 floating point values. Some instructions use a vector of four 32 bit
7649 integers, these use @code{V4SI}. Finally, some instructions operate on an
7650 entire vector register, interpreting it as a 128 bit integer, these use mode
7653 The following built-in functions are made available by @option{-mmmx}:
7655 @item v8qi __builtin_ia32_paddb (v8qi, v8qi)
7656 Generates the @code{paddb} machine instruction.
7657 @item v4hi __builtin_ia32_paddw (v4hi, v4hi)
7658 Generates the @code{paddw} machine instruction.
7659 @item v2si __builtin_ia32_paddd (v2si, v2si)
7660 Generates the @code{paddd} machine instruction.
7661 @item v8qi __builtin_ia32_psubb (v8qi, v8qi)
7662 Generates the @code{psubb} machine instruction.
7663 @item v4hi __builtin_ia32_psubw (v4hi, v4hi)
7664 Generates the @code{psubw} machine instruction.
7665 @item v2si __builtin_ia32_psubd (v2si, v2si)
7666 Generates the @code{psubd} machine instruction.
7668 @item v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7669 Generates the @code{paddsb} machine instruction.
7670 @item v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7671 Generates the @code{paddsw} machine instruction.
7672 @item v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7673 Generates the @code{psubsb} machine instruction.
7674 @item v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7675 Generates the @code{psubsw} machine instruction.
7677 @item v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7678 Generates the @code{paddusb} machine instruction.
7679 @item v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7680 Generates the @code{paddusw} machine instruction.
7681 @item v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7682 Generates the @code{psubusb} machine instruction.
7683 @item v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7684 Generates the @code{psubusw} machine instruction.
7686 @item v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7687 Generates the @code{pmullw} machine instruction.
7688 @item v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7689 Generates the @code{pmulhw} machine instruction.
7691 @item di __builtin_ia32_pand (di, di)
7692 Generates the @code{pand} machine instruction.
7693 @item di __builtin_ia32_pandn (di,di)
7694 Generates the @code{pandn} machine instruction.
7695 @item di __builtin_ia32_por (di, di)
7696 Generates the @code{por} machine instruction.
7697 @item di __builtin_ia32_pxor (di, di)
7698 Generates the @code{pxor} machine instruction.
7700 @item v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7701 Generates the @code{pcmpeqb} machine instruction.
7702 @item v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7703 Generates the @code{pcmpeqw} machine instruction.
7704 @item v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7705 Generates the @code{pcmpeqd} machine instruction.
7706 @item v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7707 Generates the @code{pcmpgtb} machine instruction.
7708 @item v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7709 Generates the @code{pcmpgtw} machine instruction.
7710 @item v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7711 Generates the @code{pcmpgtd} machine instruction.
7713 @item v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7714 Generates the @code{punpckhbw} machine instruction.
7715 @item v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7716 Generates the @code{punpckhwd} machine instruction.
7717 @item v2si __builtin_ia32_punpckhdq (v2si, v2si)
7718 Generates the @code{punpckhdq} machine instruction.
7719 @item v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7720 Generates the @code{punpcklbw} machine instruction.
7721 @item v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7722 Generates the @code{punpcklwd} machine instruction.
7723 @item v2si __builtin_ia32_punpckldq (v2si, v2si)
7724 Generates the @code{punpckldq} machine instruction.
7726 @item v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7727 Generates the @code{packsswb} machine instruction.
7728 @item v4hi __builtin_ia32_packssdw (v2si, v2si)
7729 Generates the @code{packssdw} machine instruction.
7730 @item v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7731 Generates the @code{packuswb} machine instruction.
7735 The following built-in functions are made available either with @option{-msse}, or
7736 with a combination of @option{-m3dnow} and @option{-march=athlon}.
7739 @item v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7740 Generates the @code{pmulhuw} machine instruction.
7742 @item v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7743 Generates the @code{pavgb} machine instruction.
7744 @item v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7745 Generates the @code{pavgw} machine instruction.
7746 @item v4hi __builtin_ia32_psadbw (v8qi, v8qi)
7747 Generates the @code{psadbw} machine instruction.
7749 @item v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7750 Generates the @code{pmaxub} machine instruction.
7751 @item v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7752 Generates the @code{pmaxsw} machine instruction.
7753 @item v8qi __builtin_ia32_pminub (v8qi, v8qi)
7754 Generates the @code{pminub} machine instruction.
7755 @item v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7756 Generates the @code{pminsw} machine instruction.
7758 @item int __builtin_ia32_pextrw (v4hi, int)
7759 Generates the @code{pextrw} machine instruction.
7760 @item v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7761 Generates the @code{pinsrw} machine instruction.
7763 @item int __builtin_ia32_pmovmskb (v8qi)
7764 Generates the @code{pmovmskb} machine instruction.
7765 @item void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7766 Generates the @code{maskmovq} machine instruction.
7767 @item void __builtin_ia32_movntq (di *, di)
7768 Generates the @code{movntq} machine instruction.
7769 @item void __builtin_ia32_sfence (void)
7770 Generates the @code{sfence} machine instruction.
7771 @item void __builtin_ia32_prefetch (char *, int selector)
7772 Generates a prefetch machine instruction, depending on the value of
7773 selector. If @code{selector} is 0, it generates @code{prefetchnta}; for
7774 a value of 1, it generates @code{prefetcht0}; for a value of 2, it generates
7775 @code{prefetcht1}; and for a value of 3 it generates @code{prefetcht2}.
7779 The following built-in functions are available when @option{-msse} is used.
7782 @item int __builtin_ia32_comieq (v4sf, v4sf)
7783 Generates the @code{comiss} machine instruction and performs an equality
7784 comparison. The return value is the truth value of that comparison.
7785 @item int __builtin_ia32_comineq (v4sf, v4sf)
7786 Generates the @code{comiss} machine instruction and performs an inequality
7787 comparison. The return value is the truth value of that comparison.
7788 @item int __builtin_ia32_comilt (v4sf, v4sf)
7789 Generates the @code{comiss} machine instruction and performs a ``less than''
7790 comparison. The return value is the truth value of that comparison.
7791 @item int __builtin_ia32_comile (v4sf, v4sf)
7792 Generates the @code{comiss} machine instruction and performs a ``less or
7793 equal'' comparison. The return value is the truth value of that comparison.
7794 @item int __builtin_ia32_comigt (v4sf, v4sf)
7795 Generates the @code{comiss} machine instruction and performs a ``greater than''
7796 comparison. The return value is the truth value of that comparison.
7797 @item int __builtin_ia32_comige (v4sf, v4sf)
7798 Generates the @code{comiss} machine instruction and performs a ``greater or
7799 equal'' comparison. The return value is the truth value of that comparison.
7801 @item int __builtin_ia32_ucomieq (v4sf, v4sf)
7802 Generates the @code{ucomiss} machine instruction and performs an equality
7803 comparison. The return value is the truth value of that comparison.
7804 @item int __builtin_ia32_ucomineq (v4sf, v4sf)
7805 Generates the @code{ucomiss} machine instruction and performs an inequality
7806 comparison. The return value is the truth value of that comparison.
7807 @item int __builtin_ia32_ucomilt (v4sf, v4sf)
7808 Generates the @code{ucomiss} machine instruction and performs a ``less than''
7809 comparison. The return value is the truth value of that comparison.
7810 @item int __builtin_ia32_ucomile (v4sf, v4sf)
7811 Generates the @code{ucomiss} machine instruction and performs a ``less or
7812 equal'' comparison. The return value is the truth value of that comparison.
7813 @item int __builtin_ia32_ucomigt (v4sf, v4sf)
7814 Generates the @code{ucomiss} machine instruction and performs a ``greater than''
7815 comparison. The return value is the truth value of that comparison.
7816 @item int __builtin_ia32_ucomige (v4sf, v4sf)
7817 Generates the @code{ucomiss} machine instruction and performs a ``greater or
7818 equal'' comparison. The return value is the truth value of that comparison.
7820 @item v4sf __builtin_ia32_addps (v4sf, v4sf)
7821 Generates the @code{addps} machine instruction.
7822 @item v4sf __builtin_ia32_addss (v4sf, v4sf)
7823 Generates the @code{addss} machine instruction.
7824 @item v4sf __builtin_ia32_subps (v4sf, v4sf)
7825 Generates the @code{subps} machine instruction.
7826 @item v4sf __builtin_ia32_subss (v4sf, v4sf)
7827 Generates the @code{subss} machine instruction.
7828 @item v4sf __builtin_ia32_mulps (v4sf, v4sf)
7829 Generates the @code{mulps} machine instruction.
7830 @item v4sf __builtin_ia32_mulss (v4sf, v4sf)
7831 Generates the @code{mulss} machine instruction.
7832 @item v4sf __builtin_ia32_divps (v4sf, v4sf)
7833 Generates the @code{divps} machine instruction.
7834 @item v4sf __builtin_ia32_divss (v4sf, v4sf)
7835 Generates the @code{divss} machine instruction.
7837 @item v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7838 Generates the @code{cmpeqps} machine instruction.
7839 @item v4si __builtin_ia32_cmplts (v4sf, v4sf)
7840 Generates the @code{cmpltps} machine instruction.
7841 @item v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7842 Generates the @code{cmpleps} machine instruction.
7843 @item v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7844 Generates the @code{cmpgtps} machine instruction.
7845 @item v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7846 Generates the @code{cmpgeps} machine instruction.
7847 @item v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7848 Generates the @code{cmpunodps} machine instruction.
7849 @item v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7850 Generates the @code{cmpeqps} machine instruction.
7851 @item v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7852 Generates the @code{cmpltps} machine instruction.
7853 @item v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7854 Generates the @code{cmpleps} machine instruction.
7855 @item v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7856 Generates the @code{cmpgtps} machine instruction.
7857 @item v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7858 Generates the @code{cmpgeps} machine instruction.
7859 @item v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7860 Generates the @code{cmpunodps} machine instruction.
7862 @item v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7863 Generates the @code{cmpeqss} machine instruction.
7864 @item v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7865 Generates the @code{cmpltss} machine instruction.
7866 @item v4si __builtin_ia32_cmpless (v4sf, v4sf)
7867 Generates the @code{cmpless} machine instruction.
7868 @item v4si __builtin_ia32_cmpgtss (v4sf, v4sf)
7869 Generates the @code{cmpgtss} machine instruction.
7870 @item v4si __builtin_ia32_cmpgess (v4sf, v4sf)
7871 Generates the @code{cmpgess} machine instruction.
7872 @item v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7873 Generates the @code{cmpunodss} machine instruction.
7874 @item v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7875 Generates the @code{cmpeqss} machine instruction.
7876 @item v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7877 Generates the @code{cmpltss} machine instruction.
7878 @item v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7879 Generates the @code{cmpless} machine instruction.
