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 [@samp{-c}|@samp{-S}|@samp{-E}] [@samp{-std=}@var{standard}]
31 [@samp{-g}] [@samp{-pg}] [@samp{-O}@var{level}]
32 [@samp{-W}@var{warn}...] [@samp{-pedantic}]
33 [@samp{-I}@var{dir}...] [@samp{-L}@var{dir}...]
34 [@samp{-D}@var{macro}[=@var{defn}]...] [@samp{-U}@var{macro}]
35 [@samp{-f}@var{option}...] [@samp{-m}@var{machine-option}...]
36 [@samp{-o} @var{outfile}] @var{infile}...
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 @samp{-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: @samp{-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 @samp{-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, @samp{-fforce-mem},
103 @samp{-fstrength-reduce}, @samp{-Wformat} and so on. Most of
104 these have both positive and negative forms; the negative form of
105 @samp{-ffoo} would be @samp{-fno-foo}. This manual documents
106 only one of these two forms, whichever one is not the default.
111 * Option Summary:: Brief list of all options, without explanations.
112 * Overall Options:: Controlling the kind of output:
113 an executable, object files, assembler files,
114 or preprocessed source.
115 * Invoking G++:: Compiling C++ programs.
116 * C Dialect Options:: Controlling the variant of C language compiled.
117 * C++ Dialect Options:: Variations on C++.
118 * Objective-C Dialect Options:: Variations on Objective-C.
119 * Language Independent Options:: Controlling how diagnostics should be
121 * Warning Options:: How picky should the compiler be?
122 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
123 * Optimize Options:: How much optimization?
124 * Preprocessor Options:: Controlling header files and macro definitions.
125 Also, getting dependency information for Make.
126 * Assembler Options:: Passing options to the assembler.
127 * Link Options:: Specifying libraries and so on.
128 * Directory Options:: Where to find header files and libraries.
129 Where to find the compiler executable files.
130 * Spec Files:: How to pass switches to sub-processes.
131 * Target Options:: Running a cross-compiler, or an old version of GCC.
132 * Submodel Options:: Specifying minor hardware or convention variations,
133 such as 68010 vs 68020.
134 * Code Gen Options:: Specifying conventions for function calls, data layout
136 * Environment Variables:: Env vars that affect GCC.
137 * Running Protoize:: Automatically adding or removing function prototypes.
143 @section Option Summary
145 Here is a summary of all the options, grouped by type. Explanations are
146 in the following sections.
149 @item Overall Options
150 @xref{Overall Options,,Options Controlling the Kind of Output}.
152 -c -S -E -o @var{file} -pipe -pass-exit-codes -x @var{language} @gol
153 -v --target-help --help}
155 @item C Language Options
156 @xref{C Dialect Options,,Options Controlling C Dialect}.
158 -ansi -std=@var{standard} -fno-asm -fno-builtin @gol
159 -fhosted -ffreestanding @gol
160 -trigraphs -traditional -traditional-cpp @gol
161 -fallow-single-precision -fcond-mismatch @gol
162 -fsigned-bitfields -fsigned-char @gol
163 -funsigned-bitfields -funsigned-char @gol
164 -fwritable-strings -fshort-wchar}
166 @item C++ Language Options
167 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
169 -fno-access-control -fcheck-new -fconserve-space @gol
170 -fno-const-strings -fdollars-in-identifiers @gol
171 -fno-elide-constructors @gol
172 -fno-enforce-eh-specs -fexternal-templates @gol
173 -falt-external-templates @gol
174 -ffor-scope -fno-for-scope -fno-gnu-keywords -fhonor-std @gol
175 -fno-implicit-templates @gol
176 -fno-implicit-inline-templates @gol
177 -fno-implement-inlines -fms-extensions @gol
178 -fno-nonansi-builtins -fno-operator-names @gol
179 -fno-optional-diags -fpermissive @gol
180 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
181 -fuse-cxa-atexit -fvtable-gc -fno-weak -nostdinc++ @gol
182 -fno-default-inline -Wctor-dtor-privacy @gol
183 -Wnon-virtual-dtor -Wreorder @gol
184 -Weffc++ -Wno-deprecated @gol
185 -Wno-non-template-friend -Wold-style-cast @gol
186 -Woverloaded-virtual -Wno-pmf-conversions @gol
187 -Wsign-promo -Wsynth}
189 @item Objective-C Language Options
190 @xref{Objective-C Dialect Options,,Options Controlling Objective-C Dialect}.
192 -fconstant-string-class=@var{class name} @gol
193 -fgnu-runtime -fnext-runtime -gen-decls
194 -Wno-protocol -Wselector}
196 @item Language Independent Options
197 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
199 -fmessage-length=@var{n} @gol
200 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
202 @item Warning Options
203 @xref{Warning Options,,Options to Request or Suppress Warnings}.
205 -fsyntax-only -pedantic -pedantic-errors @gol
206 -w -W -Wall -Waggregate-return @gol
207 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
208 -Wconversion -Wdisabled-optimization -Werror @gol
209 -Wfloat-equal -Wformat -Wformat=2 @gol
210 -Wformat-nonliteral -Wformat-security @gol
211 -Wid-clash-@var{len} -Wimplicit -Wimplicit-int @gol
212 -Wimplicit-function-declaration @gol
213 -Werror-implicit-function-declaration @gol
214 -Wimport -Winline @gol
215 -Wlarger-than-@var{len} -Wlong-long @gol
216 -Wmain -Wmissing-braces -Wmissing-declarations @gol
217 -Wmissing-format-attribute -Wmissing-noreturn @gol
218 -Wmultichar -Wno-format-extra-args -Wno-format-y2k @gol
219 -Wno-import -Wpacked -Wpadded @gol
220 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
221 -Wreturn-type -Wsequence-point -Wshadow @gol
222 -Wsign-compare -Wswitch -Wsystem-headers @gol
223 -Wtrigraphs -Wundef -Wuninitialized @gol
224 -Wunknown-pragmas -Wunreachable-code @gol
225 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
226 -Wunused-value -Wunused-variable -Wwrite-strings}
228 @item C-only Warning Options
230 -Wbad-function-cast -Wmissing-prototypes -Wnested-externs @gol
231 -Wstrict-prototypes -Wtraditional}
233 @item Debugging Options
234 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
236 -a -ax -d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
237 -fdump-unnumbered -fdump-translation-unit=@var{file} @gol
238 -fdump-class-layout=@var{file} -fmem-report -fpretend-float @gol
239 -fprofile-arcs -ftest-coverage -ftime-report @gol
240 -g -g@var{level} -gcoff -gdwarf -gdwarf-1 -gdwarf-1+ -gdwarf-2 @gol
241 -ggdb -gstabs -gstabs+ -gxcoff -gxcoff+ @gol
242 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
243 -print-prog-name=@var{program} -print-search-dirs -Q @gol
246 @item Optimization Options
247 @xref{Optimize Options,,Options that Control Optimization}.
249 -falign-functions=@var{n} -falign-jumps=@var{n} @gol
250 -falign-labels=@var{n} -falign-loops=@var{n} @gol
251 -fbranch-probabilities -fcaller-saves @gol
252 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections -fdce @gol
253 -fdelayed-branch -fdelete-null-pointer-checks @gol
254 -fexpensive-optimizations -ffast-math -ffloat-store @gol
255 -fforce-addr -fforce-mem -ffunction-sections -fgcse -fgcse-lm -fgcse-sm @gol
256 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
257 -fkeep-static-consts -fmove-all-movables @gol
258 -fno-default-inline -fno-defer-pop @gol
259 -fno-function-cse -fno-guess-branch-probability @gol
260 -fno-inline -fno-math-errno -fno-peephole @gol
261 -funsafe-math-optimizations -fno-trapping-math @gol
262 -fomit-frame-pointer -foptimize-register-move @gol
263 -foptimize-sibling-calls -freduce-all-givs @gol
264 -fregmove -frename-registers @gol
265 -frerun-cse-after-loop -frerun-loop-opt @gol
266 -fschedule-insns -fschedule-insns2 @gol
267 -fsingle-precision-constant -fssa @gol
268 -fstrength-reduce -fstrict-aliasing -fthread-jumps -ftrapv @gol
269 -funroll-all-loops -funroll-loops @gol
270 --param @var{name}=@var{value}
271 -O -O0 -O1 -O2 -O3 -Os}
273 @item Preprocessor Options
274 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
276 -$ -A@var{question}=@var{answer} -A-@var{question}@r{[}=@var{answer}@r{]} @gol
277 -C -dD -dI -dM -dN @gol
278 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
279 -idirafter @var{dir} @gol
280 -include @var{file} -imacros @var{file} @gol
281 -iprefix @var{file} -iwithprefix @var{dir} @gol
282 -iwithprefixbefore @var{dir} -isystem @var{dir} -isystem-c++ @var{dir} @gol
283 -M -MM -MF -MG -MP -MQ -MT -nostdinc -P -remap @gol
284 -trigraphs -undef -U@var{macro} -Wp\,@var{option}}
286 @item Assembler Option
287 @xref{Assembler Options,,Passing Options to the Assembler}.
292 @xref{Link Options,,Options for Linking}.
294 @var{object-file-name} -l@var{library} @gol
295 -nostartfiles -nodefaultlibs -nostdlib @gol
296 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
297 -Wl\,@var{option} -Xlinker @var{option} @gol
300 @item Directory Options
301 @xref{Directory Options,,Options for Directory Search}.
303 -B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}}
306 @c I wrote this xref this way to avoid overfull hbox. -- rms
307 @xref{Target Options}.
309 -b @var{machine} -V @var{version}}
311 @item Machine Dependent Options
312 @xref{Submodel Options,,Hardware Models and Configurations}.
313 @emph{M680x0 Options}
315 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
316 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
317 -mfpa -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
318 -malign-int -mstrict-align}
320 @emph{M68hc1x Options}
322 -m6811 -m6812 -m68hc11 -m68hc12 @gol
323 -mauto-incdec -mshort -msoft-reg-count=@var{count}}
331 -mcpu=@var{cpu type} @gol
332 -mtune=@var{cpu type} @gol
333 -mcmodel=@var{code model} @gol
335 -mapp-regs -mbroken-saverestore -mcypress @gol
336 -mepilogue -mfaster-structs -mflat @gol
337 -mfpu -mhard-float -mhard-quad-float @gol
338 -mimpure-text -mlive-g0 -mno-app-regs @gol
339 -mno-epilogue -mno-faster-structs -mno-flat -mno-fpu @gol
340 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles @gol
341 -msoft-float -msoft-quad-float -msparclite -mstack-bias @gol
342 -msupersparc -munaligned-doubles -mv8}
344 @emph{Convex Options}
346 -mc1 -mc2 -mc32 -mc34 -mc38 @gol
347 -margcount -mnoargcount @gol
348 -mlong32 -mlong64 @gol
349 -mvolatile-cache -mvolatile-nocache}
351 @emph{AMD29K Options}
353 -m29000 -m29050 -mbw -mnbw -mdw -mndw @gol
354 -mlarge -mnormal -msmall @gol
355 -mkernel-registers -mno-reuse-arg-regs @gol
356 -mno-stack-check -mno-storem-bug @gol
357 -mreuse-arg-regs -msoft-float -mstack-check @gol
358 -mstorem-bug -muser-registers}
362 -mapcs-frame -mno-apcs-frame @gol
363 -mapcs-26 -mapcs-32 @gol
364 -mapcs-stack-check -mno-apcs-stack-check @gol
365 -mapcs-float -mno-apcs-float @gol
366 -mapcs-reentrant -mno-apcs-reentrant @gol
367 -msched-prolog -mno-sched-prolog @gol
368 -mlittle-endian -mbig-endian -mwords-little-endian @gol
369 -malignment-traps -mno-alignment-traps @gol
370 -msoft-float -mhard-float -mfpe @gol
371 -mthumb-interwork -mno-thumb-interwork @gol
372 -mcpu=@var{name} -march=@var{name} -mfpe=@var{name} @gol
373 -mstructure-size-boundary=@var{n} @gol
374 -mbsd -mxopen -mno-symrename @gol
375 -mabort-on-noreturn @gol
376 -mlong-calls -mno-long-calls @gol
377 -msingle-pic-base -mno-single-pic-base @gol
378 -mpic-register=@var{reg} @gol
379 -mnop-fun-dllimport @gol
380 -mpoke-function-name @gol
382 -mtpcs-frame -mtpcs-leaf-frame @gol
383 -mcaller-super-interworking -mcallee-super-interworking }
385 @emph{MN10200 Options}
389 @emph{MN10300 Options}
397 @emph{M32R/D Options}
399 -mcode-model=@var{model type} -msdata=@var{sdata type} @gol
404 -m88000 -m88100 -m88110 -mbig-pic @gol
405 -mcheck-zero-division -mhandle-large-shift @gol
406 -midentify-revision -mno-check-zero-division @gol
407 -mno-ocs-debug-info -mno-ocs-frame-position @gol
408 -mno-optimize-arg-area -mno-serialize-volatile @gol
409 -mno-underscores -mocs-debug-info @gol
410 -mocs-frame-position -moptimize-arg-area @gol
411 -mserialize-volatile -mshort-data-@var{num} -msvr3 @gol
412 -msvr4 -mtrap-large-shift -muse-div-instruction @gol
413 -mversion-03.00 -mwarn-passed-structs}
415 @emph{RS/6000 and PowerPC Options}
417 -mcpu=@var{cpu type} @gol
418 -mtune=@var{cpu type} @gol
419 -mpower -mno-power -mpower2 -mno-power2 @gol
420 -mpowerpc -mpowerpc64 -mno-powerpc @gol
421 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
422 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
423 -mnew-mnemonics -mold-mnemonics @gol
424 -mfull-toc -mminimal-toc -mno-fop-in-toc -mno-sum-in-toc @gol
425 -m64 -m32 -mxl-call -mno-xl-call -mthreads -mpe @gol
426 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
427 -mstring -mno-string -mupdate -mno-update @gol
428 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
429 -mstrict-align -mno-strict-align -mrelocatable @gol
430 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
431 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
432 -mcall-aix -mcall-sysv -mprototype -mno-prototype @gol
433 -msim -mmvme -mads -myellowknife -memb -msdata @gol
434 -msdata=@var{opt} -mvxworks -G @var{num}}
438 -mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs @gol
439 -mfull-fp-blocks -mhc-struct-return -min-line-mul @gol
440 -mminimum-fp-blocks -mnohc-struct-return}
444 -mabicalls -mcpu=@var{cpu type}
445 -membedded-data -muninit-const-in-rodata @gol
446 -membedded-pic -mfp32 -mfp64 -mgas -mgp32 -mgp64 @gol
447 -mgpopt -mhalf-pic -mhard-float -mint64 -mips1 @gol
448 -mips2 -mips3 -mips4 -mlong64 -mlong32 -mlong-calls -mmemcpy @gol
449 -mmips-as -mmips-tfile -mno-abicalls @gol
450 -mno-embedded-data -mno-uninit-const-in-rodata -mno-embedded-pic @gol
451 -mno-gpopt -mno-long-calls @gol
452 -mno-memcpy -mno-mips-tfile -mno-rnames -mno-stats @gol
453 -mrnames -msoft-float @gol
454 -m4650 -msingle-float -mmad @gol
455 -mstats -EL -EB -G @var{num} -nocpp @gol
456 -mabi=32 -mabi=n32 -mabi=64 -mabi=eabi @gol
461 -mcpu=@var{cpu type} -march=@var{cpu type} @gol
462 -mintel-syntax -mieee-fp -mno-fancy-math-387 @gol
463 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
464 -mno-wide-multiply -mrtd -malign-double @gol
465 -mpreferred-stack-boundary=@var{num} @gol
466 -mthreads -mno-align-stringops -minline-all-stringops @gol
467 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
468 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer}
472 -march=@var{architecture type} @gol
473 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
474 -mfast-indirect-calls -mgas -mjump-in-delay @gol
475 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
476 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
477 -mno-jump-in-delay -mno-long-load-store @gol
478 -mno-portable-runtime -mno-soft-float @gol
479 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
480 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
481 -mschedule=@var{cpu type} -mspace-regs}
483 @emph{Intel 960 Options}
485 -m@var{cpu type} -masm-compat -mclean-linkage @gol
486 -mcode-align -mcomplex-addr -mleaf-procedures @gol
487 -mic-compat -mic2.0-compat -mic3.0-compat @gol
488 -mintel-asm -mno-clean-linkage -mno-code-align @gol
489 -mno-complex-addr -mno-leaf-procedures @gol
490 -mno-old-align -mno-strict-align -mno-tail-call @gol
491 -mnumerics -mold-align -msoft-float -mstrict-align @gol
494 @emph{DEC Alpha Options}
496 -mfp-regs -mno-fp-regs -mno-soft-float -msoft-float @gol
497 -malpha-as -mgas @gol
498 -mieee -mieee-with-inexact -mieee-conformant @gol
499 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
500 -mtrap-precision=@var{mode} -mbuild-constants @gol
501 -mcpu=@var{cpu type} @gol
502 -mbwx -mno-bwx -mcix -mno-cix -mmax -mno-max @gol
503 -mmemory-latency=@var{time}}
505 @emph{Clipper Options}
509 @emph{H8/300 Options}
511 -mrelax -mh -ms -mint32 -malign-300}
515 -m1 -m2 -m3 -m3e @gol
516 -m4-nofpu -m4-single-only -m4-single -m4 @gol
517 -mb -ml -mdalign -mrelax @gol
518 -mbigtable -mfmovd -mhitachi -mnomacsave @gol
519 -misize -mpadstruct -mspace @gol
523 @emph{System V Options}
525 -Qy -Qn -YP\,@var{paths} -Ym\,@var{dir}}
530 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text section} @gol
531 -mdata=@var{data section} -mrodata=@var{readonly data section}}
533 @emph{TMS320C3x/C4x Options}
535 -mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
536 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
537 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
538 -mparallel-insns -mparallel-mpy -mpreserve-float}
542 -mlong-calls -mno-long-calls -mep -mno-ep @gol
543 -mprolog-function -mno-prolog-function -mspace @gol
544 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
549 -m32032 -m32332 -m32532 -m32081 -m32381 -mmult-add -mnomult-add @gol
550 -msoft-float -mrtd -mnortd -mregparam -mnoregparam -msb -mnosb @gol
551 -mbitfield -mnobitfield -mhimem -mnohimem}
555 -mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
556 -mcall-prologues -mno-tablejump -mtiny-stack}
560 -mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
561 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
562 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
563 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
564 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
568 -mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
569 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
570 -mconstant-gp -mauto-pic -minline-divide-min-latency @gol
571 -minline-divide-max-throughput -mno-dwarf2-asm @gol
572 -mfixed-range=@var{register range}}
574 @item Code Generation Options
575 @xref{Code Gen Options,,Options for Code Generation Conventions}.
577 -fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
578 -fexceptions -funwind-tables -ffixed-@var{reg} @gol
579 -finhibit-size-directive -finstrument-functions @gol
580 -fcheck-memory-usage -fprefix-function-name @gol
581 -fno-common -fno-ident -fno-gnu-linker @gol
582 -fpcc-struct-return -fpic -fPIC @gol
583 -freg-struct-return -fshared-data -fshort-enums @gol
584 -fshort-double -fvolatile -fvolatile-global -fvolatile-static @gol
585 -fverbose-asm -fpack-struct -fstack-check @gol
586 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
587 -fargument-alias -fargument-noalias @gol
588 -fargument-noalias-global @gol
589 -fleading-underscore}
593 * Overall Options:: Controlling the kind of output:
594 an executable, object files, assembler files,
595 or preprocessed source.
596 * C Dialect Options:: Controlling the variant of C language compiled.
597 * C++ Dialect Options:: Variations on C++.
598 * Objective-C Dialect Options:: Variations on Objective-C.
599 * Language Independent Options:: Controlling how diagnostics should be
601 * Warning Options:: How picky should the compiler be?
602 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
603 * Optimize Options:: How much optimization?
604 * Preprocessor Options:: Controlling header files and macro definitions.
605 Also, getting dependency information for Make.
606 * Assembler Options:: Passing options to the assembler.
607 * Link Options:: Specifying libraries and so on.
608 * Directory Options:: Where to find header files and libraries.
609 Where to find the compiler executable files.
610 * Spec Files:: How to pass switches to sub-processes.
611 * Target Options:: Running a cross-compiler, or an old version of GCC.
614 @node Overall Options
615 @section Options Controlling the Kind of Output
617 Compilation can involve up to four stages: preprocessing, compilation
618 proper, assembly and linking, always in that order. The first three
619 stages apply to an individual source file, and end by producing an
620 object file; linking combines all the object files (those newly
621 compiled, and those specified as input) into an executable file.
623 @cindex file name suffix
624 For any given input file, the file name suffix determines what kind of
629 C source code which must be preprocessed.
632 C source code which should not be preprocessed.
635 C++ source code which should not be preprocessed.
638 Objective-C source code. Note that you must link with the library
639 @file{libobjc.a} to make an Objective-C program work.
642 Objective-C source code which should not be preprocessed.
645 C header file (not to be compiled or linked).
649 @itemx @var{file}.cxx
650 @itemx @var{file}.cpp
651 @itemx @var{file}.c++
653 C++ source code which must be preprocessed. Note that in @samp{.cxx},
654 the last two letters must both be literally @samp{x}. Likewise,
655 @samp{.C} refers to a literal capital C.
658 @itemx @var{file}.for
659 @itemx @var{file}.FOR
660 Fortran source code which should not be preprocessed.
663 @itemx @var{file}.fpp
664 @itemx @var{file}.FPP
665 Fortran source code which must be preprocessed (with the traditional
669 Fortran source code which must be preprocessed with a RATFOR
670 preprocessor (not included with GCC).
672 @xref{Overall Options,,Options Controlling the Kind of Output, g77,
673 Using and Porting GNU Fortran}, for more details of the handling of
676 @c FIXME: Descriptions of Java file types.
682 @c GCC also knows about some suffixes for languages not yet included:
692 @itemx @var{file}.chi
693 CHILL source code (preprocessed with the traditional preprocessor).
699 Assembler code which must be preprocessed.
702 An object file to be fed straight into linking.
703 Any file name with no recognized suffix is treated this way.
706 You can specify the input language explicitly with the @samp{-x} option:
709 @item -x @var{language}
710 Specify explicitly the @var{language} for the following input files
711 (rather than letting the compiler choose a default based on the file
712 name suffix). This option applies to all following input files until
713 the next @samp{-x} option. Possible values for @var{language} are:
715 c c-header cpp-output
717 objective-c objc-cpp-output
718 assembler assembler-with-cpp
719 f77 f77-cpp-input ratfor
722 @c Also f77-version, for internal use only.
725 Turn off any specification of a language, so that subsequent files are
726 handled according to their file name suffixes (as they are if @samp{-x}
727 has not been used at all).
729 @item -pass-exit-codes
730 Normally the @command{gcc} program will exit with the code of 1 if any
731 phase of the compiler returns a non-success return code. If you specify
732 @samp{-pass-exit-codes}, the @command{gcc} program will instead return with
733 numerically highest error produced by any phase that returned an error
737 If you only want some of the stages of compilation, you can use
738 @samp{-x} (or filename suffixes) to tell @command{gcc} where to start, and
739 one of the options @samp{-c}, @samp{-S}, or @samp{-E} to say where
740 @command{gcc} is to stop. Note that some combinations (for example,
741 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
745 Compile or assemble the source files, but do not link. The linking
746 stage simply is not done. The ultimate output is in the form of an
747 object file for each source file.
749 By default, the object file name for a source file is made by replacing
750 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
752 Unrecognized input files, not requiring compilation or assembly, are
756 Stop after the stage of compilation proper; do not assemble. The output
757 is in the form of an assembler code file for each non-assembler input
760 By default, the assembler file name for a source file is made by
761 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
763 Input files that don't require compilation are ignored.
766 Stop after the preprocessing stage; do not run the compiler proper. The
767 output is in the form of preprocessed source code, which is sent to the
770 Input files which don't require preprocessing are ignored.
772 @cindex output file option
774 Place output in file @var{file}. This applies regardless to whatever
775 sort of output is being produced, whether it be an executable file,
776 an object file, an assembler file or preprocessed C code.
778 Since only one output file can be specified, it does not make sense to
779 use @samp{-o} when compiling more than one input file, unless you are
780 producing an executable file as output.
782 If @samp{-o} is not specified, the default is to put an executable file
783 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
784 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
785 all preprocessed C source on standard output.@refill
788 Print (on standard error output) the commands executed to run the stages
789 of compilation. Also print the version number of the compiler driver
790 program and of the preprocessor and the compiler proper.
793 Use pipes rather than temporary files for communication between the
794 various stages of compilation. This fails to work on some systems where
795 the assembler is unable to read from a pipe; but the GNU assembler has
799 Print (on the standard output) a description of the command line options
800 understood by @command{gcc}. If the @option{-v} option is also specified
801 then @option{--help} will also be passed on to the various processes
802 invoked by @command{gcc}, so that they can display the command line options
803 they accept. If the @option{-W} option is also specified then command
804 line options which have no documentation associated with them will also
808 Print (on the standard output) a description of target specific command
809 line options for each tool.
813 @section Compiling C++ Programs
815 @cindex suffixes for C++ source
816 @cindex C++ source file suffixes
817 C++ source files conventionally use one of the suffixes @samp{.C},
818 @samp{.cc}, @samp{.cpp}, @samp{.c++}, @samp{.cp}, or @samp{.cxx};
819 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
820 files with these names and compiles them as C++ programs even if you
821 call the compiler the same way as for compiling C programs (usually with
822 the name @command{gcc}).
826 However, C++ programs often require class libraries as well as a
827 compiler that understands the C++ language---and under some
828 circumstances, you might want to compile programs from standard input,
829 or otherwise without a suffix that flags them as C++ programs.
830 @command{g++} is a program that calls GCC with the default language
831 set to C++, and automatically specifies linking against the C++
832 library. On many systems, @command{g++} is also
833 installed with the name @command{c++}.
835 @cindex invoking @command{g++}
836 When you compile C++ programs, you may specify many of the same
837 command-line options that you use for compiling programs in any
838 language; or command-line options meaningful for C and related
839 languages; or options that are meaningful only for C++ programs.
840 @xref{C Dialect Options,,Options Controlling C Dialect}, for
841 explanations of options for languages related to C.
842 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
843 explanations of options that are meaningful only for C++ programs.
845 @node C Dialect Options
846 @section Options Controlling C Dialect
847 @cindex dialect options
848 @cindex language dialect options
849 @cindex options, dialect
851 The following options control the dialect of C (or languages derived
852 from C, such as C++ and Objective C) that the compiler accepts:
858 In C mode, support all ISO C89 programs. In C++ mode,
859 remove GNU extensions that conflict with ISO C++.
861 This turns off certain features of GCC that are incompatible with ISO
862 C (when compiling C code), or of standard C++ (when compiling C++ code),
863 such as the @code{asm} and @code{typeof} keywords, and
864 predefined macros such as @code{unix} and @code{vax} that identify the
865 type of system you are using. It also enables the undesirable and
866 rarely used ISO trigraph feature. For the C compiler,
867 it disables recognition of C++ style @samp{//} comments as well as
868 the @code{inline} keyword.
870 The alternate keywords @code{__asm__}, @code{__extension__},
871 @code{__inline__} and @code{__typeof__} continue to work despite
872 @samp{-ansi}. You would not want to use them in an ISO C program, of
873 course, but it is useful to put them in header files that might be included
874 in compilations done with @samp{-ansi}. Alternate predefined macros
875 such as @code{__unix__} and @code{__vax__} are also available, with or
876 without @samp{-ansi}.
878 The @samp{-ansi} option does not cause non-ISO programs to be
879 rejected gratuitously. For that, @samp{-pedantic} is required in
880 addition to @samp{-ansi}. @xref{Warning Options}.
882 The macro @code{__STRICT_ANSI__} is predefined when the @samp{-ansi}
883 option is used. Some header files may notice this macro and refrain
884 from declaring certain functions or defining certain macros that the
885 ISO standard doesn't call for; this is to avoid interfering with any
886 programs that might use these names for other things.
888 Functions which would normally be builtin but do not have semantics
889 defined by ISO C (such as @code{alloca} and @code{ffs}) are not builtin
890 functions with @samp{-ansi} is used. @xref{Other Builtins,,Other
891 built-in functions provided by GNU CC}, for details of the functions
895 Determine the language standard. A value for this option must be provided;
900 Same as @option{-ansi}
903 ISO C as modified in amend. 1
906 ISO C99. Note that this standard is not yet fully supported; see
907 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information.
910 same as @option{-std=iso9899:1990}
913 same as @option{-std=iso9899:1999}
916 default, iso9899:1990 + gnu extensions
919 iso9899:1999 + gnu extensions
922 same as @option{-std=iso9899:1999}, deprecated
925 same as @option{-std=iso9899:1999}, deprecated
928 same as @option{-std=gnu99}, deprecated
932 Even when this option is not specified, you can still use some of the
933 features of newer standards in so far as they do not conflict with
934 previous C standards. For example, you may use @code{__restrict__} even
935 when @option{-std=c99} is not specified.
937 The @option{-std} options specifying some version of ISO C have the same
938 effects as @option{-ansi}, except that features that were not in ISO C89
939 but are in the specified version (for example, @samp{//} comments and
940 the @code{inline} keyword in ISO C99) are not disabled.
942 @xref{Standards,,Language Standards Supported by GCC}, for details of
943 these standard versions.
946 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
947 keyword, so that code can use these words as identifiers. You can use
948 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
949 instead. @samp{-ansi} implies @samp{-fno-asm}.
951 In C++, this switch only affects the @code{typeof} keyword, since
952 @code{asm} and @code{inline} are standard keywords. You may want to
953 use the @samp{-fno-gnu-keywords} flag instead, which has the same
954 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
955 switch only affects the @code{asm} and @code{typeof} keywords, since
956 @code{inline} is a standard keyword in ISO C99.
959 @cindex builtin functions
960 Don't recognize builtin functions that do not begin with
961 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
962 functions provided by GNU CC}, for details of the functions affected,
963 including those which are not builtin functions when @option{-ansi} or
964 @option{-std} options for strict ISO C conformance are used because they
965 do not have an ISO standard meaning.
967 GCC normally generates special code to handle certain builtin functions
968 more efficiently; for instance, calls to @code{alloca} may become single
969 instructions that adjust the stack directly, and calls to @code{memcpy}
970 may become inline copy loops. The resulting code is often both smaller
971 and faster, but since the function calls no longer appear as such, you
972 cannot set a breakpoint on those calls, nor can you change the behavior
973 of the functions by linking with a different library.
975 In C++, @samp{-fno-builtin} is always in effect. The @samp{-fbuiltin}
976 option has no effect. Therefore, in C++, the only way to get the
977 optimization benefits of builtin functions is to call the function
978 using the @samp{__builtin_} prefix. The GNU C++ Standard Library uses
979 builtin functions to implement many functions (like
980 @code{std::strchr}), so that you automatically get efficient code.
983 @cindex hosted environment
985 Assert that compilation takes place in a hosted environment. This implies
986 @samp{-fbuiltin}. A hosted environment is one in which the
987 entire standard library is available, and in which @code{main} has a return
988 type of @code{int}. Examples are nearly everything except a kernel.
989 This is equivalent to @samp{-fno-freestanding}.
992 @cindex hosted environment
994 Assert that compilation takes place in a freestanding environment. This
995 implies @samp{-fno-builtin}. A freestanding environment
996 is one in which the standard library may not exist, and program startup may
997 not necessarily be at @code{main}. The most obvious example is an OS kernel.
998 This is equivalent to @samp{-fno-hosted}.
1000 @xref{Standards,,Language Standards Supported by GCC}, for details of
1001 freestanding and hosted environments.
1004 Support ISO C trigraphs. You don't want to know about this
1005 brain-damage. The @option{-ansi} option (and @option{-std} options for
1006 strict ISO C conformance) implies @option{-trigraphs}.
1008 @cindex traditional C language
1009 @cindex C language, traditional
1011 Attempt to support some aspects of traditional C compilers.
1016 All @code{extern} declarations take effect globally even if they
1017 are written inside of a function definition. This includes implicit
1018 declarations of functions.
1021 The newer keywords @code{typeof}, @code{inline}, @code{signed}, @code{const}
1022 and @code{volatile} are not recognized. (You can still use the
1023 alternative keywords such as @code{__typeof__}, @code{__inline__}, and
1027 Comparisons between pointers and integers are always allowed.
1030 Integer types @code{unsigned short} and @code{unsigned char} promote
1031 to @code{unsigned int}.
1034 Out-of-range floating point literals are not an error.
1037 Certain constructs which ISO regards as a single invalid preprocessing
1038 number, such as @samp{0xe-0xd}, are treated as expressions instead.
1041 String ``constants'' are not necessarily constant; they are stored in
1042 writable space, and identical looking constants are allocated
1043 separately. (This is the same as the effect of
1044 @samp{-fwritable-strings}.)
1046 @cindex @code{longjmp} and automatic variables
1048 All automatic variables not declared @code{register} are preserved by
1049 @code{longjmp}. Ordinarily, GNU C follows ISO C: automatic variables
1050 not declared @code{volatile} may be clobbered.
1055 @cindex escape sequences, traditional
1056 The character escape sequences @samp{\x} and @samp{\a} evaluate as the
1057 literal characters @samp{x} and @samp{a} respectively. Without
1058 @w{@samp{-traditional}}, @samp{\x} is a prefix for the hexadecimal
1059 representation of a character, and @samp{\a} produces a bell.
1062 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
1063 if your program uses names that are normally GNU C builtin functions for
1064 other purposes of its own.
1066 You cannot use @samp{-traditional} if you include any header files that
1067 rely on ISO C features. Some vendors are starting to ship systems with
1068 ISO C header files and you cannot use @samp{-traditional} on such
1069 systems to compile files that include any system headers.
1071 The @samp{-traditional} option also enables @samp{-traditional-cpp},
1072 which is described next.
1074 @item -traditional-cpp
1075 Attempt to support some aspects of traditional C preprocessors.
1080 Comments convert to nothing at all, rather than to a space. This allows
1081 traditional token concatenation.
1084 In a preprocessing directive, the @samp{#} symbol must appear as the first
1085 character of a line.
1088 Macro arguments are recognized within string constants in a macro
1089 definition (and their values are stringified, though without additional
1090 quote marks, when they appear in such a context). The preprocessor
1091 always considers a string constant to end at a newline.
1094 @cindex detecting @w{@samp{-traditional}}
1095 The predefined macro @code{__STDC__} is not defined when you use
1096 @samp{-traditional}, but @code{__GNUC__} is (since the GNU extensions
1097 which @code{__GNUC__} indicates are not affected by
1098 @samp{-traditional}). If you need to write header files that work
1099 differently depending on whether @samp{-traditional} is in use, by
1100 testing both of these predefined macros you can distinguish four
1101 situations: GNU C, traditional GNU C, other ISO C compilers, and other
1102 old C compilers. The predefined macro @code{__STDC_VERSION__} is also
1103 not defined when you use @samp{-traditional}. @xref{Standard
1104 Predefined,,Standard Predefined Macros,cpp.info,The C Preprocessor},
1105 for more discussion of these and other predefined macros.
1108 @cindex string constants vs newline
1109 @cindex newline vs string constants
1110 The preprocessor considers a string constant to end at a newline (unless
1111 the newline is escaped with @samp{\}). (Without @w{@samp{-traditional}},
1112 string constants can contain the newline character as typed.)
1115 @item -fcond-mismatch
1116 Allow conditional expressions with mismatched types in the second and
1117 third arguments. The value of such an expression is void. This option
1118 is not supported for C++.
1120 @item -funsigned-char
1121 Let the type @code{char} be unsigned, like @code{unsigned char}.
1123 Each kind of machine has a default for what @code{char} should
1124 be. It is either like @code{unsigned char} by default or like
1125 @code{signed char} by default.
1127 Ideally, a portable program should always use @code{signed char} or
1128 @code{unsigned char} when it depends on the signedness of an object.
1129 But many programs have been written to use plain @code{char} and
1130 expect it to be signed, or expect it to be unsigned, depending on the
1131 machines they were written for. This option, and its inverse, let you
1132 make such a program work with the opposite default.
1134 The type @code{char} is always a distinct type from each of
1135 @code{signed char} or @code{unsigned char}, even though its behavior
1136 is always just like one of those two.
1139 Let the type @code{char} be signed, like @code{signed char}.
1141 Note that this is equivalent to @samp{-fno-unsigned-char}, which is
1142 the negative form of @samp{-funsigned-char}. Likewise, the option
1143 @samp{-fno-signed-char} is equivalent to @samp{-funsigned-char}.
1145 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
1146 if your program uses names that are normally GNU C builtin functions for
1147 other purposes of its own.
1149 You cannot use @samp{-traditional} if you include any header files that
1150 rely on ISO C features. Some vendors are starting to ship systems with
1151 ISO C header files and you cannot use @samp{-traditional} on such
1152 systems to compile files that include any system headers.
1154 @item -fsigned-bitfields
1155 @itemx -funsigned-bitfields
1156 @itemx -fno-signed-bitfields
1157 @itemx -fno-unsigned-bitfields
1158 These options control whether a bitfield is signed or unsigned, when the
1159 declaration does not use either @code{signed} or @code{unsigned}. By
1160 default, such a bitfield is signed, because this is consistent: the
1161 basic integer types such as @code{int} are signed types.
