1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000 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 multi-letter 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 * Language Independent Options:: Controlling how diagnostics should be
64 * Warning Options:: How picky should the compiler be?
65 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
66 * Optimize Options:: How much optimization?
67 * Preprocessor Options:: Controlling header files and macro definitions.
68 Also, getting dependency information for Make.
69 * Assembler Options:: Passing options to the assembler.
70 * Link Options:: Specifying libraries and so on.
71 * Directory Options:: Where to find header files and libraries.
72 Where to find the compiler executable files.
73 * Spec Files:: How to pass switches to sub-processes.
74 * Target Options:: Running a cross-compiler, or an old version of GCC.
75 * Submodel Options:: Specifying minor hardware or convention variations,
76 such as 68010 vs 68020.
77 * Code Gen Options:: Specifying conventions for function calls, data layout
79 * Environment Variables:: Env vars that affect GCC.
80 * Running Protoize:: Automatically adding or removing function prototypes.
84 @section Option Summary
86 Here is a summary of all the options, grouped by type. Explanations are
87 in the following sections.
91 @xref{Overall Options,,Options Controlling the Kind of Output}.
93 -c -S -E -o @var{file} -pipe -pass-exit-codes -v --help -x @var{language}
96 @item C Language Options
97 @xref{C Dialect Options,,Options Controlling C Dialect}.
99 -ansi -std -fallow-single-precision -fcond-mismatch -fno-asm
100 -fno-builtin -ffreestanding -fhosted -fsigned-bitfields -fsigned-char
101 -funsigned-bitfields -funsigned-char -fwritable-strings
102 -traditional -traditional-cpp -trigraphs -fsingle-precision-constant
105 @item C++ Language Options
106 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
108 -fno-access-control -fcheck-new -fconserve-space
109 -fdollars-in-identifiers -fno-elide-constructors -fexternal-templates
110 -ffor-scope -fno-for-scope -fno-gnu-keywords -fhonor-std
111 -fhuge-objects -fno-implicit-templates -finit-priority
112 -fno-implement-inlines -fname-mangling-version-@var{n}
113 -fno-default-inline -fno-operator-names -fno-optional-diags -fpermissive
114 -frepo -fsquangle -ftemplate-depth-@var{n}
115 -fuse-cxa-atexit -fvtable-thunks -nostdinc++ -Wctor-dtor-privacy
116 -Wno-deprecated -Weffc++ -Wno-non-template-friend -Wnon-virtual-dtor
117 -Wold-style-cast -Woverloaded-virtual -Wno-pmf-conversions -Wreorder
121 @item Language Independent Options
122 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
124 -fmessage-length=@var{n}
125 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}
128 @item Warning Options
129 @xref{Warning Options,,Options to Request or Suppress Warnings}.
131 -fsyntax-only -pedantic -pedantic-errors
132 -w -W -Wall -Waggregate-return
133 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment
134 -Wconversion -Wdisabled-optimization -Werror -Wformat
135 -Wid-clash-@var{len} -Wimplicit -Wimplicit-int
136 -Wimplicit-function-declaration -Wimport
137 -Werror-implicit-function-declaration -Wfloat-equal -Winline
138 -Wlarger-than-@var{len} -Wlong-long
139 -Wmain -Wmissing-declarations -Wmissing-noreturn
140 -Wmultichar -Wno-import -Wpacked -Wpadded
141 -Wparentheses -Wpointer-arith -Wredundant-decls
142 -Wreturn-type -Wshadow -Wsign-compare -Wswitch
143 -Wtrigraphs -Wundef -Wuninitialized -Wunknown-pragmas -Wunreachable-code
144 -Wunused -Wunused-function -Wunused-label -Wunused-parameter
145 -Wunused-variable -Wunused-value -Wwrite-strings
148 @item C-only Warning Options
150 -Wbad-function-cast -Wmissing-prototypes -Wnested-externs
151 -Wstrict-prototypes -Wtraditional
154 @item Debugging Options
155 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
157 -a -ax -d@var{letters} -fdump-unnumbered -fdump-translation-unit-@var{file}
158 -fpretend-float -fprofile-arcs -ftest-coverage
159 -g -g@var{level} -gcoff -gdwarf -gdwarf-1 -gdwarf-1+ -gdwarf-2
160 -ggdb -gstabs -gstabs+ -gxcoff -gxcoff+
161 -p -pg -print-file-name=@var{library} -print-libgcc-file-name
162 -print-prog-name=@var{program} -print-search-dirs -save-temps -time
165 @item Optimization Options
166 @xref{Optimize Options,,Options that Control Optimization}.
168 -falign-functions=@var{n} -falign-labels=@var{n} -falign-loops=@var{n}
169 -falign-jumps=@var{n} -fbranch-probabilities
170 -fcaller-saves -fcse-follow-jumps -fcse-skip-blocks
171 -fdce -fdelayed-branch -fdelete-null-pointer-checks -fexpensive-optimizations
172 -ffast-math -ffloat-store -fforce-addr -fforce-mem -fno-math-errno
173 -fdata-sections -ffunction-sections -fgcse
174 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions
175 -fmove-all-movables -fno-default-inline -fno-defer-pop
176 -fno-function-cse -fno-inline -fno-peephole
177 -fomit-frame-pointer -foptimize-register-moves -foptimize-sibling-calls
178 -fregmove -frerun-cse-after-loop -frerun-loop-opt -freduce-all-givs
179 -fschedule-insns -fschedule-insns2 -fssa -fstrength-reduce
180 -fstrict-aliasing -fthread-jumps -funroll-all-loops
182 -O -O0 -O1 -O2 -O3 -Os
185 @item Preprocessor Options
186 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
188 -A@var{question}(@var{answer}) -C -dD -dM -dN
189 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H
191 -include @var{file} -imacros @var{file}
192 -iprefix @var{file} -iwithprefix @var{dir}
193 -iwithprefixbefore @var{dir} -isystem @var{dir} -isystem-c++ @var{dir}
194 -M -MD -MM -MMD -MG -nostdinc -P -trigraphs
195 -undef -U@var{macro} -Wp,@var{option}
198 @item Assembler Option
199 @xref{Assembler Options,,Passing Options to the Assembler}.
205 @xref{Link Options,,Options for Linking}.
207 @var{object-file-name} -l@var{library}
208 -nostartfiles -nodefaultlibs -nostdlib
209 -s -static -shared -symbolic
210 -Wl,@var{option} -Xlinker @var{option}
214 @item Directory Options
215 @xref{Directory Options,,Options for Directory Search}.
217 -B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}
221 @c I wrote this xref this way to avoid overfull hbox. -- rms
222 @xref{Target Options}.
224 -b @var{machine} -V @var{version}
227 @item Machine Dependent Options
228 @xref{Submodel Options,,Hardware Models and Configurations}.
230 @emph{M680x0 Options}
231 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040
232 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020
233 -mfpa -mnobitfield -mrtd -mshort -msoft-float -mpcrel
234 -malign-int -mstrict-align
236 @emph{M68hc1x Options}
237 -m6811 -m6812 -m68hc11 -m68hc12
238 -mauto-incdec -mshort -msoft-reg-count=@var{count}
245 -mtune=@var{cpu type}
246 -mcmodel=@var{code model}
248 -mapp-regs -mbroken-saverestore -mcypress
249 -mepilogue -mfaster-structs -mflat
250 -mfpu -mhard-float -mhard-quad-float
251 -mimpure-text -mlive-g0 -mno-app-regs
252 -mno-epilogue -mno-faster-structs -mno-flat -mno-fpu
253 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles
254 -msoft-float -msoft-quad-float -msparclite -mstack-bias
255 -msupersparc -munaligned-doubles -mv8
257 @emph{Convex Options}
258 -mc1 -mc2 -mc32 -mc34 -mc38
259 -margcount -mnoargcount
261 -mvolatile-cache -mvolatile-nocache
263 @emph{AMD29K Options}
264 -m29000 -m29050 -mbw -mnbw -mdw -mndw
265 -mlarge -mnormal -msmall
266 -mkernel-registers -mno-reuse-arg-regs
267 -mno-stack-check -mno-storem-bug
268 -mreuse-arg-regs -msoft-float -mstack-check
269 -mstorem-bug -muser-registers
272 -mapcs-frame -mno-apcs-frame
274 -mapcs-stack-check -mno-apcs-stack-check
275 -mapcs-float -mno-apcs-float
276 -mapcs-reentrant -mno-apcs-reentrant
277 -msched-prolog -mno-sched-prolog
278 -mlittle-endian -mbig-endian -mwords-little-endian
279 -malignment-traps -mno-alignment-traps
280 -msoft-float -mhard-float -mfpe
281 -mthumb-interwork -mno-thumb-interwork
282 -mcpu= -march= -mfpe=
283 -mstructure-size-boundary=
284 -mbsd -mxopen -mno-symrename
286 -mlong-calls -mno-long-calls
287 -mnop-fun-dllimport -mno-nop-fun-dllimport
288 -msingle-pic-base -mno-single-pic-base
292 -mtpcs-frame -mno-tpcs-frame
293 -mtpcs-leaf-frame -mno-tpcs-leaf-frame
294 -mlittle-endian -mbig-endian
295 -mthumb-interwork -mno-thumb-interwork
296 -mstructure-size-boundary=
297 -mnop-fun-dllimport -mno-nop-fun-dllimport
298 -mcallee-super-interworking -mno-callee-super-interworking
299 -mcaller-super-interworking -mno-caller-super-interworking
300 -msingle-pic-base -mno-single-pic-base
303 @emph{MN10200 Options}
306 @emph{MN10300 Options}
313 @emph{M32R/D Options}
314 -mcode-model=@var{model type} -msdata=@var{sdata type}
318 -m88000 -m88100 -m88110 -mbig-pic
319 -mcheck-zero-division -mhandle-large-shift
320 -midentify-revision -mno-check-zero-division
321 -mno-ocs-debug-info -mno-ocs-frame-position
322 -mno-optimize-arg-area -mno-serialize-volatile
323 -mno-underscores -mocs-debug-info
324 -mocs-frame-position -moptimize-arg-area
325 -mserialize-volatile -mshort-data-@var{num} -msvr3
326 -msvr4 -mtrap-large-shift -muse-div-instruction
327 -mversion-03.00 -mwarn-passed-structs
329 @emph{RS/6000 and PowerPC Options}
331 -mtune=@var{cpu type}
332 -mpower -mno-power -mpower2 -mno-power2
333 -mpowerpc -mpowerpc64 -mno-powerpc
334 -mpowerpc-gpopt -mno-powerpc-gpopt
335 -mpowerpc-gfxopt -mno-powerpc-gfxopt
336 -mnew-mnemonics -mold-mnemonics
337 -mfull-toc -mminimal-toc -mno-fop-in-toc -mno-sum-in-toc
338 -m64 -m32 -mxl-call -mno-xl-call -mthreads -mpe
339 -msoft-float -mhard-float -mmultiple -mno-multiple
340 -mstring -mno-string -mupdate -mno-update
341 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align
342 -mstrict-align -mno-strict-align -mrelocatable
343 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib
344 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian
345 -mcall-aix -mcall-sysv -mprototype -mno-prototype
346 -msim -mmvme -mads -myellowknife -memb -msdata
347 -msdata=@var{opt} -mvxworks -G @var{num}
350 -mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs
351 -mfull-fp-blocks -mhc-struct-return -min-line-mul
352 -mminimum-fp-blocks -mnohc-struct-return
355 -mabicalls -mcpu=@var{cpu type} -membedded-data -muninit-const-in-rodata
356 -membedded-pic -mfp32 -mfp64 -mgas -mgp32 -mgp64
357 -mgpopt -mhalf-pic -mhard-float -mint64 -mips1
358 -mips2 -mips3 -mips4 -mlong64 -mlong32 -mlong-calls -mmemcpy
359 -mmips-as -mmips-tfile -mno-abicalls
360 -mno-embedded-data -mno-uninit-const-in-rodata -mno-embedded-pic
361 -mno-gpopt -mno-long-calls
362 -mno-memcpy -mno-mips-tfile -mno-rnames -mno-stats
363 -mrnames -msoft-float
364 -m4650 -msingle-float -mmad
365 -mstats -EL -EB -G @var{num} -nocpp
366 -mabi=32 -mabi=n32 -mabi=64 -mabi=eabi
371 -march=@var{cpu type}
372 -mieee-fp -mno-fancy-math-387
373 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib
374 -mno-wide-multiply -mrtd -malign-double
375 -mreg-alloc=@var{list} -mregparm=@var{num}
376 -malign-jumps=@var{num} -malign-loops=@var{num}
377 -malign-functions=@var{num} -mpreferred-stack-boundary=@var{num}
378 -mthreads -mno-align-stringops -minline-all-stringops
379 -mpush-args -maccumulate-outgoing-args
382 -march=@var{architecture type}
383 -mbig-switch -mdisable-fpregs -mdisable-indexing
384 -mfast-indirect-calls -mgas -mjump-in-delay
385 -mlong-load-store -mno-big-switch -mno-disable-fpregs
386 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas
387 -mno-jump-in-delay -mno-long-load-store
388 -mno-portable-runtime -mno-soft-float
389 -mno-space-regs -msoft-float -mpa-risc-1-0
390 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime
391 -mschedule=@var{cpu type} -mspace-regs
393 @emph{Intel 960 Options}
394 -m@var{cpu type} -masm-compat -mclean-linkage
395 -mcode-align -mcomplex-addr -mleaf-procedures
396 -mic-compat -mic2.0-compat -mic3.0-compat
397 -mintel-asm -mno-clean-linkage -mno-code-align
398 -mno-complex-addr -mno-leaf-procedures
399 -mno-old-align -mno-strict-align -mno-tail-call
400 -mnumerics -mold-align -msoft-float -mstrict-align
403 @emph{DEC Alpha Options}
404 -mfp-regs -mno-fp-regs -mno-soft-float -msoft-float
406 -mieee -mieee-with-inexact -mieee-conformant
407 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode}
408 -mtrap-precision=@var{mode} -mbuild-constants
410 -mbwx -mno-bwx -mcix -mno-cix -mmax -mno-max
411 -mmemory-latency=@var{time}
413 @emph{Clipper Options}
416 @emph{H8/300 Options}
417 -mrelax -mh -ms -mint32 -malign-300
421 -m4-nofpu -m4-single-only -m4-single -m4
422 -mb -ml -mdalign -mrelax
423 -mbigtable -mfmovd -mhitachi -mnomacsave
424 -misize -mpadstruct -mspace
428 @emph{System V Options}
429 -Qy -Qn -YP,@var{paths} -Ym,@var{dir}
433 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text section}
434 -mdata=@var{data section} -mrodata=@var{readonly data section}
436 @emph{TMS320C3x/C4x Options}
437 -mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm
438 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload
439 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned
440 -mparallel-insns -mparallel-mpy -mpreserve-float
443 -mlong-calls -mno-long-calls -mep -mno-ep
444 -mprolog-function -mno-prolog-function -mspace
445 -mtda=@var{n} -msda=@var{n} -mzda=@var{n}
449 -m32032 -m32332 -m32532 -m32081 -m32381 -mmult-add -mnomult-add
450 -msoft-float -mrtd -mnortd -mregparam -mnoregparam -msb -mnosb
451 -mbitfield -mnobitfield -mhimem -mnohimem
454 -mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts
458 -mhardlit, -mno-hardlit -mdiv -mno-div -mrelax-immediates
459 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields
460 -m4byte-functions -mno-4byte-functions -mcallgraph-data
461 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim
462 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment
465 @item Code Generation Options
466 @xref{Code Gen Options,,Options for Code Generation Conventions}.
468 -fcall-saved-@var{reg} -fcall-used-@var{reg}
469 -fexceptions -funwind-tables -ffixed-@var{reg} -finhibit-size-directive
470 -fcheck-memory-usage -fprefix-function-name
471 -fno-common -fno-ident -fno-gnu-linker
472 -fpcc-struct-return -fpic -fPIC
473 -freg-struct-return -fshared-data -fshort-enums
474 -fshort-double -fvolatile -fvolatile-global -fvolatile-static
475 -fverbose-asm -fpack-struct -fstack-check
476 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym}
477 -fargument-alias -fargument-noalias
478 -fargument-noalias-global
484 * Overall Options:: Controlling the kind of output:
485 an executable, object files, assembler files,
486 or preprocessed source.
487 * C Dialect Options:: Controlling the variant of C language compiled.
488 * C++ Dialect Options:: Variations on C++.
489 * Language Independent Options:: Controlling how diagnostics should be
491 * Warning Options:: How picky should the compiler be?
492 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
493 * Optimize Options:: How much optimization?
494 * Preprocessor Options:: Controlling header files and macro definitions.
495 Also, getting dependency information for Make.
496 * Assembler Options:: Passing options to the assembler.
497 * Link Options:: Specifying libraries and so on.
498 * Directory Options:: Where to find header files and libraries.
499 Where to find the compiler executable files.
500 * Spec Files:: How to pass switches to sub-processes.
501 * Target Options:: Running a cross-compiler, or an old version of GCC.
504 @node Overall Options
505 @section Options Controlling the Kind of Output
507 Compilation can involve up to four stages: preprocessing, compilation
508 proper, assembly and linking, always in that order. The first three
509 stages apply to an individual source file, and end by producing an
510 object file; linking combines all the object files (those newly
511 compiled, and those specified as input) into an executable file.
513 @cindex file name suffix
514 For any given input file, the file name suffix determines what kind of
519 C source code which must be preprocessed.
522 C source code which should not be preprocessed.
525 C++ source code which should not be preprocessed.
528 Objective-C source code. Note that you must link with the library
529 @file{libobjc.a} to make an Objective-C program work.
532 C header file (not to be compiled or linked).
535 @itemx @var{file}.cxx
536 @itemx @var{file}.cpp
538 C++ source code which must be preprocessed. Note that in @samp{.cxx},
539 the last two letters must both be literally @samp{x}. Likewise,
540 @samp{.C} refers to a literal capital C.
546 Assembler code which must be preprocessed.
549 An object file to be fed straight into linking.
550 Any file name with no recognized suffix is treated this way.
553 You can specify the input language explicitly with the @samp{-x} option:
556 @item -x @var{language}
557 Specify explicitly the @var{language} for the following input files
558 (rather than letting the compiler choose a default based on the file
559 name suffix). This option applies to all following input files until
560 the next @samp{-x} option. Possible values for @var{language} are:
563 c-header cpp-output c++-cpp-output
564 assembler assembler-with-cpp
568 Turn off any specification of a language, so that subsequent files are
569 handled according to their file name suffixes (as they are if @samp{-x}
570 has not been used at all).
572 @item -pass-exit-codes
573 Normally the @code{gcc} program will exit with the code of 1 if any
574 phase of the compiler returns a non-success return code. If you specify
575 @samp{-pass-exit-codes}, the @code{gcc} program will instead return with
576 numerically highest error produced by any phase that returned an error
580 If you only want some of the stages of compilation, you can use
581 @samp{-x} (or filename suffixes) to tell @code{gcc} where to start, and
582 one of the options @samp{-c}, @samp{-S}, or @samp{-E} to say where
583 @code{gcc} is to stop. Note that some combinations (for example,
584 @samp{-x cpp-output -E} instruct @code{gcc} to do nothing at all.
588 Compile or assemble the source files, but do not link. The linking
589 stage simply is not done. The ultimate output is in the form of an
590 object file for each source file.
592 By default, the object file name for a source file is made by replacing
593 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
595 Unrecognized input files, not requiring compilation or assembly, are
599 Stop after the stage of compilation proper; do not assemble. The output
600 is in the form of an assembler code file for each non-assembler input
603 By default, the assembler file name for a source file is made by
604 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
606 Input files that don't require compilation are ignored.
609 Stop after the preprocessing stage; do not run the compiler proper. The
610 output is in the form of preprocessed source code, which is sent to the
613 Input files which don't require preprocessing are ignored.
615 @cindex output file option
617 Place output in file @var{file}. This applies regardless to whatever
618 sort of output is being produced, whether it be an executable file,
619 an object file, an assembler file or preprocessed C code.
621 Since only one output file can be specified, it does not make sense to
622 use @samp{-o} when compiling more than one input file, unless you are
623 producing an executable file as output.
625 If @samp{-o} is not specified, the default is to put an executable file
626 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
627 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
628 all preprocessed C source on standard output.@refill
631 Print (on standard error output) the commands executed to run the stages
632 of compilation. Also print the version number of the compiler driver
633 program and of the preprocessor and the compiler proper.
636 Use pipes rather than temporary files for communication between the
637 various stages of compilation. This fails to work on some systems where
638 the assembler is unable to read from a pipe; but the GNU assembler has
642 Print (on the standard output) a description of the command line options
643 understood by @code{gcc}. If the @code{-v} option is also specified
644 then @code{--help} will also be passed on to the various processes
645 invoked by @code{gcc}, so that they can display the command line options
646 they accept. If the @code{-W} option is also specified then command
647 line options which have no documentation associated with them will also
652 @section Compiling C++ Programs
654 @cindex suffixes for C++ source
655 @cindex C++ source file suffixes
656 C++ source files conventionally use one of the suffixes @samp{.C},
657 @samp{.cc}, @samp{.cpp}, @samp{.c++}, @samp{.cp}, or @samp{.cxx};
658 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
659 files with these names and compiles them as C++ programs even if you
660 call the compiler the same way as for compiling C programs (usually with
661 the name @code{gcc}).
665 However, C++ programs often require class libraries as well as a
666 compiler that understands the C++ language---and under some
667 circumstances, you might want to compile programs from standard input,
668 or otherwise without a suffix that flags them as C++ programs.
669 @code{g++} is a program that calls GCC with the default language
670 set to C++, and automatically specifies linking against the C++
671 library. On many systems, @code{g++} is also
672 installed with the name @code{c++}.
674 @cindex invoking @code{g++}
675 When you compile C++ programs, you may specify many of the same
676 command-line options that you use for compiling programs in any
677 language; or command-line options meaningful for C and related
678 languages; or options that are meaningful only for C++ programs.
679 @xref{C Dialect Options,,Options Controlling C Dialect}, for
680 explanations of options for languages related to C.
681 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
682 explanations of options that are meaningful only for C++ programs.
684 @node C Dialect Options
685 @section Options Controlling C Dialect
686 @cindex dialect options
687 @cindex language dialect options
688 @cindex options, dialect
690 The following options control the dialect of C (or languages derived
691 from C, such as C++ and Objective C) that the compiler accepts:
696 In C mode, support all ANSI standard C programs. In C++ mode,
697 remove GNU extensions that conflict with ISO C++.