7880 @item v4si __builtin_ia32_cmpngtss (v4sf, v4sf)
7881 Generates the @code{cmpgtss} machine instruction.
7882 @item v4si __builtin_ia32_cmpngess (v4sf, v4sf)
7883 Generates the @code{cmpgess} machine instruction.
7884 @item v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7885 Generates the @code{cmpunodss} machine instruction.
7887 @item v4sf __builtin_ia32_maxps (v4sf, v4sf)
7888 Generates the @code{maxps} machine instruction.
7889 @item v4sf __builtin_ia32_maxsss (v4sf, v4sf)
7890 Generates the @code{maxss} machine instruction.
7891 @item v4sf __builtin_ia32_minps (v4sf, v4sf)
7892 Generates the @code{minps} machine instruction.
7893 @item v4sf __builtin_ia32_minsss (v4sf, v4sf)
7894 Generates the @code{minss} machine instruction.
7896 @item ti __builtin_ia32_andps (ti, ti)
7897 Generates the @code{andps} machine instruction.
7898 @item ti __builtin_ia32_andnps (ti, ti)
7899 Generates the @code{andnps} machine instruction.
7900 @item ti __builtin_ia32_orps (ti, ti)
7901 Generates the @code{orps} machine instruction.
7902 @item ti __builtin_ia32_xorps (ti, ti)
7903 Generates the @code{xorps} machine instruction.
7905 @item v4sf __builtin_ia32_movps (v4sf, v4sf)
7906 Generates the @code{movps} machine instruction.
7907 @item v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7908 Generates the @code{movhlps} machine instruction.
7909 @item v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7910 Generates the @code{movlhps} machine instruction.
7911 @item v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7912 Generates the @code{unpckhps} machine instruction.
7913 @item v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7914 Generates the @code{unpcklps} machine instruction.
7916 @item v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7917 Generates the @code{cvtpi2ps} machine instruction.
7918 @item v2si __builtin_ia32_cvtps2pi (v4sf)
7919 Generates the @code{cvtps2pi} machine instruction.
7920 @item v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7921 Generates the @code{cvtsi2ss} machine instruction.
7922 @item int __builtin_ia32_cvtss2si (v4sf)
7923 Generates the @code{cvtsi2ss} machine instruction.
7924 @item v2si __builtin_ia32_cvttps2pi (v4sf)
7925 Generates the @code{cvttps2pi} machine instruction.
7926 @item int __builtin_ia32_cvttss2si (v4sf)
7927 Generates the @code{cvttsi2ss} machine instruction.
7929 @item v4sf __builtin_ia32_rcpps (v4sf)
7930 Generates the @code{rcpps} machine instruction.
7931 @item v4sf __builtin_ia32_rsqrtps (v4sf)
7932 Generates the @code{rsqrtps} machine instruction.
7933 @item v4sf __builtin_ia32_sqrtps (v4sf)
7934 Generates the @code{sqrtps} machine instruction.
7935 @item v4sf __builtin_ia32_rcpss (v4sf)
7936 Generates the @code{rcpss} machine instruction.
7937 @item v4sf __builtin_ia32_rsqrtss (v4sf)
7938 Generates the @code{rsqrtss} machine instruction.
7939 @item v4sf __builtin_ia32_sqrtss (v4sf)
7940 Generates the @code{sqrtss} machine instruction.
7942 @item v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7943 Generates the @code{shufps} machine instruction.
7945 @item v4sf __builtin_ia32_loadaps (float *)
7946 Generates the @code{movaps} machine instruction as a load from memory.
7947 @item void __builtin_ia32_storeaps (float *, v4sf)
7948 Generates the @code{movaps} machine instruction as a store to memory.
7949 @item v4sf __builtin_ia32_loadups (float *)
7950 Generates the @code{movups} machine instruction as a load from memory.
7951 @item void __builtin_ia32_storeups (float *, v4sf)
7952 Generates the @code{movups} machine instruction as a store to memory.
7953 @item v4sf __builtin_ia32_loadsss (float *)
7954 Generates the @code{movss} machine instruction as a load from memory.
7955 @item void __builtin_ia32_storess (float *, v4sf)
7956 Generates the @code{movss} machine instruction as a store to memory.
7958 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7959 Generates the @code{movhps} machine instruction as a load from memory.
7960 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7961 Generates the @code{movlps} machine instruction as a load from memory
7962 @item void __builtin_ia32_storehps (v4sf, v2si *)
7963 Generates the @code{movhps} machine instruction as a store to memory.
7964 @item void __builtin_ia32_storelps (v4sf, v2si *)
7965 Generates the @code{movlps} machine instruction as a store to memory.
7967 @item void __builtin_ia32_movntps (float *, v4sf)
7968 Generates the @code{movntps} machine instruction.
7969 @item int __builtin_ia32_movmskps (v4sf)
7970 Generates the @code{movntps} machine instruction.
7972 @item void __builtin_ia32_storeps1 (float *, v4sf)
7973 Generates the @code{movaps} machine instruction as a store to memory.
7974 Before storing, the value is modified with a @code{shufps} instruction
7975 so that the lowest of the four floating point elements is replicated
7976 across the entire vector that is stored.
7977 @item void __builtin_ia32_storerps (float *, v4sf)
7978 Generates the @code{movaps} machine instruction as a store to memory.
7979 Before storing, the value is modified with a @code{shufps} instruction
7980 so that the order of the four floating point elements in the vector is
7982 @item v4sf __builtin_ia32_loadps1 (float *)
7983 Generates a @code{movss} machine instruction to load a floating point
7984 value from memory, and a @code{shufps} instruction to replicate the
7985 loaded value across all four elements of the result vector.
7986 @item v4sf __builtin_ia32_loadrps (float *)
7987 Generates a @code{movaps} machine instruction to load a vector from
7988 memory, and a @code{shufps} instruction to reverse the order of the
7989 four floating point elements in the result vector.
7990 @item v4sf __builtin_ia32_setps (float, float, float, float)
7991 Constructs a vector from four single floating point values. The return
7992 value is equal to the value that would result from storing the four
7993 arguments into consecutive memory locations and then executing a
7994 @code{movaps} to load the vector from memory.
7995 @item v4sf __builtin_ia32_setps1 (float)
7996 Constructs a vector from a single floating point value by replicating
7997 it across all four elements of the result vector.
8001 @itemx -mno-push-args
8003 @opindex mno-push-args
8004 Use PUSH operations to store outgoing parameters. This method is shorter
8005 and usually equally fast as method using SUB/MOV operations and is enabled
8006 by default. In some cases disabling it may improve performance because of
8007 improved scheduling and reduced dependencies.
8009 @item -maccumulate-outgoing-args
8010 @opindex maccumulate-outgoing-args
8011 If enabled, the maximum amount of space required for outgoing arguments will be
8012 computed in the function prologue. This is faster on most modern CPUs
8013 because of reduced dependencies, improved scheduling and reduced stack usage
8014 when preferred stack boundary is not equal to 2. The drawback is a notable
8015 increase in code size. This switch implies @option{-mno-push-args}.
8019 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8020 on thread-safe exception handling must compile and link all code with the
8021 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8022 @option{-D_MT}; when linking, it links in a special thread helper library
8023 @option{-lmingwthrd} which cleans up per thread exception handling data.
8025 @item -mno-align-stringops
8026 @opindex mno-align-stringops
8027 Do not align destination of inlined string operations. This switch reduces
8028 code size and improves performance in case the destination is already aligned,
8029 but gcc don't know about it.
8031 @item -minline-all-stringops
8032 @opindex minline-all-stringops
8033 By default GCC inlines string operations only when destination is known to be
8034 aligned at least to 4 byte boundary. This enables more inlining, increase code
8035 size, but may improve performance of code that depends on fast memcpy, strlen
8036 and memset for short lengths.
8038 @item -momit-leaf-frame-pointer
8039 @opindex momit-leaf-frame-pointer
8040 Don't keep the frame pointer in a register for leaf functions. This
8041 avoids the instructions to save, set up and restore frame pointers and
8042 makes an extra register available in leaf functions. The option
8043 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8044 which might make debugging harder.
8047 These @samp{-m} switches are supported in addition to the above
8048 on AMD x86-64 processors in 64-bit environments.
8055 Generate code for a 32-bit or 64-bit environment.
8056 The 32-bit environment sets int, long and pointer to 32 bits and
8057 generates code that runs on any i386 system.
8058 The 64-bit environment sets int to 32 bits and long and pointer
8059 to 64 bits and generates code for AMD's x86-64 architecture.
8062 @opindex no-red-zone
8063 Do not use a so called red zone for x86-64 code. The red zone is mandated
8064 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8065 stack pointer that will not be modified by signal or interrupt handlers
8066 and therefore can be used for temporary data without adjusting the stack
8067 pointer. The flag @option{-mno-red-zone} disables this red zone.
8071 @subsection HPPA Options
8072 @cindex HPPA Options
8074 These @samp{-m} options are defined for the HPPA family of computers:
8077 @item -march=@var{architecture-type}
8079 Generate code for the specified architecture. The choices for
8080 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8081 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8082 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8083 architecture option for your machine. Code compiled for lower numbered
8084 architectures will run on higher numbered architectures, but not the
8087 PA 2.0 support currently requires gas snapshot 19990413 or later. The
8088 next release of binutils (current is 2.9.1) will probably contain PA 2.0
8092 @itemx -mpa-risc-1-1
8093 @itemx -mpa-risc-2-0
8094 @opindex mpa-risc-1-0
8095 @opindex mpa-risc-1-1
8096 @opindex mpa-risc-2-0
8097 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8100 @opindex mbig-switch
8101 Generate code suitable for big switch tables. Use this option only if
8102 the assembler/linker complain about out of range branches within a switch
8105 @item -mjump-in-delay
8106 @opindex mjump-in-delay
8107 Fill delay slots of function calls with unconditional jump instructions
8108 by modifying the return pointer for the function call to be the target
8109 of the conditional jump.
8111 @item -mdisable-fpregs
8112 @opindex mdisable-fpregs
8113 Prevent floating point registers from being used in any manner. This is
8114 necessary for compiling kernels which perform lazy context switching of
8115 floating point registers. If you use this option and attempt to perform
8116 floating point operations, the compiler will abort.
8118 @item -mdisable-indexing
8119 @opindex mdisable-indexing
8120 Prevent the compiler from using indexing address modes. This avoids some
8121 rather obscure problems when compiling MIG generated code under MACH@.
8123 @item -mno-space-regs
8124 @opindex mno-space-regs
8125 Generate code that assumes the target has no space registers. This allows
8126 GCC to generate faster indirect calls and use unscaled index address modes.
8128 Such code is suitable for level 0 PA systems and kernels.
8130 @item -mfast-indirect-calls
8131 @opindex mfast-indirect-calls
8132 Generate code that assumes calls never cross space boundaries. This
8133 allows GCC to emit code which performs faster indirect calls.
