1 @c Copyright (C) 1988, 89, 92-98, 1999 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter GCC Command Options
7 @cindex GCC command options
8 @cindex command options
9 @cindex options, GCC command
11 When you invoke GCC, it normally does preprocessing, compilation,
12 assembly and linking. The ``overall options'' allow you to stop this
13 process at an intermediate stage. For example, the @samp{-c} option
14 says not to run the linker. Then the output consists of object files
15 output by the assembler.
17 Other options are passed on to one stage of processing. Some options
18 control the preprocessor and others the compiler itself. Yet other
19 options control the assembler and linker; most of these are not
20 documented here, since you rarely need to use any of them.
22 @cindex C compilation options
23 Most of the command line options that you can use with GCC are useful
24 for C programs; when an option is only useful with another language
25 (usually C++), the explanation says so explicitly. If the description
26 for a particular option does not mention a source language, you can use
27 that option with all supported languages.
29 @cindex C++ compilation options
30 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
31 options for compiling C++ programs.
33 @cindex grouping options
34 @cindex options, grouping
35 The @code{gcc} program accepts options and file names as operands. Many
36 options have multiletter names; therefore multiple single-letter options
37 may @emph{not} be grouped: @samp{-dr} is very different from @w{@samp{-d
40 @cindex order of options
41 @cindex options, order
42 You can mix options and other arguments. For the most part, the order
43 you use doesn't matter. Order does matter when you use several options
44 of the same kind; for example, if you specify @samp{-L} more than once,
45 the directories are searched in the order specified.
47 Many options have long names starting with @samp{-f} or with
48 @samp{-W}---for example, @samp{-fforce-mem},
49 @samp{-fstrength-reduce}, @samp{-Wformat} and so on. Most of
50 these have both positive and negative forms; the negative form of
51 @samp{-ffoo} would be @samp{-fno-foo}. This manual documents
52 only one of these two forms, whichever one is not the default.
55 * Option Summary:: Brief list of all options, without explanations.
56 * Overall Options:: Controlling the kind of output:
57 an executable, object files, assembler files,
58 or preprocessed source.
59 * Invoking G++:: Compiling C++ programs.
60 * C Dialect Options:: Controlling the variant of C language compiled.
61 * C++ Dialect Options:: Variations on C++.
62 * Warning Options:: How picky should the compiler be?
63 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
64 * Optimize Options:: How much optimization?
65 * Preprocessor Options:: Controlling header files and macro definitions.
66 Also, getting dependency information for Make.
67 * Assembler Options:: Passing options to the assembler.
68 * Link Options:: Specifying libraries and so on.
69 * Directory Options:: Where to find header files and libraries.
70 Where to find the compiler executable files.
71 * Spec Files:: How to pass switches to sub-processes.
72 * Target Options:: Running a cross-compiler, or an old version of GCC.
73 * Submodel Options:: Specifying minor hardware or convention variations,
74 such as 68010 vs 68020.
75 * Code Gen Options:: Specifying conventions for function calls, data layout
77 * Environment Variables:: Env vars that affect GCC.
78 * Running Protoize:: Automatically adding or removing function prototypes.
82 @section Option Summary
84 Here is a summary of all the options, grouped by type. Explanations are
85 in the following sections.
89 @xref{Overall Options,,Options Controlling the Kind of Output}.
91 -c -S -E -o @var{file} -pipe -pass-exit-codes -v --help -x @var{language}
94 @item C Language Options
95 @xref{C Dialect Options,,Options Controlling C Dialect}.
97 -ansi -flang-isoc9x -fallow-single-precision -fcond-mismatch -fno-asm
98 -fno-builtin -ffreestanding -fhosted -fsigned-bitfields -fsigned-char
99 -funsigned-bitfields -funsigned-char -fwritable-strings
100 -traditional -traditional-cpp -trigraphs
103 @item C++ Language Options
104 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
106 -fno-access-control -fcheck-new -fconserve-space -fdollars-in-identifiers
107 -fno-elide-constructors -fexternal-templates -ffor-scope
108 -fno-for-scope -fno-gnu-keywords -fguiding-decls
109 -fhonor-std -fhuge-objects -fno-implicit-templates -finit-priority
110 -fno-implement-inlines -fname-mangling-version-@var{n} -fno-default-inline
111 -foperator-names -fno-optional-diags -fpermissive -frepo -fstrict-prototype
112 -fsquangle -ftemplate-depth-@var{n} -fvtable-thunks
113 -nostdinc++ -Wctor-dtor-privacy -Wno-deprecated -Weffc++
114 -Wno-non-template-friend
115 -Wnon-virtual-dtor -Wold-style-cast -Woverloaded-virtual
116 -Wno-pmf-conversions -Wreorder -Wsign-promo -Wsynth
119 @item Warning Options
120 @xref{Warning Options,,Options to Request or Suppress Warnings}.
122 -fsyntax-only -pedantic -pedantic-errors
123 -w -W -Wall -Waggregate-return
124 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment
125 -Wconversion -Werror -Wformat
126 -Wid-clash-@var{len} -Wimplicit -Wimplicit-int
127 -Wimplicit-function-declaration -Wimport
128 -Werror-implicit-function-declaration -Wfloat-equal -Winline
129 -Wlarger-than-@var{len} -Wlong-long
130 -Wmain -Wmissing-declarations -Wmissing-noreturn
131 -Wmultichar -Wno-import
132 -Wparentheses -Wpointer-arith -Wredundant-decls
133 -Wreturn-type -Wshadow -Wsign-compare -Wswitch
134 -Wtrigraphs -Wundef -Wuninitialized -Wunknown-pragmas -Wunreachable-code
135 -Wunused -Wwrite-strings
138 @item C-only Warning Options
140 -Wbad-function-cast -Wmissing-prototypes -Wnested-externs
141 -Wstrict-prototypes -Wtraditional
144 @item Debugging Options
145 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
147 -a -ax -d@var{letters} -fdump-unnumbered -fdump-translation-unit-@var{file}
148 -fpretend-float -fprofile-arcs -ftest-coverage
149 -g -g@var{level} -gcoff -gdwarf -gdwarf-1 -gdwarf-1+ -gdwarf-2
150 -ggdb -gstabs -gstabs+ -gxcoff -gxcoff+
151 -p -pg -print-file-name=@var{library} -print-libgcc-file-name
152 -print-prog-name=@var{program} -print-search-dirs -save-temps -time
155 @item Optimization Options
156 @xref{Optimize Options,,Options that Control Optimization}.
158 -falign-functions=@var{n} -falign-labels=@var{n} -falign-loops=@var{n}
159 -falign-jumps=@var{n} -fbranch-probabilities
160 -fcaller-saves -fcse-follow-jumps -fcse-skip-blocks
161 -fdelayed-branch -fdelete-null-pointer-checks -fexpensive-optimizations
162 -ffast-math -ffloat-store -fforce-addr -fforce-mem -fno-math-errno
163 -fdata-sections -ffunction-sections -fgcse
164 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions
165 -fmove-all-movables -fno-default-inline -fno-defer-pop
166 -fno-function-cse -fno-inline -fno-peephole
167 -fomit-frame-pointer -foptimize-register-moves -fregmove
168 -frerun-cse-after-loop -frerun-loop-opt -freduce-all-givs
169 -fschedule-insns -fschedule-insns2 -fstrength-reduce
170 -fstrict-aliasing -fthread-jumps -funroll-all-loops
172 -O -O0 -O1 -O2 -O3 -Os
175 @item Preprocessor Options
176 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
178 -A@var{question}(@var{answer}) -C -dD -dM -dN
179 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H
181 -include @var{file} -imacros @var{file}
182 -iprefix @var{file} -iwithprefix @var{dir}
183 -iwithprefixbefore @var{dir} -isystem @var{dir} -isystem-c++ @var{dir}
184 -M -MD -MM -MMD -MG -nostdinc -P -trigraphs
185 -undef -U@var{macro} -Wp,@var{option}
188 @item Assembler Option
189 @xref{Assembler Options,,Passing Options to the Assembler}.
195 @xref{Link Options,,Options for Linking}.
197 @var{object-file-name} -l@var{library}
198 -nostartfiles -nodefaultlibs -nostdlib
199 -s -static -shared -symbolic
200 -Wl,@var{option} -Xlinker @var{option}
204 @item Directory Options
205 @xref{Directory Options,,Options for Directory Search}.
207 -B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}
211 @c I wrote this xref this way to avoid overfull hbox. -- rms
212 @xref{Target Options}.
214 -b @var{machine} -V @var{version}
217 @item Machine Dependent Options
218 @xref{Submodel Options,,Hardware Models and Configurations}.
220 @emph{M680x0 Options}
221 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040
222 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020
223 -mfpa -mnobitfield -mrtd -mshort -msoft-float -mpcrel
224 -malign-int -mstrict-align
231 -mtune=@var{cpu type}
232 -mcmodel=@var{code model}
234 -mapp-regs -mbroken-saverestore -mcypress
235 -mepilogue -mfaster-structs -mflat
236 -mfpu -mhard-float -mhard-quad-float
237 -mimpure-text -mlive-g0 -mno-app-regs
238 -mno-epilogue -mno-faster-structs -mno-flat -mno-fpu
239 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles
240 -msoft-float -msoft-quad-float -msparclite -mstack-bias
241 -msupersparc -munaligned-doubles -mv8
243 @emph{Convex Options}
244 -mc1 -mc2 -mc32 -mc34 -mc38
245 -margcount -mnoargcount
247 -mvolatile-cache -mvolatile-nocache
249 @emph{AMD29K Options}
250 -m29000 -m29050 -mbw -mnbw -mdw -mndw
251 -mlarge -mnormal -msmall
252 -mkernel-registers -mno-reuse-arg-regs
253 -mno-stack-check -mno-storem-bug
254 -mreuse-arg-regs -msoft-float -mstack-check
255 -mstorem-bug -muser-registers
258 -mapcs-frame -mno-apcs-frame
260 -mapcs-stack-check -mno-apcs-stack-check
261 -mapcs-float -mno-apcs-float
262 -mapcs-reentrant -mno-apcs-reentrant
263 -msched-prolog -mno-sched-prolog
264 -mlittle-endian -mbig-endian -mwords-little-endian
265 -malignment-traps -mno-alignment-traps
266 -msoft-float -mhard-float -mfpe
267 -mthumb-interwork -mno-thumb-interwork
268 -mcpu= -march= -mfpe=
269 -mstructure-size-boundary=
270 -mbsd -mxopen -mno-symrename
272 -mnop-fun-dllimport -mno-nop-fun-dllimport
273 -msingle-pic-base -mno-single-pic-base
277 -mtpcs-frame -mno-tpcs-frame
278 -mtpcs-leaf-frame -mno-tpcs-leaf-frame
279 -mlittle-endian -mbig-endian
280 -mthumb-interwork -mno-thumb-interwork
281 -mstructure-size-boundary=
282 -mnop-fun-dllimport -mno-nop-fun-dllimport
283 -mcallee-super-interworking -mno-callee-super-interworking
284 -mcaller-super-interworking -mno-caller-super-interworking
285 -msingle-pic-base -mno-single-pic-base
288 @emph{MN10200 Options}
291 @emph{MN10300 Options}
296 @emph{M32R/D Options}
297 -mcode-model=@var{model type} -msdata=@var{sdata type}
301 -m88000 -m88100 -m88110 -mbig-pic
302 -mcheck-zero-division -mhandle-large-shift
303 -midentify-revision -mno-check-zero-division
304 -mno-ocs-debug-info -mno-ocs-frame-position
305 -mno-optimize-arg-area -mno-serialize-volatile
306 -mno-underscores -mocs-debug-info
307 -mocs-frame-position -moptimize-arg-area
308 -mserialize-volatile -mshort-data-@var{num} -msvr3
309 -msvr4 -mtrap-large-shift -muse-div-instruction
310 -mversion-03.00 -mwarn-passed-structs
312 @emph{RS/6000 and PowerPC Options}
314 -mtune=@var{cpu type}
315 -mpower -mno-power -mpower2 -mno-power2
316 -mpowerpc -mpowerpc64 -mno-powerpc
317 -mpowerpc-gpopt -mno-powerpc-gpopt
318 -mpowerpc-gfxopt -mno-powerpc-gfxopt
319 -mnew-mnemonics -mno-new-mnemonics
320 -mfull-toc -mminimal-toc -mno-fop-in-toc -mno-sum-in-toc
321 -m64 -m32 -mxl-call -mno-xl-call -mthreads -mpe
322 -msoft-float -mhard-float -mmultiple -mno-multiple
323 -mstring -mno-string -mupdate -mno-update
324 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align
325 -mstrict-align -mno-strict-align -mrelocatable
326 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib
327 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian
328 -mcall-aix -mcall-sysv -mprototype -mno-prototype
329 -msim -mmvme -mads -myellowknife -memb -msdata
330 -msdata=@var{opt} -G @var{num}
333 -mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs
334 -mfull-fp-blocks -mhc-struct-return -min-line-mul
335 -mminimum-fp-blocks -mnohc-struct-return
338 -mabicalls -mcpu=@var{cpu type} -membedded-data -muninit-const-in-rodata
339 -membedded-pic -mfp32 -mfp64 -mgas -mgp32 -mgp64
340 -mgpopt -mhalf-pic -mhard-float -mint64 -mips1
341 -mips2 -mips3 -mips4 -mlong64 -mlong32 -mlong-calls -mmemcpy
342 -mmips-as -mmips-tfile -mno-abicalls
343 -mno-embedded-data -mno-uninit-const-in-rodata -mno-embedded-pic
344 -mno-gpopt -mno-long-calls
345 -mno-memcpy -mno-mips-tfile -mno-rnames -mno-stats
346 -mrnames -msoft-float
347 -m4650 -msingle-float -mmad
348 -mstats -EL -EB -G @var{num} -nocpp
349 -mabi=32 -mabi=n32 -mabi=64 -mabi=eabi
353 -march=@var{cpu type}
354 -mieee-fp -mno-fancy-math-387
355 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib
356 -mno-wide-multiply -mrtd -malign-double
357 -mreg-alloc=@var{list} -mregparm=@var{num}
358 -malign-jumps=@var{num} -malign-loops=@var{num}
359 -malign-functions=@var{num} -mpreferred-stack-boundary=@var{num}
362 -march=@var{architecture type}
363 -mbig-switch -mdisable-fpregs -mdisable-indexing
364 -mfast-indirect-calls -mgas -mjump-in-delay
365 -mlong-load-store -mno-big-switch -mno-disable-fpregs
366 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas
367 -mno-jump-in-delay -mno-long-load-store
368 -mno-portable-runtime -mno-soft-float
369 -mno-space-regs -msoft-float -mpa-risc-1-0
370 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime
371 -mschedule=@var{cpu type} -mspace-regs
373 @emph{Intel 960 Options}
374 -m@var{cpu type} -masm-compat -mclean-linkage
375 -mcode-align -mcomplex-addr -mleaf-procedures
376 -mic-compat -mic2.0-compat -mic3.0-compat
377 -mintel-asm -mno-clean-linkage -mno-code-align
378 -mno-complex-addr -mno-leaf-procedures
379 -mno-old-align -mno-strict-align -mno-tail-call
380 -mnumerics -mold-align -msoft-float -mstrict-align
383 @emph{DEC Alpha Options}
384 -mfp-regs -mno-fp-regs -mno-soft-float -msoft-float
386 -mieee -mieee-with-inexact -mieee-conformant
387 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode}
388 -mtrap-precision=@var{mode} -mbuild-constants
390 -mbwx -mno-bwx -mcix -mno-cix -mmax -mno-max
391 -mmemory-latency=@var{time}
393 @emph{Clipper Options}
396 @emph{H8/300 Options}
397 -mrelax -mh -ms -mint32 -malign-300
400 -m1 -m2 -m3 -m3e -mb -ml -mdalign -mrelax
402 @emph{System V Options}
403 -Qy -Qn -YP,@var{paths} -Ym,@var{dir}
407 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text section}
408 -mdata=@var{data section} -mrodata=@var{readonly data section}
410 @emph{TMS320C3x/C4x Options}
411 -mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm
412 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload
413 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned
414 -mparallel-insns -mparallel-mpy -mpreserve-float
417 -mlong-calls -mno-long-calls -mep -mno-ep
418 -mprolog-function -mno-prolog-function -mspace
419 -mtda=@var{n} -msda=@var{n} -mzda=@var{n}
423 -m32032 -m32332 -m32532 -m32081 -m32381 -mmult-add -mnomult-add
424 -msoft-float -mrtd -mnortd -mregparam -mnoregparam -msb -mnosb
425 -mbitfield -mnobitfield -mhimem -mnohimem
428 @item Code Generation Options
429 @xref{Code Gen Options,,Options for Code Generation Conventions}.
431 -fcall-saved-@var{reg} -fcall-used-@var{reg}
432 -fexceptions -funwind-tables -ffixed-@var{reg} -finhibit-size-directive
433 -fcheck-memory-usage -fprefix-function-name
434 -fno-common -fno-ident -fno-gnu-linker
435 -fpcc-struct-return -fpic -fPIC
436 -freg-struct-return -fshared-data -fshort-enums
437 -fshort-double -fvolatile -fvolatile-global -fvolatile-static
438 -fverbose-asm -fpack-struct -fstack-check
439 -fargument-alias -fargument-noalias
440 -fargument-noalias-global
446 * Overall Options:: Controlling the kind of output:
447 an executable, object files, assembler files,
448 or preprocessed source.
449 * C Dialect Options:: Controlling the variant of C language compiled.
450 * C++ Dialect Options:: Variations on C++.
451 * Warning Options:: How picky should the compiler be?
452 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
453 * Optimize Options:: How much optimization?
454 * Preprocessor Options:: Controlling header files and macro definitions.
455 Also, getting dependency information for Make.
456 * Assembler Options:: Passing options to the assembler.
457 * Link Options:: Specifying libraries and so on.
458 * Directory Options:: Where to find header files and libraries.
459 Where to find the compiler executable files.
460 * Spec Files:: How to pass switches to sub-processes.
461 * Target Options:: Running a cross-compiler, or an old version of GCC.
464 @node Overall Options
465 @section Options Controlling the Kind of Output
467 Compilation can involve up to four stages: preprocessing, compilation
468 proper, assembly and linking, always in that order. The first three
469 stages apply to an individual source file, and end by producing an
470 object file; linking combines all the object files (those newly
471 compiled, and those specified as input) into an executable file.
473 @cindex file name suffix
474 For any given input file, the file name suffix determines what kind of
479 C source code which must be preprocessed.
482 C source code which should not be preprocessed.
485 C++ source code which should not be preprocessed.
488 Objective-C source code. Note that you must link with the library
489 @file{libobjc.a} to make an Objective-C program work.
492 C header file (not to be compiled or linked).
495 @itemx @var{file}.cxx
496 @itemx @var{file}.cpp
498 C++ source code which must be preprocessed. Note that in @samp{.cxx},
499 the last two letters must both be literally @samp{x}. Likewise,
500 @samp{.C} refers to a literal capital C.
506 Assembler code which must be preprocessed.
509 An object file to be fed straight into linking.
510 Any file name with no recognized suffix is treated this way.
513 You can specify the input language explicitly with the @samp{-x} option:
516 @item -x @var{language}
517 Specify explicitly the @var{language} for the following input files
518 (rather than letting the compiler choose a default based on the file
519 name suffix). This option applies to all following input files until
520 the next @samp{-x} option. Possible values for @var{language} are:
523 c-header cpp-output c++-cpp-output
524 assembler assembler-with-cpp
528 Turn off any specification of a language, so that subsequent files are
529 handled according to their file name suffixes (as they are if @samp{-x}
530 has not been used at all).
532 @item -pass-exit-codes
533 Normally the @code{gcc} program will exit with the code of 1 if any
534 phase of the compiler returns a non-success return code. If you specify
535 @samp{-pass-exit-codes}, the @code{gcc} program will instead return with
536 numerically highest error produced by any phase that returned an error
540 If you only want some of the stages of compilation, you can use
541 @samp{-x} (or filename suffixes) to tell @code{gcc} where to start, and
542 one of the options @samp{-c}, @samp{-S}, or @samp{-E} to say where
543 @code{gcc} is to stop. Note that some combinations (for example,
544 @samp{-x cpp-output -E} instruct @code{gcc} to do nothing at all.
548 Compile or assemble the source files, but do not link. The linking
549 stage simply is not done. The ultimate output is in the form of an
550 object file for each source file.
552 By default, the object file name for a source file is made by replacing
553 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
555 Unrecognized input files, not requiring compilation or assembly, are
559 Stop after the stage of compilation proper; do not assemble. The output
560 is in the form of an assembler code file for each non-assembler input
563 By default, the assembler file name for a source file is made by
564 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
566 Input files that don't require compilation are ignored.
569 Stop after the preprocessing stage; do not run the compiler proper. The
570 output is in the form of preprocessed source code, which is sent to the
573 Input files which don't require preprocessing are ignored.
575 @cindex output file option
577 Place output in file @var{file}. This applies regardless to whatever
578 sort of output is being produced, whether it be an executable file,
579 an object file, an assembler file or preprocessed C code.
581 Since only one output file can be specified, it does not make sense to
582 use @samp{-o} when compiling more than one input file, unless you are
583 producing an executable file as output.
585 If @samp{-o} is not specified, the default is to put an executable file
586 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
587 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
588 all preprocessed C source on standard output.@refill
591 Print (on standard error output) the commands executed to run the stages
592 of compilation. Also print the version number of the compiler driver
593 program and of the preprocessor and the compiler proper.
596 Use pipes rather than temporary files for communication between the
597 various stages of compilation. This fails to work on some systems where
598 the assembler is unable to read from a pipe; but the GNU assembler has
602 Print (on the standard output) a description of the command line options
603 understood by @code{gcc}. If the @code{-v} option is also specified
604 then @code{--help} will also be passed on to the various processes
605 invoked by @code{gcc}, so that they can display the command line options
606 they accept. If the @code{-W} option is also specified then command
607 line options which have no documentation associated with them will also
612 @section Compiling C++ Programs
614 @cindex suffixes for C++ source
615 @cindex C++ source file suffixes
616 C++ source files conventionally use one of the suffixes @samp{.C},
617 @samp{.cc}, @samp{.cpp}, @samp{.c++}, @samp{.cp}, or @samp{.cxx};
618 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
619 files with these names and compiles them as C++ programs even if you
620 call the compiler the same way as for compiling C programs (usually with
621 the name @code{gcc}).
625 However, C++ programs often require class libraries as well as a
626 compiler that understands the C++ language---and under some
627 circumstances, you might want to compile programs from standard input,
628 or otherwise without a suffix that flags them as C++ programs.
629 @code{g++} is a program that calls GCC with the default language
630 set to C++, and automatically specifies linking against the C++
631 library. On many systems, the script @code{g++} is also
632 installed with the name @code{c++}.
634 @cindex invoking @code{g++}
635 When you compile C++ programs, you may specify many of the same
636 command-line options that you use for compiling programs in any
637 language; or command-line options meaningful for C and related
638 languages; or options that are meaningful only for C++ programs.
639 @xref{C Dialect Options,,Options Controlling C Dialect}, for
640 explanations of options for languages related to C.
641 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
642 explanations of options that are meaningful only for C++ programs.
644 @node C Dialect Options
645 @section Options Controlling C Dialect
646 @cindex dialect options
647 @cindex language dialect options
648 @cindex options, dialect
650 The following options control the dialect of C (or languages derived
651 from C, such as C++ and Objective C) that the compiler accepts:
656 In C mode, support all ANSI standard C programs. In C++ mode,
657 remove GNU extensions that conflict with ANSI C++.
658 @c shouldn't we be saying "ISO"?
660 This turns off certain features of GCC that are incompatible with ANSI
661 C (when compiling C code), or of ANSI standard C++ (when compiling C++ code),
662 such as the @code{asm} and @code{typeof} keywords, and
663 predefined macros such as @code{unix} and @code{vax} that identify the
664 type of system you are using. It also enables the undesirable and
665 rarely used ANSI trigraph feature. For the C compiler,
666 it disables recognition of C++ style @samp{//} comments as well as
667 the @code{inline} keyword. For the C++ compiler,
668 @samp{-foperator-names} is enabled as well.
671 The alternate keywords @code{__asm__}, @code{__extension__},
672 @code{__inline__} and @code{__typeof__} continue to work despite
673 @samp{-ansi}. You would not want to use them in an ANSI C program, of
674 course, but it is useful to put them in header files that might be included
675 in compilations done with @samp{-ansi}. Alternate predefined macros
676 such as @code{__unix__} and @code{__vax__} are also available, with or
677 without @samp{-ansi}.
679 The @samp{-ansi} option does not cause non-ANSI programs to be
680 rejected gratuitously. For that, @samp{-pedantic} is required in
681 addition to @samp{-ansi}. @xref{Warning Options}.
683 The macro @code{__STRICT_ANSI__} is predefined when the @samp{-ansi}
684 option is used. Some header files may notice this macro and refrain
685 from declaring certain functions or defining certain macros that the
686 ANSI standard doesn't call for; this is to avoid interfering with any
687 programs that might use these names for other things.
689 The functions @code{alloca}, @code{abort}, @code{exit}, and
690 @code{_exit} are not builtin functions when @samp{-ansi} is used.
693 Enable support for features found in the C9X standard. In particular,
694 enable support for the C9X @code{restrict} keyword.
696 Even when this option is not specified, you can still use some C9X
697 features in so far as they do not conflict with previous C standards.
698 For example, you may use @code{__restrict__} even when -flang-isoc9x
702 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
703 keyword, so that code can use these words as identifiers. You can use
704 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
705 instead. @samp{-ansi} implies @samp{-fno-asm}.
707 In C++, this switch only affects the @code{typeof} keyword, since
708 @code{asm} and @code{inline} are standard keywords. You may want to
709 use the @samp{-fno-gnu-keywords} flag instead, as it also disables the
710 other, C++-specific, extension keywords such as @code{headof}.
713 @cindex builtin functions
739 Don't recognize builtin functions that do not begin with @samp{__builtin_}
740 as prefix. Currently, the functions affected include @code{abort},
741 @code{abs}, @code{alloca}, @code{cos}, @code{cosf}, @code{cosl},
742 @code{exit}, @code{_exit}, @code{fabs}, @code{fabsf}, @code{fabsl},
743 @code{ffs}, @code{labs}, @code{memcmp}, @code{memcpy}, @code{memset},
744 @code{sin}, @code{sinf}, @code{sinl}, @code{sqrt}, @code{sqrtf},
745 @code{sqrtl}, @code{strcmp}, @code{strcpy}, and @code{strlen}.
747 GCC normally generates special code to handle certain builtin functions
748 more efficiently; for instance, calls to @code{alloca} may become single
749 instructions that adjust the stack directly, and calls to @code{memcpy}
750 may become inline copy loops. The resulting code is often both smaller
751 and faster, but since the function calls no longer appear as such, you
752 cannot set a breakpoint on those calls, nor can you change the behavior
753 of the functions by linking with a different library.
755 The @samp{-ansi} option prevents @code{alloca}, @code{ffs} and @code{_exit}
756 from being builtin functions, since these functions do not have an ANSI
760 @cindex hosted environment
762 Assert that compilation takes place in a hosted environment. This implies
763 @samp{-fbuiltin}. A hosted environment is one in which the
764 entire standard library is available, and in which @code{main} has a return
765 type of @code{int}. Examples are nearly everything except a kernel.
766 This is equivalent to @samp{-fno-freestanding}.
769 @cindex hosted environment
771 Assert that compilation takes place in a freestanding environment. This
772 implies @samp{-fno-builtin}. A freestanding environment
773 is one in which the standard library may not exist, and program startup may
774 not necessarily be at @code{main}. The most obvious example is an OS kernel.
775 This is equivalent to @samp{-fno-hosted}.
778 Support ANSI C trigraphs. You don't want to know about this
779 brain-damage. The @samp{-ansi} option implies @samp{-trigraphs}.
781 @cindex traditional C language
782 @cindex C language, traditional
784 Attempt to support some aspects of traditional C compilers.
789 All @code{extern} declarations take effect globally even if they
790 are written inside of a function definition. This includes implicit
791 declarations of functions.
794 The newer keywords @code{typeof}, @code{inline}, @code{signed}, @code{const}
795 and @code{volatile} are not recognized. (You can still use the
796 alternative keywords such as @code{__typeof__}, @code{__inline__}, and
800 Comparisons between pointers and integers are always allowed.
803 Integer types @code{unsigned short} and @code{unsigned char} promote
804 to @code{unsigned int}.
807 Out-of-range floating point literals are not an error.
810 Certain constructs which ANSI regards as a single invalid preprocessing
811 number, such as @samp{0xe-0xd}, are treated as expressions instead.
814 String ``constants'' are not necessarily constant; they are stored in
815 writable space, and identical looking constants are allocated
816 separately. (This is the same as the effect of
817 @samp{-fwritable-strings}.)
819 @cindex @code{longjmp} and automatic variables
821 All automatic variables not declared @code{register} are preserved by
822 @code{longjmp}. Ordinarily, GNU C follows ANSI C: automatic variables
823 not declared @code{volatile} may be clobbered.
828 @cindex escape sequences, traditional
829 The character escape sequences @samp{\x} and @samp{\a} evaluate as the
830 literal characters @samp{x} and @samp{a} respectively. Without
831 @w{@samp{-traditional}}, @samp{\x} is a prefix for the hexadecimal
832 representation of a character, and @samp{\a} produces a bell.
835 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
836 if your program uses names that are normally GNU C builtin functions for
837 other purposes of its own.
839 You cannot use @samp{-traditional} if you include any header files that
840 rely on ANSI C features. Some vendors are starting to ship systems with
841 ANSI C header files and you cannot use @samp{-traditional} on such
842 systems to compile files that include any system headers.
844 The @samp{-traditional} option also enables @samp{-traditional-cpp},
845 which is described next.
847 @item -traditional-cpp
848 Attempt to support some aspects of traditional C preprocessors.
853 Comments convert to nothing at all, rather than to a space. This allows
854 traditional token concatenation.
857 In a preprocessing directive, the @samp{#} symbol must appear as the first
861 Macro arguments are recognized within string constants in a macro
862 definition (and their values are stringified, though without additional
863 quote marks, when they appear in such a context). The preprocessor
864 always considers a string constant to end at a newline.
867 @cindex detecting @w{@samp{-traditional}}
868 The predefined macro @code{__STDC__} is not defined when you use
869 @samp{-traditional}, but @code{__GNUC__} is (since the GNU extensions
870 which @code{__GNUC__} indicates are not affected by
871 @samp{-traditional}). If you need to write header files that work
872 differently depending on whether @samp{-traditional} is in use, by
873 testing both of these predefined macros you can distinguish four
874 situations: GNU C, traditional GNU C, other ANSI C compilers, and other
875 old C compilers. The predefined macro @code{__STDC_VERSION__} is also
876 not defined when you use @samp{-traditional}. @xref{Standard
877 Predefined,,Standard Predefined Macros,cpp.info,The C Preprocessor},
878 for more discussion of these and other predefined macros.
881 @cindex string constants vs newline
882 @cindex newline vs string constants
883 The preprocessor considers a string constant to end at a newline (unless
884 the newline is escaped with @samp{\}). (Without @w{@samp{-traditional}},
885 string constants can contain the newline character as typed.)
888 @item -fcond-mismatch
889 Allow conditional expressions with mismatched types in the second and
890 third arguments. The value of such an expression is void.
892 @item -funsigned-char
893 Let the type @code{char} be unsigned, like @code{unsigned char}.
895 Each kind of machine has a default for what @code{char} should
896 be. It is either like @code{unsigned char} by default or like
897 @code{signed char} by default.
899 Ideally, a portable program should always use @code{signed char} or
900 @code{unsigned char} when it depends on the signedness of an object.
901 But many programs have been written to use plain @code{char} and
902 expect it to be signed, or expect it to be unsigned, depending on the
903 machines they were written for. This option, and its inverse, let you
904 make such a program work with the opposite default.
906 The type @code{char} is always a distinct type from each of
907 @code{signed char} or @code{unsigned char}, even though its behavior
908 is always just like one of those two.
911 Let the type @code{char} be signed, like @code{signed char}.
913 Note that this is equivalent to @samp{-fno-unsigned-char}, which is
914 the negative form of @samp{-funsigned-char}. Likewise, the option
915 @samp{-fno-signed-char} is equivalent to @samp{-funsigned-char}.
917 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
918 if your program uses names that are normally GNU C builtin functions for
919 other purposes of its own.