698 @c shouldn't we be saying "ISO"?
700 This turns off certain features of GCC that are incompatible with ANSI
701 C (when compiling C code), or of standard C++ (when compiling C++ code),
702 such as the @code{asm} and @code{typeof} keywords, and
703 predefined macros such as @code{unix} and @code{vax} that identify the
704 type of system you are using. It also enables the undesirable and
705 rarely used ANSI trigraph feature. For the C compiler,
706 it disables recognition of C++ style @samp{//} comments as well as
707 the @code{inline} keyword.
709 The alternate keywords @code{__asm__}, @code{__extension__},
710 @code{__inline__} and @code{__typeof__} continue to work despite
711 @samp{-ansi}. You would not want to use them in an ANSI C program, of
712 course, but it is useful to put them in header files that might be included
713 in compilations done with @samp{-ansi}. Alternate predefined macros
714 such as @code{__unix__} and @code{__vax__} are also available, with or
715 without @samp{-ansi}.
717 The @samp{-ansi} option does not cause non-ANSI programs to be
718 rejected gratuitously. For that, @samp{-pedantic} is required in
719 addition to @samp{-ansi}. @xref{Warning Options}.
721 The macro @code{__STRICT_ANSI__} is predefined when the @samp{-ansi}
722 option is used. Some header files may notice this macro and refrain
723 from declaring certain functions or defining certain macros that the
724 ANSI standard doesn't call for; this is to avoid interfering with any
725 programs that might use these names for other things.
727 The functions @code{alloca}, @code{abort}, @code{exit}, and
728 @code{_exit} are not builtin functions when @samp{-ansi} is used.
731 Determine the language standard. A value for this option must be provided;
739 ISO C as modified in amend. 1
742 ISO C99. Note that this standard is not yet fully supported; see
743 @url{http://gcc.gnu.org/c99status.html} for more information.
746 same as -std=iso9899:1990
749 same as -std=iso9899:1999
752 default, iso9899:1990 + gnu extensions
755 iso9899:1999 + gnu extensions
758 same as -std=iso9899:1999, deprecated
761 same as -std=iso9899:1999, deprecated
764 same as -std=gnu99, deprecated
768 Even when this option is not specified, you can still use some of the
769 features of newer standards in so far as they do not conflict with
770 previous C standards. For example, you may use @code{__restrict__} even
771 when -std=c99 is not specified.
774 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
775 keyword, so that code can use these words as identifiers. You can use
776 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
777 instead. @samp{-ansi} implies @samp{-fno-asm}.
779 In C++, this switch only affects the @code{typeof} keyword, since
780 @code{asm} and @code{inline} are standard keywords. You may want to
781 use the @samp{-fno-gnu-keywords} flag instead, which has the same effect.
784 @cindex builtin functions
810 Don't recognize builtin functions that do not begin with @samp{__builtin_}
811 as prefix. Currently, the functions affected include @code{abort},
812 @code{abs}, @code{alloca}, @code{cos}, @code{cosf}, @code{cosl},
813 @code{exit}, @code{_exit}, @code{fabs}, @code{fabsf}, @code{fabsl},
814 @code{ffs}, @code{labs}, @code{memcmp}, @code{memcpy}, @code{memset},
815 @code{sin}, @code{sinf}, @code{sinl}, @code{sqrt}, @code{sqrtf},
816 @code{sqrtl}, @code{strcmp}, @code{strcpy}, and @code{strlen}.
818 GCC normally generates special code to handle certain builtin functions
819 more efficiently; for instance, calls to @code{alloca} may become single
820 instructions that adjust the stack directly, and calls to @code{memcpy}
821 may become inline copy loops. The resulting code is often both smaller
822 and faster, but since the function calls no longer appear as such, you
823 cannot set a breakpoint on those calls, nor can you change the behavior
824 of the functions by linking with a different library.
826 The @samp{-ansi} option prevents @code{alloca}, @code{ffs} and @code{_exit}
827 from being builtin functions, since these functions do not have an ANSI
831 @cindex hosted environment
833 Assert that compilation takes place in a hosted environment. This implies
834 @samp{-fbuiltin}. A hosted environment is one in which the
835 entire standard library is available, and in which @code{main} has a return
836 type of @code{int}. Examples are nearly everything except a kernel.
837 This is equivalent to @samp{-fno-freestanding}.
840 @cindex hosted environment
842 Assert that compilation takes place in a freestanding environment. This
843 implies @samp{-fno-builtin}. A freestanding environment
844 is one in which the standard library may not exist, and program startup may
845 not necessarily be at @code{main}. The most obvious example is an OS kernel.
846 This is equivalent to @samp{-fno-hosted}.
849 Support ANSI C trigraphs. You don't want to know about this
850 brain-damage. The @samp{-ansi} option implies @samp{-trigraphs}.
852 @cindex traditional C language
853 @cindex C language, traditional
855 Attempt to support some aspects of traditional C compilers.
860 All @code{extern} declarations take effect globally even if they
861 are written inside of a function definition. This includes implicit
862 declarations of functions.
865 The newer keywords @code{typeof}, @code{inline}, @code{signed}, @code{const}
866 and @code{volatile} are not recognized. (You can still use the
867 alternative keywords such as @code{__typeof__}, @code{__inline__}, and
871 Comparisons between pointers and integers are always allowed.
874 Integer types @code{unsigned short} and @code{unsigned char} promote
875 to @code{unsigned int}.
878 Out-of-range floating point literals are not an error.
881 Certain constructs which ANSI regards as a single invalid preprocessing
882 number, such as @samp{0xe-0xd}, are treated as expressions instead.
885 String ``constants'' are not necessarily constant; they are stored in
886 writable space, and identical looking constants are allocated
887 separately. (This is the same as the effect of
888 @samp{-fwritable-strings}.)
890 @cindex @code{longjmp} and automatic variables
892 All automatic variables not declared @code{register} are preserved by
893 @code{longjmp}. Ordinarily, GNU C follows ANSI C: automatic variables
894 not declared @code{volatile} may be clobbered.
899 @cindex escape sequences, traditional
900 The character escape sequences @samp{\x} and @samp{\a} evaluate as the
901 literal characters @samp{x} and @samp{a} respectively. Without
902 @w{@samp{-traditional}}, @samp{\x} is a prefix for the hexadecimal
903 representation of a character, and @samp{\a} produces a bell.
906 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
907 if your program uses names that are normally GNU C builtin functions for
908 other purposes of its own.
910 You cannot use @samp{-traditional} if you include any header files that
911 rely on ANSI C features. Some vendors are starting to ship systems with
912 ANSI C header files and you cannot use @samp{-traditional} on such
913 systems to compile files that include any system headers.
915 The @samp{-traditional} option also enables @samp{-traditional-cpp},
916 which is described next.
918 @item -traditional-cpp
919 Attempt to support some aspects of traditional C preprocessors.
924 Comments convert to nothing at all, rather than to a space. This allows
925 traditional token concatenation.
928 In a preprocessing directive, the @samp{#} symbol must appear as the first
932 Macro arguments are recognized within string constants in a macro
933 definition (and their values are stringified, though without additional
934 quote marks, when they appear in such a context). The preprocessor
935 always considers a string constant to end at a newline.
938 @cindex detecting @w{@samp{-traditional}}
939 The predefined macro @code{__STDC__} is not defined when you use
940 @samp{-traditional}, but @code{__GNUC__} is (since the GNU extensions
941 which @code{__GNUC__} indicates are not affected by
942 @samp{-traditional}). If you need to write header files that work
943 differently depending on whether @samp{-traditional} is in use, by
944 testing both of these predefined macros you can distinguish four
945 situations: GNU C, traditional GNU C, other ANSI C compilers, and other
946 old C compilers. The predefined macro @code{__STDC_VERSION__} is also
947 not defined when you use @samp{-traditional}. @xref{Standard
948 Predefined,,Standard Predefined Macros,cpp.info,The C Preprocessor},
949 for more discussion of these and other predefined macros.
952 @cindex string constants vs newline
953 @cindex newline vs string constants
954 The preprocessor considers a string constant to end at a newline (unless
955 the newline is escaped with @samp{\}). (Without @w{@samp{-traditional}},
956 string constants can contain the newline character as typed.)
959 @item -fcond-mismatch
960 Allow conditional expressions with mismatched types in the second and
961 third arguments. The value of such an expression is void.
963 @item -funsigned-char
964 Let the type @code{char} be unsigned, like @code{unsigned char}.
966 Each kind of machine has a default for what @code{char} should
967 be. It is either like @code{unsigned char} by default or like
968 @code{signed char} by default.
970 Ideally, a portable program should always use @code{signed char} or
971 @code{unsigned char} when it depends on the signedness of an object.
972 But many programs have been written to use plain @code{char} and
973 expect it to be signed, or expect it to be unsigned, depending on the
974 machines they were written for. This option, and its inverse, let you
975 make such a program work with the opposite default.
977 The type @code{char} is always a distinct type from each of
978 @code{signed char} or @code{unsigned char}, even though its behavior
979 is always just like one of those two.
982 Let the type @code{char} be signed, like @code{signed char}.
984 Note that this is equivalent to @samp{-fno-unsigned-char}, which is
985 the negative form of @samp{-funsigned-char}. Likewise, the option
986 @samp{-fno-signed-char} is equivalent to @samp{-funsigned-char}.
988 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
989 if your program uses names that are normally GNU C builtin functions for
990 other purposes of its own.
992 You cannot use @samp{-traditional} if you include any header files that
993 rely on ANSI C features. Some vendors are starting to ship systems with
994 ANSI C header files and you cannot use @samp{-traditional} on such
995 systems to compile files that include any system headers.
997 @item -fsigned-bitfields
998 @itemx -funsigned-bitfields
999 @itemx -fno-signed-bitfields
1000 @itemx -fno-unsigned-bitfields
1001 These options control whether a bitfield is signed or unsigned, when the
1002 declaration does not use either @code{signed} or @code{unsigned}. By
1003 default, such a bitfield is signed, because this is consistent: the
1004 basic integer types such as @code{int} are signed types.
1006 However, when @samp{-traditional} is used, bitfields are all unsigned
1009 @item -fwritable-strings
1010 Store string constants in the writable data segment and don't uniquize
1011 them. This is for compatibility with old programs which assume they can
1012 write into string constants. The option @samp{-traditional} also has
1015 Writing into string constants is a very bad idea; ``constants'' should
1018 @item -fallow-single-precision
1019 Do not promote single precision math operations to double precision,
1020 even when compiling with @samp{-traditional}.
1022 Traditional K&R C promotes all floating point operations to double
1023 precision, regardless of the sizes of the operands. On the
1024 architecture for which you are compiling, single precision may be faster
1025 than double precision. If you must use @samp{-traditional}, but want
1026 to use single precision operations when the operands are single
1027 precision, use this option. This option has no effect when compiling
1028 with ANSI or GNU C conventions (the default).
1031 Override the underlying type for @samp{wchar_t} to be @samp{short
1032 unsigned int} instead of the default for the target. This option is
1033 useful for building programs to run under WINE.
1036 @node C++ Dialect Options
1037 @section Options Controlling C++ Dialect
1039 @cindex compiler options, C++
1040 @cindex C++ options, command line
1041 @cindex options, C++
1042 This section describes the command-line options that are only meaningful
1043 for C++ programs; but you can also use most of the GNU compiler options
1044 regardless of what language your program is in. For example, you
1045 might compile a file @code{firstClass.C} like this:
1048 g++ -g -frepo -O -c firstClass.C
1052 In this example, only @samp{-frepo} is an option meant
1053 only for C++ programs; you can use the other options with any
1054 language supported by GCC.
1056 Here is a list of options that are @emph{only} for compiling C++ programs:
1059 @item -fno-access-control
1060 Turn off all access checking. This switch is mainly useful for working
1061 around bugs in the access control code.
1064 Check that the pointer returned by @code{operator new} is non-null
1065 before attempting to modify the storage allocated. The current Working
1066 Paper requires that @code{operator new} never return a null pointer, so
1067 this check is normally unnecessary.
1069 An alternative to using this option is to specify that your
1070 @code{operator new} does not throw any exceptions; if you declare it
1071 @samp{throw()}, g++ will check the return value. See also @samp{new
1074 @item -fconserve-space
1075 Put uninitialized or runtime-initialized global variables into the
1076 common segment, as C does. This saves space in the executable at the
1077 cost of not diagnosing duplicate definitions. If you compile with this
1078 flag and your program mysteriously crashes after @code{main()} has
1079 completed, you may have an object that is being destroyed twice because
1080 two definitions were merged.
1082 This option is no longer useful on most targets, now that support has
1083 been added for putting variables into BSS without making them common.
1085 @item -fdollars-in-identifiers
1086 Accept @samp{$} in identifiers. You can also explicitly prohibit use of
1087 @samp{$} with the option @samp{-fno-dollars-in-identifiers}. (GNU C allows
1088 @samp{$} by default on most target systems, but there are a few exceptions.)
1089 Traditional C allowed the character @samp{$} to form part of
1090 identifiers. However, ANSI C and C++ forbid @samp{$} in identifiers.
1092 @item -fno-elide-constructors
1093 The C++ standard allows an implementation to omit creating a temporary
1094 which is only used to initialize another object of the same type.
1095 Specifying this option disables that optimization, and forces g++ to
1096 call the copy constructor in all cases.
1098 @item -fno-enforce-eh-specs
1099 Don't check for violation of exception specifications at runtime. This
1100 option violates the C++ standard, but may be useful for reducing code
1101 size in production builds, much like defining @samp{NDEBUG}. The compiler
1102 will still optimize based on the exception specifications.
1104 @item -fexternal-templates
1105 Cause template instantiations to obey @samp{#pragma interface} and
1106 @samp{implementation}; template instances are emitted or not according
1107 to the location of the template definition. @xref{Template
1108 Instantiation}, for more information.
1110 This option is deprecated.
1112 @item -falt-external-templates
1113 Similar to -fexternal-templates, but template instances are emitted or
1114 not according to the place where they are first instantiated.
1115 @xref{Template Instantiation}, for more information.
1117 This option is deprecated.
1120 @itemx -fno-for-scope
1121 If -ffor-scope is specified, the scope of variables declared in
1122 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1123 as specified by the C++ standard.
1124 If -fno-for-scope is specified, the scope of variables declared in
1125 a @i{for-init-statement} extends to the end of the enclosing scope,
1126 as was the case in old versions of gcc, and other (traditional)
1127 implementations of C++.
1129 The default if neither flag is given to follow the standard,
1130 but to allow and give a warning for old-style code that would
1131 otherwise be invalid, or have different behavior.
1133 @item -fno-gnu-keywords
1134 Do not recognize @code{typeof} as a keyword, so that code can use this
1135 word as an identifier. You can use the keyword @code{__typeof__} instead.
1136 @samp{-ansi} implies @samp{-fno-gnu-keywords}.
1139 Treat the @code{namespace std} as a namespace, instead of ignoring
1140 it. For compatibility with earlier versions of g++, the compiler will,
1141 by default, ignore @code{namespace-declarations},
1142 @code{using-declarations}, @code{using-directives}, and
1143 @code{namespace-names}, if they involve @code{std}.
1145 @item -fhuge-objects
1146 Support virtual function calls for objects that exceed the size
1147 representable by a @samp{short int}. Users should not use this flag by
1148 default; if you need to use it, the compiler will tell you so.
1150 This flag is not useful when compiling with -fvtable-thunks.
1152 Like all options that change the ABI, all C++ code, @emph{including
1153 libgcc} must be built with the same setting of this option.
1155 @item -fno-implicit-templates
1156 Never emit code for non-inline templates which are instantiated
1157 implicitly (i.e. by use); only emit code for explicit instantiations.
1158 @xref{Template Instantiation}, for more information.
1160 @item -fno-implicit-inline-templates
1161 Don't emit code for implicit instantiations of inline templates, either.
1162 The default is to handle inlines differently so that compiles with and
1163 without optimization will need the same set of explicit instantiations.
1165 @item -finit-priority
1166 Support @samp{__attribute__ ((init_priority (n)))} for controlling the
1167 order of initialization of file-scope objects. On ELF targets, this
1168 requires GNU ld 2.10 or later.
1170 @item -fno-implement-inlines
1171 To save space, do not emit out-of-line copies of inline functions
1172 controlled by @samp{#pragma implementation}. This will cause linker
1173 errors if these functions are not inlined everywhere they are called.
1175 @item -fms-extensions
1176 Disable pedwarns about constructs used in MFC, such as implicit int and
1177 getting a pointer to member function via non-standard syntax.
1179 @item -fname-mangling-version-@var{n}
1180 Control the way in which names are mangled. Version 0 is compatible
1181 with versions of g++ before 2.8. Version 1 is the default. Version 1
1182 will allow correct mangling of function templates. For example,
1183 version 0 mangling does not mangle foo<int, double> and foo<int, char>
1184 given this declaration:
1187 template <class T, class U> void foo(T t);
1190 Like all options that change the ABI, all C++ code, @emph{including
1191 libgcc} must be built with the same setting of this option.
1193 @item -fno-operator-names
1194 Do not treat the operator name keywords @code{and}, @code{bitand},
1195 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1196 synonyms as keywords.
1198 @item -fno-optional-diags
1199 Disable diagnostics that the standard says a compiler does not need to
1200 issue. Currently, the only such diagnostic issued by g++ is the one for
1201 a name having multiple meanings within a class.
1204 Downgrade messages about nonconformant code from errors to warnings. By
1205 default, g++ effectively sets @samp{-pedantic-errors} without
1206 @samp{-pedantic}; this option reverses that. This behavior and this
1207 option are superseded by @samp{-pedantic}, which works as it does for GNU C.
1210 Enable automatic template instantiation. This option also implies
1211 @samp{-fno-implicit-templates}. @xref{Template Instantiation}, for more
1215 Disable generation of information about every class with virtual
1216 functions for use by the C++ runtime type identification features
1217 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1218 of the language, you can save some space by using this flag. Note that
1219 exception handling uses the same information, but it will generate it as
1223 @itemx -fno-squangle
1224 @samp{-fsquangle} will enable a compressed form of name mangling for
1225 identifiers. In particular, it helps to shorten very long names by recognizing
1226 types and class names which occur more than once, replacing them with special
1227 short ID codes. This option also requires any C++ libraries being used to
1228 be compiled with this option as well. The compiler has this disabled (the
1229 equivalent of @samp{-fno-squangle}) by default.
1231 Like all options that change the ABI, all C++ code, @emph{including
1232 libgcc.a} must be built with the same setting of this option.
1234 @item -ftemplate-depth-@var{n}
1235 Set the maximum instantiation depth for template classes to @var{n}.
1236 A limit on the template instantiation depth is needed to detect
1237 endless recursions during template class instantiation. ANSI/ISO C++
1238 conforming programs must not rely on a maximum depth greater than 17.
1240 @item -fuse-cxa-atexit
1241 Register destructors for objects with static storage duration with the
1242 @code{__cxa_atexit} function rather than the @code{atexit} function.
1243 This option is required for fully standards-compliant handling of static
1244 destructors, but will only work if your C library supports
1245 @code{__cxa_atexit}.
1247 @item -fvtable-thunks
1248 Use @samp{thunks} to implement the virtual function dispatch table
1249 (@samp{vtable}). The traditional (cfront-style) approach to
1250 implementing vtables was to store a pointer to the function and two
1251 offsets for adjusting the @samp{this} pointer at the call site. Newer
1252 implementations store a single pointer to a @samp{thunk} function which
1253 does any necessary adjustment and then calls the target function.
1255 This option also enables a heuristic for controlling emission of
1256 vtables; if a class has any non-inline virtual functions, the vtable
1257 will be emitted in the translation unit containing the first one of
1260 Like all options that change the ABI, all C++ code, @emph{including
1261 libgcc.a} must be built with the same setting of this option.
1264 Do not search for header files in the standard directories specific to
1265 C++, but do still search the other standard directories. (This option
1266 is used when building the C++ library.)
1269 In addition, these optimization, warning, and code generation options
1270 have meanings only for C++ programs:
1273 @item -fno-default-inline
1274 Do not assume @samp{inline} for functions defined inside a class scope.
1275 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1276 functions will have linkage like inline functions; they just won't be
1279 @item -Wctor-dtor-privacy (C++ only)
1280 Warn when a class seems unusable, because all the constructors or
1281 destructors in a class are private and the class has no friends or
1282 public static member functions.
1284 @item -Wnon-virtual-dtor (C++ only)
1285 Warn when a class declares a non-virtual destructor that should probably
1286 be virtual, because it looks like the class will be used polymorphically.
1288 @item -Wreorder (C++ only)
1289 @cindex reordering, warning
1290 @cindex warning for reordering of member initializers
1291 Warn when the order of member initializers given in the code does not
1292 match the order in which they must be executed. For instance:
1298 A(): j (0), i (1) @{ @}
1302 Here the compiler will warn that the member initializers for @samp{i}
1303 and @samp{j} will be rearranged to match the declaration order of the
1307 The following @samp{-W@dots{}} options are not affected by @samp{-Wall}.
1310 @item -Weffc++ (C++ only)
1311 Warn about violations of various style guidelines from Scott Meyers'
1312 @cite{Effective C++} books. If you use this option, you should be aware
1313 that the standard library headers do not obey all of these guidelines;
1314 you can use @samp{grep -v} to filter out those warnings.
1316 @item -Wno-deprecated (C++ only)
1317 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1319 @item -Wno-non-template-friend (C++ only)
1320 Disable warnings when non-templatized friend functions are declared
1321 within a template. With the advent of explicit template specification
1322 support in g++, if the name of the friend is an unqualified-id (ie,
1323 @samp{friend foo(int)}), the C++ language specification demands that the
1324 friend declare or define an ordinary, nontemplate function. (Section
1325 14.5.3). Before g++ implemented explicit specification, unqualified-ids
1326 could be interpreted as a particular specialization of a templatized
1327 function. Because this non-conforming behavior is no longer the default
1328 behavior for g++, @samp{-Wnon-template-friend} allows the compiler to
1329 check existing code for potential trouble spots, and is on by default.
1330 This new compiler behavior can be turned off with
1331 @samp{-Wno-non-template-friend} which keeps the conformant compiler code
1332 but disables the helpful warning.
1334 @item -Wold-style-cast (C++ only)
1335 Warn if an old-style (C-style) cast is used within a C++ program. The
1336 new-style casts (@samp{static_cast}, @samp{reinterpret_cast}, and
1337 @samp{const_cast}) are less vulnerable to unintended effects.