8135 This option will not work in the presence of shared libraries or nested
8138 @item -mlong-load-store
8139 @opindex mlong-load-store
8140 Generate 3-instruction load and store sequences as sometimes required by
8141 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8144 @item -mportable-runtime
8145 @opindex mportable-runtime
8146 Use the portable calling conventions proposed by HP for ELF systems.
8150 Enable the use of assembler directives only GAS understands.
8152 @item -mschedule=@var{cpu-type}
8154 Schedule code according to the constraints for the machine type
8155 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8156 @samp{7100}, @samp{7100LC}, @samp{7200}, and @samp{8000}. Refer to
8157 @file{/usr/lib/sched.models} on an HP-UX system to determine the
8158 proper scheduling option for your machine.
8161 @opindex mlinker-opt
8162 Enable the optimization pass in the HPUX linker. Note this makes symbolic
8163 debugging impossible. It also triggers a bug in the HPUX 8 and HPUX 9 linkers
8164 in which they give bogus error messages when linking some programs.
8167 @opindex msoft-float
8168 Generate output containing library calls for floating point.
8169 @strong{Warning:} the requisite libraries are not available for all HPPA
8170 targets. Normally the facilities of the machine's usual C compiler are
8171 used, but this cannot be done directly in cross-compilation. You must make
8172 your own arrangements to provide suitable library functions for
8173 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8174 does provide software floating point support.
8176 @option{-msoft-float} changes the calling convention in the output file;
8177 therefore, it is only useful if you compile @emph{all} of a program with
8178 this option. In particular, you need to compile @file{libgcc.a}, the
8179 library that comes with GCC, with @option{-msoft-float} in order for
8183 @node Intel 960 Options
8184 @subsection Intel 960 Options
8186 These @samp{-m} options are defined for the Intel 960 implementations:
8189 @item -m@var{cpu-type}
8197 Assume the defaults for the machine type @var{cpu-type} for some of
8198 the other options, including instruction scheduling, floating point
8199 support, and addressing modes. The choices for @var{cpu-type} are
8200 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
8201 @samp{sa}, and @samp{sb}.
8208 @opindex msoft-float
8209 The @option{-mnumerics} option indicates that the processor does support
8210 floating-point instructions. The @option{-msoft-float} option indicates
8211 that floating-point support should not be assumed.
8213 @item -mleaf-procedures
8214 @itemx -mno-leaf-procedures
8215 @opindex mleaf-procedures
8216 @opindex mno-leaf-procedures
8217 Do (or do not) attempt to alter leaf procedures to be callable with the
8218 @code{bal} instruction as well as @code{call}. This will result in more
8219 efficient code for explicit calls when the @code{bal} instruction can be
8220 substituted by the assembler or linker, but less efficient code in other
8221 cases, such as calls via function pointers, or using a linker that doesn't
8222 support this optimization.
8225 @itemx -mno-tail-call
8227 @opindex mno-tail-call
8228 Do (or do not) make additional attempts (beyond those of the
8229 machine-independent portions of the compiler) to optimize tail-recursive
8230 calls into branches. You may not want to do this because the detection of
8231 cases where this is not valid is not totally complete. The default is
8232 @option{-mno-tail-call}.
8234 @item -mcomplex-addr
8235 @itemx -mno-complex-addr
8236 @opindex mcomplex-addr
8237 @opindex mno-complex-addr
8238 Assume (or do not assume) that the use of a complex addressing mode is a
8239 win on this implementation of the i960. Complex addressing modes may not
8240 be worthwhile on the K-series, but they definitely are on the C-series.
8241 The default is currently @option{-mcomplex-addr} for all processors except
8245 @itemx -mno-code-align
8246 @opindex mcode-align
8247 @opindex mno-code-align
8248 Align code to 8-byte boundaries for faster fetching (or don't bother).
8249 Currently turned on by default for C-series implementations only.
8252 @item -mclean-linkage
8253 @itemx -mno-clean-linkage
8254 @opindex mclean-linkage
8255 @opindex mno-clean-linkage
8256 These options are not fully implemented.
8260 @itemx -mic2.0-compat
8261 @itemx -mic3.0-compat
8263 @opindex mic2.0-compat
8264 @opindex mic3.0-compat
8265 Enable compatibility with iC960 v2.0 or v3.0.
8269 @opindex masm-compat
8271 Enable compatibility with the iC960 assembler.
8273 @item -mstrict-align
8274 @itemx -mno-strict-align
8275 @opindex mstrict-align
8276 @opindex mno-strict-align
8277 Do not permit (do permit) unaligned accesses.
8281 Enable structure-alignment compatibility with Intel's gcc release version
8282 1.3 (based on gcc 1.37). This option implies @option{-mstrict-align}.
8284 @item -mlong-double-64
8285 @opindex mlong-double-64
8286 Implement type @samp{long double} as 64-bit floating point numbers.
8287 Without the option @samp{long double} is implemented by 80-bit
8288 floating point numbers. The only reason we have it because there is
8289 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
8290 is only useful for people using soft-float targets. Otherwise, we
8291 should recommend against use of it.
8295 @node DEC Alpha Options
8296 @subsection DEC Alpha Options
8298 These @samp{-m} options are defined for the DEC Alpha implementations:
8301 @item -mno-soft-float
8303 @opindex mno-soft-float
8304 @opindex msoft-float
8305 Use (do not use) the hardware floating-point instructions for
8306 floating-point operations. When @option{-msoft-float} is specified,
8307 functions in @file{libgcc.a} will be used to perform floating-point
8308 operations. Unless they are replaced by routines that emulate the
8309 floating-point operations, or compiled in such a way as to call such
8310 emulations routines, these routines will issue floating-point
8311 operations. If you are compiling for an Alpha without floating-point
8312 operations, you must ensure that the library is built so as not to call
8315 Note that Alpha implementations without floating-point operations are
8316 required to have floating-point registers.
8321 @opindex mno-fp-regs
8322 Generate code that uses (does not use) the floating-point register set.
8323 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8324 register set is not used, floating point operands are passed in integer
8325 registers as if they were integers and floating-point results are passed
8326 in $0 instead of $f0. This is a non-standard calling sequence, so any
8327 function with a floating-point argument or return value called by code
8328 compiled with @option{-mno-fp-regs} must also be compiled with that
8331 A typical use of this option is building a kernel that does not use,
8332 and hence need not save and restore, any floating-point registers.
8336 The Alpha architecture implements floating-point hardware optimized for
8337 maximum performance. It is mostly compliant with the IEEE floating
8338 point standard. However, for full compliance, software assistance is
8339 required. This option generates code fully IEEE compliant code
8340 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8341 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8342 defined during compilation. The resulting code is less efficient but is
8343 able to correctly support denormalized numbers and exceptional IEEE
8344 values such as not-a-number and plus/minus infinity. Other Alpha
8345 compilers call this option @option{-ieee_with_no_inexact}.
8347 @item -mieee-with-inexact
8348 @opindex mieee-with-inexact
8349 This is like @option{-mieee} except the generated code also maintains
8350 the IEEE @var{inexact-flag}. Turning on this option causes the
8351 generated code to implement fully-compliant IEEE math. In addition to
8352 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8353 macro. On some Alpha implementations the resulting code may execute
8354 significantly slower than the code generated by default. Since there is
8355 very little code that depends on the @var{inexact-flag}, you should
8356 normally not specify this option. Other Alpha compilers call this
8357 option @option{-ieee_with_inexact}.
8359 @item -mfp-trap-mode=@var{trap-mode}
8360 @opindex mfp-trap-mode
8361 This option controls what floating-point related traps are enabled.
8362 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8363 The trap mode can be set to one of four values:
8367 This is the default (normal) setting. The only traps that are enabled
8368 are the ones that cannot be disabled in software (e.g., division by zero
8372 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8376 Like @samp{su}, but the instructions are marked to be safe for software
8377 completion (see Alpha architecture manual for details).
8380 Like @samp{su}, but inexact traps are enabled as well.
8383 @item -mfp-rounding-mode=@var{rounding-mode}
8384 @opindex mfp-rounding-mode
8385 Selects the IEEE rounding mode. Other Alpha compilers call this option
8386 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8391 Normal IEEE rounding mode. Floating point numbers are rounded towards
8392 the nearest machine number or towards the even machine number in case
8396 Round towards minus infinity.
8399 Chopped rounding mode. Floating point numbers are rounded towards zero.
8402 Dynamic rounding mode. A field in the floating point control register
8403 (@var{fpcr}, see Alpha architecture reference manual) controls the
8404 rounding mode in effect. The C library initializes this register for
8405 rounding towards plus infinity. Thus, unless your program modifies the
8406 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8409 @item -mtrap-precision=@var{trap-precision}
8410 @opindex mtrap-precision
8411 In the Alpha architecture, floating point traps are imprecise. This
8412 means without software assistance it is impossible to recover from a
8413 floating trap and program execution normally needs to be terminated.
8414 GCC can generate code that can assist operating system trap handlers
8415 in determining the exact location that caused a floating point trap.
8416 Depending on the requirements of an application, different levels of
8417 precisions can be selected:
8421 Program precision. This option is the default and means a trap handler
8422 can only identify which program caused a floating point exception.
8425 Function precision. The trap handler can determine the function that
8426 caused a floating point exception.
8429 Instruction precision. The trap handler can determine the exact
8430 instruction that caused a floating point exception.
8433 Other Alpha compilers provide the equivalent options called
8434 @option{-scope_safe} and @option{-resumption_safe}.
8436 @item -mieee-conformant
8437 @opindex mieee-conformant
8438 This option marks the generated code as IEEE conformant. You must not
8439 use this option unless you also specify @option{-mtrap-precision=i} and either
8440 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8441 is to emit the line @samp{.eflag 48} in the function prologue of the
8442 generated assembly file. Under DEC Unix, this has the effect that
8443 IEEE-conformant math library routines will be linked in.
8445 @item -mbuild-constants
8446 @opindex mbuild-constants
8447 Normally GCC examines a 32- or 64-bit integer constant to
8448 see if it can construct it from smaller constants in two or three
8449 instructions. If it cannot, it will output the constant as a literal and
8450 generate code to load it from the data segment at runtime.
8452 Use this option to require GCC to construct @emph{all} integer constants
8453 using code, even if it takes more instructions (the maximum is six).
8455 You would typically use this option to build a shared library dynamic
8456 loader. Itself a shared library, it must relocate itself in memory
8457 before it can find the variables and constants in its own data segment.
8463 Select whether to generate code to be assembled by the vendor-supplied
8464 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8478 Indicate whether GCC should generate code to use the optional BWX,
8479 CIX, and MAX instruction sets. The default is to use the instruction sets
8480 supported by the CPU type specified via @option{-mcpu=} option or that
8481 of the CPU on which GCC was built if none was specified.