921 You cannot use @samp{-traditional} if you include any header files that
922 rely on ANSI C features. Some vendors are starting to ship systems with
923 ANSI C header files and you cannot use @samp{-traditional} on such
924 systems to compile files that include any system headers.
926 @item -fsigned-bitfields
927 @itemx -funsigned-bitfields
928 @itemx -fno-signed-bitfields
929 @itemx -fno-unsigned-bitfields
930 These options control whether a bitfield is signed or unsigned, when the
931 declaration does not use either @code{signed} or @code{unsigned}. By
932 default, such a bitfield is signed, because this is consistent: the
933 basic integer types such as @code{int} are signed types.
935 However, when @samp{-traditional} is used, bitfields are all unsigned
938 @item -fwritable-strings
939 Store string constants in the writable data segment and don't uniquize
940 them. This is for compatibility with old programs which assume they can
941 write into string constants. The option @samp{-traditional} also has
944 Writing into string constants is a very bad idea; ``constants'' should
947 @item -fallow-single-precision
948 Do not promote single precision math operations to double precision,
949 even when compiling with @samp{-traditional}.
951 Traditional K&R C promotes all floating point operations to double
952 precision, regardless of the sizes of the operands. On the
953 architecture for which you are compiling, single precision may be faster
954 than double precision. If you must use @samp{-traditional}, but want
955 to use single precision operations when the operands are single
956 precision, use this option. This option has no effect when compiling
957 with ANSI or GNU C conventions (the default).
960 Override the underlying type for @samp{wchar_t} to be @samp{short
961 unsigned int} instead of the default for the target. This option is
962 useful for building programs to run under WINE.
965 @node C++ Dialect Options
966 @section Options Controlling C++ Dialect
968 @cindex compiler options, C++
969 @cindex C++ options, command line
971 This section describes the command-line options that are only meaningful
972 for C++ programs; but you can also use most of the GNU compiler options
973 regardless of what language your program is in. For example, you
974 might compile a file @code{firstClass.C} like this:
977 g++ -g -frepo -O -c firstClass.C
981 In this example, only @samp{-frepo} is an option meant
982 only for C++ programs; you can use the other options with any
983 language supported by GCC.
985 Here is a list of options that are @emph{only} for compiling C++ programs:
988 @item -fno-access-control
989 Turn off all access checking. This switch is mainly useful for working
990 around bugs in the access control code.
993 Check that the pointer returned by @code{operator new} is non-null
994 before attempting to modify the storage allocated. The current Working
995 Paper requires that @code{operator new} never return a null pointer, so
996 this check is normally unnecessary.
998 An alternative to using this option is to specify that your
999 @code{operator new} does not throw any exceptions; if you declare it
1000 @samp{throw()}, g++ will check the return value. See also @samp{new
1003 @item -fconserve-space
1004 Put uninitialized or runtime-initialized global variables into the
1005 common segment, as C does. This saves space in the executable at the
1006 cost of not diagnosing duplicate definitions. If you compile with this
1007 flag and your program mysteriously crashes after @code{main()} has
1008 completed, you may have an object that is being destroyed twice because
1009 two definitions were merged.
1011 This option is no longer useful on most targets, now that support has
1012 been added for putting variables into BSS without making them common.
1014 @item -fdollars-in-identifiers
1015 Accept @samp{$} in identifiers. You can also explicitly prohibit use of
1016 @samp{$} with the option @samp{-fno-dollars-in-identifiers}. (GNU C allows
1017 @samp{$} by default on most target systems, but there are a few exceptions.)
1018 Traditional C allowed the character @samp{$} to form part of
1019 identifiers. However, ANSI C and C++ forbid @samp{$} in identifiers.
1021 @item -fno-elide-constructors
1022 The C++ standard allows an implementation to omit creating a temporary
1023 which is only used to initialize another object of the same type.
1024 Specifying this option disables that optimization, and forces g++ to
1025 call the copy constructor in all cases.
1027 @item -fexternal-templates
1028 Cause template instantiations to obey @samp{#pragma interface} and
1029 @samp{implementation}; template instances are emitted or not according
1030 to the location of the template definition. @xref{Template
1031 Instantiation}, for more information.
1033 This option is deprecated.
1035 @item -falt-external-templates
1036 Similar to -fexternal-templates, but template instances are emitted or
1037 not according to the place where they are first instantiated.
1038 @xref{Template Instantiation}, for more information.
1040 This option is deprecated.
1043 @itemx -fno-for-scope
1044 If -ffor-scope is specified, the scope of variables declared in
1045 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1046 as specified by the draft C++ standard.
1047 If -fno-for-scope is specified, the scope of variables declared in
1048 a @i{for-init-statement} extends to the end of the enclosing scope,
1049 as was the case in old versions of gcc, and other (traditional)
1050 implementations of C++.
1052 The default if neither flag is given to follow the standard,
1053 but to allow and give a warning for old-style code that would
1054 otherwise be invalid, or have different behavior.
1056 @item -fno-gnu-keywords
1057 Do not recognize @code{classof}, @code{headof}, or @code{typeof} as a
1058 keyword, so that code can use these words as identifiers. You can use
1059 the keywords @code{__classof__}, @code{__headof__}, and
1060 @code{__typeof__} instead. @samp{-ansi} implies
1061 @samp{-fno-gnu-keywords}.
1063 @item -fguiding-decls
1064 Treat a function declaration with the same type as a potential function
1065 template instantiation as though it declares that instantiation, not a
1066 normal function. If a definition is given for the function later in the
1067 translation unit (or another translation unit if the target supports
1068 weak symbols), that definition will be used; otherwise the template will
1069 be instantiated. This behavior reflects the C++ language prior to
1070 September 1996, when guiding declarations were removed.
1072 This option implies @samp{-fname-mangling-version-0}, and will not work
1073 with other name mangling versions. Like all options that change the
1074 ABI, all C++ code, @emph{including libgcc.a} must be built with the same
1075 setting of this option.
1078 Treat the @code{namespace std} as a namespace, instead of ignoring
1079 it. For compatibility with earlier versions of g++, the compiler will,
1080 by default, ignore @code{namespace-declarations},
1081 @code{using-declarations}, @code{using-directives}, and
1082 @code{namespace-names}, if they involve @code{std}.
1084 @item -fhuge-objects
1085 Support virtual function calls for objects that exceed the size
1086 representable by a @samp{short int}. Users should not use this flag by
1087 default; if you need to use it, the compiler will tell you so.
1089 This flag is not useful when compiling with -fvtable-thunks.
1091 Like all options that change the ABI, all C++ code, @emph{including
1092 libgcc} must be built with the same setting of this option.
1094 @item -fno-implicit-templates
1095 Never emit code for non-inline templates which are instantiated
1096 implicitly (i.e. by use); only emit code for explicit instantiations.
1097 @xref{Template Instantiation}, for more information.
1099 @item -fno-implicit-inline-templates
1100 Don't emit code for implicit instantiations of inline templates, either.
1101 The default is to handle inlines differently so that compiles with and
1102 without optimization will need the same set of explicit instantiations.
1104 @item -finit-priority
1105 Support @samp{__attribute__ ((init_priority (n)))} for controlling the
1106 order of initialization of file-scope objects. On ELF targets, this
1107 requires GNU ld 2.10 or later.
1109 @item -fno-implement-inlines
1110 To save space, do not emit out-of-line copies of inline functions
1111 controlled by @samp{#pragma implementation}. This will cause linker
1112 errors if these functions are not inlined everywhere they are called.
1114 @item -fms-extensions
1115 Disable pedwarns about constructs used in MFC, such as implicit int and
1116 getting a pointer to member function via non-standard syntax.
1118 @item -fname-mangling-version-@var{n}
1119 Control the way in which names are mangled. Version 0 is compatible
1120 with versions of g++ before 2.8. Version 1 is the default. Version 1
1121 will allow correct mangling of function templates. For example,
1122 version 0 mangling does not mangle foo<int, double> and foo<int, char>
1123 given this declaration:
1126 template <class T, class U> void foo(T t);
1129 Like all options that change the ABI, all C++ code, @emph{including
1130 libgcc} must be built with the same setting of this option.
1132 @item -foperator-names
1133 Recognize the operator name keywords @code{and}, @code{bitand},
1134 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1135 synonyms for the symbols they refer to. @samp{-ansi} implies
1136 @samp{-foperator-names}.
1138 @item -fno-optional-diags
1139 Disable diagnostics that the standard says a compiler does not need to
1140 issue. Currently, the only such diagnostic issued by g++ is the one for
1141 a name having multiple meanings within a class.
1144 Downgrade messages about nonconformant code from errors to warnings. By
1145 default, g++ effectively sets @samp{-pedantic-errors} without
1146 @samp{-pedantic}; this option reverses that. This behavior and this
1147 option are superseded by @samp{-pedantic}, which works as it does for GNU C.
1150 Enable automatic template instantiation. This option also implies
1151 @samp{-fno-implicit-templates}. @xref{Template Instantiation}, for more
1155 Disable generation of information about every class with virtual
1156 functions for use by the C++ runtime type identification features
1157 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1158 of the language, you can save some space by using this flag. Note that
1159 exception handling uses the same information, but it will generate it as
1162 @item -fstrict-prototype
1163 Within an @samp{extern "C"} linkage specification, treat a function
1164 declaration with no arguments, such as @samp{int foo ();}, as declaring
1165 the function to take no arguments. Normally, such a declaration means
1166 that the function @code{foo} can take any combination of arguments, as
1167 in C. @samp{-pedantic} implies @samp{-fstrict-prototype} unless
1168 overridden with @samp{-fno-strict-prototype}.
1170 Specifying this option will also suppress implicit declarations of
1173 This flag no longer affects declarations with C++ linkage.
1176 @itemx -fno-squangle
1177 @samp{-fsquangle} will enable a compressed form of name mangling for
1178 identifiers. In particular, it helps to shorten very long names by recognizing
1179 types and class names which occur more than once, replacing them with special
1180 short ID codes. This option also requires any C++ libraries being used to
1181 be compiled with this option as well. The compiler has this disabled (the
1182 equivalent of @samp{-fno-squangle}) by default.
1184 Like all options that change the ABI, all C++ code, @emph{including
1185 libgcc.a} must be built with the same setting of this option.
1187 @item -ftemplate-depth-@var{n}
1188 Set the maximum instantiation depth for template classes to @var{n}.
1189 A limit on the template instantiation depth is needed to detect
1190 endless recursions during template class instantiation. ANSI/ISO C++
1191 conforming programs must not rely on a maximum depth greater than 17.
1193 @item -fvtable-thunks
1194 Use @samp{thunks} to implement the virtual function dispatch table
1195 (@samp{vtable}). The traditional (cfront-style) approach to
1196 implementing vtables was to store a pointer to the function and two
1197 offsets for adjusting the @samp{this} pointer at the call site. Newer
1198 implementations store a single pointer to a @samp{thunk} function which
1199 does any necessary adjustment and then calls the target function.
1201 This option also enables a heuristic for controlling emission of
1202 vtables; if a class has any non-inline virtual functions, the vtable
1203 will be emitted in the translation unit containing the first one of
1206 Like all options that change the ABI, all C++ code, @emph{including
1207 libgcc.a} must be built with the same setting of this option.
1210 Do not search for header files in the standard directories specific to
1211 C++, but do still search the other standard directories. (This option
1212 is used when building the C++ library.)
1215 In addition, these optimization, warning, and code generation options
1216 have meanings only for C++ programs:
1219 @item -fno-default-inline
1220 Do not assume @samp{inline} for functions defined inside a class scope.
1221 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1222 functions will have linkage like inline functions; they just won't be
1225 @item -Wctor-dtor-privacy (C++ only)
1226 Warn when a class seems unusable, because all the constructors or
1227 destructors in a class are private and the class has no friends or
1228 public static member functions.
1230 @item -Wnon-virtual-dtor (C++ only)
1231 Warn when a class declares a non-virtual destructor that should probably
1232 be virtual, because it looks like the class will be used polymorphically.
1234 @item -Wreorder (C++ only)
1235 @cindex reordering, warning
1236 @cindex warning for reordering of member initializers
1237 Warn when the order of member initializers given in the code does not
1238 match the order in which they must be executed. For instance:
1244 A(): j (0), i (1) @{ @}
1248 Here the compiler will warn that the member initializers for @samp{i}
1249 and @samp{j} will be rearranged to match the declaration order of the
1253 The following @samp{-W@dots{}} options are not affected by @samp{-Wall}.
1256 @item -Weffc++ (C++ only)
1257 Warn about violations of various style guidelines from Scott Meyers'
1258 @cite{Effective C++} books. If you use this option, you should be aware
1259 that the standard library headers do not obey all of these guidelines;
1260 you can use @samp{grep -v} to filter out those warnings.
1262 @item -Wno-deprecated (C++ only)
1263 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1265 @item -Wno-non-template-friend (C++ only)
1266 Disable warnings when non-templatized friend functions are declared
1267 within a template. With the advent of explicit template specification
1268 support in g++, if the name of the friend is an unqualified-id (ie,
1269 @samp{friend foo(int)}), the C++ language specification demands that the
1270 friend declare or define an ordinary, nontemplate function. (Section
1271 14.5.3). Before g++ implemented explicit specification, unqualified-ids
1272 could be interpreted as a particular specialization of a templatized
1273 function. Because this non-conforming behavior is no longer the default
1274 behavior for g++, @samp{-Wnon-template-friend} allows the compiler to
1275 check existing code for potential trouble spots, and is on by default.
1276 This new compiler behavior can also be turned off with the flag
1277 @samp{-fguiding-decls}, which activates the older, non-specification
1278 compiler code, or with @samp{-Wno-non-template-friend} which keeps the
1279 conformant compiler code but disables the helpful warning.
1281 @item -Wold-style-cast (C++ only)
1282 Warn if an old-style (C-style) cast is used within a C++ program. The
1283 new-style casts (@samp{static_cast}, @samp{reinterpret_cast}, and
1284 @samp{const_cast}) are less vulnerable to unintended effects.
1286 @item -Woverloaded-virtual (C++ only)
1287 @cindex overloaded virtual fn, warning
1288 @cindex warning for overloaded virtual fn
1289 Warn when a derived class function declaration may be an error in
1290 defining a virtual function. In a derived class, the
1291 definitions of virtual functions must match the type signature of a
1292 virtual function declared in the base class. With this option, the
1293 compiler warns when you define a function with the same name as a
1294 virtual function, but with a type signature that does not match any
1295 declarations from the base class.
1297 @item -Wno-pmf-conversions (C++ only)
1298 Disable the diagnostic for converting a bound pointer to member function
1301 @item -Wsign-promo (C++ only)
1302 Warn when overload resolution chooses a promotion from unsigned or
1303 enumeral type to a signed type over a conversion to an unsigned type of
1304 the same size. Previous versions of g++ would try to preserve
1305 unsignedness, but the standard mandates the current behavior.
1307 @item -Wsynth (C++ only)
1308 @cindex warning for synthesized methods
1309 @cindex synthesized methods, warning
1310 Warn when g++'s synthesis behavior does not match that of cfront. For
1316 A& operator = (int);
1326 In this example, g++ will synthesize a default @samp{A& operator =
1327 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1330 @node Warning Options
1331 @section Options to Request or Suppress Warnings
1332 @cindex options to control warnings
1333 @cindex warning messages
1334 @cindex messages, warning
1335 @cindex suppressing warnings
1337 Warnings are diagnostic messages that report constructions which
1338 are not inherently erroneous but which are risky or suggest there
1339 may have been an error.
1341 You can request many specific warnings with options beginning @samp{-W},
1342 for example @samp{-Wimplicit} to request warnings on implicit
1343 declarations. Each of these specific warning options also has a
1344 negative form beginning @samp{-Wno-} to turn off warnings;
1345 for example, @samp{-Wno-implicit}. This manual lists only one of the
1346 two forms, whichever is not the default.
1348 These options control the amount and kinds of warnings produced by GCC:
1351 @cindex syntax checking
1353 Check the code for syntax errors, but don't do anything beyond that.
1356 Issue all the warnings demanded by strict ANSI C and ISO C++;
1357 reject all programs that use forbidden extensions.
1359 Valid ANSI C and ISO C++ programs should compile properly with or without
1360 this option (though a rare few will require @samp{-ansi}). However,
1361 without this option, certain GNU extensions and traditional C and C++
1362 features are supported as well. With this option, they are rejected.
1364 @samp{-pedantic} does not cause warning messages for use of the
1365 alternate keywords whose names begin and end with @samp{__}. Pedantic
1366 warnings are also disabled in the expression that follows
1367 @code{__extension__}. However, only system header files should use
1368 these escape routes; application programs should avoid them.
1369 @xref{Alternate Keywords}.
1371 This option is not intended to be @i{useful}; it exists only to satisfy
1372 pedants who would otherwise claim that GCC fails to support the ANSI
1375 Some users try to use @samp{-pedantic} to check programs for strict ANSI
1376 C conformance. They soon find that it does not do quite what they want:
1377 it finds some non-ANSI practices, but not all---only those for which
1378 ANSI C @emph{requires} a diagnostic.
1380 A feature to report any failure to conform to ANSI C might be useful in
1381 some instances, but would require considerable additional work and would
1382 be quite different from @samp{-pedantic}. We don't have plans to
1383 support such a feature in the near future.
1385 @item -pedantic-errors
1386 Like @samp{-pedantic}, except that errors are produced rather than
1390 Inhibit all warning messages.
1393 Inhibit warning messages about the use of @samp{#import}.
1395 @item -Wchar-subscripts
1396 Warn if an array subscript has type @code{char}. This is a common cause
1397 of error, as programmers often forget that this type is signed on some
1401 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
1402 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
1405 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
1406 the arguments supplied have types appropriate to the format string
1409 @item -Wimplicit-int
1410 Warn when a declaration does not specify a type.
1412 @item -Wimplicit-function-declaration
1413 @itemx -Werror-implicit-function-declaration
1414 Give a warning (or error) whenever a function is used before being
1418 Same as @samp{-Wimplicit-int} and @samp{-Wimplicit-function-}@*
1422 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
1423 function with external linkage, returning int, taking either zero
1424 arguments, two, or three arguments of appropriate types.
1427 Warn if a multicharacter constant (@samp{'FOOF'}) is used. Usually they
1428 indicate a typo in the user's code, as they have implementation-defined
1429 values, and should not be used in portable code.
1432 Warn if parentheses are omitted in certain contexts, such
1433 as when there is an assignment in a context where a truth value
1434 is expected, or when operators are nested whose precedence people
1435 often get confused about.
1437 Also warn about constructions where there may be confusion to which
1438 @code{if} statement an @code{else} branch belongs. Here is an example of
1451 In C, every @code{else} branch belongs to the innermost possible @code{if}
1452 statement, which in this example is @code{if (b)}. This is often not
1453 what the programmer expected, as illustrated in the above example by
1454 indentation the programmer chose. When there is the potential for this
1455 confusion, GNU C will issue a warning when this flag is specified.
1456 To eliminate the warning, add explicit braces around the innermost
1457 @code{if} statement so there is no way the @code{else} could belong to
1458 the enclosing @code{if}. The resulting code would look like this:
1473 Warn whenever a function is defined with a return-type that defaults
1474 to @code{int}. Also warn about any @code{return} statement with no
1475 return-value in a function whose return-type is not @code{void}.
1478 Warn whenever a @code{switch} statement has an index of enumeral type
1479 and lacks a @code{case} for one or more of the named codes of that
1480 enumeration. (The presence of a @code{default} label prevents this
1481 warning.) @code{case} labels outside the enumeration range also
1482 provoke warnings when this option is used.
1485 Warn if any trigraphs are encountered (assuming they are enabled).
1488 Warn whenever a variable is unused aside from its declaration,
1489 whenever a function is declared static but never defined, whenever a
1490 label is declared but not used, and whenever a statement computes a
1491 result that is explicitly not used.
1493 In order to get a warning about an unused function parameter, you must
1494 specify both @samp{-W} and @samp{-Wunused}.
1496 To suppress this warning for an expression, simply cast it to void. For
1497 unused variables, parameters and labels, use the @samp{unused} attribute
1498 (@pxref{Variable Attributes}).
1500 @item -Wuninitialized
1501 Warn if an automatic variable is used without first being initialized or
1502 if a variable may be clobbered by a @code{setjmp} call.
1504 These warnings are possible only in optimizing compilation,
1505 because they require data flow information that is computed only
1506 when optimizing. If you don't specify @samp{-O}, you simply won't
1509 These warnings occur only for variables that are candidates for
1510 register allocation. Therefore, they do not occur for a variable that
1511 is declared @code{volatile}, or whose address is taken, or whose size
1512 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
1513 structures, unions or arrays, even when they are in registers.
1515 Note that there may be no warning about a variable that is used only
1516 to compute a value that itself is never used, because such
1517 computations may be deleted by data flow analysis before the warnings
1520 These warnings are made optional because GCC is not smart
1521 enough to see all the reasons why the code might be correct
1522 despite appearing to have an error. Here is one example of how
1541 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
1542 always initialized, but GCC doesn't know this. Here is
1543 another common case:
1548 if (change_y) save_y = y, y = new_y;
1550 if (change_y) y = save_y;
1555 This has no bug because @code{save_y} is used only if it is set.
1557 @cindex @code{longjmp} warnings
1558 This option also warns when a nonvolatile automatic variable might be
1559 changed by a call to @code{longjmp}. These warnings as well are possible
1560 only in optimizing compilation.
1562 The compiler sees only the calls to @code{setjmp}. It cannot know
1563 where @code{longjmp} will be called; in fact, a signal handler could
1564 call it at any point in the code. As a result, you may get a warning
1565 even when there is in fact no problem because @code{longjmp} cannot
1566 in fact be called at the place which would cause a problem.
1568 Some spurious warnings can be avoided if you declare all the functions
1569 you use that never return as @code{noreturn}. @xref{Function
1572 @item -Wreorder (C++ only)
1573 @cindex reordering, warning
1574 @cindex warning for reordering of member initializers
1575 Warn when the order of member initializers given in the code does not
1576 match the order in which they must be executed. For instance:
1578 @item -Wunknown-pragmas
1579 @cindex warning for unknown pragmas
1580 @cindex unknown pragmas, warning
1581 @cindex pragmas, warning of unknown
1582 Warn when a #pragma directive is encountered which is not understood by
1583 GCC. If this command line option is used, warnings will even be issued
1584 for unknown pragmas in system header files. This is not the case if
1585 the warnings were only enabled by the @samp{-Wall} command line option.
1588 All of the above @samp{-W} options combined. This enables all the
1589 warnings about constructions that some users consider questionable, and
1590 that are easy to avoid (or modify to prevent the warning), even in
1591 conjunction with macros.
1594 The following @samp{-W@dots{}} options are not implied by @samp{-Wall}.
1595 Some of them warn about constructions that users generally do not
1596 consider questionable, but which occasionally you might wish to check
1597 for; others warn about constructions that are necessary or hard to avoid
1598 in some cases, and there is no simple way to modify the code to suppress
1603 Print extra warning messages for these events:
1607 A function can return either with or without a value. (Falling
1608 off the end of the function body is considered returning without
1609 a value.) For example, this function would evoke such a
1623 An expression-statement or the left-hand side of a comma expression
1624 contains no side effects.
1625 To suppress the warning, cast the unused expression to void.
1626 For example, an expression such as @samp{x[i,j]} will cause a warning,
1627 but @samp{x[(void)i,j]} will not.
1630 An unsigned value is compared against zero with @samp{<} or @samp{<=}.
1633 A comparison like @samp{x<=y<=z} appears; this is equivalent to
1634 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
1635 that of ordinary mathematical notation.
1638 Storage-class specifiers like @code{static} are not the first things in
1639 a declaration. According to the C Standard, this usage is obsolescent.
1642 If @samp{-Wall} or @samp{-Wunused} is also specified, warn about unused
1646 A comparison between signed and unsigned values could produce an
1647 incorrect result when the signed value is converted to unsigned.
1648 (But don't warn if @samp{-Wno-sign-compare} is also specified.)
1651 An aggregate has a partly bracketed initializer.
1652 For example, the following code would evoke such a warning,
1653 because braces are missing around the initializer for @code{x.h}:
1656 struct s @{ int f, g; @};
1657 struct t @{ struct s h; int i; @};
1658 struct t x = @{ 1, 2, 3 @};
1662 An aggregate has an initializer which does not initialize all members.
1663 For example, the following code would cause such a warning, because
1664 @code{x.h} would be implicitly initialized to zero:
1667 struct s @{ int f, g, h; @};
1668 struct s x = @{ 3, 4 @};
1673 Warn if floating point values are used in equality comparisons.
1675 @item -Wtraditional (C only)
1676 Warn about certain constructs that behave differently in traditional and
1681 Macro arguments occurring within string constants in the macro body.
1682 These would substitute the argument in traditional C, but are part of
1683 the constant in ANSI C.
1686 A function declared external in one block and then used after the end of
1690 A @code{switch} statement has an operand of type @code{long}.
1693 A non-@code{static} function declaration follows a @code{static} one.
1694 This construct is not accepted by some traditional C compilers.
1698 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
1701 Warn whenever a local variable shadows another local variable.
1703 @item -Wid-clash-@var{len}
1704 Warn whenever two distinct identifiers match in the first @var{len}
1705 characters. This may help you prepare a program that will compile
1706 with certain obsolete, brain-damaged compilers.
1708 @item -Wlarger-than-@var{len}
1709 Warn whenever an object of larger than @var{len} bytes is defined.
1711 @item -Wpointer-arith
1712 Warn about anything that depends on the ``size of'' a function type or
1713 of @code{void}. GNU C assigns these types a size of 1, for
1714 convenience in calculations with @code{void *} pointers and pointers
1717 @item -Wbad-function-cast (C only)
1718 Warn whenever a function call is cast to a non-matching type.
1719 For example, warn if @code{int malloc()} is cast to @code{anything *}.
1722 Warn whenever a pointer is cast so as to remove a type qualifier from
1723 the target type. For example, warn if a @code{const char *} is cast
1724 to an ordinary @code{char *}.
1727 Warn whenever a pointer is cast such that the required alignment of the
1728 target is increased. For example, warn if a @code{char *} is cast to
1729 an @code{int *} on machines where integers can only be accessed at
1730 two- or four-byte boundaries.
1732 @item -Wwrite-strings
1733 Give string constants the type @code{const char[@var{length}]} so that
1734 copying the address of one into a non-@code{const} @code{char *}
1735 pointer will get a warning. These warnings will help you find at
1736 compile time code that can try to write into a string constant, but
1737 only if you have been very careful about using @code{const} in
1738 declarations and prototypes. Otherwise, it will just be a nuisance;
1739 this is why we did not make @samp{-Wall} request these warnings.
1742 Warn if a prototype causes a type conversion that is different from what
1743 would happen to the same argument in the absence of a prototype. This
1744 includes conversions of fixed point to floating and vice versa, and
1745 conversions changing the width or signedness of a fixed point argument
1746 except when the same as the default promotion.
1748 Also, warn if a negative integer constant expression is implicitly
1749 converted to an unsigned type. For example, warn about the assignment
1750 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
1751 casts like @code{(unsigned) -1}.
1753 @item -Wsign-compare
1754 @cindex warning for comparison of signed and unsigned values
1755 @cindex comparison of signed and unsigned values, warning
1756 @cindex signed and unsigned values, comparison warning
1757 Warn when a comparison between signed and unsigned values could produce
1758 an incorrect result when the signed value is converted to unsigned.
1759 This warning is also enabled by @samp{-W}; to get the other warnings
1760 of @samp{-W} without this warning, use @samp{-W -Wno-sign-compare}.
1762 @item -Waggregate-return
1763 Warn if any functions that return structures or unions are defined or
1764 called. (In languages where you can return an array, this also elicits
1767 @item -Wstrict-prototypes (C only)
1768 Warn if a function is declared or defined without specifying the
1769 argument types. (An old-style function definition is permitted without
1770 a warning if preceded by a declaration which specifies the argument
1773 @item -Wmissing-prototypes (C only)
1774 Warn if a global function is defined without a previous prototype
1775 declaration. This warning is issued even if the definition itself
1776 provides a prototype. The aim is to detect global functions that fail
1777 to be declared in header files.
1779 @item -Wmissing-declarations
1780 Warn if a global function is defined without a previous declaration.
1781 Do so even if the definition itself provides a prototype.
1782 Use this option to detect global functions that are not declared in
1785 @item -Wmissing-noreturn
1786 Warn about functions which might be candidates for attribute @code{noreturn}.
1787 Note these are only possible candidates, not absolute ones. Care should
1788 be taken to manually verify functions actually do not ever return before
1789 adding the @code{noreturn} attribute, otherwise subtle code generation
1790 bugs could be introduced.
1792 @item -Wredundant-decls
1793 Warn if anything is declared more than once in the same scope, even in
1794 cases where multiple declaration is valid and changes nothing.
1796 @item -Wnested-externs (C only)
1797 Warn if an @code{extern} declaration is encountered within a function.
1799 @item -Wunreachable-code
1800 Warn if the compiler detects that code will never be executed.
1802 This option is intended to warn when the compiler detects that at
1803 least a whole line of source code will never be executed, because
1804 some condition is never satisfied or because it is after a
1805 procedure that never returns.
1807 It is possible for this option to produce a warning even though there
1808 are circumstances under which part of the affected line can be executed,
1809 so care should be taken when removing apparently-unreachable code.
1811 For instance, when a function is inlined, a warning may mean that the
1812 line is unreachable in only one inlined copy of the function.
1814 This option is not made part of @samp{-Wall} because in a debugging
1815 version of a program there is often substantial code which checks
1816 correct functioning of the program and is, hopefully, unreachable
1817 because the program does work. Another common use of unreachable
1818 code is to provide behaviour which is selectable at compile-time.
1821 Warn if a function can not be inlined and it was declared as inline.
1824 Warn if @samp{long long} type is used. This is default. To inhibit
1825 the warning messages, use @samp{-Wno-long-long}. Flags
1826 @samp{-Wlong-long} and @samp{-Wno-long-long} are taken into account
1827 only when @samp{-pedantic} flag is used.
1830 Make all warnings into errors.
1833 @node Debugging Options
1834 @section Options for Debugging Your Program or GCC
1835 @cindex options, debugging
1836 @cindex debugging information options
1838 GCC has various special options that are used for debugging
1839 either your program or GCC:
1843 Produce debugging information in the operating system's native format
1844 (stabs, COFF, XCOFF, or DWARF). GDB can work with this debugging
1847 On most systems that use stabs format, @samp{-g} enables use of extra
1848 debugging information that only GDB can use; this extra information
1849 makes debugging work better in GDB but will probably make other debuggers
1851 refuse to read the program. If you want to control for certain whether
1852 to generate the extra information, use @samp{-gstabs+}, @samp{-gstabs},
1853 @samp{-gxcoff+}, @samp{-gxcoff}, @samp{-gdwarf-1+}, or @samp{-gdwarf-1}
1856 Unlike most other C compilers, GCC allows you to use @samp{-g} with
1857 @samp{-O}. The shortcuts taken by optimized code may occasionally
1858 produce surprising results: some variables you declared may not exist
1859 at all; flow of control may briefly move where you did not expect it;
1860 some statements may not be executed because they compute constant
1861 results or their values were already at hand; some statements may
1862 execute in different places because they were moved out of loops.
1864 Nevertheless it proves possible to debug optimized output. This makes
1865 it reasonable to use the optimizer for programs that might have bugs.
1867 The following options are useful when GCC is generated with the
1868 capability for more than one debugging format.
1871 Produce debugging information for use by GDB. This means to use the
1872 most expressive format available (DWARF 2, stabs, or the native format
1873 if neither of those are supported), including GDB extensions if at all
1877 Produce debugging information in stabs format (if that is supported),
1878 without GDB extensions. This is the format used by DBX on most BSD
1879 systems. On MIPS, Alpha and System V Release 4 systems this option
1880 produces stabs debugging output which is not understood by DBX or SDB.
1881 On System V Release 4 systems this option requires the GNU assembler.
1884 Produce debugging information in stabs format (if that is supported),
1885 using GNU extensions understood only by the GNU debugger (GDB). The
1886 use of these extensions is likely to make other debuggers crash or
1887 refuse to read the program.