1339 @item -Woverloaded-virtual (C++ only)
1340 @cindex overloaded virtual fn, warning
1341 @cindex warning for overloaded virtual fn
1342 Warn when a derived class function declaration may be an error in
1343 defining a virtual function. In a derived class, the
1344 definitions of virtual functions must match the type signature of a
1345 virtual function declared in the base class. With this option, the
1346 compiler warns when you define a function with the same name as a
1347 virtual function, but with a type signature that does not match any
1348 declarations from the base class.
1350 @item -Wno-pmf-conversions (C++ only)
1351 Disable the diagnostic for converting a bound pointer to member function
1354 @item -Wsign-promo (C++ only)
1355 Warn when overload resolution chooses a promotion from unsigned or
1356 enumeral type to a signed type over a conversion to an unsigned type of
1357 the same size. Previous versions of g++ would try to preserve
1358 unsignedness, but the standard mandates the current behavior.
1360 @item -Wsynth (C++ only)
1361 @cindex warning for synthesized methods
1362 @cindex synthesized methods, warning
1363 Warn when g++'s synthesis behavior does not match that of cfront. For
1369 A& operator = (int);
1379 In this example, g++ will synthesize a default @samp{A& operator =
1380 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1383 @node Language Independent Options
1384 @section Options to Control Diagnostic Messages Formatting
1385 @cindex options to control diagnostics formatting
1386 @cindex diagnostic messages
1387 @cindex message formatting
1389 Traditionally, diagnostic messages have been formatted irrespective of
1390 the output device's aspect (e.g. its width, ...). The options described
1391 below can be used to control the diagnostic messages formatting
1392 algorithm, e.g. how many characters per line, how often source location
1393 information should be reported. Right now, only the C++ front-end can
1394 honor these options. However it is expected, in the near future, that
1395 the remaining front-ends would be able to digest them correctly.
1398 @item -fmessage-length=@var{n}
1399 Try to format error messages so that they fit on lines of about @var{n}
1400 characters. The default is 72 characters for g++ and 0 for the rest of
1401 the front-ends supported by GCC. If @var{n} is zero, then no
1402 line-wrapping will be done; each error message will appear on a single
1405 @item -fdiagnostics-show-location=once
1406 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1407 reporter to emit @emph{once} source location information; that is, in
1408 case the message is too long to fit on a single physical line and has to
1409 be wrapped, the source location won't be emitted (as prefix) again,
1410 over and over, in subsequent continuation lines. This is the default
1413 @item -fdiagnostics-show-location=every-line
1414 Only meaningful in line-wrapping mode. Instructs the diagnostic
1415 messages reporter to emit the same source location information (as
1416 prefix) for physical lines that result from the process of breaking a
1417 a message which is too long to fit on a single line.
1421 @node Warning Options
1422 @section Options to Request or Suppress Warnings
1423 @cindex options to control warnings
1424 @cindex warning messages
1425 @cindex messages, warning
1426 @cindex suppressing warnings
1428 Warnings are diagnostic messages that report constructions which
1429 are not inherently erroneous but which are risky or suggest there
1430 may have been an error.
1432 You can request many specific warnings with options beginning @samp{-W},
1433 for example @samp{-Wimplicit} to request warnings on implicit
1434 declarations. Each of these specific warning options also has a
1435 negative form beginning @samp{-Wno-} to turn off warnings;
1436 for example, @samp{-Wno-implicit}. This manual lists only one of the
1437 two forms, whichever is not the default.
1439 These options control the amount and kinds of warnings produced by GCC:
1442 @cindex syntax checking
1444 Check the code for syntax errors, but don't do anything beyond that.
1447 Issue all the warnings demanded by strict ISO C and ISO C++;
1448 reject all programs that use forbidden extensions, and some other
1449 programs that do not follow ISO C and ISO C++. For ISO C, follows the
1450 version of the ISO C standard specified by any @samp{-std} option used.
1452 Valid ISO C and ISO C++ programs should compile properly with or without
1453 this option (though a rare few will require @samp{-ansi}). However,
1454 without this option, certain GNU extensions and traditional C and C++
1455 features are supported as well. With this option, they are rejected.
1457 @samp{-pedantic} does not cause warning messages for use of the
1458 alternate keywords whose names begin and end with @samp{__}. Pedantic
1459 warnings are also disabled in the expression that follows
1460 @code{__extension__}. However, only system header files should use
1461 these escape routes; application programs should avoid them.
1462 @xref{Alternate Keywords}.
1464 Some users try to use @samp{-pedantic} to check programs for strict ISO
1465 C conformance. They soon find that it does not do quite what they want:
1466 it finds some non-ANSI practices, but not all---only those for which
1467 ISO C @emph{requires} a diagnostic, and some others for which
1468 diagnostics have been added.
1470 A feature to report any failure to conform to ISO C might be useful in
1471 some instances, but would require considerable additional work and would
1472 be quite different from @samp{-pedantic}. We don't have plans to
1473 support such a feature in the near future.
1475 @item -pedantic-errors
1476 Like @samp{-pedantic}, except that errors are produced rather than
1480 Inhibit all warning messages.
1483 Inhibit warning messages about the use of @samp{#import}.
1485 @item -Wchar-subscripts
1486 Warn if an array subscript has type @code{char}. This is a common cause
1487 of error, as programmers often forget that this type is signed on some
1491 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
1492 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
1495 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
1496 the arguments supplied have types appropriate to the format string
1499 @item -Wimplicit-int
1500 Warn when a declaration does not specify a type.
1502 @item -Wimplicit-function-declaration
1503 @itemx -Werror-implicit-function-declaration
1504 Give a warning (or error) whenever a function is used before being
1508 Same as @samp{-Wimplicit-int} and @samp{-Wimplicit-function-}@*
1512 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
1513 function with external linkage, returning int, taking either zero
1514 arguments, two, or three arguments of appropriate types.
1517 Warn if a multicharacter constant (@samp{'FOOF'}) is used. Usually they
1518 indicate a typo in the user's code, as they have implementation-defined
1519 values, and should not be used in portable code.
1522 Warn if parentheses are omitted in certain contexts, such
1523 as when there is an assignment in a context where a truth value
1524 is expected, or when operators are nested whose precedence people
1525 often get confused about.
1527 Also warn about constructions where there may be confusion to which
1528 @code{if} statement an @code{else} branch belongs. Here is an example of
1541 In C, every @code{else} branch belongs to the innermost possible @code{if}
1542 statement, which in this example is @code{if (b)}. This is often not
1543 what the programmer expected, as illustrated in the above example by
1544 indentation the programmer chose. When there is the potential for this
1545 confusion, GNU C will issue a warning when this flag is specified.
1546 To eliminate the warning, add explicit braces around the innermost
1547 @code{if} statement so there is no way the @code{else} could belong to
1548 the enclosing @code{if}. The resulting code would look like this:
1563 Warn whenever a function is defined with a return-type that defaults to
1564 @code{int}. Also warn about any @code{return} statement with no
1565 return-value in a function whose return-type is not @code{void}.
1567 For C++, a function without return type always produces a diagnostic
1568 message, even when @samp{-Wno-return-type} is specified. The only
1569 exceptions are @samp{main} and functions defined in system headers.
1572 Warn whenever a @code{switch} statement has an index of enumeral type
1573 and lacks a @code{case} for one or more of the named codes of that
1574 enumeration. (The presence of a @code{default} label prevents this
1575 warning.) @code{case} labels outside the enumeration range also
1576 provoke warnings when this option is used.
1579 Warn if any trigraphs are encountered (assuming they are enabled).
1581 @item -Wunused-function
1582 Warn whenever a static function is declared but not defined or a
1583 non\-inline static function is unused.
1585 @item -Wunused-label
1586 Warn whenever a label is declared but not used.
1588 To suppress this warning use the @samp{unused} attribute
1589 (@pxref{Variable Attributes}).
1591 @item -Wunused-parameter
1592 Warn whenever a function parameter is unused aside from its declaration.
1594 To suppress this warning use the @samp{unused} attribute
1595 (@pxref{Variable Attributes}).
1597 @item -Wunused-variable
1598 Warn whenever a local variable or non-constant static variable is unused
1599 aside from its declaration
1601 To suppress this warning use the @samp{unused} attribute
1602 (@pxref{Variable Attributes}).
1604 @item -Wunused-value
1605 Warn whenever a statement computes a result that is explicitly not used.
1607 To suppress this warning cast the expression to @samp{void}.
1610 All all the above @samp{-Wunused} options combined.
1612 In order to get a warning about an unused function parameter, you must
1613 either specify @samp{-W -Wunused} or separately specify
1614 @samp{-Wunused-parameter}.
1616 @item -Wuninitialized
1617 Warn if an automatic variable is used without first being initialized or
1618 if a variable may be clobbered by a @code{setjmp} call.
1620 These warnings are possible only in optimizing compilation,
1621 because they require data flow information that is computed only
1622 when optimizing. If you don't specify @samp{-O}, you simply won't
1625 These warnings occur only for variables that are candidates for
1626 register allocation. Therefore, they do not occur for a variable that
1627 is declared @code{volatile}, or whose address is taken, or whose size
1628 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
1629 structures, unions or arrays, even when they are in registers.
1631 Note that there may be no warning about a variable that is used only
1632 to compute a value that itself is never used, because such
1633 computations may be deleted by data flow analysis before the warnings
1636 These warnings are made optional because GCC is not smart
1637 enough to see all the reasons why the code might be correct
1638 despite appearing to have an error. Here is one example of how
1657 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
1658 always initialized, but GCC doesn't know this. Here is
1659 another common case:
1664 if (change_y) save_y = y, y = new_y;
1666 if (change_y) y = save_y;
1671 This has no bug because @code{save_y} is used only if it is set.
1673 @cindex @code{longjmp} warnings
1674 This option also warns when a non-volatile automatic variable might be
1675 changed by a call to @code{longjmp}. These warnings as well are possible
1676 only in optimizing compilation.
1678 The compiler sees only the calls to @code{setjmp}. It cannot know
1679 where @code{longjmp} will be called; in fact, a signal handler could
1680 call it at any point in the code. As a result, you may get a warning
1681 even when there is in fact no problem because @code{longjmp} cannot
1682 in fact be called at the place which would cause a problem.
1684 Some spurious warnings can be avoided if you declare all the functions
1685 you use that never return as @code{noreturn}. @xref{Function
1688 @item -Wreorder (C++ only)
1689 @cindex reordering, warning
1690 @cindex warning for reordering of member initializers
1691 Warn when the order of member initializers given in the code does not
1692 match the order in which they must be executed. For instance:
1694 @item -Wunknown-pragmas
1695 @cindex warning for unknown pragmas
1696 @cindex unknown pragmas, warning
1697 @cindex pragmas, warning of unknown
1698 Warn when a #pragma directive is encountered which is not understood by
1699 GCC. If this command line option is used, warnings will even be issued
1700 for unknown pragmas in system header files. This is not the case if
1701 the warnings were only enabled by the @samp{-Wall} command line option.
1704 All of the above @samp{-W} options combined. This enables all the
1705 warnings about constructions that some users consider questionable, and
1706 that are easy to avoid (or modify to prevent the warning), even in
1707 conjunction with macros.
1710 The following @samp{-W@dots{}} options are not implied by @samp{-Wall}.
1711 Some of them warn about constructions that users generally do not
1712 consider questionable, but which occasionally you might wish to check
1713 for; others warn about constructions that are necessary or hard to avoid
1714 in some cases, and there is no simple way to modify the code to suppress
1719 Print extra warning messages for these events:
1723 A function can return either with or without a value. (Falling
1724 off the end of the function body is considered returning without
1725 a value.) For example, this function would evoke such a
1739 An expression-statement or the left-hand side of a comma expression
1740 contains no side effects.
1741 To suppress the warning, cast the unused expression to void.
1742 For example, an expression such as @samp{x[i,j]} will cause a warning,
1743 but @samp{x[(void)i,j]} will not.
1746 An unsigned value is compared against zero with @samp{<} or @samp{<=}.
1749 A comparison like @samp{x<=y<=z} appears; this is equivalent to
1750 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
1751 that of ordinary mathematical notation.
1754 Storage-class specifiers like @code{static} are not the first things in
1755 a declaration. According to the C Standard, this usage is obsolescent.
1758 If @samp{-Wall} or @samp{-Wunused} is also specified, warn about unused
1762 A comparison between signed and unsigned values could produce an
1763 incorrect result when the signed value is converted to unsigned.
1764 (But don't warn if @samp{-Wno-sign-compare} is also specified.)
1767 An aggregate has a partly bracketed initializer.
1768 For example, the following code would evoke such a warning,
1769 because braces are missing around the initializer for @code{x.h}:
1772 struct s @{ int f, g; @};
1773 struct t @{ struct s h; int i; @};
1774 struct t x = @{ 1, 2, 3 @};
1778 An aggregate has an initializer which does not initialize all members.
1779 For example, the following code would cause such a warning, because
1780 @code{x.h} would be implicitly initialized to zero:
1783 struct s @{ int f, g, h; @};
1784 struct s x = @{ 3, 4 @};
1789 Warn if floating point values are used in equality comparisons.
1791 The idea behind this is that sometimes it is convenient (for the
1792 programmer) to consider floating-point values as approximations to
1793 infinitely precise real numbers. If you are doing this, then you need
1794 to compute (by analysing the code, or in some other way) the maximum or
1795 likely maximum error that the computation introduces, and allow for it
1796 when performing comparisons (and when producing output, but that's a
1797 different problem). In particular, instead of testing for equality, you
1798 would check to see whether the two values have ranges that overlap; and
1799 this is done with the relational operators, so equality comparisons are
1802 @item -Wtraditional (C only)
1803 Warn about certain constructs that behave differently in traditional and
1808 Macro arguments occurring within string constants in the macro body.
1809 These would substitute the argument in traditional C, but are part of
1810 the constant in ANSI C.
1813 A function declared external in one block and then used after the end of
1817 A @code{switch} statement has an operand of type @code{long}.
1820 A non-@code{static} function declaration follows a @code{static} one.
1821 This construct is not accepted by some traditional C compilers.
1824 The ANSI type of an integer constant has a different width or
1825 signedness from its traditional type. This warning is only issued if
1826 the base of the constant is ten. I.e. hexadecimal or octal values, which
1827 typically represent bit patterns, are not warned about.
1830 Usage of ANSI string concatenation is detected.
1833 A function macro appears without arguments.
1836 The unary plus operator.
1839 Initialization of automatic aggregates.
1842 Identifier conflicts with labels. Traditional C lacks a separate
1843 namespace for labels.
1846 Initialization of unions. If the initializer is zero, the warning is
1847 omitted. This is done under the assumption that the zero initializer in
1848 user code appears conditioned on e.g. @code{__STDC__} to avoid missing
1849 initializer warnings and relies on default initialization to zero in the
1853 The `U' integer constant suffix, or the `F' or `L' floating point
1854 constant suffixes. (Traditonal C does support the `L' suffix on integer
1855 constants.) Note, these suffixes appear in macros defined in the system
1856 headers of most modern systems, e.g. the _MIN/_MAX macros in limits.h.
1857 Use of these macros can lead to spurious warnings as they do not
1858 necessarily reflect whether the code in question is any less portable to
1859 traditional C given that suitable backup definitions are provided.
1863 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
1866 Warn whenever a local variable shadows another local variable.
1868 @item -Wid-clash-@var{len}
1869 Warn whenever two distinct identifiers match in the first @var{len}
1870 characters. This may help you prepare a program that will compile
1871 with certain obsolete, brain-damaged compilers.
1873 @item -Wlarger-than-@var{len}
1874 Warn whenever an object of larger than @var{len} bytes is defined.
1876 @item -Wpointer-arith
1877 Warn about anything that depends on the ``size of'' a function type or
1878 of @code{void}. GNU C assigns these types a size of 1, for
1879 convenience in calculations with @code{void *} pointers and pointers
1882 @item -Wbad-function-cast (C only)
1883 Warn whenever a function call is cast to a non-matching type.
1884 For example, warn if @code{int malloc()} is cast to @code{anything *}.
1887 Warn whenever a pointer is cast so as to remove a type qualifier from
1888 the target type. For example, warn if a @code{const char *} is cast
1889 to an ordinary @code{char *}.
1892 Warn whenever a pointer is cast such that the required alignment of the
1893 target is increased. For example, warn if a @code{char *} is cast to
1894 an @code{int *} on machines where integers can only be accessed at
1895 two- or four-byte boundaries.
1897 @item -Wwrite-strings
1898 Give string constants the type @code{const char[@var{length}]} so that
1899 copying the address of one into a non-@code{const} @code{char *}
1900 pointer will get a warning. These warnings will help you find at
1901 compile time code that can try to write into a string constant, but
1902 only if you have been very careful about using @code{const} in
1903 declarations and prototypes. Otherwise, it will just be a nuisance;
1904 this is why we did not make @samp{-Wall} request these warnings.
1907 Warn if a prototype causes a type conversion that is different from what
1908 would happen to the same argument in the absence of a prototype. This
1909 includes conversions of fixed point to floating and vice versa, and
1910 conversions changing the width or signedness of a fixed point argument
1911 except when the same as the default promotion.
1913 Also, warn if a negative integer constant expression is implicitly
1914 converted to an unsigned type. For example, warn about the assignment
1915 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
1916 casts like @code{(unsigned) -1}.
1918 @item -Wsign-compare
1919 @cindex warning for comparison of signed and unsigned values
1920 @cindex comparison of signed and unsigned values, warning
1921 @cindex signed and unsigned values, comparison warning
1922 Warn when a comparison between signed and unsigned values could produce
1923 an incorrect result when the signed value is converted to unsigned.
1924 This warning is also enabled by @samp{-W}; to get the other warnings
1925 of @samp{-W} without this warning, use @samp{-W -Wno-sign-compare}.
1927 @item -Waggregate-return
1928 Warn if any functions that return structures or unions are defined or
1929 called. (In languages where you can return an array, this also elicits
1932 @item -Wstrict-prototypes (C only)
1933 Warn if a function is declared or defined without specifying the
1934 argument types. (An old-style function definition is permitted without
1935 a warning if preceded by a declaration which specifies the argument
1938 @item -Wmissing-prototypes (C only)
1939 Warn if a global function is defined without a previous prototype
1940 declaration. This warning is issued even if the definition itself
1941 provides a prototype. The aim is to detect global functions that fail
1942 to be declared in header files.
1944 @item -Wmissing-declarations
1945 Warn if a global function is defined without a previous declaration.
1946 Do so even if the definition itself provides a prototype.
1947 Use this option to detect global functions that are not declared in
1950 @item -Wmissing-noreturn
1951 Warn about functions which might be candidates for attribute @code{noreturn}.
1952 Note these are only possible candidates, not absolute ones. Care should
1953 be taken to manually verify functions actually do not ever return before
1954 adding the @code{noreturn} attribute, otherwise subtle code generation
1955 bugs could be introduced.
1958 Warn if a structure is given the packed attribute, but the packed
1959 attribute has no effect on the layout or size of the structure.
1960 Such structures may be mis-aligned for little benefit. For
1961 instance, in this code, the variable @code{f.x} in @code{struct bar}
1962 will be misaligned even though @code{struct bar} does not itself
1963 have the packed attribute:
1970 @} __attribute__((packed));
1979 Warn if padding is included in a structure, either to align an element
1980 of the structure or to align the whole structure. Sometimes when this
1981 happens it is possible to rearrange the fields of the structure to
1982 reduce the padding and so make the structure smaller.
1984 @item -Wredundant-decls
1985 Warn if anything is declared more than once in the same scope, even in
1986 cases where multiple declaration is valid and changes nothing.
1988 @item -Wnested-externs (C only)
1989 Warn if an @code{extern} declaration is encountered within a function.
1991 @item -Wunreachable-code
1992 Warn if the compiler detects that code will never be executed.
1994 This option is intended to warn when the compiler detects that at
1995 least a whole line of source code will never be executed, because
1996 some condition is never satisfied or because it is after a
1997 procedure that never returns.
1999 It is possible for this option to produce a warning even though there
2000 are circumstances under which part of the affected line can be executed,
2001 so care should be taken when removing apparently-unreachable code.
2003 For instance, when a function is inlined, a warning may mean that the
2004 line is unreachable in only one inlined copy of the function.
2006 This option is not made part of @samp{-Wall} because in a debugging
2007 version of a program there is often substantial code which checks
2008 correct functioning of the program and is, hopefully, unreachable
2009 because the program does work. Another common use of unreachable
2010 code is to provide behaviour which is selectable at compile-time.
2013 Warn if a function can not be inlined and it was declared as inline.
2016 Warn if @samp{long long} type is used. This is default. To inhibit
2017 the warning messages, use @samp{-Wno-long-long}. Flags
2018 @samp{-Wlong-long} and @samp{-Wno-long-long} are taken into account
2019 only when @samp{-pedantic} flag is used.
2021 @item -Wdisabled-optimization
2022 Warn if a requested optimization pass is disabled. This warning does
2023 not generally indicate that there is anything wrong with your code; it
2024 merely indicates that GCC's optimizers were unable to handle the code
2025 effectively. Often, the problem is that your code is too big or too
2026 complex; GCC will refuse to optimize programs when the optimization
2027 itself is likely to take inordinate amounts of time.
2030 Make all warnings into errors.
2033 @node Debugging Options
2034 @section Options for Debugging Your Program or GCC
2035 @cindex options, debugging
2036 @cindex debugging information options
2038 GCC has various special options that are used for debugging
2039 either your program or GCC:
2043 Produce debugging information in the operating system's native format
2044 (stabs, COFF, XCOFF, or DWARF). GDB can work with this debugging
2047 On most systems that use stabs format, @samp{-g} enables use of extra
2048 debugging information that only GDB can use; this extra information
2049 makes debugging work better in GDB but will probably make other debuggers
2051 refuse to read the program. If you want to control for certain whether
2052 to generate the extra information, use @samp{-gstabs+}, @samp{-gstabs},
2053 @samp{-gxcoff+}, @samp{-gxcoff}, @samp{-gdwarf-1+}, or @samp{-gdwarf-1}
2056 Unlike most other C compilers, GCC allows you to use @samp{-g} with
2057 @samp{-O}. The shortcuts taken by optimized code may occasionally
2058 produce surprising results: some variables you declared may not exist
2059 at all; flow of control may briefly move where you did not expect it;
2060 some statements may not be executed because they compute constant
2061 results or their values were already at hand; some statements may
2062 execute in different places because they were moved out of loops.
2064 Nevertheless it proves possible to debug optimized output. This makes
2065 it reasonable to use the optimizer for programs that might have bugs.