8483 @item -mcpu=@var{cpu_type}
8485 Set the instruction set, register set, and instruction scheduling
8486 parameters for machine type @var{cpu_type}. You can specify either the
8487 @samp{EV} style name or the corresponding chip number. GCC
8488 supports scheduling parameters for the EV4 and EV5 family of processors
8489 and will choose the default values for the instruction set from
8490 the processor you specify. If you do not specify a processor type,
8491 GCC will default to the processor on which the compiler was built.
8493 Supported values for @var{cpu_type} are
8498 Schedules as an EV4 and has no instruction set extensions.
8502 Schedules as an EV5 and has no instruction set extensions.
8506 Schedules as an EV5 and supports the BWX extension.
8511 Schedules as an EV5 and supports the BWX and MAX extensions.
8515 Schedules as an EV5 (until Digital releases the scheduling parameters
8516 for the EV6) and supports the BWX, CIX, and MAX extensions.
8519 @item -mmemory-latency=@var{time}
8520 @opindex mmemory-latency
8521 Sets the latency the scheduler should assume for typical memory
8522 references as seen by the application. This number is highly
8523 dependent on the memory access patterns used by the application
8524 and the size of the external cache on the machine.
8526 Valid options for @var{time} are
8530 A decimal number representing clock cycles.
8536 The compiler contains estimates of the number of clock cycles for
8537 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8538 (also called Dcache, Scache, and Bcache), as well as to main memory.
8539 Note that L3 is only valid for EV5.
8544 @node Clipper Options
8545 @subsection Clipper Options
8547 These @samp{-m} options are defined for the Clipper implementations:
8552 Produce code for a C300 Clipper processor. This is the default.
8556 Produce code for a C400 Clipper processor, i.e.@: use floating point
8560 @node H8/300 Options
8561 @subsection H8/300 Options
8563 These @samp{-m} options are defined for the H8/300 implementations:
8568 Shorten some address references at link time, when possible; uses the
8569 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8570 ld.info, Using ld}, for a fuller description.
8574 Generate code for the H8/300H@.
8578 Generate code for the H8/S@.
8582 Generate code for the H8/S2600. This switch must be used with @option{-ms}.
8586 Make @code{int} data 32 bits by default.
8590 On the H8/300H and H8/S, use the same alignment rules as for the H8/300.
8591 The default for the H8/300H and H8/S is to align longs and floats on 4
8593 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8594 This option has no effect on the H8/300.
8598 @subsection SH Options
8600 These @samp{-m} options are defined for the SH implementations:
8605 Generate code for the SH1.
8609 Generate code for the SH2.
8613 Generate code for the SH3.
8617 Generate code for the SH3e.
8621 Generate code for the SH4 without a floating-point unit.
8623 @item -m4-single-only
8624 @opindex m4-single-only
8625 Generate code for the SH4 with a floating-point unit that only
8626 supports single-precision arithmetic.
8630 Generate code for the SH4 assuming the floating-point unit is in
8631 single-precision mode by default.
8635 Generate code for the SH4.
8639 Compile code for the processor in big endian mode.
8643 Compile code for the processor in little endian mode.
8647 Align doubles at 64-bit boundaries. Note that this changes the calling
8648 conventions, and thus some functions from the standard C library will
8649 not work unless you recompile it first with @option{-mdalign}.
8653 Shorten some address references at link time, when possible; uses the
8654 linker option @option{-relax}.
8658 Use 32-bit offsets in @code{switch} tables. The default is to use
8663 Enable the use of the instruction @code{fmovd}.
8667 Comply with the calling conventions defined by Hitachi.
8671 Mark the @code{MAC} register as call-clobbered, even if
8672 @option{-mhitachi} is given.
8676 Increase IEEE-compliance of floating-point code.
8680 Dump instruction size and location in the assembly code.
8684 This option is deprecated. It pads structures to multiple of 4 bytes,
8685 which is incompatible with the SH ABI@.
8689 Optimize for space instead of speed. Implied by @option{-Os}.
8693 When generating position-independent code, emit function calls using
8694 the Global Offset Table instead of the Procedure Linkage Table.
8698 Generate a library function call to invalidate instruction cache
8699 entries, after fixing up a trampoline. This library function call
8700 doesn't assume it can write to the whole memory address space. This
8701 is the default when the target is @code{sh-*-linux*}.
8704 @node System V Options
8705 @subsection Options for System V
8707 These additional options are available on System V Release 4 for
8708 compatibility with other compilers on those systems:
8713 Create a shared object.
8714 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
8718 Identify the versions of each tool used by the compiler, in a
8719 @code{.ident} assembler directive in the output.
8723 Refrain from adding @code{.ident} directives to the output file (this is
8726 @item -YP,@var{dirs}
8728 Search the directories @var{dirs}, and no others, for libraries
8729 specified with @option{-l}.
8733 Look in the directory @var{dir} to find the M4 preprocessor.
8734 The assembler uses this option.
8735 @c This is supposed to go with a -Yd for predefined M4 macro files, but
8736 @c the generic assembler that comes with Solaris takes just -Ym.
8739 @node TMS320C3x/C4x Options
8740 @subsection TMS320C3x/C4x Options
8741 @cindex TMS320C3x/C4x Options
8743 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
8747 @item -mcpu=@var{cpu_type}
8749 Set the instruction set, register set, and instruction scheduling
8750 parameters for machine type @var{cpu_type}. Supported values for
8751 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
8752 @samp{c44}. The default is @samp{c40} to generate code for the
8757 @itemx -msmall-memory
8759 @opindex mbig-memory
8761 @opindex msmall-memory
8763 Generates code for the big or small memory model. The small memory
8764 model assumed that all data fits into one 64K word page. At run-time
8765 the data page (DP) register must be set to point to the 64K page
8766 containing the .bss and .data program sections. The big memory model is
8767 the default and requires reloading of the DP register for every direct
8774 Allow (disallow) allocation of general integer operands into the block
8781 Enable (disable) generation of code using decrement and branch,
8782 DBcond(D), instructions. This is enabled by default for the C4x. To be
8783 on the safe side, this is disabled for the C3x, since the maximum
8784 iteration count on the C3x is @math{2^23 + 1} (but who iterates loops more than
8785 @math{2^23} times on the C3x?). Note that GCC will try to reverse a loop so
8786 that it can utilise the decrement and branch instruction, but will give
8787 up if there is more than one memory reference in the loop. Thus a loop
8788 where the loop counter is decremented can generate slightly more
8789 efficient code, in cases where the RPTB instruction cannot be utilised.
8791 @item -mdp-isr-reload
8793 @opindex mdp-isr-reload
8795 Force the DP register to be saved on entry to an interrupt service
8796 routine (ISR), reloaded to point to the data section, and restored on
8797 exit from the ISR@. This should not be required unless someone has
8798 violated the small memory model by modifying the DP register, say within
8805 For the C3x use the 24-bit MPYI instruction for integer multiplies
8806 instead of a library call to guarantee 32-bit results. Note that if one
8807 of the operands is a constant, then the multiplication will be performed
8808 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
8809 then squaring operations are performed inline instead of a library call.
8812 @itemx -mno-fast-fix
8814 @opindex mno-fast-fix
8815 The C3x/C4x FIX instruction to convert a floating point value to an
8816 integer value chooses the nearest integer less than or equal to the
8817 floating point value rather than to the nearest integer. Thus if the
8818 floating point number is negative, the result will be incorrectly
8819 truncated an additional code is necessary to detect and correct this
8820 case. This option can be used to disable generation of the additional
8821 code required to correct the result.
8827 Enable (disable) generation of repeat block sequences using the RPTB
8828 instruction for zero overhead looping. The RPTB construct is only used
8829 for innermost loops that do not call functions or jump across the loop
8830 boundaries. There is no advantage having nested RPTB loops due to the
8831 overhead required to save and restore the RC, RS, and RE registers.
8832 This is enabled by default with @option{-O2}.
8834 @item -mrpts=@var{count}
8838 Enable (disable) the use of the single instruction repeat instruction
8839 RPTS@. If a repeat block contains a single instruction, and the loop
8840 count can be guaranteed to be less than the value @var{count}, GCC will
8841 emit a RPTS instruction instead of a RPTB@. If no value is specified,
8842 then a RPTS will be emitted even if the loop count cannot be determined
8843 at compile time. Note that the repeated instruction following RPTS does
8844 not have to be reloaded from memory each iteration, thus freeing up the
8845 CPU buses for operands. However, since interrupts are blocked by this
8846 instruction, it is disabled by default.
8848 @item -mloop-unsigned
8849 @itemx -mno-loop-unsigned
8850 @opindex mloop-unsigned
8851 @opindex mno-loop-unsigned
8852 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
8853 is @math{2^31 + 1} since these instructions test if the iteration count is
8854 negative to terminate the loop. If the iteration count is unsigned
8855 there is a possibility than the @math{2^31 + 1} maximum iteration count may be
8856 exceeded. This switch allows an unsigned iteration count.
8860 Try to emit an assembler syntax that the TI assembler (asm30) is happy
8861 with. This also enforces compatibility with the API employed by the TI
8862 C3x C compiler. For example, long doubles are passed as structures
8863 rather than in floating point registers.
8869 Generate code that uses registers (stack) for passing arguments to functions.
8870 By default, arguments are passed in registers where possible rather
8871 than by pushing arguments on to the stack.
8873 @item -mparallel-insns
8874 @itemx -mno-parallel-insns
8875 @opindex mparallel-insns
8876 @opindex mno-parallel-insns
8877 Allow the generation of parallel instructions. This is enabled by
8878 default with @option{-O2}.
8880 @item -mparallel-mpy
8881 @itemx -mno-parallel-mpy
8882 @opindex mparallel-mpy
8883 @opindex mno-parallel-mpy
8884 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
8885 provided @option{-mparallel-insns} is also specified. These instructions have
8886 tight register constraints which can pessimize the code generation
8892 @subsection V850 Options
8893 @cindex V850 Options
8895 These @samp{-m} options are defined for V850 implementations:
8899 @itemx -mno-long-calls
8900 @opindex mlong-calls
8901 @opindex mno-long-calls
8902 Treat all calls as being far away (near). If calls are assumed to be
8903 far away, the compiler will always load the functions address up into a
8904 register, and call indirect through the pointer.
8910 Do not optimize (do optimize) basic blocks that use the same index
8911 pointer 4 or more times to copy pointer into the @code{ep} register, and
8912 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
8913 option is on by default if you optimize.
8915 @item -mno-prolog-function
8916 @itemx -mprolog-function
8917 @opindex mno-prolog-function
8918 @opindex mprolog-function
8919 Do not use (do use) external functions to save and restore registers at
8920 the prolog and epilog of a function. The external functions are slower,
8921 but use less code space if more than one function saves the same number
8922 of registers. The @option{-mprolog-function} option is on by default if
8927 Try to make the code as small as possible. At present, this just turns
8928 on the @option{-mep} and @option{-mprolog-function} options.
8932 Put static or global variables whose size is @var{n} bytes or less into
8933 the tiny data area that register @code{ep} points to. The tiny data
8934 area can hold up to 256 bytes in total (128 bytes for byte references).