1890 Produce debugging information in COFF format (if that is supported).
1891 This is the format used by SDB on most System V systems prior to
1895 Produce debugging information in XCOFF format (if that is supported).
1896 This is the format used by the DBX debugger on IBM RS/6000 systems.
1899 Produce debugging information in XCOFF format (if that is supported),
1900 using GNU extensions understood only by the GNU debugger (GDB). The
1901 use of these extensions is likely to make other debuggers crash or
1902 refuse to read the program, and may cause assemblers other than the GNU
1903 assembler (GAS) to fail with an error.
1906 Produce debugging information in DWARF version 1 format (if that is
1907 supported). This is the format used by SDB on most System V Release 4
1911 Produce debugging information in DWARF version 1 format (if that is
1912 supported), using GNU extensions understood only by the GNU debugger
1913 (GDB). The use of these extensions is likely to make other debuggers
1914 crash or refuse to read the program.
1917 Produce debugging information in DWARF version 2 format (if that is
1918 supported). This is the format used by DBX on IRIX 6.
1921 @itemx -ggdb@var{level}
1922 @itemx -gstabs@var{level}
1923 @itemx -gcoff@var{level}
1924 @itemx -gxcoff@var{level}
1925 @itemx -gdwarf@var{level}
1926 @itemx -gdwarf-2@var{level}
1927 Request debugging information and also use @var{level} to specify how
1928 much information. The default level is 2.
1930 Level 1 produces minimal information, enough for making backtraces in
1931 parts of the program that you don't plan to debug. This includes
1932 descriptions of functions and external variables, but no information
1933 about local variables and no line numbers.
1935 Level 3 includes extra information, such as all the macro definitions
1936 present in the program. Some debuggers support macro expansion when
1941 Generate extra code to write profile information suitable for the
1942 analysis program @code{prof}. You must use this option when compiling
1943 the source files you want data about, and you must also use it when
1946 @cindex @code{gprof}
1948 Generate extra code to write profile information suitable for the
1949 analysis program @code{gprof}. You must use this option when compiling
1950 the source files you want data about, and you must also use it when
1955 Generate extra code to write profile information for basic blocks, which will
1956 record the number of times each basic block is executed, the basic block start
1957 address, and the function name containing the basic block. If @samp{-g} is
1958 used, the line number and filename of the start of the basic block will also be
1959 recorded. If not overridden by the machine description, the default action is
1960 to append to the text file @file{bb.out}.
1962 This data could be analyzed by a program like @code{tcov}. Note,
1963 however, that the format of the data is not what @code{tcov} expects.
1964 Eventually GNU @code{gprof} should be extended to process this data.
1967 Makes the compiler print out each function name as it is compiled, and
1968 print some statistics about each pass when it finishes.
1971 Generate extra code to profile basic blocks. Your executable will
1972 produce output that is a superset of that produced when @samp{-a} is
1973 used. Additional output is the source and target address of the basic
1974 blocks where a jump takes place, the number of times a jump is executed,
1975 and (optionally) the complete sequence of basic blocks being executed.
1976 The output is appended to file @file{bb.out}.
1978 You can examine different profiling aspects without recompilation. Your
1979 executable will read a list of function names from file @file{bb.in}.
1980 Profiling starts when a function on the list is entered and stops when
1981 that invocation is exited. To exclude a function from profiling, prefix
1982 its name with `-'. If a function name is not unique, you can
1983 disambiguate it by writing it in the form
1984 @samp{/path/filename.d:functionname}. Your executable will write the
1985 available paths and filenames in file @file{bb.out}.
1987 Several function names have a special meaning:
1990 Write source, target and frequency of jumps to file @file{bb.out}.
1991 @item __bb_hidecall__
1992 Exclude function calls from frequency count.
1993 @item __bb_showret__
1994 Include function returns in frequency count.
1996 Write the sequence of basic blocks executed to file @file{bbtrace.gz}.
1997 The file will be compressed using the program @samp{gzip}, which must
1998 exist in your @code{PATH}. On systems without the @samp{popen}
1999 function, the file will be named @file{bbtrace} and will not be
2000 compressed. @strong{Profiling for even a few seconds on these systems
2001 will produce a very large file.} Note: @code{__bb_hidecall__} and
2002 @code{__bb_showret__} will not affect the sequence written to
2006 Here's a short example using different profiling parameters
2007 in file @file{bb.in}. Assume function @code{foo} consists of basic blocks
2008 1 and 2 and is called twice from block 3 of function @code{main}. After
2009 the calls, block 3 transfers control to block 4 of @code{main}.
2011 With @code{__bb_trace__} and @code{main} contained in file @file{bb.in},
2012 the following sequence of blocks is written to file @file{bbtrace.gz}:
2013 0 3 1 2 1 2 4. The return from block 2 to block 3 is not shown, because
2014 the return is to a point inside the block and not to the top. The
2015 block address 0 always indicates, that control is transferred
2016 to the trace from somewhere outside the observed functions. With
2017 @samp{-foo} added to @file{bb.in}, the blocks of function
2018 @code{foo} are removed from the trace, so only 0 3 4 remains.
2020 With @code{__bb_jumps__} and @code{main} contained in file @file{bb.in},
2021 jump frequencies will be written to file @file{bb.out}. The
2022 frequencies are obtained by constructing a trace of blocks
2023 and incrementing a counter for every neighbouring pair of blocks
2024 in the trace. The trace 0 3 1 2 1 2 4 displays the following
2028 Jump from block 0x0 to block 0x3 executed 1 time(s)
2029 Jump from block 0x3 to block 0x1 executed 1 time(s)
2030 Jump from block 0x1 to block 0x2 executed 2 time(s)
2031 Jump from block 0x2 to block 0x1 executed 1 time(s)
2032 Jump from block 0x2 to block 0x4 executed 1 time(s)
2035 With @code{__bb_hidecall__}, control transfer due to call instructions
2036 is removed from the trace, that is the trace is cut into three parts: 0
2037 3 4, 0 1 2 and 0 1 2. With @code{__bb_showret__}, control transfer due
2038 to return instructions is added to the trace. The trace becomes: 0 3 1
2039 2 3 1 2 3 4. Note, that this trace is not the same, as the sequence
2040 written to @file{bbtrace.gz}. It is solely used for counting jump
2043 @item -fprofile-arcs
2044 Instrument @dfn{arcs} during compilation. For each function of your
2045 program, GCC creates a program flow graph, then finds a spanning tree
2046 for the graph. Only arcs that are not on the spanning tree have to be
2047 instrumented: the compiler adds code to count the number of times that these
2048 arcs are executed. When an arc is the only exit or only entrance to a
2049 block, the instrumentation code can be added to the block; otherwise, a
2050 new basic block must be created to hold the instrumentation code.
2052 Since not every arc in the program must be instrumented, programs
2053 compiled with this option run faster than programs compiled with
2054 @samp{-a}, which adds instrumentation code to every basic block in the
2055 program. The tradeoff: since @code{gcov} does not have
2056 execution counts for all branches, it must start with the execution
2057 counts for the instrumented branches, and then iterate over the program
2058 flow graph until the entire graph has been solved. Hence, @code{gcov}
2059 runs a little more slowly than a program which uses information from
2062 @samp{-fprofile-arcs} also makes it possible to estimate branch
2063 probabilities, and to calculate basic block execution counts. In
2064 general, basic block execution counts do not give enough information to
2065 estimate all branch probabilities. When the compiled program exits, it
2066 saves the arc execution counts to a file called
2067 @file{@var{sourcename}.da}. Use the compiler option
2068 @samp{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
2069 Control Optimization}) when recompiling, to optimize using estimated
2070 branch probabilities.
2073 @item -ftest-coverage
2074 Create data files for the @code{gcov} code-coverage utility
2075 (@pxref{Gcov,, @code{gcov}: a GCC Test Coverage Program}).
2076 The data file names begin with the name of your source file:
2079 @item @var{sourcename}.bb
2080 A mapping from basic blocks to line numbers, which @code{gcov} uses to
2081 associate basic block execution counts with line numbers.
2083 @item @var{sourcename}.bbg
2084 A list of all arcs in the program flow graph. This allows @code{gcov}
2085 to reconstruct the program flow graph, so that it can compute all basic
2086 block and arc execution counts from the information in the
2087 @code{@var{sourcename}.da} file (this last file is the output from
2088 @samp{-fprofile-arcs}).
2091 @item -d@var{letters}
2092 Says to make debugging dumps during compilation at times specified by
2093 @var{letters}. This is used for debugging the compiler. The file names
2094 for most of the dumps are made by appending a pass number and a word to
2095 the source file name (e.g. @file{foo.c.00.rtl} or @file{foo.c.01.jump}).
2096 Here are the possible letters for use in @var{letters}, and their meanings:
2100 Annotate the assembler output with miscellaneous debugging information.
2102 Dump after computing branch probabilities, to @file{@var{file}.07.bp}.
2104 Dump after instruction combination, to the file @file{@var{file}.09.combine}.
2106 Dump after delayed branch scheduling, to @file{@var{file}.19.dbr}.
2108 Dump all macro definitions, at the end of preprocessing, in addition to
2111 Dump after purging ADDRESSOF, to @file{@var{file}.03.addressof}.
2113 Dump after flow analysis, to @file{@var{file}.08.flow}.
2115 Dump after global register allocation, to @file{@var{file}.13.greg}.
2117 Dump after GCSE, to @file{@var{file}.04.gcse}.
2119 Dump after first jump optimization, to @file{@var{file}.01.jump}.
2121 Dump after last jump optimization, to @file{@var{file}.17.jump2}.
2123 Dump after conversion from registers to stack, to @file{@var{file}.20.stack}.
2125 Dump after local register allocation, to @file{@var{file}.12.lreg}.
2127 Dump after loop optimization, to @file{@var{file}.05.loop}.
2129 Dump after performing the machine dependent reorganisation pass, to
2130 @file{@var{file}.18.mach}.
2132 Dump after the register move pass, to @file{@var{file}.10.regmove}.
2134 Dump after RTL generation, to @file{@var{file}.00.rtl}.
2136 Dump after the second instruction scheduling pass, to
2137 @file{@var{file}.16.sched2}.
2139 Dump after CSE (including the jump optimization that sometimes follows
2140 CSE), to @file{@var{file}.02.cse}.
2142 Dump after the first instruction scheduling pass, to
2143 @file{@var{file}.11.sched}.
2145 Dump after the second CSE pass (including the jump optimization that
2146 sometimes follows CSE), to @file{@var{file}.06.cse2}.
2148 Produce all the dumps listed above.
2150 Print statistics on memory usage, at the end of the run, to
2153 Annotate the assembler output with a comment indicating which
2154 pattern and alternative was used. The length of each instruction is
2157 For each of the other indicated dump files (except for
2158 @file{@var{file}.00.rtl}), dump a representation of the control flow graph
2159 suitible for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
2161 Dump after the second flow pass to @file{@var{file}.14.flow2}.
2163 Just generate RTL for a function instead of compiling it. Usually used
2166 Dump debugging information during parsing, to standard error.
2168 Dump after the peephole2 pass to @file{@var{file}.15.peephole2}.
2171 @item -fdump-unnumbered
2172 When doing debugging dumps (see -d option above), suppress instruction
2173 numbers and line number note output. This makes it more feasible to
2174 use diff on debugging dumps for compiler invokations with different
2175 options, in particular with and without -g.
2177 @item -fdump-translation-unit-@var{file} (C++ only)
2178 Dump a representation of the tree structure for the entire translation
2181 @item -fpretend-float
2182 When running a cross-compiler, pretend that the target machine uses the
2183 same floating point format as the host machine. This causes incorrect
2184 output of the actual floating constants, but the actual instruction
2185 sequence will probably be the same as GCC would make when running on
2189 Store the usual ``temporary'' intermediate files permanently; place them
2190 in the current directory and name them based on the source file. Thus,
2191 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
2192 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}.
2195 Report the CPU time taken by each subprocess in the compilation
2196 sequence. For C source files, this is the preprocessor, compiler
2197 proper, and assembler. The output looks like this:
2205 The first number on each line is the ``user time,'' that is time spent
2206 executing the program itself. The second number is ``system time,''
2207 time spent executing operating system routines on behalf of the program.
2208 Both numbers are in seconds.
2210 @item -print-file-name=@var{library}
2211 Print the full absolute name of the library file @var{library} that
2212 would be used when linking---and don't do anything else. With this
2213 option, GCC does not compile or link anything; it just prints the
2216 @item -print-prog-name=@var{program}
2217 Like @samp{-print-file-name}, but searches for a program such as @samp{cpp}.
2219 @item -print-libgcc-file-name
2220 Same as @samp{-print-file-name=libgcc.a}.
2222 This is useful when you use @samp{-nostdlib} or @samp{-nodefaultlibs}
2223 but you do want to link with @file{libgcc.a}. You can do
2226 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
2229 @item -print-search-dirs
2230 Print the name of the configured installation directory and a list of
2231 program and library directories gcc will search---and don't do anything else.
2233 This is useful when gcc prints the error message
2234 @samp{installation problem, cannot exec cpp: No such file or directory}.
2235 To resolve this you either need to put @file{cpp} and the other compiler
2236 components where gcc expects to find them, or you can set the environment
2237 variable @code{GCC_EXEC_PREFIX} to the directory where you installed them.
2238 Don't forget the trailing '/'.
2239 @xref{Environment Variables}.
2242 @node Optimize Options
2243 @section Options That Control Optimization
2244 @cindex optimize options
2245 @cindex options, optimization
2247 These options control various sorts of optimizations:
2252 Optimize. Optimizing compilation takes somewhat more time, and a lot
2253 more memory for a large function.
2255 Without @samp{-O}, the compiler's goal is to reduce the cost of
2256 compilation and to make debugging produce the expected results.
2257 Statements are independent: if you stop the program with a breakpoint
2258 between statements, you can then assign a new value to any variable or
2259 change the program counter to any other statement in the function and
2260 get exactly the results you would expect from the source code.
2262 Without @samp{-O}, the compiler only allocates variables declared
2263 @code{register} in registers. The resulting compiled code is a little
2264 worse than produced by PCC without @samp{-O}.
2266 With @samp{-O}, the compiler tries to reduce code size and execution
2269 When you specify @samp{-O}, the compiler turns on @samp{-fthread-jumps}
2270 and @samp{-fdefer-pop} on all machines. The compiler turns on
2271 @samp{-fdelayed-branch} on machines that have delay slots, and
2272 @samp{-fomit-frame-pointer} on machines that can support debugging even
2273 without a frame pointer. On some machines the compiler also turns
2274 on other flags.@refill
2277 Optimize even more. GCC performs nearly all supported optimizations
2278 that do not involve a space-speed tradeoff. The compiler does not
2279 perform loop unrolling or function inlining when you specify @samp{-O2}.
2280 As compared to @samp{-O}, this option increases both compilation time
2281 and the performance of the generated code.
2283 @samp{-O2} turns on all optional optimizations except for loop unrolling
2284 and function inlining. It also turns on the @samp{-fforce-mem} option
2285 on all machines and frame pointer elimination on machines where doing so
2286 does not interfere with debugging.
2289 Optimize yet more. @samp{-O3} turns on all optimizations specified by
2290 @samp{-O2} and also turns on the @samp{inline-functions} option.
2296 Optimize for size. @samp{-Os} enables all @samp{-O2} optimizations that
2297 do not typically increase code size. It also performs further
2298 optimizations designed to reduce code size.
2300 If you use multiple @samp{-O} options, with or without level numbers,
2301 the last such option is the one that is effective.
2304 Options of the form @samp{-f@var{flag}} specify machine-independent
2305 flags. Most flags have both positive and negative forms; the negative
2306 form of @samp{-ffoo} would be @samp{-fno-foo}. In the table below,
2307 only one of the forms is listed---the one which is not the default.
2308 You can figure out the other form by either removing @samp{no-} or
2313 Do not store floating point variables in registers, and inhibit other
2314 options that might change whether a floating point value is taken from a
2317 @cindex floating point precision
2318 This option prevents undesirable excess precision on machines such as
2319 the 68000 where the floating registers (of the 68881) keep more
2320 precision than a @code{double} is supposed to have. Similarly for the
2321 x86 architecture. For most programs, the excess precision does only
2322 good, but a few programs rely on the precise definition of IEEE floating
2323 point. Use @samp{-ffloat-store} for such programs, after modifying
2324 them to store all pertinent intermediate computations into variables.
2326 @item -fno-default-inline
2327 Do not make member functions inline by default merely because they are
2328 defined inside the class scope (C++ only). Otherwise, when you specify
2329 @w{@samp{-O}}, member functions defined inside class scope are compiled
2330 inline by default; i.e., you don't need to add @samp{inline} in front of
2331 the member function name.
2333 @item -fno-defer-pop
2334 Always pop the arguments to each function call as soon as that function
2335 returns. For machines which must pop arguments after a function call,
2336 the compiler normally lets arguments accumulate on the stack for several
2337 function calls and pops them all at once.
2340 Force memory operands to be copied into registers before doing
2341 arithmetic on them. This produces better code by making all memory
2342 references potential common subexpressions. When they are not common
2343 subexpressions, instruction combination should eliminate the separate
2344 register-load. The @samp{-O2} option turns on this option.
2347 Force memory address constants to be copied into registers before
2348 doing arithmetic on them. This may produce better code just as
2349 @samp{-fforce-mem} may.
2351 @item -fomit-frame-pointer
2352 Don't keep the frame pointer in a register for functions that
2353 don't need one. This avoids the instructions to save, set up and
2354 restore frame pointers; it also makes an extra register available
2355 in many functions. @strong{It also makes debugging impossible on
2359 On some machines, such as the Vax, this flag has no effect, because
2360 the standard calling sequence automatically handles the frame pointer
2361 and nothing is saved by pretending it doesn't exist. The
2362 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2363 whether a target machine supports this flag. @xref{Registers}.@refill
2366 On some machines, such as the Vax, this flag has no effect, because
2367 the standard calling sequence automatically handles the frame pointer
2368 and nothing is saved by pretending it doesn't exist. The
2369 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2370 whether a target machine supports this flag. @xref{Registers,,Register
2371 Usage, gcc.info, Using and Porting GCC}.@refill
2375 Don't pay attention to the @code{inline} keyword. Normally this option
2376 is used to keep the compiler from expanding any functions inline.
2377 Note that if you are not optimizing, no functions can be expanded inline.
2379 @item -finline-functions
2380 Integrate all simple functions into their callers. The compiler
2381 heuristically decides which functions are simple enough to be worth
2382 integrating in this way.
2384 If all calls to a given function are integrated, and the function is
2385 declared @code{static}, then the function is normally not output as
2386 assembler code in its own right.
2388 @item -finline-limit=@var{n}
2389 By default, gcc limits the size of functions that can be inlined. This flag
2390 allows the control of this limit for functions that are explicitly marked as
2391 inline (ie marked with the inline keyword or defined within the class
2392 definition in c++). @var{n} is the size of functions that can be inlined in
2393 number of pseudo instructions (not counting parameter handling). The default
2394 value of n is 10000. Increasing this value can result in more inlined code at
2395 the cost of compilation time and memory consumption. Decreasing usually makes
2396 the compilation faster and less code will be inlined (which presumably
2397 means slower programs). This option is particularly useful for programs that
2398 use inlining heavily such as those based on recursive templates with c++.
2400 @emph{Note:} pseudo instruction represents, in this particular context, an
2401 abstract measurement of function's size. In no way, it represents a count
2402 of assembly instructions and as such its exact meaning might change from one
2403 release to an another.
2405 @item -fkeep-inline-functions
2406 Even if all calls to a given function are integrated, and the function
2407 is declared @code{static}, nevertheless output a separate run-time
2408 callable version of the function. This switch does not affect
2409 @code{extern inline} functions.
2411 @item -fkeep-static-consts
2412 Emit variables declared @code{static const} when optimization isn't turned
2413 on, even if the variables aren't referenced.
2415 GCC enables this option by default. If you want to force the compiler to
2416 check if the variable was referenced, regardless of whether or not
2417 optimization is turned on, use the @samp{-fno-keep-static-consts} option.
2419 @item -fno-function-cse
2420 Do not put function addresses in registers; make each instruction that
2421 calls a constant function contain the function's address explicitly.
2423 This option results in less efficient code, but some strange hacks
2424 that alter the assembler output may be confused by the optimizations
2425 performed when this option is not used.
2428 This option allows GCC to violate some ANSI or IEEE rules and/or
2429 specifications in the interest of optimizing code for speed. For
2430 example, it allows the compiler to assume arguments to the @code{sqrt}
2431 function are non-negative numbers and that no floating-point values
2434 This option should never be turned on by any @samp{-O} option since
2435 it can result in incorrect output for programs which depend on
2436 an exact implementation of IEEE or ANSI rules/specifications for
2439 @item -fno-math-errno
2440 Do not set ERRNO after calling math functions that are executed
2441 with a single instruction, e.g., sqrt. A program that relies on
2442 IEEE exceptions for math error handling may want to use this flag
2443 for speed while maintaining IEEE arithmetic compatibility.
2445 The default is @samp{-fmath-errno}. The @samp{-ffast-math} option
2446 sets @samp{-fno-math-errno}.
2449 @c following causes underfulls.. they don't look great, but we deal.
2451 The following options control specific optimizations. The @samp{-O2}
2452 option turns on all of these optimizations except @samp{-funroll-loops}
2453 and @samp{-funroll-all-loops}. On most machines, the @samp{-O} option
2454 turns on the @samp{-fthread-jumps} and @samp{-fdelayed-branch} options,
2455 but specific machines may handle it differently.
2457 You can use the following flags in the rare cases when ``fine-tuning''
2458 of optimizations to be performed is desired.
2461 @item -fstrength-reduce
2462 Perform the optimizations of loop strength reduction and
2463 elimination of iteration variables.
2465 @item -fthread-jumps
2466 Perform optimizations where we check to see if a jump branches to a
2467 location where another comparison subsumed by the first is found. If
2468 so, the first branch is redirected to either the destination of the
2469 second branch or a point immediately following it, depending on whether
2470 the condition is known to be true or false.
2472 @item -fcse-follow-jumps
2473 In common subexpression elimination, scan through jump instructions
2474 when the target of the jump is not reached by any other path. For
2475 example, when CSE encounters an @code{if} statement with an
2476 @code{else} clause, CSE will follow the jump when the condition
2479 @item -fcse-skip-blocks
2480 This is similar to @samp{-fcse-follow-jumps}, but causes CSE to
2481 follow jumps which conditionally skip over blocks. When CSE
2482 encounters a simple @code{if} statement with no else clause,
2483 @samp{-fcse-skip-blocks} causes CSE to follow the jump around the
2484 body of the @code{if}.
2486 @item -frerun-cse-after-loop
2487 Re-run common subexpression elimination after loop optimizations has been
2490 @item -frerun-loop-opt
2491 Run the loop optimizer twice.
2494 Perform a global common subexpression elimination pass.
2495 This pass also performs global constant and copy propagation.
2497 @item -fdelete-null-pointer-checks
2498 Use global dataflow analysis to identify and eliminate useless null
2499 pointer checks. Programs which rely on NULL pointer dereferences @emph{not}
2500 halting the program may not work properly with this option. Use
2501 -fno-delete-null-pointer-checks to disable this optimizing for programs
2502 which depend on that behavior.
2505 @item -fexpensive-optimizations
2506 Perform a number of minor optimizations that are relatively expensive.
2508 @item -foptimize-register-moves
2510 Attempt to reassign register numbers in move instructions and as
2511 operands of other simple instructions in order to maximize the amount of
2512 register tying. This is especially helpful on machines with two-operand
2513 instructions. GCC enables this optimization by default with @samp{-O2}
2516 Note @code{-fregmove} and @code{-foptimize-register-moves} are the same
2519 @item -fdelayed-branch
2520 If supported for the target machine, attempt to reorder instructions
2521 to exploit instruction slots available after delayed branch
2524 @item -fschedule-insns
2525 If supported for the target machine, attempt to reorder instructions to
2526 eliminate execution stalls due to required data being unavailable. This
2527 helps machines that have slow floating point or memory load instructions
2528 by allowing other instructions to be issued until the result of the load
2529 or floating point instruction is required.
2531 @item -fschedule-insns2
2532 Similar to @samp{-fschedule-insns}, but requests an additional pass of
2533 instruction scheduling after register allocation has been done. This is
2534 especially useful on machines with a relatively small number of
2535 registers and where memory load instructions take more than one cycle.
2537 @item -ffunction-sections
2538 @itemx -fdata-sections
2539 Place each function or data item into its own section in the output
2540 file if the target supports arbitrary sections. The name of the
2541 function or the name of the data item determines the section's name
2544 Use these options on systems where the linker can perform optimizations
2545 to improve locality of reference in the instruction space. HPPA
2546 processors running HP-UX and Sparc processors running Solaris 2 have
2547 linkers with such optimizations. Other systems using the ELF object format
2548 as well as AIX may have these optimizations in the future.
2550 Only use these options when there are significant benefits from doing
2551 so. When you specify these options, the assembler and linker will
2552 create larger object and executable files and will also be slower.
2553 You will not be able to use @code{gprof} on all systems if you
2554 specify this option and you may have problems with debugging if
2555 you specify both this option and @samp{-g}.
2557 @item -fcaller-saves
2558 Enable values to be allocated in registers that will be clobbered by
2559 function calls, by emitting extra instructions to save and restore the
2560 registers around such calls. Such allocation is done only when it
2561 seems to result in better code than would otherwise be produced.
2563 This option is always enabled by default on certain machines, usually
2564 those which have no call-preserved registers to use instead.
2566 For all machines, optimization level 2 and higher enables this flag by
2569 @item -funroll-loops
2570 Perform the optimization of loop unrolling. This is only done for loops
2571 whose number of iterations can be determined at compile time or run time.
2572 @samp{-funroll-loops} implies both @samp{-fstrength-reduce} and
2573 @samp{-frerun-cse-after-loop}.
2575 @item -funroll-all-loops
2576 Perform the optimization of loop unrolling. This is done for all loops
2577 and usually makes programs run more slowly. @samp{-funroll-all-loops}
2578 implies @samp{-fstrength-reduce} as well as @samp{-frerun-cse-after-loop}.
2580 @item -fmove-all-movables
2581 Forces all invariant computations in loops to be moved
2584 @item -freduce-all-givs
2585 Forces all general-induction variables in loops to be
2588 @emph{Note:} When compiling programs written in Fortran,
2589 @samp{-fmove-all-movables} and @samp{-freduce-all-givs} are enabled
2590 by default when you use the optimizer.
2592 These options may generate better or worse code; results are highly
2593 dependent on the structure of loops within the source code.
2595 These two options are intended to be removed someday, once
2596 they have helped determine the efficacy of various
2597 approaches to improving loop optimizations.
2599 Please let us (@code{gcc@@gcc.gnu.org} and @code{fortran@@gnu.org})
2600 know how use of these options affects
2601 the performance of your production code.
2602 We're very interested in code that runs @emph{slower}
2603 when these options are @emph{enabled}.
2606 Disable any machine-specific peephole optimizations.
2608 @item -fbranch-probabilities
2609 After running a program compiled with @samp{-fprofile-arcs}
2610 (@pxref{Debugging Options,, Options for Debugging Your Program or
2611 @code{gcc}}), you can compile it a second time using
2612 @samp{-fbranch-probabilities}, to improve optimizations based on
2613 guessing the path a branch might take.
2616 With @samp{-fbranch-probabilities}, GCC puts a @samp{REG_EXEC_COUNT}
2617 note on the first instruction of each basic block, and a
2618 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
2619 These can be used to improve optimization. Currently, they are only
2620 used in one place: in @file{reorg.c}, instead of guessing which path a
2621 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
2622 exactly determine which path is taken more often.
2625 @item -fstrict-aliasing
2626 Allows the compiler to assume the strictest aliasing rules applicable to
2627 the language being compiled. For C (and C++), this activates
2628 optimizations based on the type of expressions. In particular, an
2629 object of one type is assumed never to reside at the same address as an
2630 object of a different type, unless the types are almost the same. For
2631 example, an @code{unsigned int} can alias an @code{int}, but not a
2632 @code{void*} or a @code{double}. A character type may alias any other
2635 Pay special attention to code like this:
2648 The practice of reading from a different union member than the one most
2649 recently written to (called ``type-punning'') is common. Even with
2650 @samp{-fstrict-aliasing}, type-punning is allowed, provided the memory
2651 is accessed through the union type. So, the code above will work as
2652 expected. However, this code might not:
2664 Every language that wishes to perform language-specific alias analysis
2665 should define a function that computes, given an @code{tree}
2666 node, an alias set for the node. Nodes in different alias sets are not
2667 allowed to alias. For an example, see the C front-end function
2668 @code{c_get_alias_set}.
2671 @item -falign-functions
2672 @itemx -falign-functions=@var{n}
2673 Align the start of functions to the next power-of-two greater than
2674 @var{n}, skipping up to @var{n} bytes. For instance,
2675 @samp{-falign-functions=32} aligns functions to the next 32-byte
2676 boundary, but @samp{-falign-functions=24} would align to the next
2677 32-byte boundary only if this can be done by skipping 23 bytes or less.
2679 @samp{-fno-align-functions} and @samp{-falign-functions=1} are
2680 equivalent and mean that functions will not be aligned.
2682 Some assemblers only support this flag when @var{n} is a power of two;
2683 in that case, it is rounded up.
2685 If @var{n} is not specified, use a machine-dependent default.
2687 @item -falign-labels
2688 @itemx -falign-labels=@var{n}
2689 Align all branch targets to a power-of-two boundary, skipping up to
2690 @var{n} bytes like @samp{-falign-functions}. This option can easily
2691 make code slower, because it must insert dummy operations for when the
2692 branch target is reached in the usual flow of the code.
2694 If @samp{-falign-loops} or @samp{-falign-jumps} are applicable and
2695 are greater than this value, then their values are used instead.
2697 If @var{n} is not specified, use a machine-dependent default which is
2698 very likely to be @samp{1}, meaning no alignment.
2701 @itemx -falign-loops=@var{n}
2702 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
2703 like @samp{-falign-functions}. The hope is that the loop will be
2704 executed many times, which will make up for any execution of the dummy
2707 If @var{n} is not specified, use a machine-dependent default.
2710 @itemx -falign-jumps=@var{n}
2711 Align branch targets to a power-of-two boundary, for branch targets
2712 where the targets can only be reached by jumping, skipping up to @var{n}
2713 bytes like @samp{-falign-functions}. In this case, no dummy operations
2716 If @var{n} is not specified, use a machine-dependent default.
2720 @node Preprocessor Options
2721 @section Options Controlling the Preprocessor
2722 @cindex preprocessor options
2723 @cindex options, preprocessor
2725 These options control the C preprocessor, which is run on each C source
2726 file before actual compilation.
2728 If you use the @samp{-E} option, nothing is done except preprocessing.
2729 Some of these options make sense only together with @samp{-E} because
2730 they cause the preprocessor output to be unsuitable for actual
2734 @item -include @var{file}
2735 Process @var{file} as input before processing the regular input file.
2736 In effect, the contents of @var{file} are compiled first. Any @samp{-D}
2737 and @samp{-U} options on the command line are always processed before
2738 @samp{-include @var{file}}, regardless of the order in which they are
2739 written. All the @samp{-include} and @samp{-imacros} options are
2740 processed in the order in which they are written.
2742 @item -imacros @var{file}
2743 Process @var{file} as input, discarding the resulting output, before
2744 processing the regular input file. Because the output generated from
2745 @var{file} is discarded, the only effect of @samp{-imacros @var{file}}
2746 is to make the macros defined in @var{file} available for use in the
2749 Any @samp{-D} and @samp{-U} options on the command line are always
2750 processed before @samp{-imacros @var{file}}, regardless of the order in
2751 which they are written. All the @samp{-include} and @samp{-imacros}
2752 options are processed in the order in which they are written.
2754 @item -idirafter @var{dir}
2755 @cindex second include path
2756 Add the directory @var{dir} to the second include path. The directories
2757 on the second include path are searched when a header file is not found
2758 in any of the directories in the main include path (the one that
2761 @item -iprefix @var{prefix}
2762 Specify @var{prefix} as the prefix for subsequent @samp{-iwithprefix}
2765 @item -iwithprefix @var{dir}
2766 Add a directory to the second include path. The directory's name is
2767 made by concatenating @var{prefix} and @var{dir}, where @var{prefix} was
2768 specified previously with @samp{-iprefix}. If you have not specified a
2769 prefix yet, the directory containing the installed passes of the
2770 compiler is used as the default.
2772 @item -iwithprefixbefore @var{dir}
2773 Add a directory to the main include path. The directory's name is made
2774 by concatenating @var{prefix} and @var{dir}, as in the case of
2775 @samp{-iwithprefix}.