2067 The following options are useful when GCC is generated with the
2068 capability for more than one debugging format.
2071 Produce debugging information for use by GDB. This means to use the
2072 most expressive format available (DWARF 2, stabs, or the native format
2073 if neither of those are supported), including GDB extensions if at all
2077 Produce debugging information in stabs format (if that is supported),
2078 without GDB extensions. This is the format used by DBX on most BSD
2079 systems. On MIPS, Alpha and System V Release 4 systems this option
2080 produces stabs debugging output which is not understood by DBX or SDB.
2081 On System V Release 4 systems this option requires the GNU assembler.
2084 Produce debugging information in stabs format (if that is supported),
2085 using GNU extensions understood only by the GNU debugger (GDB). The
2086 use of these extensions is likely to make other debuggers crash or
2087 refuse to read the program.
2090 Produce debugging information in COFF format (if that is supported).
2091 This is the format used by SDB on most System V systems prior to
2095 Produce debugging information in XCOFF format (if that is supported).
2096 This is the format used by the DBX debugger on IBM RS/6000 systems.
2099 Produce debugging information in XCOFF format (if that is supported),
2100 using GNU extensions understood only by the GNU debugger (GDB). The
2101 use of these extensions is likely to make other debuggers crash or
2102 refuse to read the program, and may cause assemblers other than the GNU
2103 assembler (GAS) to fail with an error.
2106 Produce debugging information in DWARF version 1 format (if that is
2107 supported). This is the format used by SDB on most System V Release 4
2111 Produce debugging information in DWARF version 1 format (if that is
2112 supported), using GNU extensions understood only by the GNU debugger
2113 (GDB). The use of these extensions is likely to make other debuggers
2114 crash or refuse to read the program.
2117 Produce debugging information in DWARF version 2 format (if that is
2118 supported). This is the format used by DBX on IRIX 6.
2121 @itemx -ggdb@var{level}
2122 @itemx -gstabs@var{level}
2123 @itemx -gcoff@var{level}
2124 @itemx -gxcoff@var{level}
2125 @itemx -gdwarf@var{level}
2126 @itemx -gdwarf-2@var{level}
2127 Request debugging information and also use @var{level} to specify how
2128 much information. The default level is 2.
2130 Level 1 produces minimal information, enough for making backtraces in
2131 parts of the program that you don't plan to debug. This includes
2132 descriptions of functions and external variables, but no information
2133 about local variables and no line numbers.
2135 Level 3 includes extra information, such as all the macro definitions
2136 present in the program. Some debuggers support macro expansion when
2141 Generate extra code to write profile information suitable for the
2142 analysis program @code{prof}. You must use this option when compiling
2143 the source files you want data about, and you must also use it when
2146 @cindex @code{gprof}
2148 Generate extra code to write profile information suitable for the
2149 analysis program @code{gprof}. You must use this option when compiling
2150 the source files you want data about, and you must also use it when
2155 Generate extra code to write profile information for basic blocks, which will
2156 record the number of times each basic block is executed, the basic block start
2157 address, and the function name containing the basic block. If @samp{-g} is
2158 used, the line number and filename of the start of the basic block will also be
2159 recorded. If not overridden by the machine description, the default action is
2160 to append to the text file @file{bb.out}.
2162 This data could be analyzed by a program like @code{tcov}. Note,
2163 however, that the format of the data is not what @code{tcov} expects.
2164 Eventually GNU @code{gprof} should be extended to process this data.
2167 Makes the compiler print out each function name as it is compiled, and
2168 print some statistics about each pass when it finishes.
2171 Generate extra code to profile basic blocks. Your executable will
2172 produce output that is a superset of that produced when @samp{-a} is
2173 used. Additional output is the source and target address of the basic
2174 blocks where a jump takes place, the number of times a jump is executed,
2175 and (optionally) the complete sequence of basic blocks being executed.
2176 The output is appended to file @file{bb.out}.
2178 You can examine different profiling aspects without recompilation. Your
2179 executable will read a list of function names from file @file{bb.in}.
2180 Profiling starts when a function on the list is entered and stops when
2181 that invocation is exited. To exclude a function from profiling, prefix
2182 its name with `-'. If a function name is not unique, you can
2183 disambiguate it by writing it in the form
2184 @samp{/path/filename.d:functionname}. Your executable will write the
2185 available paths and filenames in file @file{bb.out}.
2187 Several function names have a special meaning:
2190 Write source, target and frequency of jumps to file @file{bb.out}.
2191 @item __bb_hidecall__
2192 Exclude function calls from frequency count.
2193 @item __bb_showret__
2194 Include function returns in frequency count.
2196 Write the sequence of basic blocks executed to file @file{bbtrace.gz}.
2197 The file will be compressed using the program @samp{gzip}, which must
2198 exist in your @code{PATH}. On systems without the @samp{popen}
2199 function, the file will be named @file{bbtrace} and will not be
2200 compressed. @strong{Profiling for even a few seconds on these systems
2201 will produce a very large file.} Note: @code{__bb_hidecall__} and
2202 @code{__bb_showret__} will not affect the sequence written to
2206 Here's a short example using different profiling parameters
2207 in file @file{bb.in}. Assume function @code{foo} consists of basic blocks
2208 1 and 2 and is called twice from block 3 of function @code{main}. After
2209 the calls, block 3 transfers control to block 4 of @code{main}.
2211 With @code{__bb_trace__} and @code{main} contained in file @file{bb.in},
2212 the following sequence of blocks is written to file @file{bbtrace.gz}:
2213 0 3 1 2 1 2 4. The return from block 2 to block 3 is not shown, because
2214 the return is to a point inside the block and not to the top. The
2215 block address 0 always indicates, that control is transferred
2216 to the trace from somewhere outside the observed functions. With
2217 @samp{-foo} added to @file{bb.in}, the blocks of function
2218 @code{foo} are removed from the trace, so only 0 3 4 remains.
2220 With @code{__bb_jumps__} and @code{main} contained in file @file{bb.in},
2221 jump frequencies will be written to file @file{bb.out}. The
2222 frequencies are obtained by constructing a trace of blocks
2223 and incrementing a counter for every neighbouring pair of blocks
2224 in the trace. The trace 0 3 1 2 1 2 4 displays the following
2228 Jump from block 0x0 to block 0x3 executed 1 time(s)
2229 Jump from block 0x3 to block 0x1 executed 1 time(s)
2230 Jump from block 0x1 to block 0x2 executed 2 time(s)
2231 Jump from block 0x2 to block 0x1 executed 1 time(s)
2232 Jump from block 0x2 to block 0x4 executed 1 time(s)
2235 With @code{__bb_hidecall__}, control transfer due to call instructions
2236 is removed from the trace, that is the trace is cut into three parts: 0
2237 3 4, 0 1 2 and 0 1 2. With @code{__bb_showret__}, control transfer due
2238 to return instructions is added to the trace. The trace becomes: 0 3 1
2239 2 3 1 2 3 4. Note, that this trace is not the same, as the sequence
2240 written to @file{bbtrace.gz}. It is solely used for counting jump
2243 @item -fprofile-arcs
2244 Instrument @dfn{arcs} during compilation. For each function of your
2245 program, GCC creates a program flow graph, then finds a spanning tree
2246 for the graph. Only arcs that are not on the spanning tree have to be
2247 instrumented: the compiler adds code to count the number of times that these
2248 arcs are executed. When an arc is the only exit or only entrance to a
2249 block, the instrumentation code can be added to the block; otherwise, a
2250 new basic block must be created to hold the instrumentation code.
2252 Since not every arc in the program must be instrumented, programs
2253 compiled with this option run faster than programs compiled with
2254 @samp{-a}, which adds instrumentation code to every basic block in the
2255 program. The tradeoff: since @code{gcov} does not have
2256 execution counts for all branches, it must start with the execution
2257 counts for the instrumented branches, and then iterate over the program
2258 flow graph until the entire graph has been solved. Hence, @code{gcov}
2259 runs a little more slowly than a program which uses information from
2262 @samp{-fprofile-arcs} also makes it possible to estimate branch
2263 probabilities, and to calculate basic block execution counts. In
2264 general, basic block execution counts do not give enough information to
2265 estimate all branch probabilities. When the compiled program exits, it
2266 saves the arc execution counts to a file called
2267 @file{@var{sourcename}.da}. Use the compiler option
2268 @samp{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
2269 Control Optimization}) when recompiling, to optimize using estimated
2270 branch probabilities.
2273 @item -ftest-coverage
2274 Create data files for the @code{gcov} code-coverage utility
2275 (@pxref{Gcov,, @code{gcov}: a GCC Test Coverage Program}).
2276 The data file names begin with the name of your source file:
2279 @item @var{sourcename}.bb
2280 A mapping from basic blocks to line numbers, which @code{gcov} uses to
2281 associate basic block execution counts with line numbers.
2283 @item @var{sourcename}.bbg
2284 A list of all arcs in the program flow graph. This allows @code{gcov}
2285 to reconstruct the program flow graph, so that it can compute all basic
2286 block and arc execution counts from the information in the
2287 @code{@var{sourcename}.da} file (this last file is the output from
2288 @samp{-fprofile-arcs}).
2291 @item -d@var{letters}
2292 Says to make debugging dumps during compilation at times specified by
2293 @var{letters}. This is used for debugging the compiler. The file names
2294 for most of the dumps are made by appending a pass number and a word to
2295 the source file name (e.g. @file{foo.c.00.rtl} or @file{foo.c.01.sibling}).
2296 Here are the possible letters for use in @var{letters}, and their meanings:
2300 Annotate the assembler output with miscellaneous debugging information.
2302 Dump after computing branch probabilities, to @file{@var{file}.11.bp}.
2304 Dump after block reordering, to @file{@var{file}.25.bbro}.
2306 Dump after instruction combination, to the file @file{@var{file}.14.combine}.
2308 Dump after the first if conversion, to the file @file{@var{file}.15.ce}.
2310 Dump after delayed branch scheduling, to @file{@var{file}.28.dbr}.
2312 Dump all macro definitions, at the end of preprocessing, in addition to
2315 Dump after SSA optimizations, to @file{@var{file}.05.ssa} and
2316 @file{@var{file}.06.ussa}.
2318 Dump after the second if conversion, to @file{@var{file}.21.ce2}.
2320 Dump after life analysis, to @file{@var{file}.13.life}.
2322 Dump after purging @code{ADDRESSOF} codes, to @file{@var{file}.04.addressof}.
2324 Dump after global register allocation, to @file{@var{file}.19.greg}.
2326 Dump after GCSE, to @file{@var{file}.08.gcse}.
2328 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
2330 Dump after the first jump optimization, to @file{@var{file}.02.jump}.
2332 Dump after the last jump optimization, to @file{@var{file}.26.jump2}.
2334 Dump after conversion from registers to stack, to @file{@var{file}.29.stack}.
2336 Dump after local register allocation, to @file{@var{file}.18.lreg}.
2338 Dump after loop optimization, to @file{@var{file}.09.loop}.
2340 Dump after performing the machine dependent reorganisation pass, to
2341 @file{@var{file}.27.mach}.
2343 Dump after register renumbering, to @file{@var{file}.23.rnreg}.
2345 Dump after the register move pass, to @file{@var{file}.16.regmove}.
2347 Dump after RTL generation, to @file{@var{file}.00.rtl}.
2349 Dump after the second instruction scheduling pass, to
2350 @file{@var{file}.24.sched2}.
2352 Dump after CSE (including the jump optimization that sometimes follows
2353 CSE), to @file{@var{file}.03.cse}.
2355 Dump after the first instruction scheduling pass, to
2356 @file{@var{file}.17.sched}.
2358 Dump after the second CSE pass (including the jump optimization that
2359 sometimes follows CSE), to @file{@var{file}.10.cse2}.
2361 Dump after the second flow pass, to @file{@var{file}.20.flow2}.
2363 Dump after dead code elimination, to @file{@var{file}.06.dce}.
2365 Dump after the peephole pass, to @file{@var{file}.22.peephole2}.
2367 Produce all the dumps listed above.
2369 Print statistics on memory usage, at the end of the run, to
2372 Annotate the assembler output with a comment indicating which
2373 pattern and alternative was used. The length of each instruction is
2376 Dump the RTL in the assembler output as a comment before each instruction.
2377 Also turns on @samp{-dp} annotation.
2379 For each of the other indicated dump files (except for
2380 @file{@var{file}.00.rtl}), dump a representation of the control flow graph
2381 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
2383 Just generate RTL for a function instead of compiling it. Usually used
2386 Dump debugging information during parsing, to standard error.
2389 @item -fdump-unnumbered
2390 When doing debugging dumps (see -d option above), suppress instruction
2391 numbers and line number note output. This makes it more feasible to
2392 use diff on debugging dumps for compiler invocations with different
2393 options, in particular with and without -g.
2395 @item -fdump-translation-unit-@var{file} (C++ only)
2396 Dump a representation of the tree structure for the entire translation
2399 @item -fpretend-float
2400 When running a cross-compiler, pretend that the target machine uses the
2401 same floating point format as the host machine. This causes incorrect
2402 output of the actual floating constants, but the actual instruction
2403 sequence will probably be the same as GCC would make when running on
2407 Store the usual ``temporary'' intermediate files permanently; place them
2408 in the current directory and name them based on the source file. Thus,
2409 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
2410 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}.
2413 Report the CPU time taken by each subprocess in the compilation
2414 sequence. For C source files, this is the preprocessor, compiler
2415 proper, and assembler. The output looks like this:
2423 The first number on each line is the ``user time,'' that is time spent
2424 executing the program itself. The second number is ``system time,''
2425 time spent executing operating system routines on behalf of the program.
2426 Both numbers are in seconds.
2428 @item -print-file-name=@var{library}
2429 Print the full absolute name of the library file @var{library} that
2430 would be used when linking---and don't do anything else. With this
2431 option, GCC does not compile or link anything; it just prints the
2434 @item -print-prog-name=@var{program}
2435 Like @samp{-print-file-name}, but searches for a program such as @samp{cpp}.
2437 @item -print-libgcc-file-name
2438 Same as @samp{-print-file-name=libgcc.a}.
2440 This is useful when you use @samp{-nostdlib} or @samp{-nodefaultlibs}
2441 but you do want to link with @file{libgcc.a}. You can do
2444 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
2447 @item -print-search-dirs
2448 Print the name of the configured installation directory and a list of
2449 program and library directories gcc will search---and don't do anything else.
2451 This is useful when gcc prints the error message
2452 @samp{installation problem, cannot exec cpp: No such file or directory}.
2453 To resolve this you either need to put @file{cpp} and the other compiler
2454 components where gcc expects to find them, or you can set the environment
2455 variable @code{GCC_EXEC_PREFIX} to the directory where you installed them.
2456 Don't forget the trailing '/'.
2457 @xref{Environment Variables}.
2460 @node Optimize Options
2461 @section Options That Control Optimization
2462 @cindex optimize options
2463 @cindex options, optimization
2465 These options control various sorts of optimizations:
2470 Optimize. Optimizing compilation takes somewhat more time, and a lot
2471 more memory for a large function.
2473 Without @samp{-O}, the compiler's goal is to reduce the cost of
2474 compilation and to make debugging produce the expected results.
2475 Statements are independent: if you stop the program with a breakpoint
2476 between statements, you can then assign a new value to any variable or
2477 change the program counter to any other statement in the function and
2478 get exactly the results you would expect from the source code.
2480 Without @samp{-O}, the compiler only allocates variables declared
2481 @code{register} in registers. The resulting compiled code is a little
2482 worse than produced by PCC without @samp{-O}.
2484 With @samp{-O}, the compiler tries to reduce code size and execution
2487 When you specify @samp{-O}, the compiler turns on @samp{-fthread-jumps}
2488 and @samp{-fdefer-pop} on all machines. The compiler turns on
2489 @samp{-fdelayed-branch} on machines that have delay slots, and
2490 @samp{-fomit-frame-pointer} on machines that can support debugging even
2491 without a frame pointer. On some machines the compiler also turns
2492 on other flags.@refill
2495 Optimize even more. GCC performs nearly all supported optimizations
2496 that do not involve a space-speed tradeoff. The compiler does not
2497 perform loop unrolling or function inlining when you specify @samp{-O2}.
2498 As compared to @samp{-O}, this option increases both compilation time
2499 and the performance of the generated code.
2501 @samp{-O2} turns on all optional optimizations except for loop unrolling
2502 and function inlining. It also turns on the @samp{-fforce-mem} option
2503 on all machines and frame pointer elimination on machines where doing so
2504 does not interfere with debugging.
2507 Optimize yet more. @samp{-O3} turns on all optimizations specified by
2508 @samp{-O2} and also turns on the @samp{inline-functions} option.
2514 Optimize for size. @samp{-Os} enables all @samp{-O2} optimizations that
2515 do not typically increase code size. It also performs further
2516 optimizations designed to reduce code size.
2518 If you use multiple @samp{-O} options, with or without level numbers,
2519 the last such option is the one that is effective.
2522 Options of the form @samp{-f@var{flag}} specify machine-independent
2523 flags. Most flags have both positive and negative forms; the negative
2524 form of @samp{-ffoo} would be @samp{-fno-foo}. In the table below,
2525 only one of the forms is listed---the one which is not the default.
2526 You can figure out the other form by either removing @samp{no-} or
2531 Do not store floating point variables in registers, and inhibit other
2532 options that might change whether a floating point value is taken from a
2535 @cindex floating point precision
2536 This option prevents undesirable excess precision on machines such as
2537 the 68000 where the floating registers (of the 68881) keep more
2538 precision than a @code{double} is supposed to have. Similarly for the
2539 x86 architecture. For most programs, the excess precision does only
2540 good, but a few programs rely on the precise definition of IEEE floating
2541 point. Use @samp{-ffloat-store} for such programs, after modifying
2542 them to store all pertinent intermediate computations into variables.
2544 @item -fno-default-inline
2545 Do not make member functions inline by default merely because they are
2546 defined inside the class scope (C++ only). Otherwise, when you specify
2547 @w{@samp{-O}}, member functions defined inside class scope are compiled
2548 inline by default; i.e., you don't need to add @samp{inline} in front of
2549 the member function name.
2551 @item -fno-defer-pop
2552 Always pop the arguments to each function call as soon as that function
2553 returns. For machines which must pop arguments after a function call,
2554 the compiler normally lets arguments accumulate on the stack for several
2555 function calls and pops them all at once.
2558 Force memory operands to be copied into registers before doing
2559 arithmetic on them. This produces better code by making all memory
2560 references potential common subexpressions. When they are not common
2561 subexpressions, instruction combination should eliminate the separate
2562 register-load. The @samp{-O2} option turns on this option.
2565 Force memory address constants to be copied into registers before
2566 doing arithmetic on them. This may produce better code just as
2567 @samp{-fforce-mem} may.
2569 @item -fomit-frame-pointer
2570 Don't keep the frame pointer in a register for functions that
2571 don't need one. This avoids the instructions to save, set up and
2572 restore frame pointers; it also makes an extra register available
2573 in many functions. @strong{It also makes debugging impossible on
2577 On some machines, such as the Vax, this flag has no effect, because
2578 the standard calling sequence automatically handles the frame pointer
2579 and nothing is saved by pretending it doesn't exist. The
2580 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2581 whether a target machine supports this flag. @xref{Registers}.@refill
2584 On some machines, such as the Vax, this flag has no effect, because
2585 the standard calling sequence automatically handles the frame pointer
2586 and nothing is saved by pretending it doesn't exist. The
2587 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2588 whether a target machine supports this flag. @xref{Registers,,Register
2589 Usage, gcc.info, Using and Porting GCC}.@refill
2592 @item -foptimize-sibling-calls
2593 Optimize sibling and tail recursive calls.
2596 Don't pay attention to the @code{inline} keyword. Normally this option
2597 is used to keep the compiler from expanding any functions inline.
2598 Note that if you are not optimizing, no functions can be expanded inline.
2600 @item -finline-functions
2601 Integrate all simple functions into their callers. The compiler
2602 heuristically decides which functions are simple enough to be worth
2603 integrating in this way.
2605 If all calls to a given function are integrated, and the function is
2606 declared @code{static}, then the function is normally not output as
2607 assembler code in its own right.
2609 @item -finline-limit=@var{n}
2610 By default, gcc limits the size of functions that can be inlined. This flag
2611 allows the control of this limit for functions that are explicitly marked as
2612 inline (ie marked with the inline keyword or defined within the class
2613 definition in c++). @var{n} is the size of functions that can be inlined in
2614 number of pseudo instructions (not counting parameter handling). The default
2615 value of n is 10000. Increasing this value can result in more inlined code at
2616 the cost of compilation time and memory consumption. Decreasing usually makes
2617 the compilation faster and less code will be inlined (which presumably
2618 means slower programs). This option is particularly useful for programs that
2619 use inlining heavily such as those based on recursive templates with c++.
2621 @emph{Note:} pseudo instruction represents, in this particular context, an
2622 abstract measurement of function's size. In no way, it represents a count
2623 of assembly instructions and as such its exact meaning might change from one
2624 release to an another.
2626 @item -fkeep-inline-functions
2627 Even if all calls to a given function are integrated, and the function
2628 is declared @code{static}, nevertheless output a separate run-time
2629 callable version of the function. This switch does not affect
2630 @code{extern inline} functions.
2632 @item -fkeep-static-consts
2633 Emit variables declared @code{static const} when optimization isn't turned
2634 on, even if the variables aren't referenced.
2636 GCC enables this option by default. If you want to force the compiler to
2637 check if the variable was referenced, regardless of whether or not
2638 optimization is turned on, use the @samp{-fno-keep-static-consts} option.
2640 @item -fno-function-cse
2641 Do not put function addresses in registers; make each instruction that
2642 calls a constant function contain the function's address explicitly.
2644 This option results in less efficient code, but some strange hacks
2645 that alter the assembler output may be confused by the optimizations
2646 performed when this option is not used.
2649 This option allows GCC to violate some ANSI or IEEE rules and/or
2650 specifications in the interest of optimizing code for speed. For
2651 example, it allows the compiler to assume arguments to the @code{sqrt}
2652 function are non-negative numbers and that no floating-point values
2655 This option should never be turned on by any @samp{-O} option since
2656 it can result in incorrect output for programs which depend on
2657 an exact implementation of IEEE or ANSI rules/specifications for
2660 @item -fno-math-errno
2661 Do not set ERRNO after calling math functions that are executed
2662 with a single instruction, e.g., sqrt. A program that relies on
2663 IEEE exceptions for math error handling may want to use this flag
2664 for speed while maintaining IEEE arithmetic compatibility.