1163 However, when @samp{-traditional} is used, bitfields are all unsigned
1166 @item -fwritable-strings
1167 Store string constants in the writable data segment and don't uniquize
1168 them. This is for compatibility with old programs which assume they can
1169 write into string constants. The option @samp{-traditional} also has
1172 Writing into string constants is a very bad idea; ``constants'' should
1175 @item -fallow-single-precision
1176 Do not promote single precision math operations to double precision,
1177 even when compiling with @samp{-traditional}.
1179 Traditional K&R C promotes all floating point operations to double
1180 precision, regardless of the sizes of the operands. On the
1181 architecture for which you are compiling, single precision may be faster
1182 than double precision. If you must use @samp{-traditional}, but want
1183 to use single precision operations when the operands are single
1184 precision, use this option. This option has no effect when compiling
1185 with ISO or GNU C conventions (the default).
1188 Override the underlying type for @samp{wchar_t} to be @samp{short
1189 unsigned int} instead of the default for the target. This option is
1190 useful for building programs to run under WINE.
1193 @node C++ Dialect Options
1194 @section Options Controlling C++ Dialect
1196 @cindex compiler options, C++
1197 @cindex C++ options, command line
1198 @cindex options, C++
1199 This section describes the command-line options that are only meaningful
1200 for C++ programs; but you can also use most of the GNU compiler options
1201 regardless of what language your program is in. For example, you
1202 might compile a file @code{firstClass.C} like this:
1205 g++ -g -frepo -O -c firstClass.C
1209 In this example, only @samp{-frepo} is an option meant
1210 only for C++ programs; you can use the other options with any
1211 language supported by GCC.
1213 Here is a list of options that are @emph{only} for compiling C++ programs:
1216 @item -fno-access-control
1217 Turn off all access checking. This switch is mainly useful for working
1218 around bugs in the access control code.
1221 Check that the pointer returned by @code{operator new} is non-null
1222 before attempting to modify the storage allocated. The current Working
1223 Paper requires that @code{operator new} never return a null pointer, so
1224 this check is normally unnecessary.
1226 An alternative to using this option is to specify that your
1227 @code{operator new} does not throw any exceptions; if you declare it
1228 @samp{throw()}, g++ will check the return value. See also @samp{new
1231 @item -fconserve-space
1232 Put uninitialized or runtime-initialized global variables into the
1233 common segment, as C does. This saves space in the executable at the
1234 cost of not diagnosing duplicate definitions. If you compile with this
1235 flag and your program mysteriously crashes after @code{main()} has
1236 completed, you may have an object that is being destroyed twice because
1237 two definitions were merged.
1239 This option is no longer useful on most targets, now that support has
1240 been added for putting variables into BSS without making them common.
1242 @item -fno-const-strings
1243 Give string constants type @code{char *} instead of type @code{const
1244 char *}. By default, G++ uses type @code{const char *} as required by
1245 the standard. Even if you use @samp{-fno-const-strings}, you cannot
1246 actually modify the value of a string constant, unless you also use
1247 @samp{-fwritable-strings}.
1249 This option might be removed in a future release of G++. For maximum
1250 portability, you should structure your code so that it works with
1251 string constants that have type @code{const char *}.
1253 @item -fdollars-in-identifiers
1254 Accept @samp{$} in identifiers. You can also explicitly prohibit use of
1255 @samp{$} with the option @samp{-fno-dollars-in-identifiers}. (GNU C allows
1256 @samp{$} by default on most target systems, but there are a few exceptions.)
1257 Traditional C allowed the character @samp{$} to form part of
1258 identifiers. However, ISO C and C++ forbid @samp{$} in identifiers.
1260 @item -fno-elide-constructors
1261 The C++ standard allows an implementation to omit creating a temporary
1262 which is only used to initialize another object of the same type.
1263 Specifying this option disables that optimization, and forces g++ to
1264 call the copy constructor in all cases.
1266 @item -fno-enforce-eh-specs
1267 Don't check for violation of exception specifications at runtime. This
1268 option violates the C++ standard, but may be useful for reducing code
1269 size in production builds, much like defining @samp{NDEBUG}. The compiler
1270 will still optimize based on the exception specifications.
1272 @item -fexternal-templates
1273 Cause template instantiations to obey @samp{#pragma interface} and
1274 @samp{implementation}; template instances are emitted or not according
1275 to the location of the template definition. @xref{Template
1276 Instantiation}, for more information.
1278 This option is deprecated.
1280 @item -falt-external-templates
1281 Similar to -fexternal-templates, but template instances are emitted or
1282 not according to the place where they are first instantiated.
1283 @xref{Template Instantiation}, for more information.
1285 This option is deprecated.
1288 @itemx -fno-for-scope
1289 If -ffor-scope is specified, the scope of variables declared in
1290 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1291 as specified by the C++ standard.
1292 If -fno-for-scope is specified, the scope of variables declared in
1293 a @i{for-init-statement} extends to the end of the enclosing scope,
1294 as was the case in old versions of gcc, and other (traditional)
1295 implementations of C++.
1297 The default if neither flag is given to follow the standard,
1298 but to allow and give a warning for old-style code that would
1299 otherwise be invalid, or have different behavior.
1301 @item -fno-gnu-keywords
1302 Do not recognize @code{typeof} as a keyword, so that code can use this
1303 word as an identifier. You can use the keyword @code{__typeof__} instead.
1304 @samp{-ansi} implies @samp{-fno-gnu-keywords}.
1307 Treat the @code{namespace std} as a namespace, instead of ignoring
1308 it. For compatibility with earlier versions of g++, the compiler will,
1309 by default, ignore @code{namespace-declarations},
1310 @code{using-declarations}, @code{using-directives}, and
1311 @code{namespace-names}, if they involve @code{std}.
1313 @item -fno-implicit-templates
1314 Never emit code for non-inline templates which are instantiated
1315 implicitly (i.e. by use); only emit code for explicit instantiations.
1316 @xref{Template Instantiation}, for more information.
1318 @item -fno-implicit-inline-templates
1319 Don't emit code for implicit instantiations of inline templates, either.
1320 The default is to handle inlines differently so that compiles with and
1321 without optimization will need the same set of explicit instantiations.
1323 @item -fno-implement-inlines
1324 To save space, do not emit out-of-line copies of inline functions
1325 controlled by @samp{#pragma implementation}. This will cause linker
1326 errors if these functions are not inlined everywhere they are called.
1328 @item -fms-extensions
1329 Disable pedantic warnings about constructs used in MFC, such as implicit
1330 int and getting a pointer to member function via non-standard syntax.
1332 @item -fno-nonansi-builtins
1333 Disable builtin declarations of functions that are not mandated by
1334 ANSI/ISO C. These include @code{ffs}, @code{alloca}, @code{_exit},
1335 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1337 @item -fno-operator-names
1338 Do not treat the operator name keywords @code{and}, @code{bitand},
1339 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1340 synonyms as keywords.
1342 @item -fno-optional-diags
1343 Disable diagnostics that the standard says a compiler does not need to
1344 issue. Currently, the only such diagnostic issued by g++ is the one for
1345 a name having multiple meanings within a class.
1348 Downgrade messages about nonconformant code from errors to warnings. By
1349 default, g++ effectively sets @samp{-pedantic-errors} without
1350 @samp{-pedantic}; this option reverses that. This behavior and this
1351 option are superseded by @samp{-pedantic}, which works as it does for GNU C.
1354 Enable automatic template instantiation. This option also implies
1355 @samp{-fno-implicit-templates}. @xref{Template Instantiation}, for more
1359 Disable generation of information about every class with virtual
1360 functions for use by the C++ runtime type identification features
1361 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1362 of the language, you can save some space by using this flag. Note that
1363 exception handling uses the same information, but it will generate it as
1367 Emit statistics about front-end processing at the end of the compilation.
1368 This information is generally only useful to the G++ development team.
1370 @item -ftemplate-depth-@var{n}
1371 Set the maximum instantiation depth for template classes to @var{n}.
1372 A limit on the template instantiation depth is needed to detect
1373 endless recursions during template class instantiation. ANSI/ISO C++
1374 conforming programs must not rely on a maximum depth greater than 17.
1376 @item -fuse-cxa-atexit
1377 Register destructors for objects with static storage duration with the
1378 @code{__cxa_atexit} function rather than the @code{atexit} function.
1379 This option is required for fully standards-compliant handling of static
1380 destructors, but will only work if your C library supports
1381 @code{__cxa_atexit}.
1384 Emit special relocations for vtables and virtual function references
1385 so that the linker can identify unused virtual functions and zero out
1386 vtable slots that refer to them. This is most useful with
1387 @samp{-ffunction-sections} and @samp{-Wl,--gc-sections}, in order to
1388 also discard the functions themselves.
1390 This optimization requires GNU as and GNU ld. Not all systems support
1391 this option. @samp{-Wl,--gc-sections} is ignored without @samp{-static}.
1394 Do not use weak symbol support, even if it is provided by the linker.
1395 By default, G++ will use weak symbols if they are available. This
1396 option exists only for testing, and should not be used by end-users;
1397 it will result in inferior code and has no benefits. This option may
1398 be removed in a future release of G++.
1401 Do not search for header files in the standard directories specific to
1402 C++, but do still search the other standard directories. (This option
1403 is used when building the C++ library.)
1406 In addition, these optimization, warning, and code generation options
1407 have meanings only for C++ programs:
1410 @item -fno-default-inline
1411 Do not assume @samp{inline} for functions defined inside a class scope.
1412 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1413 functions will have linkage like inline functions; they just won't be
1416 @item -Wctor-dtor-privacy (C++ only)
1417 Warn when a class seems unusable, because all the constructors or
1418 destructors in a class are private and the class has no friends or
1419 public static member functions.
1421 @item -Wnon-virtual-dtor (C++ only)
1422 Warn when a class declares a non-virtual destructor that should probably
1423 be virtual, because it looks like the class will be used polymorphically.
1425 @item -Wreorder (C++ only)
1426 @cindex reordering, warning
1427 @cindex warning for reordering of member initializers
1428 Warn when the order of member initializers given in the code does not
1429 match the order in which they must be executed. For instance:
1435 A(): j (0), i (1) @{ @}
1439 Here the compiler will warn that the member initializers for @samp{i}
1440 and @samp{j} will be rearranged to match the declaration order of the
1444 The following @samp{-W@dots{}} options are not affected by @samp{-Wall}.
1447 @item -Weffc++ (C++ only)
1448 Warn about violations of various style guidelines from Scott Meyers'
1449 @cite{Effective C++} books. If you use this option, you should be aware
1450 that the standard library headers do not obey all of these guidelines;
1451 you can use @samp{grep -v} to filter out those warnings.
1453 @item -Wno-deprecated (C++ only)
1454 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1456 @item -Wno-non-template-friend (C++ only)
1457 Disable warnings when non-templatized friend functions are declared
1458 within a template. With the advent of explicit template specification
1459 support in g++, if the name of the friend is an unqualified-id (ie,
1460 @samp{friend foo(int)}), the C++ language specification demands that the
1461 friend declare or define an ordinary, nontemplate function. (Section
1462 14.5.3). Before g++ implemented explicit specification, unqualified-ids
1463 could be interpreted as a particular specialization of a templatized
1464 function. Because this non-conforming behavior is no longer the default
1465 behavior for g++, @samp{-Wnon-template-friend} allows the compiler to
1466 check existing code for potential trouble spots, and is on by default.
1467 This new compiler behavior can be turned off with
1468 @samp{-Wno-non-template-friend} which keeps the conformant compiler code
1469 but disables the helpful warning.
1471 @item -Wold-style-cast (C++ only)
1472 Warn if an old-style (C-style) cast is used within a C++ program. The
1473 new-style casts (@samp{static_cast}, @samp{reinterpret_cast}, and
1474 @samp{const_cast}) are less vulnerable to unintended effects, and much
1477 @item -Woverloaded-virtual (C++ only)
1478 @cindex overloaded virtual fn, warning
1479 @cindex warning for overloaded virtual fn
1480 Warn when a function declaration hides virtual functions from a
1481 base class. For example, in:
1488 struct B: public A @{
1493 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1501 will fail to compile.
1503 @item -Wno-pmf-conversions (C++ only)
1504 Disable the diagnostic for converting a bound pointer to member function
1507 @item -Wsign-promo (C++ only)
1508 Warn when overload resolution chooses a promotion from unsigned or
1509 enumeral type to a signed type over a conversion to an unsigned type of
1510 the same size. Previous versions of g++ would try to preserve
1511 unsignedness, but the standard mandates the current behavior.
1513 @item -Wsynth (C++ only)
1514 @cindex warning for synthesized methods
1515 @cindex synthesized methods, warning
1516 Warn when g++'s synthesis behavior does not match that of cfront. For
1522 A& operator = (int);
1532 In this example, g++ will synthesize a default @samp{A& operator =
1533 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1536 @node Objective-C Dialect Options
1537 @section Options Controlling Objective-C Dialect
1539 @cindex compiler options, Objective-C
1540 @cindex Objective-C options, command line
1541 @cindex options, Objective-C
1542 This section describes the command-line options that are only meaningful
1543 for Objective-C programs; but you can also use most of the GNU compiler
1544 options regardless of what language your program is in. For example,
1545 you might compile a file @code{some_class.m} like this:
1548 gcc -g -fgnu-runtime -O -c some_class.m
1552 In this example, only @samp{-fgnu-runtime} is an option meant only for
1553 Objective-C programs; you can use the other options with any language
1556 Here is a list of options that are @emph{only} for compiling Objective-C
1560 @item -fconstant-string-class=@var{class name}
1561 Use @var{class name} as the name of the class to instantiate for each
1562 literal string specified with the syntax @code{@@"..."}. The default
1563 class name is @code{NXConstantString}.
1566 Generate object code compatible with the standard GNU Objective-C
1567 runtime. This is the default for most types of systems.
1569 @item -fnext-runtime
1570 Generate output compatible with the NeXT runtime. This is the default
1571 for NeXT-based systems, including Darwin and Mac OS X.
1574 Dump interface declarations for all classes seen in the source file to a
1575 file named @file{@var{sourcename}.decl}.
1578 Do not warn if methods required by a protocol are not implemented
1579 in the class adopting it.
1582 Warn if a selector has multiple methods of different types defined.
1584 @c not documented because only avail via -Wp
1585 @c @item -print-objc-runtime-info
1589 @node Language Independent Options
1590 @section Options to Control Diagnostic Messages Formatting
1591 @cindex options to control diagnostics formatting
1592 @cindex diagnostic messages
1593 @cindex message formatting
1595 Traditionally, diagnostic messages have been formatted irrespective of
1596 the output device's aspect (e.g. its width, ...). The options described
1597 below can be used to control the diagnostic messages formatting
1598 algorithm, e.g. how many characters per line, how often source location
1599 information should be reported. Right now, only the C++ front-end can
1600 honor these options. However it is expected, in the near future, that
1601 the remaining front-ends would be able to digest them correctly.
1604 @item -fmessage-length=@var{n}
1605 Try to format error messages so that they fit on lines of about @var{n}
1606 characters. The default is 72 characters for g++ and 0 for the rest of
1607 the front-ends supported by GCC. If @var{n} is zero, then no
1608 line-wrapping will be done; each error message will appear on a single
1611 @item -fdiagnostics-show-location=once
1612 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1613 reporter to emit @emph{once} source location information; that is, in
1614 case the message is too long to fit on a single physical line and has to
1615 be wrapped, the source location won't be emitted (as prefix) again,
1616 over and over, in subsequent continuation lines. This is the default
1619 @item -fdiagnostics-show-location=every-line
1620 Only meaningful in line-wrapping mode. Instructs the diagnostic
1621 messages reporter to emit the same source location information (as
1622 prefix) for physical lines that result from the process of breaking a
1623 a message which is too long to fit on a single line.
1627 @node Warning Options
1628 @section Options to Request or Suppress Warnings
1629 @cindex options to control warnings
1630 @cindex warning messages
1631 @cindex messages, warning
1632 @cindex suppressing warnings
1634 Warnings are diagnostic messages that report constructions which
1635 are not inherently erroneous but which are risky or suggest there
1636 may have been an error.
1638 You can request many specific warnings with options beginning @samp{-W},
1639 for example @samp{-Wimplicit} to request warnings on implicit
1640 declarations. Each of these specific warning options also has a
1641 negative form beginning @samp{-Wno-} to turn off warnings;
1642 for example, @samp{-Wno-implicit}. This manual lists only one of the
1643 two forms, whichever is not the default.
1645 These options control the amount and kinds of warnings produced by GCC:
1648 @cindex syntax checking
1650 Check the code for syntax errors, but don't do anything beyond that.
1653 Issue all the warnings demanded by strict ISO C and ISO C++;
1654 reject all programs that use forbidden extensions, and some other
1655 programs that do not follow ISO C and ISO C++. For ISO C, follows the
1656 version of the ISO C standard specified by any @samp{-std} option used.
1658 Valid ISO C and ISO C++ programs should compile properly with or without
1659 this option (though a rare few will require @option{-ansi} or a
1660 @option{-std} option specifying the required version of ISO C). However,
1661 without this option, certain GNU extensions and traditional C and C++
1662 features are supported as well. With this option, they are rejected.
1664 @samp{-pedantic} does not cause warning messages for use of the
1665 alternate keywords whose names begin and end with @samp{__}. Pedantic
1666 warnings are also disabled in the expression that follows
1667 @code{__extension__}. However, only system header files should use
1668 these escape routes; application programs should avoid them.
1669 @xref{Alternate Keywords}.
1671 Some users try to use @samp{-pedantic} to check programs for strict ISO
1672 C conformance. They soon find that it does not do quite what they want:
1673 it finds some non-ISO practices, but not all---only those for which
1674 ISO C @emph{requires} a diagnostic, and some others for which
1675 diagnostics have been added.
1677 A feature to report any failure to conform to ISO C might be useful in
1678 some instances, but would require considerable additional work and would
1679 be quite different from @samp{-pedantic}. We don't have plans to
1680 support such a feature in the near future.
1682 Where the standard specified with @option{-std} represents a GNU
1683 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
1684 corresponding @dfn{base standard}, the version of ISO C on which the GNU
1685 extended dialect is based. Warnings from @option{-pedantic} are given
1686 where they are required by the base standard. (It would not make sense
1687 for such warnings to be given only for features not in the specified GNU
1688 C dialect, since by definition the GNU dialects of C include all
1689 features the compiler supports with the given option, and there would be
1690 nothing to warn about.)
1692 @item -pedantic-errors
1693 Like @samp{-pedantic}, except that errors are produced rather than
1697 Inhibit all warning messages.
1700 Inhibit warning messages about the use of @samp{#import}.
1702 @item -Wchar-subscripts
1703 Warn if an array subscript has type @code{char}. This is a common cause
1704 of error, as programmers often forget that this type is signed on some
1708 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
1709 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
1712 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
1713 the arguments supplied have types appropriate to the format string
1714 specified, and that the conversions specified in the format string make
1715 sense. This includes standard functions, and others specified by format
1716 attributes (@pxref{Function Attributes}), in the @code{printf},
1717 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
1718 not in the C standard) families.
1720 The formats are checked against the format features supported by GNU
1721 libc version 2.2. These include all ISO C89 and C99 features, as well
1722 as features from the Single Unix Specification and some BSD and GNU
1723 extensions. Other library implementations may not support all these
1724 features; GCC does not support warning about features that go beyond a
1725 particular library's limitations. However, if @samp{-pedantic} is used
1726 with @samp{-Wformat}, warnings will be given about format features not
1727 in the selected standard version (but not for @code{strfmon} formats,
1728 since those are not in any version of the C standard). @xref{C Dialect
1729 Options,,Options Controlling C Dialect}.
1731 @samp{-Wformat} is included in @samp{-Wall}. For more control over some
1732 aspects of format checking, the options @samp{-Wno-format-y2k},
1733 @samp{-Wno-format-extra-args}, @samp{-Wformat-nonliteral},
1734 @samp{-Wformat-security} and @samp{-Wformat=2} are available, but are
1735 not included in @samp{-Wall}.
1737 @item -Wno-format-y2k
1738 If @samp{-Wformat} is specified, do not warn about @code{strftime}
1739 formats which may yield only a two-digit year.
1741 @item -Wno-format-extra-args
1742 If @samp{-Wformat} is specified, do not warn about excess arguments to a
1743 @code{printf} or @code{scanf} format function. The C standard specifies
1744 that such arguments are ignored.
1746 @item -Wformat-nonliteral
1747 If @samp{-Wformat} is specified, also warn if the format string is not a
1748 string literal and so cannot be checked, unless the format function
1749 takes its format arguments as a @code{va_list}.
1751 @item -Wformat-security
1752 If @samp{-Wformat} is specified, also warn about uses of format
1753 functions that represent possible security problems. At present, this
1754 warns about calls to @code{printf} and @code{scanf} functions where the
1755 format string is not a string literal and there are no format arguments,
1756 as in @code{printf (foo);}. This may be a security hole if the format
1757 string came from untrusted input and contains @samp{%n}. (This is
1758 currently a subset of what @samp{-Wformat-nonliteral} warns about, but
1759 in future warnings may be added to @samp{-Wformat-security} that are not
1760 included in @samp{-Wformat-nonliteral}.)
1763 Enable @samp{-Wformat} plus format checks not included in
1764 @samp{-Wformat}. Currently equivalent to @samp{-Wformat
1765 -Wformat-nonliteral -Wformat-security}.
1767 @item -Wimplicit-int
1768 Warn when a declaration does not specify a type.
1770 @item -Wimplicit-function-declaration
1771 @itemx -Werror-implicit-function-declaration
1772 Give a warning (or error) whenever a function is used before being
1776 Same as @samp{-Wimplicit-int} and @samp{-Wimplicit-function-}@*
1780 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
1781 function with external linkage, returning int, taking either zero
1782 arguments, two, or three arguments of appropriate types.
1784 @item -Wmissing-braces
1785 Warn if an aggregate or union initializer is not fully bracketed. In
1786 the following example, the initializer for @samp{a} is not fully
1787 bracketed, but that for @samp{b} is fully bracketed.
1790 int a[2][2] = @{ 0, 1, 2, 3 @};
1791 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
1795 Warn if a multicharacter constant (@samp{'FOOF'}) is used. Usually they
1796 indicate a typo in the user's code, as they have implementation-defined
1797 values, and should not be used in portable code.
1800 Warn if parentheses are omitted in certain contexts, such
1801 as when there is an assignment in a context where a truth value
1802 is expected, or when operators are nested whose precedence people
1803 often get confused about.
1805 Also warn about constructions where there may be confusion to which
1806 @code{if} statement an @code{else} branch belongs. Here is an example of
1819 In C, every @code{else} branch belongs to the innermost possible @code{if}
1820 statement, which in this example is @code{if (b)}. This is often not
1821 what the programmer expected, as illustrated in the above example by
1822 indentation the programmer chose. When there is the potential for this
1823 confusion, GNU C will issue a warning when this flag is specified.
1824 To eliminate the warning, add explicit braces around the innermost
1825 @code{if} statement so there is no way the @code{else} could belong to
1826 the enclosing @code{if}. The resulting code would look like this:
1840 @item -Wsequence-point
1841 Warn about code that may have undefined semantics because of violations
1842 of sequence point rules in the C standard.
1844 The C standard defines the order in which expressions in a C program are
1845 evaluated in terms of @dfn{sequence points}, which represent a partial
1846 ordering between the execution of parts of the program: those executed
1847 before the sequence point, and those executed after it. These occur
1848 after the evaluation of a full expression (one which is not part of a
1849 larger expression), after the evaluation of the first operand of a
1850 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
1851 function is called (but after the evaluation of its arguments and the
1852 expression denoting the called function), and in certain other places.
1853 Other than as expressed by the sequence point rules, the order of
1854 evaluation of subexpressions of an expression is not specified. All
1855 these rules describe only a partial order rather than a total order,
1856 since, for example, if two functions are called within one expression
1857 with no sequence point between them, the order in which the functions
1858 are called is not specified. However, the standards committee have
1859 ruled that function calls do not overlap.
1861 It is not specified when between sequence points modifications to the
1862 values of objects take effect. Programs whose behavior depends on this
1863 have undefined behavior; the C standard specifies that ``Between the
1864 previous and next sequence point an object shall have its stored value
1865 modified at most once by the evaluation of an expression. Furthermore,
1866 the prior value shall be read only to determine the value to be
1867 stored.''. If a program breaks these rules, the results on any
1868 particular implementation are entirely unpredictable.
1870 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
1871 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
1872 diagnosed by this option, and it may give an occasional false positive
1873 result, but in general it has been found fairly effective at detecting
1874 this sort of problem in programs.
1876 The present implementation of this option only works for C programs. A
1877 future implementation may also work for C++ programs.
1879 There is some controversy over the precise meaning of the sequence point
1880 rules in subtle cases. Alternative formal definitions may be found in
1881 Clive Feather's ``Annex S''
1882 @w{@uref{http://wwwold.dkuug.dk/JTC1/SC22/WG14/www/docs/n925.htm}} and in
1883 Michael Norrish's thesis
1884 @w{@uref{http://www.cl.cam.ac.uk/users/mn200/PhD/thesis-report.ps.gz}}.
1885 Other discussions are by Raymond Mak
1886 @w{@uref{http://wwwold.dkuug.dk/JTC1/SC22/WG14/www/docs/n926.htm}} and
1888 @w{@uref{http://wwwold.dkuug.dk/JTC1/SC22/WG14/www/docs/n927.htm}}.
1891 Warn whenever a function is defined with a return-type that defaults to
1892 @code{int}. Also warn about any @code{return} statement with no
1893 return-value in a function whose return-type is not @code{void}.
1895 For C++, a function without return type always produces a diagnostic
1896 message, even when @samp{-Wno-return-type} is specified. The only
1897 exceptions are @samp{main} and functions defined in system headers.
1900 Warn whenever a @code{switch} statement has an index of enumeral type
1901 and lacks a @code{case} for one or more of the named codes of that
1902 enumeration. (The presence of a @code{default} label prevents this
1903 warning.) @code{case} labels outside the enumeration range also
1904 provoke warnings when this option is used.
1907 Warn if any trigraphs are encountered that might change the meaning of
1908 the program (trigraphs within comments are not warned about).
1910 @item -Wunused-function
1911 Warn whenever a static function is declared but not defined or a
1912 non\-inline static function is unused.
1914 @item -Wunused-label
1915 Warn whenever a label is declared but not used.
1917 To suppress this warning use the @samp{unused} attribute
1918 (@pxref{Variable Attributes}).
1920 @item -Wunused-parameter
1921 Warn whenever a function parameter is unused aside from its declaration.
1923 To suppress this warning use the @samp{unused} attribute
1924 (@pxref{Variable Attributes}).
1926 @item -Wunused-variable
1927 Warn whenever a local variable or non-constant static variable is unused
1928 aside from its declaration
1930 To suppress this warning use the @samp{unused} attribute
1931 (@pxref{Variable Attributes}).
1933 @item -Wunused-value
1934 Warn whenever a statement computes a result that is explicitly not used.
1936 To suppress this warning cast the expression to @samp{void}.
1939 All all the above @samp{-Wunused} options combined.
1941 In order to get a warning about an unused function parameter, you must
1942 either specify @samp{-W -Wunused} or separately specify
1943 @samp{-Wunused-parameter}.
1945 @item -Wuninitialized
1946 Warn if an automatic variable is used without first being initialized or
1947 if a variable may be clobbered by a @code{setjmp} call.
1949 These warnings are possible only in optimizing compilation,
1950 because they require data flow information that is computed only
1951 when optimizing. If you don't specify @samp{-O}, you simply won't
1954 These warnings occur only for variables that are candidates for
1955 register allocation. Therefore, they do not occur for a variable that
1956 is declared @code{volatile}, or whose address is taken, or whose size
1957 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
1958 structures, unions or arrays, even when they are in registers.
1960 Note that there may be no warning about a variable that is used only
1961 to compute a value that itself is never used, because such
1962 computations may be deleted by data flow analysis before the warnings
1965 These warnings are made optional because GCC is not smart
1966 enough to see all the reasons why the code might be correct
1967 despite appearing to have an error. Here is one example of how
1986 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
1987 always initialized, but GCC doesn't know this. Here is
1988 another common case:
1993 if (change_y) save_y = y, y = new_y;
1995 if (change_y) y = save_y;
2000 This has no bug because @code{save_y} is used only if it is set.
2002 @cindex @code{longjmp} warnings
2003 This option also warns when a non-volatile automatic variable might be
2004 changed by a call to @code{longjmp}. These warnings as well are possible
2005 only in optimizing compilation.
2007 The compiler sees only the calls to @code{setjmp}. It cannot know
2008 where @code{longjmp} will be called; in fact, a signal handler could
2009 call it at any point in the code. As a result, you may get a warning
2010 even when there is in fact no problem because @code{longjmp} cannot
2011 in fact be called at the place which would cause a problem.
2013 Some spurious warnings can be avoided if you declare all the functions
2014 you use that never return as @code{noreturn}. @xref{Function
2017 @item -Wreorder (C++ only)
2018 @cindex reordering, warning
2019 @cindex warning for reordering of member initializers
2020 Warn when the order of member initializers given in the code does not
2021 match the order in which they must be executed. For instance:
2023 @item -Wunknown-pragmas
2024 @cindex warning for unknown pragmas
2025 @cindex unknown pragmas, warning
2026 @cindex pragmas, warning of unknown
2027 Warn when a #pragma directive is encountered which is not understood by
2028 GCC. If this command line option is used, warnings will even be issued
2029 for unknown pragmas in system header files. This is not the case if
2030 the warnings were only enabled by the @samp{-Wall} command line option.
2033 All of the above @samp{-W} options combined. This enables all the
2034 warnings about constructions that some users consider questionable, and
2035 that are easy to avoid (or modify to prevent the warning), even in
2036 conjunction with macros.
2038 @item -Wsystem-headers
2039 @cindex warnings from system headers
2040 @cindex system headers, warnings from
2041 Print warning messages for constructs found in system header files.
2042 Warnings from system headers are normally suppressed, on the assumption
2043 that they usually do not indicate real problems and would only make the
2044 compiler output harder to read. Using this command line option tells
2045 GCC to emit warnings from system headers as if they occurred in user
2046 code. However, note that using @samp{-Wall} in conjunction with this
2047 option will @emph{not} warn about unknown pragmas in system
2048 headers---for that, @samp{-Wunknown-pragmas} must also be used.
2051 The following @samp{-W@dots{}} options are not implied by @samp{-Wall}.
2052 Some of them warn about constructions that users generally do not
2053 consider questionable, but which occasionally you might wish to check
2054 for; others warn about constructions that are necessary or hard to avoid
2055 in some cases, and there is no simple way to modify the code to suppress
2060 Print extra warning messages for these events:
2064 A function can return either with or without a value. (Falling
2065 off the end of the function body is considered returning without
2066 a value.) For example, this function would evoke such a
2080 An expression-statement or the left-hand side of a comma expression
2081 contains no side effects.
2082 To suppress the warning, cast the unused expression to void.
2083 For example, an expression such as @samp{x[i,j]} will cause a warning,
2084 but @samp{x[(void)i,j]} will not.
2087 An unsigned value is compared against zero with @samp{<} or @samp{<=}.
2090 A comparison like @samp{x<=y<=z} appears; this is equivalent to
2091 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
2092 that of ordinary mathematical notation.
2095 Storage-class specifiers like @code{static} are not the first things in
2096 a declaration. According to the C Standard, this usage is obsolescent.
2099 The return type of a function has a type qualifier such as @code{const}.
2100 Such a type qualifier has no effect, since the value returned by a
2101 function is not an lvalue. (But don't warn about the GNU extension of
2102 @code{volatile void} return types. That extension will be warned about
2103 if @samp{-pedantic} is specified.)
2106 If @samp{-Wall} or @samp{-Wunused} is also specified, warn about unused
2110 A comparison between signed and unsigned values could produce an
2111 incorrect result when the signed value is converted to unsigned.
2112 (But don't warn if @samp{-Wno-sign-compare} is also specified.)
2115 An aggregate has a partly bracketed initializer.
2116 For example, the following code would evoke such a warning,
2117 because braces are missing around the initializer for @code{x.h}:
2120 struct s @{ int f, g; @};
2121 struct t @{ struct s h; int i; @};
2122 struct t x = @{ 1, 2, 3 @};
2126 An aggregate has an initializer which does not initialize all members.
2127 For example, the following code would cause such a warning, because
2128 @code{x.h} would be implicitly initialized to zero:
2131 struct s @{ int f, g, h; @};
2132 struct s x = @{ 3, 4 @};
2137 Warn if floating point values are used in equality comparisons.
2139 The idea behind this is that sometimes it is convenient (for the
2140 programmer) to consider floating-point values as approximations to
2141 infinitely precise real numbers. If you are doing this, then you need
2142 to compute (by analysing the code, or in some other way) the maximum or
2143 likely maximum error that the computation introduces, and allow for it
2144 when performing comparisons (and when producing output, but that's a
2145 different problem). In particular, instead of testing for equality, you
2146 would check to see whether the two values have ranges that overlap; and
2147 this is done with the relational operators, so equality comparisons are
2150 @item -Wtraditional (C only)
2151 Warn about certain constructs that behave differently in traditional and
2152 ISO C. Also warn about ISO C constructs that have no traditional C
2153 equivalent, and/or problematic constructs which should be avoided.
2157 Macro parameters that appear within string literals in the macro body.
2158 In traditional C macro replacement takes place within string literals,
2159 but does not in ISO C.
2162 In traditional C, some preprocessor directives did not exist.
2163 Traditional preprocessors would only consider a line to be a directive
2164 if the @samp{#} appeared in column 1 on the line. Therefore
2165 @samp{-Wtraditional} warns about directives that traditional C
2166 understands but would ignore because the @samp{#} does not appear as the
2167 first character on the line. It also suggests you hide directives like
2168 @samp{#pragma} not understood by traditional C by indenting them. Some
2169 traditional implementations would not recognise @samp{#elif}, so it
2170 suggests avoiding it altogether.
2173 A function-like macro that appears without arguments.
2176 The unary plus operator.
2179 The `U' integer constant suffix, or the `F' or `L' floating point
2180 constant suffixes. (Traditonal C does support the `L' suffix on integer
2181 constants.) Note, these suffixes appear in macros defined in the system
2182 headers of most modern systems, e.g. the _MIN/_MAX macros in limits.h.
2183 Use of these macros in user code might normally lead to spurious
2184 warnings, however gcc's integrated preprocessor has enough context to
2185 avoid warning in these cases.
2188 A function declared external in one block and then used after the end of
2192 A @code{switch} statement has an operand of type @code{long}.
2195 A non-@code{static} function declaration follows a @code{static} one.
2196 This construct is not accepted by some traditional C compilers.
2199 The ISO type of an integer constant has a different width or
2200 signedness from its traditional type. This warning is only issued if
2201 the base of the constant is ten. I.e. hexadecimal or octal values, which
2202 typically represent bit patterns, are not warned about.
2205 Usage of ISO string concatenation is detected.
2208 Initialization of automatic aggregates.
2211 Identifier conflicts with labels. Traditional C lacks a separate
2212 namespace for labels.
2215 Initialization of unions. If the initializer is zero, the warning is
2216 omitted. This is done under the assumption that the zero initializer in
2217 user code appears conditioned on e.g. @code{__STDC__} to avoid missing
2218 initializer warnings and relies on default initialization to zero in the
2222 Conversions by prototypes between fixed/floating point values and vice
2223 versa. The absence of these prototypes when compiling with traditional
2224 C would cause serious problems. This is a subset of the possible
2225 conversion warnings, for the full set use @samp{-Wconversion}.
2229 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2232 Warn whenever a local variable shadows another local variable, parameter or
2233 global variable or whenever a built-in function is shadowed.
2235 @item -Wid-clash-@var{len}
2236 Warn whenever two distinct identifiers match in the first @var{len}
2237 characters. This may help you prepare a program that will compile
2238 with certain obsolete, brain-damaged compilers.
2240 @item -Wlarger-than-@var{len}
2241 Warn whenever an object of larger than @var{len} bytes is defined.
2243 @item -Wpointer-arith
2244 Warn about anything that depends on the ``size of'' a function type or
2245 of @code{void}. GNU C assigns these types a size of 1, for
2246 convenience in calculations with @code{void *} pointers and pointers
2249 @item -Wbad-function-cast (C only)
2250 Warn whenever a function call is cast to a non-matching type.
2251 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2254 Warn whenever a pointer is cast so as to remove a type qualifier from
2255 the target type. For example, warn if a @code{const char *} is cast
2256 to an ordinary @code{char *}.
2259 Warn whenever a pointer is cast such that the required alignment of the
2260 target is increased. For example, warn if a @code{char *} is cast to
2261 an @code{int *} on machines where integers can only be accessed at
2262 two- or four-byte boundaries.
2264 @item -Wwrite-strings
2265 Give string constants the type @code{const char[@var{length}]} so that
2266 copying the address of one into a non-@code{const} @code{char *}
2267 pointer will get a warning. These warnings will help you find at
2268 compile time code that can try to write into a string constant, but
2269 only if you have been very careful about using @code{const} in
2270 declarations and prototypes. Otherwise, it will just be a nuisance;
2271 this is why we did not make @samp{-Wall} request these warnings.