8938 Put static or global variables whose size is @var{n} bytes or less into
8939 the small data area that register @code{gp} points to. The small data
8940 area can hold up to 64 kilobytes.
8944 Put static or global variables whose size is @var{n} bytes or less into
8945 the first 32 kilobytes of memory.
8949 Specify that the target processor is the V850.
8952 @opindex mbig-switch
8953 Generate code suitable for big switch tables. Use this option only if
8954 the assembler/linker complain about out of range branches within a switch
8959 @subsection ARC Options
8962 These options are defined for ARC implementations:
8967 Compile code for little endian mode. This is the default.
8971 Compile code for big endian mode.
8974 @opindex mmangle-cpu
8975 Prepend the name of the cpu to all public symbol names.
8976 In multiple-processor systems, there are many ARC variants with different
8977 instruction and register set characteristics. This flag prevents code
8978 compiled for one cpu to be linked with code compiled for another.
8979 No facility exists for handling variants that are ``almost identical''.
8980 This is an all or nothing option.
8982 @item -mcpu=@var{cpu}
8984 Compile code for ARC variant @var{cpu}.
8985 Which variants are supported depend on the configuration.
8986 All variants support @option{-mcpu=base}, this is the default.
8988 @item -mtext=@var{text-section}
8989 @itemx -mdata=@var{data-section}
8990 @itemx -mrodata=@var{readonly-data-section}
8994 Put functions, data, and readonly data in @var{text-section},
8995 @var{data-section}, and @var{readonly-data-section} respectively
8996 by default. This can be overridden with the @code{section} attribute.
8997 @xref{Variable Attributes}.
9002 @subsection NS32K Options
9003 @cindex NS32K options
9005 These are the @samp{-m} options defined for the 32000 series. The default
9006 values for these options depends on which style of 32000 was selected when
9007 the compiler was configured; the defaults for the most common choices are
9015 Generate output for a 32032. This is the default
9016 when the compiler is configured for 32032 and 32016 based systems.
9022 Generate output for a 32332. This is the default
9023 when the compiler is configured for 32332-based systems.
9029 Generate output for a 32532. This is the default
9030 when the compiler is configured for 32532-based systems.
9034 Generate output containing 32081 instructions for floating point.
9035 This is the default for all systems.
9039 Generate output containing 32381 instructions for floating point. This
9040 also implies @option{-m32081}. The 32381 is only compatible with the 32332
9041 and 32532 cpus. This is the default for the pc532-netbsd configuration.
9045 Try and generate multiply-add floating point instructions @code{polyF}
9046 and @code{dotF}. This option is only available if the @option{-m32381}
9047 option is in effect. Using these instructions requires changes to
9048 register allocation which generally has a negative impact on
9049 performance. This option should only be enabled when compiling code
9050 particularly likely to make heavy use of multiply-add instructions.
9053 @opindex mnomulti-add
9054 Do not try and generate multiply-add floating point instructions
9055 @code{polyF} and @code{dotF}. This is the default on all platforms.
9058 @opindex msoft-float
9059 Generate output containing library calls for floating point.
9060 @strong{Warning:} the requisite libraries may not be available.
9063 @opindex mnobitfield
9064 Do not use the bit-field instructions. On some machines it is faster to
9065 use shifting and masking operations. This is the default for the pc532.
9069 Do use the bit-field instructions. This is the default for all platforms
9074 Use a different function-calling convention, in which functions
9075 that take a fixed number of arguments return pop their
9076 arguments on return with the @code{ret} instruction.
9078 This calling convention is incompatible with the one normally
9079 used on Unix, so you cannot use it if you need to call libraries
9080 compiled with the Unix compiler.
9082 Also, you must provide function prototypes for all functions that
9083 take variable numbers of arguments (including @code{printf});
9084 otherwise incorrect code will be generated for calls to those
9087 In addition, seriously incorrect code will result if you call a
9088 function with too many arguments. (Normally, extra arguments are
9089 harmlessly ignored.)
9091 This option takes its name from the 680x0 @code{rtd} instruction.
9096 Use a different function-calling convention where the first two arguments
9097 are passed in registers.
9099 This calling convention is incompatible with the one normally
9100 used on Unix, so you cannot use it if you need to call libraries
9101 compiled with the Unix compiler.
9104 @opindex mnoregparam
9105 Do not pass any arguments in registers. This is the default for all
9110 It is OK to use the sb as an index register which is always loaded with
9111 zero. This is the default for the pc532-netbsd target.
9115 The sb register is not available for use or has not been initialized to
9116 zero by the run time system. This is the default for all targets except
9117 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
9118 @option{-fpic} is set.
9122 Many ns32000 series addressing modes use displacements of up to 512MB@.
9123 If an address is above 512MB then displacements from zero can not be used.
9124 This option causes code to be generated which can be loaded above 512MB@.
9125 This may be useful for operating systems or ROM code.
9129 Assume code will be loaded in the first 512MB of virtual address space.
9130 This is the default for all platforms.
9136 @subsection AVR Options
9139 These options are defined for AVR implementations:
9142 @item -mmcu=@var{mcu}
9144 Specify ATMEL AVR instruction set or MCU type.
9146 Instruction set avr1 is for the minimal AVR core, not supported by the C
9147 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9148 attiny11, attiny12, attiny15, attiny28).
9150 Instruction set avr2 (default) is for the classic AVR core with up to
9151 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9152 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9153 at90c8534, at90s8535).
9155 Instruction set avr3 is for the classic AVR core with up to 128K program
9156 memory space (MCU types: atmega103, atmega603).
9158 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9159 memory space (MCU types: atmega83, atmega85).
9161 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9162 memory space (MCU types: atmega161, atmega163, atmega32, at94k).
9166 Output instruction sizes to the asm file.
9168 @item -minit-stack=@var{N}
9169 @opindex minit-stack
9170 Specify the initial stack address, which may be a symbol or numeric value,
9171 @samp{__stack} is the default.
9173 @item -mno-interrupts
9174 @opindex mno-interrupts
9175 Generated code is not compatible with hardware interrupts.
9176 Code size will be smaller.
9178 @item -mcall-prologues
9179 @opindex mcall-prologues
9180 Functions prologues/epilogues expanded as call to appropriate
9181 subroutines. Code size will be smaller.
9183 @item -mno-tablejump
9184 @opindex mno-tablejump
9185 Do not generate tablejump insns which sometimes increase code size.
9188 @opindex mtiny-stack
9189 Change only the low 8 bits of the stack pointer.
9193 @subsection MCore Options
9194 @cindex MCore options
9196 These are the @samp{-m} options defined for the Motorola M*Core
9206 @opindex mno-hardlit
9207 Inline constants into the code stream if it can be done in two
9208 instructions or less.
9216 Use the divide instruction. (Enabled by default).
9218 @item -mrelax-immediate
9219 @itemx -mrelax-immediate
9220 @itemx -mno-relax-immediate
9221 @opindex mrelax-immediate
9222 @opindex mrelax-immediate
9223 @opindex mno-relax-immediate
9224 Allow arbitrary sized immediates in bit operations.
9226 @item -mwide-bitfields
9227 @itemx -mwide-bitfields
9228 @itemx -mno-wide-bitfields
9229 @opindex mwide-bitfields
9230 @opindex mwide-bitfields
9231 @opindex mno-wide-bitfields
9232 Always treat bit-fields as int-sized.
9234 @item -m4byte-functions
9235 @itemx -m4byte-functions
9236 @itemx -mno-4byte-functions
9237 @opindex m4byte-functions
9238 @opindex m4byte-functions
9239 @opindex mno-4byte-functions
9240 Force all functions to be aligned to a four byte boundary.
9242 @item -mcallgraph-data
9243 @itemx -mcallgraph-data
9244 @itemx -mno-callgraph-data
9245 @opindex mcallgraph-data
9246 @opindex mcallgraph-data
9247 @opindex mno-callgraph-data
9248 Emit callgraph information.
9252 @itemx -mno-slow-bytes
9253 @opindex mslow-bytes
9254 @opindex mslow-bytes
9255 @opindex mno-slow-bytes
9256 Prefer word access when reading byte quantities.
9258 @item -mlittle-endian
9259 @itemx -mlittle-endian
9261 @opindex mlittle-endian
9262 @opindex mlittle-endian
9263 @opindex mbig-endian
9264 Generate code for a little endian target.
9272 Generate code for the 210 processor.
9276 @subsection IA-64 Options
9277 @cindex IA-64 Options
9279 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9283 @opindex mbig-endian
9284 Generate code for a big endian target. This is the default for HPUX@.
9286 @item -mlittle-endian
9287 @opindex mlittle-endian
9288 Generate code for a little endian target. This is the default for AIX5
9295 Generate (or don't) code for the GNU assembler. This is the default.
9296 @c Also, this is the default if the configure option @option{--with-gnu-as}
9303 Generate (or don't) code for the GNU linker. This is the default.
9304 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9309 Generate code that does not use a global pointer register. The result
9310 is not position independent code, and violates the IA-64 ABI@.
9312 @item -mvolatile-asm-stop
9313 @itemx -mno-volatile-asm-stop
9314 @opindex mvolatile-asm-stop
9315 @opindex mno-volatile-asm-stop
9316 Generate (or don't) a stop bit immediately before and after volatile asm
9321 Generate code that works around Itanium B step errata.
9323 @item -mregister-names
9324 @itemx -mno-register-names
9325 @opindex mregister-names
9326 @opindex mno-register-names
9327 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9328 the stacked registers. This may make assembler output more readable.
9334 Disable (or enable) optimizations that use the small data section. This may
9335 be useful for working around optimizer bugs.
9338 @opindex mconstant-gp
9339 Generate code that uses a single constant global pointer value. This is
9340 useful when compiling kernel code.
9344 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9345 This is useful when compiling firmware code.
9347 @item -minline-divide-min-latency
9348 @opindex minline-divide-min-latency
9349 Generate code for inline divides using the minimum latency algorithm.
9351 @item -minline-divide-max-throughput
9352 @opindex minline-divide-max-throughput
9353 Generate code for inline divides using the maximum throughput algorithm.
9355 @item -mno-dwarf2-asm
9357 @opindex mno-dwarf2-asm
9358 @opindex mdwarf2-asm
9359 Don't (or do) generate assembler code for the DWARF2 line number debugging
9360 info. This may be useful when not using the GNU assembler.
9362 @item -mfixed-range=@var{register-range}
9363 @opindex mfixed-range
9364 Generate code treating the given register range as fixed registers.
9365 A fixed register is one that the register allocator can not use. This is
9366 useful when compiling kernel code. A register range is specified as
9367 two registers separated by a dash. Multiple register ranges can be
9368 specified separated by a comma.
9372 @subsection D30V Options
9373 @cindex D30V Options
9375 These @samp{-m} options are defined for D30V implementations:
9380 Link the @samp{.text}, @samp{.data}, @samp{.bss}, @samp{.strings},
9381 @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections into external
9382 memory, which starts at location @code{0x80000000}.