2666 The default is @samp{-fmath-errno}. The @samp{-ffast-math} option
2667 sets @samp{-fno-math-errno}.
2670 @c following causes underfulls.. they don't look great, but we deal.
2672 The following options control specific optimizations. The @samp{-O2}
2673 option turns on all of these optimizations except @samp{-funroll-loops}
2674 and @samp{-funroll-all-loops}. On most machines, the @samp{-O} option
2675 turns on the @samp{-fthread-jumps} and @samp{-fdelayed-branch} options,
2676 but specific machines may handle it differently.
2678 You can use the following flags in the rare cases when ``fine-tuning''
2679 of optimizations to be performed is desired.
2682 @item -fstrength-reduce
2683 Perform the optimizations of loop strength reduction and
2684 elimination of iteration variables.
2686 @item -fthread-jumps
2687 Perform optimizations where we check to see if a jump branches to a
2688 location where another comparison subsumed by the first is found. If
2689 so, the first branch is redirected to either the destination of the
2690 second branch or a point immediately following it, depending on whether
2691 the condition is known to be true or false.
2693 @item -fcse-follow-jumps
2694 In common subexpression elimination, scan through jump instructions
2695 when the target of the jump is not reached by any other path. For
2696 example, when CSE encounters an @code{if} statement with an
2697 @code{else} clause, CSE will follow the jump when the condition
2700 @item -fcse-skip-blocks
2701 This is similar to @samp{-fcse-follow-jumps}, but causes CSE to
2702 follow jumps which conditionally skip over blocks. When CSE
2703 encounters a simple @code{if} statement with no else clause,
2704 @samp{-fcse-skip-blocks} causes CSE to follow the jump around the
2705 body of the @code{if}.
2707 @item -frerun-cse-after-loop
2708 Re-run common subexpression elimination after loop optimizations has been
2711 @item -frerun-loop-opt
2712 Run the loop optimizer twice.
2715 Perform a global common subexpression elimination pass.
2716 This pass also performs global constant and copy propagation.
2718 @item -fdelete-null-pointer-checks
2719 Use global dataflow analysis to identify and eliminate useless null
2720 pointer checks. Programs which rely on NULL pointer dereferences @emph{not}
2721 halting the program may not work properly with this option. Use
2722 -fno-delete-null-pointer-checks to disable this optimizing for programs
2723 which depend on that behavior.
2726 @item -fexpensive-optimizations
2727 Perform a number of minor optimizations that are relatively expensive.
2729 @item -foptimize-register-moves
2731 Attempt to reassign register numbers in move instructions and as
2732 operands of other simple instructions in order to maximize the amount of
2733 register tying. This is especially helpful on machines with two-operand
2734 instructions. GCC enables this optimization by default with @samp{-O2}
2737 Note @code{-fregmove} and @code{-foptimize-register-moves} are the same
2740 @item -fdelayed-branch
2741 If supported for the target machine, attempt to reorder instructions
2742 to exploit instruction slots available after delayed branch
2745 @item -fschedule-insns
2746 If supported for the target machine, attempt to reorder instructions to
2747 eliminate execution stalls due to required data being unavailable. This
2748 helps machines that have slow floating point or memory load instructions
2749 by allowing other instructions to be issued until the result of the load
2750 or floating point instruction is required.
2752 @item -fschedule-insns2
2753 Similar to @samp{-fschedule-insns}, but requests an additional pass of
2754 instruction scheduling after register allocation has been done. This is
2755 especially useful on machines with a relatively small number of
2756 registers and where memory load instructions take more than one cycle.
2758 @item -ffunction-sections
2759 @itemx -fdata-sections
2760 Place each function or data item into its own section in the output
2761 file if the target supports arbitrary sections. The name of the
2762 function or the name of the data item determines the section's name
2765 Use these options on systems where the linker can perform optimizations
2766 to improve locality of reference in the instruction space. HPPA
2767 processors running HP-UX and Sparc processors running Solaris 2 have
2768 linkers with such optimizations. Other systems using the ELF object format
2769 as well as AIX may have these optimizations in the future.
2771 Only use these options when there are significant benefits from doing
2772 so. When you specify these options, the assembler and linker will
2773 create larger object and executable files and will also be slower.
2774 You will not be able to use @code{gprof} on all systems if you
2775 specify this option and you may have problems with debugging if
2776 you specify both this option and @samp{-g}.
2778 @item -fcaller-saves
2779 Enable values to be allocated in registers that will be clobbered by
2780 function calls, by emitting extra instructions to save and restore the
2781 registers around such calls. Such allocation is done only when it
2782 seems to result in better code than would otherwise be produced.
2784 This option is always enabled by default on certain machines, usually
2785 those which have no call-preserved registers to use instead.
2787 For all machines, optimization level 2 and higher enables this flag by
2790 @item -funroll-loops
2791 Perform the optimization of loop unrolling. This is only done for loops
2792 whose number of iterations can be determined at compile time or run time.
2793 @samp{-funroll-loops} implies both @samp{-fstrength-reduce} and
2794 @samp{-frerun-cse-after-loop}.
2796 @item -funroll-all-loops
2797 Perform the optimization of loop unrolling. This is done for all loops
2798 and usually makes programs run more slowly. @samp{-funroll-all-loops}
2799 implies @samp{-fstrength-reduce} as well as @samp{-frerun-cse-after-loop}.
2801 @item -fmove-all-movables
2802 Forces all invariant computations in loops to be moved
2805 @item -freduce-all-givs
2806 Forces all general-induction variables in loops to be
2809 @emph{Note:} When compiling programs written in Fortran,
2810 @samp{-fmove-all-movables} and @samp{-freduce-all-givs} are enabled
2811 by default when you use the optimizer.
2813 These options may generate better or worse code; results are highly
2814 dependent on the structure of loops within the source code.
2816 These two options are intended to be removed someday, once
2817 they have helped determine the efficacy of various
2818 approaches to improving loop optimizations.
2820 Please let us (@code{gcc@@gcc.gnu.org} and @code{fortran@@gnu.org})
2821 know how use of these options affects
2822 the performance of your production code.
2823 We're very interested in code that runs @emph{slower}
2824 when these options are @emph{enabled}.
2827 Disable any machine-specific peephole optimizations.
2829 @item -fbranch-probabilities
2830 After running a program compiled with @samp{-fprofile-arcs}
2831 (@pxref{Debugging Options,, Options for Debugging Your Program or
2832 @code{gcc}}), you can compile it a second time using
2833 @samp{-fbranch-probabilities}, to improve optimizations based on
2834 guessing the path a branch might take.
2837 With @samp{-fbranch-probabilities}, GCC puts a @samp{REG_EXEC_COUNT}
2838 note on the first instruction of each basic block, and a
2839 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
2840 These can be used to improve optimization. Currently, they are only
2841 used in one place: in @file{reorg.c}, instead of guessing which path a
2842 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
2843 exactly determine which path is taken more often.
2846 @item -fstrict-aliasing
2847 Allows the compiler to assume the strictest aliasing rules applicable to
2848 the language being compiled. For C (and C++), this activates
2849 optimizations based on the type of expressions. In particular, an
2850 object of one type is assumed never to reside at the same address as an
2851 object of a different type, unless the types are almost the same. For
2852 example, an @code{unsigned int} can alias an @code{int}, but not a
2853 @code{void*} or a @code{double}. A character type may alias any other
2856 Pay special attention to code like this:
2869 The practice of reading from a different union member than the one most
2870 recently written to (called ``type-punning'') is common. Even with
2871 @samp{-fstrict-aliasing}, type-punning is allowed, provided the memory
2872 is accessed through the union type. So, the code above will work as
2873 expected. However, this code might not:
2885 Every language that wishes to perform language-specific alias analysis
2886 should define a function that computes, given an @code{tree}
2887 node, an alias set for the node. Nodes in different alias sets are not
2888 allowed to alias. For an example, see the C front-end function
2889 @code{c_get_alias_set}.
2892 @item -falign-functions
2893 @itemx -falign-functions=@var{n}
2894 Align the start of functions to the next power-of-two greater than
2895 @var{n}, skipping up to @var{n} bytes. For instance,
2896 @samp{-falign-functions=32} aligns functions to the next 32-byte
2897 boundary, but @samp{-falign-functions=24} would align to the next
2898 32-byte boundary only if this can be done by skipping 23 bytes or less.
2900 @samp{-fno-align-functions} and @samp{-falign-functions=1} are
2901 equivalent and mean that functions will not be aligned.
2903 Some assemblers only support this flag when @var{n} is a power of two;
2904 in that case, it is rounded up.
2906 If @var{n} is not specified, use a machine-dependent default.
2908 @item -falign-labels
2909 @itemx -falign-labels=@var{n}
2910 Align all branch targets to a power-of-two boundary, skipping up to
2911 @var{n} bytes like @samp{-falign-functions}. This option can easily
2912 make code slower, because it must insert dummy operations for when the
2913 branch target is reached in the usual flow of the code.
2915 If @samp{-falign-loops} or @samp{-falign-jumps} are applicable and
2916 are greater than this value, then their values are used instead.
2918 If @var{n} is not specified, use a machine-dependent default which is
2919 very likely to be @samp{1}, meaning no alignment.
2922 @itemx -falign-loops=@var{n}
2923 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
2924 like @samp{-falign-functions}. The hope is that the loop will be
2925 executed many times, which will make up for any execution of the dummy
2928 If @var{n} is not specified, use a machine-dependent default.
2931 @itemx -falign-jumps=@var{n}
2932 Align branch targets to a power-of-two boundary, for branch targets
2933 where the targets can only be reached by jumping, skipping up to @var{n}
2934 bytes like @samp{-falign-functions}. In this case, no dummy operations
2937 If @var{n} is not specified, use a machine-dependent default.
2940 Perform optimizations in static single assignment form. Each function's
2941 flow graph is translated into SSA form, optimizations are performed, and
2942 the flow graph is translated back from SSA form. User's should not
2943 specify this option, since it is not yet ready for production use.
2946 Perform dead-code elimination in SSA form. Requires @samp{-fssa}. Like
2947 @samp{-fssa}, this is an experimental feature.
2949 @item -fsingle-precision-constant
2950 Treat floating point constant as single precision constant instead of
2951 implicitly converting it to double precision constant.
2955 @node Preprocessor Options
2956 @section Options Controlling the Preprocessor
2957 @cindex preprocessor options
2958 @cindex options, preprocessor
2960 These options control the C preprocessor, which is run on each C source
2961 file before actual compilation.
2963 If you use the @samp{-E} option, nothing is done except preprocessing.
2964 Some of these options make sense only together with @samp{-E} because
2965 they cause the preprocessor output to be unsuitable for actual
2969 @item -include @var{file}
2970 Process @var{file} as input before processing the regular input file.
2971 In effect, the contents of @var{file} are compiled first. Any @samp{-D}
2972 and @samp{-U} options on the command line are always processed before
2973 @samp{-include @var{file}}, regardless of the order in which they are
2974 written. All the @samp{-include} and @samp{-imacros} options are
2975 processed in the order in which they are written.
2977 @item -imacros @var{file}
2978 Process @var{file} as input, discarding the resulting output, before
2979 processing the regular input file. Because the output generated from
2980 @var{file} is discarded, the only effect of @samp{-imacros @var{file}}
2981 is to make the macros defined in @var{file} available for use in the
2984 Any @samp{-D} and @samp{-U} options on the command line are always
2985 processed before @samp{-imacros @var{file}}, regardless of the order in
2986 which they are written. All the @samp{-include} and @samp{-imacros}
2987 options are processed in the order in which they are written.
2989 @item -idirafter @var{dir}
2990 @cindex second include path
2991 Add the directory @var{dir} to the second include path. The directories
2992 on the second include path are searched when a header file is not found
2993 in any of the directories in the main include path (the one that
2996 @item -iprefix @var{prefix}
2997 Specify @var{prefix} as the prefix for subsequent @samp{-iwithprefix}
3000 @item -iwithprefix @var{dir}
3001 Add a directory to the second include path. The directory's name is
3002 made by concatenating @var{prefix} and @var{dir}, where @var{prefix} was
3003 specified previously with @samp{-iprefix}. If you have not specified a
3004 prefix yet, the directory containing the installed passes of the
3005 compiler is used as the default.
3007 @item -iwithprefixbefore @var{dir}
3008 Add a directory to the main include path. The directory's name is made
3009 by concatenating @var{prefix} and @var{dir}, as in the case of
3010 @samp{-iwithprefix}.
3012 @item -isystem @var{dir}
3013 Add a directory to the beginning of the second include path, marking it
3014 as a system directory, so that it gets the same special treatment as
3015 is applied to the standard system directories.
3018 Do not search the standard system directories for header files. Only
3019 the directories you have specified with @samp{-I} options (and the
3020 current directory, if appropriate) are searched. @xref{Directory
3021 Options}, for information on @samp{-I}.
3023 By using both @samp{-nostdinc} and @samp{-I-}, you can limit the include-file
3024 search path to only those directories you specify explicitly.
3027 Do not predefine any nonstandard macros. (Including architecture flags).
3030 Run only the C preprocessor. Preprocess all the C source files
3031 specified and output the results to standard output or to the
3032 specified output file.
3035 Tell the preprocessor not to discard comments. Used with the
3039 Tell the preprocessor not to generate @samp{#line} directives.
3040 Used with the @samp{-E} option.
3043 @cindex dependencies, make
3045 Tell the preprocessor to output a rule suitable for @code{make}
3046 describing the dependencies of each object file. For each source file,
3047 the preprocessor outputs one @code{make}-rule whose target is the object
3048 file name for that source file and whose dependencies are all the
3049 @code{#include} header files it uses. This rule may be a single line or
3050 may be continued with @samp{\}-newline if it is long. The list of rules
3051 is printed on standard output instead of the preprocessed C program.
3053 @samp{-M} implies @samp{-E}.
3055 Another way to specify output of a @code{make} rule is by setting
3056 the environment variable @code{DEPENDENCIES_OUTPUT} (@pxref{Environment
3060 Like @samp{-M} but the output mentions only the user header files
3061 included with @samp{#include "@var{file}"}. System header files
3062 included with @samp{#include <@var{file}>} are omitted.
3065 Like @samp{-M} but the dependency information is written to a file made by
3066 replacing ".c" with ".d" at the end of the input file names.
3067 This is in addition to compiling the file as specified---@samp{-MD} does
3068 not inhibit ordinary compilation the way @samp{-M} does.
3070 In Mach, you can use the utility @code{md} to merge multiple dependency
3071 files into a single dependency file suitable for using with the @samp{make}
3075 Like @samp{-MD} except mention only user header files, not system
3079 Treat missing header files as generated files and assume they live in the
3080 same directory as the source file. If you specify @samp{-MG}, you
3081 must also specify either @samp{-M} or @samp{-MM}. @samp{-MG} is not
3082 supported with @samp{-MD} or @samp{-MMD}.
3085 Print the name of each header file used, in addition to other normal
3088 @item -A@var{question}(@var{answer})
3089 Assert the answer @var{answer} for @var{question}, in case it is tested
3090 with a preprocessing conditional such as @samp{#if
3091 #@var{question}(@var{answer})}. @samp{-A-} disables the standard
3092 assertions that normally describe the target machine.
3095 Define macro @var{macro} with the string @samp{1} as its definition.
3097 @item -D@var{macro}=@var{defn}
3098 Define macro @var{macro} as @var{defn}. All instances of @samp{-D} on
3099 the command line are processed before any @samp{-U} options.
3102 Undefine macro @var{macro}. @samp{-U} options are evaluated after all
3103 @samp{-D} options, but before any @samp{-include} and @samp{-imacros}
3107 Tell the preprocessor to output only a list of the macro definitions
3108 that are in effect at the end of preprocessing. Used with the @samp{-E}
3112 Tell the preprocessing to pass all macro definitions into the output, in
3113 their proper sequence in the rest of the output.
3116 Like @samp{-dD} except that the macro arguments and contents are omitted.
3117 Only @samp{#define @var{name}} is included in the output.
3120 Support ANSI C trigraphs. The @samp{-ansi} option also has this effect.
3122 @item -Wp,@var{option}
3123 Pass @var{option} as an option to the preprocessor. If @var{option}
3124 contains commas, it is split into multiple options at the commas.
3127 @node Assembler Options
3128 @section Passing Options to the Assembler
3130 @c prevent bad page break with this line
3131 You can pass options to the assembler.
3134 @item -Wa,@var{option}
3135 Pass @var{option} as an option to the assembler. If @var{option}
3136 contains commas, it is split into multiple options at the commas.
3140 @section Options for Linking
3141 @cindex link options
3142 @cindex options, linking
3144 These options come into play when the compiler links object files into
3145 an executable output file. They are meaningless if the compiler is
3146 not doing a link step.
3150 @item @var{object-file-name}
3151 A file name that does not end in a special recognized suffix is
3152 considered to name an object file or library. (Object files are
3153 distinguished from libraries by the linker according to the file
3154 contents.) If linking is done, these object files are used as input
3160 If any of these options is used, then the linker is not run, and
3161 object file names should not be used as arguments. @xref{Overall
3165 @item -l@var{library}
3166 Search the library named @var{library} when linking.
3168 It makes a difference where in the command you write this option; the
3169 linker searches processes libraries and object files in the order they
3170 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
3171 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
3172 to functions in @samp{z}, those functions may not be loaded.
3174 The linker searches a standard list of directories for the library,
3175 which is actually a file named @file{lib@var{library}.a}. The linker
3176 then uses this file as if it had been specified precisely by name.
3178 The directories searched include several standard system directories
3179 plus any that you specify with @samp{-L}.
3181 Normally the files found this way are library files---archive files
3182 whose members are object files. The linker handles an archive file by
3183 scanning through it for members which define symbols that have so far
3184 been referenced but not defined. But if the file that is found is an
3185 ordinary object file, it is linked in the usual fashion. The only
3186 difference between using an @samp{-l} option and specifying a file name
3187 is that @samp{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
3188 and searches several directories.
3191 You need this special case of the @samp{-l} option in order to
3192 link an Objective C program.
3195 Do not use the standard system startup files when linking.
3196 The standard system libraries are used normally, unless @code{-nostdlib}
3197 or @code{-nodefaultlibs} is used.
3199 @item -nodefaultlibs
3200 Do not use the standard system libraries when linking.
3201 Only the libraries you specify will be passed to the linker.
3202 The standard startup files are used normally, unless @code{-nostartfiles}
3203 is used. The compiler may generate calls to memcmp, memset, and memcpy
3204 for System V (and ANSI C) environments or to bcopy and bzero for
3205 BSD environments. These entries are usually resolved by entries in
3206 libc. These entry points should be supplied through some other
3207 mechanism when this option is specified.
3210 Do not use the standard system startup files or libraries when linking.
3211 No startup files and only the libraries you specify will be passed to
3212 the linker. The compiler may generate calls to memcmp, memset, and memcpy
3213 for System V (and ANSI C) environments or to bcopy and bzero for
3214 BSD environments. These entries are usually resolved by entries in
3215 libc. These entry points should be supplied through some other
3216 mechanism when this option is specified.
3218 @cindex @code{-lgcc}, use with @code{-nostdlib}
3219 @cindex @code{-nostdlib} and unresolved references
3220 @cindex unresolved references and @code{-nostdlib}
3221 @cindex @code{-lgcc}, use with @code{-nodefaultlibs}
3222 @cindex @code{-nodefaultlibs} and unresolved references
3223 @cindex unresolved references and @code{-nodefaultlibs}
3224 One of the standard libraries bypassed by @samp{-nostdlib} and
3225 @samp{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
3226 that GCC uses to overcome shortcomings of particular machines, or special
3227 needs for some languages.
3229 (@xref{Interface,,Interfacing to GCC Output}, for more discussion of
3233 (@xref{Interface,,Interfacing to GCC Output,gcc.info,Porting GCC},
3234 for more discussion of @file{libgcc.a}.)
3236 In most cases, you need @file{libgcc.a} even when you want to avoid
3237 other standard libraries. In other words, when you specify @samp{-nostdlib}
3238 or @samp{-nodefaultlibs} you should usually specify @samp{-lgcc} as well.
3239 This ensures that you have no unresolved references to internal GCC
3240 library subroutines. (For example, @samp{__main}, used to ensure C++
3241 constructors will be called; @pxref{Collect2,,@code{collect2}}.)
3244 Remove all symbol table and relocation information from the executable.
3247 On systems that support dynamic linking, this prevents linking with the shared
3248 libraries. On other systems, this option has no effect.
3251 Produce a shared object which can then be linked with other objects to
3252 form an executable. Not all systems support this option. You must
3253 also specify @samp{-fpic} or @samp{-fPIC} on some systems when
3254 you specify this option.
3257 Bind references to global symbols when building a shared object. Warn
3258 about any unresolved references (unless overridden by the link editor
3259 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
3262 @item -Xlinker @var{option}
3263 Pass @var{option} as an option to the linker. You can use this to
3264 supply system-specific linker options which GCC does not know how to
3267 If you want to pass an option that takes an argument, you must use
3268 @samp{-Xlinker} twice, once for the option and once for the argument.
3269 For example, to pass @samp{-assert definitions}, you must write
3270 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
3271 @samp{-Xlinker "-assert definitions"}, because this passes the entire
3272 string as a single argument, which is not what the linker expects.
3274 @item -Wl,@var{option}
3275 Pass @var{option} as an option to the linker. If @var{option} contains
3276 commas, it is split into multiple options at the commas.
3278 @item -u @var{symbol}
3279 Pretend the symbol @var{symbol} is undefined, to force linking of
3280 library modules to define it. You can use @samp{-u} multiple times with
3281 different symbols to force loading of additional library modules.
3284 @node Directory Options
3285 @section Options for Directory Search
3286 @cindex directory options
3287 @cindex options, directory search
3290 These options specify directories to search for header files, for
3291 libraries and for parts of the compiler:
3295 Add the directory @var{dir} to the head of the list of directories to be
3296 searched for header files. This can be used to override a system header
3297 file, substituting your own version, since these directories are
3298 searched before the system header file directories. If you use more
3299 than one @samp{-I} option, the directories are scanned in left-to-right
3300 order; the standard system directories come after.
3303 Any directories you specify with @samp{-I} options before the @samp{-I-}
3304 option are searched only for the case of @samp{#include "@var{file}"};
3305 they are not searched for @samp{#include <@var{file}>}.