2274 Warn if a prototype causes a type conversion that is different from what
2275 would happen to the same argument in the absence of a prototype. This
2276 includes conversions of fixed point to floating and vice versa, and
2277 conversions changing the width or signedness of a fixed point argument
2278 except when the same as the default promotion.
2280 Also, warn if a negative integer constant expression is implicitly
2281 converted to an unsigned type. For example, warn about the assignment
2282 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2283 casts like @code{(unsigned) -1}.
2285 @item -Wsign-compare
2286 @cindex warning for comparison of signed and unsigned values
2287 @cindex comparison of signed and unsigned values, warning
2288 @cindex signed and unsigned values, comparison warning
2289 Warn when a comparison between signed and unsigned values could produce
2290 an incorrect result when the signed value is converted to unsigned.
2291 This warning is also enabled by @samp{-W}; to get the other warnings
2292 of @samp{-W} without this warning, use @samp{-W -Wno-sign-compare}.
2294 @item -Waggregate-return
2295 Warn if any functions that return structures or unions are defined or
2296 called. (In languages where you can return an array, this also elicits
2299 @item -Wstrict-prototypes (C only)
2300 Warn if a function is declared or defined without specifying the
2301 argument types. (An old-style function definition is permitted without
2302 a warning if preceded by a declaration which specifies the argument
2305 @item -Wmissing-prototypes (C only)
2306 Warn if a global function is defined without a previous prototype
2307 declaration. This warning is issued even if the definition itself
2308 provides a prototype. The aim is to detect global functions that fail
2309 to be declared in header files.
2311 @item -Wmissing-declarations
2312 Warn if a global function is defined without a previous declaration.
2313 Do so even if the definition itself provides a prototype.
2314 Use this option to detect global functions that are not declared in
2317 @item -Wmissing-noreturn
2318 Warn about functions which might be candidates for attribute @code{noreturn}.
2319 Note these are only possible candidates, not absolute ones. Care should
2320 be taken to manually verify functions actually do not ever return before
2321 adding the @code{noreturn} attribute, otherwise subtle code generation
2322 bugs could be introduced. You will not get a warning for @code{main} in
2323 hosted C environments.
2325 @item -Wmissing-format-attribute
2326 If @samp{-Wformat} is enabled, also warn about functions which might be
2327 candidates for @code{format} attributes. Note these are only possible
2328 candidates, not absolute ones. GCC will guess that @code{format}
2329 attributes might be appropriate for any function that calls a function
2330 like @code{vprintf} or @code{vscanf}, but this might not always be the
2331 case, and some functions for which @code{format} attributes are
2332 appropriate may not be detected. This option has no effect unless
2333 @samp{-Wformat} is enabled (possibly by @samp{-Wall}).
2336 Warn if a structure is given the packed attribute, but the packed
2337 attribute has no effect on the layout or size of the structure.
2338 Such structures may be mis-aligned for little benefit. For
2339 instance, in this code, the variable @code{f.x} in @code{struct bar}
2340 will be misaligned even though @code{struct bar} does not itself
2341 have the packed attribute:
2348 @} __attribute__((packed));
2357 Warn if padding is included in a structure, either to align an element
2358 of the structure or to align the whole structure. Sometimes when this
2359 happens it is possible to rearrange the fields of the structure to
2360 reduce the padding and so make the structure smaller.
2362 @item -Wredundant-decls
2363 Warn if anything is declared more than once in the same scope, even in
2364 cases where multiple declaration is valid and changes nothing.
2366 @item -Wnested-externs (C only)
2367 Warn if an @code{extern} declaration is encountered within a function.
2369 @item -Wunreachable-code
2370 Warn if the compiler detects that code will never be executed.
2372 This option is intended to warn when the compiler detects that at
2373 least a whole line of source code will never be executed, because
2374 some condition is never satisfied or because it is after a
2375 procedure that never returns.
2377 It is possible for this option to produce a warning even though there
2378 are circumstances under which part of the affected line can be executed,
2379 so care should be taken when removing apparently-unreachable code.
2381 For instance, when a function is inlined, a warning may mean that the
2382 line is unreachable in only one inlined copy of the function.
2384 This option is not made part of @samp{-Wall} because in a debugging
2385 version of a program there is often substantial code which checks
2386 correct functioning of the program and is, hopefully, unreachable
2387 because the program does work. Another common use of unreachable
2388 code is to provide behaviour which is selectable at compile-time.
2391 Warn if a function can not be inlined and it was declared as inline.
2394 Warn if @samp{long long} type is used. This is default. To inhibit
2395 the warning messages, use @samp{-Wno-long-long}. Flags
2396 @samp{-Wlong-long} and @samp{-Wno-long-long} are taken into account
2397 only when @samp{-pedantic} flag is used.
2399 @item -Wdisabled-optimization
2400 Warn if a requested optimization pass is disabled. This warning does
2401 not generally indicate that there is anything wrong with your code; it
2402 merely indicates that GCC's optimizers were unable to handle the code
2403 effectively. Often, the problem is that your code is too big or too
2404 complex; GCC will refuse to optimize programs when the optimization
2405 itself is likely to take inordinate amounts of time.
2408 Make all warnings into errors.
2411 @node Debugging Options
2412 @section Options for Debugging Your Program or GCC
2413 @cindex options, debugging
2414 @cindex debugging information options
2416 GCC has various special options that are used for debugging
2417 either your program or GCC:
2421 Produce debugging information in the operating system's native format
2422 (stabs, COFF, XCOFF, or DWARF). GDB can work with this debugging
2425 On most systems that use stabs format, @samp{-g} enables use of extra
2426 debugging information that only GDB can use; this extra information
2427 makes debugging work better in GDB but will probably make other debuggers
2429 refuse to read the program. If you want to control for certain whether
2430 to generate the extra information, use @samp{-gstabs+}, @samp{-gstabs},
2431 @samp{-gxcoff+}, @samp{-gxcoff}, @samp{-gdwarf-1+}, or @samp{-gdwarf-1}
2434 Unlike most other C compilers, GCC allows you to use @samp{-g} with
2435 @samp{-O}. The shortcuts taken by optimized code may occasionally
2436 produce surprising results: some variables you declared may not exist
2437 at all; flow of control may briefly move where you did not expect it;
2438 some statements may not be executed because they compute constant
2439 results or their values were already at hand; some statements may
2440 execute in different places because they were moved out of loops.
2442 Nevertheless it proves possible to debug optimized output. This makes
2443 it reasonable to use the optimizer for programs that might have bugs.
2445 The following options are useful when GCC is generated with the
2446 capability for more than one debugging format.
2449 Produce debugging information for use by GDB. This means to use the
2450 most expressive format available (DWARF 2, stabs, or the native format
2451 if neither of those are supported), including GDB extensions if at all
2455 Produce debugging information in stabs format (if that is supported),
2456 without GDB extensions. This is the format used by DBX on most BSD
2457 systems. On MIPS, Alpha and System V Release 4 systems this option
2458 produces stabs debugging output which is not understood by DBX or SDB.
2459 On System V Release 4 systems this option requires the GNU assembler.
2462 Produce debugging information in stabs format (if that is supported),
2463 using GNU extensions understood only by the GNU debugger (GDB). The
2464 use of these extensions is likely to make other debuggers crash or
2465 refuse to read the program.
2468 Produce debugging information in COFF format (if that is supported).
2469 This is the format used by SDB on most System V systems prior to
2473 Produce debugging information in XCOFF format (if that is supported).
2474 This is the format used by the DBX debugger on IBM RS/6000 systems.
2477 Produce debugging information in XCOFF format (if that is supported),
2478 using GNU extensions understood only by the GNU debugger (GDB). The
2479 use of these extensions is likely to make other debuggers crash or
2480 refuse to read the program, and may cause assemblers other than the GNU
2481 assembler (GAS) to fail with an error.
2484 Produce debugging information in DWARF version 1 format (if that is
2485 supported). This is the format used by SDB on most System V Release 4
2489 Produce debugging information in DWARF version 1 format (if that is
2490 supported), using GNU extensions understood only by the GNU debugger
2491 (GDB). The use of these extensions is likely to make other debuggers
2492 crash or refuse to read the program.
2495 Produce debugging information in DWARF version 2 format (if that is
2496 supported). This is the format used by DBX on IRIX 6.
2499 @itemx -ggdb@var{level}
2500 @itemx -gstabs@var{level}
2501 @itemx -gcoff@var{level}
2502 @itemx -gxcoff@var{level}
2503 @itemx -gdwarf@var{level}
2504 @itemx -gdwarf-2@var{level}
2505 Request debugging information and also use @var{level} to specify how
2506 much information. The default level is 2.
2508 Level 1 produces minimal information, enough for making backtraces in
2509 parts of the program that you don't plan to debug. This includes
2510 descriptions of functions and external variables, but no information
2511 about local variables and no line numbers.
2513 Level 3 includes extra information, such as all the macro definitions
2514 present in the program. Some debuggers support macro expansion when
2519 Generate extra code to write profile information suitable for the
2520 analysis program @code{prof}. You must use this option when compiling
2521 the source files you want data about, and you must also use it when
2524 @cindex @code{gprof}
2526 Generate extra code to write profile information suitable for the
2527 analysis program @code{gprof}. You must use this option when compiling
2528 the source files you want data about, and you must also use it when
2533 Generate extra code to write profile information for basic blocks, which will
2534 record the number of times each basic block is executed, the basic block start
2535 address, and the function name containing the basic block. If @samp{-g} is
2536 used, the line number and filename of the start of the basic block will also be
2537 recorded. If not overridden by the machine description, the default action is
2538 to append to the text file @file{bb.out}.
2540 This data could be analyzed by a program like @code{tcov}. Note,
2541 however, that the format of the data is not what @code{tcov} expects.
2542 Eventually GNU @code{gprof} should be extended to process this data.
2545 Makes the compiler print out each function name as it is compiled, and
2546 print some statistics about each pass when it finishes.
2549 Makes the compiler print some statistics about the time consumed by each
2550 pass when it finishes.
2553 Makes the compiler print some statistics about permanent memory
2554 allocation when it finishes.
2557 Generate extra code to profile basic blocks. Your executable will
2558 produce output that is a superset of that produced when @samp{-a} is
2559 used. Additional output is the source and target address of the basic
2560 blocks where a jump takes place, the number of times a jump is executed,
2561 and (optionally) the complete sequence of basic blocks being executed.
2562 The output is appended to file @file{bb.out}.
2564 You can examine different profiling aspects without recompilation. Your
2565 executable will read a list of function names from file @file{bb.in}.
2566 Profiling starts when a function on the list is entered and stops when
2567 that invocation is exited. To exclude a function from profiling, prefix
2568 its name with `-'. If a function name is not unique, you can
2569 disambiguate it by writing it in the form
2570 @samp{/path/filename.d:functionname}. Your executable will write the
2571 available paths and filenames in file @file{bb.out}.
2573 Several function names have a special meaning:
2576 Write source, target and frequency of jumps to file @file{bb.out}.
2577 @item __bb_hidecall__
2578 Exclude function calls from frequency count.
2579 @item __bb_showret__
2580 Include function returns in frequency count.
2582 Write the sequence of basic blocks executed to file @file{bbtrace.gz}.
2583 The file will be compressed using the program @samp{gzip}, which must
2584 exist in your @env{PATH}. On systems without the @samp{popen}
2585 function, the file will be named @file{bbtrace} and will not be
2586 compressed. @strong{Profiling for even a few seconds on these systems
2587 will produce a very large file.} Note: @code{__bb_hidecall__} and
2588 @code{__bb_showret__} will not affect the sequence written to
2592 Here's a short example using different profiling parameters
2593 in file @file{bb.in}. Assume function @code{foo} consists of basic blocks
2594 1 and 2 and is called twice from block 3 of function @code{main}. After
2595 the calls, block 3 transfers control to block 4 of @code{main}.
2597 With @code{__bb_trace__} and @code{main} contained in file @file{bb.in},
2598 the following sequence of blocks is written to file @file{bbtrace.gz}:
2599 0 3 1 2 1 2 4. The return from block 2 to block 3 is not shown, because
2600 the return is to a point inside the block and not to the top. The
2601 block address 0 always indicates, that control is transferred
2602 to the trace from somewhere outside the observed functions. With
2603 @samp{-foo} added to @file{bb.in}, the blocks of function
2604 @code{foo} are removed from the trace, so only 0 3 4 remains.
2606 With @code{__bb_jumps__} and @code{main} contained in file @file{bb.in},
2607 jump frequencies will be written to file @file{bb.out}. The
2608 frequencies are obtained by constructing a trace of blocks
2609 and incrementing a counter for every neighbouring pair of blocks
2610 in the trace. The trace 0 3 1 2 1 2 4 displays the following
2614 Jump from block 0x0 to block 0x3 executed 1 time(s)
2615 Jump from block 0x3 to block 0x1 executed 1 time(s)
2616 Jump from block 0x1 to block 0x2 executed 2 time(s)
2617 Jump from block 0x2 to block 0x1 executed 1 time(s)
2618 Jump from block 0x2 to block 0x4 executed 1 time(s)
2621 With @code{__bb_hidecall__}, control transfer due to call instructions
2622 is removed from the trace, that is the trace is cut into three parts: 0
2623 3 4, 0 1 2 and 0 1 2. With @code{__bb_showret__}, control transfer due
2624 to return instructions is added to the trace. The trace becomes: 0 3 1
2625 2 3 1 2 3 4. Note, that this trace is not the same, as the sequence
2626 written to @file{bbtrace.gz}. It is solely used for counting jump
2629 @item -fprofile-arcs
2630 Instrument @dfn{arcs} during compilation. For each function of your
2631 program, GCC creates a program flow graph, then finds a spanning tree
2632 for the graph. Only arcs that are not on the spanning tree have to be
2633 instrumented: the compiler adds code to count the number of times that these
2634 arcs are executed. When an arc is the only exit or only entrance to a
2635 block, the instrumentation code can be added to the block; otherwise, a
2636 new basic block must be created to hold the instrumentation code.
2638 Since not every arc in the program must be instrumented, programs
2639 compiled with this option run faster than programs compiled with
2640 @samp{-a}, which adds instrumentation code to every basic block in the
2641 program. The tradeoff: since @code{gcov} does not have
2642 execution counts for all branches, it must start with the execution
2643 counts for the instrumented branches, and then iterate over the program
2644 flow graph until the entire graph has been solved. Hence, @code{gcov}
2645 runs a little more slowly than a program which uses information from
2648 @samp{-fprofile-arcs} also makes it possible to estimate branch
2649 probabilities, and to calculate basic block execution counts. In
2650 general, basic block execution counts do not give enough information to
2651 estimate all branch probabilities. When the compiled program exits, it
2652 saves the arc execution counts to a file called
2653 @file{@var{sourcename}.da}. Use the compiler option
2654 @samp{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
2655 Control Optimization}) when recompiling, to optimize using estimated
2656 branch probabilities.
2659 @item -ftest-coverage
2660 Create data files for the @code{gcov} code-coverage utility
2661 (@pxref{Gcov,, @code{gcov}: a GCC Test Coverage Program}).
2662 The data file names begin with the name of your source file:
2665 @item @var{sourcename}.bb
2666 A mapping from basic blocks to line numbers, which @code{gcov} uses to
2667 associate basic block execution counts with line numbers.
2669 @item @var{sourcename}.bbg
2670 A list of all arcs in the program flow graph. This allows @code{gcov}
2671 to reconstruct the program flow graph, so that it can compute all basic
2672 block and arc execution counts from the information in the
2673 @code{@var{sourcename}.da} file (this last file is the output from
2674 @samp{-fprofile-arcs}).
2677 @item -d@var{letters}
2678 Says to make debugging dumps during compilation at times specified by
2679 @var{letters}. This is used for debugging the compiler. The file names
2680 for most of the dumps are made by appending a pass number and a word to
2681 the source file name (e.g. @file{foo.c.00.rtl} or @file{foo.c.01.sibling}).
2682 Here are the possible letters for use in @var{letters}, and their meanings:
2686 Annotate the assembler output with miscellaneous debugging information.
2688 Dump after computing branch probabilities, to @file{@var{file}.11.bp}.
2690 Dump after block reordering, to @file{@var{file}.26.bbro}.
2692 Dump after instruction combination, to the file @file{@var{file}.14.combine}.
2694 Dump after the first if conversion, to the file @file{@var{file}.15.ce}.
2696 Dump after delayed branch scheduling, to @file{@var{file}.29.dbr}.
2698 Dump all macro definitions, at the end of preprocessing, in addition to
2701 Dump after SSA optimizations, to @file{@var{file}.05.ssa} and
2702 @file{@var{file}.06.ussa}.
2704 Dump after the second if conversion, to @file{@var{file}.24.ce2}.
2706 Dump after life analysis, to @file{@var{file}.13.life}.
2708 Dump after purging @code{ADDRESSOF} codes, to @file{@var{file}.04.addressof}.
2710 Dump after global register allocation, to @file{@var{file}.19.greg}.
2712 Dump after post-reload CSE and other optimizations, to @file{@var{file}.20.postreload}.
2714 Dump after GCSE, to @file{@var{file}.08.gcse}.
2716 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
2718 Dump after the first jump optimization, to @file{@var{file}.02.jump}.
2720 Dump after the last jump optimization, to @file{@var{file}.27.jump2}.
2722 Dump after conversion from registers to stack, to @file{@var{file}.29.stack}.
2724 Dump after local register allocation, to @file{@var{file}.18.lreg}.
2726 Dump after loop optimization, to @file{@var{file}.09.loop}.
2728 Dump after performing the machine dependent reorganisation pass, to
2729 @file{@var{file}.28.mach}.
2731 Dump after register renumbering, to @file{@var{file}.23.rnreg}.
2733 Dump after the register move pass, to @file{@var{file}.16.regmove}.
2735 Dump after RTL generation, to @file{@var{file}.00.rtl}.
2737 Dump after the second instruction scheduling pass, to
2738 @file{@var{file}.25.sched2}.
2740 Dump after CSE (including the jump optimization that sometimes follows
2741 CSE), to @file{@var{file}.03.cse}.
2743 Dump after the first instruction scheduling pass, to
2744 @file{@var{file}.17.sched}.
2746 Dump after the second CSE pass (including the jump optimization that
2747 sometimes follows CSE), to @file{@var{file}.10.cse2}.
2749 Dump after the second flow pass, to @file{@var{file}.21.flow2}.
2751 Dump after dead code elimination, to @file{@var{file}.06.dce}.
2753 Dump after the peephole pass, to @file{@var{file}.22.peephole2}.
2755 Produce all the dumps listed above.
2757 Print statistics on memory usage, at the end of the run, to
2760 Annotate the assembler output with a comment indicating which
2761 pattern and alternative was used. The length of each instruction is
2764 Dump the RTL in the assembler output as a comment before each instruction.
2765 Also turns on @samp{-dp} annotation.
2767 For each of the other indicated dump files (except for
2768 @file{@var{file}.00.rtl}), dump a representation of the control flow graph
2769 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
2771 Just generate RTL for a function instead of compiling it. Usually used
2774 Dump debugging information during parsing, to standard error.
2777 @item -fdump-unnumbered
2778 When doing debugging dumps (see -d option above), suppress instruction
2779 numbers and line number note output. This makes it more feasible to
2780 use diff on debugging dumps for compiler invocations with different
2781 options, in particular with and without -g.
2783 @item -fdump-translation-unit=@var{file} (C and C++ only)
2784 Dump a representation of the tree structure for the entire translation
2787 @item -fdump-class_layout=@var{file} (C++ only)
2788 @item -fdump-class_layout (C++ only)
2789 Dump a representation of each class's heirarchy to @var{file}, or
2790 @code{stderr} if not specified.
2792 @item -fpretend-float
2793 When running a cross-compiler, pretend that the target machine uses the
2794 same floating point format as the host machine. This causes incorrect
2795 output of the actual floating constants, but the actual instruction
2796 sequence will probably be the same as GCC would make when running on
2800 Store the usual ``temporary'' intermediate files permanently; place them
2801 in the current directory and name them based on the source file. Thus,
2802 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
2803 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
2804 preprocessed @file{foo.i} output file even though the compiler now
2805 normally uses an integrated preprocessor.
2808 Report the CPU time taken by each subprocess in the compilation
2809 sequence. For C source files, this is the compiler proper and assembler
2810 (plus the linker if linking is done). The output looks like this:
2817 The first number on each line is the ``user time,'' that is time spent
2818 executing the program itself. The second number is ``system time,''
2819 time spent executing operating system routines on behalf of the program.
2820 Both numbers are in seconds.
2822 @item -print-file-name=@var{library}
2823 Print the full absolute name of the library file @var{library} that
2824 would be used when linking---and don't do anything else. With this
2825 option, GCC does not compile or link anything; it just prints the
2828 @item -print-prog-name=@var{program}
2829 Like @samp{-print-file-name}, but searches for a program such as @samp{cpp}.
2831 @item -print-libgcc-file-name
2832 Same as @samp{-print-file-name=libgcc.a}.
2834 This is useful when you use @samp{-nostdlib} or @samp{-nodefaultlibs}
2835 but you do want to link with @file{libgcc.a}. You can do
2838 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
2841 @item -print-search-dirs
2842 Print the name of the configured installation directory and a list of
2843 program and library directories gcc will search---and don't do anything else.
2845 This is useful when gcc prints the error message
2846 @samp{installation problem, cannot exec cpp0: No such file or directory}.
2847 To resolve this you either need to put @file{cpp0} and the other compiler
2848 components where gcc expects to find them, or you can set the environment
2849 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
2850 Don't forget the trailing '/'.
2851 @xref{Environment Variables}.
2854 Print the compiler's target machine (for example,
2855 @samp{i686-pc-linux-gnu})---and don't do anything else.
2858 Print the compiler version (for example, @samp{3.0})---and don't do
2862 Print the compiler's built-in specs---and don't do anything else. (This
2863 is used when GCC itself is being built.) @xref{Spec Files}.
2866 @node Optimize Options
2867 @section Options That Control Optimization
2868 @cindex optimize options
2869 @cindex options, optimization
2871 These options control various sorts of optimizations:
2876 Optimize. Optimizing compilation takes somewhat more time, and a lot
2877 more memory for a large function.
2879 Without @samp{-O}, the compiler's goal is to reduce the cost of
2880 compilation and to make debugging produce the expected results.
2881 Statements are independent: if you stop the program with a breakpoint
2882 between statements, you can then assign a new value to any variable or
2883 change the program counter to any other statement in the function and
2884 get exactly the results you would expect from the source code.
2886 Without @samp{-O}, the compiler only allocates variables declared
2887 @code{register} in registers. The resulting compiled code is a little
2888 worse than produced by PCC without @samp{-O}.
2890 With @samp{-O}, the compiler tries to reduce code size and execution
2893 When you specify @samp{-O}, the compiler turns on @samp{-fthread-jumps}
2894 and @samp{-fdefer-pop} on all machines. The compiler turns on
2895 @samp{-fdelayed-branch} on machines that have delay slots, and
2896 @samp{-fomit-frame-pointer} on machines that can support debugging even
2897 without a frame pointer. On some machines the compiler also turns
2898 on other flags.@refill
2901 Optimize even more. GCC performs nearly all supported optimizations
2902 that do not involve a space-speed tradeoff. The compiler does not
2903 perform loop unrolling or function inlining when you specify @samp{-O2}.
2904 As compared to @samp{-O}, this option increases both compilation time
2905 and the performance of the generated code.
2907 @samp{-O2} turns on all optional optimizations except for loop unrolling,
2908 function inlining, and register renaming. It also turns on the
2909 @samp{-fforce-mem} option on all machines and frame pointer elimination
2910 on machines where doing so does not interfere with debugging.
2913 Optimize yet more. @samp{-O3} turns on all optimizations specified by
2914 @samp{-O2} and also turns on the @samp{-finline-functions} and
2915 @samp{-frename-registers} options.
2921 Optimize for size. @samp{-Os} enables all @samp{-O2} optimizations that
2922 do not typically increase code size. It also performs further
2923 optimizations designed to reduce code size.
2925 If you use multiple @samp{-O} options, with or without level numbers,
2926 the last such option is the one that is effective.
2929 Options of the form @samp{-f@var{flag}} specify machine-independent
2930 flags. Most flags have both positive and negative forms; the negative
2931 form of @samp{-ffoo} would be @samp{-fno-foo}. In the table below,
2932 only one of the forms is listed---the one which is not the default.
2933 You can figure out the other form by either removing @samp{no-} or
2938 Do not store floating point variables in registers, and inhibit other
2939 options that might change whether a floating point value is taken from a
2942 @cindex floating point precision
2943 This option prevents undesirable excess precision on machines such as
2944 the 68000 where the floating registers (of the 68881) keep more
2945 precision than a @code{double} is supposed to have. Similarly for the
2946 x86 architecture. For most programs, the excess precision does only
2947 good, but a few programs rely on the precise definition of IEEE floating
2948 point. Use @samp{-ffloat-store} for such programs, after modifying
2949 them to store all pertinent intermediate computations into variables.
2951 @item -fno-default-inline
2952 Do not make member functions inline by default merely because they are
2953 defined inside the class scope (C++ only). Otherwise, when you specify
2954 @w{@samp{-O}}, member functions defined inside class scope are compiled
2955 inline by default; i.e., you don't need to add @samp{inline} in front of
2956 the member function name.
2958 @item -fno-defer-pop
2959 Always pop the arguments to each function call as soon as that function
2960 returns. For machines which must pop arguments after a function call,
2961 the compiler normally lets arguments accumulate on the stack for several
2962 function calls and pops them all at once.
2965 Force memory operands to be copied into registers before doing
2966 arithmetic on them. This produces better code by making all memory
2967 references potential common subexpressions. When they are not common
2968 subexpressions, instruction combination should eliminate the separate
2969 register-load. The @samp{-O2} option turns on this option.
2972 Force memory address constants to be copied into registers before
2973 doing arithmetic on them. This may produce better code just as
2974 @samp{-fforce-mem} may.
2976 @item -fomit-frame-pointer
2977 Don't keep the frame pointer in a register for functions that
2978 don't need one. This avoids the instructions to save, set up and
2979 restore frame pointers; it also makes an extra register available
2980 in many functions. @strong{It also makes debugging impossible on
2984 On some machines, such as the Vax, this flag has no effect, because
2985 the standard calling sequence automatically handles the frame pointer
2986 and nothing is saved by pretending it doesn't exist. The
2987 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2988 whether a target machine supports this flag. @xref{Registers}.@refill
2991 On some machines, such as the Vax, this flag has no effect, because
2992 the standard calling sequence automatically handles the frame pointer
2993 and nothing is saved by pretending it doesn't exist. The
2994 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2995 whether a target machine supports this flag. @xref{Registers,,Register
2996 Usage, gcc.info, Using and Porting GCC}.@refill
2999 @item -foptimize-sibling-calls
3000 Optimize sibling and tail recursive calls.
3003 This option generates traps for signed overflow on addition, subtraction,
3004 multiplication operations.
3007 Don't pay attention to the @code{inline} keyword. Normally this option
3008 is used to keep the compiler from expanding any functions inline.
3009 Note that if you are not optimizing, no functions can be expanded inline.
3011 @item -finline-functions
3012 Integrate all simple functions into their callers. The compiler
3013 heuristically decides which functions are simple enough to be worth
3014 integrating in this way.
3016 If all calls to a given function are integrated, and the function is
3017 declared @code{static}, then the function is normally not output as
3018 assembler code in its own right.
3020 @item -finline-limit=@var{n}
3021 By default, gcc limits the size of functions that can be inlined. This flag
3022 allows the control of this limit for functions that are explicitly marked as
3023 inline (ie marked with the inline keyword or defined within the class
3024 definition in c++). @var{n} is the size of functions that can be inlined in
3025 number of pseudo instructions (not counting parameter handling). The default
3026 value of n is 10000. Increasing this value can result in more inlined code at
3027 the cost of compilation time and memory consumption. Decreasing usually makes
3028 the compilation faster and less code will be inlined (which presumably
3029 means slower programs). This option is particularly useful for programs that
3030 use inlining heavily such as those based on recursive templates with c++.
3032 @emph{Note:} pseudo instruction represents, in this particular context, an
3033 abstract measurement of function's size. In no way, it represents a count
3034 of assembly instructions and as such its exact meaning might change from one
3035 release to an another.
3037 @item -fkeep-inline-functions
3038 Even if all calls to a given function are integrated, and the function
3039 is declared @code{static}, nevertheless output a separate run-time
3040 callable version of the function. This switch does not affect
3041 @code{extern inline} functions.
3043 @item -fkeep-static-consts
3044 Emit variables declared @code{static const} when optimization isn't turned
3045 on, even if the variables aren't referenced.
3047 GCC enables this option by default. If you want to force the compiler to
3048 check if the variable was referenced, regardless of whether or not
3049 optimization is turned on, use the @samp{-fno-keep-static-consts} option.
3051 @item -fno-function-cse
3052 Do not put function addresses in registers; make each instruction that
3053 calls a constant function contain the function's address explicitly.
3055 This option results in less efficient code, but some strange hacks
3056 that alter the assembler output may be confused by the optimizations
3057 performed when this option is not used.
3060 Sets @samp{-fno-math-errno}, @samp{-funsafe-math-optimizations},
3061 and @samp{-fno-trapping-math}.
3063 This option causes the preprocessor macro __FAST_MATH__ to be defined.
3065 This option should never be turned on by any @samp{-O} option since
3066 it can result in incorrect output for programs which depend on
3067 an exact implementation of IEEE or ISO rules/specifications for
3070 @item -fno-math-errno
3071 Do not set ERRNO after calling math functions that are executed
3072 with a single instruction, e.g., sqrt. A program that relies on
3073 IEEE exceptions for math error handling may want to use this flag
3074 for speed while maintaining IEEE arithmetic compatibility.
3076 This option should never be turned on by any @samp{-O} option since
3077 it can result in incorrect output for programs which depend on
3078 an exact implementation of IEEE or ISO rules/specifications for
3081 The default is @samp{-fmath-errno}. The @samp{-ffast-math} option
3082 sets @samp{-fno-math-errno}.
3084 @item -funsafe-math-optimizations
3085 Allow optimizations for floating-point arithmetic that (a) assume
3086 that arguments and results are valid and (b) may violate IEEE or
3089 This option should never be turned on by any @samp{-O} option since
3090 it can result in incorrect output for programs which depend on
3091 an exact implementation of IEEE or ISO rules/specifications for
3094 The default is @samp{-fno-unsafe-math-optimizations}. The
3095 @samp{-ffast-math} option sets @samp{-funsafe-math-optimizations}.
3097 @item -fno-trapping-math
3098 Compile code assuming that floating-point operations cannot generate
3099 user-visible traps. Setting this option may allow faster code
3100 if one relies on ``non-stop'' IEEE arithmetic, for example.
3102 This option should never be turned on by any @samp{-O} option since
3103 it can result in incorrect output for programs which depend on
3104 an exact implementation of IEEE or ISO rules/specifications for
3107 The default is @samp{-ftrapping-math}. The @samp{-ffast-math}
3108 option sets @samp{-fno-trapping-math}.
3111 @c following causes underfulls.. they don't look great, but we deal.
3113 The following options control specific optimizations. The @samp{-O2}
3114 option turns on all of these optimizations except @samp{-funroll-loops}
3115 and @samp{-funroll-all-loops}. On most machines, the @samp{-O} option
3116 turns on the @samp{-fthread-jumps} and @samp{-fdelayed-branch} options,
3117 but specific machines may handle it differently.
3119 You can use the following flags in the rare cases when ``fine-tuning''
3120 of optimizations to be performed is desired.
3123 @item -fstrength-reduce
3124 Perform the optimizations of loop strength reduction and
3125 elimination of iteration variables.
3127 @item -fthread-jumps
3128 Perform optimizations where we check to see if a jump branches to a
3129 location where another comparison subsumed by the first is found. If
3130 so, the first branch is redirected to either the destination of the
3131 second branch or a point immediately following it, depending on whether
3132 the condition is known to be true or false.
3134 @item -fcse-follow-jumps
3135 In common subexpression elimination, scan through jump instructions
3136 when the target of the jump is not reached by any other path. For
3137 example, when CSE encounters an @code{if} statement with an
3138 @code{else} clause, CSE will follow the jump when the condition
3141 @item -fcse-skip-blocks
3142 This is similar to @samp{-fcse-follow-jumps}, but causes CSE to
3143 follow jumps which conditionally skip over blocks. When CSE
3144 encounters a simple @code{if} statement with no else clause,
3145 @samp{-fcse-skip-blocks} causes CSE to follow the jump around the
3146 body of the @code{if}.
3148 @item -frerun-cse-after-loop
3149 Re-run common subexpression elimination after loop optimizations has been
3152 @item -frerun-loop-opt
3153 Run the loop optimizer twice.
3156 Perform a global common subexpression elimination pass.
3157 This pass also performs global constant and copy propagation.
3160 When -fgcse-lm is enabled, global common subexpression elimination will
3161 attempt to move loads which are only killed by stores into themselves. This
3162 allows a loop containing a load/store sequence to be changed to a load outside
3163 the loop, and a copy/store within the loop.
3166 When -fgcse-sm is enabled, A store motion pass is run after global common
3167 subexpression elimination. This pass will attempt to move stores out of loops.
3168 When used in conjunction with -fgcse-lm, loops containing a load/store sequence
3169 can be changed to a load before the loop and a store after the loop.
3171 @item -fdelete-null-pointer-checks
3172 Use global dataflow analysis to identify and eliminate useless null
3173 pointer checks. Programs which rely on NULL pointer dereferences @emph{not}
3174 halting the program may not work properly with this option. Use
3175 -fno-delete-null-pointer-checks to disable this optimizing for programs
3176 which depend on that behavior.
3178 @item -fexpensive-optimizations
3179 Perform a number of minor optimizations that are relatively expensive.
3181 @item -foptimize-register-move
3183 Attempt to reassign register numbers in move instructions and as
3184 operands of other simple instructions in order to maximize the amount of
3185 register tying. This is especially helpful on machines with two-operand
3186 instructions. GCC enables this optimization by default with @samp{-O2}
3189 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
3192 @item -fdelayed-branch
3193 If supported for the target machine, attempt to reorder instructions
3194 to exploit instruction slots available after delayed branch
3197 @item -fschedule-insns
3198 If supported for the target machine, attempt to reorder instructions to
3199 eliminate execution stalls due to required data being unavailable. This
3200 helps machines that have slow floating point or memory load instructions
3201 by allowing other instructions to be issued until the result of the load
3202 or floating point instruction is required.
3204 @item -fschedule-insns2
3205 Similar to @samp{-fschedule-insns}, but requests an additional pass of
3206 instruction scheduling after register allocation has been done. This is
3207 especially useful on machines with a relatively small number of
3208 registers and where memory load instructions take more than one cycle.
3210 @item -ffunction-sections
3211 @itemx -fdata-sections
3212 Place each function or data item into its own section in the output
3213 file if the target supports arbitrary sections. The name of the
3214 function or the name of the data item determines the section's name
3217 Use these options on systems where the linker can perform optimizations
3218 to improve locality of reference in the instruction space. HPPA
3219 processors running HP-UX and Sparc processors running Solaris 2 have
3220 linkers with such optimizations. Other systems using the ELF object format
3221 as well as AIX may have these optimizations in the future.
3223 Only use these options when there are significant benefits from doing
3224 so. When you specify these options, the assembler and linker will
3225 create larger object and executable files and will also be slower.
3226 You will not be able to use @code{gprof} on all systems if you
3227 specify this option and you may have problems with debugging if
3228 you specify both this option and @samp{-g}.
3230 @item -fcaller-saves
3231 Enable values to be allocated in registers that will be clobbered by
3232 function calls, by emitting extra instructions to save and restore the
3233 registers around such calls. Such allocation is done only when it
3234 seems to result in better code than would otherwise be produced.
3236 This option is always enabled by default on certain machines, usually
3237 those which have no call-preserved registers to use instead.
3239 For all machines, optimization level 2 and higher enables this flag by
3242 @item -funroll-loops
3243 Perform the optimization of loop unrolling. This is only done for loops
3244 whose number of iterations can be determined at compile time or run time.
3245 @samp{-funroll-loops} implies both @samp{-fstrength-reduce} and
3246 @samp{-frerun-cse-after-loop}.
3248 @item -funroll-all-loops
3249 Perform the optimization of loop unrolling. This is done for all loops
3250 and usually makes programs run more slowly. @samp{-funroll-all-loops}
3251 implies @samp{-fstrength-reduce} as well as @samp{-frerun-cse-after-loop}.
3253 @item -fmove-all-movables
3254 Forces all invariant computations in loops to be moved
3257 @item -freduce-all-givs
3258 Forces all general-induction variables in loops to be
3261 @emph{Note:} When compiling programs written in Fortran,
3262 @samp{-fmove-all-movables} and @samp{-freduce-all-givs} are enabled
3263 by default when you use the optimizer.
3265 These options may generate better or worse code; results are highly
3266 dependent on the structure of loops within the source code.
3268 These two options are intended to be removed someday, once
3269 they have helped determine the efficacy of various
3270 approaches to improving loop optimizations.