2777 @item -isystem @var{dir}
2778 Add a directory to the beginning of the second include path, marking it
2779 as a system directory, so that it gets the same special treatment as
2780 is applied to the standard system directories.
2783 Do not search the standard system directories for header files. Only
2784 the directories you have specified with @samp{-I} options (and the
2785 current directory, if appropriate) are searched. @xref{Directory
2786 Options}, for information on @samp{-I}.
2788 By using both @samp{-nostdinc} and @samp{-I-}, you can limit the include-file
2789 search path to only those directories you specify explicitly.
2792 Do not predefine any nonstandard macros. (Including architecture flags).
2795 Run only the C preprocessor. Preprocess all the C source files
2796 specified and output the results to standard output or to the
2797 specified output file.
2800 Tell the preprocessor not to discard comments. Used with the
2804 Tell the preprocessor not to generate @samp{#line} directives.
2805 Used with the @samp{-E} option.
2808 @cindex dependencies, make
2810 Tell the preprocessor to output a rule suitable for @code{make}
2811 describing the dependencies of each object file. For each source file,
2812 the preprocessor outputs one @code{make}-rule whose target is the object
2813 file name for that source file and whose dependencies are all the
2814 @code{#include} header files it uses. This rule may be a single line or
2815 may be continued with @samp{\}-newline if it is long. The list of rules
2816 is printed on standard output instead of the preprocessed C program.
2818 @samp{-M} implies @samp{-E}.
2820 Another way to specify output of a @code{make} rule is by setting
2821 the environment variable @code{DEPENDENCIES_OUTPUT} (@pxref{Environment
2825 Like @samp{-M} but the output mentions only the user header files
2826 included with @samp{#include "@var{file}"}. System header files
2827 included with @samp{#include <@var{file}>} are omitted.
2830 Like @samp{-M} but the dependency information is written to a file made by
2831 replacing ".c" with ".d" at the end of the input file names.
2832 This is in addition to compiling the file as specified---@samp{-MD} does
2833 not inhibit ordinary compilation the way @samp{-M} does.
2835 In Mach, you can use the utility @code{md} to merge multiple dependency
2836 files into a single dependency file suitable for using with the @samp{make}
2840 Like @samp{-MD} except mention only user header files, not system
2844 Treat missing header files as generated files and assume they live in the
2845 same directory as the source file. If you specify @samp{-MG}, you
2846 must also specify either @samp{-M} or @samp{-MM}. @samp{-MG} is not
2847 supported with @samp{-MD} or @samp{-MMD}.
2850 Print the name of each header file used, in addition to other normal
2853 @item -A@var{question}(@var{answer})
2854 Assert the answer @var{answer} for @var{question}, in case it is tested
2855 with a preprocessing conditional such as @samp{#if
2856 #@var{question}(@var{answer})}. @samp{-A-} disables the standard
2857 assertions that normally describe the target machine.
2860 Define macro @var{macro} with the string @samp{1} as its definition.
2862 @item -D@var{macro}=@var{defn}
2863 Define macro @var{macro} as @var{defn}. All instances of @samp{-D} on
2864 the command line are processed before any @samp{-U} options.
2867 Undefine macro @var{macro}. @samp{-U} options are evaluated after all
2868 @samp{-D} options, but before any @samp{-include} and @samp{-imacros}
2872 Tell the preprocessor to output only a list of the macro definitions
2873 that are in effect at the end of preprocessing. Used with the @samp{-E}
2877 Tell the preprocessing to pass all macro definitions into the output, in
2878 their proper sequence in the rest of the output.
2881 Like @samp{-dD} except that the macro arguments and contents are omitted.
2882 Only @samp{#define @var{name}} is included in the output.
2885 Support ANSI C trigraphs. The @samp{-ansi} option also has this effect.
2887 @item -Wp,@var{option}
2888 Pass @var{option} as an option to the preprocessor. If @var{option}
2889 contains commas, it is split into multiple options at the commas.
2892 @node Assembler Options
2893 @section Passing Options to the Assembler
2895 @c prevent bad page break with this line
2896 You can pass options to the assembler.
2899 @item -Wa,@var{option}
2900 Pass @var{option} as an option to the assembler. If @var{option}
2901 contains commas, it is split into multiple options at the commas.
2905 @section Options for Linking
2906 @cindex link options
2907 @cindex options, linking
2909 These options come into play when the compiler links object files into
2910 an executable output file. They are meaningless if the compiler is
2911 not doing a link step.
2915 @item @var{object-file-name}
2916 A file name that does not end in a special recognized suffix is
2917 considered to name an object file or library. (Object files are
2918 distinguished from libraries by the linker according to the file
2919 contents.) If linking is done, these object files are used as input
2925 If any of these options is used, then the linker is not run, and
2926 object file names should not be used as arguments. @xref{Overall
2930 @item -l@var{library}
2931 Search the library named @var{library} when linking.
2933 It makes a difference where in the command you write this option; the
2934 linker searches processes libraries and object files in the order they
2935 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
2936 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
2937 to functions in @samp{z}, those functions may not be loaded.
2939 The linker searches a standard list of directories for the library,
2940 which is actually a file named @file{lib@var{library}.a}. The linker
2941 then uses this file as if it had been specified precisely by name.
2943 The directories searched include several standard system directories
2944 plus any that you specify with @samp{-L}.
2946 Normally the files found this way are library files---archive files
2947 whose members are object files. The linker handles an archive file by
2948 scanning through it for members which define symbols that have so far
2949 been referenced but not defined. But if the file that is found is an
2950 ordinary object file, it is linked in the usual fashion. The only
2951 difference between using an @samp{-l} option and specifying a file name
2952 is that @samp{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
2953 and searches several directories.
2956 You need this special case of the @samp{-l} option in order to
2957 link an Objective C program.
2960 Do not use the standard system startup files when linking.
2961 The standard system libraries are used normally, unless @code{-nostdlib}
2962 or @code{-nodefaultlibs} is used.
2964 @item -nodefaultlibs
2965 Do not use the standard system libraries when linking.
2966 Only the libraries you specify will be passed to the linker.
2967 The standard startup files are used normally, unless @code{-nostartfiles}
2968 is used. The compiler may generate calls to memcmp, memset, and memcpy
2969 for System V (and ANSI C) environments or to bcopy and bzero for
2970 BSD environments. These entries are usually resolved by entries in
2971 libc. These entry points should be supplied through some other
2972 mechanism when this option is specified.
2975 Do not use the standard system startup files or libraries when linking.
2976 No startup files and only the libraries you specify will be passed to
2977 the linker. The compiler may generate calls to memcmp, memset, and memcpy
2978 for System V (and ANSI C) environments or to bcopy and bzero for
2979 BSD environments. These entries are usually resolved by entries in
2980 libc. These entry points should be supplied through some other
2981 mechanism when this option is specified.
2983 @cindex @code{-lgcc}, use with @code{-nostdlib}
2984 @cindex @code{-nostdlib} and unresolved references
2985 @cindex unresolved references and @code{-nostdlib}
2986 @cindex @code{-lgcc}, use with @code{-nodefaultlibs}
2987 @cindex @code{-nodefaultlibs} and unresolved references
2988 @cindex unresolved references and @code{-nodefaultlibs}
2989 One of the standard libraries bypassed by @samp{-nostdlib} and
2990 @samp{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
2991 that GCC uses to overcome shortcomings of particular machines, or special
2992 needs for some languages.
2994 (@xref{Interface,,Interfacing to GCC Output}, for more discussion of
2998 (@xref{Interface,,Interfacing to GCC Output,gcc.info,Porting GCC},
2999 for more discussion of @file{libgcc.a}.)
3001 In most cases, you need @file{libgcc.a} even when you want to avoid
3002 other standard libraries. In other words, when you specify @samp{-nostdlib}
3003 or @samp{-nodefaultlibs} you should usually specify @samp{-lgcc} as well.
3004 This ensures that you have no unresolved references to internal GCC
3005 library subroutines. (For example, @samp{__main}, used to ensure C++
3006 constructors will be called; @pxref{Collect2,,@code{collect2}}.)
3009 Remove all symbol table and relocation information from the executable.
3012 On systems that support dynamic linking, this prevents linking with the shared
3013 libraries. On other systems, this option has no effect.
3016 Produce a shared object which can then be linked with other objects to
3017 form an executable. Not all systems support this option. You must
3018 also specify @samp{-fpic} or @samp{-fPIC} on some systems when
3019 you specify this option.
3022 Bind references to global symbols when building a shared object. Warn
3023 about any unresolved references (unless overridden by the link editor
3024 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
3027 @item -Xlinker @var{option}
3028 Pass @var{option} as an option to the linker. You can use this to
3029 supply system-specific linker options which GCC does not know how to
3032 If you want to pass an option that takes an argument, you must use
3033 @samp{-Xlinker} twice, once for the option and once for the argument.
3034 For example, to pass @samp{-assert definitions}, you must write
3035 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
3036 @samp{-Xlinker "-assert definitions"}, because this passes the entire
3037 string as a single argument, which is not what the linker expects.
3039 @item -Wl,@var{option}
3040 Pass @var{option} as an option to the linker. If @var{option} contains
3041 commas, it is split into multiple options at the commas.
3043 @item -u @var{symbol}
3044 Pretend the symbol @var{symbol} is undefined, to force linking of
3045 library modules to define it. You can use @samp{-u} multiple times with
3046 different symbols to force loading of additional library modules.
3049 @node Directory Options
3050 @section Options for Directory Search
3051 @cindex directory options
3052 @cindex options, directory search
3055 These options specify directories to search for header files, for
3056 libraries and for parts of the compiler:
3060 Add the directory @var{dir} to the head of the list of directories to be
3061 searched for header files. This can be used to override a system header
3062 file, substituting your own version, since these directories are
3063 searched before the system header file directories. If you use more
3064 than one @samp{-I} option, the directories are scanned in left-to-right
3065 order; the standard system directories come after.
3068 Any directories you specify with @samp{-I} options before the @samp{-I-}
3069 option are searched only for the case of @samp{#include "@var{file}"};
3070 they are not searched for @samp{#include <@var{file}>}.
3072 If additional directories are specified with @samp{-I} options after
3073 the @samp{-I-}, these directories are searched for all @samp{#include}
3074 directives. (Ordinarily @emph{all} @samp{-I} directories are used
3077 In addition, the @samp{-I-} option inhibits the use of the current
3078 directory (where the current input file came from) as the first search
3079 directory for @samp{#include "@var{file}"}. There is no way to
3080 override this effect of @samp{-I-}. With @samp{-I.} you can specify
3081 searching the directory which was current when the compiler was
3082 invoked. That is not exactly the same as what the preprocessor does
3083 by default, but it is often satisfactory.
3085 @samp{-I-} does not inhibit the use of the standard system directories
3086 for header files. Thus, @samp{-I-} and @samp{-nostdinc} are
3090 Add directory @var{dir} to the list of directories to be searched
3093 @item -B@var{prefix}
3094 This option specifies where to find the executables, libraries,
3095 include files, and data files of the compiler itself.
3097 The compiler driver program runs one or more of the subprograms
3098 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
3099 @var{prefix} as a prefix for each program it tries to run, both with and
3100 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
3102 For each subprogram to be run, the compiler driver first tries the
3103 @samp{-B} prefix, if any. If that name is not found, or if @samp{-B}
3104 was not specified, the driver tries two standard prefixes, which are
3105 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc-lib/}. If neither of
3106 those results in a file name that is found, the unmodified program
3107 name is searched for using the directories specified in your
3108 @samp{PATH} environment variable.
3110 @samp{-B} prefixes that effectively specify directory names also apply
3111 to libraries in the linker, because the compiler translates these
3112 options into @samp{-L} options for the linker. They also apply to
3113 includes files in the preprocessor, because the compiler translates these
3114 options into @samp{-isystem} options for the preprocessor. In this case,
3115 the compiler appends @samp{include} to the prefix.
3117 The run-time support file @file{libgcc.a} can also be searched for using
3118 the @samp{-B} prefix, if needed. If it is not found there, the two
3119 standard prefixes above are tried, and that is all. The file is left
3120 out of the link if it is not found by those means.
3122 Another way to specify a prefix much like the @samp{-B} prefix is to use
3123 the environment variable @code{GCC_EXEC_PREFIX}. @xref{Environment
3126 @item -specs=@var{file}
3127 Process @var{file} after the compiler reads in the standard @file{specs}
3128 file, in order to override the defaults that the @file{gcc} driver
3129 program uses when determining what switches to pass to @file{cc1},
3130 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
3131 @samp{-specs=}@var{file} can be specified on the command line, and they
3132 are processed in order, from left to right.
3136 @section Specifying subprocesses and the switches to pass to them
3138 @code{GCC} is a driver program. It performs its job by invoking a
3139 sequence of other programs to do the work of compiling, assembling and
3140 linking. GCC interprets its command-line parameters and uses these to
3141 deduce which programs it should invoke, and which command-line options
3142 it ought to place on their command lines. This behaviour is controlled
3143 by @dfn{spec strings}. In most cases there is one spec string for each
3144 program that GCC can invoke, but a few programs have multiple spec
3145 strings to control their behaviour. The spec strings built into GCC can
3146 be overridden by using the @samp{-specs=} command-line switch to specify
3149 @dfn{Spec files} are plaintext files that are used to construct spec
3150 strings. They consist of a sequence of directives separated by blank
3151 lines. The type of directive is determined by the first non-whitespace
3152 character on the line and it can be one of the following:
3155 @item %@var{command}
3156 Issues a @var{command} to the spec file processor. The commands that can
3160 @item %include <@var{file}>
3162 Search for @var{file} and insert its text at the current point in the
3165 @item %include_noerr <@var{file}>
3166 @cindex %include_noerr
3167 Just like @samp{%include}, but do not generate an error message if the include
3168 file cannot be found.
3170 @item %rename @var{old_name} @var{new_name}
3172 Rename the spec string @var{old_name} to @var{new_name}.
3176 @item *[@var{spec_name}]:
3177 This tells the compiler to create, override or delete the named spec
3178 string. All lines after this directive up to the next directive or
3179 blank line are considered to be the text for the spec string. If this
3180 results in an empty string then the spec will be deleted. (Or, if the
3181 spec did not exist, then nothing will happened.) Otherwise, if the spec
3182 does not currently exist a new spec will be created. If the spec does
3183 exist then its contents will be overridden by the text of this
3184 directive, unless the first character of that text is the @samp{+}
3185 character, in which case the text will be appended to the spec.
3187 @item [@var{suffix}]:
3188 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
3189 and up to the next directive or blank line are considered to make up the
3190 spec string for the indicated suffix. When the compiler encounters an
3191 input file with the named suffix, it will processes the spec string in
3192 order to work out how to compile that file. For example:
3199 This says that any input file whose name ends in @samp{.ZZ} should be
3200 passed to the program @samp{z-compile}, which should be invoked with the
3201 command-line switch @samp{-input} and with the result of performing the
3202 @samp{%i} substitution. (See below.)
3204 As an alternative to providing a spec string, the text that follows a
3205 suffix directive can be one of the following:
3208 @item @@@var{language}
3209 This says that the suffix is an alias for a known @var{language}. This is
3210 similar to using the @code{-x} command-line switch to GCC to specify a
3211 language explicitly. For example:
3218 Says that .ZZ files are, in fact, C++ source files.
3221 This causes an error messages saying:
3224 @var{name} compiler not installed on this system.
3228 GCC already has an extensive list of suffixes built into it.
3229 This directive will add an entry to the end of the list of suffixes, but
3230 since the list is searched from the end backwards, it is effectively
3231 possible to override earlier entries using this technique.
3235 GCC has the following spec strings built into it. Spec files can
3236 override these strings or create their own. Note that individual
3237 targets can also add their own spec strings to this list.
3240 asm Options to pass to the assembler
3241 asm_final Options to pass to the assembler post-processor
3242 cpp Options to pass to the C preprocessor
3243 cc1 Options to pass to the C compiler
3244 cc1plus Options to pass to the C++ compiler
3245 endfile Object files to include at the end of the link
3246 link Options to pass to the linker
3247 lib Libraries to include on the command line to the linker
3248 libgcc Decides which GCC support library to pass to the linker
3249 linker Sets the name of the linker
3250 predefines Defines to be passed to the C preprocessor
3251 signed_char Defines to pass to CPP to say whether @code{char} is signed by default
3252 startfile Object files to include at the start of the link
3255 Here is a small example of a spec file:
3261 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
3264 This example renames the spec called @samp{lib} to @samp{old_lib} and
3265 then overrides the previous definition of @samp{lib} with a new one.
3266 The new definition adds in some extra command-line options before
3267 including the text of the old definition.
3269 @dfn{Spec strings} are a list of command-line options to be passed to their
3270 corresponding program. In addition, the spec strings can contain
3271 @samp{%}-prefixed sequences to substitute variable text or to
3272 conditionally insert text into the command line. Using these constructs
3273 it is possible to generate quite complex command lines.
3275 Here is a table of all defined @samp{%}-sequences for spec
3276 strings. Note that spaces are not generated automatically around the
3277 results of expanding these sequences. Therefore you can concatenate them
3278 together or combine them with constant text in a single argument.
3282 Substitute one @samp{%} into the program name or argument.
3285 Substitute the name of the input file being processed.
3288 Substitute the basename of the input file being processed.
3289 This is the substring up to (and not including) the last period
3290 and not including the directory.
3293 Marks the argument containing or following the @samp{%d} as a
3294 temporary file name, so that that file will be deleted if GCC exits
3295 successfully. Unlike @samp{%g}, this contributes no text to the
3298 @item %g@var{suffix}
3299 Substitute a file name that has suffix @var{suffix} and is chosen
3300 once per compilation, and mark the argument in the same way as
3301 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
3302 name is now chosen in a way that is hard to predict even when previously
3303 chosen file names are known. For example, @samp{%g.s ... %g.o ... %g.s}
3304 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
3305 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
3306 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
3307 was simply substituted with a file name chosen once per compilation,
3308 without regard to any appended suffix (which was therefore treated
3309 just like ordinary text), making such attacks more likely to succeed.
3311 @item %u@var{suffix}
3312 Like @samp{%g}, but generates a new temporary file name even if
3313 @samp{%u@var{suffix}} was already seen.
3315 @item %U@var{suffix}
3316 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
3317 new one if there is no such last file name. In the absence of any
3318 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
3319 the same suffix @emph{space}, so @samp{%g.s ... %U.s ... %g.s ... %U.s}
3320 would involve the generation of two distinct file names, one
3321 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
3322 simply substituted with a file name chosen for the previous @samp{%u},
3323 without regard to any appended suffix.
3326 Marks the argument containing or following the @samp{%w} as the
3327 designated output file of this compilation. This puts the argument
3328 into the sequence of arguments that @samp{%o} will substitute later.
3331 Substitutes the names of all the output files, with spaces
3332 automatically placed around them. You should write spaces
3333 around the @samp{%o} as well or the results are undefined.
3334 @samp{%o} is for use in the specs for running the linker.
3335 Input files whose names have no recognized suffix are not compiled
3336 at all, but they are included among the output files, so they will
3340 Substitutes the suffix for object files. Note that this is
3341 handled specially when it immediately follows @samp{%g, %u, or %U},
3342 because of the need for those to form complete file names. The
3343 handling is such that @samp{%O} is treated exactly as if it had already
3344 been substituted, except that @samp{%g, %u, and %U} do not currently
3345 support additional @var{suffix} characters following @samp{%O} as they would
3346 following, for example, @samp{.o}.
3349 Substitutes the standard macro predefinitions for the
3350 current target machine. Use this when running @code{cpp}.
3353 Like @samp{%p}, but puts @samp{__} before and after the name of each
3354 predefined macro, except for macros that start with @samp{__} or with
3355 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ANSI
3359 Substitute a @samp{-iprefix} option made from GCC_EXEC_PREFIX.
3362 Current argument is the name of a library or startup file of some sort.
3363 Search for that file in a standard list of directories and substitute
3364 the full name found.
3367 Print @var{str} as an error message. @var{str} is terminated by a newline.
3368 Use this when inconsistent options are detected.
3371 Output @samp{-} if the input for the current command is coming from a pipe.
3374 Substitute the contents of spec string @var{name} at this point.
3377 Like @samp{%(...)} but put @samp{__} around @samp{-D} arguments.
3379 @item %x@{@var{option}@}
3380 Accumulate an option for @samp{%X}.
3383 Output the accumulated linker options specified by @samp{-Wl} or a @samp{%x}
3387 Output the accumulated assembler options specified by @samp{-Wa}.
3390 Output the accumulated preprocessor options specified by @samp{-Wp}.
3393 Substitute the major version number of GCC.
3394 (For version 2.9.5, this is 2.)
3397 Substitute the minor version number of GCC.
3398 (For version 2.9.5, this is 9.)
3401 Process the @code{asm} spec. This is used to compute the
3402 switches to be passed to the assembler.
3405 Process the @code{asm_final} spec. This is a spec string for
3406 passing switches to an assembler post-processor, if such a program is
3410 Process the @code{link} spec. This is the spec for computing the
3411 command line passed to the linker. Typically it will make use of the
3412 @samp{%L %G %S %D and %E} sequences.
3415 Dump out a @samp{-L} option for each directory that GCC believes might
3416 contain startup files. If the target supports multilibs then the
3417 current multilib directory will be prepended to each of these paths.
3420 Process the @code{lib} spec. This is a spec string for deciding which
3421 libraries should be included on the command line to the linker.
3424 Process the @code{libgcc} spec. This is a spec string for deciding
3425 which GCC support library should be included on the command line to the linker.
3428 Process the @code{startfile} spec. This is a spec for deciding which
3429 object files should be the first ones passed to the linker. Typically
3430 this might be a file named @file{crt0.o}.
3433 Process the @code{endfile} spec. This is a spec string that specifies
3434 the last object files that will be passed to the linker.
3437 Process the @code{cpp} spec. This is used to construct the arguments
3438 to be passed to the C preprocessor.
3441 Process the @code{signed_char} spec. This is intended to be used
3442 to tell cpp whether a char is signed. It typically has the definition:
3444 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
3448 Process the @code{cc1} spec. This is used to construct the options to be
3449 passed to the actual C compiler (@samp{cc1}).
3452 Process the @code{cc1plus} spec. This is used to construct the options to be
3453 passed to the actual C++ compiler (@samp{cc1plus}).
3456 Substitute the variable part of a matched option. See below.
3457 Note that each comma in the substituted string is replaced by
3461 Substitutes the @code{-S} switch, if that switch was given to GCC.
3462 If that switch was not specified, this substitutes nothing. Note that
3463 the leading dash is omitted when specifying this option, and it is
3464 automatically inserted if the substitution is performed. Thus the spec
3465 string @samp{%@{foo@}} would match the command-line option @samp{-foo}
3466 and would output the command line option @samp{-foo}.
3468 @item %W@{@code{S}@}
3469 Like %@{@code{S}@} but mark last argument supplied within as a file to be
3472 @item %@{@code{S}*@}
3473 Substitutes all the switches specified to GCC whose names start
3474 with @code{-S}, but which also take an argument. This is used for
3475 switches like @samp{-o, -D, -I}, etc. GCC considers @samp{-o foo} as being
3476 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
3477 text, including the space. Thus two arguments would be generated.
3479 @item %@{^@code{S}*@}
3480 Like %@{@code{S}*@}, but don't put a blank between a switch and its
3481 argument. Thus %@{^o*@} would only generate one argument, not two.
3483 @item %@{@code{S}*:@code{X}@}
3484 Substitutes @code{X} if one or more switches whose names start with
3485 @code{-S} are specified to GCC. Note that the tail part of the
3486 @code{-S} option (i.e. the part matched by the @samp{*}) will be substituted
3487 for each occurrence of @samp{%*} within @code{X}.
3489 @item %@{@code{S}:@code{X}@}
3490 Substitutes @code{X}, but only if the @samp{-S} switch was given to GCC.
3492 @item %@{!@code{S}:@code{X}@}
3493 Substitutes @code{X}, but only if the @samp{-S} switch was @emph{not} given to GCC.
3495 @item %@{|@code{S}:@code{X}@}
3496 Like %@{@code{S}:@code{X}@}, but if no @code{S} switch, substitute @samp{-}.
3498 @item %@{|!@code{S}:@code{X}@}
3499 Like %@{!@code{S}:@code{X}@}, but if there is an @code{S} switch, substitute @samp{-}.
3501 @item %@{.@code{S}:@code{X}@}
3502 Substitutes @code{X}, but only if processing a file with suffix @code{S}.
3504 @item %@{!.@code{S}:@code{X}@}
3505 Substitutes @code{X}, but only if @emph{not} processing a file with suffix @code{S}.
3507 @item %@{@code{S}|@code{P}:@code{X}@}
3508 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC. This may be
3509 combined with @samp{!} and @samp{.} sequences as well, although they
3510 have a stronger binding than the @samp{|}. For example a spec string
3514 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
3517 will output the following command-line options from the following input
3518 command-line options:
3523 -d fred.c -foo -baz -boggle
3524 -d jim.d -bar -baz -boggle
3529 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or
3530 %@{!@code{S}:@code{X}@} construct may contain other nested @samp{%} constructs
3531 or spaces, or even newlines. They are processed as usual, as described
3534 The @samp{-O, -f, -m, and -W} switches are handled specifically in these
3535 constructs. If another value of @samp{-O} or the negated form of a @samp{-f, -m, or
3536 -W} switch is found later in the command line, the earlier switch
3537 value is ignored, except with @{@code{S}*@} where @code{S} is just one
3538 letter, which passes all matching options.
3540 The character @samp{|} at the beginning of the predicate text is used to indicate
3541 that a command should be piped to the following command, but only if @samp{-pipe}
3544 It is built into GCC which switches take arguments and which do not.
3545 (You might think it would be useful to generalize this to allow each
3546 compiler's spec to say which switches take arguments. But this cannot
3547 be done in a consistent fashion. GCC cannot even decide which input
3548 files have been specified without knowing which switches take arguments,
3549 and it must know which input files to compile in order to tell which
3552 GCC also knows implicitly that arguments starting in @samp{-l} are to be
3553 treated as compiler output files, and passed to the linker in their
3554 proper position among the other output files.
3556 @node Target Options
3557 @section Specifying Target Machine and Compiler Version
3558 @cindex target options
3559 @cindex cross compiling
3560 @cindex specifying machine version
3561 @cindex specifying compiler version and target machine
3562 @cindex compiler version, specifying
3563 @cindex target machine, specifying
3565 By default, GCC compiles code for the same type of machine that you
3566 are using. However, it can also be installed as a cross-compiler, to
3567 compile for some other type of machine. In fact, several different
3568 configurations of GCC, for different target machines, can be
3569 installed side by side. Then you specify which one to use with the
3572 In addition, older and newer versions of GCC can be installed side
3573 by side. One of them (probably the newest) will be the default, but
3574 you may sometimes wish to use another.
3577 @item -b @var{machine}
3578 The argument @var{machine} specifies the target machine for compilation.
3579 This is useful when you have installed GCC as a cross-compiler.
3581 The value to use for @var{machine} is the same as was specified as the
3582 machine type when configuring GCC as a cross-compiler. For
3583 example, if a cross-compiler was configured with @samp{configure
3584 i386v}, meaning to compile for an 80386 running System V, then you
3585 would specify @samp{-b i386v} to run that cross compiler.
3587 When you do not specify @samp{-b}, it normally means to compile for
3588 the same type of machine that you are using.
3590 @item -V @var{version}
3591 The argument @var{version} specifies which version of GCC to run.
3592 This is useful when multiple versions are installed. For example,
3593 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
3595 The default version, when you do not specify @samp{-V}, is the last
3596 version of GCC that you installed.
3599 The @samp{-b} and @samp{-V} options actually work by controlling part of
3600 the file name used for the executable files and libraries used for
3601 compilation. A given version of GCC, for a given target machine, is
3602 normally kept in the directory @file{/usr/local/lib/gcc-lib/@var{machine}/@var{version}}.@refill
3604 Thus, sites can customize the effect of @samp{-b} or @samp{-V} either by
3605 changing the names of these directories or adding alternate names (or
3606 symbolic links). If in directory @file{/usr/local/lib/gcc-lib/} the
3607 file @file{80386} is a link to the file @file{i386v}, then @samp{-b
3608 80386} becomes an alias for @samp{-b i386v}.
3610 In one respect, the @samp{-b} or @samp{-V} do not completely change
3611 to a different compiler: the top-level driver program @code{gcc}
3612 that you originally invoked continues to run and invoke the other
3613 executables (preprocessor, compiler per se, assembler and linker)
3614 that do the real work. However, since no real work is done in the
3615 driver program, it usually does not matter that the driver program
3616 in use is not the one for the specified target and version.
3618 The only way that the driver program depends on the target machine is
3619 in the parsing and handling of special machine-specific options.
3620 However, this is controlled by a file which is found, along with the
3621 other executables, in the directory for the specified version and
3622 target machine. As a result, a single installed driver program adapts
3623 to any specified target machine and compiler version.
3625 The driver program executable does control one significant thing,
3626 however: the default version and target machine. Therefore, you can
3627 install different instances of the driver program, compiled for
3628 different targets or versions, under different names.
3630 For example, if the driver for version 2.0 is installed as @code{ogcc}
3631 and that for version 2.1 is installed as @code{gcc}, then the command
3632 @code{gcc} will use version 2.1 by default, while @code{ogcc} will use
3633 2.0 by default. However, you can choose either version with either
3634 command with the @samp{-V} option.
3636 @node Submodel Options
3637 @section Hardware Models and Configurations
3638 @cindex submodel options
3639 @cindex specifying hardware config
3640 @cindex hardware models and configurations, specifying
3641 @cindex machine dependent options
3643 Earlier we discussed the standard option @samp{-b} which chooses among
3644 different installed compilers for completely different target
3645 machines, such as Vax vs. 68000 vs. 80386.
3647 In addition, each of these target machine types can have its own
3648 special options, starting with @samp{-m}, to choose among various
3649 hardware models or configurations---for example, 68010 vs 68020,
3650 floating coprocessor or none. A single installed version of the
3651 compiler can compile for any model or configuration, according to the
3654 Some configurations of the compiler also support additional special
3655 options, usually for compatibility with other compilers on the same
3659 These options are defined by the macro @code{TARGET_SWITCHES} in the
3660 machine description. The default for the options is also defined by
3661 that macro, which enables you to change the defaults.
3676 * RS/6000 and PowerPC Options::
3681 * Intel 960 Options::
3682 * DEC Alpha Options::
3686 * System V Options::
3687 * TMS320C3x/C4x Options::
3693 @node M680x0 Options
3694 @subsection M680x0 Options
3695 @cindex M680x0 options
3697 These are the @samp{-m} options defined for the 68000 series. The default
3698 values for these options depends on which style of 68000 was selected when
3699 the compiler was configured; the defaults for the most common choices are
3705 Generate output for a 68000. This is the default
3706 when the compiler is configured for 68000-based systems.
3708 Use this option for microcontrollers with a 68000 or EC000 core,
3709 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
3713 Generate output for a 68020. This is the default
3714 when the compiler is configured for 68020-based systems.
3717 Generate output containing 68881 instructions for floating point.
3718 This is the default for most 68020 systems unless @samp{-nfp} was
3719 specified when the compiler was configured.
3722 Generate output for a 68030. This is the default when the compiler is
3723 configured for 68030-based systems.
3726 Generate output for a 68040. This is the default when the compiler is
3727 configured for 68040-based systems.
3729 This option inhibits the use of 68881/68882 instructions that have to be
3730 emulated by software on the 68040. Use this option if your 68040 does not
3731 have code to emulate those instructions.
3734 Generate output for a 68060. This is the default when the compiler is
3735 configured for 68060-based systems.
3737 This option inhibits the use of 68020 and 68881/68882 instructions that
3738 have to be emulated by software on the 68060. Use this option if your 68060
3739 does not have code to emulate those instructions.
3742 Generate output for a CPU32. This is the default
3743 when the compiler is configured for CPU32-based systems.
3745 Use this option for microcontrollers with a
3746 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
3747 68336, 68340, 68341, 68349 and 68360.
3750 Generate output for a 520X "coldfire" family cpu. This is the default
3751 when the compiler is configured for 520X-based systems.
3753 Use this option for microcontroller with a 5200 core, including
3754 the MCF5202, MCF5203, MCF5204 and MCF5202.