3307 If additional directories are specified with @samp{-I} options after
3308 the @samp{-I-}, these directories are searched for all @samp{#include}
3309 directives. (Ordinarily @emph{all} @samp{-I} directories are used
3312 In addition, the @samp{-I-} option inhibits the use of the current
3313 directory (where the current input file came from) as the first search
3314 directory for @samp{#include "@var{file}"}. There is no way to
3315 override this effect of @samp{-I-}. With @samp{-I.} you can specify
3316 searching the directory which was current when the compiler was
3317 invoked. That is not exactly the same as what the preprocessor does
3318 by default, but it is often satisfactory.
3320 @samp{-I-} does not inhibit the use of the standard system directories
3321 for header files. Thus, @samp{-I-} and @samp{-nostdinc} are
3325 Add directory @var{dir} to the list of directories to be searched
3328 @item -B@var{prefix}
3329 This option specifies where to find the executables, libraries,
3330 include files, and data files of the compiler itself.
3332 The compiler driver program runs one or more of the subprograms
3333 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
3334 @var{prefix} as a prefix for each program it tries to run, both with and
3335 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
3337 For each subprogram to be run, the compiler driver first tries the
3338 @samp{-B} prefix, if any. If that name is not found, or if @samp{-B}
3339 was not specified, the driver tries two standard prefixes, which are
3340 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc-lib/}. If neither of
3341 those results in a file name that is found, the unmodified program
3342 name is searched for using the directories specified in your
3343 @samp{PATH} environment variable.
3345 @samp{-B} prefixes that effectively specify directory names also apply
3346 to libraries in the linker, because the compiler translates these
3347 options into @samp{-L} options for the linker. They also apply to
3348 includes files in the preprocessor, because the compiler translates these
3349 options into @samp{-isystem} options for the preprocessor. In this case,
3350 the compiler appends @samp{include} to the prefix.
3352 The run-time support file @file{libgcc.a} can also be searched for using
3353 the @samp{-B} prefix, if needed. If it is not found there, the two
3354 standard prefixes above are tried, and that is all. The file is left
3355 out of the link if it is not found by those means.
3357 Another way to specify a prefix much like the @samp{-B} prefix is to use
3358 the environment variable @code{GCC_EXEC_PREFIX}. @xref{Environment
3361 @item -specs=@var{file}
3362 Process @var{file} after the compiler reads in the standard @file{specs}
3363 file, in order to override the defaults that the @file{gcc} driver
3364 program uses when determining what switches to pass to @file{cc1},
3365 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
3366 @samp{-specs=}@var{file} can be specified on the command line, and they
3367 are processed in order, from left to right.
3371 @section Specifying subprocesses and the switches to pass to them
3373 @code{GCC} is a driver program. It performs its job by invoking a
3374 sequence of other programs to do the work of compiling, assembling and
3375 linking. GCC interprets its command-line parameters and uses these to
3376 deduce which programs it should invoke, and which command-line options
3377 it ought to place on their command lines. This behaviour is controlled
3378 by @dfn{spec strings}. In most cases there is one spec string for each
3379 program that GCC can invoke, but a few programs have multiple spec
3380 strings to control their behaviour. The spec strings built into GCC can
3381 be overridden by using the @samp{-specs=} command-line switch to specify
3384 @dfn{Spec files} are plaintext files that are used to construct spec
3385 strings. They consist of a sequence of directives separated by blank
3386 lines. The type of directive is determined by the first non-whitespace
3387 character on the line and it can be one of the following:
3390 @item %@var{command}
3391 Issues a @var{command} to the spec file processor. The commands that can
3395 @item %include <@var{file}>
3397 Search for @var{file} and insert its text at the current point in the
3400 @item %include_noerr <@var{file}>
3401 @cindex %include_noerr
3402 Just like @samp{%include}, but do not generate an error message if the include
3403 file cannot be found.
3405 @item %rename @var{old_name} @var{new_name}
3407 Rename the spec string @var{old_name} to @var{new_name}.
3411 @item *[@var{spec_name}]:
3412 This tells the compiler to create, override or delete the named spec
3413 string. All lines after this directive up to the next directive or
3414 blank line are considered to be the text for the spec string. If this
3415 results in an empty string then the spec will be deleted. (Or, if the
3416 spec did not exist, then nothing will happened.) Otherwise, if the spec
3417 does not currently exist a new spec will be created. If the spec does
3418 exist then its contents will be overridden by the text of this
3419 directive, unless the first character of that text is the @samp{+}
3420 character, in which case the text will be appended to the spec.
3422 @item [@var{suffix}]:
3423 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
3424 and up to the next directive or blank line are considered to make up the
3425 spec string for the indicated suffix. When the compiler encounters an
3426 input file with the named suffix, it will processes the spec string in
3427 order to work out how to compile that file. For example:
3434 This says that any input file whose name ends in @samp{.ZZ} should be
3435 passed to the program @samp{z-compile}, which should be invoked with the
3436 command-line switch @samp{-input} and with the result of performing the
3437 @samp{%i} substitution. (See below.)
3439 As an alternative to providing a spec string, the text that follows a
3440 suffix directive can be one of the following:
3443 @item @@@var{language}
3444 This says that the suffix is an alias for a known @var{language}. This is
3445 similar to using the @code{-x} command-line switch to GCC to specify a
3446 language explicitly. For example:
3453 Says that .ZZ files are, in fact, C++ source files.
3456 This causes an error messages saying:
3459 @var{name} compiler not installed on this system.
3463 GCC already has an extensive list of suffixes built into it.
3464 This directive will add an entry to the end of the list of suffixes, but
3465 since the list is searched from the end backwards, it is effectively
3466 possible to override earlier entries using this technique.
3470 GCC has the following spec strings built into it. Spec files can
3471 override these strings or create their own. Note that individual
3472 targets can also add their own spec strings to this list.
3475 asm Options to pass to the assembler
3476 asm_final Options to pass to the assembler post-processor
3477 cpp Options to pass to the C preprocessor
3478 cc1 Options to pass to the C compiler
3479 cc1plus Options to pass to the C++ compiler
3480 endfile Object files to include at the end of the link
3481 link Options to pass to the linker
3482 lib Libraries to include on the command line to the linker
3483 libgcc Decides which GCC support library to pass to the linker
3484 linker Sets the name of the linker
3485 predefines Defines to be passed to the C preprocessor
3486 signed_char Defines to pass to CPP to say whether @code{char} is signed by default
3487 startfile Object files to include at the start of the link
3490 Here is a small example of a spec file:
3496 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
3499 This example renames the spec called @samp{lib} to @samp{old_lib} and
3500 then overrides the previous definition of @samp{lib} with a new one.
3501 The new definition adds in some extra command-line options before
3502 including the text of the old definition.
3504 @dfn{Spec strings} are a list of command-line options to be passed to their
3505 corresponding program. In addition, the spec strings can contain
3506 @samp{%}-prefixed sequences to substitute variable text or to
3507 conditionally insert text into the command line. Using these constructs
3508 it is possible to generate quite complex command lines.
3510 Here is a table of all defined @samp{%}-sequences for spec
3511 strings. Note that spaces are not generated automatically around the
3512 results of expanding these sequences. Therefore you can concatenate them
3513 together or combine them with constant text in a single argument.
3517 Substitute one @samp{%} into the program name or argument.
3520 Substitute the name of the input file being processed.
3523 Substitute the basename of the input file being processed.
3524 This is the substring up to (and not including) the last period
3525 and not including the directory.
3528 Marks the argument containing or following the @samp{%d} as a
3529 temporary file name, so that that file will be deleted if GCC exits
3530 successfully. Unlike @samp{%g}, this contributes no text to the
3533 @item %g@var{suffix}
3534 Substitute a file name that has suffix @var{suffix} and is chosen
3535 once per compilation, and mark the argument in the same way as
3536 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
3537 name is now chosen in a way that is hard to predict even when previously
3538 chosen file names are known. For example, @samp{%g.s ... %g.o ... %g.s}
3539 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
3540 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
3541 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
3542 was simply substituted with a file name chosen once per compilation,
3543 without regard to any appended suffix (which was therefore treated
3544 just like ordinary text), making such attacks more likely to succeed.
3546 @item %u@var{suffix}
3547 Like @samp{%g}, but generates a new temporary file name even if
3548 @samp{%u@var{suffix}} was already seen.
3550 @item %U@var{suffix}
3551 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
3552 new one if there is no such last file name. In the absence of any
3553 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
3554 the same suffix @emph{space}, so @samp{%g.s ... %U.s ... %g.s ... %U.s}
3555 would involve the generation of two distinct file names, one
3556 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
3557 simply substituted with a file name chosen for the previous @samp{%u},
3558 without regard to any appended suffix.
3561 Marks the argument containing or following the @samp{%w} as the
3562 designated output file of this compilation. This puts the argument
3563 into the sequence of arguments that @samp{%o} will substitute later.
3566 Substitutes the names of all the output files, with spaces
3567 automatically placed around them. You should write spaces
3568 around the @samp{%o} as well or the results are undefined.
3569 @samp{%o} is for use in the specs for running the linker.
3570 Input files whose names have no recognized suffix are not compiled
3571 at all, but they are included among the output files, so they will
3575 Substitutes the suffix for object files. Note that this is
3576 handled specially when it immediately follows @samp{%g, %u, or %U},
3577 because of the need for those to form complete file names. The
3578 handling is such that @samp{%O} is treated exactly as if it had already
3579 been substituted, except that @samp{%g, %u, and %U} do not currently
3580 support additional @var{suffix} characters following @samp{%O} as they would
3581 following, for example, @samp{.o}.
3584 Substitutes the standard macro predefinitions for the
3585 current target machine. Use this when running @code{cpp}.
3588 Like @samp{%p}, but puts @samp{__} before and after the name of each
3589 predefined macro, except for macros that start with @samp{__} or with
3590 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ANSI
3594 Substitute a @samp{-iprefix} option made from GCC_EXEC_PREFIX.
3597 Current argument is the name of a library or startup file of some sort.
3598 Search for that file in a standard list of directories and substitute
3599 the full name found.
3602 Print @var{str} as an error message. @var{str} is terminated by a newline.
3603 Use this when inconsistent options are detected.
3606 Output @samp{-} if the input for the current command is coming from a pipe.
3609 Substitute the contents of spec string @var{name} at this point.
3612 Like @samp{%(...)} but put @samp{__} around @samp{-D} arguments.
3614 @item %x@{@var{option}@}
3615 Accumulate an option for @samp{%X}.
3618 Output the accumulated linker options specified by @samp{-Wl} or a @samp{%x}
3622 Output the accumulated assembler options specified by @samp{-Wa}.
3625 Output the accumulated preprocessor options specified by @samp{-Wp}.
3628 Substitute the major version number of GCC.
3629 (For version 2.9.5, this is 2.)
3632 Substitute the minor version number of GCC.
3633 (For version 2.9.5, this is 9.)
3636 Process the @code{asm} spec. This is used to compute the
3637 switches to be passed to the assembler.
3640 Process the @code{asm_final} spec. This is a spec string for
3641 passing switches to an assembler post-processor, if such a program is
3645 Process the @code{link} spec. This is the spec for computing the
3646 command line passed to the linker. Typically it will make use of the
3647 @samp{%L %G %S %D and %E} sequences.
3650 Dump out a @samp{-L} option for each directory that GCC believes might
3651 contain startup files. If the target supports multilibs then the
3652 current multilib directory will be prepended to each of these paths.
3655 Process the @code{lib} spec. This is a spec string for deciding which
3656 libraries should be included on the command line to the linker.
3659 Process the @code{libgcc} spec. This is a spec string for deciding
3660 which GCC support library should be included on the command line to the linker.
3663 Process the @code{startfile} spec. This is a spec for deciding which
3664 object files should be the first ones passed to the linker. Typically
3665 this might be a file named @file{crt0.o}.
3668 Process the @code{endfile} spec. This is a spec string that specifies
3669 the last object files that will be passed to the linker.
3672 Process the @code{cpp} spec. This is used to construct the arguments
3673 to be passed to the C preprocessor.
3676 Process the @code{signed_char} spec. This is intended to be used
3677 to tell cpp whether a char is signed. It typically has the definition:
3679 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
3683 Process the @code{cc1} spec. This is used to construct the options to be
3684 passed to the actual C compiler (@samp{cc1}).
3687 Process the @code{cc1plus} spec. This is used to construct the options to be
3688 passed to the actual C++ compiler (@samp{cc1plus}).
3691 Substitute the variable part of a matched option. See below.
3692 Note that each comma in the substituted string is replaced by
3696 Substitutes the @code{-S} switch, if that switch was given to GCC.
3697 If that switch was not specified, this substitutes nothing. Note that
3698 the leading dash is omitted when specifying this option, and it is
3699 automatically inserted if the substitution is performed. Thus the spec
3700 string @samp{%@{foo@}} would match the command-line option @samp{-foo}
3701 and would output the command line option @samp{-foo}.
3703 @item %W@{@code{S}@}
3704 Like %@{@code{S}@} but mark last argument supplied within as a file to be
3707 @item %@{@code{S}*@}
3708 Substitutes all the switches specified to GCC whose names start
3709 with @code{-S}, but which also take an argument. This is used for
3710 switches like @samp{-o, -D, -I}, etc. GCC considers @samp{-o foo} as being
3711 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
3712 text, including the space. Thus two arguments would be generated.
3714 @item %@{^@code{S}*@}
3715 Like %@{@code{S}*@}, but don't put a blank between a switch and its
3716 argument. Thus %@{^o*@} would only generate one argument, not two.
3718 @item %@{<@code{S}@}
3719 Remove all occurrences of @code{-S} from the command line. Note - this
3720 command is position dependent. @samp{%} commands in the spec string
3721 before this option will see @code{-S}, @samp{%} commands in the spec
3722 string after this option will not.
3724 @item %@{@code{S}*:@code{X}@}
3725 Substitutes @code{X} if one or more switches whose names start with
3726 @code{-S} are specified to GCC. Note that the tail part of the
3727 @code{-S} option (i.e. the part matched by the @samp{*}) will be substituted
3728 for each occurrence of @samp{%*} within @code{X}.
3730 @item %@{@code{S}:@code{X}@}
3731 Substitutes @code{X}, but only if the @samp{-S} switch was given to GCC.
3733 @item %@{!@code{S}:@code{X}@}
3734 Substitutes @code{X}, but only if the @samp{-S} switch was @emph{not} given to GCC.
3736 @item %@{|@code{S}:@code{X}@}
3737 Like %@{@code{S}:@code{X}@}, but if no @code{S} switch, substitute @samp{-}.
3739 @item %@{|!@code{S}:@code{X}@}
3740 Like %@{!@code{S}:@code{X}@}, but if there is an @code{S} switch, substitute @samp{-}.
3742 @item %@{.@code{S}:@code{X}@}
3743 Substitutes @code{X}, but only if processing a file with suffix @code{S}.
3745 @item %@{!.@code{S}:@code{X}@}
3746 Substitutes @code{X}, but only if @emph{not} processing a file with suffix @code{S}.
3748 @item %@{@code{S}|@code{P}:@code{X}@}
3749 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC. This may be
3750 combined with @samp{!} and @samp{.} sequences as well, although they
3751 have a stronger binding than the @samp{|}. For example a spec string
3755 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
3758 will output the following command-line options from the following input
3759 command-line options:
3764 -d fred.c -foo -baz -boggle
3765 -d jim.d -bar -baz -boggle
3770 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or
3771 %@{!@code{S}:@code{X}@} construct may contain other nested @samp{%} constructs
3772 or spaces, or even newlines. They are processed as usual, as described
3775 The @samp{-O, -f, -m, and -W} switches are handled specifically in these
3776 constructs. If another value of @samp{-O} or the negated form of a @samp{-f, -m, or
3777 -W} switch is found later in the command line, the earlier switch
3778 value is ignored, except with @{@code{S}*@} where @code{S} is just one
3779 letter, which passes all matching options.
3781 The character @samp{|} at the beginning of the predicate text is used to indicate
3782 that a command should be piped to the following command, but only if @samp{-pipe}
3785 It is built into GCC which switches take arguments and which do not.
3786 (You might think it would be useful to generalize this to allow each
3787 compiler's spec to say which switches take arguments. But this cannot
3788 be done in a consistent fashion. GCC cannot even decide which input
3789 files have been specified without knowing which switches take arguments,
3790 and it must know which input files to compile in order to tell which
3793 GCC also knows implicitly that arguments starting in @samp{-l} are to be
3794 treated as compiler output files, and passed to the linker in their
3795 proper position among the other output files.
3797 @node Target Options
3798 @section Specifying Target Machine and Compiler Version
3799 @cindex target options
3800 @cindex cross compiling
3801 @cindex specifying machine version
3802 @cindex specifying compiler version and target machine
3803 @cindex compiler version, specifying
3804 @cindex target machine, specifying
3806 By default, GCC compiles code for the same type of machine that you
3807 are using. However, it can also be installed as a cross-compiler, to
3808 compile for some other type of machine. In fact, several different
3809 configurations of GCC, for different target machines, can be
3810 installed side by side. Then you specify which one to use with the
3813 In addition, older and newer versions of GCC can be installed side
3814 by side. One of them (probably the newest) will be the default, but
3815 you may sometimes wish to use another.
3818 @item -b @var{machine}
3819 The argument @var{machine} specifies the target machine for compilation.
3820 This is useful when you have installed GCC as a cross-compiler.
3822 The value to use for @var{machine} is the same as was specified as the
3823 machine type when configuring GCC as a cross-compiler. For
3824 example, if a cross-compiler was configured with @samp{configure
3825 i386v}, meaning to compile for an 80386 running System V, then you
3826 would specify @samp{-b i386v} to run that cross compiler.
3828 When you do not specify @samp{-b}, it normally means to compile for
3829 the same type of machine that you are using.
3831 @item -V @var{version}
3832 The argument @var{version} specifies which version of GCC to run.
3833 This is useful when multiple versions are installed. For example,
3834 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
3836 The default version, when you do not specify @samp{-V}, is the last
3837 version of GCC that you installed.
3840 The @samp{-b} and @samp{-V} options actually work by controlling part of
3841 the file name used for the executable files and libraries used for
3842 compilation. A given version of GCC, for a given target machine, is
3843 normally kept in the directory @file{/usr/local/lib/gcc-lib/@var{machine}/@var{version}}.@refill
3845 Thus, sites can customize the effect of @samp{-b} or @samp{-V} either by
3846 changing the names of these directories or adding alternate names (or
3847 symbolic links). If in directory @file{/usr/local/lib/gcc-lib/} the
3848 file @file{80386} is a link to the file @file{i386v}, then @samp{-b
3849 80386} becomes an alias for @samp{-b i386v}.
3851 In one respect, the @samp{-b} or @samp{-V} do not completely change
3852 to a different compiler: the top-level driver program @code{gcc}
3853 that you originally invoked continues to run and invoke the other
3854 executables (preprocessor, compiler per se, assembler and linker)
3855 that do the real work. However, since no real work is done in the
3856 driver program, it usually does not matter that the driver program
3857 in use is not the one for the specified target and version.
3859 The only way that the driver program depends on the target machine is
3860 in the parsing and handling of special machine-specific options.
3861 However, this is controlled by a file which is found, along with the
3862 other executables, in the directory for the specified version and
3863 target machine. As a result, a single installed driver program adapts
3864 to any specified target machine and compiler version.
3866 The driver program executable does control one significant thing,
3867 however: the default version and target machine. Therefore, you can
3868 install different instances of the driver program, compiled for
3869 different targets or versions, under different names.
3871 For example, if the driver for version 2.0 is installed as @code{ogcc}
3872 and that for version 2.1 is installed as @code{gcc}, then the command
3873 @code{gcc} will use version 2.1 by default, while @code{ogcc} will use
3874 2.0 by default. However, you can choose either version with either
3875 command with the @samp{-V} option.
3877 @node Submodel Options
3878 @section Hardware Models and Configurations
3879 @cindex submodel options
3880 @cindex specifying hardware config
3881 @cindex hardware models and configurations, specifying
3882 @cindex machine dependent options
3884 Earlier we discussed the standard option @samp{-b} which chooses among
3885 different installed compilers for completely different target
3886 machines, such as Vax vs. 68000 vs. 80386.
3888 In addition, each of these target machine types can have its own
3889 special options, starting with @samp{-m}, to choose among various
3890 hardware models or configurations---for example, 68010 vs 68020,
3891 floating coprocessor or none. A single installed version of the
3892 compiler can compile for any model or configuration, according to the
3895 Some configurations of the compiler also support additional special
3896 options, usually for compatibility with other compilers on the same
3900 These options are defined by the macro @code{TARGET_SWITCHES} in the
3901 machine description. The default for the options is also defined by
3902 that macro, which enables you to change the defaults.
3918 * RS/6000 and PowerPC Options::
3923 * Intel 960 Options::
3924 * DEC Alpha Options::
3928 * System V Options::
3929 * TMS320C3x/C4x Options::
3938 @node M680x0 Options
3939 @subsection M680x0 Options
3940 @cindex M680x0 options
3942 These are the @samp{-m} options defined for the 68000 series. The default
3943 values for these options depends on which style of 68000 was selected when
3944 the compiler was configured; the defaults for the most common choices are
3950 Generate output for a 68000. This is the default
3951 when the compiler is configured for 68000-based systems.
3953 Use this option for microcontrollers with a 68000 or EC000 core,
3954 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
3958 Generate output for a 68020. This is the default
3959 when the compiler is configured for 68020-based systems.
3962 Generate output containing 68881 instructions for floating point.
3963 This is the default for most 68020 systems unless @samp{-nfp} was
3964 specified when the compiler was configured.
3967 Generate output for a 68030. This is the default when the compiler is
3968 configured for 68030-based systems.
3971 Generate output for a 68040. This is the default when the compiler is
3972 configured for 68040-based systems.
3974 This option inhibits the use of 68881/68882 instructions that have to be
3975 emulated by software on the 68040. Use this option if your 68040 does not
3976 have code to emulate those instructions.
3979 Generate output for a 68060. This is the default when the compiler is
3980 configured for 68060-based systems.
3982 This option inhibits the use of 68020 and 68881/68882 instructions that
3983 have to be emulated by software on the 68060. Use this option if your 68060
3984 does not have code to emulate those instructions.
3987 Generate output for a CPU32. This is the default
3988 when the compiler is configured for CPU32-based systems.
3990 Use this option for microcontrollers with a
3991 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
3992 68336, 68340, 68341, 68349 and 68360.
3995 Generate output for a 520X "coldfire" family cpu. This is the default
3996 when the compiler is configured for 520X-based systems.