3272 Please let us (@w{@email{gcc@@gcc.gnu.org}} and @w{@email{fortran@@gnu.org}})
3273 know how use of these options affects
3274 the performance of your production code.
3275 We're very interested in code that runs @emph{slower}
3276 when these options are @emph{enabled}.
3279 Disable any machine-specific peephole optimizations.
3281 @item -fbranch-probabilities
3282 After running a program compiled with @samp{-fprofile-arcs}
3283 (@pxref{Debugging Options,, Options for Debugging Your Program or
3284 @command{gcc}}), you can compile it a second time using
3285 @samp{-fbranch-probabilities}, to improve optimizations based on
3286 guessing the path a branch might take.
3289 With @samp{-fbranch-probabilities}, GCC puts a @samp{REG_EXEC_COUNT}
3290 note on the first instruction of each basic block, and a
3291 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
3292 These can be used to improve optimization. Currently, they are only
3293 used in one place: in @file{reorg.c}, instead of guessing which path a
3294 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
3295 exactly determine which path is taken more often.
3298 @item -fno-guess-branch-probability
3299 Sometimes gcc will opt to guess branch probabilities when none are
3300 available from either profile directed feedback (@samp{-fprofile-arcs})
3301 or @samp{__builtin_expect}. In a hard real-time system, people don't
3302 want different runs of the compiler to produce code that has different
3303 behavior; minimizing non-determinism is of paramount import. This
3304 switch allows users to reduce non-determinism, possibly at the expense
3305 of inferior optimization.
3307 @item -fstrict-aliasing
3308 Allows the compiler to assume the strictest aliasing rules applicable to
3309 the language being compiled. For C (and C++), this activates
3310 optimizations based on the type of expressions. In particular, an
3311 object of one type is assumed never to reside at the same address as an
3312 object of a different type, unless the types are almost the same. For
3313 example, an @code{unsigned int} can alias an @code{int}, but not a
3314 @code{void*} or a @code{double}. A character type may alias any other
3317 Pay special attention to code like this:
3330 The practice of reading from a different union member than the one most
3331 recently written to (called ``type-punning'') is common. Even with
3332 @samp{-fstrict-aliasing}, type-punning is allowed, provided the memory
3333 is accessed through the union type. So, the code above will work as
3334 expected. However, this code might not:
3346 Every language that wishes to perform language-specific alias analysis
3347 should define a function that computes, given an @code{tree}
3348 node, an alias set for the node. Nodes in different alias sets are not
3349 allowed to alias. For an example, see the C front-end function
3350 @code{c_get_alias_set}.
3353 @item -falign-functions
3354 @itemx -falign-functions=@var{n}
3355 Align the start of functions to the next power-of-two greater than
3356 @var{n}, skipping up to @var{n} bytes. For instance,
3357 @samp{-falign-functions=32} aligns functions to the next 32-byte
3358 boundary, but @samp{-falign-functions=24} would align to the next
3359 32-byte boundary only if this can be done by skipping 23 bytes or less.
3361 @samp{-fno-align-functions} and @samp{-falign-functions=1} are
3362 equivalent and mean that functions will not be aligned.
3364 Some assemblers only support this flag when @var{n} is a power of two;
3365 in that case, it is rounded up.
3367 If @var{n} is not specified, use a machine-dependent default.
3369 @item -falign-labels
3370 @itemx -falign-labels=@var{n}
3371 Align all branch targets to a power-of-two boundary, skipping up to
3372 @var{n} bytes like @samp{-falign-functions}. This option can easily
3373 make code slower, because it must insert dummy operations for when the
3374 branch target is reached in the usual flow of the code.
3376 If @samp{-falign-loops} or @samp{-falign-jumps} are applicable and
3377 are greater than this value, then their values are used instead.
3379 If @var{n} is not specified, use a machine-dependent default which is
3380 very likely to be @samp{1}, meaning no alignment.
3383 @itemx -falign-loops=@var{n}
3384 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
3385 like @samp{-falign-functions}. The hope is that the loop will be
3386 executed many times, which will make up for any execution of the dummy
3389 If @var{n} is not specified, use a machine-dependent default.
3392 @itemx -falign-jumps=@var{n}
3393 Align branch targets to a power-of-two boundary, for branch targets
3394 where the targets can only be reached by jumping, skipping up to @var{n}
3395 bytes like @samp{-falign-functions}. In this case, no dummy operations
3398 If @var{n} is not specified, use a machine-dependent default.
3401 Perform optimizations in static single assignment form. Each function's
3402 flow graph is translated into SSA form, optimizations are performed, and
3403 the flow graph is translated back from SSA form. Users should not
3404 specify this option, since it is not yet ready for production use.
3407 Perform dead-code elimination in SSA form. Requires @samp{-fssa}. Like
3408 @samp{-fssa}, this is an experimental feature.
3410 @item -fsingle-precision-constant
3411 Treat floating point constant as single precision constant instead of
3412 implicitly converting it to double precision constant.
3414 @item -frename-registers
3415 Attempt to avoid false dependancies in scheduled code by making use
3416 of registers left over after register allocation. This optimization
3417 will most benefit processors with lots of registers. It can, however,
3418 make debugging impossible, since variables will no longer stay in
3419 a ``home register''.
3421 @item --param @var{name}=@var{value}
3422 In some places, GCC uses various constants to control the amount of
3423 optimization that is done. For example, GCC will not inline functions
3424 that contain more that a certain number of instructions. You can
3425 control some of these constants on the command-line using the
3426 @samp{--param} option.
3428 In each case, the @var{value} is a integer. The allowable choices for
3429 @var{name} are given in the following table:
3432 @item max-delay-slot-insn-search
3433 The maximum number of instructions to consider when looking for an
3434 instruction to fill a delay slot. If more than this arbitrary number of
3435 instructions is searched, the time savings from filling the delay slot
3436 will be minimal so stop searching. Increasing values mean more
3437 aggressive optimization, making the compile time increase with probably
3438 small improvement in executable run time.
3440 @item max-delay-slot-live-search
3441 When trying to fill delay slots, the maximum number of instructions to
3442 consider when searching for a block with valid live register
3443 information. Increasing this arbitrarily chosen value means more
3444 aggressive optimization, increasing the compile time. This parameter
3445 should be removed when the delay slot code is rewritten to maintain the
3448 @item max-gcse-memory
3449 The approximate maximum amount of memory that will be allocated in
3450 order to perform the global common subexpression elimination
3451 optimization. If more memory than specified is required, the
3452 optimization will not be done.
3454 @item max-inline-insns
3455 If an function contains more than this many instructions, it
3456 will not be inlined. This option is precisely equivalent to
3457 @samp{-finline-limit}.
3462 @node Preprocessor Options
3463 @section Options Controlling the Preprocessor
3464 @cindex preprocessor options
3465 @cindex options, preprocessor
3467 These options control the C preprocessor, which is run on each C source
3468 file before actual compilation.
3470 If you use the @samp{-E} option, nothing is done except preprocessing.
3471 Some of these options make sense only together with @samp{-E} because
3472 they cause the preprocessor output to be unsuitable for actual
3476 @item -include @var{file}
3477 Process @var{file} as input before processing the regular input file.
3478 In effect, the contents of @var{file} are compiled first. Any @samp{-D}
3479 and @samp{-U} options on the command line are always processed before
3480 @samp{-include @var{file}}, regardless of the order in which they are
3481 written. All the @samp{-include} and @samp{-imacros} options are
3482 processed in the order in which they are written.
3484 @item -imacros @var{file}
3485 Process @var{file} as input, discarding the resulting output, before
3486 processing the regular input file. Because the output generated from
3487 @var{file} is discarded, the only effect of @samp{-imacros @var{file}}
3488 is to make the macros defined in @var{file} available for use in the
3489 main input. All the @samp{-include} and @samp{-imacros} options are
3490 processed in the order in which they are written.
3492 @item -idirafter @var{dir}
3493 @cindex second include path
3494 Add the directory @var{dir} to the second include path. The directories
3495 on the second include path are searched when a header file is not found
3496 in any of the directories in the main include path (the one that
3499 @item -iprefix @var{prefix}
3500 Specify @var{prefix} as the prefix for subsequent @samp{-iwithprefix}
3503 @item -iwithprefix @var{dir}
3504 Add a directory to the second include path. The directory's name is
3505 made by concatenating @var{prefix} and @var{dir}, where @var{prefix} was
3506 specified previously with @samp{-iprefix}. If you have not specified a
3507 prefix yet, the directory containing the installed passes of the
3508 compiler is used as the default.
3510 @item -iwithprefixbefore @var{dir}
3511 Add a directory to the main include path. The directory's name is made
3512 by concatenating @var{prefix} and @var{dir}, as in the case of
3513 @samp{-iwithprefix}.
3515 @item -isystem @var{dir}
3516 Add a directory to the beginning of the second include path, marking it
3517 as a system directory, so that it gets the same special treatment as
3518 is applied to the standard system directories.
3521 Do not search the standard system directories for header files. Only
3522 the directories you have specified with @samp{-I} options (and the
3523 current directory, if appropriate) are searched. @xref{Directory
3524 Options}, for information on @samp{-I}.
3526 By using both @samp{-nostdinc} and @samp{-I-}, you can limit the include-file
3527 search path to only those directories you specify explicitly.
3531 When searching for a header file in a directory, remap file names if a
3532 file named @file{header.gcc} exists in that directory. This can be used
3533 to work around limitations of file systems with file name restrictions.
3534 The @file{header.gcc} file should contain a series of lines with two
3535 tokens on each line: the first token is the name to map, and the second
3536 token is the actual name to use.
3539 Do not predefine any nonstandard macros. (Including architecture flags).
3542 Run only the C preprocessor. Preprocess all the C source files
3543 specified and output the results to standard output or to the
3544 specified output file.
3547 Tell the preprocessor not to discard comments. Used with the
3551 Tell the preprocessor not to generate @samp{#line} directives.
3552 Used with the @samp{-E} option.
3555 @cindex dependencies, make
3558 Instead of outputting the result of preprocessing, output a rule
3559 suitable for @code{make} describing the dependencies of the main source
3560 file. The preprocessor outputs one @code{make} rule containing the
3561 object file name for that source file, a colon, and the names of all the
3562 included files. Unless overridden explicitly, the object file name
3563 consists of the basename of the source file with any suffix replaced with
3564 object file suffix. If there are many included files then the
3565 rule is split into several lines using @samp{\}-newline.
3567 @samp{-M} implies @samp{-E}.
3571 Like @samp{-M}, but mention only the files included with @samp{#include
3572 "@var{file}"}. System header files included with @samp{#include
3573 <@var{file}>} are omitted.
3577 Like @samp{-M} but the dependency information is written to a file
3578 rather than stdout. @code{gcc} will use the same file name and
3579 directory as the object file, but with the suffix ".d" instead.
3581 This is in addition to compiling the main file as specified ---
3582 @samp{-MD} does not inhibit ordinary compilation the way @samp{-M} does,
3583 unless you also specify @samp{-MG}.
3585 With Mach, you can use the utility @code{md} to merge multiple
3586 dependency files into a single dependency file suitable for using with
3587 the @samp{make} command.
3591 Like @samp{-MD} except mention only user header files, not system
3594 @item -MF @var{file}
3596 When used with @samp{-M} or @samp{-MM}, specifies a file to write the
3597 dependencies to. This allows the preprocessor to write the preprocessed
3598 file to stdout normally. If no @samp{-MF} switch is given, CPP sends
3599 the rules to stdout and suppresses normal preprocessed output.
3601 Another way to specify output of a @code{make} rule is by setting
3602 the environment variable @env{DEPENDENCIES_OUTPUT} (@pxref{Environment
3607 When used with @samp{-M} or @samp{-MM}, @samp{-MG} says to treat missing
3608 header files as generated files and assume they live in the same
3609 directory as the source file. It suppresses preprocessed output, as a
3610 missing header file is ordinarily an error.
3612 This feature is used in automatic updating of makefiles.
3616 This option instructs CPP to add a phony target for each dependency
3617 other than the main file, causing each to depend on nothing. These
3618 dummy rules work around errors @code{make} gives if you remove header
3619 files without updating the @code{Makefile} to match.
3621 This is typical output:-
3624 /tmp/test.o: /tmp/test.c /tmp/test.h
3629 @item -MQ @var{target}
3630 @item -MT @var{target}
3633 By default CPP uses the main file name, including any path, and appends
3634 the object suffix, normally ``.o'', to it to obtain the name of the
3635 target for dependency generation. With @samp{-MT} you can specify a
3636 target yourself, overriding the default one.
3638 If you want multiple targets, you can specify them as a single argument
3639 to @samp{-MT}, or use multiple @samp{-MT} options.
3641 The targets you specify are output in the order they appear on the
3642 command line. @samp{-MQ} is identical to @samp{-MT}, except that the
3643 target name is quoted for Make, but with @samp{-MT} it isn't. For
3644 example, -MT '$(objpfx)foo.o' gives
3647 $(objpfx)foo.o: /tmp/foo.c
3650 but -MQ '$(objpfx)foo.o' gives
3653 $$(objpfx)foo.o: /tmp/foo.c
3656 The default target is automatically quoted, as if it were given with
3660 Print the name of each header file used, in addition to other normal
3663 @item -A@var{question}(@var{answer})
3664 Assert the answer @var{answer} for @var{question}, in case it is tested
3665 with a preprocessing conditional such as @samp{#if
3666 #@var{question}(@var{answer})}. @samp{-A-} disables the standard
3667 assertions that normally describe the target machine.
3670 Define macro @var{macro} with the string @samp{1} as its definition.
3672 @item -D@var{macro}=@var{defn}
3673 Define macro @var{macro} as @var{defn}. All instances of @samp{-D} on
3674 the command line are processed before any @samp{-U} options.
3676 Any @samp{-D} and @samp{-U} options on the command line are processed in
3677 order, and always before @samp{-imacros @var{file}}, regardless of the
3678 order in which they are written.
3681 Undefine macro @var{macro}. @samp{-U} options are evaluated after all
3682 @samp{-D} options, but before any @samp{-include} and @samp{-imacros}
3685 Any @samp{-D} and @samp{-U} options on the command line are processed in
3686 order, and always before @samp{-imacros @var{file}}, regardless of the
3687 order in which they are written.
3690 Tell the preprocessor to output only a list of the macro definitions
3691 that are in effect at the end of preprocessing. Used with the @samp{-E}
3695 Tell the preprocessing to pass all macro definitions into the output, in
3696 their proper sequence in the rest of the output.
3699 Like @samp{-dD} except that the macro arguments and contents are omitted.
3700 Only @samp{#define @var{name}} is included in the output.
3704 Output @samp{#include} directives in addition to the result of
3709 Process ISO standard trigraph sequences. These are three-character
3710 sequences, all starting with @samp{??}, that are defined by ISO C to
3711 stand for single characters. For example, @samp{??/} stands for
3712 @samp{\}, so @samp{'??/n'} is a character constant for a newline. By
3713 default, GCC ignores trigraphs, but in standard-conforming modes it
3714 converts them. See the @samp{-std} and @samp{-ansi} options.
3716 The nine trigraph sequences are
3747 Trigraph support is not popular, so many compilers do not implement it
3748 properly. Portable code should not rely on trigraphs being either
3749 converted or ignored.
3751 @item -Wp\,@var{option}
3752 Pass @var{option} as an option to the preprocessor. If @var{option}
3753 contains commas, it is split into multiple options at the commas.
3756 @node Assembler Options
3757 @section Passing Options to the Assembler
3759 @c prevent bad page break with this line
3760 You can pass options to the assembler.
3763 @item -Wa\,@var{option}
3764 Pass @var{option} as an option to the assembler. If @var{option}
3765 contains commas, it is split into multiple options at the commas.
3769 @section Options for Linking
3770 @cindex link options
3771 @cindex options, linking
3773 These options come into play when the compiler links object files into
3774 an executable output file. They are meaningless if the compiler is
3775 not doing a link step.
3779 @item @var{object-file-name}
3780 A file name that does not end in a special recognized suffix is
3781 considered to name an object file or library. (Object files are
3782 distinguished from libraries by the linker according to the file
3783 contents.) If linking is done, these object files are used as input
3789 If any of these options is used, then the linker is not run, and
3790 object file names should not be used as arguments. @xref{Overall
3794 @item -l@var{library}
3795 @itemx -l @var{library}
3796 Search the library named @var{library} when linking. (The second
3797 alternative with the library as a separate argument is only for
3798 POSIX compliance and is not recommended.)
3800 It makes a difference where in the command you write this option; the
3801 linker searches and processes libraries and object files in the order they
3802 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
3803 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
3804 to functions in @samp{z}, those functions may not be loaded.
3806 The linker searches a standard list of directories for the library,
3807 which is actually a file named @file{lib@var{library}.a}. The linker
3808 then uses this file as if it had been specified precisely by name.
3810 The directories searched include several standard system directories
3811 plus any that you specify with @samp{-L}.
3813 Normally the files found this way are library files---archive files
3814 whose members are object files. The linker handles an archive file by
3815 scanning through it for members which define symbols that have so far
3816 been referenced but not defined. But if the file that is found is an
3817 ordinary object file, it is linked in the usual fashion. The only
3818 difference between using an @samp{-l} option and specifying a file name
3819 is that @samp{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
3820 and searches several directories.
3823 You need this special case of the @samp{-l} option in order to
3824 link an Objective C program.
3827 Do not use the standard system startup files when linking.
3828 The standard system libraries are used normally, unless @option{-nostdlib}
3829 or @option{-nodefaultlibs} is used.
3831 @item -nodefaultlibs
3832 Do not use the standard system libraries when linking.
3833 Only the libraries you specify will be passed to the linker.
3834 The standard startup files are used normally, unless @option{-nostartfiles}
3835 is used. The compiler may generate calls to memcmp, memset, and memcpy
3836 for System V (and ISO C) environments or to bcopy and bzero for
3837 BSD environments. These entries are usually resolved by entries in
3838 libc. These entry points should be supplied through some other
3839 mechanism when this option is specified.
3842 Do not use the standard system startup files or libraries when linking.
3843 No startup files and only the libraries you specify will be passed to
3844 the linker. The compiler may generate calls to memcmp, memset, and memcpy
3845 for System V (and ISO C) environments or to bcopy and bzero for
3846 BSD environments. These entries are usually resolved by entries in
3847 libc. These entry points should be supplied through some other
3848 mechanism when this option is specified.
3850 @cindex @code{-lgcc}, use with @code{-nostdlib}
3851 @cindex @code{-nostdlib} and unresolved references
3852 @cindex unresolved references and @code{-nostdlib}
3853 @cindex @code{-lgcc}, use with @code{-nodefaultlibs}
3854 @cindex @code{-nodefaultlibs} and unresolved references
3855 @cindex unresolved references and @code{-nodefaultlibs}
3856 One of the standard libraries bypassed by @samp{-nostdlib} and
3857 @samp{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
3858 that GCC uses to overcome shortcomings of particular machines, or special
3859 needs for some languages.
3861 (@xref{Interface,,Interfacing to GCC Output}, for more discussion of
3865 (@xref{Interface,,Interfacing to GCC Output,gcc.info,Porting GCC},
3866 for more discussion of @file{libgcc.a}.)
3868 In most cases, you need @file{libgcc.a} even when you want to avoid
3869 other standard libraries. In other words, when you specify @samp{-nostdlib}
3870 or @samp{-nodefaultlibs} you should usually specify @samp{-lgcc} as well.
3871 This ensures that you have no unresolved references to internal GCC
3872 library subroutines. (For example, @samp{__main}, used to ensure C++
3873 constructors will be called; @pxref{Collect2,,@command{collect2}}.)
3876 Remove all symbol table and relocation information from the executable.
3879 On systems that support dynamic linking, this prevents linking with the shared
3880 libraries. On other systems, this option has no effect.
3883 Produce a shared object which can then be linked with other objects to
3884 form an executable. Not all systems support this option. For predictable
3885 results, you must also specify the same set of options that were used to
3886 generate code (@samp{-fpic}, @samp{-fPIC}, or model suboptions)
3887 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
3888 needs to build supplementary stub code for constructors to work. On
3889 multi-libbed systems, @samp{gcc -shared} must select the correct support
3890 libraries to link against. Failing to supply the correct flags may lead
3891 to subtle defects. Supplying them in cases where they are not necessary
3894 @item -shared-libgcc
3895 @itemx -static-libgcc
3896 On systems that provide @file{libgcc} as a shared library, these options
3897 force the use of either the shared or static version respectively.
3898 If no shared version of @file{libgcc} was built when the compiler was
3899 configured, these options have no effect.
3901 There are several situations in which an application should use the
3902 shared @file{libgcc} instead of the static version. The most common
3903 of these is when the application wishes to throw and catch exceptions
3904 across different shared libraries. In that case, each of the libraries
3905 as well as the application itself should use the shared @file{libgcc}.
3907 Therefore, whenever you specify the @samp{-shared} option, the GCC
3908 driver automatically adds @samp{-shared-libgcc}, unless you explicitly
3909 specify @samp{-static-libgcc}. The G++ driver automatically adds
3910 @samp{-shared-libgcc} when you build a main executable as well because
3911 for C++ programs that is typically the right thing to do.
3912 (Exception-handling will not work reliably otherwise.)
3914 However, when linking a main executable written in C, you must
3915 explicitly say @samp{-shared-libgcc} if you want to use the shared
3919 Bind references to global symbols when building a shared object. Warn
3920 about any unresolved references (unless overridden by the link editor
3921 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
3924 @item -Xlinker @var{option}
3925 Pass @var{option} as an option to the linker. You can use this to
3926 supply system-specific linker options which GCC does not know how to
3929 If you want to pass an option that takes an argument, you must use
3930 @samp{-Xlinker} twice, once for the option and once for the argument.
3931 For example, to pass @samp{-assert definitions}, you must write
3932 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
3933 @samp{-Xlinker "-assert definitions"}, because this passes the entire
3934 string as a single argument, which is not what the linker expects.
3936 @item -Wl\,@var{option}
3937 Pass @var{option} as an option to the linker. If @var{option} contains
3938 commas, it is split into multiple options at the commas.
3940 @item -u @var{symbol}
3941 Pretend the symbol @var{symbol} is undefined, to force linking of
3942 library modules to define it. You can use @samp{-u} multiple times with
3943 different symbols to force loading of additional library modules.
3946 @node Directory Options
3947 @section Options for Directory Search
3948 @cindex directory options
3949 @cindex options, directory search
3952 These options specify directories to search for header files, for
3953 libraries and for parts of the compiler:
3957 Add the directory @var{dir} to the head of the list of directories to be
3958 searched for header files. This can be used to override a system header
3959 file, substituting your own version, since these directories are
3960 searched before the system header file directories. However, you should
3961 not use this option to add directories that contain vendor-supplied
3962 system header files (use @samp{-isystem} for that). If you use more than
3963 one @samp{-I} option, the directories are scanned in left-to-right
3964 order; the standard system directories come after.
3967 Any directories you specify with @samp{-I} options before the @samp{-I-}
3968 option are searched only for the case of @samp{#include "@var{file}"};
3969 they are not searched for @samp{#include <@var{file}>}.
3971 If additional directories are specified with @samp{-I} options after
3972 the @samp{-I-}, these directories are searched for all @samp{#include}
3973 directives. (Ordinarily @emph{all} @samp{-I} directories are used
3976 In addition, the @samp{-I-} option inhibits the use of the current
3977 directory (where the current input file came from) as the first search
3978 directory for @samp{#include "@var{file}"}. There is no way to
3979 override this effect of @samp{-I-}. With @samp{-I.} you can specify
3980 searching the directory which was current when the compiler was
3981 invoked. That is not exactly the same as what the preprocessor does
3982 by default, but it is often satisfactory.
3984 @samp{-I-} does not inhibit the use of the standard system directories
3985 for header files. Thus, @samp{-I-} and @samp{-nostdinc} are
3989 Add directory @var{dir} to the list of directories to be searched
3992 @item -B@var{prefix}
3993 This option specifies where to find the executables, libraries,
3994 include files, and data files of the compiler itself.
3996 The compiler driver program runs one or more of the subprograms
3997 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
3998 @var{prefix} as a prefix for each program it tries to run, both with and
3999 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
4001 For each subprogram to be run, the compiler driver first tries the
4002 @samp{-B} prefix, if any. If that name is not found, or if @samp{-B}
4003 was not specified, the driver tries two standard prefixes, which are
4004 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc-lib/}. If neither of
4005 those results in a file name that is found, the unmodified program
4006 name is searched for using the directories specified in your
4007 @env{PATH} environment variable.
4009 @samp{-B} prefixes that effectively specify directory names also apply
4010 to libraries in the linker, because the compiler translates these
4011 options into @samp{-L} options for the linker. They also apply to
4012 includes files in the preprocessor, because the compiler translates these
4013 options into @samp{-isystem} options for the preprocessor. In this case,
4014 the compiler appends @samp{include} to the prefix.
4016 The run-time support file @file{libgcc.a} can also be searched for using
4017 the @samp{-B} prefix, if needed. If it is not found there, the two
4018 standard prefixes above are tried, and that is all. The file is left
4019 out of the link if it is not found by those means.
4021 Another way to specify a prefix much like the @samp{-B} prefix is to use
4022 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
4025 @item -specs=@var{file}
4026 Process @var{file} after the compiler reads in the standard @file{specs}
4027 file, in order to override the defaults that the @file{gcc} driver
4028 program uses when determining what switches to pass to @file{cc1},
4029 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
4030 @samp{-specs=}@var{file} can be specified on the command line, and they
4031 are processed in order, from left to right.
4037 @section Specifying subprocesses and the switches to pass to them
4039 @command{gcc} is a driver program. It performs its job by invoking a
4040 sequence of other programs to do the work of compiling, assembling and
4041 linking. GCC interprets its command-line parameters and uses these to
4042 deduce which programs it should invoke, and which command-line options
4043 it ought to place on their command lines. This behaviour is controlled
4044 by @dfn{spec strings}. In most cases there is one spec string for each
4045 program that GCC can invoke, but a few programs have multiple spec
4046 strings to control their behaviour. The spec strings built into GCC can
4047 be overridden by using the @samp{-specs=} command-line switch to specify
4050 @dfn{Spec files} are plaintext files that are used to construct spec
4051 strings. They consist of a sequence of directives separated by blank
4052 lines. The type of directive is determined by the first non-whitespace
4053 character on the line and it can be one of the following:
4056 @item %@var{command}
4057 Issues a @var{command} to the spec file processor. The commands that can
4061 @item %include <@var{file}>
4063 Search for @var{file} and insert its text at the current point in the
4066 @item %include_noerr <@var{file}>
4067 @cindex %include_noerr
4068 Just like @samp{%include}, but do not generate an error message if the include
4069 file cannot be found.
4071 @item %rename @var{old_name} @var{new_name}
4073 Rename the spec string @var{old_name} to @var{new_name}.
4077 @item *[@var{spec_name}]:
4078 This tells the compiler to create, override or delete the named spec
4079 string. All lines after this directive up to the next directive or
4080 blank line are considered to be the text for the spec string. If this
4081 results in an empty string then the spec will be deleted. (Or, if the
4082 spec did not exist, then nothing will happened.) Otherwise, if the spec
4083 does not currently exist a new spec will be created. If the spec does
4084 exist then its contents will be overridden by the text of this
4085 directive, unless the first character of that text is the @samp{+}
4086 character, in which case the text will be appended to the spec.
4088 @item [@var{suffix}]:
4089 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
4090 and up to the next directive or blank line are considered to make up the
4091 spec string for the indicated suffix. When the compiler encounters an
4092 input file with the named suffix, it will processes the spec string in
4093 order to work out how to compile that file. For example:
4100 This says that any input file whose name ends in @samp{.ZZ} should be
4101 passed to the program @samp{z-compile}, which should be invoked with the
4102 command-line switch @samp{-input} and with the result of performing the
4103 @samp{%i} substitution. (See below.)
4105 As an alternative to providing a spec string, the text that follows a
4106 suffix directive can be one of the following:
4109 @item @@@var{language}
4110 This says that the suffix is an alias for a known @var{language}. This is
4111 similar to using the @option{-x} command-line switch to GCC to specify a
4112 language explicitly. For example:
4119 Says that .ZZ files are, in fact, C++ source files.
4122 This causes an error messages saying:
4125 @var{name} compiler not installed on this system.
4129 GCC already has an extensive list of suffixes built into it.
4130 This directive will add an entry to the end of the list of suffixes, but
4131 since the list is searched from the end backwards, it is effectively
4132 possible to override earlier entries using this technique.
4136 GCC has the following spec strings built into it. Spec files can
4137 override these strings or create their own. Note that individual
4138 targets can also add their own spec strings to this list.
4141 asm Options to pass to the assembler
4142 asm_final Options to pass to the assembler post-processor
4143 cpp Options to pass to the C preprocessor
4144 cc1 Options to pass to the C compiler
4145 cc1plus Options to pass to the C++ compiler
4146 endfile Object files to include at the end of the link
4147 link Options to pass to the linker
4148 lib Libraries to include on the command line to the linker
4149 libgcc Decides which GCC support library to pass to the linker
4150 linker Sets the name of the linker
4151 predefines Defines to be passed to the C preprocessor
4152 signed_char Defines to pass to CPP to say whether @code{char} is signed by default
4153 startfile Object files to include at the start of the link
4156 Here is a small example of a spec file:
4162 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
4165 This example renames the spec called @samp{lib} to @samp{old_lib} and
4166 then overrides the previous definition of @samp{lib} with a new one.
4167 The new definition adds in some extra command-line options before
4168 including the text of the old definition.
4170 @dfn{Spec strings} are a list of command-line options to be passed to their
4171 corresponding program. In addition, the spec strings can contain
4172 @samp{%}-prefixed sequences to substitute variable text or to
4173 conditionally insert text into the command line. Using these constructs
4174 it is possible to generate quite complex command lines.
4176 Here is a table of all defined @samp{%}-sequences for spec
4177 strings. Note that spaces are not generated automatically around the
4178 results of expanding these sequences. Therefore you can concatenate them
4179 together or combine them with constant text in a single argument.
4183 Substitute one @samp{%} into the program name or argument.
4186 Substitute the name of the input file being processed.
4189 Substitute the basename of the input file being processed.
4190 This is the substring up to (and not including) the last period
4191 and not including the directory.
4194 This is the same as @samp{%b}, but include the file suffix (text after
4198 Marks the argument containing or following the @samp{%d} as a
4199 temporary file name, so that that file will be deleted if GCC exits
4200 successfully. Unlike @samp{%g}, this contributes no text to the
4203 @item %g@var{suffix}
4204 Substitute a file name that has suffix @var{suffix} and is chosen
4205 once per compilation, and mark the argument in the same way as
4206 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
4207 name is now chosen in a way that is hard to predict even when previously
4208 chosen file names are known. For example, @samp{%g.s ... %g.o ... %g.s}
4209 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
4210 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
4211 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
4212 was simply substituted with a file name chosen once per compilation,
4213 without regard to any appended suffix (which was therefore treated
4214 just like ordinary text), making such attacks more likely to succeed.
4216 @item %u@var{suffix}
4217 Like @samp{%g}, but generates a new temporary file name even if
4218 @samp{%u@var{suffix}} was already seen.
4220 @item %U@var{suffix}
4221 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
4222 new one if there is no such last file name. In the absence of any
4223 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
4224 the same suffix @emph{space}, so @samp{%g.s ... %U.s ... %g.s ... %U.s}
4225 would involve the generation of two distinct file names, one
4226 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
4227 simply substituted with a file name chosen for the previous @samp{%u},
4228 without regard to any appended suffix.
4230 @item %j@var{SUFFIX}
4231 Substitutes the name of the HOST_BIT_BUCKET, if any, and if it is
4232 writable, and if save-temps is off; otherwise, substitute the name
4233 of a temporary file, just like @samp{%u}. This temporary file is not
4234 meant for communication between processes, but rather as a junk
4237 @item %.@var{SUFFIX}
4238 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
4239 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
4240 terminated by the next space or %.
4243 Marks the argument containing or following the @samp{%w} as the
4244 designated output file of this compilation. This puts the argument
4245 into the sequence of arguments that @samp{%o} will substitute later.
4248 Substitutes the names of all the output files, with spaces
4249 automatically placed around them. You should write spaces
4250 around the @samp{%o} as well or the results are undefined.
4251 @samp{%o} is for use in the specs for running the linker.
4252 Input files whose names have no recognized suffix are not compiled
4253 at all, but they are included among the output files, so they will
4257 Substitutes the suffix for object files. Note that this is
4258 handled specially when it immediately follows @samp{%g, %u, or %U},
4259 because of the need for those to form complete file names. The
4260 handling is such that @samp{%O} is treated exactly as if it had already
4261 been substituted, except that @samp{%g, %u, and %U} do not currently
4262 support additional @var{suffix} characters following @samp{%O} as they would
4263 following, for example, @samp{.o}.
4266 Substitutes the standard macro predefinitions for the
4267 current target machine. Use this when running @code{cpp}.
4270 Like @samp{%p}, but puts @samp{__} before and after the name of each
4271 predefined macro, except for macros that start with @samp{__} or with
4272 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
4276 Substitute a @samp{-iprefix} option made from GCC_EXEC_PREFIX.
4279 Current argument is the name of a library or startup file of some sort.
4280 Search for that file in a standard list of directories and substitute
4281 the full name found.
4284 Print @var{str} as an error message. @var{str} is terminated by a newline.
4285 Use this when inconsistent options are detected.
4288 Output @samp{-} if the input for the current command is coming from a pipe.
4291 Substitute the contents of spec string @var{name} at this point.
4294 Like @samp{%(...)} but put @samp{__} around @samp{-D} arguments.
4296 @item %x@{@var{option}@}
4297 Accumulate an option for @samp{%X}.
4300 Output the accumulated linker options specified by @samp{-Wl} or a @samp{%x}
4304 Output the accumulated assembler options specified by @samp{-Wa}.
4307 Output the accumulated preprocessor options specified by @samp{-Wp}.
4310 Substitute the major version number of GCC.
4311 (For version 2.9.5, this is 2.)
4314 Substitute the minor version number of GCC.
4315 (For version 2.9.5, this is 9.)
4318 Substitute the patch level number of GCC.
4319 (For version 2.9.5, this is 5.)
4322 Process the @code{asm} spec. This is used to compute the
4323 switches to be passed to the assembler.
4326 Process the @code{asm_final} spec. This is a spec string for
4327 passing switches to an assembler post-processor, if such a program is
4331 Process the @code{link} spec. This is the spec for computing the
4332 command line passed to the linker. Typically it will make use of the
4333 @samp{%L %G %S %D and %E} sequences.
4336 Dump out a @samp{-L} option for each directory that GCC believes might
4337 contain startup files. If the target supports multilibs then the
4338 current multilib directory will be prepended to each of these paths.
4341 Output the multilib directory with directory seperators replaced with
4342 "_". If multilib directories are not set, or the multilib directory is
4343 "." then this option emits nothing.
4346 Process the @code{lib} spec. This is a spec string for deciding which
4347 libraries should be included on the command line to the linker.
4350 Process the @code{libgcc} spec. This is a spec string for deciding
4351 which GCC support library should be included on the command line to the linker.
4354 Process the @code{startfile} spec. This is a spec for deciding which
4355 object files should be the first ones passed to the linker. Typically
4356 this might be a file named @file{crt0.o}.
4359 Process the @code{endfile} spec. This is a spec string that specifies
4360 the last object files that will be passed to the linker.
4363 Process the @code{cpp} spec. This is used to construct the arguments
4364 to be passed to the C preprocessor.
4367 Process the @code{signed_char} spec. This is intended to be used
4368 to tell cpp whether a char is signed. It typically has the definition:
4370 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
4374 Process the @code{cc1} spec. This is used to construct the options to be
4375 passed to the actual C compiler (@samp{cc1}).
4378 Process the @code{cc1plus} spec. This is used to construct the options to be
4379 passed to the actual C++ compiler (@samp{cc1plus}).
4382 Substitute the variable part of a matched option. See below.
4383 Note that each comma in the substituted string is replaced by
4387 Substitutes the @code{-S} switch, if that switch was given to GCC.
4388 If that switch was not specified, this substitutes nothing. Note that
4389 the leading dash is omitted when specifying this option, and it is
4390 automatically inserted if the substitution is performed. Thus the spec
4391 string @samp{%@{foo@}} would match the command-line option @samp{-foo}
4392 and would output the command line option @samp{-foo}.
4394 @item %W@{@code{S}@}
4395 Like %@{@code{S}@} but mark last argument supplied within as a file to be
4398 @item %@{@code{S}*@}
4399 Substitutes all the switches specified to GCC whose names start
4400 with @code{-S}, but which also take an argument. This is used for
4401 switches like @samp{-o, -D, -I}, etc. GCC considers @samp{-o foo} as being
4402 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
4403 text, including the space. Thus two arguments would be generated.
4405 @item %@{^@code{S}*@}
4406 Like %@{@code{S}*@}, but don't put a blank between a switch and its
4407 argument. Thus %@{^o*@} would only generate one argument, not two.
4409 @item %@{@code{S}*&@code{T}*@}
4410 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
4411 (the order of @code{S} and @code{T} in the spec is not significant).