3758 Generate output for a 68040, without using any of the new instructions.
3759 This results in code which can run relatively efficiently on either a
3760 68020/68881 or a 68030 or a 68040. The generated code does use the
3761 68881 instructions that are emulated on the 68040.
3764 Generate output for a 68060, without using any of the new instructions.
3765 This results in code which can run relatively efficiently on either a
3766 68020/68881 or a 68030 or a 68040. The generated code does use the
3767 68881 instructions that are emulated on the 68060.
3770 Generate output containing Sun FPA instructions for floating point.
3773 Generate output containing library calls for floating point.
3774 @strong{Warning:} the requisite libraries are not available for all m68k
3775 targets. Normally the facilities of the machine's usual C compiler are
3776 used, but this can't be done directly in cross-compilation. You must
3777 make your own arrangements to provide suitable library functions for
3778 cross-compilation. The embedded targets @samp{m68k-*-aout} and
3779 @samp{m68k-*-coff} do provide software floating point support.
3782 Consider type @code{int} to be 16 bits wide, like @code{short int}.
3785 Do not use the bit-field instructions. The @samp{-m68000}, @samp{-mcpu32}
3786 and @samp{-m5200} options imply @w{@samp{-mnobitfield}}.
3789 Do use the bit-field instructions. The @samp{-m68020} option implies
3790 @samp{-mbitfield}. This is the default if you use a configuration
3791 designed for a 68020.
3794 Use a different function-calling convention, in which functions
3795 that take a fixed number of arguments return with the @code{rtd}
3796 instruction, which pops their arguments while returning. This
3797 saves one instruction in the caller since there is no need to pop
3798 the arguments there.
3800 This calling convention is incompatible with the one normally
3801 used on Unix, so you cannot use it if you need to call libraries
3802 compiled with the Unix compiler.
3804 Also, you must provide function prototypes for all functions that
3805 take variable numbers of arguments (including @code{printf});
3806 otherwise incorrect code will be generated for calls to those
3809 In addition, seriously incorrect code will result if you call a
3810 function with too many arguments. (Normally, extra arguments are
3811 harmlessly ignored.)
3813 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
3814 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
3817 @itemx -mno-align-int
3818 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
3819 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
3820 boundary (@samp{-malign-int}) or a 16-bit boundary (@samp{-mno-align-int}).
3821 Aligning variables on 32-bit boundaries produces code that runs somewhat
3822 faster on processors with 32-bit busses at the expense of more memory.
3824 @strong{Warning:} if you use the @samp{-malign-int} switch, GCC will
3825 align structures containing the above types differently than
3826 most published application binary interface specifications for the m68k.
3829 Use the pc-relative addressing mode of the 68000 directly, instead of
3830 using a global offset table. At present, this option implies -fpic,
3831 allowing at most a 16-bit offset for pc-relative addressing. -fPIC is
3832 not presently supported with -mpcrel, though this could be supported for
3833 68020 and higher processors.
3835 @item -mno-strict-align
3836 @itemx -mstrict-align
3837 @kindex -mstrict-align
3838 Do not (do) assume that unaligned memory references will be handled by
3844 @subsection VAX Options
3847 These @samp{-m} options are defined for the Vax:
3851 Do not output certain jump instructions (@code{aobleq} and so on)
3852 that the Unix assembler for the Vax cannot handle across long
3856 Do output those jump instructions, on the assumption that you
3857 will assemble with the GNU assembler.
3860 Output code for g-format floating point numbers instead of d-format.
3864 @subsection SPARC Options
3865 @cindex SPARC options
3867 These @samp{-m} switches are supported on the SPARC:
3872 Specify @samp{-mapp-regs} to generate output using the global registers
3873 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
3876 To be fully SVR4 ABI compliant at the cost of some performance loss,
3877 specify @samp{-mno-app-regs}. You should compile libraries and system
3878 software with this option.
3882 Generate output containing floating point instructions. This is the
3887 Generate output containing library calls for floating point.
3888 @strong{Warning:} the requisite libraries are not available for all SPARC
3889 targets. Normally the facilities of the machine's usual C compiler are
3890 used, but this cannot be done directly in cross-compilation. You must make
3891 your own arrangements to provide suitable library functions for
3892 cross-compilation. The embedded targets @samp{sparc-*-aout} and
3893 @samp{sparclite-*-*} do provide software floating point support.
3895 @samp{-msoft-float} changes the calling convention in the output file;
3896 therefore, it is only useful if you compile @emph{all} of a program with
3897 this option. In particular, you need to compile @file{libgcc.a}, the
3898 library that comes with GCC, with @samp{-msoft-float} in order for
3901 @item -mhard-quad-float
3902 Generate output containing quad-word (long double) floating point
3905 @item -msoft-quad-float
3906 Generate output containing library calls for quad-word (long double)
3907 floating point instructions. The functions called are those specified
3908 in the SPARC ABI. This is the default.
3910 As of this writing, there are no sparc implementations that have hardware
3911 support for the quad-word floating point instructions. They all invoke
3912 a trap handler for one of these instructions, and then the trap handler
3913 emulates the effect of the instruction. Because of the trap handler overhead,
3914 this is much slower than calling the ABI library routines. Thus the
3915 @samp{-msoft-quad-float} option is the default.
3919 With @samp{-mepilogue} (the default), the compiler always emits code for
3920 function exit at the end of each function. Any function exit in
3921 the middle of the function (such as a return statement in C) will
3922 generate a jump to the exit code at the end of the function.
3924 With @samp{-mno-epilogue}, the compiler tries to emit exit code inline
3925 at every function exit.
3929 With @samp{-mflat}, the compiler does not generate save/restore instructions
3930 and will use a "flat" or single register window calling convention.
3931 This model uses %i7 as the frame pointer and is compatible with the normal
3932 register window model. Code from either may be intermixed.
3933 The local registers and the input registers (0-5) are still treated as
3934 "call saved" registers and will be saved on the stack as necessary.
3936 With @samp{-mno-flat} (the default), the compiler emits save/restore
3937 instructions (except for leaf functions) and is the normal mode of operation.
3939 @item -mno-unaligned-doubles
3940 @itemx -munaligned-doubles
3941 Assume that doubles have 8 byte alignment. This is the default.
3943 With @samp{-munaligned-doubles}, GCC assumes that doubles have 8 byte
3944 alignment only if they are contained in another type, or if they have an
3945 absolute address. Otherwise, it assumes they have 4 byte alignment.
3946 Specifying this option avoids some rare compatibility problems with code
3947 generated by other compilers. It is not the default because it results
3948 in a performance loss, especially for floating point code.
3950 @item -mno-faster-structs
3951 @itemx -mfaster-structs
3952 With @samp{-mfaster-structs}, the compiler assumes that structures
3953 should have 8 byte alignment. This enables the use of pairs of
3954 @code{ldd} and @code{std} instructions for copies in structure
3955 assignment, in place of twice as many @code{ld} and @code{st} pairs.
3956 However, the use of this changed alignment directly violates the Sparc
3957 ABI. Thus, it's intended only for use on targets where the developer
3958 acknowledges that their resulting code will not be directly in line with
3959 the rules of the ABI.
3963 These two options select variations on the SPARC architecture.
3965 By default (unless specifically configured for the Fujitsu SPARClite),
3966 GCC generates code for the v7 variant of the SPARC architecture.
3968 @samp{-mv8} will give you SPARC v8 code. The only difference from v7
3969 code is that the compiler emits the integer multiply and integer
3970 divide instructions which exist in SPARC v8 but not in SPARC v7.
3972 @samp{-msparclite} will give you SPARClite code. This adds the integer
3973 multiply, integer divide step and scan (@code{ffs}) instructions which
3974 exist in SPARClite but not in SPARC v7.
3976 These options are deprecated and will be deleted in a future GCC release.
3977 They have been replaced with @samp{-mcpu=xxx}.
3981 These two options select the processor for which the code is optimised.
3983 With @samp{-mcypress} (the default), the compiler optimizes code for the
3984 Cypress CY7C602 chip, as used in the SparcStation/SparcServer 3xx series.
3985 This is also appropriate for the older SparcStation 1, 2, IPX etc.
3987 With @samp{-msupersparc} the compiler optimizes code for the SuperSparc cpu, as
3988 used in the SparcStation 10, 1000 and 2000 series. This flag also enables use
3989 of the full SPARC v8 instruction set.
3991 These options are deprecated and will be deleted in a future GCC release.
3992 They have been replaced with @samp{-mcpu=xxx}.
3994 @item -mcpu=@var{cpu_type}
3995 Set the instruction set, register set, and instruction scheduling parameters
3996 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
3997 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
3998 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
3999 @samp{sparclet}, @samp{tsc701}, @samp{v9}, and @samp{ultrasparc}.
4001 Default instruction scheduling parameters are used for values that select
4002 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
4003 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
4005 Here is a list of each supported architecture and their supported
4010 v8: supersparc, hypersparc
4011 sparclite: f930, f934, sparclite86x
4016 @item -mtune=@var{cpu_type}
4017 Set the instruction scheduling parameters for machine type
4018 @var{cpu_type}, but do not set the instruction set or register set that the
4019 option @samp{-mcpu=}@var{cpu_type} would.
4021 The same values for @samp{-mcpu=}@var{cpu_type} are used for
4022 @samp{-mtune=}@*@var{cpu_type}, though the only useful values are those that
4023 select a particular cpu implementation: @samp{cypress}, @samp{supersparc},
4024 @samp{hypersparc}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
4025 @samp{tsc701}, @samp{ultrasparc}.
4029 These @samp{-m} switches are supported in addition to the above
4030 on the SPARCLET processor.
4033 @item -mlittle-endian
4034 Generate code for a processor running in little-endian mode.
4037 Treat register @code{%g0} as a normal register.
4038 GCC will continue to clobber it as necessary but will not assume
4039 it always reads as 0.
4041 @item -mbroken-saverestore
4042 Generate code that does not use non-trivial forms of the @code{save} and
4043 @code{restore} instructions. Early versions of the SPARCLET processor do
4044 not correctly handle @code{save} and @code{restore} instructions used with
4045 arguments. They correctly handle them used without arguments. A @code{save}
4046 instruction used without arguments increments the current window pointer
4047 but does not allocate a new stack frame. It is assumed that the window
4048 overflow trap handler will properly handle this case as will interrupt
4052 These @samp{-m} switches are supported in addition to the above
4053 on SPARC V9 processors in 64 bit environments.
4056 @item -mlittle-endian
4057 Generate code for a processor running in little-endian mode.
4061 Generate code for a 32 bit or 64 bit environment.
4062 The 32 bit environment sets int, long and pointer to 32 bits.
4063 The 64 bit environment sets int to 32 bits and long and pointer
4066 @item -mcmodel=medlow
4067 Generate code for the Medium/Low code model: the program must be linked
4068 in the low 32 bits of the address space. Pointers are 64 bits.
4069 Programs can be statically or dynamically linked.
4071 @item -mcmodel=medmid
4072 Generate code for the Medium/Middle code model: the program must be linked
4073 in the low 44 bits of the address space, the text segment must be less than
4074 2G bytes, and data segment must be within 2G of the text segment.
4075 Pointers are 64 bits.
4077 @item -mcmodel=medany
4078 Generate code for the Medium/Anywhere code model: the program may be linked
4079 anywhere in the address space, the text segment must be less than
4080 2G bytes, and data segment must be within 2G of the text segment.
4081 Pointers are 64 bits.
4083 @item -mcmodel=embmedany
4084 Generate code for the Medium/Anywhere code model for embedded systems:
4085 assume a 32 bit text and a 32 bit data segment, both starting anywhere
4086 (determined at link time). Register %g4 points to the base of the
4087 data segment. Pointers still 64 bits.
4088 Programs are statically linked, PIC is not supported.
4091 @itemx -mno-stack-bias
4092 With @samp{-mstack-bias}, GCC assumes that the stack pointer, and
4093 frame pointer if present, are offset by -2047 which must be added back
4094 when making stack frame references.
4095 Otherwise, assume no such offset is present.
4098 @node Convex Options
4099 @subsection Convex Options
4100 @cindex Convex options
4102 These @samp{-m} options are defined for Convex:
4106 Generate output for C1. The code will run on any Convex machine.
4107 The preprocessor symbol @code{__convex__c1__} is defined.
4110 Generate output for C2. Uses instructions not available on C1.
4111 Scheduling and other optimizations are chosen for max performance on C2.
4112 The preprocessor symbol @code{__convex_c2__} is defined.
4115 Generate output for C32xx. Uses instructions not available on C1.
4116 Scheduling and other optimizations are chosen for max performance on C32.
4117 The preprocessor symbol @code{__convex_c32__} is defined.
4120 Generate output for C34xx. Uses instructions not available on C1.
4121 Scheduling and other optimizations are chosen for max performance on C34.
4122 The preprocessor symbol @code{__convex_c34__} is defined.
4125 Generate output for C38xx. Uses instructions not available on C1.
4126 Scheduling and other optimizations are chosen for max performance on C38.
4127 The preprocessor symbol @code{__convex_c38__} is defined.
4130 Generate code which puts an argument count in the word preceding each
4131 argument list. This is compatible with regular CC, and a few programs
4132 may need the argument count word. GDB and other source-level debuggers
4133 do not need it; this info is in the symbol table.
4136 Omit the argument count word. This is the default.
4138 @item -mvolatile-cache
4139 Allow volatile references to be cached. This is the default.
4141 @item -mvolatile-nocache
4142 Volatile references bypass the data cache, going all the way to memory.
4143 This is only needed for multi-processor code that does not use standard
4144 synchronization instructions. Making non-volatile references to volatile
4145 locations will not necessarily work.
4148 Type long is 32 bits, the same as type int. This is the default.
4151 Type long is 64 bits, the same as type long long. This option is useless,
4152 because no library support exists for it.
4155 @node AMD29K Options
4156 @subsection AMD29K Options
4157 @cindex AMD29K options
4159 These @samp{-m} options are defined for the AMD Am29000:
4164 @cindex DW bit (29k)
4165 Generate code that assumes the @code{DW} bit is set, i.e., that byte and
4166 halfword operations are directly supported by the hardware. This is the
4171 Generate code that assumes the @code{DW} bit is not set.
4175 @cindex byte writes (29k)
4176 Generate code that assumes the system supports byte and halfword write
4177 operations. This is the default.
4181 Generate code that assumes the systems does not support byte and
4182 halfword write operations. @samp{-mnbw} implies @samp{-mndw}.
4186 @cindex memory model (29k)
4187 Use a small memory model that assumes that all function addresses are
4188 either within a single 256 KB segment or at an absolute address of less
4189 than 256k. This allows the @code{call} instruction to be used instead
4190 of a @code{const}, @code{consth}, @code{calli} sequence.
4194 Use the normal memory model: Generate @code{call} instructions only when
4195 calling functions in the same file and @code{calli} instructions
4196 otherwise. This works if each file occupies less than 256 KB but allows
4197 the entire executable to be larger than 256 KB. This is the default.
4200 Always use @code{calli} instructions. Specify this option if you expect
4201 a single file to compile into more than 256 KB of code.
4205 @cindex processor selection (29k)
4206 Generate code for the Am29050.
4210 Generate code for the Am29000. This is the default.
4212 @item -mkernel-registers
4213 @kindex -mkernel-registers
4214 @cindex kernel and user registers (29k)
4215 Generate references to registers @code{gr64-gr95} instead of to
4216 registers @code{gr96-gr127}. This option can be used when compiling
4217 kernel code that wants a set of global registers disjoint from that used
4220 Note that when this option is used, register names in @samp{-f} flags
4221 must use the normal, user-mode, names.
4223 @item -muser-registers
4224 @kindex -muser-registers
4225 Use the normal set of global registers, @code{gr96-gr127}. This is the
4229 @itemx -mno-stack-check
4230 @kindex -mstack-check
4231 @cindex stack checks (29k)
4232 Insert (or do not insert) a call to @code{__msp_check} after each stack
4233 adjustment. This is often used for kernel code.
4236 @itemx -mno-storem-bug
4237 @kindex -mstorem-bug
4238 @cindex storem bug (29k)
4239 @samp{-mstorem-bug} handles 29k processors which cannot handle the
4240 separation of a mtsrim insn and a storem instruction (most 29000 chips
4241 to date, but not the 29050).
4243 @item -mno-reuse-arg-regs
4244 @itemx -mreuse-arg-regs
4245 @kindex -mreuse-arg-regs
4246 @samp{-mno-reuse-arg-regs} tells the compiler to only use incoming argument
4247 registers for copying out arguments. This helps detect calling a function
4248 with fewer arguments than it was declared with.
4250 @item -mno-impure-text
4251 @itemx -mimpure-text
4252 @kindex -mimpure-text
4253 @samp{-mimpure-text}, used in addition to @samp{-shared}, tells the compiler to
4254 not pass @samp{-assert pure-text} to the linker when linking a shared object.
4257 @kindex -msoft-float
4258 Generate output containing library calls for floating point.
4259 @strong{Warning:} the requisite libraries are not part of GCC.
4260 Normally the facilities of the machine's usual C compiler are used, but
4261 this can't be done directly in cross-compilation. You must make your
4262 own arrangements to provide suitable library functions for
4267 Do not generate multm or multmu instructions. This is useful for some embedded
4268 systems which do not have trap handlers for these instructions.
4272 @subsection ARM Options
4275 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
4280 @kindex -mapcs-frame
4281 Generate a stack frame that is compliant with the ARM Procedure Call
4282 Standard for all functions, even if this is not strictly necessary for
4283 correct execution of the code. Specifying @samp{-fomit-frame-pointer}
4284 with this option will cause the stack frames not to be generated for
4285 leaf functions. The default is @samp{-mno-apcs-frame}.
4289 This is a synonym for @samp{-mapcs-frame}.
4293 Generate code for a processor running with a 26-bit program counter,
4294 and conforming to the function calling standards for the APCS 26-bit
4295 option. This option replaces the @samp{-m2} and @samp{-m3} options
4296 of previous releases of the compiler.
4300 Generate code for a processor running with a 32-bit program counter,
4301 and conforming to the function calling standards for the APCS 32-bit
4302 option. This option replaces the @samp{-m6} option of previous releases
4305 @item -mapcs-stack-check
4306 @kindex -mapcs-stack-check
4307 @kindex -mno-apcs-stack-check
4308 Generate code to check the amount of stack space available upon entry to
4309 every function (that actually uses some stack space). If there is
4310 insufficient space available then either the function
4311 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
4312 called, depending upon the amount of stack space required. The run time
4313 system is required to provide these functions. The default is
4314 @samp{-mno-apcs-stack-check}, since this produces smaller code.
4317 @kindex -mapcs-float
4318 @kindex -mno-apcs-float
4319 Pass floating point arguments using the float point registers. This is
4320 one of the variants of the APCS. This option is recommended if the
4321 target hardware has a floating point unit or if a lot of floating point
4322 arithmetic is going to be performed by the code. The default is
4323 @samp{-mno-apcs-float}, since integer only code is slightly increased in
4324 size if @samp{-mapcs-float} is used.
4326 @item -mapcs-reentrant
4327 @kindex -mapcs-reentrant
4328 @kindex -mno-apcs-reentrant
4329 Generate reentrant, position independent code. This is the equivalent
4330 to specifying the @samp{-fpic} option. The default is
4331 @samp{-mno-apcs-reentrant}.
4333 @item -mthumb-interwork
4334 @kindex -mthumb-interwork
4335 @kindex -mno-thumb-interwork
4336 Generate code which supports calling between the ARM and THUMB
4337 instruction sets. Without this option the two instruction sets cannot
4338 be reliably used inside one program. The default is
4339 @samp{-mno-thumb-interwork}, since slightly larger code is generated
4340 when @samp{-mthumb-interwork} is specified.
4342 @item -mno-sched-prolog
4343 @kindex -mno-sched-prolog
4344 @kindex -msched-prolog
4345 Prevent the reordering of instructions in the function prolog, or the
4346 merging of those instruction with the instructions in the function's
4347 body. This means that all functions will start with a recognizable set
4348 of instructions (or in fact one of a choice from a small set of
4349 different function prologues), and this information can be used to
4350 locate the start if functions inside an executable piece of code. The
4351 default is @samp{-msched-prolog}.
4354 Generate output containing floating point instructions. This is the
4358 Generate output containing library calls for floating point.
4359 @strong{Warning:} the requisite libraries are not available for all ARM
4360 targets. Normally the facilities of the machine's usual C compiler are
4361 used, but this cannot be done directly in cross-compilation. You must make
4362 your own arrangements to provide suitable library functions for
4365 @samp{-msoft-float} changes the calling convention in the output file;
4366 therefore, it is only useful if you compile @emph{all} of a program with
4367 this option. In particular, you need to compile @file{libgcc.a}, the
4368 library that comes with GCC, with @samp{-msoft-float} in order for
4371 @item -mlittle-endian
4372 Generate code for a processor running in little-endian mode. This is
4373 the default for all standard configurations.
4376 Generate code for a processor running in big-endian mode; the default is
4377 to compile code for a little-endian processor.
4379 @item -mwords-little-endian
4380 This option only applies when generating code for big-endian processors.
4381 Generate code for a little-endian word order but a big-endian byte
4382 order. That is, a byte order of the form @samp{32107654}. Note: this
4383 option should only be used if you require compatibility with code for
4384 big-endian ARM processors generated by versions of the compiler prior to
4387 @item -malignment-traps
4388 @kindex -malignment-traps
4389 Generate code that will not trap if the MMU has alignment traps enabled.
4390 On ARM architectures prior to ARMv4, there were no instructions to
4391 access half-word objects stored in memory. However, when reading from
4392 memory a feature of the ARM architecture allows a word load to be used,
4393 even if the address is unaligned, and the processor core will rotate the
4394 data as it is being loaded. This option tells the compiler that such
4395 misaligned accesses will cause a MMU trap and that it should instead
4396 synthesise the access as a series of byte accesses. The compiler can
4397 still use word accesses to load half-word data if it knows that the
4398 address is aligned to a word boundary.
4400 This option is ignored when compiling for ARM architecture 4 or later,
4401 since these processors have instructions to directly access half-word
4404 @item -mno-alignment-traps
4405 @kindex -mno-alignment-traps
4406 Generate code that assumes that the MMU will not trap unaligned
4407 accesses. This produces better code when the target instruction set
4408 does not have half-word memory operations (implementations prior to
4411 Note that you cannot use this option to access unaligned word objects,
4412 since the processor will only fetch one 32-bit aligned object from
4415 The default setting for most targets is -mno-alignment-traps, since
4416 this produces better code when there are no half-word memory
4417 instructions available.
4419 @item -mshort-load-bytes
4420 @kindex -mshort-load-bytes
4421 This is a depreciated alias for @samp{-malignment-traps}.
4423 @item -mno-short-load-bytes
4424 @kindex -mno-short-load-bytes
4425 This is a depreciated alias for @samp{-mno-alignment-traps}.
4427 @item -mshort-load-words
4428 @kindex -mshort-load-words
4429 This is a depreciated alias for @samp{-mno-alignment-traps}.
4431 @item -mno-short-load-words
4432 @kindex -mno-short-load-words
4433 This is a depreciated alias for @samp{-malignment-traps}.
4437 This option only applies to RISC iX. Emulate the native BSD-mode
4438 compiler. This is the default if @samp{-ansi} is not specified.
4442 This option only applies to RISC iX. Emulate the native X/Open-mode
4445 @item -mno-symrename
4446 @kindex -mno-symrename
4447 This option only applies to RISC iX. Do not run the assembler
4448 post-processor, @samp{symrename}, after code has been assembled.
4449 Normally it is necessary to modify some of the standard symbols in
4450 preparation for linking with the RISC iX C library; this option
4451 suppresses this pass. The post-processor is never run when the
4452 compiler is built for cross-compilation.
4456 This specifies the name of the target ARM processor. GCC uses this name
4457 to determine what kind of instructions it can use when generating
4458 assembly code. Permissible names are: arm2, arm250, arm3, arm6, arm60,
4459 arm600, arm610, arm620, arm7, arm7m, arm7d, arm7dm, arm7di, arm7dmi,
4460 arm70, arm700, arm700i, arm710, arm710c, arm7100, arm7500, arm7500fe,
4461 arm7tdmi, arm8, strongarm, strongarm110, strongarm1100, arm8, arm810,
4462 arm9, arm920, arm920t, arm9tdmi.
4464 @itemx -mtune=<name>
4466 This option is very similar to the @samp{-mcpu=} option, except that
4467 instead of specifying the actual target processor type, and hence
4468 restricting which instructions can be used, it specifies that GCC should
4469 tune the performance of the code as if the target were of the type
4470 specified in this option, but still choosing the instructions that it
4471 will generate based on the cpu specified by a @samp{-mcpu=} option.
4472 For some arm implementations better performance can be obtained by using
4477 This specifies the name of the target ARM architecture. GCC uses this
4478 name to determine what kind of instructions it can use when generating
4479 assembly code. This option can be used in conjunction with or instead
4480 of the @samp{-mcpu=} option. Permissible names are: armv2, armv2a,
4481 armv3, armv3m, armv4, armv4t, armv5.
4483 @item -mfpe=<number>
4484 @itemx -mfp=<number>
4487 This specifes the version of the floating point emulation available on
4488 the target. Permissible values are 2 and 3. @samp{-mfp=} is a synonym
4489 for @samp{-mfpe=} to support older versions of GCC.
4491 @item -mstructure-size-boundary=<n>
4492 @kindex -mstructure-size-boundary
4493 The size of all structures and unions will be rounded up to a multiple
4494 of the number of bits set by this option. Permissible values are 8 and
4495 32. The default value varies for different toolchains. For the COFF
4496 targeted toolchain the default value is 8. Specifying the larger number
4497 can produce faster, more efficient code, but can also increase the size
4498 of the program. The two values are potentially incompatible. Code
4499 compiled with one value cannot necessarily expect to work with code or
4500 libraries compiled with the other value, if they exchange information
4501 using structures or unions. Programmers are encouraged to use the 32
4502 value as future versions of the toolchain may default to this value.
4504 @item -mabort-on-noreturn
4505 @kindex -mabort-on-noreturn
4506 @kindex -mnoabort-on-noreturn
4507 Generate a call to the function abort at the end of a noreturn function.
4508 It will be executed if the function tries to return.
4510 @item -mnop-fun-dllimport
4511 @kindex -mnop-fun-dllimport
4512 Disable the support for the @emph{dllimport} attribute.
4514 @item -msingle-pic-base
4515 @kindex -msingle-pic-base
4516 Treat the register used for PIC addressing as read-only, rather than
4517 loading it in the prologue for each function. The run-time system is
4518 responsible for initialising this register with an appropriate value
4519 before execution begins.
4521 @item -mpic-register=<reg>
4522 @kindex -mpic-register=
4523 Specify the register to be used for PIC addressing. The default is R10
4524 unless stack-checking is enabled, when R9 is used.
4529 @subsection Thumb Options
4530 @cindex Thumb Options
4534 @item -mthumb-interwork
4535 @kindex -mthumb-interwork
4536 @kindex -mno-thumb-interwork
4537 Generate code which supports calling between the THUMB and ARM
4538 instruction sets. Without this option the two instruction sets cannot
4539 be reliably used inside one program. The default is
4540 @samp{-mno-thumb-interwork}, since slightly smaller code is generated
4544 @kindex -mtpcs-frame
4545 @kindex -mno-tpcs-frame
4546 Generate a stack frame that is compliant with the Thumb Procedure Call
4547 Standard for all non-leaf functions. (A leaf function is one that does
4548 not call any other functions). The default is @samp{-mno-apcs-frame}.
4550 @item -mtpcs-leaf-frame
4551 @kindex -mtpcs-leaf-frame
4552 @kindex -mno-tpcs-leaf-frame
4553 Generate a stack frame that is compliant with the Thumb Procedure Call
4554 Standard for all leaf functions. (A leaf function is one that does
4555 not call any other functions). The default is @samp{-mno-apcs-leaf-frame}.
4557 @item -mlittle-endian
4558 @kindex -mlittle-endian
4559 Generate code for a processor running in little-endian mode. This is
4560 the default for all standard configurations.
4563 @kindex -mbig-endian
4564 Generate code for a processor running in big-endian mode.
4566 @item -mstructure-size-boundary=<n>
4567 @kindex -mstructure-size-boundary
4568 The size of all structures and unions will be rounded up to a multiple
4569 of the number of bits set by this option. Permissible values are 8 and
4570 32. The default value varies for different toolchains. For the COFF
4571 targeted toolchain the default value is 8. Specifying the larger number
4572 can produced faster, more efficient code, but can also increase the size
4573 of the program. The two values are potentially incompatible. Code
4574 compiled with one value cannot necessarily expect to work with code or
4575 libraries compiled with the other value, if they exchange information
4576 using structures or unions. Programmers are encouraged to use the 32
4577 value as future versions of the toolchain may default to this value.
4579 @item -mnop-fun-dllimport
4580 @kindex -mnop-fun-dllimport
4581 Disable the support for the @emph{dllimport} attribute.
4583 @item -mcallee-super-interworking
4584 @kindex -mcallee-super-interworking
4585 Gives all externally visible functions in the file being compiled an ARM
4586 instruction set header which switches to Thumb mode before executing the
4587 rest of the function. This allows these functions to be called from
4588 non-interworking code.
4590 @item -mcaller-super-interworking
4591 @kindex -mcaller-super-interworking
4592 Allows calls via function pointers (including virtual functions) to
4593 execute correctly regardless of whether the target code has been
4594 compiled for interworking or not. There is a small overhead in the cost
4595 of executing a function pointer if this option is enabled.
4597 @item -msingle-pic-base
4598 @kindex -msingle-pic-base
4599 Treat the register used for PIC addressing as read-only, rather than
4600 loading it in the prologue for each function. The run-time system is
4601 responsible for initialising this register with an appropriate value
4602 before execution begins.
4604 @item -mpic-register=<reg>
4605 @kindex -mpic-register=
4606 Specify the register to be used for PIC addressing. The default is R10.
4610 @node MN10200 Options
4611 @subsection MN10200 Options
4612 @cindex MN10200 options
4613 These @samp{-m} options are defined for Matsushita MN10200 architectures:
4617 Indicate to the linker that it should perform a relaxation optimization pass
4618 to shorten branches, calls and absolute memory addresses. This option only
4619 has an effect when used on the command line for the final link step.
4621 This option makes symbolic debugging impossible.
4624 @node MN10300 Options
4625 @subsection MN10300 Options
4626 @cindex MN10300 options
4627 These @samp{-m} options are defined for Matsushita MN10300 architectures:
4631 Generate code to avoid bugs in the multiply instructions for the MN10300
4632 processors. This is the default.
4635 Do not generate code to avoid bugs in the multiply instructions for the
4639 Indicate to the linker that it should perform a relaxation optimization pass
4640 to shorten branches, calls and absolute memory addresses. This option only
4641 has an effect when used on the command line for the final link step.
4643 This option makes symbolic debugging impossible.
4647 @node M32R/D Options
4648 @subsection M32R/D Options
4649 @cindex M32R/D options
4651 These @samp{-m} options are defined for Mitsubishi M32R/D architectures:
4654 @item -mcode-model=small
4655 Assume all objects live in the lower 16MB of memory (so that their addresses
4656 can be loaded with the @code{ld24} instruction), and assume all subroutines
4657 are reachable with the @code{bl} instruction.
4658 This is the default.
4660 The addressability of a particular object can be set with the
4661 @code{model} attribute.
4663 @item -mcode-model=medium
4664 Assume objects may be anywhere in the 32 bit address space (the compiler
4665 will generate @code{seth/add3} instructions to load their addresses), and
4666 assume all subroutines are reachable with the @code{bl} instruction.
4668 @item -mcode-model=large
4669 Assume objects may be anywhere in the 32 bit address space (the compiler
4670 will generate @code{seth/add3} instructions to load their addresses), and
4671 assume subroutines may not be reachable with the @code{bl} instruction
4672 (the compiler will generate the much slower @code{seth/add3/jl}
4673 instruction sequence).
4676 Disable use of the small data area. Variables will be put into
4677 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
4678 @code{section} attribute has been specified).
4679 This is the default.
4681 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
4682 Objects may be explicitly put in the small data area with the
4683 @code{section} attribute using one of these sections.
4686 Put small global and static data in the small data area, but do not
4687 generate special code to reference them.
4690 Put small global and static data in the small data area, and generate
4691 special instructions to reference them.
4694 @cindex smaller data references
4695 Put global and static objects less than or equal to @var{num} bytes
4696 into the small data or bss sections instead of the normal data or bss
4697 sections. The default value of @var{num} is 8.