9386 Same as the @option{-mextmem} switch.
9390 Link the @samp{.text} section into onchip text memory, which starts at
9391 location @code{0x0}. Also link @samp{.data}, @samp{.bss},
9392 @samp{.strings}, @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections
9393 into onchip data memory, which starts at location @code{0x20000000}.
9395 @item -mno-asm-optimize
9396 @itemx -masm-optimize
9397 @opindex mno-asm-optimize
9398 @opindex masm-optimize
9399 Disable (enable) passing @option{-O} to the assembler when optimizing.
9400 The assembler uses the @option{-O} option to automatically parallelize
9401 adjacent short instructions where possible.
9403 @item -mbranch-cost=@var{n}
9404 @opindex mbranch-cost
9405 Increase the internal costs of branches to @var{n}. Higher costs means
9406 that the compiler will issue more instructions to avoid doing a branch.
9409 @item -mcond-exec=@var{n}
9411 Specify the maximum number of conditionally executed instructions that
9412 replace a branch. The default is 4.
9415 @node S/390 and zSeries Options
9416 @subsection S/390 and zSeries Options
9417 @cindex S/390 and zSeries Options
9419 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
9424 @opindex mhard-float
9425 @opindex msoft-float
9426 Use (do not use) the hardware floating-point instructions and registers
9427 for floating-point operations. When @option{-msoft-float} is specified,
9428 functions in @file{libgcc.a} will be used to perform floating-point
9429 operations. When @option{-mhard-float} is specified, the compiler
9430 generates IEEE floating-point instructions. This is the default.
9433 @itemx -mno-backchain
9435 @opindex mno-backchain
9436 Generate (or do not generate) code which maintains an explicit
9437 backchain within the stack frame that points to the caller's frame.
9438 This is currently needed to allow debugging. The default is to
9439 generate the backchain.
9442 @itemx -mno-small-exec
9443 @opindex msmall-exec
9444 @opindex mno-small-exec
9445 Generate (or do not generate) code using the @code{bras} instruction
9446 to do subroutine calls.
9447 This only works reliably if the total executable size does not
9448 exceed 64k. The default is to use the @code{basr} instruction instead,
9449 which does not have this limitation.
9455 When @option{-m31} is specified, generate code compliant to the
9456 Linux for S/390 ABI@. When @option{-m64} is specified, generate
9457 code compliant to the Linux for zSeries ABI@. This allows GCC in
9458 particular to generate 64-bit instructions. For the @samp{s390}
9459 targets, the default is @option{-m31}, while the @samp{s390x}
9460 targets default to @option{-m64}.
9466 Generate (or do not generate) code using the @code{mvcle} instruction
9467 to perform block moves. When @option{-mno-mvcle} is specifed,
9468 use a @code{mvc} loop instead. This is the default.
9474 Print (or do not print) additional debug information when compiling.
9475 The default is to not print debug information.
9480 @subsection CRIS Options
9481 @cindex CRIS Options
9483 These options are defined specifically for the CRIS ports.
9486 @item -march=@var{architecture-type}
9487 @itemx -mcpu=@var{architecture-type}
9490 Generate code for the specified architecture. The choices for
9491 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9492 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX.
9493 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9496 @item -mtune=@var{architecture-type}
9498 Tune to @var{architecture-type} everything applicable about the generated
9499 code, except for the ABI and the set of available instructions. The
9500 choices for @var{architecture-type} are the same as for
9501 @option{-march=@var{architecture-type}}.
9503 @item -mmax-stack-frame=@var{n}
9504 @opindex mmax-stack-frame
9505 Warn when the stack frame of a function exceeds @var{n} bytes.
9507 @item -melinux-stacksize=@var{n}
9508 @opindex melinux-stacksize
9509 Only available with the @samp{cris-axis-aout} target. Arranges for
9510 indications in the program to the kernel loader that the stack of the
9511 program should be set to @var{n} bytes.
9517 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9518 @option{-march=v3} and @option{-march=v8} respectively.
9522 Enable CRIS-specific verbose debug-related information in the assembly
9523 code. This option also has the effect to turn off the @samp{#NO_APP}
9524 formatted-code indicator to the assembler at the beginning of the
9529 Do not use condition-code results from previous instruction; always emit
9530 compare and test instructions before use of condition codes.
9532 @item -mno-side-effects
9533 @opindex mno-side-effects
9534 Do not emit instructions with side-effects in addressing modes other than
9538 @itemx -mno-stack-align
9540 @itemx -mno-data-align
9541 @itemx -mconst-align
9542 @itemx -mno-const-align
9543 @opindex mstack-align
9544 @opindex mno-stack-align
9545 @opindex mdata-align
9546 @opindex mno-data-align
9547 @opindex mconst-align
9548 @opindex mno-const-align
9549 These options (no-options) arranges (eliminate arrangements) for the
9550 stack-frame, individual data and constants to be aligned for the maximum
9551 single data access size for the chosen CPU model. The default is to
9552 arrange for 32-bit alignment. ABI details such as structure layout are
9553 not affected by these options.
9561 Similar to the stack- data- and const-align options above, these options
9562 arrange for stack-frame, writable data and constants to all be 32-bit,
9563 16-bit or 8-bit aligned. The default is 32-bit alignment.
9565 @item -mno-prologue-epilogue
9566 @itemx -mprologue-epilogue
9567 @opindex mno-prologue-epilogue
9568 @opindex mprologue-epilogue
9569 With @option{-mno-prologue-epilogue}, the normal function prologue and
9570 epilogue that sets up the stack-frame are omitted and no return
9571 instructions or return sequences are generated in the code. Use this
9572 option only together with visual inspection of the compiled code: no
9573 warnings or errors are generated when call-saved registers must be saved,
9574 or storage for local variable needs to be allocated.
9580 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9581 instruction sequences that load addresses for functions from the PLT part
9582 of the GOT rather than (traditional on other architectures) calls to the
9583 PLT. The default is @option{-mgotplt}.
9587 Legacy no-op option only recognized with the cris-axis-aout target.
9591 Legacy no-op option only recognized with the cris-axis-elf and
9592 cris-axis-linux-gnu targets.
9596 Only recognized with the cris-axis-aout target, where it selects a
9597 GNU/linux-like multilib, include files and instruction set for
9602 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9606 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
9607 to link with input-output functions from a simulator library. Code,
9608 initialized data and zero-initialized data are allocated consecutively.
9612 Like @option{-sim}, but pass linker options to locate initialized data at
9613 0x40000000 and zero-initialized data at 0x80000000.
9617 @node Code Gen Options
9618 @section Options for Code Generation Conventions
9619 @cindex code generation conventions
9620 @cindex options, code generation
9621 @cindex run-time options
9623 These machine-independent options control the interface conventions
9624 used in code generation.
9626 Most of them have both positive and negative forms; the negative form
9627 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
9628 one of the forms is listed---the one which is not the default. You
9629 can figure out the other form by either removing @samp{no-} or adding
9634 @opindex fexceptions
9635 Enable exception handling. Generates extra code needed to propagate
9636 exceptions. For some targets, this implies GCC will generate frame
9637 unwind information for all functions, which can produce significant data
9638 size overhead, although it does not affect execution. If you do not
9639 specify this option, GCC will enable it by default for languages like
9640 C++ which normally require exception handling, and disable it for
9641 languages like C that do not normally require it. However, you may need
9642 to enable this option when compiling C code that needs to interoperate
9643 properly with exception handlers written in C++. You may also wish to
9644 disable this option if you are compiling older C++ programs that don't
9645 use exception handling.
9647 @item -fnon-call-exceptions
9648 @opindex fnon-call-exceptions
9649 Generate code that allows trapping instructions to throw exceptions.
9650 Note that this requires platform-specific runtime support that does
9651 not exist everywhere. Moreover, it only allows @emph{trapping}
9652 instructions to throw exceptions, i.e.@: memory references or floating
9653 point instructions. It does not allow exceptions to be thrown from
9654 arbitrary signal handlers such as @code{SIGALRM}.
9656 @item -funwind-tables
9657 @opindex funwind-tables
9658 Similar to @option{-fexceptions}, except that it will just generate any needed
9659 static data, but will not affect the generated code in any other way.
9660 You will normally not enable this option; instead, a language processor
9661 that needs this handling would enable it on your behalf.
9663 @item -fasynchronous-unwind-tables
9664 @opindex funwind-tables
9665 Generate unwind table in dwarf2 format, if supported by target machine. The
9666 table is exact at each instruction boundary, so it can be used for stack
9667 unwinding from asynchronous events (such as debugger or garbage collector).
9669 @item -fpcc-struct-return
9670 @opindex fpcc-struct-return
9671 Return ``short'' @code{struct} and @code{union} values in memory like
9672 longer ones, rather than in registers. This convention is less
9673 efficient, but it has the advantage of allowing intercallability between
9674 GCC-compiled files and files compiled with other compilers.
9676 The precise convention for returning structures in memory depends
9677 on the target configuration macros.
9679 Short structures and unions are those whose size and alignment match
9680 that of some integer type.
9682 @item -freg-struct-return
9683 @opindex freg-struct-return
9684 Return @code{struct} and @code{union} values in registers when possible.
9685 This is more efficient for small structures than
9686 @option{-fpcc-struct-return}.
9688 If you specify neither @option{-fpcc-struct-return} nor
9689 @option{-freg-struct-return}, GCC defaults to whichever convention is
9690 standard for the target. If there is no standard convention, GCC
9691 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
9692 the principal compiler. In those cases, we can choose the standard, and
9693 we chose the more efficient register return alternative.
9696 @opindex fshort-enums
9697 Allocate to an @code{enum} type only as many bytes as it needs for the
9698 declared range of possible values. Specifically, the @code{enum} type
9699 will be equivalent to the smallest integer type which has enough room.
9701 @item -fshort-double
9702 @opindex fshort-double
9703 Use the same size for @code{double} as for @code{float}.
9706 @opindex fshared-data
9707 Requests that the data and non-@code{const} variables of this
9708 compilation be shared data rather than private data. The distinction
9709 makes sense only on certain operating systems, where shared data is
9710 shared between processes running the same program, while private data
9711 exists in one copy per process.
9715 In C, allocate even uninitialized global variables in the data section of the
9716 object file, rather than generating them as common blocks. This has the
9717 effect that if the same variable is declared (without @code{extern}) in
9718 two different compilations, you will get an error when you link them.
9719 The only reason this might be useful is if you wish to verify that the
9720 program will work on other systems which always work this way.