4412 There can be any number of ampersand-separated variables; for each the
4413 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
4415 @item %@{<@code{S}@}
4416 Remove all occurrences of @code{-S} from the command line. Note - this
4417 command is position dependent. @samp{%} commands in the spec string
4418 before this option will see @code{-S}, @samp{%} commands in the spec
4419 string after this option will not.
4421 @item %@{@code{S}*:@code{X}@}
4422 Substitutes @code{X} if one or more switches whose names start with
4423 @code{-S} are specified to GCC. Note that the tail part of the
4424 @code{-S} option (i.e. the part matched by the @samp{*}) will be substituted
4425 for each occurrence of @samp{%*} within @code{X}.
4427 @item %@{@code{S}:@code{X}@}
4428 Substitutes @code{X}, but only if the @samp{-S} switch was given to GCC.
4430 @item %@{!@code{S}:@code{X}@}
4431 Substitutes @code{X}, but only if the @samp{-S} switch was @emph{not} given to GCC.
4433 @item %@{|@code{S}:@code{X}@}
4434 Like %@{@code{S}:@code{X}@}, but if no @code{S} switch, substitute @samp{-}.
4436 @item %@{|!@code{S}:@code{X}@}
4437 Like %@{!@code{S}:@code{X}@}, but if there is an @code{S} switch, substitute @samp{-}.
4439 @item %@{.@code{S}:@code{X}@}
4440 Substitutes @code{X}, but only if processing a file with suffix @code{S}.
4442 @item %@{!.@code{S}:@code{X}@}
4443 Substitutes @code{X}, but only if @emph{not} processing a file with suffix @code{S}.
4445 @item %@{@code{S}|@code{P}:@code{X}@}
4446 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC. This may be
4447 combined with @samp{!} and @samp{.} sequences as well, although they
4448 have a stronger binding than the @samp{|}. For example a spec string
4452 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
4455 will output the following command-line options from the following input
4456 command-line options:
4461 -d fred.c -foo -baz -boggle
4462 -d jim.d -bar -baz -boggle
4467 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or
4468 %@{!@code{S}:@code{X}@} construct may contain other nested @samp{%} constructs
4469 or spaces, or even newlines. They are processed as usual, as described
4472 The @samp{-O, -f, -m, and -W} switches are handled specifically in these
4473 constructs. If another value of @samp{-O} or the negated form of a @samp{-f, -m, or
4474 -W} switch is found later in the command line, the earlier switch
4475 value is ignored, except with @{@code{S}*@} where @code{S} is just one
4476 letter, which passes all matching options.
4478 The character @samp{|} at the beginning of the predicate text is used to indicate
4479 that a command should be piped to the following command, but only if @samp{-pipe}
4482 It is built into GCC which switches take arguments and which do not.
4483 (You might think it would be useful to generalize this to allow each
4484 compiler's spec to say which switches take arguments. But this cannot
4485 be done in a consistent fashion. GCC cannot even decide which input
4486 files have been specified without knowing which switches take arguments,
4487 and it must know which input files to compile in order to tell which
4490 GCC also knows implicitly that arguments starting in @samp{-l} are to be
4491 treated as compiler output files, and passed to the linker in their
4492 proper position among the other output files.
4494 @c man begin OPTIONS
4496 @node Target Options
4497 @section Specifying Target Machine and Compiler Version
4498 @cindex target options
4499 @cindex cross compiling
4500 @cindex specifying machine version
4501 @cindex specifying compiler version and target machine
4502 @cindex compiler version, specifying
4503 @cindex target machine, specifying
4505 By default, GCC compiles code for the same type of machine that you
4506 are using. However, it can also be installed as a cross-compiler, to
4507 compile for some other type of machine. In fact, several different
4508 configurations of GCC, for different target machines, can be
4509 installed side by side. Then you specify which one to use with the
4512 In addition, older and newer versions of GCC can be installed side
4513 by side. One of them (probably the newest) will be the default, but
4514 you may sometimes wish to use another.
4517 @item -b @var{machine}
4518 The argument @var{machine} specifies the target machine for compilation.
4519 This is useful when you have installed GCC as a cross-compiler.
4521 The value to use for @var{machine} is the same as was specified as the
4522 machine type when configuring GCC as a cross-compiler. For
4523 example, if a cross-compiler was configured with @samp{configure
4524 i386v}, meaning to compile for an 80386 running System V, then you
4525 would specify @samp{-b i386v} to run that cross compiler.
4527 When you do not specify @samp{-b}, it normally means to compile for
4528 the same type of machine that you are using.
4530 @item -V @var{version}
4531 The argument @var{version} specifies which version of GCC to run.
4532 This is useful when multiple versions are installed. For example,
4533 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
4535 The default version, when you do not specify @samp{-V}, is the last
4536 version of GCC that you installed.
4539 The @samp{-b} and @samp{-V} options actually work by controlling part of
4540 the file name used for the executable files and libraries used for
4541 compilation. A given version of GCC, for a given target machine, is
4542 normally kept in the directory @file{/usr/local/lib/gcc-lib/@var{machine}/@var{version}}.@refill
4544 Thus, sites can customize the effect of @samp{-b} or @samp{-V} either by
4545 changing the names of these directories or adding alternate names (or
4546 symbolic links). If in directory @file{/usr/local/lib/gcc-lib/} the
4547 file @file{80386} is a link to the file @file{i386v}, then @samp{-b
4548 80386} becomes an alias for @samp{-b i386v}.
4550 In one respect, the @samp{-b} or @samp{-V} do not completely change
4551 to a different compiler: the top-level driver program @command{gcc}
4552 that you originally invoked continues to run and invoke the other
4553 executables (preprocessor, compiler per se, assembler and linker)
4554 that do the real work. However, since no real work is done in the
4555 driver program, it usually does not matter that the driver program
4556 in use is not the one for the specified target. It is common for the
4557 interface to the other executables to change incompatibly between
4558 compiler versions, so unless the version specified is very close to that
4559 of the driver (for example, @samp{-V 3.0} with a driver program from GCC
4560 version 3.0.1), use of @samp{-V} may not work; for example, using
4561 @samp{-V 2.95.2} will not work with a driver program from GCC 3.0.
4563 The only way that the driver program depends on the target machine is
4564 in the parsing and handling of special machine-specific options.
4565 However, this is controlled by a file which is found, along with the
4566 other executables, in the directory for the specified version and
4567 target machine. As a result, a single installed driver program adapts
4568 to any specified target machine, and sufficiently similar compiler
4571 The driver program executable does control one significant thing,
4572 however: the default version and target machine. Therefore, you can
4573 install different instances of the driver program, compiled for
4574 different targets or versions, under different names.
4576 For example, if the driver for version 2.0 is installed as @command{ogcc}
4577 and that for version 2.1 is installed as @command{gcc}, then the command
4578 @command{gcc} will use version 2.1 by default, while @command{ogcc} will use
4579 2.0 by default. However, you can choose either version with either
4580 command with the @samp{-V} option.
4582 @node Submodel Options
4583 @section Hardware Models and Configurations
4584 @cindex submodel options
4585 @cindex specifying hardware config
4586 @cindex hardware models and configurations, specifying
4587 @cindex machine dependent options
4589 Earlier we discussed the standard option @samp{-b} which chooses among
4590 different installed compilers for completely different target
4591 machines, such as Vax vs. 68000 vs. 80386.
4593 In addition, each of these target machine types can have its own
4594 special options, starting with @samp{-m}, to choose among various
4595 hardware models or configurations---for example, 68010 vs 68020,
4596 floating coprocessor or none. A single installed version of the
4597 compiler can compile for any model or configuration, according to the
4600 Some configurations of the compiler also support additional special
4601 options, usually for compatibility with other compilers on the same
4605 These options are defined by the macro @code{TARGET_SWITCHES} in the
4606 machine description. The default for the options is also defined by
4607 that macro, which enables you to change the defaults.
4622 * RS/6000 and PowerPC Options::
4627 * Intel 960 Options::
4628 * DEC Alpha Options::
4632 * System V Options::
4633 * TMS320C3x/C4x Options::
4643 @node M680x0 Options
4644 @subsection M680x0 Options
4645 @cindex M680x0 options
4647 These are the @samp{-m} options defined for the 68000 series. The default
4648 values for these options depends on which style of 68000 was selected when
4649 the compiler was configured; the defaults for the most common choices are
4655 Generate output for a 68000. This is the default
4656 when the compiler is configured for 68000-based systems.
4658 Use this option for microcontrollers with a 68000 or EC000 core,
4659 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
4663 Generate output for a 68020. This is the default
4664 when the compiler is configured for 68020-based systems.
4667 Generate output containing 68881 instructions for floating point.
4668 This is the default for most 68020 systems unless @samp{-nfp} was
4669 specified when the compiler was configured.
4672 Generate output for a 68030. This is the default when the compiler is
4673 configured for 68030-based systems.
4676 Generate output for a 68040. This is the default when the compiler is
4677 configured for 68040-based systems.
4679 This option inhibits the use of 68881/68882 instructions that have to be
4680 emulated by software on the 68040. Use this option if your 68040 does not
4681 have code to emulate those instructions.
4684 Generate output for a 68060. This is the default when the compiler is
4685 configured for 68060-based systems.
4687 This option inhibits the use of 68020 and 68881/68882 instructions that
4688 have to be emulated by software on the 68060. Use this option if your 68060
4689 does not have code to emulate those instructions.
4692 Generate output for a CPU32. This is the default
4693 when the compiler is configured for CPU32-based systems.
4695 Use this option for microcontrollers with a
4696 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
4697 68336, 68340, 68341, 68349 and 68360.
4700 Generate output for a 520X "coldfire" family cpu. This is the default
4701 when the compiler is configured for 520X-based systems.
4703 Use this option for microcontroller with a 5200 core, including
4704 the MCF5202, MCF5203, MCF5204 and MCF5202.
4708 Generate output for a 68040, without using any of the new instructions.
4709 This results in code which can run relatively efficiently on either a
4710 68020/68881 or a 68030 or a 68040. The generated code does use the
4711 68881 instructions that are emulated on the 68040.
4714 Generate output for a 68060, without using any of the new instructions.
4715 This results in code which can run relatively efficiently on either a
4716 68020/68881 or a 68030 or a 68040. The generated code does use the
4717 68881 instructions that are emulated on the 68060.
4720 Generate output containing Sun FPA instructions for floating point.
4723 Generate output containing library calls for floating point.
4724 @strong{Warning:} the requisite libraries are not available for all m68k
4725 targets. Normally the facilities of the machine's usual C compiler are
4726 used, but this can't be done directly in cross-compilation. You must
4727 make your own arrangements to provide suitable library functions for
4728 cross-compilation. The embedded targets @samp{m68k-*-aout} and
4729 @samp{m68k-*-coff} do provide software floating point support.
4732 Consider type @code{int} to be 16 bits wide, like @code{short int}.
4735 Do not use the bit-field instructions. The @samp{-m68000}, @samp{-mcpu32}
4736 and @samp{-m5200} options imply @w{@samp{-mnobitfield}}.
4739 Do use the bit-field instructions. The @samp{-m68020} option implies
4740 @samp{-mbitfield}. This is the default if you use a configuration
4741 designed for a 68020.
4744 Use a different function-calling convention, in which functions
4745 that take a fixed number of arguments return with the @code{rtd}
4746 instruction, which pops their arguments while returning. This
4747 saves one instruction in the caller since there is no need to pop
4748 the arguments there.
4750 This calling convention is incompatible with the one normally
4751 used on Unix, so you cannot use it if you need to call libraries
4752 compiled with the Unix compiler.
4754 Also, you must provide function prototypes for all functions that
4755 take variable numbers of arguments (including @code{printf});
4756 otherwise incorrect code will be generated for calls to those
4759 In addition, seriously incorrect code will result if you call a
4760 function with too many arguments. (Normally, extra arguments are
4761 harmlessly ignored.)
4763 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
4764 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
4767 @itemx -mno-align-int
4768 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
4769 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
4770 boundary (@samp{-malign-int}) or a 16-bit boundary (@samp{-mno-align-int}).
4771 Aligning variables on 32-bit boundaries produces code that runs somewhat
4772 faster on processors with 32-bit busses at the expense of more memory.
4774 @strong{Warning:} if you use the @samp{-malign-int} switch, GCC will
4775 align structures containing the above types differently than
4776 most published application binary interface specifications for the m68k.
4779 Use the pc-relative addressing mode of the 68000 directly, instead of
4780 using a global offset table. At present, this option implies -fpic,
4781 allowing at most a 16-bit offset for pc-relative addressing. -fPIC is
4782 not presently supported with -mpcrel, though this could be supported for
4783 68020 and higher processors.
4785 @item -mno-strict-align
4786 @itemx -mstrict-align
4787 @kindex -mstrict-align
4788 Do not (do) assume that unaligned memory references will be handled by
4793 @node M68hc1x Options
4794 @subsection M68hc1x Options
4795 @cindex M68hc1x options
4797 These are the @samp{-m} options defined for the 68hc11 and 68hc12
4798 microcontrollers. The default values for these options depends on
4799 which style of microcontroller was selected when the compiler was configured;
4800 the defaults for the most common choices are given below.
4805 Generate output for a 68HC11. This is the default
4806 when the compiler is configured for 68HC11-based systems.
4810 Generate output for a 68HC12. This is the default
4811 when the compiler is configured for 68HC12-based systems.
4814 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
4818 Consider type @code{int} to be 16 bits wide, like @code{short int}.
4820 @item -msoft-reg-count=@var{count}
4821 Specify the number of pseudo-soft registers which are used for the
4822 code generation. The maximum number is 32. Using more pseudo-soft
4823 register may or may not result in better code depending on the program.
4824 The default is 4 for 68HC11 and 2 for 68HC12.
4829 @subsection VAX Options
4832 These @samp{-m} options are defined for the Vax:
4836 Do not output certain jump instructions (@code{aobleq} and so on)
4837 that the Unix assembler for the Vax cannot handle across long
4841 Do output those jump instructions, on the assumption that you
4842 will assemble with the GNU assembler.
4845 Output code for g-format floating point numbers instead of d-format.
4849 @subsection SPARC Options
4850 @cindex SPARC options
4852 These @samp{-m} switches are supported on the SPARC:
4857 Specify @samp{-mapp-regs} to generate output using the global registers
4858 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
4861 To be fully SVR4 ABI compliant at the cost of some performance loss,
4862 specify @samp{-mno-app-regs}. You should compile libraries and system
4863 software with this option.
4867 Generate output containing floating point instructions. This is the
4872 Generate output containing library calls for floating point.
4873 @strong{Warning:} the requisite libraries are not available for all SPARC
4874 targets. Normally the facilities of the machine's usual C compiler are
4875 used, but this cannot be done directly in cross-compilation. You must make
4876 your own arrangements to provide suitable library functions for
4877 cross-compilation. The embedded targets @samp{sparc-*-aout} and
4878 @samp{sparclite-*-*} do provide software floating point support.
4880 @samp{-msoft-float} changes the calling convention in the output file;
4881 therefore, it is only useful if you compile @emph{all} of a program with
4882 this option. In particular, you need to compile @file{libgcc.a}, the
4883 library that comes with GCC, with @samp{-msoft-float} in order for
4886 @item -mhard-quad-float
4887 Generate output containing quad-word (long double) floating point
4890 @item -msoft-quad-float
4891 Generate output containing library calls for quad-word (long double)
4892 floating point instructions. The functions called are those specified
4893 in the SPARC ABI. This is the default.
4895 As of this writing, there are no sparc implementations that have hardware
4896 support for the quad-word floating point instructions. They all invoke
4897 a trap handler for one of these instructions, and then the trap handler
4898 emulates the effect of the instruction. Because of the trap handler overhead,
4899 this is much slower than calling the ABI library routines. Thus the
4900 @samp{-msoft-quad-float} option is the default.
4904 With @samp{-mepilogue} (the default), the compiler always emits code for
4905 function exit at the end of each function. Any function exit in
4906 the middle of the function (such as a return statement in C) will
4907 generate a jump to the exit code at the end of the function.
4909 With @samp{-mno-epilogue}, the compiler tries to emit exit code inline
4910 at every function exit.
4914 With @samp{-mflat}, the compiler does not generate save/restore instructions
4915 and will use a "flat" or single register window calling convention.
4916 This model uses %i7 as the frame pointer and is compatible with the normal
4917 register window model. Code from either may be intermixed.
4918 The local registers and the input registers (0-5) are still treated as
4919 "call saved" registers and will be saved on the stack as necessary.
4921 With @samp{-mno-flat} (the default), the compiler emits save/restore
4922 instructions (except for leaf functions) and is the normal mode of operation.
4924 @item -mno-unaligned-doubles
4925 @itemx -munaligned-doubles
4926 Assume that doubles have 8 byte alignment. This is the default.
4928 With @samp{-munaligned-doubles}, GCC assumes that doubles have 8 byte
4929 alignment only if they are contained in another type, or if they have an
4930 absolute address. Otherwise, it assumes they have 4 byte alignment.
4931 Specifying this option avoids some rare compatibility problems with code
4932 generated by other compilers. It is not the default because it results
4933 in a performance loss, especially for floating point code.
4935 @item -mno-faster-structs
4936 @itemx -mfaster-structs
4937 With @samp{-mfaster-structs}, the compiler assumes that structures
4938 should have 8 byte alignment. This enables the use of pairs of
4939 @code{ldd} and @code{std} instructions for copies in structure
4940 assignment, in place of twice as many @code{ld} and @code{st} pairs.
4941 However, the use of this changed alignment directly violates the Sparc
4942 ABI. Thus, it's intended only for use on targets where the developer
4943 acknowledges that their resulting code will not be directly in line with
4944 the rules of the ABI.
4948 These two options select variations on the SPARC architecture.
4950 By default (unless specifically configured for the Fujitsu SPARClite),
4951 GCC generates code for the v7 variant of the SPARC architecture.
4953 @samp{-mv8} will give you SPARC v8 code. The only difference from v7
4954 code is that the compiler emits the integer multiply and integer
4955 divide instructions which exist in SPARC v8 but not in SPARC v7.
4957 @samp{-msparclite} will give you SPARClite code. This adds the integer
4958 multiply, integer divide step and scan (@code{ffs}) instructions which
4959 exist in SPARClite but not in SPARC v7.
4961 These options are deprecated and will be deleted in a future GCC release.
4962 They have been replaced with @samp{-mcpu=xxx}.
4966 These two options select the processor for which the code is optimised.
4968 With @samp{-mcypress} (the default), the compiler optimizes code for the
4969 Cypress CY7C602 chip, as used in the SparcStation/SparcServer 3xx series.
4970 This is also appropriate for the older SparcStation 1, 2, IPX etc.
4972 With @samp{-msupersparc} the compiler optimizes code for the SuperSparc cpu, as
4973 used in the SparcStation 10, 1000 and 2000 series. This flag also enables use
4974 of the full SPARC v8 instruction set.
4976 These options are deprecated and will be deleted in a future GCC release.
4977 They have been replaced with @samp{-mcpu=xxx}.
4979 @item -mcpu=@var{cpu_type}
4980 Set the instruction set, register set, and instruction scheduling parameters
4981 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
4982 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
4983 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
4984 @samp{sparclet}, @samp{tsc701}, @samp{v9}, and @samp{ultrasparc}.
4986 Default instruction scheduling parameters are used for values that select
4987 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
4988 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
4990 Here is a list of each supported architecture and their supported
4995 v8: supersparc, hypersparc
4996 sparclite: f930, f934, sparclite86x
5001 @item -mtune=@var{cpu_type}
5002 Set the instruction scheduling parameters for machine type
5003 @var{cpu_type}, but do not set the instruction set or register set that the
5004 option @samp{-mcpu=}@var{cpu_type} would.
5006 The same values for @samp{-mcpu=}@var{cpu_type} are used for
5007 @samp{-mtune=}@*@var{cpu_type}, though the only useful values are those that
5008 select a particular cpu implementation: @samp{cypress}, @samp{supersparc},
5009 @samp{hypersparc}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
5010 @samp{tsc701}, @samp{ultrasparc}.
5014 These @samp{-m} switches are supported in addition to the above
5015 on the SPARCLET processor.
5018 @item -mlittle-endian
5019 Generate code for a processor running in little-endian mode.
5022 Treat register @code{%g0} as a normal register.
5023 GCC will continue to clobber it as necessary but will not assume
5024 it always reads as 0.
5026 @item -mbroken-saverestore
5027 Generate code that does not use non-trivial forms of the @code{save} and
5028 @code{restore} instructions. Early versions of the SPARCLET processor do
5029 not correctly handle @code{save} and @code{restore} instructions used with
5030 arguments. They correctly handle them used without arguments. A @code{save}
5031 instruction used without arguments increments the current window pointer
5032 but does not allocate a new stack frame. It is assumed that the window
5033 overflow trap handler will properly handle this case as will interrupt
5037 These @samp{-m} switches are supported in addition to the above
5038 on SPARC V9 processors in 64-bit environments.
5041 @item -mlittle-endian
5042 Generate code for a processor running in little-endian mode.
5046 Generate code for a 32-bit or 64-bit environment.
5047 The 32-bit environment sets int, long and pointer to 32 bits.
5048 The 64-bit environment sets int to 32 bits and long and pointer
5051 @item -mcmodel=medlow
5052 Generate code for the Medium/Low code model: the program must be linked
5053 in the low 32 bits of the address space. Pointers are 64 bits.
5054 Programs can be statically or dynamically linked.
5056 @item -mcmodel=medmid
5057 Generate code for the Medium/Middle code model: the program must be linked
5058 in the low 44 bits of the address space, the text segment must be less than
5059 2G bytes, and data segment must be within 2G of the text segment.
5060 Pointers are 64 bits.
5062 @item -mcmodel=medany
5063 Generate code for the Medium/Anywhere code model: the program may be linked
5064 anywhere in the address space, the text segment must be less than
5065 2G bytes, and data segment must be within 2G of the text segment.
5066 Pointers are 64 bits.
5068 @item -mcmodel=embmedany
5069 Generate code for the Medium/Anywhere code model for embedded systems:
5070 assume a 32-bit text and a 32-bit data segment, both starting anywhere
5071 (determined at link time). Register %g4 points to the base of the
5072 data segment. Pointers are still 64 bits.
5073 Programs are statically linked, PIC is not supported.
5076 @itemx -mno-stack-bias
5077 With @samp{-mstack-bias}, GCC assumes that the stack pointer, and
5078 frame pointer if present, are offset by -2047 which must be added back
5079 when making stack frame references.
5080 Otherwise, assume no such offset is present.
5083 @node Convex Options
5084 @subsection Convex Options
5085 @cindex Convex options
5087 These @samp{-m} options are defined for Convex:
5091 Generate output for C1. The code will run on any Convex machine.
5092 The preprocessor symbol @code{__convex__c1__} is defined.
5095 Generate output for C2. Uses instructions not available on C1.
5096 Scheduling and other optimizations are chosen for max performance on C2.
5097 The preprocessor symbol @code{__convex_c2__} is defined.
5100 Generate output for C32xx. Uses instructions not available on C1.
5101 Scheduling and other optimizations are chosen for max performance on C32.
5102 The preprocessor symbol @code{__convex_c32__} is defined.
5105 Generate output for C34xx. Uses instructions not available on C1.
5106 Scheduling and other optimizations are chosen for max performance on C34.
5107 The preprocessor symbol @code{__convex_c34__} is defined.
5110 Generate output for C38xx. Uses instructions not available on C1.
5111 Scheduling and other optimizations are chosen for max performance on C38.
5112 The preprocessor symbol @code{__convex_c38__} is defined.
5115 Generate code which puts an argument count in the word preceding each
5116 argument list. This is compatible with regular CC, and a few programs
5117 may need the argument count word. GDB and other source-level debuggers
5118 do not need it; this info is in the symbol table.
5121 Omit the argument count word. This is the default.
5123 @item -mvolatile-cache
5124 Allow volatile references to be cached. This is the default.
5126 @item -mvolatile-nocache
5127 Volatile references bypass the data cache, going all the way to memory.
5128 This is only needed for multi-processor code that does not use standard
5129 synchronization instructions. Making non-volatile references to volatile
5130 locations will not necessarily work.
5133 Type long is 32 bits, the same as type int. This is the default.
5136 Type long is 64 bits, the same as type long long. This option is useless,
5137 because no library support exists for it.
5140 @node AMD29K Options
5141 @subsection AMD29K Options
5142 @cindex AMD29K options
5144 These @samp{-m} options are defined for the AMD Am29000:
5149 @cindex DW bit (29k)
5150 Generate code that assumes the @code{DW} bit is set, i.e., that byte and
5151 halfword operations are directly supported by the hardware. This is the
5156 Generate code that assumes the @code{DW} bit is not set.
5160 @cindex byte writes (29k)
5161 Generate code that assumes the system supports byte and halfword write
5162 operations. This is the default.
5166 Generate code that assumes the systems does not support byte and
5167 halfword write operations. @samp{-mnbw} implies @samp{-mndw}.
5171 @cindex memory model (29k)
5172 Use a small memory model that assumes that all function addresses are
5173 either within a single 256 KB segment or at an absolute address of less
5174 than 256k. This allows the @code{call} instruction to be used instead
5175 of a @code{const}, @code{consth}, @code{calli} sequence.
5179 Use the normal memory model: Generate @code{call} instructions only when
5180 calling functions in the same file and @code{calli} instructions
5181 otherwise. This works if each file occupies less than 256 KB but allows
5182 the entire executable to be larger than 256 KB. This is the default.
5185 Always use @code{calli} instructions. Specify this option if you expect
5186 a single file to compile into more than 256 KB of code.
5190 @cindex processor selection (29k)
5191 Generate code for the Am29050.
5195 Generate code for the Am29000. This is the default.
5197 @item -mkernel-registers
5198 @kindex -mkernel-registers
5199 @cindex kernel and user registers (29k)
5200 Generate references to registers @code{gr64-gr95} instead of to
5201 registers @code{gr96-gr127}. This option can be used when compiling
5202 kernel code that wants a set of global registers disjoint from that used
5205 Note that when this option is used, register names in @samp{-f} flags
5206 must use the normal, user-mode, names.
5208 @item -muser-registers
5209 @kindex -muser-registers
5210 Use the normal set of global registers, @code{gr96-gr127}. This is the
5214 @itemx -mno-stack-check
5215 @kindex -mstack-check
5216 @cindex stack checks (29k)
5217 Insert (or do not insert) a call to @code{__msp_check} after each stack
5218 adjustment. This is often used for kernel code.
5221 @itemx -mno-storem-bug
5222 @kindex -mstorem-bug
5223 @cindex storem bug (29k)
5224 @samp{-mstorem-bug} handles 29k processors which cannot handle the
5225 separation of a mtsrim insn and a storem instruction (most 29000 chips
5226 to date, but not the 29050).
5228 @item -mno-reuse-arg-regs
5229 @itemx -mreuse-arg-regs
5230 @kindex -mreuse-arg-regs
5231 @samp{-mno-reuse-arg-regs} tells the compiler to only use incoming argument
5232 registers for copying out arguments. This helps detect calling a function
5233 with fewer arguments than it was declared with.
5235 @item -mno-impure-text
5236 @itemx -mimpure-text
5237 @kindex -mimpure-text
5238 @samp{-mimpure-text}, used in addition to @samp{-shared}, tells the compiler to
5239 not pass @samp{-assert pure-text} to the linker when linking a shared object.
5242 @kindex -msoft-float
5243 Generate output containing library calls for floating point.
5244 @strong{Warning:} the requisite libraries are not part of GCC.
5245 Normally the facilities of the machine's usual C compiler are used, but
5246 this can't be done directly in cross-compilation. You must make your
5247 own arrangements to provide suitable library functions for
5252 Do not generate multm or multmu instructions. This is useful for some embedded
5253 systems which do not have trap handlers for these instructions.
5257 @subsection ARM Options
5260 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
5265 @kindex -mapcs-frame
5266 Generate a stack frame that is compliant with the ARM Procedure Call
5267 Standard for all functions, even if this is not strictly necessary for
5268 correct execution of the code. Specifying @samp{-fomit-frame-pointer}
5269 with this option will cause the stack frames not to be generated for
5270 leaf functions. The default is @samp{-mno-apcs-frame}.
5274 This is a synonym for @samp{-mapcs-frame}.
5278 Generate code for a processor running with a 26-bit program counter,
5279 and conforming to the function calling standards for the APCS 26-bit
5280 option. This option replaces the @samp{-m2} and @samp{-m3} options
5281 of previous releases of the compiler.
5285 Generate code for a processor running with a 32-bit program counter,
5286 and conforming to the function calling standards for the APCS 32-bit
5287 option. This option replaces the @samp{-m6} option of previous releases
5291 @c not currently implemented
5292 @item -mapcs-stack-check
5293 @kindex -mapcs-stack-check
5294 @kindex -mno-apcs-stack-check
5295 Generate code to check the amount of stack space available upon entry to
5296 every function (that actually uses some stack space). If there is
5297 insufficient space available then either the function
5298 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
5299 called, depending upon the amount of stack space required. The run time
5300 system is required to provide these functions. The default is
5301 @samp{-mno-apcs-stack-check}, since this produces smaller code.
5303 @c not currently implemented
5305 @kindex -mapcs-float
5306 @kindex -mno-apcs-float
5307 Pass floating point arguments using the float point registers. This is
5308 one of the variants of the APCS. This option is recommended if the
5309 target hardware has a floating point unit or if a lot of floating point
5310 arithmetic is going to be performed by the code. The default is
5311 @samp{-mno-apcs-float}, since integer only code is slightly increased in
5312 size if @samp{-mapcs-float} is used.
5314 @c not currently implemented
5315 @item -mapcs-reentrant
5316 @kindex -mapcs-reentrant
5317 @kindex -mno-apcs-reentrant
5318 Generate reentrant, position independent code. The default is
5319 @samp{-mno-apcs-reentrant}.
5322 @item -mthumb-interwork
5323 @kindex -mthumb-interwork
5324 @kindex -mno-thumb-interwork
5325 Generate code which supports calling between the ARM and Thumb
5326 instruction sets. Without this option the two instruction sets cannot
5327 be reliably used inside one program. The default is
5328 @samp{-mno-thumb-interwork}, since slightly larger code is generated
5329 when @samp{-mthumb-interwork} is specified.
5331 @item -mno-sched-prolog
5332 @kindex -mno-sched-prolog
5333 @kindex -msched-prolog
5334 Prevent the reordering of instructions in the function prolog, or the
5335 merging of those instruction with the instructions in the function's
5336 body. This means that all functions will start with a recognizable set
5337 of instructions (or in fact one of a choice from a small set of
5338 different function prologues), and this information can be used to
5339 locate the start if functions inside an executable piece of code. The
5340 default is @samp{-msched-prolog}.
5343 Generate output containing floating point instructions. This is the
5347 Generate output containing library calls for floating point.
5348 @strong{Warning:} the requisite libraries are not available for all ARM
5349 targets. Normally the facilities of the machine's usual C compiler are
5350 used, but this cannot be done directly in cross-compilation. You must make
5351 your own arrangements to provide suitable library functions for
5354 @samp{-msoft-float} changes the calling convention in the output file;
5355 therefore, it is only useful if you compile @emph{all} of a program with
5356 this option. In particular, you need to compile @file{libgcc.a}, the
5357 library that comes with GCC, with @samp{-msoft-float} in order for
5360 @item -mlittle-endian
5361 Generate code for a processor running in little-endian mode. This is
5362 the default for all standard configurations.
5365 Generate code for a processor running in big-endian mode; the default is
5366 to compile code for a little-endian processor.
5368 @item -mwords-little-endian
5369 This option only applies when generating code for big-endian processors.
5370 Generate code for a little-endian word order but a big-endian byte
5371 order. That is, a byte order of the form @samp{32107654}. Note: this
5372 option should only be used if you require compatibility with code for
5373 big-endian ARM processors generated by versions of the compiler prior to
5376 @item -malignment-traps
5377 @kindex -malignment-traps
5378 Generate code that will not trap if the MMU has alignment traps enabled.
5379 On ARM architectures prior to ARMv4, there were no instructions to
5380 access half-word objects stored in memory. However, when reading from
5381 memory a feature of the ARM architecture allows a word load to be used,
5382 even if the address is unaligned, and the processor core will rotate the
5383 data as it is being loaded. This option tells the compiler that such
5384 misaligned accesses will cause a MMU trap and that it should instead
5385 synthesise the access as a series of byte accesses. The compiler can
5386 still use word accesses to load half-word data if it knows that the
5387 address is aligned to a word boundary.
5389 This option is ignored when compiling for ARM architecture 4 or later,
5390 since these processors have instructions to directly access half-word
5393 @item -mno-alignment-traps
5394 @kindex -mno-alignment-traps
5395 Generate code that assumes that the MMU will not trap unaligned
5396 accesses. This produces better code when the target instruction set
5397 does not have half-word memory operations (i.e. implementations prior to
5400 Note that you cannot use this option to access unaligned word objects,
5401 since the processor will only fetch one 32-bit aligned object from
5404 The default setting for most targets is -mno-alignment-traps, since
5405 this produces better code when there are no half-word memory
5406 instructions available.
5408 @item -mshort-load-bytes
5409 @itemx -mno-short-load-words
5410 @kindex -mshort-load-bytes
5411 @kindex -mno-short-load-words
5412 These are deprecated aliases for @samp{-malignment-traps}.
5414 @item -mno-short-load-bytes
5415 @itemx -mshort-load-words
5416 @kindex -mno-short-load-bytes
5417 @kindex -mshort-load-words
5418 This are deprecated aliases for @samp{-mno-alignment-traps}.
5422 This option only applies to RISC iX. Emulate the native BSD-mode
5423 compiler. This is the default if @samp{-ansi} is not specified.
5427 This option only applies to RISC iX. Emulate the native X/Open-mode
5430 @item -mno-symrename
5431 @kindex -mno-symrename
5432 This option only applies to RISC iX. Do not run the assembler
5433 post-processor, @samp{symrename}, after code has been assembled.
5434 Normally it is necessary to modify some of the standard symbols in
5435 preparation for linking with the RISC iX C library; this option
5436 suppresses this pass. The post-processor is never run when the
5437 compiler is built for cross-compilation.
5441 This specifies the name of the target ARM processor. GCC uses this name
5442 to determine what kind of instructions it can emit when generating
5443 assembly code. Permissible names are: arm2, arm250, arm3, arm6, arm60,
5444 arm600, arm610, arm620, arm7, arm7m, arm7d, arm7dm, arm7di, arm7dmi,
5445 arm70, arm700, arm700i, arm710, arm710c, arm7100, arm7500, arm7500fe,
5446 arm7tdmi, arm8, strongarm, strongarm110, strongarm1100, arm8, arm810,
5447 arm9, arm9e, arm920, arm920t, arm940t, arm9tdmi, arm10tdmi, arm1020t,
5450 @itemx -mtune=<name>
5452 This option is very similar to the @samp{-mcpu=} option, except that
5453 instead of specifying the actual target processor type, and hence
5454 restricting which instructions can be used, it specifies that GCC should
5455 tune the performance of the code as if the target were of the type
5456 specified in this option, but still choosing the instructions that it
5457 will generate based on the cpu specified by a @samp{-mcpu=} option.
5458 For some ARM implementations better performance can be obtained by using
5463 This specifies the name of the target ARM architecture. GCC uses this
5464 name to determine what kind of instructions it can emit when generating
5465 assembly code. This option can be used in conjunction with or instead
5466 of the @samp{-mcpu=} option. Permissible names are: armv2, armv2a,
5467 armv3, armv3m, armv4, armv4t, armv5, armv5t, armv5te.
5469 @item -mfpe=<number>
5470 @itemx -mfp=<number>
5473 This specifies the version of the floating point emulation available on
5474 the target. Permissible values are 2 and 3. @samp{-mfp=} is a synonym
5475 for @samp{-mfpe=}, for compatibility with older versions of GCC.
5477 @item -mstructure-size-boundary=<n>
5478 @kindex -mstructure-size-boundary
5479 The size of all structures and unions will be rounded up to a multiple
5480 of the number of bits set by this option. Permissible values are 8 and
5481 32. The default value varies for different toolchains. For the COFF
5482 targeted toolchain the default value is 8. Specifying the larger number
5483 can produce faster, more efficient code, but can also increase the size
5484 of the program. The two values are potentially incompatible. Code
5485 compiled with one value cannot necessarily expect to work with code or
5486 libraries compiled with the other value, if they exchange information
5487 using structures or unions.
5489 @item -mabort-on-noreturn
5490 @kindex -mabort-on-noreturn
5491 @kindex -mnoabort-on-noreturn
5492 Generate a call to the function @code{abort} at the end of a
5493 @code{noreturn} function. It will be executed if the function tries to
5497 @itemx -mno-long-calls
5498 Tells the compiler to perform function calls by first loading the
5499 address of the function into a register and then performing a subroutine
5500 call on this register. This switch is needed if the target function
5501 will lie outside of the 64 megabyte addressing range of the offset based
5502 version of subroutine call instruction.
5504 Even if this switch is enabled, not all function calls will be turned
5505 into long calls. The heuristic is that static functions, functions
5506 which have the @samp{short-call} attribute, functions that are inside
5507 the scope of a @samp{#pragma no_long_calls} directive and functions whose
5508 definitions have already been compiled within the current compilation
5509 unit, will not be turned into long calls. The exception to this rule is
5510 that weak function definitions, functions with the @samp{long-call}
5511 attribute or the @samp{section} attribute, and functions that are within
5512 the scope of a @samp{#pragma long_calls} directive, will always be
5513 turned into long calls.