4698 The @samp{-msdata} option must be set to one of @samp{sdata} or @samp{use}
4699 for this option to have any effect.
4701 All modules should be compiled with the same @samp{-G @var{num}} value.
4702 Compiling with different values of @var{num} may or may not work; if it
4703 doesn't the linker will give an error message - incorrect code will not be
4709 @subsection M88K Options
4710 @cindex M88k options
4712 These @samp{-m} options are defined for Motorola 88k architectures:
4717 Generate code that works well on both the m88100 and the
4722 Generate code that works best for the m88100, but that also
4727 Generate code that works best for the m88110, and may not run
4732 Obsolete option to be removed from the next revision.
4735 @item -midentify-revision
4736 @kindex -midentify-revision
4738 @cindex identifying source, compiler (88k)
4739 Include an @code{ident} directive in the assembler output recording the
4740 source file name, compiler name and version, timestamp, and compilation
4743 @item -mno-underscores
4744 @kindex -mno-underscores
4745 @cindex underscores, avoiding (88k)
4746 In assembler output, emit symbol names without adding an underscore
4747 character at the beginning of each name. The default is to use an
4748 underscore as prefix on each name.
4750 @item -mocs-debug-info
4751 @itemx -mno-ocs-debug-info
4752 @kindex -mocs-debug-info
4753 @kindex -mno-ocs-debug-info
4755 @cindex debugging, 88k OCS
4756 Include (or omit) additional debugging information (about registers used
4757 in each stack frame) as specified in the 88open Object Compatibility
4758 Standard, ``OCS''. This extra information allows debugging of code that
4759 has had the frame pointer eliminated. The default for DG/UX, SVr4, and
4760 Delta 88 SVr3.2 is to include this information; other 88k configurations
4761 omit this information by default.
4763 @item -mocs-frame-position
4764 @kindex -mocs-frame-position
4765 @cindex register positions in frame (88k)
4766 When emitting COFF debugging information for automatic variables and
4767 parameters stored on the stack, use the offset from the canonical frame
4768 address, which is the stack pointer (register 31) on entry to the
4769 function. The DG/UX, SVr4, Delta88 SVr3.2, and BCS configurations use
4770 @samp{-mocs-frame-position}; other 88k configurations have the default
4771 @samp{-mno-ocs-frame-position}.
4773 @item -mno-ocs-frame-position
4774 @kindex -mno-ocs-frame-position
4775 @cindex register positions in frame (88k)
4776 When emitting COFF debugging information for automatic variables and
4777 parameters stored on the stack, use the offset from the frame pointer
4778 register (register 30). When this option is in effect, the frame
4779 pointer is not eliminated when debugging information is selected by the
4782 @item -moptimize-arg-area
4783 @itemx -mno-optimize-arg-area
4784 @kindex -moptimize-arg-area
4785 @kindex -mno-optimize-arg-area
4786 @cindex arguments in frame (88k)
4787 Control how function arguments are stored in stack frames.
4788 @samp{-moptimize-arg-area} saves space by optimizing them, but this
4789 conflicts with the 88open specifications. The opposite alternative,
4790 @samp{-mno-optimize-arg-area}, agrees with 88open standards. By default
4791 GCC does not optimize the argument area.
4793 @item -mshort-data-@var{num}
4794 @kindex -mshort-data-@var{num}
4795 @cindex smaller data references (88k)
4796 @cindex r0-relative references (88k)
4797 Generate smaller data references by making them relative to @code{r0},
4798 which allows loading a value using a single instruction (rather than the
4799 usual two). You control which data references are affected by
4800 specifying @var{num} with this option. For example, if you specify
4801 @samp{-mshort-data-512}, then the data references affected are those
4802 involving displacements of less than 512 bytes.
4803 @samp{-mshort-data-@var{num}} is not effective for @var{num} greater
4806 @item -mserialize-volatile
4807 @kindex -mserialize-volatile
4808 @itemx -mno-serialize-volatile
4809 @kindex -mno-serialize-volatile
4810 @cindex sequential consistency on 88k
4811 Do, or don't, generate code to guarantee sequential consistency
4812 of volatile memory references. By default, consistency is
4815 The order of memory references made by the MC88110 processor does
4816 not always match the order of the instructions requesting those
4817 references. In particular, a load instruction may execute before
4818 a preceding store instruction. Such reordering violates
4819 sequential consistency of volatile memory references, when there
4820 are multiple processors. When consistency must be guaranteed,
4821 GNU C generates special instructions, as needed, to force
4822 execution in the proper order.
4824 The MC88100 processor does not reorder memory references and so
4825 always provides sequential consistency. However, by default, GNU
4826 C generates the special instructions to guarantee consistency
4827 even when you use @samp{-m88100}, so that the code may be run on an
4828 MC88110 processor. If you intend to run your code only on the
4829 MC88100 processor, you may use @samp{-mno-serialize-volatile}.
4831 The extra code generated to guarantee consistency may affect the
4832 performance of your application. If you know that you can safely
4833 forgo this guarantee, you may use @samp{-mno-serialize-volatile}.
4839 @cindex assembler syntax, 88k
4841 Turn on (@samp{-msvr4}) or off (@samp{-msvr3}) compiler extensions
4842 related to System V release 4 (SVr4). This controls the following:
4846 Which variant of the assembler syntax to emit.
4848 @samp{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
4849 that is used on System V release 4.
4851 @samp{-msvr4} makes GCC issue additional declaration directives used in
4855 @samp{-msvr4} is the default for the m88k-motorola-sysv4 and
4856 m88k-dg-dgux m88k configurations. @samp{-msvr3} is the default for all
4857 other m88k configurations.
4859 @item -mversion-03.00
4860 @kindex -mversion-03.00
4861 This option is obsolete, and is ignored.
4862 @c ??? which asm syntax better for GAS? option there too?
4864 @item -mno-check-zero-division
4865 @itemx -mcheck-zero-division
4866 @kindex -mno-check-zero-division
4867 @kindex -mcheck-zero-division
4868 @cindex zero division on 88k
4869 Do, or don't, generate code to guarantee that integer division by
4870 zero will be detected. By default, detection is guaranteed.
4872 Some models of the MC88100 processor fail to trap upon integer
4873 division by zero under certain conditions. By default, when
4874 compiling code that might be run on such a processor, GNU C
4875 generates code that explicitly checks for zero-valued divisors
4876 and traps with exception number 503 when one is detected. Use of
4877 mno-check-zero-division suppresses such checking for code
4878 generated to run on an MC88100 processor.
4880 GNU C assumes that the MC88110 processor correctly detects all
4881 instances of integer division by zero. When @samp{-m88110} is
4882 specified, both @samp{-mcheck-zero-division} and
4883 @samp{-mno-check-zero-division} are ignored, and no explicit checks for
4884 zero-valued divisors are generated.
4886 @item -muse-div-instruction
4887 @kindex -muse-div-instruction
4888 @cindex divide instruction, 88k
4889 Use the div instruction for signed integer division on the
4890 MC88100 processor. By default, the div instruction is not used.
4892 On the MC88100 processor the signed integer division instruction
4893 div) traps to the operating system on a negative operand. The
4894 operating system transparently completes the operation, but at a
4895 large cost in execution time. By default, when compiling code
4896 that might be run on an MC88100 processor, GNU C emulates signed
4897 integer division using the unsigned integer division instruction
4898 divu), thereby avoiding the large penalty of a trap to the
4899 operating system. Such emulation has its own, smaller, execution
4900 cost in both time and space. To the extent that your code's
4901 important signed integer division operations are performed on two
4902 nonnegative operands, it may be desirable to use the div
4903 instruction directly.
4905 On the MC88110 processor the div instruction (also known as the
4906 divs instruction) processes negative operands without trapping to
4907 the operating system. When @samp{-m88110} is specified,
4908 @samp{-muse-div-instruction} is ignored, and the div instruction is used
4909 for signed integer division.
4911 Note that the result of dividing INT_MIN by -1 is undefined. In
4912 particular, the behavior of such a division with and without
4913 @samp{-muse-div-instruction} may differ.
4915 @item -mtrap-large-shift
4916 @itemx -mhandle-large-shift
4917 @kindex -mtrap-large-shift
4918 @kindex -mhandle-large-shift
4919 @cindex bit shift overflow (88k)
4920 @cindex large bit shifts (88k)
4921 Include code to detect bit-shifts of more than 31 bits; respectively,
4922 trap such shifts or emit code to handle them properly. By default GCC
4923 makes no special provision for large bit shifts.
4925 @item -mwarn-passed-structs
4926 @kindex -mwarn-passed-structs
4927 @cindex structure passing (88k)
4928 Warn when a function passes a struct as an argument or result.
4929 Structure-passing conventions have changed during the evolution of the C
4930 language, and are often the source of portability problems. By default,
4931 GCC issues no such warning.
4934 @node RS/6000 and PowerPC Options
4935 @subsection IBM RS/6000 and PowerPC Options
4936 @cindex RS/6000 and PowerPC Options
4937 @cindex IBM RS/6000 and PowerPC Options
4939 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
4947 @itemx -mpowerpc-gpopt
4948 @itemx -mno-powerpc-gpopt
4949 @itemx -mpowerpc-gfxopt
4950 @itemx -mno-powerpc-gfxopt
4952 @itemx -mno-powerpc64
4956 @kindex -mpowerpc-gpopt
4957 @kindex -mpowerpc-gfxopt
4959 GCC supports two related instruction set architectures for the
4960 RS/6000 and PowerPC. The @dfn{POWER} instruction set are those
4961 instructions supported by the @samp{rios} chip set used in the original
4962 RS/6000 systems and the @dfn{PowerPC} instruction set is the
4963 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
4964 the IBM 4xx microprocessors.
4966 Neither architecture is a subset of the other. However there is a
4967 large common subset of instructions supported by both. An MQ
4968 register is included in processors supporting the POWER architecture.
4970 You use these options to specify which instructions are available on the
4971 processor you are using. The default value of these options is
4972 determined when configuring GCC. Specifying the
4973 @samp{-mcpu=@var{cpu_type}} overrides the specification of these
4974 options. We recommend you use the @samp{-mcpu=@var{cpu_type}} option
4975 rather than the options listed above.
4977 The @samp{-mpower} option allows GCC to generate instructions that
4978 are found only in the POWER architecture and to use the MQ register.
4979 Specifying @samp{-mpower2} implies @samp{-power} and also allows GCC
4980 to generate instructions that are present in the POWER2 architecture but
4981 not the original POWER architecture.
4983 The @samp{-mpowerpc} option allows GCC to generate instructions that
4984 are found only in the 32-bit subset of the PowerPC architecture.
4985 Specifying @samp{-mpowerpc-gpopt} implies @samp{-mpowerpc} and also allows
4986 GCC to use the optional PowerPC architecture instructions in the
4987 General Purpose group, including floating-point square root. Specifying
4988 @samp{-mpowerpc-gfxopt} implies @samp{-mpowerpc} and also allows GCC to
4989 use the optional PowerPC architecture instructions in the Graphics
4990 group, including floating-point select.
4992 The @samp{-mpowerpc64} option allows GCC to generate the additional
4993 64-bit instructions that are found in the full PowerPC64 architecture
4994 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
4995 @samp{-mno-powerpc64}.
4997 If you specify both @samp{-mno-power} and @samp{-mno-powerpc}, GCC
4998 will use only the instructions in the common subset of both
4999 architectures plus some special AIX common-mode calls, and will not use
5000 the MQ register. Specifying both @samp{-mpower} and @samp{-mpowerpc}
5001 permits GCC to use any instruction from either architecture and to
5002 allow use of the MQ register; specify this for the Motorola MPC601.
5004 @item -mnew-mnemonics
5005 @itemx -mold-mnemonics
5006 @kindex -mnew-mnemonics
5007 @kindex -mold-mnemonics
5008 Select which mnemonics to use in the generated assembler code.
5009 @samp{-mnew-mnemonics} requests output that uses the assembler mnemonics
5010 defined for the PowerPC architecture, while @samp{-mold-mnemonics}
5011 requests the assembler mnemonics defined for the POWER architecture.
5012 Instructions defined in only one architecture have only one mnemonic;
5013 GCC uses that mnemonic irrespective of which of these options is
5016 GCC defaults to the mnemonics appropriate for the architecture in
5017 use. Specifying @samp{-mcpu=@var{cpu_type}} sometimes overrides the
5018 value of these option. Unless you are building a cross-compiler, you
5019 should normally not specify either @samp{-mnew-mnemonics} or
5020 @samp{-mold-mnemonics}, but should instead accept the default.
5022 @item -mcpu=@var{cpu_type}
5024 Set architecture type, register usage, choice of mnemonics, and
5025 instruction scheduling parameters for machine type @var{cpu_type}.
5026 Supported values for @var{cpu_type} are @samp{rios}, @samp{rios1},
5027 @samp{rsc}, @samp{rios2}, @samp{rs64a}, @samp{601}, @samp{602},
5028 @samp{603}, @samp{603e}, @samp{604}, @samp{604e}, @samp{620},
5029 @samp{630}, @samp{740}, @samp{750}, @samp{power}, @samp{power2},
5030 @samp{powerpc}, @samp{403}, @samp{505}, @samp{801}, @samp{821},
5031 @samp{823}, and @samp{860} and @samp{common}. @samp{-mcpu=power},
5032 @samp{-mcpu=power2}, @samp{-mcpu=powerpc}, and @samp{-mcpu=powerpc64}
5033 specify generic POWER, POWER2, pure 32-bit PowerPC (i.e., not MPC601),
5034 and 64-bit PowerPC architecture machine types, with an appropriate,
5035 generic processor model assumed for scheduling purposes.@refill
5037 Specifying any of the following options:
5038 @samp{-mcpu=rios1}, @samp{-mcpu=rios2}, @samp{-mcpu=rsc},
5039 @samp{-mcpu=power}, or @samp{-mcpu=power2}
5040 enables the @samp{-mpower} option and disables the @samp{-mpowerpc} option;
5041 @samp{-mcpu=601} enables both the @samp{-mpower} and @samp{-mpowerpc} options.
5042 All of @samp{-mcpu=rs64a}, @samp{-mcpu=602}, @samp{-mcpu=603},
5043 @samp{-mcpu=603e}, @samp{-mcpu=604}, @samp{-mcpu=620}, @samp{-mcpu=630},
5044 @samp{-mcpu=740}, and @samp{-mcpu=750}
5045 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
5046 Exactly similarly, all of @samp{-mcpu=403},
5047 @samp{-mcpu=505}, @samp{-mcpu=821}, @samp{-mcpu=860} and @samp{-mcpu=powerpc}
5048 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
5049 @samp{-mcpu=common} disables both the
5050 @samp{-mpower} and @samp{-mpowerpc} options.@refill
5052 AIX versions 4 or greater selects @samp{-mcpu=common} by default, so
5053 that code will operate on all members of the RS/6000 POWER and PowerPC
5054 families. In that case, GCC will use only the instructions in the
5055 common subset of both architectures plus some special AIX common-mode
5056 calls, and will not use the MQ register. GCC assumes a generic
5057 processor model for scheduling purposes.
5059 Specifying any of the options @samp{-mcpu=rios1}, @samp{-mcpu=rios2},
5060 @samp{-mcpu=rsc}, @samp{-mcpu=power}, or @samp{-mcpu=power2} also
5061 disables the @samp{new-mnemonics} option. Specifying @samp{-mcpu=601},
5062 @samp{-mcpu=602}, @samp{-mcpu=603}, @samp{-mcpu=603e}, @samp{-mcpu=604},
5063 @samp{-mcpu=620}, @samp{-mcpu=630}, @samp{-mcpu=403}, @samp{-mcpu=505},
5064 @samp{-mcpu=821}, @samp{-mcpu=860} or @samp{-mcpu=powerpc} also enables
5065 the @samp{new-mnemonics} option.@refill
5067 Specifying @samp{-mcpu=403}, @samp{-mcpu=821}, or @samp{-mcpu=860} also
5068 enables the @samp{-msoft-float} option.
5070 @item -mtune=@var{cpu_type}
5071 Set the instruction scheduling parameters for machine type
5072 @var{cpu_type}, but do not set the architecture type, register usage,
5073 choice of mnemonics like @samp{-mcpu=}@var{cpu_type} would. The same
5074 values for @var{cpu_type} are used for @samp{-mtune=}@var{cpu_type} as
5075 for @samp{-mcpu=}@var{cpu_type}. The @samp{-mtune=}@var{cpu_type}
5076 option overrides the @samp{-mcpu=}@var{cpu_type} option in terms of
5077 instruction scheduling parameters.
5080 @itemx -mno-fp-in-toc
5081 @itemx -mno-sum-in-toc
5082 @itemx -mminimal-toc
5083 @kindex -mminimal-toc
5084 Modify generation of the TOC (Table Of Contents), which is created for
5085 every executable file. The @samp{-mfull-toc} option is selected by
5086 default. In that case, GCC will allocate at least one TOC entry for
5087 each unique non-automatic variable reference in your program. GCC
5088 will also place floating-point constants in the TOC. However, only
5089 16,384 entries are available in the TOC.
5091 If you receive a linker error message that saying you have overflowed
5092 the available TOC space, you can reduce the amount of TOC space used
5093 with the @samp{-mno-fp-in-toc} and @samp{-mno-sum-in-toc} options.
5094 @samp{-mno-fp-in-toc} prevents GCC from putting floating-point
5095 constants in the TOC and @samp{-mno-sum-in-toc} forces GCC to
5096 generate code to calculate the sum of an address and a constant at
5097 run-time instead of putting that sum into the TOC. You may specify one
5098 or both of these options. Each causes GCC to produce very slightly
5099 slower and larger code at the expense of conserving TOC space.
5101 If you still run out of space in the TOC even when you specify both of
5102 these options, specify @samp{-mminimal-toc} instead. This option causes
5103 GCC to make only one TOC entry for every file. When you specify this
5104 option, GCC will produce code that is slower and larger but which
5105 uses extremely little TOC space. You may wish to use this option
5106 only on files that contain less frequently executed code. @refill
5112 Enable 64-bit PowerPC ABI and calling convention: 64-bit pointers, 64-bit
5113 @code{long} type, and the infrastructure needed to support them.
5114 Specifying @samp{-m64} implies @samp{-mpowerpc64} and
5115 @samp{-mpowerpc}, while @samp{-m32} disables the 64-bit ABI and
5116 implies @samp{-mno-powerpc64}. GCC defaults to @samp{-m32}.
5121 On AIX, pass floating-point arguments to prototyped functions beyond the
5122 register save area (RSA) on the stack in addition to argument FPRs. The
5123 AIX calling convention was extended but not initially documented to
5124 handle an obscure K&R C case of calling a function that takes the
5125 address of its arguments with fewer arguments than declared. AIX XL
5126 compilers access floating point arguments which do not fit in the
5127 RSA from the stack when a subroutine is compiled without
5128 optimization. Because always storing floating-point arguments on the
5129 stack is inefficient and rarely needed, this option is not enabled by
5130 default and only is necessary when calling subroutines compiled by AIX
5131 XL compilers without optimization.
5135 Support @dfn{AIX Threads}. Link an application written to use
5136 @dfn{pthreads} with special libraries and startup code to enable the
5141 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE). Link an
5142 application written to use message passing with special startup code to
5143 enable the application to run. The system must have PE installed in the
5144 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
5145 must be overridden with the @samp{-specs=} option to specify the
5146 appropriate directory location. The Parallel Environment does not
5147 support threads, so the @samp{-mpe} option and the @samp{-mthreads}
5148 option are incompatible.
5152 @kindex -msoft-float
5153 Generate code that does not use (uses) the floating-point register set.
5154 Software floating point emulation is provided if you use the
5155 @samp{-msoft-float} option, and pass the option to GCC when linking.
5158 @itemx -mno-multiple
5159 Generate code that uses (does not use) the load multiple word
5160 instructions and the store multiple word instructions. These
5161 instructions are generated by default on POWER systems, and not
5162 generated on PowerPC systems. Do not use @samp{-mmultiple} on little
5163 endian PowerPC systems, since those instructions do not work when the
5164 processor is in little endian mode. The exceptions are PPC740 and
5165 PPC750 which permit the instructions usage in little endian mode.
5170 Generate code that uses (does not use) the load string instructions
5171 and the store string word instructions to save multiple registers and
5172 do small block moves. These instructions are generated by default on
5173 POWER systems, and not generated on PowerPC systems. Do not use
5174 @samp{-mstring} on little endian PowerPC systems, since those
5175 instructions do not work when the processor is in little endian mode.
5176 The exceptions are PPC740 and PPC750 which permit the instructions
5177 usage in little endian mode.
5182 Generate code that uses (does not use) the load or store instructions
5183 that update the base register to the address of the calculated memory
5184 location. These instructions are generated by default. If you use
5185 @samp{-mno-update}, there is a small window between the time that the
5186 stack pointer is updated and the address of the previous frame is
5187 stored, which means code that walks the stack frame across interrupts or
5188 signals may get corrupted data.
5191 @itemx -mno-fused-madd
5192 @kindex -mfused-madd
5193 Generate code that uses (does not use) the floating point multiply and
5194 accumulate instructions. These instructions are generated by default if
5195 hardware floating is used.
5197 @item -mno-bit-align
5200 On System V.4 and embedded PowerPC systems do not (do) force structures
5201 and unions that contain bit fields to be aligned to the base type of the
5204 For example, by default a structure containing nothing but 8
5205 @code{unsigned} bitfields of length 1 would be aligned to a 4 byte
5206 boundary and have a size of 4 bytes. By using @samp{-mno-bit-align},
5207 the structure would be aligned to a 1 byte boundary and be one byte in
5210 @item -mno-strict-align
5211 @itemx -mstrict-align
5212 @kindex -mstrict-align
5213 On System V.4 and embedded PowerPC systems do not (do) assume that
5214 unaligned memory references will be handled by the system.
5217 @itemx -mno-relocatable
5218 @kindex -mrelocatable
5219 On embedded PowerPC systems generate code that allows (does not allow)
5220 the program to be relocated to a different address at runtime. If you
5221 use @samp{-mrelocatable} on any module, all objects linked together must
5222 be compiled with @samp{-mrelocatable} or @samp{-mrelocatable-lib}.
5224 @item -mrelocatable-lib
5225 @itemx -mno-relocatable-lib
5226 On embedded PowerPC systems generate code that allows (does not allow)
5227 the program to be relocated to a different address at runtime. Modules
5228 compiled with @samp{-mrelocatable-lib} can be linked with either modules
5229 compiled without @samp{-mrelocatable} and @samp{-mrelocatable-lib} or
5230 with modules compiled with the @samp{-mrelocatable} options.
5234 On System V.4 and embedded PowerPC systems do not (do) assume that
5235 register 2 contains a pointer to a global area pointing to the addresses
5236 used in the program.
5239 @itemx -mlittle-endian
5240 On System V.4 and embedded PowerPC systems compile code for the
5241 processor in little endian mode. The @samp{-mlittle-endian} option is
5242 the same as @samp{-mlittle}.
5246 On System V.4 and embedded PowerPC systems compile code for the
5247 processor in big endian mode. The @samp{-mbig-endian} option is
5248 the same as @samp{-mbig}.
5251 On System V.4 and embedded PowerPC systems compile code using calling
5252 conventions that adheres to the March 1995 draft of the System V
5253 Application Binary Interface, PowerPC processor supplement. This is the
5254 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
5256 @item -mcall-sysv-eabi
5257 Specify both @samp{-mcall-sysv} and @samp{-meabi} options.
5259 @item -mcall-sysv-noeabi
5260 Specify both @samp{-mcall-sysv} and @samp{-mno-eabi} options.
5263 On System V.4 and embedded PowerPC systems compile code using calling
5264 conventions that are similar to those used on AIX. This is the
5265 default if you configured GCC using @samp{powerpc-*-eabiaix}.
5267 @item -mcall-solaris
5268 On System V.4 and embedded PowerPC systems compile code for the Solaris
5272 On System V.4 and embedded PowerPC systems compile code for the
5273 Linux-based GNU system.
5276 @itemx -mno-prototype
5277 On System V.4 and embedded PowerPC systems assume that all calls to
5278 variable argument functions are properly prototyped. Otherwise, the
5279 compiler must insert an instruction before every non prototyped call to
5280 set or clear bit 6 of the condition code register (@var{CR}) to
5281 indicate whether floating point values were passed in the floating point
5282 registers in case the function takes a variable arguments. With
5283 @samp{-mprototype}, only calls to prototyped variable argument functions
5284 will set or clear the bit.
5287 On embedded PowerPC systems, assume that the startup module is called
5288 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
5289 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
5293 On embedded PowerPC systems, assume that the startup module is called
5294 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
5298 On embedded PowerPC systems, assume that the startup module is called
5299 @file{crt0.o} and the standard C libraries are @file{libads.a} and
5303 On embedded PowerPC systems, assume that the startup module is called
5304 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
5308 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
5309 header to indicate that @samp{eabi} extended relocations are used.
5313 On System V.4 and embedded PowerPC systems do (do not) adhere to the
5314 Embedded Applications Binary Interface (eabi) which is a set of
5315 modifications to the System V.4 specifications. Selecting @code{-meabi}
5316 means that the stack is aligned to an 8 byte boundary, a function
5317 @code{__eabi} is called to from @code{main} to set up the eabi
5318 environment, and the @samp{-msdata} option can use both @code{r2} and
5319 @code{r13} to point to two separate small data areas. Selecting
5320 @code{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
5321 do not call an initialization function from @code{main}, and the
5322 @samp{-msdata} option will only use @code{r13} to point to a single
5323 small data area. The @samp{-meabi} option is on by default if you
5324 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
5327 On System V.4 and embedded PowerPC systems, put small initialized
5328 @code{const} global and static data in the @samp{.sdata2} section, which
5329 is pointed to by register @code{r2}. Put small initialized
5330 non-@code{const} global and static data in the @samp{.sdata} section,
5331 which is pointed to by register @code{r13}. Put small uninitialized
5332 global and static data in the @samp{.sbss} section, which is adjacent to
5333 the @samp{.sdata} section. The @samp{-msdata=eabi} option is
5334 incompatible with the @samp{-mrelocatable} option. The
5335 @samp{-msdata=eabi} option also sets the @samp{-memb} option.
5338 On System V.4 and embedded PowerPC systems, put small global and static
5339 data in the @samp{.sdata} section, which is pointed to by register
5340 @code{r13}. Put small uninitialized global and static data in the
5341 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
5342 The @samp{-msdata=sysv} option is incompatible with the
5343 @samp{-mrelocatable} option.
5345 @item -msdata=default
5347 On System V.4 and embedded PowerPC systems, if @samp{-meabi} is used,
5348 compile code the same as @samp{-msdata=eabi}, otherwise compile code the
5349 same as @samp{-msdata=sysv}.
5352 On System V.4 and embedded PowerPC systems, put small global and static
5353 data in the @samp{.sdata} section. Put small uninitialized global and
5354 static data in the @samp{.sbss} section. Do not use register @code{r13}
5355 to address small data however. This is the default behavior unless
5356 other @samp{-msdata} options are used.
5360 On embedded PowerPC systems, put all initialized global and static data
5361 in the @samp{.data} section, and all uninitialized data in the
5362 @samp{.bss} section.
5365 @cindex smaller data references (PowerPC)
5366 @cindex .sdata/.sdata2 references (PowerPC)
5367 On embedded PowerPC systems, put global and static items less than or
5368 equal to @var{num} bytes into the small data or bss sections instead of
5369 the normal data or bss section. By default, @var{num} is 8. The
5370 @samp{-G @var{num}} switch is also passed to the linker.
5371 All modules should be compiled with the same @samp{-G @var{num}} value.
5374 @itemx -mno-regnames
5375 On System V.4 and embedded PowerPC systems do (do not) emit register
5376 names in the assembly language output using symbolic forms.
5381 @subsection IBM RT Options
5383 @cindex IBM RT options
5385 These @samp{-m} options are defined for the IBM RT PC:
5389 Use an in-line code sequence for integer multiplies. This is the
5392 @item -mcall-lib-mul
5393 Call @code{lmul$$} for integer multiples.
5395 @item -mfull-fp-blocks
5396 Generate full-size floating point data blocks, including the minimum
5397 amount of scratch space recommended by IBM. This is the default.
5399 @item -mminimum-fp-blocks
5400 Do not include extra scratch space in floating point data blocks. This
5401 results in smaller code, but slower execution, since scratch space must
5402 be allocated dynamically.
5404 @cindex @file{varargs.h} and RT PC
5405 @cindex @file{stdarg.h} and RT PC
5406 @item -mfp-arg-in-fpregs
5407 Use a calling sequence incompatible with the IBM calling convention in
5408 which floating point arguments are passed in floating point registers.
5409 Note that @code{varargs.h} and @code{stdargs.h} will not work with
5410 floating point operands if this option is specified.
5412 @item -mfp-arg-in-gregs
5413 Use the normal calling convention for floating point arguments. This is
5416 @item -mhc-struct-return
5417 Return structures of more than one word in memory, rather than in a
5418 register. This provides compatibility with the MetaWare HighC (hc)
5419 compiler. Use the option @samp{-fpcc-struct-return} for compatibility
5420 with the Portable C Compiler (pcc).
5422 @item -mnohc-struct-return
5423 Return some structures of more than one word in registers, when
5424 convenient. This is the default. For compatibility with the
5425 IBM-supplied compilers, use the option @samp{-fpcc-struct-return} or the
5426 option @samp{-mhc-struct-return}.
5430 @subsection MIPS Options
5431 @cindex MIPS options
5433 These @samp{-m} options are defined for the MIPS family of computers:
5436 @item -mcpu=@var{cpu type}
5437 Assume the defaults for the machine type @var{cpu type} when scheduling
5438 instructions. The choices for @var{cpu type} are @samp{r2000}, @samp{r3000},
5439 @samp{r3900}, @samp{r4000}, @samp{r4100}, @samp{r4300}, @samp{r4400},
5440 @samp{r4600}, @samp{r4650}, @samp{r5000}, @samp{r6000}, @samp{r8000},
5441 and @samp{orion}. Additionally, the @samp{r2000}, @samp{r3000},
5442 @samp{r4000}, @samp{r5000}, and @samp{r6000} can be abbreviated as
5443 @samp{r2k} (or @samp{r2K}), @samp{r3k}, etc. While picking a specific
5444 @var{cpu type} will schedule things appropriately for that particular
5445 chip, the compiler will not generate any code that does not meet level 1
5446 of the MIPS ISA (instruction set architecture) without a @samp{-mipsX}
5447 or @samp{-mabi} switch being used.
5450 Issue instructions from level 1 of the MIPS ISA. This is the default.
5451 @samp{r3000} is the default @var{cpu type} at this ISA level.
5454 Issue instructions from level 2 of the MIPS ISA (branch likely, square
5455 root instructions). @samp{r6000} is the default @var{cpu type} at this
5459 Issue instructions from level 3 of the MIPS ISA (64 bit instructions).
5460 @samp{r4000} is the default @var{cpu type} at this ISA level.
5463 Issue instructions from level 4 of the MIPS ISA (conditional move,
5464 prefetch, enhanced FPU instructions). @samp{r8000} is the default
5465 @var{cpu type} at this ISA level.
5468 Assume that 32 32-bit floating point registers are available. This is
5472 Assume that 32 64-bit floating point registers are available. This is
5473 the default when the @samp{-mips3} option is used.
5476 Assume that 32 32-bit general purpose registers are available. This is
5480 Assume that 32 64-bit general purpose registers are available. This is
5481 the default when the @samp{-mips3} option is used.
5484 Force int and long types to be 64 bits wide. See @samp{-mlong32} for an
5485 explanation of the default, and the width of pointers.
5488 Force long types to be 64 bits wide. See @samp{-mlong32} for an
5489 explanation of the default, and the width of pointers.
5492 Force long, int, and pointer types to be 32 bits wide.
5494 If none of @samp{-mlong32}, @samp{-mlong64}, or @samp{-mint64} are set,
5495 the size of ints, longs, and pointers depends on the ABI and ISA choosen.
5496 For @samp{-mabi=32}, and @samp{-mabi=n32}, ints and longs are 32 bits
5497 wide. For @samp{-mabi=64}, ints are 32 bits, and longs are 64 bits wide.
5498 For @samp{-mabi=eabi} and either @samp{-mips1} or @samp{-mips2}, ints
5499 and longs are 32 bits wide. For @samp{-mabi=eabi} and higher ISAs, ints
5500 are 32 bits, and longs are 64 bits wide. The width of pointer types is
5501 the smaller of the width of longs or the width of general purpose
5502 registers (which in turn depends on the ISA).