3998 Use this option for microcontroller with a 5200 core, including
3999 the MCF5202, MCF5203, MCF5204 and MCF5202.
4003 Generate output for a 68040, without using any of the new instructions.
4004 This results in code which can run relatively efficiently on either a
4005 68020/68881 or a 68030 or a 68040. The generated code does use the
4006 68881 instructions that are emulated on the 68040.
4009 Generate output for a 68060, without using any of the new instructions.
4010 This results in code which can run relatively efficiently on either a
4011 68020/68881 or a 68030 or a 68040. The generated code does use the
4012 68881 instructions that are emulated on the 68060.
4015 Generate output containing Sun FPA instructions for floating point.
4018 Generate output containing library calls for floating point.
4019 @strong{Warning:} the requisite libraries are not available for all m68k
4020 targets. Normally the facilities of the machine's usual C compiler are
4021 used, but this can't be done directly in cross-compilation. You must
4022 make your own arrangements to provide suitable library functions for
4023 cross-compilation. The embedded targets @samp{m68k-*-aout} and
4024 @samp{m68k-*-coff} do provide software floating point support.
4027 Consider type @code{int} to be 16 bits wide, like @code{short int}.
4030 Do not use the bit-field instructions. The @samp{-m68000}, @samp{-mcpu32}
4031 and @samp{-m5200} options imply @w{@samp{-mnobitfield}}.
4034 Do use the bit-field instructions. The @samp{-m68020} option implies
4035 @samp{-mbitfield}. This is the default if you use a configuration
4036 designed for a 68020.
4039 Use a different function-calling convention, in which functions
4040 that take a fixed number of arguments return with the @code{rtd}
4041 instruction, which pops their arguments while returning. This
4042 saves one instruction in the caller since there is no need to pop
4043 the arguments there.
4045 This calling convention is incompatible with the one normally
4046 used on Unix, so you cannot use it if you need to call libraries
4047 compiled with the Unix compiler.
4049 Also, you must provide function prototypes for all functions that
4050 take variable numbers of arguments (including @code{printf});
4051 otherwise incorrect code will be generated for calls to those
4054 In addition, seriously incorrect code will result if you call a
4055 function with too many arguments. (Normally, extra arguments are
4056 harmlessly ignored.)
4058 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
4059 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
4062 @itemx -mno-align-int
4063 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
4064 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
4065 boundary (@samp{-malign-int}) or a 16-bit boundary (@samp{-mno-align-int}).
4066 Aligning variables on 32-bit boundaries produces code that runs somewhat
4067 faster on processors with 32-bit busses at the expense of more memory.
4069 @strong{Warning:} if you use the @samp{-malign-int} switch, GCC will
4070 align structures containing the above types differently than
4071 most published application binary interface specifications for the m68k.
4074 Use the pc-relative addressing mode of the 68000 directly, instead of
4075 using a global offset table. At present, this option implies -fpic,
4076 allowing at most a 16-bit offset for pc-relative addressing. -fPIC is
4077 not presently supported with -mpcrel, though this could be supported for
4078 68020 and higher processors.
4080 @item -mno-strict-align
4081 @itemx -mstrict-align
4082 @kindex -mstrict-align
4083 Do not (do) assume that unaligned memory references will be handled by
4088 @node M68hc1x Options
4089 @subsection M68hc1x Options
4090 @cindex M68hc1x options
4092 These are the @samp{-m} options defined for the 68hc11 and 68hc12
4093 microcontrollers. The default values for these options depends on
4094 which style of microcontroller was selected when the compiler was configured;
4095 the defaults for the most common choices are given below.
4100 Generate output for a 68HC11. This is the default
4101 when the compiler is configured for 68HC11-based systems.
4105 Generate output for a 68HC12. This is the default
4106 when the compiler is configured for 68HC12-based systems.
4109 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
4113 Consider type @code{int} to be 16 bits wide, like @code{short int}.
4115 @item -msoft-reg-count=@var{count}
4116 Specify the number of pseudo-soft registers which are used for the
4117 code generation. The maximum number is 32. Using more pseudo-soft
4118 register may or may not result in better code depending on the program.
4119 The default is 4 for 68HC11 and 2 for 68HC12.
4124 @subsection VAX Options
4127 These @samp{-m} options are defined for the Vax:
4131 Do not output certain jump instructions (@code{aobleq} and so on)
4132 that the Unix assembler for the Vax cannot handle across long
4136 Do output those jump instructions, on the assumption that you
4137 will assemble with the GNU assembler.
4140 Output code for g-format floating point numbers instead of d-format.
4144 @subsection SPARC Options
4145 @cindex SPARC options
4147 These @samp{-m} switches are supported on the SPARC:
4152 Specify @samp{-mapp-regs} to generate output using the global registers
4153 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
4156 To be fully SVR4 ABI compliant at the cost of some performance loss,
4157 specify @samp{-mno-app-regs}. You should compile libraries and system
4158 software with this option.
4162 Generate output containing floating point instructions. This is the
4167 Generate output containing library calls for floating point.
4168 @strong{Warning:} the requisite libraries are not available for all SPARC
4169 targets. Normally the facilities of the machine's usual C compiler are
4170 used, but this cannot be done directly in cross-compilation. You must make
4171 your own arrangements to provide suitable library functions for
4172 cross-compilation. The embedded targets @samp{sparc-*-aout} and
4173 @samp{sparclite-*-*} do provide software floating point support.
4175 @samp{-msoft-float} changes the calling convention in the output file;
4176 therefore, it is only useful if you compile @emph{all} of a program with
4177 this option. In particular, you need to compile @file{libgcc.a}, the
4178 library that comes with GCC, with @samp{-msoft-float} in order for
4181 @item -mhard-quad-float
4182 Generate output containing quad-word (long double) floating point
4185 @item -msoft-quad-float
4186 Generate output containing library calls for quad-word (long double)
4187 floating point instructions. The functions called are those specified
4188 in the SPARC ABI. This is the default.
4190 As of this writing, there are no sparc implementations that have hardware
4191 support for the quad-word floating point instructions. They all invoke
4192 a trap handler for one of these instructions, and then the trap handler
4193 emulates the effect of the instruction. Because of the trap handler overhead,
4194 this is much slower than calling the ABI library routines. Thus the
4195 @samp{-msoft-quad-float} option is the default.
4199 With @samp{-mepilogue} (the default), the compiler always emits code for
4200 function exit at the end of each function. Any function exit in
4201 the middle of the function (such as a return statement in C) will
4202 generate a jump to the exit code at the end of the function.
4204 With @samp{-mno-epilogue}, the compiler tries to emit exit code inline
4205 at every function exit.
4209 With @samp{-mflat}, the compiler does not generate save/restore instructions
4210 and will use a "flat" or single register window calling convention.
4211 This model uses %i7 as the frame pointer and is compatible with the normal
4212 register window model. Code from either may be intermixed.
4213 The local registers and the input registers (0-5) are still treated as
4214 "call saved" registers and will be saved on the stack as necessary.
4216 With @samp{-mno-flat} (the default), the compiler emits save/restore
4217 instructions (except for leaf functions) and is the normal mode of operation.
4219 @item -mno-unaligned-doubles
4220 @itemx -munaligned-doubles
4221 Assume that doubles have 8 byte alignment. This is the default.
4223 With @samp{-munaligned-doubles}, GCC assumes that doubles have 8 byte
4224 alignment only if they are contained in another type, or if they have an
4225 absolute address. Otherwise, it assumes they have 4 byte alignment.
4226 Specifying this option avoids some rare compatibility problems with code
4227 generated by other compilers. It is not the default because it results
4228 in a performance loss, especially for floating point code.
4230 @item -mno-faster-structs
4231 @itemx -mfaster-structs
4232 With @samp{-mfaster-structs}, the compiler assumes that structures
4233 should have 8 byte alignment. This enables the use of pairs of
4234 @code{ldd} and @code{std} instructions for copies in structure
4235 assignment, in place of twice as many @code{ld} and @code{st} pairs.
4236 However, the use of this changed alignment directly violates the Sparc
4237 ABI. Thus, it's intended only for use on targets where the developer
4238 acknowledges that their resulting code will not be directly in line with
4239 the rules of the ABI.
4243 These two options select variations on the SPARC architecture.
4245 By default (unless specifically configured for the Fujitsu SPARClite),
4246 GCC generates code for the v7 variant of the SPARC architecture.
4248 @samp{-mv8} will give you SPARC v8 code. The only difference from v7
4249 code is that the compiler emits the integer multiply and integer
4250 divide instructions which exist in SPARC v8 but not in SPARC v7.
4252 @samp{-msparclite} will give you SPARClite code. This adds the integer
4253 multiply, integer divide step and scan (@code{ffs}) instructions which
4254 exist in SPARClite but not in SPARC v7.
4256 These options are deprecated and will be deleted in a future GCC release.
4257 They have been replaced with @samp{-mcpu=xxx}.
4261 These two options select the processor for which the code is optimised.
4263 With @samp{-mcypress} (the default), the compiler optimizes code for the
4264 Cypress CY7C602 chip, as used in the SparcStation/SparcServer 3xx series.
4265 This is also appropriate for the older SparcStation 1, 2, IPX etc.
4267 With @samp{-msupersparc} the compiler optimizes code for the SuperSparc cpu, as
4268 used in the SparcStation 10, 1000 and 2000 series. This flag also enables use
4269 of the full SPARC v8 instruction set.
4271 These options are deprecated and will be deleted in a future GCC release.
4272 They have been replaced with @samp{-mcpu=xxx}.
4274 @item -mcpu=@var{cpu_type}
4275 Set the instruction set, register set, and instruction scheduling parameters
4276 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
4277 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
4278 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
4279 @samp{sparclet}, @samp{tsc701}, @samp{v9}, and @samp{ultrasparc}.
4281 Default instruction scheduling parameters are used for values that select
4282 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
4283 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
4285 Here is a list of each supported architecture and their supported
4290 v8: supersparc, hypersparc
4291 sparclite: f930, f934, sparclite86x
4296 @item -mtune=@var{cpu_type}
4297 Set the instruction scheduling parameters for machine type
4298 @var{cpu_type}, but do not set the instruction set or register set that the
4299 option @samp{-mcpu=}@var{cpu_type} would.
4301 The same values for @samp{-mcpu=}@var{cpu_type} are used for
4302 @samp{-mtune=}@*@var{cpu_type}, though the only useful values are those that
4303 select a particular cpu implementation: @samp{cypress}, @samp{supersparc},
4304 @samp{hypersparc}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
4305 @samp{tsc701}, @samp{ultrasparc}.
4309 These @samp{-m} switches are supported in addition to the above
4310 on the SPARCLET processor.
4313 @item -mlittle-endian
4314 Generate code for a processor running in little-endian mode.
4317 Treat register @code{%g0} as a normal register.
4318 GCC will continue to clobber it as necessary but will not assume
4319 it always reads as 0.
4321 @item -mbroken-saverestore
4322 Generate code that does not use non-trivial forms of the @code{save} and
4323 @code{restore} instructions. Early versions of the SPARCLET processor do
4324 not correctly handle @code{save} and @code{restore} instructions used with
4325 arguments. They correctly handle them used without arguments. A @code{save}
4326 instruction used without arguments increments the current window pointer
4327 but does not allocate a new stack frame. It is assumed that the window
4328 overflow trap handler will properly handle this case as will interrupt
4332 These @samp{-m} switches are supported in addition to the above
4333 on SPARC V9 processors in 64 bit environments.
4336 @item -mlittle-endian
4337 Generate code for a processor running in little-endian mode.
4341 Generate code for a 32 bit or 64 bit environment.
4342 The 32 bit environment sets int, long and pointer to 32 bits.
4343 The 64 bit environment sets int to 32 bits and long and pointer
4346 @item -mcmodel=medlow
4347 Generate code for the Medium/Low code model: the program must be linked
4348 in the low 32 bits of the address space. Pointers are 64 bits.
4349 Programs can be statically or dynamically linked.
4351 @item -mcmodel=medmid
4352 Generate code for the Medium/Middle code model: the program must be linked
4353 in the low 44 bits of the address space, the text segment must be less than
4354 2G bytes, and data segment must be within 2G of the text segment.
4355 Pointers are 64 bits.
4357 @item -mcmodel=medany
4358 Generate code for the Medium/Anywhere code model: the program may be linked
4359 anywhere in the address space, the text segment must be less than
4360 2G bytes, and data segment must be within 2G of the text segment.
4361 Pointers are 64 bits.
4363 @item -mcmodel=embmedany
4364 Generate code for the Medium/Anywhere code model for embedded systems:
4365 assume a 32 bit text and a 32 bit data segment, both starting anywhere
4366 (determined at link time). Register %g4 points to the base of the
4367 data segment. Pointers still 64 bits.
4368 Programs are statically linked, PIC is not supported.
4371 @itemx -mno-stack-bias
4372 With @samp{-mstack-bias}, GCC assumes that the stack pointer, and
4373 frame pointer if present, are offset by -2047 which must be added back
4374 when making stack frame references.
4375 Otherwise, assume no such offset is present.
4378 @node Convex Options
4379 @subsection Convex Options
4380 @cindex Convex options
4382 These @samp{-m} options are defined for Convex:
4386 Generate output for C1. The code will run on any Convex machine.
4387 The preprocessor symbol @code{__convex__c1__} is defined.
4390 Generate output for C2. Uses instructions not available on C1.
4391 Scheduling and other optimizations are chosen for max performance on C2.
4392 The preprocessor symbol @code{__convex_c2__} is defined.
4395 Generate output for C32xx. Uses instructions not available on C1.
4396 Scheduling and other optimizations are chosen for max performance on C32.
4397 The preprocessor symbol @code{__convex_c32__} is defined.
4400 Generate output for C34xx. Uses instructions not available on C1.
4401 Scheduling and other optimizations are chosen for max performance on C34.
4402 The preprocessor symbol @code{__convex_c34__} is defined.
4405 Generate output for C38xx. Uses instructions not available on C1.
4406 Scheduling and other optimizations are chosen for max performance on C38.
4407 The preprocessor symbol @code{__convex_c38__} is defined.
4410 Generate code which puts an argument count in the word preceding each
4411 argument list. This is compatible with regular CC, and a few programs
4412 may need the argument count word. GDB and other source-level debuggers
4413 do not need it; this info is in the symbol table.
4416 Omit the argument count word. This is the default.
4418 @item -mvolatile-cache
4419 Allow volatile references to be cached. This is the default.
4421 @item -mvolatile-nocache
4422 Volatile references bypass the data cache, going all the way to memory.
4423 This is only needed for multi-processor code that does not use standard
4424 synchronization instructions. Making non-volatile references to volatile
4425 locations will not necessarily work.
4428 Type long is 32 bits, the same as type int. This is the default.
4431 Type long is 64 bits, the same as type long long. This option is useless,
4432 because no library support exists for it.
4435 @node AMD29K Options
4436 @subsection AMD29K Options
4437 @cindex AMD29K options
4439 These @samp{-m} options are defined for the AMD Am29000:
4444 @cindex DW bit (29k)
4445 Generate code that assumes the @code{DW} bit is set, i.e., that byte and
4446 halfword operations are directly supported by the hardware. This is the
4451 Generate code that assumes the @code{DW} bit is not set.
4455 @cindex byte writes (29k)
4456 Generate code that assumes the system supports byte and halfword write
4457 operations. This is the default.
4461 Generate code that assumes the systems does not support byte and
4462 halfword write operations. @samp{-mnbw} implies @samp{-mndw}.
4466 @cindex memory model (29k)
4467 Use a small memory model that assumes that all function addresses are
4468 either within a single 256 KB segment or at an absolute address of less
4469 than 256k. This allows the @code{call} instruction to be used instead
4470 of a @code{const}, @code{consth}, @code{calli} sequence.
4474 Use the normal memory model: Generate @code{call} instructions only when
4475 calling functions in the same file and @code{calli} instructions
4476 otherwise. This works if each file occupies less than 256 KB but allows
4477 the entire executable to be larger than 256 KB. This is the default.
4480 Always use @code{calli} instructions. Specify this option if you expect
4481 a single file to compile into more than 256 KB of code.
4485 @cindex processor selection (29k)
4486 Generate code for the Am29050.
4490 Generate code for the Am29000. This is the default.
4492 @item -mkernel-registers
4493 @kindex -mkernel-registers
4494 @cindex kernel and user registers (29k)
4495 Generate references to registers @code{gr64-gr95} instead of to
4496 registers @code{gr96-gr127}. This option can be used when compiling
4497 kernel code that wants a set of global registers disjoint from that used
4500 Note that when this option is used, register names in @samp{-f} flags
4501 must use the normal, user-mode, names.
4503 @item -muser-registers
4504 @kindex -muser-registers
4505 Use the normal set of global registers, @code{gr96-gr127}. This is the
4509 @itemx -mno-stack-check
4510 @kindex -mstack-check
4511 @cindex stack checks (29k)
4512 Insert (or do not insert) a call to @code{__msp_check} after each stack
4513 adjustment. This is often used for kernel code.
4516 @itemx -mno-storem-bug
4517 @kindex -mstorem-bug
4518 @cindex storem bug (29k)
4519 @samp{-mstorem-bug} handles 29k processors which cannot handle the
4520 separation of a mtsrim insn and a storem instruction (most 29000 chips
4521 to date, but not the 29050).
4523 @item -mno-reuse-arg-regs
4524 @itemx -mreuse-arg-regs
4525 @kindex -mreuse-arg-regs
4526 @samp{-mno-reuse-arg-regs} tells the compiler to only use incoming argument
4527 registers for copying out arguments. This helps detect calling a function
4528 with fewer arguments than it was declared with.
4530 @item -mno-impure-text
4531 @itemx -mimpure-text
4532 @kindex -mimpure-text
4533 @samp{-mimpure-text}, used in addition to @samp{-shared}, tells the compiler to
4534 not pass @samp{-assert pure-text} to the linker when linking a shared object.
4537 @kindex -msoft-float
4538 Generate output containing library calls for floating point.
4539 @strong{Warning:} the requisite libraries are not part of GCC.
4540 Normally the facilities of the machine's usual C compiler are used, but
4541 this can't be done directly in cross-compilation. You must make your
4542 own arrangements to provide suitable library functions for
4547 Do not generate multm or multmu instructions. This is useful for some embedded
4548 systems which do not have trap handlers for these instructions.
4552 @subsection ARM Options
4555 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
4560 @kindex -mapcs-frame
4561 Generate a stack frame that is compliant with the ARM Procedure Call
4562 Standard for all functions, even if this is not strictly necessary for
4563 correct execution of the code. Specifying @samp{-fomit-frame-pointer}
4564 with this option will cause the stack frames not to be generated for
4565 leaf functions. The default is @samp{-mno-apcs-frame}.
4569 This is a synonym for @samp{-mapcs-frame}.
4573 Generate code for a processor running with a 26-bit program counter,
4574 and conforming to the function calling standards for the APCS 26-bit
4575 option. This option replaces the @samp{-m2} and @samp{-m3} options
4576 of previous releases of the compiler.
4580 Generate code for a processor running with a 32-bit program counter,
4581 and conforming to the function calling standards for the APCS 32-bit
4582 option. This option replaces the @samp{-m6} option of previous releases
4585 @item -mapcs-stack-check
4586 @kindex -mapcs-stack-check
4587 @kindex -mno-apcs-stack-check
4588 Generate code to check the amount of stack space available upon entry to
4589 every function (that actually uses some stack space). If there is
4590 insufficient space available then either the function
4591 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
4592 called, depending upon the amount of stack space required. The run time
4593 system is required to provide these functions. The default is
4594 @samp{-mno-apcs-stack-check}, since this produces smaller code.
4597 @kindex -mapcs-float
4598 @kindex -mno-apcs-float
4599 Pass floating point arguments using the float point registers. This is
4600 one of the variants of the APCS. This option is recommended if the
4601 target hardware has a floating point unit or if a lot of floating point
4602 arithmetic is going to be performed by the code. The default is
4603 @samp{-mno-apcs-float}, since integer only code is slightly increased in
4604 size if @samp{-mapcs-float} is used.
4606 @item -mapcs-reentrant
4607 @kindex -mapcs-reentrant
4608 @kindex -mno-apcs-reentrant
4609 Generate reentrant, position independent code. This is the equivalent
4610 to specifying the @samp{-fpic} option. The default is
4611 @samp{-mno-apcs-reentrant}.
4613 @item -mthumb-interwork
4614 @kindex -mthumb-interwork
4615 @kindex -mno-thumb-interwork
4616 Generate code which supports calling between the ARM and THUMB
4617 instruction sets. Without this option the two instruction sets cannot
4618 be reliably used inside one program. The default is
4619 @samp{-mno-thumb-interwork}, since slightly larger code is generated
4620 when @samp{-mthumb-interwork} is specified.
4622 @item -mno-sched-prolog
4623 @kindex -mno-sched-prolog
4624 @kindex -msched-prolog
4625 Prevent the reordering of instructions in the function prolog, or the
4626 merging of those instruction with the instructions in the function's
4627 body. This means that all functions will start with a recognizable set
4628 of instructions (or in fact one of a choice from a small set of
4629 different function prologues), and this information can be used to
4630 locate the start if functions inside an executable piece of code. The
4631 default is @samp{-msched-prolog}.
4634 Generate output containing floating point instructions. This is the
4638 Generate output containing library calls for floating point.
4639 @strong{Warning:} the requisite libraries are not available for all ARM
4640 targets. Normally the facilities of the machine's usual C compiler are
4641 used, but this cannot be done directly in cross-compilation. You must make
4642 your own arrangements to provide suitable library functions for
4645 @samp{-msoft-float} changes the calling convention in the output file;
4646 therefore, it is only useful if you compile @emph{all} of a program with
4647 this option. In particular, you need to compile @file{libgcc.a}, the
4648 library that comes with GCC, with @samp{-msoft-float} in order for
4651 @item -mlittle-endian
4652 Generate code for a processor running in little-endian mode. This is
4653 the default for all standard configurations.
4656 Generate code for a processor running in big-endian mode; the default is
4657 to compile code for a little-endian processor.
4659 @item -mwords-little-endian
4660 This option only applies when generating code for big-endian processors.
4661 Generate code for a little-endian word order but a big-endian byte
4662 order. That is, a byte order of the form @samp{32107654}. Note: this
4663 option should only be used if you require compatibility with code for
4664 big-endian ARM processors generated by versions of the compiler prior to
4667 @item -malignment-traps
4668 @kindex -malignment-traps
4669 Generate code that will not trap if the MMU has alignment traps enabled.
4670 On ARM architectures prior to ARMv4, there were no instructions to
4671 access half-word objects stored in memory. However, when reading from
4672 memory a feature of the ARM architecture allows a word load to be used,
4673 even if the address is unaligned, and the processor core will rotate the
4674 data as it is being loaded. This option tells the compiler that such
4675 misaligned accesses will cause a MMU trap and that it should instead
4676 synthesise the access as a series of byte accesses. The compiler can
4677 still use word accesses to load half-word data if it knows that the
4678 address is aligned to a word boundary.