5515 This feature is not enabled by default. Specifying
5516 @samp{-mno-long-calls} will restore the default behaviour, as will
5517 placing the function calls within the scope of a @samp{#pragma
5518 long_calls_off} directive. Note these switches have no effect on how
5519 the compiler generates code to handle function calls via function
5522 @item -mnop-fun-dllimport
5523 @kindex -mnop-fun-dllimport
5524 Disable support for the @emph{dllimport} attribute.
5526 @item -msingle-pic-base
5527 @kindex -msingle-pic-base
5528 Treat the register used for PIC addressing as read-only, rather than
5529 loading it in the prologue for each function. The run-time system is
5530 responsible for initialising this register with an appropriate value
5531 before execution begins.
5533 @item -mpic-register=<reg>
5534 @kindex -mpic-register=
5535 Specify the register to be used for PIC addressing. The default is R10
5536 unless stack-checking is enabled, when R9 is used.
5538 @item -mpoke-function-name
5539 @kindex -mpoke-function-name
5540 Write the name of each function into the text section, directly
5541 preceding the function prologue. The generated code is similar to this:
5545 .ascii "arm_poke_function_name", 0
5548 .word 0xff000000 + (t1 - t0)
5549 arm_poke_function_name
5551 stmfd sp!, @{fp, ip, lr, pc@}
5555 When performing a stack backtrace, code can inspect the value of
5556 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
5557 location @code{pc - 12} and the top 8 bits are set, then we know that
5558 there is a function name embedded immediately preceding this location
5559 and has length @code{((pc[-3]) & 0xff000000)}.
5563 Generate code for the 16-bit Thumb instruction set. The default is to
5564 use the 32-bit ARM instruction set.
5567 @kindex -mtpcs-frame
5568 @kindex -mno-tpcs-frame
5569 Generate a stack frame that is compliant with the Thumb Procedure Call
5570 Standard for all non-leaf functions. (A leaf function is one that does
5571 not call any other functions.) The default is @samp{-mno-tpcs-frame}.
5573 @item -mtpcs-leaf-frame
5574 @kindex -mtpcs-leaf-frame
5575 @kindex -mno-tpcs-leaf-frame
5576 Generate a stack frame that is compliant with the Thumb Procedure Call
5577 Standard for all leaf functions. (A leaf function is one that does
5578 not call any other functions.) The default is @samp{-mno-apcs-leaf-frame}.
5580 @item -mcallee-super-interworking
5581 @kindex -mcallee-super-interworking
5582 Gives all externally visible functions in the file being compiled an ARM
5583 instruction set header which switches to Thumb mode before executing the
5584 rest of the function. This allows these functions to be called from
5585 non-interworking code.
5587 @item -mcaller-super-interworking
5588 @kindex -mcaller-super-interworking
5589 Allows calls via function pointers (including virtual functions) to
5590 execute correctly regardless of whether the target code has been
5591 compiled for interworking or not. There is a small overhead in the cost
5592 of executing a function pointer if this option is enabled.
5596 @node MN10200 Options
5597 @subsection MN10200 Options
5598 @cindex MN10200 options
5599 These @samp{-m} options are defined for Matsushita MN10200 architectures:
5603 Indicate to the linker that it should perform a relaxation optimization pass
5604 to shorten branches, calls and absolute memory addresses. This option only
5605 has an effect when used on the command line for the final link step.
5607 This option makes symbolic debugging impossible.
5610 @node MN10300 Options
5611 @subsection MN10300 Options
5612 @cindex MN10300 options
5613 These @samp{-m} options are defined for Matsushita MN10300 architectures:
5617 Generate code to avoid bugs in the multiply instructions for the MN10300
5618 processors. This is the default.
5621 Do not generate code to avoid bugs in the multiply instructions for the
5625 Generate code which uses features specific to the AM33 processor.
5628 Do not generate code which uses features specific to the AM33 processor. This
5632 Indicate to the linker that it should perform a relaxation optimization pass
5633 to shorten branches, calls and absolute memory addresses. This option only
5634 has an effect when used on the command line for the final link step.
5636 This option makes symbolic debugging impossible.
5640 @node M32R/D Options
5641 @subsection M32R/D Options
5642 @cindex M32R/D options
5644 These @samp{-m} options are defined for Mitsubishi M32R/D architectures:
5647 @item -mcode-model=small
5648 Assume all objects live in the lower 16MB of memory (so that their addresses
5649 can be loaded with the @code{ld24} instruction), and assume all subroutines
5650 are reachable with the @code{bl} instruction.
5651 This is the default.
5653 The addressability of a particular object can be set with the
5654 @code{model} attribute.
5656 @item -mcode-model=medium
5657 Assume objects may be anywhere in the 32-bit address space (the compiler
5658 will generate @code{seth/add3} instructions to load their addresses), and
5659 assume all subroutines are reachable with the @code{bl} instruction.
5661 @item -mcode-model=large
5662 Assume objects may be anywhere in the 32-bit address space (the compiler
5663 will generate @code{seth/add3} instructions to load their addresses), and
5664 assume subroutines may not be reachable with the @code{bl} instruction
5665 (the compiler will generate the much slower @code{seth/add3/jl}
5666 instruction sequence).
5669 Disable use of the small data area. Variables will be put into
5670 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
5671 @code{section} attribute has been specified).
5672 This is the default.
5674 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
5675 Objects may be explicitly put in the small data area with the
5676 @code{section} attribute using one of these sections.
5679 Put small global and static data in the small data area, but do not
5680 generate special code to reference them.
5683 Put small global and static data in the small data area, and generate
5684 special instructions to reference them.
5687 @cindex smaller data references
5688 Put global and static objects less than or equal to @var{num} bytes
5689 into the small data or bss sections instead of the normal data or bss
5690 sections. The default value of @var{num} is 8.
5691 The @samp{-msdata} option must be set to one of @samp{sdata} or @samp{use}
5692 for this option to have any effect.
5694 All modules should be compiled with the same @samp{-G @var{num}} value.
5695 Compiling with different values of @var{num} may or may not work; if it
5696 doesn't the linker will give an error message - incorrect code will not be
5702 @subsection M88K Options
5703 @cindex M88k options
5705 These @samp{-m} options are defined for Motorola 88k architectures:
5710 Generate code that works well on both the m88100 and the
5715 Generate code that works best for the m88100, but that also
5720 Generate code that works best for the m88110, and may not run
5725 Obsolete option to be removed from the next revision.
5728 @item -midentify-revision
5729 @kindex -midentify-revision
5731 @cindex identifying source, compiler (88k)
5732 Include an @code{ident} directive in the assembler output recording the
5733 source file name, compiler name and version, timestamp, and compilation
5736 @item -mno-underscores
5737 @kindex -mno-underscores
5738 @cindex underscores, avoiding (88k)
5739 In assembler output, emit symbol names without adding an underscore
5740 character at the beginning of each name. The default is to use an
5741 underscore as prefix on each name.
5743 @item -mocs-debug-info
5744 @itemx -mno-ocs-debug-info
5745 @kindex -mocs-debug-info
5746 @kindex -mno-ocs-debug-info
5748 @cindex debugging, 88k OCS
5749 Include (or omit) additional debugging information (about registers used
5750 in each stack frame) as specified in the 88open Object Compatibility
5751 Standard, ``OCS''. This extra information allows debugging of code that
5752 has had the frame pointer eliminated. The default for DG/UX, SVr4, and
5753 Delta 88 SVr3.2 is to include this information; other 88k configurations
5754 omit this information by default.
5756 @item -mocs-frame-position
5757 @kindex -mocs-frame-position
5758 @cindex register positions in frame (88k)
5759 When emitting COFF debugging information for automatic variables and
5760 parameters stored on the stack, use the offset from the canonical frame
5761 address, which is the stack pointer (register 31) on entry to the
5762 function. The DG/UX, SVr4, Delta88 SVr3.2, and BCS configurations use
5763 @samp{-mocs-frame-position}; other 88k configurations have the default
5764 @samp{-mno-ocs-frame-position}.
5766 @item -mno-ocs-frame-position
5767 @kindex -mno-ocs-frame-position
5768 @cindex register positions in frame (88k)
5769 When emitting COFF debugging information for automatic variables and
5770 parameters stored on the stack, use the offset from the frame pointer
5771 register (register 30). When this option is in effect, the frame
5772 pointer is not eliminated when debugging information is selected by the
5775 @item -moptimize-arg-area
5776 @itemx -mno-optimize-arg-area
5777 @kindex -moptimize-arg-area
5778 @kindex -mno-optimize-arg-area
5779 @cindex arguments in frame (88k)
5780 Control how function arguments are stored in stack frames.
5781 @samp{-moptimize-arg-area} saves space by optimizing them, but this
5782 conflicts with the 88open specifications. The opposite alternative,
5783 @samp{-mno-optimize-arg-area}, agrees with 88open standards. By default
5784 GCC does not optimize the argument area.
5786 @item -mshort-data-@var{num}
5787 @kindex -mshort-data-@var{num}
5788 @cindex smaller data references (88k)
5789 @cindex r0-relative references (88k)
5790 Generate smaller data references by making them relative to @code{r0},
5791 which allows loading a value using a single instruction (rather than the
5792 usual two). You control which data references are affected by
5793 specifying @var{num} with this option. For example, if you specify
5794 @samp{-mshort-data-512}, then the data references affected are those
5795 involving displacements of less than 512 bytes.
5796 @samp{-mshort-data-@var{num}} is not effective for @var{num} greater
5799 @item -mserialize-volatile
5800 @kindex -mserialize-volatile
5801 @itemx -mno-serialize-volatile
5802 @kindex -mno-serialize-volatile
5803 @cindex sequential consistency on 88k
5804 Do, or don't, generate code to guarantee sequential consistency
5805 of volatile memory references. By default, consistency is
5808 The order of memory references made by the MC88110 processor does
5809 not always match the order of the instructions requesting those
5810 references. In particular, a load instruction may execute before
5811 a preceding store instruction. Such reordering violates
5812 sequential consistency of volatile memory references, when there
5813 are multiple processors. When consistency must be guaranteed,
5814 GNU C generates special instructions, as needed, to force
5815 execution in the proper order.
5817 The MC88100 processor does not reorder memory references and so
5818 always provides sequential consistency. However, by default, GNU
5819 C generates the special instructions to guarantee consistency
5820 even when you use @samp{-m88100}, so that the code may be run on an
5821 MC88110 processor. If you intend to run your code only on the
5822 MC88100 processor, you may use @samp{-mno-serialize-volatile}.
5824 The extra code generated to guarantee consistency may affect the
5825 performance of your application. If you know that you can safely
5826 forgo this guarantee, you may use @samp{-mno-serialize-volatile}.
5832 @cindex assembler syntax, 88k
5834 Turn on (@samp{-msvr4}) or off (@samp{-msvr3}) compiler extensions
5835 related to System V release 4 (SVr4). This controls the following:
5839 Which variant of the assembler syntax to emit.
5841 @samp{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
5842 that is used on System V release 4.
5844 @samp{-msvr4} makes GCC issue additional declaration directives used in
5848 @samp{-msvr4} is the default for the m88k-motorola-sysv4 and
5849 m88k-dg-dgux m88k configurations. @samp{-msvr3} is the default for all
5850 other m88k configurations.
5852 @item -mversion-03.00
5853 @kindex -mversion-03.00
5854 This option is obsolete, and is ignored.
5855 @c ??? which asm syntax better for GAS? option there too?
5857 @item -mno-check-zero-division
5858 @itemx -mcheck-zero-division
5859 @kindex -mno-check-zero-division
5860 @kindex -mcheck-zero-division
5861 @cindex zero division on 88k
5862 Do, or don't, generate code to guarantee that integer division by
5863 zero will be detected. By default, detection is guaranteed.
5865 Some models of the MC88100 processor fail to trap upon integer
5866 division by zero under certain conditions. By default, when
5867 compiling code that might be run on such a processor, GNU C
5868 generates code that explicitly checks for zero-valued divisors
5869 and traps with exception number 503 when one is detected. Use of
5870 mno-check-zero-division suppresses such checking for code
5871 generated to run on an MC88100 processor.
5873 GNU C assumes that the MC88110 processor correctly detects all
5874 instances of integer division by zero. When @samp{-m88110} is
5875 specified, both @samp{-mcheck-zero-division} and
5876 @samp{-mno-check-zero-division} are ignored, and no explicit checks for
5877 zero-valued divisors are generated.
5879 @item -muse-div-instruction
5880 @kindex -muse-div-instruction
5881 @cindex divide instruction, 88k
5882 Use the div instruction for signed integer division on the
5883 MC88100 processor. By default, the div instruction is not used.
5885 On the MC88100 processor the signed integer division instruction
5886 div) traps to the operating system on a negative operand. The
5887 operating system transparently completes the operation, but at a
5888 large cost in execution time. By default, when compiling code
5889 that might be run on an MC88100 processor, GNU C emulates signed
5890 integer division using the unsigned integer division instruction
5891 divu), thereby avoiding the large penalty of a trap to the
5892 operating system. Such emulation has its own, smaller, execution
5893 cost in both time and space. To the extent that your code's
5894 important signed integer division operations are performed on two
5895 nonnegative operands, it may be desirable to use the div
5896 instruction directly.
5898 On the MC88110 processor the div instruction (also known as the
5899 divs instruction) processes negative operands without trapping to
5900 the operating system. When @samp{-m88110} is specified,
5901 @samp{-muse-div-instruction} is ignored, and the div instruction is used
5902 for signed integer division.
5904 Note that the result of dividing INT_MIN by -1 is undefined. In
5905 particular, the behavior of such a division with and without
5906 @samp{-muse-div-instruction} may differ.
5908 @item -mtrap-large-shift
5909 @itemx -mhandle-large-shift
5910 @kindex -mtrap-large-shift
5911 @kindex -mhandle-large-shift
5912 @cindex bit shift overflow (88k)
5913 @cindex large bit shifts (88k)
5914 Include code to detect bit-shifts of more than 31 bits; respectively,
5915 trap such shifts or emit code to handle them properly. By default GCC
5916 makes no special provision for large bit shifts.
5918 @item -mwarn-passed-structs
5919 @kindex -mwarn-passed-structs
5920 @cindex structure passing (88k)
5921 Warn when a function passes a struct as an argument or result.
5922 Structure-passing conventions have changed during the evolution of the C
5923 language, and are often the source of portability problems. By default,
5924 GCC issues no such warning.
5927 @node RS/6000 and PowerPC Options
5928 @subsection IBM RS/6000 and PowerPC Options
5929 @cindex RS/6000 and PowerPC Options
5930 @cindex IBM RS/6000 and PowerPC Options
5932 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
5940 @itemx -mpowerpc-gpopt
5941 @itemx -mno-powerpc-gpopt
5942 @itemx -mpowerpc-gfxopt
5943 @itemx -mno-powerpc-gfxopt
5945 @itemx -mno-powerpc64
5949 @kindex -mpowerpc-gpopt
5950 @kindex -mpowerpc-gfxopt
5952 GCC supports two related instruction set architectures for the
5953 RS/6000 and PowerPC. The @dfn{POWER} instruction set are those
5954 instructions supported by the @samp{rios} chip set used in the original
5955 RS/6000 systems and the @dfn{PowerPC} instruction set is the
5956 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
5957 the IBM 4xx microprocessors.
5959 Neither architecture is a subset of the other. However there is a
5960 large common subset of instructions supported by both. An MQ
5961 register is included in processors supporting the POWER architecture.
5963 You use these options to specify which instructions are available on the
5964 processor you are using. The default value of these options is
5965 determined when configuring GCC. Specifying the
5966 @samp{-mcpu=@var{cpu_type}} overrides the specification of these
5967 options. We recommend you use the @samp{-mcpu=@var{cpu_type}} option
5968 rather than the options listed above.
5970 The @samp{-mpower} option allows GCC to generate instructions that
5971 are found only in the POWER architecture and to use the MQ register.
5972 Specifying @samp{-mpower2} implies @samp{-power} and also allows GCC
5973 to generate instructions that are present in the POWER2 architecture but
5974 not the original POWER architecture.
5976 The @samp{-mpowerpc} option allows GCC to generate instructions that
5977 are found only in the 32-bit subset of the PowerPC architecture.
5978 Specifying @samp{-mpowerpc-gpopt} implies @samp{-mpowerpc} and also allows
5979 GCC to use the optional PowerPC architecture instructions in the
5980 General Purpose group, including floating-point square root. Specifying
5981 @samp{-mpowerpc-gfxopt} implies @samp{-mpowerpc} and also allows GCC to
5982 use the optional PowerPC architecture instructions in the Graphics
5983 group, including floating-point select.
5985 The @samp{-mpowerpc64} option allows GCC to generate the additional
5986 64-bit instructions that are found in the full PowerPC64 architecture
5987 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
5988 @samp{-mno-powerpc64}.
5990 If you specify both @samp{-mno-power} and @samp{-mno-powerpc}, GCC
5991 will use only the instructions in the common subset of both
5992 architectures plus some special AIX common-mode calls, and will not use
5993 the MQ register. Specifying both @samp{-mpower} and @samp{-mpowerpc}
5994 permits GCC to use any instruction from either architecture and to
5995 allow use of the MQ register; specify this for the Motorola MPC601.
5997 @item -mnew-mnemonics
5998 @itemx -mold-mnemonics
5999 @kindex -mnew-mnemonics
6000 @kindex -mold-mnemonics
6001 Select which mnemonics to use in the generated assembler code.
6002 @samp{-mnew-mnemonics} requests output that uses the assembler mnemonics
6003 defined for the PowerPC architecture, while @samp{-mold-mnemonics}
6004 requests the assembler mnemonics defined for the POWER architecture.
6005 Instructions defined in only one architecture have only one mnemonic;
6006 GCC uses that mnemonic irrespective of which of these options is
6009 GCC defaults to the mnemonics appropriate for the architecture in
6010 use. Specifying @samp{-mcpu=@var{cpu_type}} sometimes overrides the
6011 value of these option. Unless you are building a cross-compiler, you
6012 should normally not specify either @samp{-mnew-mnemonics} or
6013 @samp{-mold-mnemonics}, but should instead accept the default.
6015 @item -mcpu=@var{cpu_type}
6017 Set architecture type, register usage, choice of mnemonics, and
6018 instruction scheduling parameters for machine type @var{cpu_type}.
6019 Supported values for @var{cpu_type} are @samp{rios}, @samp{rios1},
6020 @samp{rsc}, @samp{rios2}, @samp{rs64a}, @samp{601}, @samp{602},
6021 @samp{603}, @samp{603e}, @samp{604}, @samp{604e}, @samp{620},
6022 @samp{630}, @samp{740}, @samp{750}, @samp{power}, @samp{power2},
6023 @samp{powerpc}, @samp{403}, @samp{505}, @samp{801}, @samp{821},
6024 @samp{823}, and @samp{860} and @samp{common}. @samp{-mcpu=power},
6025 @samp{-mcpu=power2}, @samp{-mcpu=powerpc}, and @samp{-mcpu=powerpc64}
6026 specify generic POWER, POWER2, pure 32-bit PowerPC (i.e., not MPC601),
6027 and 64-bit PowerPC architecture machine types, with an appropriate,
6028 generic processor model assumed for scheduling purposes.@refill
6030 Specifying any of the following options:
6031 @samp{-mcpu=rios1}, @samp{-mcpu=rios2}, @samp{-mcpu=rsc},
6032 @samp{-mcpu=power}, or @samp{-mcpu=power2}
6033 enables the @samp{-mpower} option and disables the @samp{-mpowerpc} option;
6034 @samp{-mcpu=601} enables both the @samp{-mpower} and @samp{-mpowerpc} options.
6035 All of @samp{-mcpu=rs64a}, @samp{-mcpu=602}, @samp{-mcpu=603},
6036 @samp{-mcpu=603e}, @samp{-mcpu=604}, @samp{-mcpu=620}, @samp{-mcpu=630},
6037 @samp{-mcpu=740}, and @samp{-mcpu=750}
6038 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
6039 Exactly similarly, all of @samp{-mcpu=403},
6040 @samp{-mcpu=505}, @samp{-mcpu=821}, @samp{-mcpu=860} and @samp{-mcpu=powerpc}
6041 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
6042 @samp{-mcpu=common} disables both the
6043 @samp{-mpower} and @samp{-mpowerpc} options.@refill
6045 AIX versions 4 or greater selects @samp{-mcpu=common} by default, so
6046 that code will operate on all members of the RS/6000 POWER and PowerPC
6047 families. In that case, GCC will use only the instructions in the
6048 common subset of both architectures plus some special AIX common-mode
6049 calls, and will not use the MQ register. GCC assumes a generic
6050 processor model for scheduling purposes.
6052 Specifying any of the options @samp{-mcpu=rios1}, @samp{-mcpu=rios2},
6053 @samp{-mcpu=rsc}, @samp{-mcpu=power}, or @samp{-mcpu=power2} also
6054 disables the @samp{new-mnemonics} option. Specifying @samp{-mcpu=601},
6055 @samp{-mcpu=602}, @samp{-mcpu=603}, @samp{-mcpu=603e}, @samp{-mcpu=604},
6056 @samp{-mcpu=620}, @samp{-mcpu=630}, @samp{-mcpu=403}, @samp{-mcpu=505},
6057 @samp{-mcpu=821}, @samp{-mcpu=860} or @samp{-mcpu=powerpc} also enables
6058 the @samp{new-mnemonics} option.@refill
6060 Specifying @samp{-mcpu=403}, @samp{-mcpu=821}, or @samp{-mcpu=860} also
6061 enables the @samp{-msoft-float} option.
6063 @item -mtune=@var{cpu_type}
6064 Set the instruction scheduling parameters for machine type
6065 @var{cpu_type}, but do not set the architecture type, register usage,
6066 choice of mnemonics like @samp{-mcpu=}@var{cpu_type} would. The same
6067 values for @var{cpu_type} are used for @samp{-mtune=}@var{cpu_type} as
6068 for @samp{-mcpu=}@var{cpu_type}. The @samp{-mtune=}@var{cpu_type}
6069 option overrides the @samp{-mcpu=}@var{cpu_type} option in terms of
6070 instruction scheduling parameters.
6073 @itemx -mno-fp-in-toc
6074 @itemx -mno-sum-in-toc
6075 @itemx -mminimal-toc
6076 @kindex -mminimal-toc
6077 Modify generation of the TOC (Table Of Contents), which is created for
6078 every executable file. The @samp{-mfull-toc} option is selected by
6079 default. In that case, GCC will allocate at least one TOC entry for
6080 each unique non-automatic variable reference in your program. GCC
6081 will also place floating-point constants in the TOC. However, only
6082 16,384 entries are available in the TOC.
6084 If you receive a linker error message that saying you have overflowed
6085 the available TOC space, you can reduce the amount of TOC space used
6086 with the @samp{-mno-fp-in-toc} and @samp{-mno-sum-in-toc} options.
6087 @samp{-mno-fp-in-toc} prevents GCC from putting floating-point
6088 constants in the TOC and @samp{-mno-sum-in-toc} forces GCC to
6089 generate code to calculate the sum of an address and a constant at
6090 run-time instead of putting that sum into the TOC. You may specify one
6091 or both of these options. Each causes GCC to produce very slightly
6092 slower and larger code at the expense of conserving TOC space.
6094 If you still run out of space in the TOC even when you specify both of
6095 these options, specify @samp{-mminimal-toc} instead. This option causes
6096 GCC to make only one TOC entry for every file. When you specify this
6097 option, GCC will produce code that is slower and larger but which
6098 uses extremely little TOC space. You may wish to use this option
6099 only on files that contain less frequently executed code. @refill
6105 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
6106 @code{long} type, and the infrastructure needed to support them.
6107 Specifying @samp{-maix64} implies @samp{-mpowerpc64} and
6108 @samp{-mpowerpc}, while @samp{-maix32} disables the 64-bit ABI and
6109 implies @samp{-mno-powerpc64}. GCC defaults to @samp{-maix32}.
6114 On AIX, pass floating-point arguments to prototyped functions beyond the
6115 register save area (RSA) on the stack in addition to argument FPRs. The
6116 AIX calling convention was extended but not initially documented to
6117 handle an obscure K&R C case of calling a function that takes the
6118 address of its arguments with fewer arguments than declared. AIX XL
6119 compilers access floating point arguments which do not fit in the
6120 RSA from the stack when a subroutine is compiled without
6121 optimization. Because always storing floating-point arguments on the
6122 stack is inefficient and rarely needed, this option is not enabled by
6123 default and only is necessary when calling subroutines compiled by AIX
6124 XL compilers without optimization.
6128 Support @dfn{AIX Threads}. Link an application written to use
6129 @dfn{pthreads} with special libraries and startup code to enable the
6134 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE). Link an
6135 application written to use message passing with special startup code to
6136 enable the application to run. The system must have PE installed in the
6137 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
6138 must be overridden with the @samp{-specs=} option to specify the
6139 appropriate directory location. The Parallel Environment does not
6140 support threads, so the @samp{-mpe} option and the @samp{-mthreads}
6141 option are incompatible.
6145 @kindex -msoft-float
6146 Generate code that does not use (uses) the floating-point register set.
6147 Software floating point emulation is provided if you use the
6148 @samp{-msoft-float} option, and pass the option to GCC when linking.
6151 @itemx -mno-multiple
6152 Generate code that uses (does not use) the load multiple word
6153 instructions and the store multiple word instructions. These
6154 instructions are generated by default on POWER systems, and not
6155 generated on PowerPC systems. Do not use @samp{-mmultiple} on little
6156 endian PowerPC systems, since those instructions do not work when the
6157 processor is in little endian mode. The exceptions are PPC740 and
6158 PPC750 which permit the instructions usage in little endian mode.
6163 Generate code that uses (does not use) the load string instructions
6164 and the store string word instructions to save multiple registers and
6165 do small block moves. These instructions are generated by default on
6166 POWER systems, and not generated on PowerPC systems. Do not use
6167 @samp{-mstring} on little endian PowerPC systems, since those
6168 instructions do not work when the processor is in little endian mode.
6169 The exceptions are PPC740 and PPC750 which permit the instructions
6170 usage in little endian mode.
6175 Generate code that uses (does not use) the load or store instructions
6176 that update the base register to the address of the calculated memory
6177 location. These instructions are generated by default. If you use
6178 @samp{-mno-update}, there is a small window between the time that the
6179 stack pointer is updated and the address of the previous frame is
6180 stored, which means code that walks the stack frame across interrupts or
6181 signals may get corrupted data.
6184 @itemx -mno-fused-madd
6185 @kindex -mfused-madd
6186 Generate code that uses (does not use) the floating point multiply and
6187 accumulate instructions. These instructions are generated by default if
6188 hardware floating is used.
6190 @item -mno-bit-align
6193 On System V.4 and embedded PowerPC systems do not (do) force structures
6194 and unions that contain bit fields to be aligned to the base type of the
6197 For example, by default a structure containing nothing but 8
6198 @code{unsigned} bitfields of length 1 would be aligned to a 4 byte
6199 boundary and have a size of 4 bytes. By using @samp{-mno-bit-align},
6200 the structure would be aligned to a 1 byte boundary and be one byte in
6203 @item -mno-strict-align
6204 @itemx -mstrict-align
6205 @kindex -mstrict-align
6206 On System V.4 and embedded PowerPC systems do not (do) assume that
6207 unaligned memory references will be handled by the system.
6210 @itemx -mno-relocatable
6211 @kindex -mrelocatable
6212 On embedded PowerPC systems generate code that allows (does not allow)
6213 the program to be relocated to a different address at runtime. If you
6214 use @samp{-mrelocatable} on any module, all objects linked together must
6215 be compiled with @samp{-mrelocatable} or @samp{-mrelocatable-lib}.
6217 @item -mrelocatable-lib
6218 @itemx -mno-relocatable-lib
6219 On embedded PowerPC systems generate code that allows (does not allow)
6220 the program to be relocated to a different address at runtime. Modules
6221 compiled with @samp{-mrelocatable-lib} can be linked with either modules
6222 compiled without @samp{-mrelocatable} and @samp{-mrelocatable-lib} or
6223 with modules compiled with the @samp{-mrelocatable} options.
6227 On System V.4 and embedded PowerPC systems do not (do) assume that
6228 register 2 contains a pointer to a global area pointing to the addresses
6229 used in the program.
6232 @itemx -mlittle-endian
6233 On System V.4 and embedded PowerPC systems compile code for the
6234 processor in little endian mode. The @samp{-mlittle-endian} option is
6235 the same as @samp{-mlittle}.
6239 On System V.4 and embedded PowerPC systems compile code for the
6240 processor in big endian mode. The @samp{-mbig-endian} option is
6241 the same as @samp{-mbig}.
6244 On System V.4 and embedded PowerPC systems compile code using calling
6245 conventions that adheres to the March 1995 draft of the System V
6246 Application Binary Interface, PowerPC processor supplement. This is the
6247 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
6249 @item -mcall-sysv-eabi
6250 Specify both @samp{-mcall-sysv} and @samp{-meabi} options.
6252 @item -mcall-sysv-noeabi
6253 Specify both @samp{-mcall-sysv} and @samp{-mno-eabi} options.
6256 On System V.4 and embedded PowerPC systems compile code using calling
6257 conventions that are similar to those used on AIX. This is the
6258 default if you configured GCC using @samp{powerpc-*-eabiaix}.
6260 @item -mcall-solaris
6261 On System V.4 and embedded PowerPC systems compile code for the Solaris
6265 On System V.4 and embedded PowerPC systems compile code for the
6266 Linux-based GNU system.
6269 @itemx -mno-prototype
6270 On System V.4 and embedded PowerPC systems assume that all calls to
6271 variable argument functions are properly prototyped. Otherwise, the
6272 compiler must insert an instruction before every non prototyped call to
6273 set or clear bit 6 of the condition code register (@var{CR}) to
6274 indicate whether floating point values were passed in the floating point
6275 registers in case the function takes a variable arguments. With
6276 @samp{-mprototype}, only calls to prototyped variable argument functions
6277 will set or clear the bit.
6280 On embedded PowerPC systems, assume that the startup module is called
6281 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
6282 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
6286 On embedded PowerPC systems, assume that the startup module is called
6287 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
6291 On embedded PowerPC systems, assume that the startup module is called
6292 @file{crt0.o} and the standard C libraries are @file{libads.a} and
6296 On embedded PowerPC systems, assume that the startup module is called
6297 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
6301 On System V.4 and embedded PowerPC systems, specify that you are
6302 compiling for a VxWorks system.
6305 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
6306 header to indicate that @samp{eabi} extended relocations are used.
6310 On System V.4 and embedded PowerPC systems do (do not) adhere to the
6311 Embedded Applications Binary Interface (eabi) which is a set of
6312 modifications to the System V.4 specifications. Selecting @option{-meabi}
6313 means that the stack is aligned to an 8 byte boundary, a function
6314 @code{__eabi} is called to from @code{main} to set up the eabi
6315 environment, and the @samp{-msdata} option can use both @code{r2} and
6316 @code{r13} to point to two separate small data areas. Selecting
6317 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
6318 do not call an initialization function from @code{main}, and the
6319 @samp{-msdata} option will only use @code{r13} to point to a single
6320 small data area. The @samp{-meabi} option is on by default if you
6321 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
6324 On System V.4 and embedded PowerPC systems, put small initialized
6325 @code{const} global and static data in the @samp{.sdata2} section, which
6326 is pointed to by register @code{r2}. Put small initialized
6327 non-@code{const} global and static data in the @samp{.sdata} section,
6328 which is pointed to by register @code{r13}. Put small uninitialized
6329 global and static data in the @samp{.sbss} section, which is adjacent to
6330 the @samp{.sdata} section. The @samp{-msdata=eabi} option is
6331 incompatible with the @samp{-mrelocatable} option. The
6332 @samp{-msdata=eabi} option also sets the @samp{-memb} option.
6335 On System V.4 and embedded PowerPC systems, put small global and static
6336 data in the @samp{.sdata} section, which is pointed to by register
6337 @code{r13}. Put small uninitialized global and static data in the
6338 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
6339 The @samp{-msdata=sysv} option is incompatible with the
6340 @samp{-mrelocatable} option.
6342 @item -msdata=default
6344 On System V.4 and embedded PowerPC systems, if @samp{-meabi} is used,
6345 compile code the same as @samp{-msdata=eabi}, otherwise compile code the
6346 same as @samp{-msdata=sysv}.
6349 On System V.4 and embedded PowerPC systems, put small global and static
6350 data in the @samp{.sdata} section. Put small uninitialized global and
6351 static data in the @samp{.sbss} section. Do not use register @code{r13}
6352 to address small data however. This is the default behavior unless
6353 other @samp{-msdata} options are used.
6357 On embedded PowerPC systems, put all initialized global and static data
6358 in the @samp{.data} section, and all uninitialized data in the
6359 @samp{.bss} section.
6362 @cindex smaller data references (PowerPC)
6363 @cindex .sdata/.sdata2 references (PowerPC)
6364 On embedded PowerPC systems, put global and static items less than or
6365 equal to @var{num} bytes into the small data or bss sections instead of
6366 the normal data or bss section. By default, @var{num} is 8. The
6367 @samp{-G @var{num}} switch is also passed to the linker.
6368 All modules should be compiled with the same @samp{-G @var{num}} value.
6371 @itemx -mno-regnames
6372 On System V.4 and embedded PowerPC systems do (do not) emit register
6373 names in the assembly language output using symbolic forms.
6378 @subsection IBM RT Options
6380 @cindex IBM RT options
6382 These @samp{-m} options are defined for the IBM RT PC:
6386 Use an in-line code sequence for integer multiplies. This is the
6389 @item -mcall-lib-mul
6390 Call @code{lmul$$} for integer multiples.
6392 @item -mfull-fp-blocks
6393 Generate full-size floating point data blocks, including the minimum
6394 amount of scratch space recommended by IBM. This is the default.
6396 @item -mminimum-fp-blocks
6397 Do not include extra scratch space in floating point data blocks. This
6398 results in smaller code, but slower execution, since scratch space must
6399 be allocated dynamically.
6401 @cindex @file{varargs.h} and RT PC
6402 @cindex @file{stdarg.h} and RT PC
6403 @item -mfp-arg-in-fpregs
6404 Use a calling sequence incompatible with the IBM calling convention in
6405 which floating point arguments are passed in floating point registers.
6406 Note that @code{varargs.h} and @code{stdargs.h} will not work with
6407 floating point operands if this option is specified.
6409 @item -mfp-arg-in-gregs
6410 Use the normal calling convention for floating point arguments. This is
6413 @item -mhc-struct-return
6414 Return structures of more than one word in memory, rather than in a
6415 register. This provides compatibility with the MetaWare HighC (hc)
6416 compiler. Use the option @samp{-fpcc-struct-return} for compatibility
6417 with the Portable C Compiler (pcc).
6419 @item -mnohc-struct-return
6420 Return some structures of more than one word in registers, when
6421 convenient. This is the default. For compatibility with the
6422 IBM-supplied compilers, use the option @samp{-fpcc-struct-return} or the
6423 option @samp{-mhc-struct-return}.
6427 @subsection MIPS Options
6428 @cindex MIPS options
6430 These @samp{-m} options are defined for the MIPS family of computers:
6433 @item -mcpu=@var{cpu type}
6434 Assume the defaults for the machine type @var{cpu type} when scheduling
6435 instructions. The choices for @var{cpu type} are @samp{r2000}, @samp{r3000},
6436 @samp{r3900}, @samp{r4000}, @samp{r4100}, @samp{r4300}, @samp{r4400},
6437 @samp{r4600}, @samp{r4650}, @samp{r5000}, @samp{r6000}, @samp{r8000},
6438 and @samp{orion}. Additionally, the @samp{r2000}, @samp{r3000},
6439 @samp{r4000}, @samp{r5000}, and @samp{r6000} can be abbreviated as
6440 @samp{r2k} (or @samp{r2K}), @samp{r3k}, etc. While picking a specific
6441 @var{cpu type} will schedule things appropriately for that particular
6442 chip, the compiler will not generate any code that does not meet level 1
6443 of the MIPS ISA (instruction set architecture) without a @samp{-mipsX}
6444 or @samp{-mabi} switch being used.
6447 Issue instructions from level 1 of the MIPS ISA. This is the default.
6448 @samp{r3000} is the default @var{cpu type} at this ISA level.
6451 Issue instructions from level 2 of the MIPS ISA (branch likely, square
6452 root instructions). @samp{r6000} is the default @var{cpu type} at this
6456 Issue instructions from level 3 of the MIPS ISA (64-bit instructions).
6457 @samp{r4000} is the default @var{cpu type} at this ISA level.
6460 Issue instructions from level 4 of the MIPS ISA (conditional move,
6461 prefetch, enhanced FPU instructions). @samp{r8000} is the default
6462 @var{cpu type} at this ISA level.
6465 Assume that 32 32-bit floating point registers are available. This is
6469 Assume that 32 64-bit floating point registers are available. This is
6470 the default when the @samp{-mips3} option is used.
6473 Assume that 32 32-bit general purpose registers are available. This is
6477 Assume that 32 64-bit general purpose registers are available. This is
6478 the default when the @samp{-mips3} option is used.