5509 Generate code for the indicated ABI. The default instruction level is
5510 @samp{-mips1} for @samp{32}, @samp{-mips3} for @samp{n32}, and
5511 @samp{-mips4} otherwise. Conversely, with @samp{-mips1} or
5512 @samp{-mips2}, the default ABI is @samp{32}; otherwise, the default ABI
5516 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
5517 add normal debug information. This is the default for all
5518 platforms except for the OSF/1 reference platform, using the OSF/rose
5519 object format. If the either of the @samp{-gstabs} or @samp{-gstabs+}
5520 switches are used, the @file{mips-tfile} program will encapsulate the
5521 stabs within MIPS ECOFF.
5524 Generate code for the GNU assembler. This is the default on the OSF/1
5525 reference platform, using the OSF/rose object format. Also, this is
5526 the default if the configure option @samp{--with-gnu-as} is used.
5528 @item -msplit-addresses
5529 @itemx -mno-split-addresses
5530 Generate code to load the high and low parts of address constants separately.
5531 This allows @code{gcc} to optimize away redundant loads of the high order
5532 bits of addresses. This optimization requires GNU as and GNU ld.
5533 This optimization is enabled by default for some embedded targets where
5534 GNU as and GNU ld are standard.
5538 The @samp{-mrnames} switch says to output code using the MIPS software
5539 names for the registers, instead of the hardware names (ie, @var{a0}
5540 instead of @var{$4}). The only known assembler that supports this option
5541 is the Algorithmics assembler.
5545 The @samp{-mgpopt} switch says to write all of the data declarations
5546 before the instructions in the text section, this allows the MIPS
5547 assembler to generate one word memory references instead of using two
5548 words for short global or static data items. This is on by default if
5549 optimization is selected.
5553 For each non-inline function processed, the @samp{-mstats} switch
5554 causes the compiler to emit one line to the standard error file to
5555 print statistics about the program (number of registers saved, stack
5560 The @samp{-mmemcpy} switch makes all block moves call the appropriate
5561 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
5562 generating inline code.
5565 @itemx -mno-mips-tfile
5566 The @samp{-mno-mips-tfile} switch causes the compiler not
5567 postprocess the object file with the @file{mips-tfile} program,
5568 after the MIPS assembler has generated it to add debug support. If
5569 @file{mips-tfile} is not run, then no local variables will be
5570 available to the debugger. In addition, @file{stage2} and
5571 @file{stage3} objects will have the temporary file names passed to the
5572 assembler embedded in the object file, which means the objects will
5573 not compare the same. The @samp{-mno-mips-tfile} switch should only
5574 be used when there are bugs in the @file{mips-tfile} program that
5575 prevents compilation.
5578 Generate output containing library calls for floating point.
5579 @strong{Warning:} the requisite libraries are not part of GCC.
5580 Normally the facilities of the machine's usual C compiler are used, but
5581 this can't be done directly in cross-compilation. You must make your
5582 own arrangements to provide suitable library functions for
5586 Generate output containing floating point instructions. This is the
5587 default if you use the unmodified sources.
5590 @itemx -mno-abicalls
5591 Emit (or do not emit) the pseudo operations @samp{.abicalls},
5592 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
5593 position independent code.
5596 @itemx -mno-long-calls
5597 Do all calls with the @samp{JALR} instruction, which requires
5598 loading up a function's address into a register before the call.
5599 You need to use this switch, if you call outside of the current
5600 512 megabyte segment to functions that are not through pointers.
5603 @itemx -mno-half-pic
5604 Put pointers to extern references into the data section and load them
5605 up, rather than put the references in the text section.
5607 @item -membedded-pic
5608 @itemx -mno-embedded-pic
5609 Generate PIC code suitable for some embedded systems. All calls are
5610 made using PC relative address, and all data is addressed using the $gp
5611 register. No more than 65536 bytes of global data may be used. This
5612 requires GNU as and GNU ld which do most of the work. This currently
5613 only works on targets which use ECOFF; it does not work with ELF.
5615 @item -membedded-data
5616 @itemx -mno-embedded-data
5617 Allocate variables to the read-only data section first if possible, then
5618 next in the small data section if possible, otherwise in data. This gives
5619 slightly slower code than the default, but reduces the amount of RAM required
5620 when executing, and thus may be preferred for some embedded systems.
5622 @item -muninit-const-in-rodata
5623 @itemx -mno-uninit-const-in-rodata
5624 When used together with -membedded-data, it will always store uninitialized
5625 const variables in the read-only data section.
5627 @item -msingle-float
5628 @itemx -mdouble-float
5629 The @samp{-msingle-float} switch tells gcc to assume that the floating
5630 point coprocessor only supports single precision operations, as on the
5631 @samp{r4650} chip. The @samp{-mdouble-float} switch permits gcc to use
5632 double precision operations. This is the default.
5636 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
5637 as on the @samp{r4650} chip.
5640 Turns on @samp{-msingle-float}, @samp{-mmad}, and, at least for now,
5645 Enable 16-bit instructions.
5648 Use the entry and exit pseudo ops. This option can only be used with
5652 Compile code for the processor in little endian mode.
5653 The requisite libraries are assumed to exist.
5656 Compile code for the processor in big endian mode.
5657 The requisite libraries are assumed to exist.
5660 @cindex smaller data references (MIPS)
5661 @cindex gp-relative references (MIPS)
5662 Put global and static items less than or equal to @var{num} bytes into
5663 the small data or bss sections instead of the normal data or bss
5664 section. This allows the assembler to emit one word memory reference
5665 instructions based on the global pointer (@var{gp} or @var{$28}),
5666 instead of the normal two words used. By default, @var{num} is 8 when
5667 the MIPS assembler is used, and 0 when the GNU assembler is used. The
5668 @samp{-G @var{num}} switch is also passed to the assembler and linker.
5669 All modules should be compiled with the same @samp{-G @var{num}}
5673 Tell the MIPS assembler to not run its preprocessor over user
5674 assembler files (with a @samp{.s} suffix) when assembling them.
5678 These options are defined by the macro
5679 @code{TARGET_SWITCHES} in the machine description. The default for the
5680 options is also defined by that macro, which enables you to change the
5685 @subsection Intel 386 Options
5686 @cindex i386 Options
5687 @cindex Intel 386 Options
5689 These @samp{-m} options are defined for the i386 family of computers:
5692 @item -mcpu=@var{cpu type}
5693 Assume the defaults for the machine type @var{cpu type} when scheduling
5694 instructions. The choices for @var{cpu type} are:
5696 @multitable @columnfractions .20 .20 .20 .20
5697 @item @samp{i386} @tab @samp{i486} @tab @samp{i586} @tab @samp{i686}
5698 @item @samp{pentium} @tab @samp{pentiumpro} @tab @samp{k6}
5701 While picking a specific @var{cpu type} will schedule things appropriately
5702 for that particular chip, the compiler will not generate any code that
5703 does not run on the i386 without the @samp{-march=@var{cpu type}} option
5704 being used. @samp{i586} is equivalent to @samp{pentium} and @samp{i686}
5705 is equivalent to @samp{pentiumpro}. @samp{k6} is the AMD chip as
5706 opposed to the Intel ones.
5708 @item -march=@var{cpu type}
5709 Generate instructions for the machine type @var{cpu type}. The choices
5710 for @var{cpu type} are the same as for @samp{-mcpu}. Moreover,
5711 specifying @samp{-march=@var{cpu type}} implies @samp{-mcpu=@var{cpu type}}.
5717 Synonyms for -mcpu=i386, -mcpu=i486, -mcpu=pentium, and -mcpu=pentiumpro
5718 respectively. These synonyms are deprecated.
5722 Control whether or not the compiler uses IEEE floating point
5723 comparisons. These handle correctly the case where the result of a
5724 comparison is unordered.
5727 Generate output containing library calls for floating point.
5728 @strong{Warning:} the requisite libraries are not part of GCC.
5729 Normally the facilities of the machine's usual C compiler are used, but
5730 this can't be done directly in cross-compilation. You must make your
5731 own arrangements to provide suitable library functions for
5734 On machines where a function returns floating point results in the 80387
5735 register stack, some floating point opcodes may be emitted even if
5736 @samp{-msoft-float} is used.
5738 @item -mno-fp-ret-in-387
5739 Do not use the FPU registers for return values of functions.
5741 The usual calling convention has functions return values of types
5742 @code{float} and @code{double} in an FPU register, even if there
5743 is no FPU. The idea is that the operating system should emulate
5746 The option @samp{-mno-fp-ret-in-387} causes such values to be returned
5747 in ordinary CPU registers instead.
5749 @item -mno-fancy-math-387
5750 Some 387 emulators do not support the @code{sin}, @code{cos} and
5751 @code{sqrt} instructions for the 387. Specify this option to avoid
5752 generating those instructions. This option is the default on FreeBSD.
5753 As of revision 2.6.1, these instructions are not generated unless you
5754 also use the @samp{-ffast-math} switch.
5756 @item -malign-double
5757 @itemx -mno-align-double
5758 Control whether GCC aligns @code{double}, @code{long double}, and
5759 @code{long long} variables on a two word boundary or a one word
5760 boundary. Aligning @code{double} variables on a two word boundary will
5761 produce code that runs somewhat faster on a @samp{Pentium} at the
5762 expense of more memory.
5764 @strong{Warning:} if you use the @samp{-malign-double} switch,
5765 structures containing the above types will be aligned differently than
5766 the published application binary interface specifications for the 386.
5769 @itemx -mno-svr3-shlib
5770 Control whether GCC places uninitialized locals into @code{bss} or
5771 @code{data}. @samp{-msvr3-shlib} places these locals into @code{bss}.
5772 These options are meaningful only on System V Release 3.
5774 @item -mno-wide-multiply
5775 @itemx -mwide-multiply
5776 Control whether GCC uses the @code{mul} and @code{imul} that produce
5777 64 bit results in @code{eax:edx} from 32 bit operands to do @code{long
5778 long} multiplies and 32-bit division by constants.
5781 Use a different function-calling convention, in which functions that
5782 take a fixed number of arguments return with the @code{ret} @var{num}
5783 instruction, which pops their arguments while returning. This saves one
5784 instruction in the caller since there is no need to pop the arguments
5787 You can specify that an individual function is called with this calling
5788 sequence with the function attribute @samp{stdcall}. You can also
5789 override the @samp{-mrtd} option by using the function attribute
5790 @samp{cdecl}. @xref{Function Attributes}.
5792 @strong{Warning:} this calling convention is incompatible with the one
5793 normally used on Unix, so you cannot use it if you need to call
5794 libraries compiled with the Unix compiler.
5796 Also, you must provide function prototypes for all functions that
5797 take variable numbers of arguments (including @code{printf});
5798 otherwise incorrect code will be generated for calls to those
5801 In addition, seriously incorrect code will result if you call a
5802 function with too many arguments. (Normally, extra arguments are
5803 harmlessly ignored.)
5805 @item -mreg-alloc=@var{regs}
5806 Control the default allocation order of integer registers. The
5807 string @var{regs} is a series of letters specifying a register. The
5808 supported letters are: @code{a} allocate EAX; @code{b} allocate EBX;
5809 @code{c} allocate ECX; @code{d} allocate EDX; @code{S} allocate ESI;
5810 @code{D} allocate EDI; @code{B} allocate EBP.
5812 @item -mregparm=@var{num}
5813 Control how many registers are used to pass integer arguments. By
5814 default, no registers are used to pass arguments, and at most 3
5815 registers can be used. You can control this behavior for a specific
5816 function by using the function attribute @samp{regparm}.
5817 @xref{Function Attributes}.
5819 @strong{Warning:} if you use this switch, and
5820 @var{num} is nonzero, then you must build all modules with the same
5821 value, including any libraries. This includes the system libraries and
5824 @item -malign-loops=@var{num}
5825 Align loops to a 2 raised to a @var{num} byte boundary. If
5826 @samp{-malign-loops} is not specified, the default is 2 unless
5827 gas 2.8 (or later) is being used in which case the default is
5828 to align the loop on a 16 byte boundary if it is less than 8
5831 @item -malign-jumps=@var{num}
5832 Align instructions that are only jumped to to a 2 raised to a @var{num}
5833 byte boundary. If @samp{-malign-jumps} is not specified, the default is
5834 2 if optimizing for a 386, and 4 if optimizing for a 486 unless
5835 gas 2.8 (or later) is being used in which case the default is
5836 to align the instruction on a 16 byte boundary if it is less
5839 @item -malign-functions=@var{num}
5840 Align the start of functions to a 2 raised to @var{num} byte boundary.
5841 If @samp{-malign-functions} is not specified, the default is 2 if optimizing
5842 for a 386, and 4 if optimizing for a 486.
5844 @item -mpreferred-stack-boundary=@var{num}
5845 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
5846 byte boundary. If @samp{-mpreferred-stack-boundary} is not specified,
5847 the default is 4 (16 bytes or 128 bits).
5849 The stack is required to be aligned on a 4 byte boundary. On Pentium
5850 and PentiumPro, @code{double} and @code{long double} values should be
5851 aligned to an 8 byte boundary (see @samp{-malign-double}) or suffer
5852 significant run time performance penalties. On Pentium III, the
5853 Streaming SIMD Extention (SSE) data type @code{__m128} suffers similar
5854 penalties if it is not 16 byte aligned.
5856 To ensure proper alignment of this values on the stack, the stack boundary
5857 must be as aligned as that required by any value stored on the stack.
5858 Further, every function must be generated such that it keeps the stack
5859 aligned. Thus calling a function compiled with a higher preferred
5860 stack boundary from a function compiled with a lower preferred stack
5861 boundary will most likely misalign the stack. It is recommended that
5862 libraries that use callbacks always use the default setting.
5864 This extra alignment does consume extra stack space. Code that is sensitive
5865 to stack space usage, such as embedded systems and operating system kernels,
5866 may want to reduce the preferred alignment to
5867 @samp{-mpreferred-stack-boundary=2}.
5871 @subsection HPPA Options
5872 @cindex HPPA Options
5874 These @samp{-m} options are defined for the HPPA family of computers:
5877 @item -march=@var{architecture type}
5878 Generate code for the specified architecture. The choices for
5879 @var{architecture type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
5880 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
5881 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
5882 architecture option for your machine. Code compiled for lower numbered
5883 architectures will run on higher numbered architectures, but not the
5886 PA 2.0 support currently requires gas snapshot 19990413 or later. The
5887 next release of binutils (current is 2.9.1) will probably contain PA 2.0
5891 @itemx -mpa-risc-1-1
5892 @itemx -mpa-risc-2-0
5893 Synonyms for -march=1.0, -march=1.1, and -march=2.0 respectively.
5896 Generate code suitable for big switch tables. Use this option only if
5897 the assembler/linker complain about out of range branches within a switch
5900 @item -mjump-in-delay
5901 Fill delay slots of function calls with unconditional jump instructions
5902 by modifying the return pointer for the function call to be the target
5903 of the conditional jump.
5905 @item -mdisable-fpregs
5906 Prevent floating point registers from being used in any manner. This is
5907 necessary for compiling kernels which perform lazy context switching of
5908 floating point registers. If you use this option and attempt to perform
5909 floating point operations, the compiler will abort.
5911 @item -mdisable-indexing
5912 Prevent the compiler from using indexing address modes. This avoids some
5913 rather obscure problems when compiling MIG generated code under MACH.
5915 @item -mno-space-regs
5916 Generate code that assumes the target has no space registers. This allows
5917 GCC to generate faster indirect calls and use unscaled index address modes.
5919 Such code is suitable for level 0 PA systems and kernels.
5921 @item -mfast-indirect-calls
5922 Generate code that assumes calls never cross space boundaries. This
5923 allows GCC to emit code which performs faster indirect calls.
5925 This option will not work in the presense of shared libraries or nested
5928 @item -mlong-load-store
5929 Generate 3-instruction load and store sequences as sometimes required by
5930 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
5933 @item -mportable-runtime
5934 Use the portable calling conventions proposed by HP for ELF systems.
5937 Enable the use of assembler directives only GAS understands.
5939 @item -mschedule=@var{cpu type}
5940 Schedule code according to the constraints for the machine type
5941 @var{cpu type}. The choices for @var{cpu type} are @samp{700}
5942 @samp{7100}, @samp{7100LC}, @samp{7200}, and @samp{8000}. Refer to
5943 @file{/usr/lib/sched.models} on an HP-UX system to determine the
5944 proper scheduling option for your machine.
5947 Enable the optimization pass in the HPUX linker. Note this makes symbolic
5948 debugging impossible. It also triggers a bug in the HPUX 8 and HPUX 9 linkers
5949 in which they give bogus error messages when linking some programs.
5952 Generate output containing library calls for floating point.
5953 @strong{Warning:} the requisite libraries are not available for all HPPA
5954 targets. Normally the facilities of the machine's usual C compiler are
5955 used, but this cannot be done directly in cross-compilation. You must make
5956 your own arrangements to provide suitable library functions for
5957 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
5958 does provide software floating point support.
5960 @samp{-msoft-float} changes the calling convention in the output file;
5961 therefore, it is only useful if you compile @emph{all} of a program with
5962 this option. In particular, you need to compile @file{libgcc.a}, the
5963 library that comes with GCC, with @samp{-msoft-float} in order for
5967 @node Intel 960 Options
5968 @subsection Intel 960 Options
5970 These @samp{-m} options are defined for the Intel 960 implementations:
5973 @item -m@var{cpu type}
5974 Assume the defaults for the machine type @var{cpu type} for some of
5975 the other options, including instruction scheduling, floating point
5976 support, and addressing modes. The choices for @var{cpu type} are
5977 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
5978 @samp{sa}, and @samp{sb}.
5984 The @samp{-mnumerics} option indicates that the processor does support
5985 floating-point instructions. The @samp{-msoft-float} option indicates
5986 that floating-point support should not be assumed.
5988 @item -mleaf-procedures
5989 @itemx -mno-leaf-procedures
5990 Do (or do not) attempt to alter leaf procedures to be callable with the
5991 @code{bal} instruction as well as @code{call}. This will result in more
5992 efficient code for explicit calls when the @code{bal} instruction can be
5993 substituted by the assembler or linker, but less efficient code in other
5994 cases, such as calls via function pointers, or using a linker that doesn't
5995 support this optimization.
5998 @itemx -mno-tail-call
5999 Do (or do not) make additional attempts (beyond those of the
6000 machine-independent portions of the compiler) to optimize tail-recursive
6001 calls into branches. You may not want to do this because the detection of
6002 cases where this is not valid is not totally complete. The default is
6003 @samp{-mno-tail-call}.
6005 @item -mcomplex-addr
6006 @itemx -mno-complex-addr
6007 Assume (or do not assume) that the use of a complex addressing mode is a
6008 win on this implementation of the i960. Complex addressing modes may not
6009 be worthwhile on the K-series, but they definitely are on the C-series.
6010 The default is currently @samp{-mcomplex-addr} for all processors except
6014 @itemx -mno-code-align
6015 Align code to 8-byte boundaries for faster fetching (or don't bother).
6016 Currently turned on by default for C-series implementations only.
6019 @item -mclean-linkage
6020 @itemx -mno-clean-linkage
6021 These options are not fully implemented.
6025 @itemx -mic2.0-compat
6026 @itemx -mic3.0-compat
6027 Enable compatibility with iC960 v2.0 or v3.0.
6031 Enable compatibility with the iC960 assembler.
6033 @item -mstrict-align
6034 @itemx -mno-strict-align
6035 Do not permit (do permit) unaligned accesses.
6038 Enable structure-alignment compatibility with Intel's gcc release version
6039 1.3 (based on gcc 1.37). This option implies @samp{-mstrict-align}.
6041 @item -mlong-double-64
6042 Implement type @samp{long double} as 64-bit floating point numbers.
6043 Without the option @samp{long double} is implemented by 80-bit
6044 floating point numbers. The only reason we have it because there is
6045 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
6046 is only useful for people using soft-float targets. Otherwise, we
6047 should recommend against use of it.
6051 @node DEC Alpha Options
6052 @subsection DEC Alpha Options
6054 These @samp{-m} options are defined for the DEC Alpha implementations:
6057 @item -mno-soft-float
6059 Use (do not use) the hardware floating-point instructions for
6060 floating-point operations. When @code{-msoft-float} is specified,
6061 functions in @file{libgcc1.c} will be used to perform floating-point
6062 operations. Unless they are replaced by routines that emulate the
6063 floating-point operations, or compiled in such a way as to call such
6064 emulations routines, these routines will issue floating-point
6065 operations. If you are compiling for an Alpha without floating-point
6066 operations, you must ensure that the library is built so as not to call
6069 Note that Alpha implementations without floating-point operations are
6070 required to have floating-point registers.
6074 Generate code that uses (does not use) the floating-point register set.
6075 @code{-mno-fp-regs} implies @code{-msoft-float}. If the floating-point
6076 register set is not used, floating point operands are passed in integer
6077 registers as if they were integers and floating-point results are passed
6078 in $0 instead of $f0. This is a non-standard calling sequence, so any
6079 function with a floating-point argument or return value called by code
6080 compiled with @code{-mno-fp-regs} must also be compiled with that
6083 A typical use of this option is building a kernel that does not use,
6084 and hence need not save and restore, any floating-point registers.
6087 The Alpha architecture implements floating-point hardware optimized for
6088 maximum performance. It is mostly compliant with the IEEE floating
6089 point standard. However, for full compliance, software assistance is
6090 required. This option generates code fully IEEE compliant code
6091 @emph{except} that the @var{inexact flag} is not maintained (see below).
6092 If this option is turned on, the CPP macro @code{_IEEE_FP} is defined
6093 during compilation. The option is a shorthand for: @samp{-D_IEEE_FP
6094 -mfp-trap-mode=su -mtrap-precision=i -mieee-conformant}. The resulting
6095 code is less efficient but is able to correctly support denormalized
6096 numbers and exceptional IEEE values such as not-a-number and plus/minus
6097 infinity. Other Alpha compilers call this option
6098 @code{-ieee_with_no_inexact}.
6100 @item -mieee-with-inexact
6101 @c overfull hbox here --bob 22 jul96
6102 @c original text between ignore ... end ignore
6104 This is like @samp{-mieee} except the generated code also maintains the
6105 IEEE @var{inexact flag}. Turning on this option causes the generated
6106 code to implement fully-compliant IEEE math. The option is a shorthand
6107 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus @samp{-mieee-conformant},
6108 @samp{-mfp-trap-mode=sui}, and @samp{-mtrap-precision=i}. On some Alpha
6109 implementations the resulting code may execute significantly slower than
6110 the code generated by default. Since there is very little code that
6111 depends on the @var{inexact flag}, you should normally not specify this
6112 option. Other Alpha compilers call this option
6113 @samp{-ieee_with_inexact}.
6115 @c changed paragraph
6116 This is like @samp{-mieee} except the generated code also maintains the
6117 IEEE @var{inexact flag}. Turning on this option causes the generated
6118 code to implement fully-compliant IEEE math. The option is a shorthand
6119 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus the three following:
6120 @samp{-mieee-conformant},
6121 @samp{-mfp-trap-mode=sui},
6122 and @samp{-mtrap-precision=i}.
6123 On some Alpha implementations the resulting code may execute
6124 significantly slower than the code generated by default. Since there
6125 is very little code that depends on the @var{inexact flag}, you should
6126 normally not specify this option. Other Alpha compilers call this
6127 option @samp{-ieee_with_inexact}.
6128 @c end changes to prevent overfull hboxes
6130 @item -mfp-trap-mode=@var{trap mode}
6131 This option controls what floating-point related traps are enabled.
6132 Other Alpha compilers call this option @samp{-fptm }@var{trap mode}.
6133 The trap mode can be set to one of four values:
6137 This is the default (normal) setting. The only traps that are enabled
6138 are the ones that cannot be disabled in software (e.g., division by zero
6142 In addition to the traps enabled by @samp{n}, underflow traps are enabled
6146 Like @samp{su}, but the instructions are marked to be safe for software
6147 completion (see Alpha architecture manual for details).
6150 Like @samp{su}, but inexact traps are enabled as well.
6153 @item -mfp-rounding-mode=@var{rounding mode}
6154 Selects the IEEE rounding mode. Other Alpha compilers call this option
6155 @samp{-fprm }@var{rounding mode}. The @var{rounding mode} can be one
6160 Normal IEEE rounding mode. Floating point numbers are rounded towards
6161 the nearest machine number or towards the even machine number in case
6165 Round towards minus infinity.
6168 Chopped rounding mode. Floating point numbers are rounded towards zero.
6171 Dynamic rounding mode. A field in the floating point control register
6172 (@var{fpcr}, see Alpha architecture reference manual) controls the
6173 rounding mode in effect. The C library initializes this register for
6174 rounding towards plus infinity. Thus, unless your program modifies the
6175 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
6178 @item -mtrap-precision=@var{trap precision}
6179 In the Alpha architecture, floating point traps are imprecise. This
6180 means without software assistance it is impossible to recover from a
6181 floating trap and program execution normally needs to be terminated.
6182 GCC can generate code that can assist operating system trap handlers
6183 in determining the exact location that caused a floating point trap.
6184 Depending on the requirements of an application, different levels of
6185 precisions can be selected:
6189 Program precision. This option is the default and means a trap handler
6190 can only identify which program caused a floating point exception.
6193 Function precision. The trap handler can determine the function that
6194 caused a floating point exception.
6197 Instruction precision. The trap handler can determine the exact
6198 instruction that caused a floating point exception.
6201 Other Alpha compilers provide the equivalent options called
6202 @samp{-scope_safe} and @samp{-resumption_safe}.
6204 @item -mieee-conformant
6205 This option marks the generated code as IEEE conformant. You must not
6206 use this option unless you also specify @samp{-mtrap-precision=i} and either
6207 @samp{-mfp-trap-mode=su} or @samp{-mfp-trap-mode=sui}. Its only effect
6208 is to emit the line @samp{.eflag 48} in the function prologue of the
6209 generated assembly file. Under DEC Unix, this has the effect that
6210 IEEE-conformant math library routines will be linked in.
6212 @item -mbuild-constants
6213 Normally GCC examines a 32- or 64-bit integer constant to
6214 see if it can construct it from smaller constants in two or three
6215 instructions. If it cannot, it will output the constant as a literal and
6216 generate code to load it from the data segment at runtime.
6218 Use this option to require GCC to construct @emph{all} integer constants
6219 using code, even if it takes more instructions (the maximum is six).
6221 You would typically use this option to build a shared library dynamic
6222 loader. Itself a shared library, it must relocate itself in memory
6223 before it can find the variables and constants in its own data segment.
6227 Select whether to generate code to be assembled by the vendor-supplied
6228 assembler (@samp{-malpha-as}) or by the GNU assembler @samp{-mgas}.
6236 Indicate whether GCC should generate code to use the optional BWX,
6237 CIX, and MAX instruction sets. The default is to use the instruction sets
6238 supported by the CPU type specified via @samp{-mcpu=} option or that
6239 of the CPU on which GCC was built if none was specified.
6241 @item -mcpu=@var{cpu_type}
6242 Set the instruction set, register set, and instruction scheduling
6243 parameters for machine type @var{cpu_type}. You can specify either the
6244 @samp{EV} style name or the corresponding chip number. GCC
6245 supports scheduling parameters for the EV4 and EV5 family of processors
6246 and will choose the default values for the instruction set from
6247 the processor you specify. If you do not specify a processor type,
6248 GCC will default to the processor on which the compiler was built.
6250 Supported values for @var{cpu_type} are
6255 Schedules as an EV4 and has no instruction set extensions.
6259 Schedules as an EV5 and has no instruction set extensions.
6263 Schedules as an EV5 and supports the BWX extension.
6268 Schedules as an EV5 and supports the BWX and MAX extensions.
6272 Schedules as an EV5 (until Digital releases the scheduling parameters
6273 for the EV6) and supports the BWX, CIX, and MAX extensions.
6276 @item -mmemory-latency=@var{time}
6277 Sets the latency the scheduler should assume for typical memory
6278 references as seen by the application. This number is highly
6279 dependant on the memory access patterns used by the application
6280 and the size of the external cache on the machine.
6282 Valid options for @var{time} are
6286 A decimal number representing clock cycles.
6292 The compiler contains estimates of the number of clock cycles for
6293 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
6294 (also called Dcache, Scache, and Bcache), as well as to main memory.
6295 Note that L3 is only valid for EV5.
6300 @node Clipper Options
6301 @subsection Clipper Options
6303 These @samp{-m} options are defined for the Clipper implementations:
6307 Produce code for a C300 Clipper processor. This is the default.
6310 Produce code for a C400 Clipper processor i.e. use floating point
6314 @node H8/300 Options
6315 @subsection H8/300 Options
6317 These @samp{-m} options are defined for the H8/300 implementations:
6321 Shorten some address references at link time, when possible; uses the
6322 linker option @samp{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
6323 ld.info, Using ld}, for a fuller description.
6326 Generate code for the H8/300H.
6329 Generate code for the H8/S.
6332 Make @code{int} data 32 bits by default.
6335 On the h8/300h, use the same alignment rules as for the h8/300.
6336 The default for the h8/300h is to align longs and floats on 4 byte boundaries.
6337 @samp{-malign-300} causes them to be aligned on 2 byte boundaries.
6338 This option has no effect on the h8/300.
6342 @subsection SH Options
6344 These @samp{-m} options are defined for the SH implementations:
6348 Generate code for the SH1.
6351 Generate code for the SH2.
6354 Generate code for the SH3.
6357 Generate code for the SH3e.
6360 Compile code for the processor in big endian mode.
6363 Compile code for the processor in little endian mode.
6366 Align doubles at 64 bit boundaries. Note that this changes the calling
6367 conventions, and thus some functions from the standard C library will
6368 not work unless you recompile it first with -mdalign.
6371 Shorten some address references at link time, when possible; uses the
6372 linker option @samp{-relax}.
6375 @node System V Options
6376 @subsection Options for System V
6378 These additional options are available on System V Release 4 for
6379 compatibility with other compilers on those systems:
6383 Create a shared object.
6384 It is recommended that @samp{-symbolic} or @samp{-shared} be used instead.
6387 Identify the versions of each tool used by the compiler, in a
6388 @code{.ident} assembler directive in the output.
6391 Refrain from adding @code{.ident} directives to the output file (this is
6394 @item -YP,@var{dirs}
6395 Search the directories @var{dirs}, and no others, for libraries
6396 specified with @samp{-l}.
6399 Look in the directory @var{dir} to find the M4 preprocessor.
6400 The assembler uses this option.
6401 @c This is supposed to go with a -Yd for predefined M4 macro files, but
6402 @c the generic assembler that comes with Solaris takes just -Ym.
6405 @node TMS320C3x/C4x Options
6406 @subsection TMS320C3x/C4x Options
6407 @cindex TMS320C3x/C4x Options
6409 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
6413 @item -mcpu=@var{cpu_type}
6414 Set the instruction set, register set, and instruction scheduling
6415 parameters for machine type @var{cpu_type}. Supported values for
6416 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
6417 @samp{c44}. The default is @samp{c40} to generate code for the
6422 @itemx -msmall-memory
6424 Generates code for the big or small memory model. The small memory
6425 model assumed that all data fits into one 64K word page. At run-time
6426 the data page (DP) register must be set to point to the 64K page
6427 containing the .bss and .data program sections. The big memory model is
6428 the default and requires reloading of the DP register for every direct
6433 Allow (disallow) allocation of general integer operands into the block
6438 Enable (disable) generation of code using decrement and branch,
6439 DBcond(D), instructions. This is enabled by default for the C4x. To be
6440 on the safe side, this is disabled for the C3x, since the maximum
6441 iteration count on the C3x is 2^23 + 1 (but who iterates loops more than
6442 2^23 times on the C3x?). Note that GCC will try to reverse a loop so
6443 that it can utilise the decrement and branch instruction, but will give
6444 up if there is more than one memory reference in the loop. Thus a loop
6445 where the loop counter is decremented can generate slightly more
6446 efficient code, in cases where the RPTB instruction cannot be utilised.
6448 @item -mdp-isr-reload
6450 Force the DP register to be saved on entry to an interrupt service
6451 routine (ISR), reloaded to point to the data section, and restored on
6452 exit from the ISR. This should not be required unless someone has
6453 violated the small memory model by modifying the DP register, say within
6458 For the C3x use the 24-bit MPYI instruction for integer multiplies
6459 instead of a library call to guarantee 32-bit results. Note that if one
6460 of the operands is a constant, then the multiplication will be performed
6461 using shifts and adds. If the -mmpyi option is not specified for the C3x,
6462 then squaring operations are performed inline instead of a library call.