9724 Ignore the @samp{#ident} directive.
9726 @item -fno-gnu-linker
9727 @opindex fno-gnu-linker
9728 Do not output global initializations (such as C++ constructors and
9729 destructors) in the form used by the GNU linker (on systems where the GNU
9730 linker is the standard method of handling them). Use this option when
9731 you want to use a non-GNU linker, which also requires using the
9732 @command{collect2} program to make sure the system linker includes
9733 constructors and destructors. (@command{collect2} is included in the GCC
9734 distribution.) For systems which @emph{must} use @command{collect2}, the
9735 compiler driver @command{gcc} is configured to do this automatically.
9737 @item -finhibit-size-directive
9738 @opindex finhibit-size-directive
9739 Don't output a @code{.size} assembler directive, or anything else that
9740 would cause trouble if the function is split in the middle, and the
9741 two halves are placed at locations far apart in memory. This option is
9742 used when compiling @file{crtstuff.c}; you should not need to use it
9746 @opindex fverbose-asm
9747 Put extra commentary information in the generated assembly code to
9748 make it more readable. This option is generally only of use to those
9749 who actually need to read the generated assembly code (perhaps while
9750 debugging the compiler itself).
9752 @option{-fno-verbose-asm}, the default, causes the
9753 extra information to be omitted and is useful when comparing two assembler
9758 Consider all memory references through pointers to be volatile.
9760 @item -fvolatile-global
9761 @opindex fvolatile-global
9762 Consider all memory references to extern and global data items to
9763 be volatile. GCC does not consider static data items to be volatile
9764 because of this switch.
9766 @item -fvolatile-static
9767 @opindex fvolatile-static
9768 Consider all memory references to static data to be volatile.
9772 @cindex global offset table
9774 Generate position-independent code (PIC) suitable for use in a shared
9775 library, if supported for the target machine. Such code accesses all
9776 constant addresses through a global offset table (GOT)@. The dynamic
9777 loader resolves the GOT entries when the program starts (the dynamic
9778 loader is not part of GCC; it is part of the operating system). If
9779 the GOT size for the linked executable exceeds a machine-specific
9780 maximum size, you get an error message from the linker indicating that
9781 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
9782 instead. (These maximums are 16k on the m88k, 8k on the Sparc, and 32k
9783 on the m68k and RS/6000. The 386 has no such limit.)
9785 Position-independent code requires special support, and therefore works
9786 only on certain machines. For the 386, GCC supports PIC for System V
9787 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
9788 position-independent.
9792 If supported for the target machine, emit position-independent code,
9793 suitable for dynamic linking and avoiding any limit on the size of the
9794 global offset table. This option makes a difference on the m68k, m88k,
9797 Position-independent code requires special support, and therefore works
9798 only on certain machines.
9800 @item -ffixed-@var{reg}
9802 Treat the register named @var{reg} as a fixed register; generated code
9803 should never refer to it (except perhaps as a stack pointer, frame
9804 pointer or in some other fixed role).
9806 @var{reg} must be the name of a register. The register names accepted
9807 are machine-specific and are defined in the @code{REGISTER_NAMES}
9808 macro in the machine description macro file.
9810 This flag does not have a negative form, because it specifies a
9813 @item -fcall-used-@var{reg}
9815 Treat the register named @var{reg} as an allocable register that is
9816 clobbered by function calls. It may be allocated for temporaries or
9817 variables that do not live across a call. Functions compiled this way
9818 will not save and restore the register @var{reg}.
9820 It is an error to used this flag with the frame pointer or stack pointer.
9821 Use of this flag for other registers that have fixed pervasive roles in
9822 the machine's execution model will produce disastrous results.
9824 This flag does not have a negative form, because it specifies a
9827 @item -fcall-saved-@var{reg}
9828 @opindex fcall-saved
9829 Treat the register named @var{reg} as an allocable register saved by
9830 functions. It may be allocated even for temporaries or variables that
9831 live across a call. Functions compiled this way will save and restore
9832 the register @var{reg} if they use it.
9834 It is an error to used this flag with the frame pointer or stack pointer.
9835 Use of this flag for other registers that have fixed pervasive roles in
9836 the machine's execution model will produce disastrous results.
9838 A different sort of disaster will result from the use of this flag for
9839 a register in which function values may be returned.
9841 This flag does not have a negative form, because it specifies a
9845 @opindex fpack-struct
9846 Pack all structure members together without holes. Usually you would
9847 not want to use this option, since it makes the code suboptimal, and
9848 the offsets of structure members won't agree with system libraries.
9850 @item -fcheck-memory-usage
9851 @opindex fcheck-memory-usage
9852 Generate extra code to check each memory access. GCC will generate
9853 code that is suitable for a detector of bad memory accesses such as
9856 Normally, you should compile all, or none, of your code with this option.
9858 If you do mix code compiled with and without this option,
9859 you must ensure that all code that has side effects
9860 and that is called by code compiled with this option
9861 is, itself, compiled with this option.
9862 If you do not, you might get erroneous messages from the detector.
9864 If you use functions from a library that have side-effects (such as
9865 @code{read}), you might not be able to recompile the library and
9866 specify this option. In that case, you can enable the
9867 @option{-fprefix-function-name} option, which requests GCC to encapsulate
9868 your code and make other functions look as if they were compiled with
9869 @option{-fcheck-memory-usage}. This is done by calling ``stubs'',
9870 which are provided by the detector. If you cannot find or build
9871 stubs for every function you call, you might have to specify
9872 @option{-fcheck-memory-usage} without @option{-fprefix-function-name}.
9874 If you specify this option, you can not use the @code{asm} or
9875 @code{__asm__} keywords in functions with memory checking enabled. GCC
9876 cannot understand what the @code{asm} statement may do, and therefore
9877 cannot generate the appropriate code, so it will reject it. However, if
9878 you specify the function attribute @code{no_check_memory_usage}
9879 (@pxref{Function Attributes}), GCC will disable memory checking within a
9880 function; you may use @code{asm} statements inside such functions. You
9881 may have an inline expansion of a non-checked function within a checked
9882 function; in that case GCC will not generate checks for the inlined
9883 function's memory accesses.
9885 If you move your @code{asm} statements to non-checked inline functions
9886 and they do access memory, you can add calls to the support code in your
9887 inline function, to indicate any reads, writes, or copies being done.
9888 These calls would be similar to those done in the stubs described above.
9890 @item -fprefix-function-name
9891 @opindex fprefix-function-name
9892 Request GCC to add a prefix to the symbols generated for function names.
9893 GCC adds a prefix to the names of functions defined as well as
9894 functions called. Code compiled with this option and code compiled
9895 without the option can't be linked together, unless stubs are used.
9897 If you compile the following code with @option{-fprefix-function-name}
9899 extern void bar (int);
9908 GCC will compile the code as if it was written:
9910 extern void prefix_bar (int);
9914 return prefix_bar (a + 5);
9917 This option is designed to be used with @option{-fcheck-memory-usage}.
9919 @item -finstrument-functions
9920 @opindex finstrument-functions
9921 Generate instrumentation calls for entry and exit to functions. Just
9922 after function entry and just before function exit, the following
9923 profiling functions will be called with the address of the current
9924 function and its call site. (On some platforms,
9925 @code{__builtin_return_address} does not work beyond the current
9926 function, so the call site information may not be available to the
9927 profiling functions otherwise.)
9930 void __cyg_profile_func_enter (void *this_fn,
9932 void __cyg_profile_func_exit (void *this_fn,
9936 The first argument is the address of the start of the current function,
9937 which may be looked up exactly in the symbol table.
9939 This instrumentation is also done for functions expanded inline in other
9940 functions. The profiling calls will indicate where, conceptually, the
9941 inline function is entered and exited. This means that addressable
9942 versions of such functions must be available. If all your uses of a
9943 function are expanded inline, this may mean an additional expansion of
9944 code size. If you use @samp{extern inline} in your C code, an
9945 addressable version of such functions must be provided. (This is
9946 normally the case anyways, but if you get lucky and the optimizer always
9947 expands the functions inline, you might have gotten away without
9948 providing static copies.)
9950 A function may be given the attribute @code{no_instrument_function}, in
9951 which case this instrumentation will not be done. This can be used, for
9952 example, for the profiling functions listed above, high-priority
9953 interrupt routines, and any functions from which the profiling functions
9954 cannot safely be called (perhaps signal handlers, if the profiling
9955 routines generate output or allocate memory).
9958 @opindex fstack-check
9959 Generate code to verify that you do not go beyond the boundary of the
9960 stack. You should specify this flag if you are running in an
9961 environment with multiple threads, but only rarely need to specify it in
9962 a single-threaded environment since stack overflow is automatically
9963 detected on nearly all systems if there is only one stack.
9965 Note that this switch does not actually cause checking to be done; the
9966 operating system must do that. The switch causes generation of code
9967 to ensure that the operating system sees the stack being extended.
9969 @item -fstack-limit-register=@var{reg}
9970 @itemx -fstack-limit-symbol=@var{sym}
9971 @itemx -fno-stack-limit
9972 @opindex fstack-limit-register
9973 @opindex fstack-limit-symbol
9974 @opindex fno-stack-limit
9975 Generate code to ensure that the stack does not grow beyond a certain value,
9976 either the value of a register or the address of a symbol. If the stack
9977 would grow beyond the value, a signal is raised. For most targets,
9978 the signal is raised before the stack overruns the boundary, so
9979 it is possible to catch the signal without taking special precautions.
9981 For instance, if the stack starts at absolute address @samp{0x80000000}
9982 and grows downwards, you can use the flags
9983 @option{-fstack-limit-symbol=__stack_limit} and
9984 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
9985 of 128KB@. Note that this may only work with the GNU linker.
9987 @cindex aliasing of parameters
9988 @cindex parameters, aliased
9989 @item -fargument-alias
9990 @itemx -fargument-noalias
9991 @itemx -fargument-noalias-global
9992 @opindex fargument-alias
9993 @opindex fargument-noalias
9994 @opindex fargument-noalias-global
9995 Specify the possible relationships among parameters and between
9996 parameters and global data.
9998 @option{-fargument-alias} specifies that arguments (parameters) may
9999 alias each other and may alias global storage.@*
10000 @option{-fargument-noalias} specifies that arguments do not alias
10001 each other, but may alias global storage.@*
10002 @option{-fargument-noalias-global} specifies that arguments do not
10003 alias each other and do not alias global storage.
10005 Each language will automatically use whatever option is required by
10006 the language standard. You should not need to use these options yourself.
10008 @item -fleading-underscore
10009 @opindex fleading-underscore
10010 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
10011 change the way C symbols are represented in the object file. One use
10012 is to help link with legacy assembly code.
10014 Be warned that you should know what you are doing when invoking this
10015 option, and that not all targets provide complete support for it.
10020 @node Environment Variables
10021 @section Environment Variables Affecting GCC
10022 @cindex environment variables
10024 @c man begin ENVIRONMENT
10026 This section describes several environment variables that affect how GCC
10027 operates. Some of them work by specifying directories or prefixes to use
10028 when searching for various kinds of files. Some are used to specify other
10029 aspects of the compilation environment.