6481 Force int and long types to be 64 bits wide. See @samp{-mlong32} for an
6482 explanation of the default, and the width of pointers.
6485 Force long types to be 64 bits wide. See @samp{-mlong32} for an
6486 explanation of the default, and the width of pointers.
6489 Force long, int, and pointer types to be 32 bits wide.
6491 If none of @samp{-mlong32}, @samp{-mlong64}, or @samp{-mint64} are set,
6492 the size of ints, longs, and pointers depends on the ABI and ISA chosen.
6493 For @samp{-mabi=32}, and @samp{-mabi=n32}, ints and longs are 32 bits
6494 wide. For @samp{-mabi=64}, ints are 32 bits, and longs are 64 bits wide.
6495 For @samp{-mabi=eabi} and either @samp{-mips1} or @samp{-mips2}, ints
6496 and longs are 32 bits wide. For @samp{-mabi=eabi} and higher ISAs, ints
6497 are 32 bits, and longs are 64 bits wide. The width of pointer types is
6498 the smaller of the width of longs or the width of general purpose
6499 registers (which in turn depends on the ISA).
6506 Generate code for the indicated ABI. The default instruction level is
6507 @samp{-mips1} for @samp{32}, @samp{-mips3} for @samp{n32}, and
6508 @samp{-mips4} otherwise. Conversely, with @samp{-mips1} or
6509 @samp{-mips2}, the default ABI is @samp{32}; otherwise, the default ABI
6513 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
6514 add normal debug information. This is the default for all
6515 platforms except for the OSF/1 reference platform, using the OSF/rose
6516 object format. If the either of the @samp{-gstabs} or @samp{-gstabs+}
6517 switches are used, the @file{mips-tfile} program will encapsulate the
6518 stabs within MIPS ECOFF.
6521 Generate code for the GNU assembler. This is the default on the OSF/1
6522 reference platform, using the OSF/rose object format. Also, this is
6523 the default if the configure option @samp{--with-gnu-as} is used.
6525 @item -msplit-addresses
6526 @itemx -mno-split-addresses
6527 Generate code to load the high and low parts of address constants separately.
6528 This allows @code{gcc} to optimize away redundant loads of the high order
6529 bits of addresses. This optimization requires GNU as and GNU ld.
6530 This optimization is enabled by default for some embedded targets where
6531 GNU as and GNU ld are standard.
6535 The @samp{-mrnames} switch says to output code using the MIPS software
6536 names for the registers, instead of the hardware names (ie, @var{a0}
6537 instead of @var{$4}). The only known assembler that supports this option
6538 is the Algorithmics assembler.
6542 The @samp{-mgpopt} switch says to write all of the data declarations
6543 before the instructions in the text section, this allows the MIPS
6544 assembler to generate one word memory references instead of using two
6545 words for short global or static data items. This is on by default if
6546 optimization is selected.
6550 For each non-inline function processed, the @samp{-mstats} switch
6551 causes the compiler to emit one line to the standard error file to
6552 print statistics about the program (number of registers saved, stack
6557 The @samp{-mmemcpy} switch makes all block moves call the appropriate
6558 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
6559 generating inline code.
6562 @itemx -mno-mips-tfile
6563 The @samp{-mno-mips-tfile} switch causes the compiler not
6564 postprocess the object file with the @file{mips-tfile} program,
6565 after the MIPS assembler has generated it to add debug support. If
6566 @file{mips-tfile} is not run, then no local variables will be
6567 available to the debugger. In addition, @file{stage2} and
6568 @file{stage3} objects will have the temporary file names passed to the
6569 assembler embedded in the object file, which means the objects will
6570 not compare the same. The @samp{-mno-mips-tfile} switch should only
6571 be used when there are bugs in the @file{mips-tfile} program that
6572 prevents compilation.
6575 Generate output containing library calls for floating point.
6576 @strong{Warning:} the requisite libraries are not part of GCC.
6577 Normally the facilities of the machine's usual C compiler are used, but
6578 this can't be done directly in cross-compilation. You must make your
6579 own arrangements to provide suitable library functions for
6583 Generate output containing floating point instructions. This is the
6584 default if you use the unmodified sources.
6587 @itemx -mno-abicalls
6588 Emit (or do not emit) the pseudo operations @samp{.abicalls},
6589 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
6590 position independent code.
6593 @itemx -mno-long-calls
6594 Do all calls with the @samp{JALR} instruction, which requires
6595 loading up a function's address into a register before the call.
6596 You need to use this switch, if you call outside of the current
6597 512 megabyte segment to functions that are not through pointers.
6600 @itemx -mno-half-pic
6601 Put pointers to extern references into the data section and load them
6602 up, rather than put the references in the text section.
6604 @item -membedded-pic
6605 @itemx -mno-embedded-pic
6606 Generate PIC code suitable for some embedded systems. All calls are
6607 made using PC relative address, and all data is addressed using the $gp
6608 register. No more than 65536 bytes of global data may be used. This
6609 requires GNU as and GNU ld which do most of the work. This currently
6610 only works on targets which use ECOFF; it does not work with ELF.
6612 @item -membedded-data
6613 @itemx -mno-embedded-data
6614 Allocate variables to the read-only data section first if possible, then
6615 next in the small data section if possible, otherwise in data. This gives
6616 slightly slower code than the default, but reduces the amount of RAM required
6617 when executing, and thus may be preferred for some embedded systems.
6619 @item -muninit-const-in-rodata
6620 @itemx -mno-uninit-const-in-rodata
6621 When used together with -membedded-data, it will always store uninitialized
6622 const variables in the read-only data section.
6624 @item -msingle-float
6625 @itemx -mdouble-float
6626 The @samp{-msingle-float} switch tells gcc to assume that the floating
6627 point coprocessor only supports single precision operations, as on the
6628 @samp{r4650} chip. The @samp{-mdouble-float} switch permits gcc to use
6629 double precision operations. This is the default.
6633 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
6634 as on the @samp{r4650} chip.
6637 Turns on @samp{-msingle-float}, @samp{-mmad}, and, at least for now,
6642 Enable 16-bit instructions.
6645 Use the entry and exit pseudo ops. This option can only be used with
6649 Compile code for the processor in little endian mode.
6650 The requisite libraries are assumed to exist.
6653 Compile code for the processor in big endian mode.
6654 The requisite libraries are assumed to exist.
6657 @cindex smaller data references (MIPS)
6658 @cindex gp-relative references (MIPS)
6659 Put global and static items less than or equal to @var{num} bytes into
6660 the small data or bss sections instead of the normal data or bss
6661 section. This allows the assembler to emit one word memory reference
6662 instructions based on the global pointer (@var{gp} or @var{$28}),
6663 instead of the normal two words used. By default, @var{num} is 8 when
6664 the MIPS assembler is used, and 0 when the GNU assembler is used. The
6665 @samp{-G @var{num}} switch is also passed to the assembler and linker.
6666 All modules should be compiled with the same @samp{-G @var{num}}
6670 Tell the MIPS assembler to not run its preprocessor over user
6671 assembler files (with a @samp{.s} suffix) when assembling them.
6674 Pass an option to gas which will cause nops to be inserted if
6675 the read of the destination register of an mfhi or mflo instruction
6676 occurs in the following two instructions.
6679 Do not include the default crt0.
6683 These options are defined by the macro
6684 @code{TARGET_SWITCHES} in the machine description. The default for the
6685 options is also defined by that macro, which enables you to change the
6690 @subsection Intel 386 Options
6691 @cindex i386 Options
6692 @cindex Intel 386 Options
6694 These @samp{-m} options are defined for the i386 family of computers:
6697 @item -mcpu=@var{cpu type}
6698 Assume the defaults for the machine type @var{cpu type} when scheduling
6699 instructions. The choices for @var{cpu type} are @samp{i386},
6700 @samp{i486}, @samp{i586}, @samp{i686}, @samp{pentium},
6701 @samp{pentiumpro}, @samp{pentium4}, @samp{k6}, and @samp{athlon}
6703 While picking a specific @var{cpu type} will schedule things appropriately
6704 for that particular chip, the compiler will not generate any code that
6705 does not run on the i386 without the @samp{-march=@var{cpu type}} option
6706 being used. @samp{i586} is equivalent to @samp{pentium} and @samp{i686}
6707 is equivalent to @samp{pentiumpro}. @samp{k6} and @samp{athlon} are the
6708 AMD chips as opposed to the Intel ones.
6710 @item -march=@var{cpu type}
6711 Generate instructions for the machine type @var{cpu type}. The choices
6712 for @var{cpu type} are the same as for @samp{-mcpu}. Moreover,
6713 specifying @samp{-march=@var{cpu type}} implies @samp{-mcpu=@var{cpu type}}.
6719 Synonyms for -mcpu=i386, -mcpu=i486, -mcpu=pentium, and -mcpu=pentiumpro
6720 respectively. These synonyms are deprecated.
6722 @item -mintel-syntax
6723 Emit assembly using Intel syntax opcodes instead of AT&T syntax.
6727 Control whether or not the compiler uses IEEE floating point
6728 comparisons. These handle correctly the case where the result of a
6729 comparison is unordered.
6732 Generate output containing library calls for floating point.
6733 @strong{Warning:} the requisite libraries are not part of GCC.
6734 Normally the facilities of the machine's usual C compiler are used, but
6735 this can't be done directly in cross-compilation. You must make your
6736 own arrangements to provide suitable library functions for
6739 On machines where a function returns floating point results in the 80387
6740 register stack, some floating point opcodes may be emitted even if
6741 @samp{-msoft-float} is used.
6743 @item -mno-fp-ret-in-387
6744 Do not use the FPU registers for return values of functions.
6746 The usual calling convention has functions return values of types
6747 @code{float} and @code{double} in an FPU register, even if there
6748 is no FPU. The idea is that the operating system should emulate
6751 The option @samp{-mno-fp-ret-in-387} causes such values to be returned
6752 in ordinary CPU registers instead.
6754 @item -mno-fancy-math-387
6755 Some 387 emulators do not support the @code{sin}, @code{cos} and
6756 @code{sqrt} instructions for the 387. Specify this option to avoid
6757 generating those instructions. This option is the default on FreeBSD.
6758 As of revision 2.6.1, these instructions are not generated unless you
6759 also use the @samp{-funsafe-math-optimizations} switch.
6761 @item -malign-double
6762 @itemx -mno-align-double
6763 Control whether GCC aligns @code{double}, @code{long double}, and
6764 @code{long long} variables on a two word boundary or a one word
6765 boundary. Aligning @code{double} variables on a two word boundary will
6766 produce code that runs somewhat faster on a @samp{Pentium} at the
6767 expense of more memory.
6769 @item -m128bit-long-double
6770 @itemx -m128bit-long-double
6771 Control the size of @code{long double} type. i386 application binary interface
6772 specify the size to be 12 bytes, while modern architectures (Pentium and newer)
6773 preffer @code{long double} aligned to 8 or 16 byte boundary. This is
6774 impossible to reach with 12 byte long doubles in the array accesses.
6776 @strong{Warning:} if you use the @samp{-m128bit-long-double} switch, the
6777 structures and arrays containing @code{long double} will change their size as
6778 well as function calling convention for function taking @code{long double}
6781 @item -m96bit-long-double
6782 @itemx -m96bit-long-double
6783 Set the size of @code{long double} to 96 bits as required by the i386
6784 application binary interface. This is the default.
6787 @itemx -mno-svr3-shlib
6788 Control whether GCC places uninitialized locals into @code{bss} or
6789 @code{data}. @samp{-msvr3-shlib} places these locals into @code{bss}.
6790 These options are meaningful only on System V Release 3.
6792 @item -mno-wide-multiply
6793 @itemx -mwide-multiply
6794 Control whether GCC uses the @code{mul} and @code{imul} that produce
6795 64-bit results in @code{eax:edx} from 32-bit operands to do @code{long
6796 long} multiplies and 32-bit division by constants.
6799 Use a different function-calling convention, in which functions that
6800 take a fixed number of arguments return with the @code{ret} @var{num}
6801 instruction, which pops their arguments while returning. This saves one
6802 instruction in the caller since there is no need to pop the arguments
6805 You can specify that an individual function is called with this calling
6806 sequence with the function attribute @samp{stdcall}. You can also
6807 override the @samp{-mrtd} option by using the function attribute
6808 @samp{cdecl}. @xref{Function Attributes}.
6810 @strong{Warning:} this calling convention is incompatible with the one
6811 normally used on Unix, so you cannot use it if you need to call
6812 libraries compiled with the Unix compiler.
6814 Also, you must provide function prototypes for all functions that
6815 take variable numbers of arguments (including @code{printf});
6816 otherwise incorrect code will be generated for calls to those
6819 In addition, seriously incorrect code will result if you call a
6820 function with too many arguments. (Normally, extra arguments are
6821 harmlessly ignored.)
6823 @item -mregparm=@var{num}
6824 Control how many registers are used to pass integer arguments. By
6825 default, no registers are used to pass arguments, and at most 3
6826 registers can be used. You can control this behavior for a specific
6827 function by using the function attribute @samp{regparm}.
6828 @xref{Function Attributes}.
6830 @strong{Warning:} if you use this switch, and
6831 @var{num} is nonzero, then you must build all modules with the same
6832 value, including any libraries. This includes the system libraries and
6835 @item -mpreferred-stack-boundary=@var{num}
6836 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
6837 byte boundary. If @samp{-mpreferred-stack-boundary} is not specified,
6838 the default is 4 (16 bytes or 128 bits).
6840 The stack is required to be aligned on a 4 byte boundary. On Pentium
6841 and PentiumPro, @code{double} and @code{long double} values should be
6842 aligned to an 8 byte boundary (see @samp{-malign-double}) or suffer
6843 significant run time performance penalties. On Pentium III, the
6844 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
6845 penalties if it is not 16 byte aligned.
6847 To ensure proper alignment of this values on the stack, the stack boundary
6848 must be as aligned as that required by any value stored on the stack.
6849 Further, every function must be generated such that it keeps the stack
6850 aligned. Thus calling a function compiled with a higher preferred
6851 stack boundary from a function compiled with a lower preferred stack
6852 boundary will most likely misalign the stack. It is recommended that
6853 libraries that use callbacks always use the default setting.
6855 This extra alignment does consume extra stack space. Code that is sensitive
6856 to stack space usage, such as embedded systems and operating system kernels,
6857 may want to reduce the preferred alignment to
6858 @samp{-mpreferred-stack-boundary=2}.
6862 Use PUSH operations to store outgoing parameters. This method is shorter
6863 and usually equally fast as method using SUB/MOV operations and is enabled
6864 by default. In some cases disabling it may improve performance because of
6865 improved scheduling and reduced dependencies.
6867 @item -maccumulate-outgoing-args
6868 @kindex -maccumulate-outgoing-args
6869 If enabled, the maximum amount of space required for outgoing arguments will be
6870 computed in the function prologue. This in faster on most modern CPUs
6871 because of reduced dependencies, improved scheduling and reduced stack usage
6872 when preferred stack boundary is not equal to 2. The drawback is a notable
6873 increase in code size. This switch implies -mno-push-args.
6877 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
6878 on thread-safe exception handling must compile and link all code with the
6879 @samp{-mthreads} option. When compiling, @samp{-mthreads} defines
6880 @samp{-D_MT}; when linking, it links in a special thread helper library
6881 @samp{-lmingwthrd} which cleans up per thread exception handling data.
6883 @item -mno-align-stringops
6884 @kindex -mno-align-stringops
6885 Do not align destination of inlined string operations. This switch reduces
6886 code size and improves performance in case the destination is already aligned,
6887 but gcc don't know about it.
6889 @item -minline-all-stringops
6890 @kindex -minline-all-stringops
6891 By default GCC inlines string operations only when destination is known to be
6892 aligned at least to 4 byte boundary. This enables more inlining, increase code
6893 size, but may improve performance of code that depends on fast memcpy, strlen
6894 and memset for short lengths.
6896 @item -momit-leaf-frame-pointer
6897 @kindex -momit-leaf-frame-pointer
6898 Don't keep the frame pointer in a register for leaf functions. This
6899 avoids the instructions to save, set up and restore frame pointers and
6900 makes an extra register available in leaf functions. The option
6901 @samp{-fomit-frame-pointer} removes the frame pointer for all functions
6902 which might make debugging harder.
6906 @subsection HPPA Options
6907 @cindex HPPA Options
6909 These @samp{-m} options are defined for the HPPA family of computers:
6912 @item -march=@var{architecture type}
6913 Generate code for the specified architecture. The choices for
6914 @var{architecture type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
6915 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
6916 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
6917 architecture option for your machine. Code compiled for lower numbered
6918 architectures will run on higher numbered architectures, but not the
6921 PA 2.0 support currently requires gas snapshot 19990413 or later. The
6922 next release of binutils (current is 2.9.1) will probably contain PA 2.0
6926 @itemx -mpa-risc-1-1
6927 @itemx -mpa-risc-2-0
6928 Synonyms for -march=1.0, -march=1.1, and -march=2.0 respectively.
6931 Generate code suitable for big switch tables. Use this option only if
6932 the assembler/linker complain about out of range branches within a switch
6935 @item -mjump-in-delay
6936 Fill delay slots of function calls with unconditional jump instructions
6937 by modifying the return pointer for the function call to be the target
6938 of the conditional jump.
6940 @item -mdisable-fpregs
6941 Prevent floating point registers from being used in any manner. This is
6942 necessary for compiling kernels which perform lazy context switching of
6943 floating point registers. If you use this option and attempt to perform
6944 floating point operations, the compiler will abort.
6946 @item -mdisable-indexing
6947 Prevent the compiler from using indexing address modes. This avoids some
6948 rather obscure problems when compiling MIG generated code under MACH.
6950 @item -mno-space-regs
6951 Generate code that assumes the target has no space registers. This allows
6952 GCC to generate faster indirect calls and use unscaled index address modes.
6954 Such code is suitable for level 0 PA systems and kernels.
6956 @item -mfast-indirect-calls
6957 Generate code that assumes calls never cross space boundaries. This
6958 allows GCC to emit code which performs faster indirect calls.
6960 This option will not work in the presence of shared libraries or nested
6963 @item -mlong-load-store
6964 Generate 3-instruction load and store sequences as sometimes required by
6965 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
6968 @item -mportable-runtime
6969 Use the portable calling conventions proposed by HP for ELF systems.
6972 Enable the use of assembler directives only GAS understands.
6974 @item -mschedule=@var{cpu type}
6975 Schedule code according to the constraints for the machine type
6976 @var{cpu type}. The choices for @var{cpu type} are @samp{700}
6977 @samp{7100}, @samp{7100LC}, @samp{7200}, and @samp{8000}. Refer to
6978 @file{/usr/lib/sched.models} on an HP-UX system to determine the
6979 proper scheduling option for your machine.
6982 Enable the optimization pass in the HPUX linker. Note this makes symbolic
6983 debugging impossible. It also triggers a bug in the HPUX 8 and HPUX 9 linkers
6984 in which they give bogus error messages when linking some programs.
6987 Generate output containing library calls for floating point.
6988 @strong{Warning:} the requisite libraries are not available for all HPPA
6989 targets. Normally the facilities of the machine's usual C compiler are
6990 used, but this cannot be done directly in cross-compilation. You must make
6991 your own arrangements to provide suitable library functions for
6992 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
6993 does provide software floating point support.
6995 @samp{-msoft-float} changes the calling convention in the output file;
6996 therefore, it is only useful if you compile @emph{all} of a program with
6997 this option. In particular, you need to compile @file{libgcc.a}, the
6998 library that comes with GCC, with @samp{-msoft-float} in order for
7002 @node Intel 960 Options
7003 @subsection Intel 960 Options
7005 These @samp{-m} options are defined for the Intel 960 implementations:
7008 @item -m@var{cpu type}
7009 Assume the defaults for the machine type @var{cpu type} for some of
7010 the other options, including instruction scheduling, floating point
7011 support, and addressing modes. The choices for @var{cpu type} are
7012 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
7013 @samp{sa}, and @samp{sb}.
7019 The @samp{-mnumerics} option indicates that the processor does support
7020 floating-point instructions. The @samp{-msoft-float} option indicates
7021 that floating-point support should not be assumed.
7023 @item -mleaf-procedures
7024 @itemx -mno-leaf-procedures
7025 Do (or do not) attempt to alter leaf procedures to be callable with the
7026 @code{bal} instruction as well as @code{call}. This will result in more
7027 efficient code for explicit calls when the @code{bal} instruction can be
7028 substituted by the assembler or linker, but less efficient code in other
7029 cases, such as calls via function pointers, or using a linker that doesn't
7030 support this optimization.
7033 @itemx -mno-tail-call
7034 Do (or do not) make additional attempts (beyond those of the
7035 machine-independent portions of the compiler) to optimize tail-recursive
7036 calls into branches. You may not want to do this because the detection of
7037 cases where this is not valid is not totally complete. The default is
7038 @samp{-mno-tail-call}.
7040 @item -mcomplex-addr
7041 @itemx -mno-complex-addr
7042 Assume (or do not assume) that the use of a complex addressing mode is a
7043 win on this implementation of the i960. Complex addressing modes may not
7044 be worthwhile on the K-series, but they definitely are on the C-series.
7045 The default is currently @samp{-mcomplex-addr} for all processors except
7049 @itemx -mno-code-align
7050 Align code to 8-byte boundaries for faster fetching (or don't bother).
7051 Currently turned on by default for C-series implementations only.
7054 @item -mclean-linkage
7055 @itemx -mno-clean-linkage
7056 These options are not fully implemented.
7060 @itemx -mic2.0-compat
7061 @itemx -mic3.0-compat
7062 Enable compatibility with iC960 v2.0 or v3.0.
7066 Enable compatibility with the iC960 assembler.
7068 @item -mstrict-align
7069 @itemx -mno-strict-align
7070 Do not permit (do permit) unaligned accesses.
7073 Enable structure-alignment compatibility with Intel's gcc release version
7074 1.3 (based on gcc 1.37). This option implies @samp{-mstrict-align}.
7076 @item -mlong-double-64
7077 Implement type @samp{long double} as 64-bit floating point numbers.
7078 Without the option @samp{long double} is implemented by 80-bit
7079 floating point numbers. The only reason we have it because there is
7080 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
7081 is only useful for people using soft-float targets. Otherwise, we
7082 should recommend against use of it.
7086 @node DEC Alpha Options
7087 @subsection DEC Alpha Options
7089 These @samp{-m} options are defined for the DEC Alpha implementations:
7092 @item -mno-soft-float
7094 Use (do not use) the hardware floating-point instructions for
7095 floating-point operations. When @option{-msoft-float} is specified,
7096 functions in @file{libgcc.a} will be used to perform floating-point
7097 operations. Unless they are replaced by routines that emulate the
7098 floating-point operations, or compiled in such a way as to call such
7099 emulations routines, these routines will issue floating-point
7100 operations. If you are compiling for an Alpha without floating-point
7101 operations, you must ensure that the library is built so as not to call
7104 Note that Alpha implementations without floating-point operations are
7105 required to have floating-point registers.
7109 Generate code that uses (does not use) the floating-point register set.
7110 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7111 register set is not used, floating point operands are passed in integer
7112 registers as if they were integers and floating-point results are passed
7113 in $0 instead of $f0. This is a non-standard calling sequence, so any
7114 function with a floating-point argument or return value called by code
7115 compiled with @option{-mno-fp-regs} must also be compiled with that
7118 A typical use of this option is building a kernel that does not use,
7119 and hence need not save and restore, any floating-point registers.
7122 The Alpha architecture implements floating-point hardware optimized for
7123 maximum performance. It is mostly compliant with the IEEE floating
7124 point standard. However, for full compliance, software assistance is
7125 required. This option generates code fully IEEE compliant code
7126 @emph{except} that the @var{inexact flag} is not maintained (see below).
7127 If this option is turned on, the CPP macro @code{_IEEE_FP} is defined
7128 during compilation. The option is a shorthand for: @samp{-D_IEEE_FP
7129 -mfp-trap-mode=su -mtrap-precision=i -mieee-conformant}. The resulting
7130 code is less efficient but is able to correctly support denormalized
7131 numbers and exceptional IEEE values such as not-a-number and plus/minus
7132 infinity. Other Alpha compilers call this option
7133 @option{-ieee_with_no_inexact}.
7135 @item -mieee-with-inexact
7136 @c overfull hbox here --bob 22 jul96
7137 @c original text between ignore ... end ignore
7139 This is like @samp{-mieee} except the generated code also maintains the
7140 IEEE @var{inexact flag}. Turning on this option causes the generated
7141 code to implement fully-compliant IEEE math. The option is a shorthand
7142 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus @samp{-mieee-conformant},
7143 @samp{-mfp-trap-mode=sui}, and @samp{-mtrap-precision=i}. On some Alpha
7144 implementations the resulting code may execute significantly slower than
7145 the code generated by default. Since there is very little code that
7146 depends on the @var{inexact flag}, you should normally not specify this
7147 option. Other Alpha compilers call this option
7148 @samp{-ieee_with_inexact}.
7150 @c changed paragraph
7151 This is like @samp{-mieee} except the generated code also maintains the
7152 IEEE @var{inexact flag}. Turning on this option causes the generated
7153 code to implement fully-compliant IEEE math. The option is a shorthand
7154 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus the three following:
7155 @samp{-mieee-conformant},
7156 @samp{-mfp-trap-mode=sui},
7157 and @samp{-mtrap-precision=i}.
7158 On some Alpha implementations the resulting code may execute
7159 significantly slower than the code generated by default. Since there
7160 is very little code that depends on the @var{inexact flag}, you should
7161 normally not specify this option. Other Alpha compilers call this
7162 option @samp{-ieee_with_inexact}.
7163 @c end changes to prevent overfull hboxes
7165 @item -mfp-trap-mode=@var{trap mode}
7166 This option controls what floating-point related traps are enabled.
7167 Other Alpha compilers call this option @samp{-fptm }@var{trap mode}.
7168 The trap mode can be set to one of four values:
7172 This is the default (normal) setting. The only traps that are enabled
7173 are the ones that cannot be disabled in software (e.g., division by zero
7177 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7181 Like @samp{su}, but the instructions are marked to be safe for software
7182 completion (see Alpha architecture manual for details).
7185 Like @samp{su}, but inexact traps are enabled as well.
7188 @item -mfp-rounding-mode=@var{rounding mode}
7189 Selects the IEEE rounding mode. Other Alpha compilers call this option
7190 @samp{-fprm }@var{rounding mode}. The @var{rounding mode} can be one
7195 Normal IEEE rounding mode. Floating point numbers are rounded towards
7196 the nearest machine number or towards the even machine number in case
7200 Round towards minus infinity.
7203 Chopped rounding mode. Floating point numbers are rounded towards zero.
7206 Dynamic rounding mode. A field in the floating point control register
7207 (@var{fpcr}, see Alpha architecture reference manual) controls the
7208 rounding mode in effect. The C library initializes this register for
7209 rounding towards plus infinity. Thus, unless your program modifies the
7210 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7213 @item -mtrap-precision=@var{trap precision}
7214 In the Alpha architecture, floating point traps are imprecise. This
7215 means without software assistance it is impossible to recover from a
7216 floating trap and program execution normally needs to be terminated.
7217 GCC can generate code that can assist operating system trap handlers
7218 in determining the exact location that caused a floating point trap.
7219 Depending on the requirements of an application, different levels of
7220 precisions can be selected:
7224 Program precision. This option is the default and means a trap handler
7225 can only identify which program caused a floating point exception.
7228 Function precision. The trap handler can determine the function that
7229 caused a floating point exception.
7232 Instruction precision. The trap handler can determine the exact
7233 instruction that caused a floating point exception.
7236 Other Alpha compilers provide the equivalent options called
7237 @samp{-scope_safe} and @samp{-resumption_safe}.
7239 @item -mieee-conformant
7240 This option marks the generated code as IEEE conformant. You must not
7241 use this option unless you also specify @samp{-mtrap-precision=i} and either
7242 @samp{-mfp-trap-mode=su} or @samp{-mfp-trap-mode=sui}. Its only effect
7243 is to emit the line @samp{.eflag 48} in the function prologue of the
7244 generated assembly file. Under DEC Unix, this has the effect that
7245 IEEE-conformant math library routines will be linked in.
7247 @item -mbuild-constants
7248 Normally GCC examines a 32- or 64-bit integer constant to
7249 see if it can construct it from smaller constants in two or three
7250 instructions. If it cannot, it will output the constant as a literal and
7251 generate code to load it from the data segment at runtime.
7253 Use this option to require GCC to construct @emph{all} integer constants
7254 using code, even if it takes more instructions (the maximum is six).
7256 You would typically use this option to build a shared library dynamic
7257 loader. Itself a shared library, it must relocate itself in memory
7258 before it can find the variables and constants in its own data segment.
7262 Select whether to generate code to be assembled by the vendor-supplied
7263 assembler (@samp{-malpha-as}) or by the GNU assembler @samp{-mgas}.
7271 Indicate whether GCC should generate code to use the optional BWX,
7272 CIX, and MAX instruction sets. The default is to use the instruction sets
7273 supported by the CPU type specified via @samp{-mcpu=} option or that
7274 of the CPU on which GCC was built if none was specified.
7276 @item -mcpu=@var{cpu_type}
7277 Set the instruction set, register set, and instruction scheduling
7278 parameters for machine type @var{cpu_type}. You can specify either the
7279 @samp{EV} style name or the corresponding chip number. GCC
7280 supports scheduling parameters for the EV4 and EV5 family of processors
7281 and will choose the default values for the instruction set from
7282 the processor you specify. If you do not specify a processor type,
7283 GCC will default to the processor on which the compiler was built.
7285 Supported values for @var{cpu_type} are
7290 Schedules as an EV4 and has no instruction set extensions.
7294 Schedules as an EV5 and has no instruction set extensions.
7298 Schedules as an EV5 and supports the BWX extension.
7303 Schedules as an EV5 and supports the BWX and MAX extensions.
7307 Schedules as an EV5 (until Digital releases the scheduling parameters
7308 for the EV6) and supports the BWX, CIX, and MAX extensions.
7311 @item -mmemory-latency=@var{time}
7312 Sets the latency the scheduler should assume for typical memory
7313 references as seen by the application. This number is highly
7314 dependent on the memory access patterns used by the application
7315 and the size of the external cache on the machine.
7317 Valid options for @var{time} are
7321 A decimal number representing clock cycles.
7327 The compiler contains estimates of the number of clock cycles for
7328 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7329 (also called Dcache, Scache, and Bcache), as well as to main memory.
7330 Note that L3 is only valid for EV5.
7335 @node Clipper Options
7336 @subsection Clipper Options
7338 These @samp{-m} options are defined for the Clipper implementations:
7342 Produce code for a C300 Clipper processor. This is the default.
7345 Produce code for a C400 Clipper processor i.e. use floating point
7349 @node H8/300 Options
7350 @subsection H8/300 Options
7352 These @samp{-m} options are defined for the H8/300 implementations:
7356 Shorten some address references at link time, when possible; uses the
7357 linker option @samp{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
7358 ld.info, Using ld}, for a fuller description.
7361 Generate code for the H8/300H.
7364 Generate code for the H8/S.
7367 Generate code for the H8/S2600. This switch must be used with -ms.
7370 Make @code{int} data 32 bits by default.
7373 On the H8/300H and H8/S, use the same alignment rules as for the H8/300.
7374 The default for the H8/300H and H8/S is to align longs and floats on 4
7376 @samp{-malign-300} causes them to be aligned on 2 byte boundaries.
7377 This option has no effect on the H8/300.
7381 @subsection SH Options
7383 These @samp{-m} options are defined for the SH implementations:
7387 Generate code for the SH1.
7390 Generate code for the SH2.
7393 Generate code for the SH3.
7396 Generate code for the SH3e.
7399 Generate code for the SH4 without a floating-point unit.
7401 @item -m4-single-only
7402 Generate code for the SH4 with a floating-point unit that only
7403 supports single-precision arithmentic.
7406 Generate code for the SH4 assuming the floating-point unit is in
7407 single-precision mode by default.
7410 Generate code for the SH4.
7413 Compile code for the processor in big endian mode.
7416 Compile code for the processor in little endian mode.
7419 Align doubles at 64-bit boundaries. Note that this changes the calling
7420 conventions, and thus some functions from the standard C library will
7421 not work unless you recompile it first with -mdalign.
7424 Shorten some address references at link time, when possible; uses the
7425 linker option @samp{-relax}.
7428 Use 32-bit offsets in @code{switch} tables. The default is to use
7432 Enable the use of the instruction @code{fmovd}.
7435 Comply with the calling conventions defined by Hitachi.
7438 Mark the @code{MAC} register as call-clobbered, even if
7439 @option{-mhitachi} is given.
7442 Dump instruction size and location in the assembly code.
7445 This option is deprecated. It pads structures to multiple of 4 bytes,
7446 which is incompatible with the SH ABI.
7449 Optimize for space instead of speed. Implied by @option{-Os}.
7452 When generating position-independent code, emit function calls using
7453 the Global Offset Table instead of the Procedure Linkage Table.
7456 Generate a library function call to invalidate instruction cache
7457 entries, after fixing up a trampoline. This library function call
7458 doesn't assume it can write to the whole memory address space. This
7459 is the default when the target is @code{sh-*-linux*}.
7462 @node System V Options
7463 @subsection Options for System V
7465 These additional options are available on System V Release 4 for
7466 compatibility with other compilers on those systems:
7470 Create a shared object.
7471 It is recommended that @samp{-symbolic} or @samp{-shared} be used instead.
7474 Identify the versions of each tool used by the compiler, in a
7475 @code{.ident} assembler directive in the output.
7478 Refrain from adding @code{.ident} directives to the output file (this is
7481 @item -YP\,@var{dirs}
7482 Search the directories @var{dirs}, and no others, for libraries
7483 specified with @samp{-l}.
7485 @item -Ym\,@var{dir}
7486 Look in the directory @var{dir} to find the M4 preprocessor.
7487 The assembler uses this option.
7488 @c This is supposed to go with a -Yd for predefined M4 macro files, but
7489 @c the generic assembler that comes with Solaris takes just -Ym.
7492 @node TMS320C3x/C4x Options
7493 @subsection TMS320C3x/C4x Options
7494 @cindex TMS320C3x/C4x Options
7496 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
7500 @item -mcpu=@var{cpu_type}
7501 Set the instruction set, register set, and instruction scheduling
7502 parameters for machine type @var{cpu_type}. Supported values for
7503 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
7504 @samp{c44}. The default is @samp{c40} to generate code for the
7509 @itemx -msmall-memory
7511 Generates code for the big or small memory model. The small memory
7512 model assumed that all data fits into one 64K word page. At run-time
7513 the data page (DP) register must be set to point to the 64K page
7514 containing the .bss and .data program sections. The big memory model is
7515 the default and requires reloading of the DP register for every direct
7520 Allow (disallow) allocation of general integer operands into the block
7525 Enable (disable) generation of code using decrement and branch,
7526 DBcond(D), instructions. This is enabled by default for the C4x. To be
7527 on the safe side, this is disabled for the C3x, since the maximum
7528 iteration count on the C3x is 2^23 + 1 (but who iterates loops more than
7529 2^23 times on the C3x?). Note that GCC will try to reverse a loop so
7530 that it can utilise the decrement and branch instruction, but will give
7531 up if there is more than one memory reference in the loop. Thus a loop
7532 where the loop counter is decremented can generate slightly more
7533 efficient code, in cases where the RPTB instruction cannot be utilised.
7535 @item -mdp-isr-reload
7537 Force the DP register to be saved on entry to an interrupt service
7538 routine (ISR), reloaded to point to the data section, and restored on
7539 exit from the ISR. This should not be required unless someone has
7540 violated the small memory model by modifying the DP register, say within
7545 For the C3x use the 24-bit MPYI instruction for integer multiplies
7546 instead of a library call to guarantee 32-bit results. Note that if one
7547 of the operands is a constant, then the multiplication will be performed
7548 using shifts and adds. If the -mmpyi option is not specified for the C3x,
7549 then squaring operations are performed inline instead of a library call.
7552 @itemx -mno-fast-fix
7553 The C3x/C4x FIX instruction to convert a floating point value to an
7554 integer value chooses the nearest integer less than or equal to the
7555 floating point value rather than to the nearest integer. Thus if the
7556 floating point number is negative, the result will be incorrectly
7557 truncated an additional code is necessary to detect and correct this
7558 case. This option can be used to disable generation of the additional
7559 code required to correct the result.
7563 Enable (disable) generation of repeat block sequences using the RPTB
7564 instruction for zero overhead looping. The RPTB construct is only used
7565 for innermost loops that do not call functions or jump across the loop
7566 boundaries. There is no advantage having nested RPTB loops due to the
7567 overhead required to save and restore the RC, RS, and RE registers.
7568 This is enabled by default with -O2.
7570 @item -mrpts=@var{count}
7572 Enable (disable) the use of the single instruction repeat instruction
7573 RPTS. If a repeat block contains a single instruction, and the loop
7574 count can be guaranteed to be less than the value @var{count}, GCC will
7575 emit a RPTS instruction instead of a RPTB. If no value is specified,
7576 then a RPTS will be emitted even if the loop count cannot be determined
7577 at compile time. Note that the repeated instruction following RPTS does
7578 not have to be reloaded from memory each iteration, thus freeing up the
7579 CPU buses for operands. However, since interrupts are blocked by this
7580 instruction, it is disabled by default.