6465 @itemx -mno-fast-fix
6466 The C3x/C4x FIX instruction to convert a floating point value to an
6467 integer value chooses the nearest integer less than or equal to the
6468 floating point value rather than to the nearest integer. Thus if the
6469 floating point number is negative, the result will be incorrectly
6470 truncated an additional code is necessary to detect and correct this
6471 case. This option can be used to disable generation of the additional
6472 code required to correct the result.
6476 Enable (disable) generation of repeat block sequences using the RPTB
6477 instruction for zero overhead looping. The RPTB construct is only used
6478 for innermost loops that do not call functions or jump across the loop
6479 boundaries. There is no advantage having nested RPTB loops due to the
6480 overhead required to save and restore the RC, RS, and RE registers.
6481 This is enabled by default with -O2.
6483 @item -mrpts=@var{count}
6485 Enable (disable) the use of the single instruction repeat instruction
6486 RPTS. If a repeat block contains a single instruction, and the loop
6487 count can be guaranteed to be less than the value @var{count}, GCC will
6488 emit a RPTS instruction instead of a RPTB. If no value is specified,
6489 then a RPTS will be emitted even if the loop count cannot be determined
6490 at compile time. Note that the repeated instruction following RPTS does
6491 not have to be reloaded from memory each iteration, thus freeing up the
6492 CPU buses for oeprands. However, since interrupts are blocked by this
6493 instruction, it is disabled by default.
6495 @item -mloop-unsigned
6496 @itemx -mno-loop-unsigned
6497 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
6498 is 2^31 + 1 since these instructions test if the iteration count is
6499 negative to terminate the loop. If the iteration count is unsigned
6500 there is a possibility than the 2^31 + 1 maximum iteration count may be
6501 exceeded. This switch allows an unsigned iteration count.
6504 Try to emit an assembler syntax that the TI assembler (asm30) is happy
6505 with. This also enforces compatibility with the API employed by the TI
6506 C3x C compiler. For example, long doubles are passed as structures
6507 rather than in floating point registers.
6511 Generate code that uses registers (stack) for passing arguments to functions.
6512 By default, arguments are passed in registers where possible rather
6513 than by pushing arguments on to the stack.
6515 @item -mparallel-insns
6516 @itemx -mno-parallel-insns
6517 Allow the generation of parallel instructions. This is enabled by
6520 @item -mparallel-mpy
6521 @itemx -mno-parallel-mpy
6522 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
6523 provided -mparallel-insns is also specified. These instructions have
6524 tight register constraints which can pessimize the code generation
6530 @subsection V850 Options
6531 @cindex V850 Options
6533 These @samp{-m} options are defined for V850 implementations:
6537 @itemx -mno-long-calls
6538 Treat all calls as being far away (near). If calls are assumed to be
6539 far away, the compiler will always load the functions address up into a
6540 register, and call indirect through the pointer.
6544 Do not optimize (do optimize) basic blocks that use the same index
6545 pointer 4 or more times to copy pointer into the @code{ep} register, and
6546 use the shorter @code{sld} and @code{sst} instructions. The @samp{-mep}
6547 option is on by default if you optimize.
6549 @item -mno-prolog-function
6550 @itemx -mprolog-function
6551 Do not use (do use) external functions to save and restore registers at
6552 the prolog and epilog of a function. The external functions are slower,
6553 but use less code space if more than one function saves the same number
6554 of registers. The @samp{-mprolog-function} option is on by default if
6558 Try to make the code as small as possible. At present, this just turns
6559 on the @samp{-mep} and @samp{-mprolog-function} options.
6562 Put static or global variables whose size is @var{n} bytes or less into
6563 the tiny data area that register @code{ep} points to. The tiny data
6564 area can hold up to 256 bytes in total (128 bytes for byte references).
6567 Put static or global variables whose size is @var{n} bytes or less into
6568 the small data area that register @code{gp} points to. The small data
6569 area can hold up to 64 kilobytes.
6572 Put static or global variables whose size is @var{n} bytes or less into
6573 the first 32 kilobytes of memory.
6576 Specify that the target processor is the V850.
6579 Generate code suitable for big switch tables. Use this option only if
6580 the assembler/linker complain about out of range branches within a switch
6585 @subsection ARC Options
6588 These options are defined for ARC implementations:
6592 Compile code for little endian mode. This is the default.
6595 Compile code for big endian mode.
6598 Prepend the name of the cpu to all public symbol names.
6599 In multiple-processor systems, there are many ARC variants with different
6600 instruction and register set characteristics. This flag prevents code
6601 compiled for one cpu to be linked with code compiled for another.
6602 No facility exists for handling variants that are "almost identical".
6603 This is an all or nothing option.
6605 @item -mcpu=@var{cpu}
6606 Compile code for ARC variant @var{cpu}.
6607 Which variants are supported depend on the configuration.
6608 All variants support @samp{-mcpu=base}, this is the default.
6610 @item -mtext=@var{text section}
6611 @itemx -mdata=@var{data section}
6612 @itemx -mrodata=@var{readonly data section}
6613 Put functions, data, and readonly data in @var{text section},
6614 @var{data section}, and @var{readonly data section} respectively
6615 by default. This can be overridden with the @code{section} attribute.
6616 @xref{Variable Attributes}.
6621 @subsection NS32K Options
6622 @cindex NS32K options
6624 These are the @samp{-m} options defined for the 32000 series. The default
6625 values for these options depends on which style of 32000 was selected when
6626 the compiler was configured; the defaults for the most common choices are
6632 Generate output for a 32032. This is the default
6633 when the compiler is configured for 32032 and 32016 based systems.
6637 Generate output for a 32332. This is the default
6638 when the compiler is configured for 32332-based systems.
6642 Generate output for a 32532. This is the default
6643 when the compiler is configured for 32532-based systems.
6646 Generate output containing 32081 instructions for floating point.
6647 This is the default for all systems.
6650 Generate output containing 32381 instructions for floating point. This
6651 also implies @samp{-m32081}. The 32381 is only compatible with the 32332
6652 and 32532 cpus. This is the default for the pc532-netbsd configuration.
6655 Try and generate multiply-add floating point instructions @code{polyF}
6656 and @code{dotF}. This option is only available if the @samp{-m32381}
6657 option is in effect. Using these instructions requires changes to to
6658 register allocation which generally has a negative impact on
6659 performance. This option should only be enabled when compiling code
6660 particularly likely to make heavy use of multiply-add instructions.
6663 Do not try and generate multiply-add floating point instructions
6664 @code{polyF} and @code{dotF}. This is the default on all platforms.
6667 Generate output containing library calls for floating point.
6668 @strong{Warning:} the requisite libraries may not be available.
6671 Do not use the bit-field instructions. On some machines it is faster to
6672 use shifting and masking operations. This is the default for the pc532.
6675 Do use the bit-field instructions. This is the default for all platforms
6679 Use a different function-calling convention, in which functions
6680 that take a fixed number of arguments return pop their
6681 arguments on return with the @code{ret} instruction.
6683 This calling convention is incompatible with the one normally
6684 used on Unix, so you cannot use it if you need to call libraries
6685 compiled with the Unix compiler.
6687 Also, you must provide function prototypes for all functions that
6688 take variable numbers of arguments (including @code{printf});
6689 otherwise incorrect code will be generated for calls to those
6692 In addition, seriously incorrect code will result if you call a
6693 function with too many arguments. (Normally, extra arguments are
6694 harmlessly ignored.)
6696 This option takes its name from the 680x0 @code{rtd} instruction.
6700 Use a different function-calling convention where the first two arguments
6701 are passed in registers.
6703 This calling convention is incompatible with the one normally
6704 used on Unix, so you cannot use it if you need to call libraries
6705 compiled with the Unix compiler.
6708 Do not pass any arguments in registers. This is the default for all
6712 It is OK to use the sb as an index register which is always loaded with
6713 zero. This is the default for the pc532-netbsd target.
6716 The sb register is not available for use or has not been initialized to
6717 zero by the run time system. This is the default for all targets except
6718 the pc532-netbsd. It is also implied whenever @samp{-mhimem} or
6719 @samp{-fpic} is set.
6722 Many ns32000 series addressing modes use displacements of up to 512MB.
6723 If an address is above 512MB then displacements from zero can not be used.
6724 This option causes code to be generated which can be loaded above 512MB.
6725 This may be useful for operating systems or ROM code.
6728 Assume code will be loaded in the first 512MB of virtual address space.
6729 This is the default for all platforms.
6736 @node Code Gen Options
6737 @section Options for Code Generation Conventions
6738 @cindex code generation conventions
6739 @cindex options, code generation
6740 @cindex run-time options
6742 These machine-independent options control the interface conventions
6743 used in code generation.
6745 Most of them have both positive and negative forms; the negative form
6746 of @samp{-ffoo} would be @samp{-fno-foo}. In the table below, only
6747 one of the forms is listed---the one which is not the default. You
6748 can figure out the other form by either removing @samp{no-} or adding
6753 Enable exception handling. Generates extra code needed to propagate
6754 exceptions. For some targets, this implies GNU CC will generate frame
6755 unwind information for all functions, which can produce significant data
6756 size overhead, although it does not affect execution. If you do not
6757 specify this option, GNU CC will enable it by default for languages like
6758 C++ which normally require exception handling, and disable itfor
6759 languages like C that do not normally require it. However, you may need
6760 to enable this option when compiling C code that needs to interoperate
6761 properly with exception handlers written in C++. You may also wish to
6762 disable this option if you are compiling older C++ programs that don't
6763 use exception handling.
6765 @item -funwind-tables
6766 Similar to @code{-fexceptions}, except that it will just generate any needed
6767 static data, but will not affect the generated code in any other way.
6768 You will normally not enable this option; instead, a language processor
6769 that needs this handling would enable it on your behalf.
6771 @item -fpcc-struct-return
6772 Return ``short'' @code{struct} and @code{union} values in memory like
6773 longer ones, rather than in registers. This convention is less
6774 efficient, but it has the advantage of allowing intercallability between
6775 GCC-compiled files and files compiled with other compilers.
6777 The precise convention for returning structures in memory depends
6778 on the target configuration macros.
6780 Short structures and unions are those whose size and alignment match
6781 that of some integer type.
6783 @item -freg-struct-return
6784 Use the convention that @code{struct} and @code{union} values are
6785 returned in registers when possible. This is more efficient for small
6786 structures than @samp{-fpcc-struct-return}.
6788 If you specify neither @samp{-fpcc-struct-return} nor its contrary
6789 @samp{-freg-struct-return}, GCC defaults to whichever convention is
6790 standard for the target. If there is no standard convention, GCC
6791 defaults to @samp{-fpcc-struct-return}, except on targets where GCC
6792 is the principal compiler. In those cases, we can choose the standard,
6793 and we chose the more efficient register return alternative.
6796 Allocate to an @code{enum} type only as many bytes as it needs for the
6797 declared range of possible values. Specifically, the @code{enum} type
6798 will be equivalent to the smallest integer type which has enough room.
6800 @item -fshort-double
6801 Use the same size for @code{double} as for @code{float}.
6804 Requests that the data and non-@code{const} variables of this
6805 compilation be shared data rather than private data. The distinction
6806 makes sense only on certain operating systems, where shared data is
6807 shared between processes running the same program, while private data
6808 exists in one copy per process.
6811 Allocate even uninitialized global variables in the data section of the
6812 object file, rather than generating them as common blocks. This has the
6813 effect that if the same variable is declared (without @code{extern}) in
6814 two different compilations, you will get an error when you link them.
6815 The only reason this might be useful is if you wish to verify that the
6816 program will work on other systems which always work this way.
6819 Ignore the @samp{#ident} directive.
6821 @item -fno-gnu-linker
6822 Do not output global initializations (such as C++ constructors and
6823 destructors) in the form used by the GNU linker (on systems where the GNU
6824 linker is the standard method of handling them). Use this option when
6825 you want to use a non-GNU linker, which also requires using the
6826 @code{collect2} program to make sure the system linker includes
6827 constructors and destructors. (@code{collect2} is included in the GCC
6828 distribution.) For systems which @emph{must} use @code{collect2}, the
6829 compiler driver @code{gcc} is configured to do this automatically.
6831 @item -finhibit-size-directive
6832 Don't output a @code{.size} assembler directive, or anything else that
6833 would cause trouble if the function is split in the middle, and the
6834 two halves are placed at locations far apart in memory. This option is
6835 used when compiling @file{crtstuff.c}; you should not need to use it
6839 Put extra commentary information in the generated assembly code to
6840 make it more readable. This option is generally only of use to those
6841 who actually need to read the generated assembly code (perhaps while
6842 debugging the compiler itself).
6844 @samp{-fno-verbose-asm}, the default, causes the
6845 extra information to be omitted and is useful when comparing two assembler
6849 Consider all memory references through pointers to be volatile.
6851 @item -fvolatile-global
6852 Consider all memory references to extern and global data items to
6853 be volatile. GCC does not consider static data items to be volatile
6854 because of this switch.
6856 @item -fvolatile-static
6857 Consider all memory references to static data to be volatile.
6860 @cindex global offset table
6862 Generate position-independent code (PIC) suitable for use in a shared
6863 library, if supported for the target machine. Such code accesses all
6864 constant addresses through a global offset table (GOT). The dynamic
6865 loader resolves the GOT entries when the program starts (the dynamic
6866 loader is not part of GCC; it is part of the operating system). If
6867 the GOT size for the linked executable exceeds a machine-specific
6868 maximum size, you get an error message from the linker indicating that
6869 @samp{-fpic} does not work; in that case, recompile with @samp{-fPIC}
6870 instead. (These maximums are 16k on the m88k, 8k on the Sparc, and 32k
6871 on the m68k and RS/6000. The 386 has no such limit.)
6873 Position-independent code requires special support, and therefore works
6874 only on certain machines. For the 386, GCC supports PIC for System V
6875 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
6876 position-independent.
6879 If supported for the target machine, emit position-independent code,
6880 suitable for dynamic linking and avoiding any limit on the size of the
6881 global offset table. This option makes a difference on the m68k, m88k,
6884 Position-independent code requires special support, and therefore works
6885 only on certain machines.
6887 @item -ffixed-@var{reg}
6888 Treat the register named @var{reg} as a fixed register; generated code
6889 should never refer to it (except perhaps as a stack pointer, frame
6890 pointer or in some other fixed role).
6892 @var{reg} must be the name of a register. The register names accepted
6893 are machine-specific and are defined in the @code{REGISTER_NAMES}
6894 macro in the machine description macro file.
6896 This flag does not have a negative form, because it specifies a
6899 @item -fcall-used-@var{reg}
6900 Treat the register named @var{reg} as an allocable register that is
6901 clobbered by function calls. It may be allocated for temporaries or
6902 variables that do not live across a call. Functions compiled this way
6903 will not save and restore the register @var{reg}.
6905 It is an error to used this flag with the frame pointer or stack pointer.
6906 Use of this flag for other registers that have fixed pervasive roles in
6907 the machine's execution model will produce disastrous results.
6909 This flag does not have a negative form, because it specifies a
6912 @item -fcall-saved-@var{reg}
6913 Treat the register named @var{reg} as an allocable register saved by
6914 functions. It may be allocated even for temporaries or variables that
6915 live across a call. Functions compiled this way will save and restore
6916 the register @var{reg} if they use it.
6918 It is an error to used this flag with the frame pointer or stack pointer.
6919 Use of this flag for other registers that have fixed pervasive roles in
6920 the machine's execution model will produce disastrous results.
6922 A different sort of disaster will result from the use of this flag for
6923 a register in which function values may be returned.
6925 This flag does not have a negative form, because it specifies a
6929 Pack all structure members together without holes. Usually you would
6930 not want to use this option, since it makes the code suboptimal, and
6931 the offsets of structure members won't agree with system libraries.
6933 @item -fcheck-memory-usage
6934 Generate extra code to check each memory access. GCC will generate
6935 code that is suitable for a detector of bad memory accesses such as
6938 Normally, you should compile all, or none, of your code with this option.
6940 If you do mix code compiled with and without this option,
6941 you must ensure that all code that has side effects
6942 and that is called by code compiled with this option
6943 is, itself, compiled with this option.
6944 If you do not, you might get erroneous messages from the detector.
6946 If you use functions from a library that have side-effects (such as
6947 @code{read}), you might not be able to recompile the library and
6948 specify this option. In that case, you can enable the
6949 @samp{-fprefix-function-name} option, which requests GCC to encapsulate
6950 your code and make other functions look as if they were compiled with
6951 @samp{-fcheck-memory-usage}. This is done by calling ``stubs'',
6952 which are provided by the detector. If you cannot find or build
6953 stubs for every function you call, you might have to specify
6954 @samp{-fcheck-memory-usage} without @samp{-fprefix-function-name}.
6956 If you specify this option, you can not use the @code{asm} or
6957 @code{__asm__} keywords in functions with memory checking enabled. GNU
6958 CC cannot understand what the @code{asm} statement may do, and therefore
6959 cannot generate the appropriate code, so it will reject it. However, if
6960 you specify the function attribute @code{no_check_memory_usage} (see
6961 @pxref{Function Attributes}, GNU CC will disable memory checking within a
6962 function; you may use @code{asm} statements inside such functions. You
6963 may have an inline expansion of a non-checked function within a checked
6964 function; in that case GNU CC will not generate checks for the inlined
6965 function's memory accesses.
6967 If you move your @code{asm} statements to non-checked inline functions
6968 and they do access memory, you can add calls to the support code in your
6969 inline function, to indicate any reads, writes, or copies being done.
6970 These calls would be similar to those done in the stubs described above.
6972 @item -fprefix-function-name
6973 Request GCC to add a prefix to the symbols generated for function names.
6974 GCC adds a prefix to the names of functions defined as well as
6975 functions called. Code compiled with this option and code compiled
6976 without the option can't be linked together, unless stubs are used.
6978 If you compile the following code with @samp{-fprefix-function-name}
6980 extern void bar (int);
6989 GCC will compile the code as if it was written:
6991 extern void prefix_bar (int);
6995 return prefix_bar (a + 5);
6998 This option is designed to be used with @samp{-fcheck-memory-usage}.
7000 @item -finstrument-functions
7001 Generate instrumentation calls for entry and exit to functions. Just
7002 after function entry and just before function exit, the following
7003 profiling functions will be called with the address of the current
7004 function and its call site. (On some platforms,
7005 @code{__builtin_return_address} does not work beyond the current
7006 function, so the call site information may not be available to the
7007 profiling functions otherwise.)
7010 void __cyg_profile_func_enter (void *this_fn, void *call_site);
7011 void __cyg_profile_func_exit (void *this_fn, void *call_site);
7014 The first argument is the address of the start of the current function,
7015 which may be looked up exactly in the symbol table.
7017 This instrumentation is also done for functions expanded inline in other
7018 functions. The profiling calls will indicate where, conceptually, the
7019 inline function is entered and exited. This means that addressable
7020 versions of such functions must be available. If all your uses of a
7021 function are expanded inline, this may mean an additional expansion of
7022 code size. If you use @samp{extern inline} in your C code, an
7023 addressable version of such functions must be provided. (This is
7024 normally the case anyways, but if you get lucky and the optimizer always
7025 expands the functions inline, you might have gotten away without
7026 providing static copies.)
7028 A function may be given the attribute @code{no_instrument_function}, in
7029 which case this instrumentation will not be done. This can be used, for
7030 example, for the profiling functions listed above, high-priority
7031 interrupt routines, and any functions from which the profiling functions
7032 cannot safely be called (perhaps signal handlers, if the profiling
7033 routines generate output or allocate memory).
7036 Generate code to verify that you do not go beyond the boundary of the
7037 stack. You should specify this flag if you are running in an
7038 environment with multiple threads, but only rarely need to specify it in
7039 a single-threaded environment since stack overflow is automatically
7040 detected on nearly all systems if there is only one stack.
7042 @cindex aliasing of parameters
7043 @cindex parameters, aliased
7044 @item -fargument-alias
7045 @itemx -fargument-noalias
7046 @itemx -fargument-noalias-global
7047 Specify the possible relationships among parameters and between
7048 parameters and global data.
7050 @samp{-fargument-alias} specifies that arguments (parameters) may
7051 alias each other and may alias global storage.
7052 @samp{-fargument-noalias} specifies that arguments do not alias
7053 each other, but may alias global storage.
7054 @samp{-fargument-noalias-global} specifies that arguments do not
7055 alias each other and do not alias global storage.
7057 Each language will automatically use whatever option is required by
7058 the language standard. You should not need to use these options yourself.
7060 @item -fleading-underscore
7061 This option and its counterpart, -fno-leading-underscore, forcibly
7062 change the way C symbols are represented in the object file. One use
7063 is to help link with legacy assembly code.
7065 Be warned that you should know what you are doing when invoking this
7066 option, and that not all targets provide complete support for it.
7069 @node Environment Variables
7070 @section Environment Variables Affecting GCC
7071 @cindex environment variables
7073 This section describes several environment variables that affect how GCC
7074 operates. Some of them work by specifying directories or prefixes to use
7075 when searching for various kinds of files. Some are used to specify other
7076 aspects of the compilation environment.
7079 Note that you can also specify places to search using options such as
7080 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
7081 take precedence over places specified using environment variables, which
7082 in turn take precedence over those specified by the configuration of GCC.
7086 Note that you can also specify places to search using options such as
7087 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
7088 take precedence over places specified using environment variables, which
7089 in turn take precedence over those specified by the configuration of GCC.
7096 @c @itemx LC_COLLATE
7098 @c @itemx LC_MONETARY
7099 @c @itemx LC_NUMERIC
7104 @c @findex LC_COLLATE
7106 @c @findex LC_MONETARY
7107 @c @findex LC_NUMERIC
7111 These environment variables control the way that GCC uses
7112 localization information that allow GCC to work with different
7113 national conventions. GCC inspects the locale categories
7114 @code{LC_CTYPE} and @code{LC_MESSAGES} if it has been configured to do
7115 so. These locale categories can be set to any value supported by your
7116 installation. A typical value is @samp{en_UK} for English in the United
7119 The @code{LC_CTYPE} environment variable specifies character
7120 classification. GCC uses it to determine the character boundaries in
7121 a string; this is needed for some multibyte encodings that contain quote
7122 and escape characters that would otherwise be interpreted as a string
7125 The @code{LC_MESSAGES} environment variable specifies the language to
7126 use in diagnostic messages.
7128 If the @code{LC_ALL} environment variable is set, it overrides the value
7129 of @code{LC_CTYPE} and @code{LC_MESSAGES}; otherwise, @code{LC_CTYPE}
7130 and @code{LC_MESSAGES} default to the value of the @code{LANG}
7131 environment variable. If none of these variables are set, GCC
7132 defaults to traditional C English behavior.
7136 If @code{TMPDIR} is set, it specifies the directory to use for temporary
7137 files. GCC uses temporary files to hold the output of one stage of
7138 compilation which is to be used as input to the next stage: for example,
7139 the output of the preprocessor, which is the input to the compiler
7142 @item GCC_EXEC_PREFIX
7143 @findex GCC_EXEC_PREFIX
7144 If @code{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
7145 names of the subprograms executed by the compiler. No slash is added
7146 when this prefix is combined with the name of a subprogram, but you can
7147 specify a prefix that ends with a slash if you wish.
7149 If @code{GCC_EXEC_PREFIX} is not set, GNU CC will attempt to figure out
7150 an appropriate prefix to use based on the pathname it was invoked with.
7152 If GCC cannot find the subprogram using the specified prefix, it
7153 tries looking in the usual places for the subprogram.
7155 The default value of @code{GCC_EXEC_PREFIX} is
7156 @file{@var{prefix}/lib/gcc-lib/} where @var{prefix} is the value
7157 of @code{prefix} when you ran the @file{configure} script.
7159 Other prefixes specified with @samp{-B} take precedence over this prefix.
7161 This prefix is also used for finding files such as @file{crt0.o} that are
7164 In addition, the prefix is used in an unusual way in finding the
7165 directories to search for header files. For each of the standard
7166 directories whose name normally begins with @samp{/usr/local/lib/gcc-lib}
7167 (more precisely, with the value of @code{GCC_INCLUDE_DIR}), GCC tries
7168 replacing that beginning with the specified prefix to produce an
7169 alternate directory name. Thus, with @samp{-Bfoo/}, GCC will search
7170 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
7171 These alternate directories are searched first; the standard directories
7175 @findex COMPILER_PATH
7176 The value of @code{COMPILER_PATH} is a colon-separated list of
7177 directories, much like @code{PATH}. GCC tries the directories thus
7178 specified when searching for subprograms, if it can't find the
7179 subprograms using @code{GCC_EXEC_PREFIX}.
7182 @findex LIBRARY_PATH
7183 The value of @code{LIBRARY_PATH} is a colon-separated list of
7184 directories, much like @code{PATH}. When configured as a native compiler,
7185 GCC tries the directories thus specified when searching for special
7186 linker files, if it can't find them using @code{GCC_EXEC_PREFIX}. Linking
7187 using GCC also uses these directories when searching for ordinary
7188 libraries for the @samp{-l} option (but directories specified with
7189 @samp{-L} come first).
7191 @item C_INCLUDE_PATH
7192 @itemx CPLUS_INCLUDE_PATH
7193 @itemx OBJC_INCLUDE_PATH
7194 @findex C_INCLUDE_PATH
7195 @findex CPLUS_INCLUDE_PATH
7196 @findex OBJC_INCLUDE_PATH
7197 @c @itemx OBJCPLUS_INCLUDE_PATH
7198 These environment variables pertain to particular languages. Each
7199 variable's value is a colon-separated list of directories, much like
7200 @code{PATH}. When GCC searches for header files, it tries the
7201 directories listed in the variable for the language you are using, after
7202 the directories specified with @samp{-I} but before the standard header
7205 @item DEPENDENCIES_OUTPUT
7206 @findex DEPENDENCIES_OUTPUT
7207 @cindex dependencies for make as output
7208 If this variable is set, its value specifies how to output dependencies
7209 for Make based on the header files processed by the compiler. This
7210 output looks much like the output from the @samp{-M} option
7211 (@pxref{Preprocessor Options}), but it goes to a separate file, and is
7212 in addition to the usual results of compilation.
7214 The value of @code{DEPENDENCIES_OUTPUT} can be just a file name, in
7215 which case the Make rules are written to that file, guessing the target
7216 name from the source file name. Or the value can have the form
7217 @samp{@var{file} @var{target}}, in which case the rules are written to
7218 file @var{file} using @var{target} as the target name.
7222 @cindex locale definition
7223 This variable is used to pass locale information to the compiler. One way in
7224 which this information is used is to determine the character set to be used
7225 when character literals, string literals and comments are parsed in C and C++.
7226 When the compiler is configured to allow multibyte characters,
7227 the following values for @code{LANG} are recognized:
7231 Recognize JIS characters.
7233 Recognize SJIS characters.
7235 Recognize EUCJP characters.
7238 If @code{LANG} is not defined, or if it has some other value, then the
7239 compiler will use mblen and mbtowc as defined by the default locale to
7240 recognize and translate multibyte characters.
7243 @node Running Protoize
7244 @section Running Protoize
7246 The program @code{protoize} is an optional part of GNU C. You can use
7247 it to add prototypes to a program, thus converting the program to ANSI
7248 C in one respect. The companion program @code{unprotoize} does the
7249 reverse: it removes argument types from any prototypes that are found.
7251 When you run these programs, you must specify a set of source files as
7252 command line arguments. The conversion programs start out by compiling
7253 these files to see what functions they define. The information gathered
7254 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
7256 After scanning comes actual conversion. The specified files are all
7257 eligible to be converted; any files they include (whether sources or
7258 just headers) are eligible as well.
7260 But not all the eligible files are converted. By default,
7261 @code{protoize} and @code{unprotoize} convert only source and header
7262 files in the current directory. You can specify additional directories
7263 whose files should be converted with the @samp{-d @var{directory}}
7264 option. You can also specify particular files to exclude with the
7265 @samp{-x @var{file}} option. A file is converted if it is eligible, its
7266 directory name matches one of the specified directory names, and its
7267 name within the directory has not been excluded.
7269 Basic conversion with @code{protoize} consists of rewriting most
7270 function definitions and function declarations to specify the types of
7271 the arguments. The only ones not rewritten are those for varargs
7274 @code{protoize} optionally inserts prototype declarations at the
7275 beginning of the source file, to make them available for any calls that
7276 precede the function's definition. Or it can insert prototype
7277 declarations with block scope in the blocks where undeclared functions
7280 Basic conversion with @code{unprotoize} consists of rewriting most
7281 function declarations to remove any argument types, and rewriting
7282 function definitions to the old-style pre-ANSI form.
7284 Both conversion programs print a warning for any function declaration or
7285 definition that they can't convert. You can suppress these warnings
7288 The output from @code{protoize} or @code{unprotoize} replaces the
7289 original source file. The original file is renamed to a name ending
7290 with @samp{.save}. If the @samp{.save} file already exists, then
7291 the source file is simply discarded.
7293 @code{protoize} and @code{unprotoize} both depend on GCC itself to
7294 scan the program and collect information about the functions it uses.
7295 So neither of these programs will work until GCC is installed.
7297 Here is a table of the options you can use with @code{protoize} and
7298 @code{unprotoize}. Each option works with both programs unless
7302 @item -B @var{directory}
7303 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
7304 usual directory (normally @file{/usr/local/lib}). This file contains
7305 prototype information about standard system functions. This option
7306 applies only to @code{protoize}.
7308 @item -c @var{compilation-options}
7309 Use @var{compilation-options} as the options when running @code{gcc} to
7310 produce the @samp{.X} files. The special option @samp{-aux-info} is
7311 always passed in addition, to tell @code{gcc} to write a @samp{.X} file.
7313 Note that the compilation options must be given as a single argument to
7314 @code{protoize} or @code{unprotoize}. If you want to specify several
7315 @code{gcc} options, you must quote the entire set of compilation options
7316 to make them a single word in the shell.
7318 There are certain @code{gcc} arguments that you cannot use, because they
7319 would produce the wrong kind of output. These include @samp{-g},
7320 @samp{-O}, @samp{-c}, @samp{-S}, and @samp{-o} If you include these in
7321 the @var{compilation-options}, they are ignored.
7324 Rename files to end in @samp{.C} instead of @samp{.c}.
7325 This is convenient if you are converting a C program to C++.
7326 This option applies only to @code{protoize}.
7329 Add explicit global declarations. This means inserting explicit
7330 declarations at the beginning of each source file for each function
7331 that is called in the file and was not declared. These declarations
7332 precede the first function definition that contains a call to an
7333 undeclared function. This option applies only to @code{protoize}.
7335 @item -i @var{string}
7336 Indent old-style parameter declarations with the string @var{string}.
7337 This option applies only to @code{protoize}.
7339 @code{unprotoize} converts prototyped function definitions to old-style
7340 function definitions, where the arguments are declared between the
7341 argument list and the initial @samp{@{}. By default, @code{unprotoize}
7342 uses five spaces as the indentation. If you want to indent with just
7343 one space instead, use @samp{-i " "}.
7346 Keep the @samp{.X} files. Normally, they are deleted after conversion
7350 Add explicit local declarations. @code{protoize} with @samp{-l} inserts
7351 a prototype declaration for each function in each block which calls the
7352 function without any declaration. This option applies only to
7356 Make no real changes. This mode just prints information about the conversions
7357 that would have been done without @samp{-n}.
7360 Make no @samp{.save} files. The original files are simply deleted.
7361 Use this option with caution.
7363 @item -p @var{program}
7364 Use the program @var{program} as the compiler. Normally, the name
7368 Work quietly. Most warnings are suppressed.
7371 Print the version number, just like @samp{-v} for @code{gcc}.
7374 If you need special compiler options to compile one of your program's
7375 source files, then you should generate that file's @samp{.X} file
7376 specially, by running @code{gcc} on that source file with the
7377 appropriate options and the option @samp{-aux-info}. Then run
7378 @code{protoize} on the entire set of files. @code{protoize} will use
7379 the existing @samp{.X} file because it is newer than the source file.
7383 gcc -Dfoo=bar file1.c -aux-info
7388 You need to include the special files along with the rest in the
7389 @code{protoize} command, even though their @samp{.X} files already
7390 exist, because otherwise they won't get converted.
7392 @xref{Protoize Caveats}, for more information on how to use
7393 @code{protoize} successfully.