10032 Note that you can also specify places to search using options such as
10033 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
10034 take precedence over places specified using environment variables, which
10035 in turn take precedence over those specified by the configuration of GCC@.
10039 Note that you can also specify places to search using options such as
10040 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
10041 take precedence over places specified using environment variables, which
10042 in turn take precedence over those specified by the configuration of GCC@.
10049 @c @itemx LC_COLLATE
10051 @c @itemx LC_MONETARY
10052 @c @itemx LC_NUMERIC
10057 @c @findex LC_COLLATE
10058 @findex LC_MESSAGES
10059 @c @findex LC_MONETARY
10060 @c @findex LC_NUMERIC
10064 These environment variables control the way that GCC uses
10065 localization information that allow GCC to work with different
10066 national conventions. GCC inspects the locale categories
10067 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
10068 so. These locale categories can be set to any value supported by your
10069 installation. A typical value is @samp{en_UK} for English in the United
10072 The @env{LC_CTYPE} environment variable specifies character
10073 classification. GCC uses it to determine the character boundaries in
10074 a string; this is needed for some multibyte encodings that contain quote
10075 and escape characters that would otherwise be interpreted as a string
10078 The @env{LC_MESSAGES} environment variable specifies the language to
10079 use in diagnostic messages.
10081 If the @env{LC_ALL} environment variable is set, it overrides the value
10082 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
10083 and @env{LC_MESSAGES} default to the value of the @env{LANG}
10084 environment variable. If none of these variables are set, GCC
10085 defaults to traditional C English behavior.
10089 If @env{TMPDIR} is set, it specifies the directory to use for temporary
10090 files. GCC uses temporary files to hold the output of one stage of
10091 compilation which is to be used as input to the next stage: for example,
10092 the output of the preprocessor, which is the input to the compiler
10095 @item GCC_EXEC_PREFIX
10096 @findex GCC_EXEC_PREFIX
10097 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
10098 names of the subprograms executed by the compiler. No slash is added
10099 when this prefix is combined with the name of a subprogram, but you can
10100 specify a prefix that ends with a slash if you wish.
10102 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
10103 an appropriate prefix to use based on the pathname it was invoked with.
10105 If GCC cannot find the subprogram using the specified prefix, it
10106 tries looking in the usual places for the subprogram.
10108 The default value of @env{GCC_EXEC_PREFIX} is
10109 @file{@var{prefix}/lib/gcc-lib/} where @var{prefix} is the value
10110 of @code{prefix} when you ran the @file{configure} script.
10112 Other prefixes specified with @option{-B} take precedence over this prefix.
10114 This prefix is also used for finding files such as @file{crt0.o} that are
10117 In addition, the prefix is used in an unusual way in finding the
10118 directories to search for header files. For each of the standard
10119 directories whose name normally begins with @samp{/usr/local/lib/gcc-lib}
10120 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
10121 replacing that beginning with the specified prefix to produce an
10122 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
10123 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
10124 These alternate directories are searched first; the standard directories
10127 @item COMPILER_PATH
10128 @findex COMPILER_PATH
10129 The value of @env{COMPILER_PATH} is a colon-separated list of
10130 directories, much like @env{PATH}. GCC tries the directories thus
10131 specified when searching for subprograms, if it can't find the
10132 subprograms using @env{GCC_EXEC_PREFIX}.
10135 @findex LIBRARY_PATH
10136 The value of @env{LIBRARY_PATH} is a colon-separated list of
10137 directories, much like @env{PATH}. When configured as a native compiler,
10138 GCC tries the directories thus specified when searching for special
10139 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
10140 using GCC also uses these directories when searching for ordinary
10141 libraries for the @option{-l} option (but directories specified with
10142 @option{-L} come first).
10144 @item C_INCLUDE_PATH
10145 @itemx CPLUS_INCLUDE_PATH
10146 @itemx OBJC_INCLUDE_PATH
10147 @findex C_INCLUDE_PATH
10148 @findex CPLUS_INCLUDE_PATH
10149 @findex OBJC_INCLUDE_PATH
10150 @c @itemx OBJCPLUS_INCLUDE_PATH
10151 These environment variables pertain to particular languages. Each
10152 variable's value is a colon-separated list of directories, much like
10153 @env{PATH}. When GCC searches for header files, it tries the
10154 directories listed in the variable for the language you are using, after
10155 the directories specified with @option{-I} but before the standard header
10158 @item DEPENDENCIES_OUTPUT
10159 @findex DEPENDENCIES_OUTPUT
10160 @cindex dependencies for make as output
10161 If this variable is set, its value specifies how to output dependencies
10162 for Make based on the header files processed by the compiler. This
10163 output looks much like the output from the @option{-M} option
10164 (@pxref{Preprocessor Options}), but it goes to a separate file, and is
10165 in addition to the usual results of compilation.
10167 The value of @env{DEPENDENCIES_OUTPUT} can be just a file name, in
10168 which case the Make rules are written to that file, guessing the target
10169 name from the source file name. Or the value can have the form
10170 @samp{@var{file} @var{target}}, in which case the rules are written to
10171 file @var{file} using @var{target} as the target name.
10175 @cindex locale definition
10176 This variable is used to pass locale information to the compiler. One way in
10177 which this information is used is to determine the character set to be used
10178 when character literals, string literals and comments are parsed in C and C++.
10179 When the compiler is configured to allow multibyte characters,
10180 the following values for @env{LANG} are recognized:
10184 Recognize JIS characters.
10186 Recognize SJIS characters.
10188 Recognize EUCJP characters.
10191 If @env{LANG} is not defined, or if it has some other value, then the
10192 compiler will use mblen and mbtowc as defined by the default locale to
10193 recognize and translate multibyte characters.
10198 @node Running Protoize
10199 @section Running Protoize
10201 The program @code{protoize} is an optional part of GCC@. You can use
10202 it to add prototypes to a program, thus converting the program to ISO
10203 C in one respect. The companion program @code{unprotoize} does the
10204 reverse: it removes argument types from any prototypes that are found.
10206 When you run these programs, you must specify a set of source files as
10207 command line arguments. The conversion programs start out by compiling
10208 these files to see what functions they define. The information gathered
10209 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
10211 After scanning comes actual conversion. The specified files are all
10212 eligible to be converted; any files they include (whether sources or
10213 just headers) are eligible as well.
10215 But not all the eligible files are converted. By default,
10216 @code{protoize} and @code{unprotoize} convert only source and header
10217 files in the current directory. You can specify additional directories
10218 whose files should be converted with the @option{-d @var{directory}}
10219 option. You can also specify particular files to exclude with the
10220 @option{-x @var{file}} option. A file is converted if it is eligible, its
10221 directory name matches one of the specified directory names, and its
10222 name within the directory has not been excluded.
10224 Basic conversion with @code{protoize} consists of rewriting most
10225 function definitions and function declarations to specify the types of
10226 the arguments. The only ones not rewritten are those for varargs
10229 @code{protoize} optionally inserts prototype declarations at the
10230 beginning of the source file, to make them available for any calls that
10231 precede the function's definition. Or it can insert prototype
10232 declarations with block scope in the blocks where undeclared functions
10235 Basic conversion with @code{unprotoize} consists of rewriting most
10236 function declarations to remove any argument types, and rewriting
10237 function definitions to the old-style pre-ISO form.
10239 Both conversion programs print a warning for any function declaration or
10240 definition that they can't convert. You can suppress these warnings
10243 The output from @code{protoize} or @code{unprotoize} replaces the
10244 original source file. The original file is renamed to a name ending
10245 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
10246 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
10247 for DOS) file already exists, then the source file is simply discarded.
10249 @code{protoize} and @code{unprotoize} both depend on GCC itself to
10250 scan the program and collect information about the functions it uses.
10251 So neither of these programs will work until GCC is installed.
10253 Here is a table of the options you can use with @code{protoize} and
10254 @code{unprotoize}. Each option works with both programs unless
10258 @item -B @var{directory}
10259 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
10260 usual directory (normally @file{/usr/local/lib}). This file contains
10261 prototype information about standard system functions. This option
10262 applies only to @code{protoize}.
10264 @item -c @var{compilation-options}
10265 Use @var{compilation-options} as the options when running @code{gcc} to
10266 produce the @samp{.X} files. The special option @option{-aux-info} is
10267 always passed in addition, to tell @code{gcc} to write a @samp{.X} file.
10269 Note that the compilation options must be given as a single argument to
10270 @code{protoize} or @code{unprotoize}. If you want to specify several
10271 @code{gcc} options, you must quote the entire set of compilation options
10272 to make them a single word in the shell.
10274 There are certain @code{gcc} arguments that you cannot use, because they
10275 would produce the wrong kind of output. These include @option{-g},
10276 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
10277 the @var{compilation-options}, they are ignored.
10280 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
10281 systems) instead of @samp{.c}. This is convenient if you are converting
10282 a C program to C++. This option applies only to @code{protoize}.
10285 Add explicit global declarations. This means inserting explicit
10286 declarations at the beginning of each source file for each function
10287 that is called in the file and was not declared. These declarations
10288 precede the first function definition that contains a call to an
10289 undeclared function. This option applies only to @code{protoize}.
10291 @item -i @var{string}
10292 Indent old-style parameter declarations with the string @var{string}.
10293 This option applies only to @code{protoize}.
10295 @code{unprotoize} converts prototyped function definitions to old-style
10296 function definitions, where the arguments are declared between the
10297 argument list and the initial @samp{@{}. By default, @code{unprotoize}
10298 uses five spaces as the indentation. If you want to indent with just
10299 one space instead, use @option{-i " "}.
10302 Keep the @samp{.X} files. Normally, they are deleted after conversion
10306 Add explicit local declarations. @code{protoize} with @option{-l} inserts
10307 a prototype declaration for each function in each block which calls the
10308 function without any declaration. This option applies only to
10312 Make no real changes. This mode just prints information about the conversions
10313 that would have been done without @option{-n}.
10316 Make no @samp{.save} files. The original files are simply deleted.
10317 Use this option with caution.
10319 @item -p @var{program}
10320 Use the program @var{program} as the compiler. Normally, the name
10321 @file{gcc} is used.
10324 Work quietly. Most warnings are suppressed.
10327 Print the version number, just like @option{-v} for @code{gcc}.
10330 If you need special compiler options to compile one of your program's
10331 source files, then you should generate that file's @samp{.X} file
10332 specially, by running @code{gcc} on that source file with the
10333 appropriate options and the option @option{-aux-info}. Then run
10334 @code{protoize} on the entire set of files. @code{protoize} will use
10335 the existing @samp{.X} file because it is newer than the source file.
10339 gcc -Dfoo=bar file1.c -aux-info file1.X
10344 You need to include the special files along with the rest in the
10345 @code{protoize} command, even though their @samp{.X} files already
10346 exist, because otherwise they won't get converted.
10348 @xref{Protoize Caveats}, for more information on how to use
10349 @code{protoize} successfully.