7582 @item -mloop-unsigned
7583 @itemx -mno-loop-unsigned
7584 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
7585 is 2^31 + 1 since these instructions test if the iteration count is
7586 negative to terminate the loop. If the iteration count is unsigned
7587 there is a possibility than the 2^31 + 1 maximum iteration count may be
7588 exceeded. This switch allows an unsigned iteration count.
7591 Try to emit an assembler syntax that the TI assembler (asm30) is happy
7592 with. This also enforces compatibility with the API employed by the TI
7593 C3x C compiler. For example, long doubles are passed as structures
7594 rather than in floating point registers.
7598 Generate code that uses registers (stack) for passing arguments to functions.
7599 By default, arguments are passed in registers where possible rather
7600 than by pushing arguments on to the stack.
7602 @item -mparallel-insns
7603 @itemx -mno-parallel-insns
7604 Allow the generation of parallel instructions. This is enabled by
7607 @item -mparallel-mpy
7608 @itemx -mno-parallel-mpy
7609 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
7610 provided -mparallel-insns is also specified. These instructions have
7611 tight register constraints which can pessimize the code generation
7617 @subsection V850 Options
7618 @cindex V850 Options
7620 These @samp{-m} options are defined for V850 implementations:
7624 @itemx -mno-long-calls
7625 Treat all calls as being far away (near). If calls are assumed to be
7626 far away, the compiler will always load the functions address up into a
7627 register, and call indirect through the pointer.
7631 Do not optimize (do optimize) basic blocks that use the same index
7632 pointer 4 or more times to copy pointer into the @code{ep} register, and
7633 use the shorter @code{sld} and @code{sst} instructions. The @samp{-mep}
7634 option is on by default if you optimize.
7636 @item -mno-prolog-function
7637 @itemx -mprolog-function
7638 Do not use (do use) external functions to save and restore registers at
7639 the prolog and epilog of a function. The external functions are slower,
7640 but use less code space if more than one function saves the same number
7641 of registers. The @samp{-mprolog-function} option is on by default if
7645 Try to make the code as small as possible. At present, this just turns
7646 on the @samp{-mep} and @samp{-mprolog-function} options.
7649 Put static or global variables whose size is @var{n} bytes or less into
7650 the tiny data area that register @code{ep} points to. The tiny data
7651 area can hold up to 256 bytes in total (128 bytes for byte references).
7654 Put static or global variables whose size is @var{n} bytes or less into
7655 the small data area that register @code{gp} points to. The small data
7656 area can hold up to 64 kilobytes.
7659 Put static or global variables whose size is @var{n} bytes or less into
7660 the first 32 kilobytes of memory.
7663 Specify that the target processor is the V850.
7666 Generate code suitable for big switch tables. Use this option only if
7667 the assembler/linker complain about out of range branches within a switch
7672 @subsection ARC Options
7675 These options are defined for ARC implementations:
7679 Compile code for little endian mode. This is the default.
7682 Compile code for big endian mode.
7685 Prepend the name of the cpu to all public symbol names.
7686 In multiple-processor systems, there are many ARC variants with different
7687 instruction and register set characteristics. This flag prevents code
7688 compiled for one cpu to be linked with code compiled for another.
7689 No facility exists for handling variants that are "almost identical".
7690 This is an all or nothing option.
7692 @item -mcpu=@var{cpu}
7693 Compile code for ARC variant @var{cpu}.
7694 Which variants are supported depend on the configuration.
7695 All variants support @samp{-mcpu=base}, this is the default.
7697 @item -mtext=@var{text section}
7698 @itemx -mdata=@var{data section}
7699 @itemx -mrodata=@var{readonly data section}
7700 Put functions, data, and readonly data in @var{text section},
7701 @var{data section}, and @var{readonly data section} respectively
7702 by default. This can be overridden with the @code{section} attribute.
7703 @xref{Variable Attributes}.
7708 @subsection NS32K Options
7709 @cindex NS32K options
7711 These are the @samp{-m} options defined for the 32000 series. The default
7712 values for these options depends on which style of 32000 was selected when
7713 the compiler was configured; the defaults for the most common choices are
7719 Generate output for a 32032. This is the default
7720 when the compiler is configured for 32032 and 32016 based systems.
7724 Generate output for a 32332. This is the default
7725 when the compiler is configured for 32332-based systems.
7729 Generate output for a 32532. This is the default
7730 when the compiler is configured for 32532-based systems.
7733 Generate output containing 32081 instructions for floating point.
7734 This is the default for all systems.
7737 Generate output containing 32381 instructions for floating point. This
7738 also implies @samp{-m32081}. The 32381 is only compatible with the 32332
7739 and 32532 cpus. This is the default for the pc532-netbsd configuration.
7742 Try and generate multiply-add floating point instructions @code{polyF}
7743 and @code{dotF}. This option is only available if the @samp{-m32381}
7744 option is in effect. Using these instructions requires changes to to
7745 register allocation which generally has a negative impact on
7746 performance. This option should only be enabled when compiling code
7747 particularly likely to make heavy use of multiply-add instructions.
7750 Do not try and generate multiply-add floating point instructions
7751 @code{polyF} and @code{dotF}. This is the default on all platforms.
7754 Generate output containing library calls for floating point.
7755 @strong{Warning:} the requisite libraries may not be available.
7758 Do not use the bit-field instructions. On some machines it is faster to
7759 use shifting and masking operations. This is the default for the pc532.
7762 Do use the bit-field instructions. This is the default for all platforms
7766 Use a different function-calling convention, in which functions
7767 that take a fixed number of arguments return pop their
7768 arguments on return with the @code{ret} instruction.
7770 This calling convention is incompatible with the one normally
7771 used on Unix, so you cannot use it if you need to call libraries
7772 compiled with the Unix compiler.
7774 Also, you must provide function prototypes for all functions that
7775 take variable numbers of arguments (including @code{printf});
7776 otherwise incorrect code will be generated for calls to those
7779 In addition, seriously incorrect code will result if you call a
7780 function with too many arguments. (Normally, extra arguments are
7781 harmlessly ignored.)
7783 This option takes its name from the 680x0 @code{rtd} instruction.
7787 Use a different function-calling convention where the first two arguments
7788 are passed in registers.
7790 This calling convention is incompatible with the one normally
7791 used on Unix, so you cannot use it if you need to call libraries
7792 compiled with the Unix compiler.
7795 Do not pass any arguments in registers. This is the default for all
7799 It is OK to use the sb as an index register which is always loaded with
7800 zero. This is the default for the pc532-netbsd target.
7803 The sb register is not available for use or has not been initialized to
7804 zero by the run time system. This is the default for all targets except
7805 the pc532-netbsd. It is also implied whenever @samp{-mhimem} or
7806 @samp{-fpic} is set.
7809 Many ns32000 series addressing modes use displacements of up to 512MB.
7810 If an address is above 512MB then displacements from zero can not be used.
7811 This option causes code to be generated which can be loaded above 512MB.
7812 This may be useful for operating systems or ROM code.
7815 Assume code will be loaded in the first 512MB of virtual address space.
7816 This is the default for all platforms.
7822 @subsection AVR Options
7825 These options are defined for AVR implementations:
7828 @item -mmcu=@var{mcu}
7829 Specify ATMEL AVR instruction set or MCU type.
7831 Instruction set avr1 is for the minimal AVR core, not supported by the C
7832 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7833 attiny11, attiny12, attiny15, attiny28).
7835 Instruction set avr2 (default) is for the classic AVR core with up to
7836 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7837 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7838 at90c8534, at90s8535).
7840 Instruction set avr3 is for the classic AVR core with up to 128K program
7841 memory space (MCU types: atmega103, atmega603).
7843 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7844 memory space (MCU types: atmega83, atmega85).
7846 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7847 memory space (MCU types: atmega161, atmega163, atmega32, at94k).
7850 Output instruction sizes to the asm file.
7852 @item -minit-stack=@var{N}
7853 Specify the initial stack address, which may be a symbol or numeric value,
7854 __stack is the default.
7856 @item -mno-interrupts
7857 Generated code is not compatible with hardware interrupts.
7858 Code size will be smaller.
7860 @item -mcall-prologues
7861 Functions prologues/epilogues expanded as call to appropriate
7862 subroutines. Code size will be smaller.
7864 @item -mno-tablejump
7865 Do not generate tablejump insns which sometimes increase code size.
7868 Change only the low 8 bits of the stack pointer.
7872 @subsection MCore Options
7873 @cindex MCore options
7875 These are the @samp{-m} options defined for the Motorola M*Core
7883 Inline constants into the code stream if it can be done in two
7884 instructions or less.
7889 Use the divide instruction. (Enabled by default).
7891 @item -mrelax-immediate
7892 @itemx -mrelax-immediate
7893 @itemx -mno-relax-immediate
7894 Allow arbitrary sized immediates in bit operations.
7896 @item -mwide-bitfields
7897 @itemx -mwide-bitfields
7898 @itemx -mno-wide-bitfields
7899 Always treat bitfields as int-sized.
7901 @item -m4byte-functions
7902 @itemx -m4byte-functions
7903 @itemx -mno-4byte-functions
7904 Force all functions to be aligned to a four byte boundary.
7906 @item -mcallgraph-data
7907 @itemx -mcallgraph-data
7908 @itemx -mno-callgraph-data
7909 Emit callgraph information.
7913 @itemx -mno-slow-bytes
7914 Prefer word access when reading byte quantities.
7916 @item -mlittle-endian
7917 @itemx -mlittle-endian
7919 Generate code for a little endian target.
7924 Generate code for the 210 processor.
7928 @subsection IA-64 Options
7929 @cindex IA-64 Options
7931 These are the @samp{-m} options defined for the Intel IA-64 architecture.
7935 Generate code for a big endian target. This is the default for HPUX.
7937 @item -mlittle-endian
7938 Generate code for a little endian target. This is the default for AIX5
7943 Generate (or don't) code for the GNU assembler. This is the default.
7944 @c Also, this is the default if the configure option @samp{--with-gnu-as}
7949 Generate (or don't) code for the GNU linker. This is the default.
7950 @c Also, this is the default if the configure option @samp{--with-gnu-ld}
7954 Generate code that does not use a global pointer register. The result
7955 is not position independent code, and violates the IA-64 ABI.
7957 @item -mvolatile-asm-stop
7958 @itemx -mno-volatile-asm-stop
7959 Generate (or don't) a stop bit immediately before and after volatile asm
7963 Generate code that works around Itanium B step errata.
7965 @item -mregister-names
7966 @itemx -mno-register-names
7967 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
7968 the stacked registers. This may make assembler output more readable.
7972 Disable (or enable) optimizations that use the small data section. This may
7973 be useful for working around optimizer bugs.
7976 Generate code that uses a single constant global pointer value. This is
7977 useful when compiling kernel code.
7980 Generate code that is self-relocatable. This implies @samp{-mconstant-gp}.
7981 This is useful when compiling firmware code.
7983 @item -minline-divide-min-latency
7984 Generate code for inline divides using the minimum latency algorithm.
7986 @item -minline-divide-max-throughput
7987 Generate code for inline divides using the maximum throughput algorithm.
7989 @item -mno-dwarf2-asm
7991 Don't (or do) generate assembler code for the DWARF2 line number debugging
7992 info. This may be useful when not using the GNU assembler.
7994 @item -mfixed-range=@var{register range}
7995 Generate code treating the given register range as fixed registers.
7996 A fixed register is one that the register allocator can not use. This is
7997 useful when compiling kernel code. A register range is specified as
7998 two registers separated by a dash. Multiple register ranges can be
7999 specified separated by a comma.
8003 @subsection D30V Options
8004 @cindex D30V Options
8006 These @samp{-m} options are defined for D30V implementations:
8010 Link the @samp{.text}, @samp{.data}, @samp{.bss}, @samp{.strings},
8011 @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections into external
8012 memory, which starts at location @code{0x80000000}.
8015 Same as the @samp{-mextmem} switch.
8018 Link the @samp{.text} section into onchip text memory, which starts at
8019 location @code{0x0}. Also link @samp{.data}, @samp{.bss},
8020 @samp{.strings}, @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections
8021 into onchip data memory, which starts at location @code{0x20000000}.
8023 @item -mno-asm-optimize
8024 @itemx -masm-optimize
8025 Disable (enable) passing @samp{-O} to the assembler when optimizing.
8026 The assembler uses the @samp{-O} option to automatically parallelize
8027 adjacent short instructions where possible.
8029 @item -mbranch-cost=@var{n}
8030 Increase the internal costs of branches to @var{n}. Higher costs means
8031 that the compiler will issue more instructions to avoid doing a branch.
8034 @item -mcond-exec=@var{n}
8035 Specify the maximum number of conditionally executed instructions that
8036 replace a branch. The default is 4.
8039 @node Code Gen Options
8040 @section Options for Code Generation Conventions
8041 @cindex code generation conventions
8042 @cindex options, code generation
8043 @cindex run-time options
8045 These machine-independent options control the interface conventions
8046 used in code generation.
8048 Most of them have both positive and negative forms; the negative form
8049 of @samp{-ffoo} would be @samp{-fno-foo}. In the table below, only
8050 one of the forms is listed---the one which is not the default. You
8051 can figure out the other form by either removing @samp{no-} or adding
8056 Enable exception handling. Generates extra code needed to propagate
8057 exceptions. For some targets, this implies GNU CC will generate frame
8058 unwind information for all functions, which can produce significant data
8059 size overhead, although it does not affect execution. If you do not
8060 specify this option, GNU CC will enable it by default for languages like
8061 C++ which normally require exception handling, and disable it for
8062 languages like C that do not normally require it. However, you may need
8063 to enable this option when compiling C code that needs to interoperate
8064 properly with exception handlers written in C++. You may also wish to
8065 disable this option if you are compiling older C++ programs that don't
8066 use exception handling.
8068 @item -funwind-tables
8069 Similar to @option{-fexceptions}, except that it will just generate any needed
8070 static data, but will not affect the generated code in any other way.
8071 You will normally not enable this option; instead, a language processor
8072 that needs this handling would enable it on your behalf.
8074 @item -fpcc-struct-return
8075 Return ``short'' @code{struct} and @code{union} values in memory like
8076 longer ones, rather than in registers. This convention is less
8077 efficient, but it has the advantage of allowing intercallability between
8078 GCC-compiled files and files compiled with other compilers.
8080 The precise convention for returning structures in memory depends
8081 on the target configuration macros.
8083 Short structures and unions are those whose size and alignment match
8084 that of some integer type.
8086 @item -freg-struct-return
8087 Use the convention that @code{struct} and @code{union} values are
8088 returned in registers when possible. This is more efficient for small
8089 structures than @samp{-fpcc-struct-return}.
8091 If you specify neither @samp{-fpcc-struct-return} nor its contrary
8092 @samp{-freg-struct-return}, GCC defaults to whichever convention is
8093 standard for the target. If there is no standard convention, GCC
8094 defaults to @samp{-fpcc-struct-return}, except on targets where GCC
8095 is the principal compiler. In those cases, we can choose the standard,
8096 and we chose the more efficient register return alternative.
8099 Allocate to an @code{enum} type only as many bytes as it needs for the
8100 declared range of possible values. Specifically, the @code{enum} type
8101 will be equivalent to the smallest integer type which has enough room.
8103 @item -fshort-double
8104 Use the same size for @code{double} as for @code{float}.
8107 Requests that the data and non-@code{const} variables of this
8108 compilation be shared data rather than private data. The distinction
8109 makes sense only on certain operating systems, where shared data is
8110 shared between processes running the same program, while private data
8111 exists in one copy per process.
8114 In C, allocate even uninitialized global variables in the data section of the
8115 object file, rather than generating them as common blocks. This has the
8116 effect that if the same variable is declared (without @code{extern}) in
8117 two different compilations, you will get an error when you link them.
8118 The only reason this might be useful is if you wish to verify that the
8119 program will work on other systems which always work this way.
8122 Ignore the @samp{#ident} directive.
8124 @item -fno-gnu-linker
8125 Do not output global initializations (such as C++ constructors and
8126 destructors) in the form used by the GNU linker (on systems where the GNU
8127 linker is the standard method of handling them). Use this option when
8128 you want to use a non-GNU linker, which also requires using the
8129 @command{collect2} program to make sure the system linker includes
8130 constructors and destructors. (@command{collect2} is included in the GCC
8131 distribution.) For systems which @emph{must} use @command{collect2}, the
8132 compiler driver @command{gcc} is configured to do this automatically.
8134 @item -finhibit-size-directive
8135 Don't output a @code{.size} assembler directive, or anything else that
8136 would cause trouble if the function is split in the middle, and the
8137 two halves are placed at locations far apart in memory. This option is
8138 used when compiling @file{crtstuff.c}; you should not need to use it
8142 Put extra commentary information in the generated assembly code to
8143 make it more readable. This option is generally only of use to those
8144 who actually need to read the generated assembly code (perhaps while
8145 debugging the compiler itself).
8147 @samp{-fno-verbose-asm}, the default, causes the
8148 extra information to be omitted and is useful when comparing two assembler
8152 Consider all memory references through pointers to be volatile.
8154 @item -fvolatile-global
8155 Consider all memory references to extern and global data items to
8156 be volatile. GCC does not consider static data items to be volatile
8157 because of this switch.
8159 @item -fvolatile-static
8160 Consider all memory references to static data to be volatile.
8163 @cindex global offset table
8165 Generate position-independent code (PIC) suitable for use in a shared
8166 library, if supported for the target machine. Such code accesses all
8167 constant addresses through a global offset table (GOT). The dynamic
8168 loader resolves the GOT entries when the program starts (the dynamic
8169 loader is not part of GCC; it is part of the operating system). If
8170 the GOT size for the linked executable exceeds a machine-specific
8171 maximum size, you get an error message from the linker indicating that
8172 @samp{-fpic} does not work; in that case, recompile with @samp{-fPIC}
8173 instead. (These maximums are 16k on the m88k, 8k on the Sparc, and 32k
8174 on the m68k and RS/6000. The 386 has no such limit.)
8176 Position-independent code requires special support, and therefore works
8177 only on certain machines. For the 386, GCC supports PIC for System V
8178 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
8179 position-independent.
8182 If supported for the target machine, emit position-independent code,
8183 suitable for dynamic linking and avoiding any limit on the size of the
8184 global offset table. This option makes a difference on the m68k, m88k,
8187 Position-independent code requires special support, and therefore works
8188 only on certain machines.
8190 @item -ffixed-@var{reg}
8191 Treat the register named @var{reg} as a fixed register; generated code
8192 should never refer to it (except perhaps as a stack pointer, frame
8193 pointer or in some other fixed role).
8195 @var{reg} must be the name of a register. The register names accepted
8196 are machine-specific and are defined in the @code{REGISTER_NAMES}
8197 macro in the machine description macro file.
8199 This flag does not have a negative form, because it specifies a
8202 @item -fcall-used-@var{reg}
8203 Treat the register named @var{reg} as an allocable register that is
8204 clobbered by function calls. It may be allocated for temporaries or
8205 variables that do not live across a call. Functions compiled this way
8206 will not save and restore the register @var{reg}.
8208 It is an error to used this flag with the frame pointer or stack pointer.
8209 Use of this flag for other registers that have fixed pervasive roles in
8210 the machine's execution model will produce disastrous results.
8212 This flag does not have a negative form, because it specifies a
8215 @item -fcall-saved-@var{reg}
8216 Treat the register named @var{reg} as an allocable register saved by
8217 functions. It may be allocated even for temporaries or variables that
8218 live across a call. Functions compiled this way will save and restore
8219 the register @var{reg} if they use it.
8221 It is an error to used this flag with the frame pointer or stack pointer.
8222 Use of this flag for other registers that have fixed pervasive roles in
8223 the machine's execution model will produce disastrous results.
8225 A different sort of disaster will result from the use of this flag for
8226 a register in which function values may be returned.
8228 This flag does not have a negative form, because it specifies a
8232 Pack all structure members together without holes. Usually you would
8233 not want to use this option, since it makes the code suboptimal, and
8234 the offsets of structure members won't agree with system libraries.
8236 @item -fcheck-memory-usage
8237 Generate extra code to check each memory access. GCC will generate
8238 code that is suitable for a detector of bad memory accesses such as
8241 Normally, you should compile all, or none, of your code with this option.
8243 If you do mix code compiled with and without this option,
8244 you must ensure that all code that has side effects
8245 and that is called by code compiled with this option
8246 is, itself, compiled with this option.
8247 If you do not, you might get erroneous messages from the detector.
8249 If you use functions from a library that have side-effects (such as
8250 @code{read}), you might not be able to recompile the library and
8251 specify this option. In that case, you can enable the
8252 @samp{-fprefix-function-name} option, which requests GCC to encapsulate
8253 your code and make other functions look as if they were compiled with
8254 @samp{-fcheck-memory-usage}. This is done by calling ``stubs'',
8255 which are provided by the detector. If you cannot find or build
8256 stubs for every function you call, you might have to specify
8257 @samp{-fcheck-memory-usage} without @samp{-fprefix-function-name}.
8259 If you specify this option, you can not use the @code{asm} or
8260 @code{__asm__} keywords in functions with memory checking enabled. GNU
8261 CC cannot understand what the @code{asm} statement may do, and therefore
8262 cannot generate the appropriate code, so it will reject it. However, if
8263 you specify the function attribute @code{no_check_memory_usage}
8264 (@pxref{Function Attributes}), GNU CC will disable memory checking within a
8265 function; you may use @code{asm} statements inside such functions. You
8266 may have an inline expansion of a non-checked function within a checked
8267 function; in that case GNU CC will not generate checks for the inlined
8268 function's memory accesses.
8270 If you move your @code{asm} statements to non-checked inline functions
8271 and they do access memory, you can add calls to the support code in your
8272 inline function, to indicate any reads, writes, or copies being done.
8273 These calls would be similar to those done in the stubs described above.
8275 @item -fprefix-function-name
8276 Request GCC to add a prefix to the symbols generated for function names.
8277 GCC adds a prefix to the names of functions defined as well as
8278 functions called. Code compiled with this option and code compiled
8279 without the option can't be linked together, unless stubs are used.
8281 If you compile the following code with @samp{-fprefix-function-name}
8283 extern void bar (int);
8292 GCC will compile the code as if it was written:
8294 extern void prefix_bar (int);
8298 return prefix_bar (a + 5);
8301 This option is designed to be used with @samp{-fcheck-memory-usage}.
8303 @item -finstrument-functions
8304 Generate instrumentation calls for entry and exit to functions. Just
8305 after function entry and just before function exit, the following
8306 profiling functions will be called with the address of the current
8307 function and its call site. (On some platforms,
8308 @code{__builtin_return_address} does not work beyond the current
8309 function, so the call site information may not be available to the
8310 profiling functions otherwise.)
8313 void __cyg_profile_func_enter (void *this_fn, void *call_site);
8314 void __cyg_profile_func_exit (void *this_fn, void *call_site);
8317 The first argument is the address of the start of the current function,
8318 which may be looked up exactly in the symbol table.
8320 This instrumentation is also done for functions expanded inline in other
8321 functions. The profiling calls will indicate where, conceptually, the
8322 inline function is entered and exited. This means that addressable
8323 versions of such functions must be available. If all your uses of a
8324 function are expanded inline, this may mean an additional expansion of
8325 code size. If you use @samp{extern inline} in your C code, an
8326 addressable version of such functions must be provided. (This is
8327 normally the case anyways, but if you get lucky and the optimizer always
8328 expands the functions inline, you might have gotten away without
8329 providing static copies.)
8331 A function may be given the attribute @code{no_instrument_function}, in
8332 which case this instrumentation will not be done. This can be used, for
8333 example, for the profiling functions listed above, high-priority
8334 interrupt routines, and any functions from which the profiling functions
8335 cannot safely be called (perhaps signal handlers, if the profiling
8336 routines generate output or allocate memory).
8339 Generate code to verify that you do not go beyond the boundary of the
8340 stack. You should specify this flag if you are running in an
8341 environment with multiple threads, but only rarely need to specify it in
8342 a single-threaded environment since stack overflow is automatically
8343 detected on nearly all systems if there is only one stack.
8345 Note that this switch does not actually cause checking to be done; the
8346 operating system must do that. The switch causes generation of code
8347 to ensure that the operating system sees the stack being extended.
8349 @item -fstack-limit-register=@var{reg}
8350 @itemx -fstack-limit-symbol=@var{sym}
8351 @itemx -fno-stack-limit
8352 Generate code to ensure that the stack does not grow beyond a certain value,
8353 either the value of a register or the address of a symbol. If the stack
8354 would grow beyond the value, a signal is raised. For most targets,
8355 the signal is raised before the stack overruns the boundary, so
8356 it is possible to catch the signal without taking special precautions.
8358 For instance, if the stack starts at address @samp{0x80000000} and grows
8359 downwards you can use the flags
8360 @samp{-fstack-limit-symbol=__stack_limit}
8361 @samp{-Wl,--defsym,__stack_limit=0x7ffe0000} which will enforce a stack
8364 @cindex aliasing of parameters
8365 @cindex parameters, aliased
8366 @item -fargument-alias
8367 @itemx -fargument-noalias
8368 @itemx -fargument-noalias-global
8369 Specify the possible relationships among parameters and between
8370 parameters and global data.
8372 @samp{-fargument-alias} specifies that arguments (parameters) may
8373 alias each other and may alias global storage.
8374 @samp{-fargument-noalias} specifies that arguments do not alias
8375 each other, but may alias global storage.
8376 @samp{-fargument-noalias-global} specifies that arguments do not
8377 alias each other and do not alias global storage.
8379 Each language will automatically use whatever option is required by
8380 the language standard. You should not need to use these options yourself.
8382 @item -fleading-underscore
8383 This option and its counterpart, -fno-leading-underscore, forcibly
8384 change the way C symbols are represented in the object file. One use
8385 is to help link with legacy assembly code.
8387 Be warned that you should know what you are doing when invoking this
8388 option, and that not all targets provide complete support for it.
8393 @node Environment Variables
8394 @section Environment Variables Affecting GCC
8395 @cindex environment variables
8397 @c man begin ENVIRONMENT
8399 This section describes several environment variables that affect how GCC
8400 operates. Some of them work by specifying directories or prefixes to use
8401 when searching for various kinds of files. Some are used to specify other
8402 aspects of the compilation environment.
8405 Note that you can also specify places to search using options such as
8406 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
8407 take precedence over places specified using environment variables, which
8408 in turn take precedence over those specified by the configuration of GCC.
8412 Note that you can also specify places to search using options such as
8413 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
8414 take precedence over places specified using environment variables, which
8415 in turn take precedence over those specified by the configuration of GCC.
8422 @c @itemx LC_COLLATE
8424 @c @itemx LC_MONETARY
8425 @c @itemx LC_NUMERIC
8430 @c @findex LC_COLLATE
8432 @c @findex LC_MONETARY
8433 @c @findex LC_NUMERIC
8437 These environment variables control the way that GCC uses
8438 localization information that allow GCC to work with different
8439 national conventions. GCC inspects the locale categories
8440 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
8441 so. These locale categories can be set to any value supported by your
8442 installation. A typical value is @samp{en_UK} for English in the United
8445 The @env{LC_CTYPE} environment variable specifies character
8446 classification. GCC uses it to determine the character boundaries in
8447 a string; this is needed for some multibyte encodings that contain quote
8448 and escape characters that would otherwise be interpreted as a string
8451 The @env{LC_MESSAGES} environment variable specifies the language to
8452 use in diagnostic messages.
8454 If the @env{LC_ALL} environment variable is set, it overrides the value
8455 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
8456 and @env{LC_MESSAGES} default to the value of the @env{LANG}
8457 environment variable. If none of these variables are set, GCC
8458 defaults to traditional C English behavior.
8462 If @env{TMPDIR} is set, it specifies the directory to use for temporary
8463 files. GCC uses temporary files to hold the output of one stage of
8464 compilation which is to be used as input to the next stage: for example,
8465 the output of the preprocessor, which is the input to the compiler
8468 @item GCC_EXEC_PREFIX
8469 @findex GCC_EXEC_PREFIX
8470 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
8471 names of the subprograms executed by the compiler. No slash is added
8472 when this prefix is combined with the name of a subprogram, but you can
8473 specify a prefix that ends with a slash if you wish.
8475 If @env{GCC_EXEC_PREFIX} is not set, GNU CC will attempt to figure out
8476 an appropriate prefix to use based on the pathname it was invoked with.
8478 If GCC cannot find the subprogram using the specified prefix, it
8479 tries looking in the usual places for the subprogram.
8481 The default value of @env{GCC_EXEC_PREFIX} is
8482 @file{@var{prefix}/lib/gcc-lib/} where @var{prefix} is the value
8483 of @code{prefix} when you ran the @file{configure} script.
8485 Other prefixes specified with @samp{-B} take precedence over this prefix.
8487 This prefix is also used for finding files such as @file{crt0.o} that are
8490 In addition, the prefix is used in an unusual way in finding the
8491 directories to search for header files. For each of the standard
8492 directories whose name normally begins with @samp{/usr/local/lib/gcc-lib}
8493 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
8494 replacing that beginning with the specified prefix to produce an
8495 alternate directory name. Thus, with @samp{-Bfoo/}, GCC will search
8496 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
8497 These alternate directories are searched first; the standard directories
8501 @findex COMPILER_PATH
8502 The value of @env{COMPILER_PATH} is a colon-separated list of
8503 directories, much like @env{PATH}. GCC tries the directories thus
8504 specified when searching for subprograms, if it can't find the
8505 subprograms using @env{GCC_EXEC_PREFIX}.
8508 @findex LIBRARY_PATH
8509 The value of @env{LIBRARY_PATH} is a colon-separated list of
8510 directories, much like @env{PATH}. When configured as a native compiler,
8511 GCC tries the directories thus specified when searching for special
8512 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
8513 using GCC also uses these directories when searching for ordinary
8514 libraries for the @samp{-l} option (but directories specified with
8515 @samp{-L} come first).
8517 @item C_INCLUDE_PATH
8518 @itemx CPLUS_INCLUDE_PATH
8519 @itemx OBJC_INCLUDE_PATH
8520 @findex C_INCLUDE_PATH
8521 @findex CPLUS_INCLUDE_PATH
8522 @findex OBJC_INCLUDE_PATH
8523 @c @itemx OBJCPLUS_INCLUDE_PATH
8524 These environment variables pertain to particular languages. Each
8525 variable's value is a colon-separated list of directories, much like
8526 @env{PATH}. When GCC searches for header files, it tries the
8527 directories listed in the variable for the language you are using, after
8528 the directories specified with @samp{-I} but before the standard header
8531 @item DEPENDENCIES_OUTPUT
8532 @findex DEPENDENCIES_OUTPUT
8533 @cindex dependencies for make as output
8534 If this variable is set, its value specifies how to output dependencies
8535 for Make based on the header files processed by the compiler. This
8536 output looks much like the output from the @samp{-M} option
8537 (@pxref{Preprocessor Options}), but it goes to a separate file, and is
8538 in addition to the usual results of compilation.
8540 The value of @env{DEPENDENCIES_OUTPUT} can be just a file name, in
8541 which case the Make rules are written to that file, guessing the target
8542 name from the source file name. Or the value can have the form
8543 @samp{@var{file} @var{target}}, in which case the rules are written to
8544 file @var{file} using @var{target} as the target name.
8548 @cindex locale definition
8549 This variable is used to pass locale information to the compiler. One way in
8550 which this information is used is to determine the character set to be used
8551 when character literals, string literals and comments are parsed in C and C++.
8552 When the compiler is configured to allow multibyte characters,
8553 the following values for @env{LANG} are recognized:
8557 Recognize JIS characters.
8559 Recognize SJIS characters.
8561 Recognize EUCJP characters.
8564 If @env{LANG} is not defined, or if it has some other value, then the
8565 compiler will use mblen and mbtowc as defined by the default locale to
8566 recognize and translate multibyte characters.
8571 @node Running Protoize
8572 @section Running Protoize
8574 The program @code{protoize} is an optional part of GNU C. You can use
8575 it to add prototypes to a program, thus converting the program to ISO
8576 C in one respect. The companion program @code{unprotoize} does the
8577 reverse: it removes argument types from any prototypes that are found.
8579 When you run these programs, you must specify a set of source files as
8580 command line arguments. The conversion programs start out by compiling
8581 these files to see what functions they define. The information gathered
8582 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
8584 After scanning comes actual conversion. The specified files are all
8585 eligible to be converted; any files they include (whether sources or
8586 just headers) are eligible as well.
8588 But not all the eligible files are converted. By default,
8589 @code{protoize} and @code{unprotoize} convert only source and header
8590 files in the current directory. You can specify additional directories
8591 whose files should be converted with the @samp{-d @var{directory}}
8592 option. You can also specify particular files to exclude with the
8593 @samp{-x @var{file}} option. A file is converted if it is eligible, its
8594 directory name matches one of the specified directory names, and its
8595 name within the directory has not been excluded.
8597 Basic conversion with @code{protoize} consists of rewriting most
8598 function definitions and function declarations to specify the types of
8599 the arguments. The only ones not rewritten are those for varargs
8602 @code{protoize} optionally inserts prototype declarations at the
8603 beginning of the source file, to make them available for any calls that
8604 precede the function's definition. Or it can insert prototype
8605 declarations with block scope in the blocks where undeclared functions
8608 Basic conversion with @code{unprotoize} consists of rewriting most
8609 function declarations to remove any argument types, and rewriting
8610 function definitions to the old-style pre-ISO form.
8612 Both conversion programs print a warning for any function declaration or
8613 definition that they can't convert. You can suppress these warnings
8616 The output from @code{protoize} or @code{unprotoize} replaces the
8617 original source file. The original file is renamed to a name ending
8618 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
8619 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
8620 for DOS) file already exists, then the source file is simply discarded.
8622 @code{protoize} and @code{unprotoize} both depend on GCC itself to
8623 scan the program and collect information about the functions it uses.
8624 So neither of these programs will work until GCC is installed.
8626 Here is a table of the options you can use with @code{protoize} and
8627 @code{unprotoize}. Each option works with both programs unless
8631 @item -B @var{directory}
8632 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
8633 usual directory (normally @file{/usr/local/lib}). This file contains
8634 prototype information about standard system functions. This option
8635 applies only to @code{protoize}.
8637 @item -c @var{compilation-options}
8638 Use @var{compilation-options} as the options when running @code{gcc} to
8639 produce the @samp{.X} files. The special option @samp{-aux-info} is
8640 always passed in addition, to tell @code{gcc} to write a @samp{.X} file.
8642 Note that the compilation options must be given as a single argument to
8643 @code{protoize} or @code{unprotoize}. If you want to specify several
8644 @code{gcc} options, you must quote the entire set of compilation options
8645 to make them a single word in the shell.
8647 There are certain @code{gcc} arguments that you cannot use, because they
8648 would produce the wrong kind of output. These include @samp{-g},
8649 @samp{-O}, @samp{-c}, @samp{-S}, and @samp{-o} If you include these in
8650 the @var{compilation-options}, they are ignored.
8653 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
8654 systems) instead of @samp{.c}. This is convenient if you are converting
8655 a C program to C++. This option applies only to @code{protoize}.
8658 Add explicit global declarations. This means inserting explicit
8659 declarations at the beginning of each source file for each function
8660 that is called in the file and was not declared. These declarations
8661 precede the first function definition that contains a call to an
8662 undeclared function. This option applies only to @code{protoize}.
8664 @item -i @var{string}
8665 Indent old-style parameter declarations with the string @var{string}.
8666 This option applies only to @code{protoize}.
8668 @code{unprotoize} converts prototyped function definitions to old-style
8669 function definitions, where the arguments are declared between the
8670 argument list and the initial @samp{@{}. By default, @code{unprotoize}
8671 uses five spaces as the indentation. If you want to indent with just
8672 one space instead, use @samp{-i " "}.
8675 Keep the @samp{.X} files. Normally, they are deleted after conversion
8679 Add explicit local declarations. @code{protoize} with @samp{-l} inserts
8680 a prototype declaration for each function in each block which calls the
8681 function without any declaration. This option applies only to
8685 Make no real changes. This mode just prints information about the conversions
8686 that would have been done without @samp{-n}.
8689 Make no @samp{.save} files. The original files are simply deleted.
8690 Use this option with caution.
8692 @item -p @var{program}
8693 Use the program @var{program} as the compiler. Normally, the name
8697 Work quietly. Most warnings are suppressed.
8700 Print the version number, just like @samp{-v} for @code{gcc}.
8703 If you need special compiler options to compile one of your program's
8704 source files, then you should generate that file's @samp{.X} file
8705 specially, by running @code{gcc} on that source file with the
8706 appropriate options and the option @samp{-aux-info}. Then run
8707 @code{protoize} on the entire set of files. @code{protoize} will use
8708 the existing @samp{.X} file because it is newer than the source file.
8712 gcc -Dfoo=bar file1.c -aux-info
8717 You need to include the special files along with the rest in the
8718 @code{protoize} command, even though their @samp{.X} files already
8719 exist, because otherwise they won't get converted.
8721 @xref{Protoize Caveats}, for more information on how to use
8722 @code{protoize} successfully.