4680 This option is ignored when compiling for ARM architecture 4 or later,
4681 since these processors have instructions to directly access half-word
4684 @item -mno-alignment-traps
4685 @kindex -mno-alignment-traps
4686 Generate code that assumes that the MMU will not trap unaligned
4687 accesses. This produces better code when the target instruction set
4688 does not have half-word memory operations (implementations prior to
4691 Note that you cannot use this option to access unaligned word objects,
4692 since the processor will only fetch one 32-bit aligned object from
4695 The default setting for most targets is -mno-alignment-traps, since
4696 this produces better code when there are no half-word memory
4697 instructions available.
4699 @item -mshort-load-bytes
4700 @kindex -mshort-load-bytes
4701 This is a depreciated alias for @samp{-malignment-traps}.
4703 @item -mno-short-load-bytes
4704 @kindex -mno-short-load-bytes
4705 This is a depreciated alias for @samp{-mno-alignment-traps}.
4707 @item -mshort-load-words
4708 @kindex -mshort-load-words
4709 This is a depreciated alias for @samp{-mno-alignment-traps}.
4711 @item -mno-short-load-words
4712 @kindex -mno-short-load-words
4713 This is a depreciated alias for @samp{-malignment-traps}.
4717 This option only applies to RISC iX. Emulate the native BSD-mode
4718 compiler. This is the default if @samp{-ansi} is not specified.
4722 This option only applies to RISC iX. Emulate the native X/Open-mode
4725 @item -mno-symrename
4726 @kindex -mno-symrename
4727 This option only applies to RISC iX. Do not run the assembler
4728 post-processor, @samp{symrename}, after code has been assembled.
4729 Normally it is necessary to modify some of the standard symbols in
4730 preparation for linking with the RISC iX C library; this option
4731 suppresses this pass. The post-processor is never run when the
4732 compiler is built for cross-compilation.
4736 This specifies the name of the target ARM processor. GCC uses this name
4737 to determine what kind of instructions it can use when generating
4738 assembly code. Permissible names are: arm2, arm250, arm3, arm6, arm60,
4739 arm600, arm610, arm620, arm7, arm7m, arm7d, arm7dm, arm7di, arm7dmi,
4740 arm70, arm700, arm700i, arm710, arm710c, arm7100, arm7500, arm7500fe,
4741 arm7tdmi, arm8, strongarm, strongarm110, strongarm1100, arm8, arm810,
4742 arm9, arm920, arm920t, arm9tdmi.
4744 @itemx -mtune=<name>
4746 This option is very similar to the @samp{-mcpu=} option, except that
4747 instead of specifying the actual target processor type, and hence
4748 restricting which instructions can be used, it specifies that GCC should
4749 tune the performance of the code as if the target were of the type
4750 specified in this option, but still choosing the instructions that it
4751 will generate based on the cpu specified by a @samp{-mcpu=} option.
4752 For some arm implementations better performance can be obtained by using
4757 This specifies the name of the target ARM architecture. GCC uses this
4758 name to determine what kind of instructions it can use when generating
4759 assembly code. This option can be used in conjunction with or instead
4760 of the @samp{-mcpu=} option. Permissible names are: armv2, armv2a,
4761 armv3, armv3m, armv4, armv4t, armv5.
4763 @item -mfpe=<number>
4764 @itemx -mfp=<number>
4767 This specifies the version of the floating point emulation available on
4768 the target. Permissible values are 2 and 3. @samp{-mfp=} is a synonym
4769 for @samp{-mfpe=} to support older versions of GCC.
4771 @item -mstructure-size-boundary=<n>
4772 @kindex -mstructure-size-boundary
4773 The size of all structures and unions will be rounded up to a multiple
4774 of the number of bits set by this option. Permissible values are 8 and
4775 32. The default value varies for different toolchains. For the COFF
4776 targeted toolchain the default value is 8. Specifying the larger number
4777 can produce faster, more efficient code, but can also increase the size
4778 of the program. The two values are potentially incompatible. Code
4779 compiled with one value cannot necessarily expect to work with code or
4780 libraries compiled with the other value, if they exchange information
4781 using structures or unions. Programmers are encouraged to use the 32
4782 value as future versions of the toolchain may default to this value.
4784 @item -mabort-on-noreturn
4785 @kindex -mabort-on-noreturn
4786 @kindex -mnoabort-on-noreturn
4787 Generate a call to the function abort at the end of a noreturn function.
4788 It will be executed if the function tries to return.
4791 @itemx -mno-long-calls
4792 Tells the compiler to perform function calls by first loading the
4793 address of the function into a register and then performing a subroutine
4794 call on this register. This switch is needed if the target function
4795 will lie outside of the 64 megabyte addressing range of the offset based
4796 version of subroutine call instruction.
4798 Even if this switch is enabled, not all function calls will be turned
4799 into long calls. The heuristic is that static functions, functions
4800 which have the @samp{short-call} attribute, functions that are inside
4801 the scope of a @samp{#pragma no_long_calls} directive and functions whose
4802 definitions have already been compiled within the current compilation
4803 unit, will not be turned into long calls. The exception to this rule is
4804 that weak function definitions, functions with the @samp{long-call}
4805 attribute or the @samp{section} attribute, and functions that are within
4806 the scope of a @samp{#pragma long_calls} directive, will always be
4807 turned into long calls.
4809 This feature is not enabled by default. Specifying
4810 @samp{--no-long-calls} will restore the default behaviour, as will
4811 placing the function calls within the scope of a @samp{#pragma
4812 long_calls_off} directive. Note these switches have no effect on how
4813 the compiler generates code to handle function calls via function
4816 @item -mnop-fun-dllimport
4817 @kindex -mnop-fun-dllimport
4818 Disable the support for the @emph{dllimport} attribute.
4820 @item -msingle-pic-base
4821 @kindex -msingle-pic-base
4822 Treat the register used for PIC addressing as read-only, rather than
4823 loading it in the prologue for each function. The run-time system is
4824 responsible for initialising this register with an appropriate value
4825 before execution begins.
4827 @item -mpic-register=<reg>
4828 @kindex -mpic-register=
4829 Specify the register to be used for PIC addressing. The default is R10
4830 unless stack-checking is enabled, when R9 is used.
4835 @subsection Thumb Options
4836 @cindex Thumb Options
4840 @item -mthumb-interwork
4841 @kindex -mthumb-interwork
4842 @kindex -mno-thumb-interwork
4843 Generate code which supports calling between the THUMB and ARM
4844 instruction sets. Without this option the two instruction sets cannot
4845 be reliably used inside one program. The default is
4846 @samp{-mno-thumb-interwork}, since slightly smaller code is generated
4850 @kindex -mtpcs-frame
4851 @kindex -mno-tpcs-frame
4852 Generate a stack frame that is compliant with the Thumb Procedure Call
4853 Standard for all non-leaf functions. (A leaf function is one that does
4854 not call any other functions). The default is @samp{-mno-apcs-frame}.
4856 @item -mtpcs-leaf-frame
4857 @kindex -mtpcs-leaf-frame
4858 @kindex -mno-tpcs-leaf-frame
4859 Generate a stack frame that is compliant with the Thumb Procedure Call
4860 Standard for all leaf functions. (A leaf function is one that does
4861 not call any other functions). The default is @samp{-mno-apcs-leaf-frame}.
4863 @item -mlittle-endian
4864 @kindex -mlittle-endian
4865 Generate code for a processor running in little-endian mode. This is
4866 the default for all standard configurations.
4869 @kindex -mbig-endian
4870 Generate code for a processor running in big-endian mode.
4872 @item -mstructure-size-boundary=<n>
4873 @kindex -mstructure-size-boundary
4874 The size of all structures and unions will be rounded up to a multiple
4875 of the number of bits set by this option. Permissible values are 8 and
4876 32. The default value varies for different toolchains. For the COFF
4877 targeted toolchain the default value is 8. Specifying the larger number
4878 can produced faster, more efficient code, but can also increase the size
4879 of the program. The two values are potentially incompatible. Code
4880 compiled with one value cannot necessarily expect to work with code or
4881 libraries compiled with the other value, if they exchange information
4882 using structures or unions. Programmers are encouraged to use the 32
4883 value as future versions of the toolchain may default to this value.
4885 @item -mnop-fun-dllimport
4886 @kindex -mnop-fun-dllimport
4887 Disable the support for the @emph{dllimport} attribute.
4889 @item -mcallee-super-interworking
4890 @kindex -mcallee-super-interworking
4891 Gives all externally visible functions in the file being compiled an ARM
4892 instruction set header which switches to Thumb mode before executing the
4893 rest of the function. This allows these functions to be called from
4894 non-interworking code.
4896 @item -mcaller-super-interworking
4897 @kindex -mcaller-super-interworking
4898 Allows calls via function pointers (including virtual functions) to
4899 execute correctly regardless of whether the target code has been
4900 compiled for interworking or not. There is a small overhead in the cost
4901 of executing a function pointer if this option is enabled.
4903 @item -msingle-pic-base
4904 @kindex -msingle-pic-base
4905 Treat the register used for PIC addressing as read-only, rather than
4906 loading it in the prologue for each function. The run-time system is
4907 responsible for initialising this register with an appropriate value
4908 before execution begins.
4910 @item -mpic-register=<reg>
4911 @kindex -mpic-register=
4912 Specify the register to be used for PIC addressing. The default is R10.
4916 @node MN10200 Options
4917 @subsection MN10200 Options
4918 @cindex MN10200 options
4919 These @samp{-m} options are defined for Matsushita MN10200 architectures:
4923 Indicate to the linker that it should perform a relaxation optimization pass
4924 to shorten branches, calls and absolute memory addresses. This option only
4925 has an effect when used on the command line for the final link step.
4927 This option makes symbolic debugging impossible.
4930 @node MN10300 Options
4931 @subsection MN10300 Options
4932 @cindex MN10300 options
4933 These @samp{-m} options are defined for Matsushita MN10300 architectures:
4937 Generate code to avoid bugs in the multiply instructions for the MN10300
4938 processors. This is the default.
4941 Do not generate code to avoid bugs in the multiply instructions for the
4945 Generate code which uses features specific to the AM33 processor.
4948 Do not generate code which uses features specific to the AM33 processor. This
4952 Indicate to the linker that it should perform a relaxation optimization pass
4953 to shorten branches, calls and absolute memory addresses. This option only
4954 has an effect when used on the command line for the final link step.
4956 This option makes symbolic debugging impossible.
4960 @node M32R/D Options
4961 @subsection M32R/D Options
4962 @cindex M32R/D options
4964 These @samp{-m} options are defined for Mitsubishi M32R/D architectures:
4967 @item -mcode-model=small
4968 Assume all objects live in the lower 16MB of memory (so that their addresses
4969 can be loaded with the @code{ld24} instruction), and assume all subroutines
4970 are reachable with the @code{bl} instruction.
4971 This is the default.
4973 The addressability of a particular object can be set with the
4974 @code{model} attribute.
4976 @item -mcode-model=medium
4977 Assume objects may be anywhere in the 32 bit address space (the compiler
4978 will generate @code{seth/add3} instructions to load their addresses), and
4979 assume all subroutines are reachable with the @code{bl} instruction.
4981 @item -mcode-model=large
4982 Assume objects may be anywhere in the 32 bit address space (the compiler
4983 will generate @code{seth/add3} instructions to load their addresses), and
4984 assume subroutines may not be reachable with the @code{bl} instruction
4985 (the compiler will generate the much slower @code{seth/add3/jl}
4986 instruction sequence).
4989 Disable use of the small data area. Variables will be put into
4990 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
4991 @code{section} attribute has been specified).
4992 This is the default.
4994 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
4995 Objects may be explicitly put in the small data area with the
4996 @code{section} attribute using one of these sections.
4999 Put small global and static data in the small data area, but do not
5000 generate special code to reference them.
5003 Put small global and static data in the small data area, and generate
5004 special instructions to reference them.
5007 @cindex smaller data references
5008 Put global and static objects less than or equal to @var{num} bytes
5009 into the small data or bss sections instead of the normal data or bss
5010 sections. The default value of @var{num} is 8.
5011 The @samp{-msdata} option must be set to one of @samp{sdata} or @samp{use}
5012 for this option to have any effect.
5014 All modules should be compiled with the same @samp{-G @var{num}} value.
5015 Compiling with different values of @var{num} may or may not work; if it
5016 doesn't the linker will give an error message - incorrect code will not be
5022 @subsection M88K Options
5023 @cindex M88k options
5025 These @samp{-m} options are defined for Motorola 88k architectures:
5030 Generate code that works well on both the m88100 and the
5035 Generate code that works best for the m88100, but that also
5040 Generate code that works best for the m88110, and may not run
5045 Obsolete option to be removed from the next revision.
5048 @item -midentify-revision
5049 @kindex -midentify-revision
5051 @cindex identifying source, compiler (88k)
5052 Include an @code{ident} directive in the assembler output recording the
5053 source file name, compiler name and version, timestamp, and compilation
5056 @item -mno-underscores
5057 @kindex -mno-underscores
5058 @cindex underscores, avoiding (88k)
5059 In assembler output, emit symbol names without adding an underscore
5060 character at the beginning of each name. The default is to use an
5061 underscore as prefix on each name.
5063 @item -mocs-debug-info
5064 @itemx -mno-ocs-debug-info
5065 @kindex -mocs-debug-info
5066 @kindex -mno-ocs-debug-info
5068 @cindex debugging, 88k OCS
5069 Include (or omit) additional debugging information (about registers used
5070 in each stack frame) as specified in the 88open Object Compatibility
5071 Standard, ``OCS''. This extra information allows debugging of code that
5072 has had the frame pointer eliminated. The default for DG/UX, SVr4, and
5073 Delta 88 SVr3.2 is to include this information; other 88k configurations
5074 omit this information by default.
5076 @item -mocs-frame-position
5077 @kindex -mocs-frame-position
5078 @cindex register positions in frame (88k)
5079 When emitting COFF debugging information for automatic variables and
5080 parameters stored on the stack, use the offset from the canonical frame
5081 address, which is the stack pointer (register 31) on entry to the
5082 function. The DG/UX, SVr4, Delta88 SVr3.2, and BCS configurations use
5083 @samp{-mocs-frame-position}; other 88k configurations have the default
5084 @samp{-mno-ocs-frame-position}.
5086 @item -mno-ocs-frame-position
5087 @kindex -mno-ocs-frame-position
5088 @cindex register positions in frame (88k)
5089 When emitting COFF debugging information for automatic variables and
5090 parameters stored on the stack, use the offset from the frame pointer
5091 register (register 30). When this option is in effect, the frame
5092 pointer is not eliminated when debugging information is selected by the
5095 @item -moptimize-arg-area
5096 @itemx -mno-optimize-arg-area
5097 @kindex -moptimize-arg-area
5098 @kindex -mno-optimize-arg-area
5099 @cindex arguments in frame (88k)
5100 Control how function arguments are stored in stack frames.
5101 @samp{-moptimize-arg-area} saves space by optimizing them, but this
5102 conflicts with the 88open specifications. The opposite alternative,
5103 @samp{-mno-optimize-arg-area}, agrees with 88open standards. By default
5104 GCC does not optimize the argument area.
5106 @item -mshort-data-@var{num}
5107 @kindex -mshort-data-@var{num}
5108 @cindex smaller data references (88k)
5109 @cindex r0-relative references (88k)
5110 Generate smaller data references by making them relative to @code{r0},
5111 which allows loading a value using a single instruction (rather than the
5112 usual two). You control which data references are affected by
5113 specifying @var{num} with this option. For example, if you specify
5114 @samp{-mshort-data-512}, then the data references affected are those
5115 involving displacements of less than 512 bytes.
5116 @samp{-mshort-data-@var{num}} is not effective for @var{num} greater
5119 @item -mserialize-volatile
5120 @kindex -mserialize-volatile
5121 @itemx -mno-serialize-volatile
5122 @kindex -mno-serialize-volatile
5123 @cindex sequential consistency on 88k
5124 Do, or don't, generate code to guarantee sequential consistency
5125 of volatile memory references. By default, consistency is
5128 The order of memory references made by the MC88110 processor does
5129 not always match the order of the instructions requesting those
5130 references. In particular, a load instruction may execute before
5131 a preceding store instruction. Such reordering violates
5132 sequential consistency of volatile memory references, when there
5133 are multiple processors. When consistency must be guaranteed,
5134 GNU C generates special instructions, as needed, to force
5135 execution in the proper order.
5137 The MC88100 processor does not reorder memory references and so
5138 always provides sequential consistency. However, by default, GNU
5139 C generates the special instructions to guarantee consistency
5140 even when you use @samp{-m88100}, so that the code may be run on an
5141 MC88110 processor. If you intend to run your code only on the
5142 MC88100 processor, you may use @samp{-mno-serialize-volatile}.
5144 The extra code generated to guarantee consistency may affect the
5145 performance of your application. If you know that you can safely
5146 forgo this guarantee, you may use @samp{-mno-serialize-volatile}.
5152 @cindex assembler syntax, 88k
5154 Turn on (@samp{-msvr4}) or off (@samp{-msvr3}) compiler extensions
5155 related to System V release 4 (SVr4). This controls the following:
5159 Which variant of the assembler syntax to emit.
5161 @samp{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
5162 that is used on System V release 4.
5164 @samp{-msvr4} makes GCC issue additional declaration directives used in
5168 @samp{-msvr4} is the default for the m88k-motorola-sysv4 and
5169 m88k-dg-dgux m88k configurations. @samp{-msvr3} is the default for all
5170 other m88k configurations.
5172 @item -mversion-03.00
5173 @kindex -mversion-03.00
5174 This option is obsolete, and is ignored.
5175 @c ??? which asm syntax better for GAS? option there too?
5177 @item -mno-check-zero-division
5178 @itemx -mcheck-zero-division
5179 @kindex -mno-check-zero-division
5180 @kindex -mcheck-zero-division
5181 @cindex zero division on 88k
5182 Do, or don't, generate code to guarantee that integer division by
5183 zero will be detected. By default, detection is guaranteed.
5185 Some models of the MC88100 processor fail to trap upon integer
5186 division by zero under certain conditions. By default, when
5187 compiling code that might be run on such a processor, GNU C
5188 generates code that explicitly checks for zero-valued divisors
5189 and traps with exception number 503 when one is detected. Use of
5190 mno-check-zero-division suppresses such checking for code
5191 generated to run on an MC88100 processor.
5193 GNU C assumes that the MC88110 processor correctly detects all
5194 instances of integer division by zero. When @samp{-m88110} is
5195 specified, both @samp{-mcheck-zero-division} and
5196 @samp{-mno-check-zero-division} are ignored, and no explicit checks for
5197 zero-valued divisors are generated.
5199 @item -muse-div-instruction
5200 @kindex -muse-div-instruction
5201 @cindex divide instruction, 88k
5202 Use the div instruction for signed integer division on the
5203 MC88100 processor. By default, the div instruction is not used.
5205 On the MC88100 processor the signed integer division instruction
5206 div) traps to the operating system on a negative operand. The
5207 operating system transparently completes the operation, but at a
5208 large cost in execution time. By default, when compiling code
5209 that might be run on an MC88100 processor, GNU C emulates signed
5210 integer division using the unsigned integer division instruction
5211 divu), thereby avoiding the large penalty of a trap to the
5212 operating system. Such emulation has its own, smaller, execution
5213 cost in both time and space. To the extent that your code's
5214 important signed integer division operations are performed on two
5215 nonnegative operands, it may be desirable to use the div
5216 instruction directly.
5218 On the MC88110 processor the div instruction (also known as the
5219 divs instruction) processes negative operands without trapping to
5220 the operating system. When @samp{-m88110} is specified,
5221 @samp{-muse-div-instruction} is ignored, and the div instruction is used
5222 for signed integer division.
5224 Note that the result of dividing INT_MIN by -1 is undefined. In
5225 particular, the behavior of such a division with and without
5226 @samp{-muse-div-instruction} may differ.
5228 @item -mtrap-large-shift
5229 @itemx -mhandle-large-shift
5230 @kindex -mtrap-large-shift
5231 @kindex -mhandle-large-shift
5232 @cindex bit shift overflow (88k)
5233 @cindex large bit shifts (88k)
5234 Include code to detect bit-shifts of more than 31 bits; respectively,
5235 trap such shifts or emit code to handle them properly. By default GCC
5236 makes no special provision for large bit shifts.
5238 @item -mwarn-passed-structs
5239 @kindex -mwarn-passed-structs
5240 @cindex structure passing (88k)
5241 Warn when a function passes a struct as an argument or result.
5242 Structure-passing conventions have changed during the evolution of the C
5243 language, and are often the source of portability problems. By default,
5244 GCC issues no such warning.
5247 @node RS/6000 and PowerPC Options
5248 @subsection IBM RS/6000 and PowerPC Options
5249 @cindex RS/6000 and PowerPC Options
5250 @cindex IBM RS/6000 and PowerPC Options
5252 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
5260 @itemx -mpowerpc-gpopt
5261 @itemx -mno-powerpc-gpopt
5262 @itemx -mpowerpc-gfxopt
5263 @itemx -mno-powerpc-gfxopt
5265 @itemx -mno-powerpc64
5269 @kindex -mpowerpc-gpopt
5270 @kindex -mpowerpc-gfxopt
5272 GCC supports two related instruction set architectures for the
5273 RS/6000 and PowerPC. The @dfn{POWER} instruction set are those
5274 instructions supported by the @samp{rios} chip set used in the original
5275 RS/6000 systems and the @dfn{PowerPC} instruction set is the
5276 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
5277 the IBM 4xx microprocessors.
5279 Neither architecture is a subset of the other. However there is a
5280 large common subset of instructions supported by both. An MQ
5281 register is included in processors supporting the POWER architecture.
5283 You use these options to specify which instructions are available on the
5284 processor you are using. The default value of these options is
5285 determined when configuring GCC. Specifying the
5286 @samp{-mcpu=@var{cpu_type}} overrides the specification of these
5287 options. We recommend you use the @samp{-mcpu=@var{cpu_type}} option
5288 rather than the options listed above.
5290 The @samp{-mpower} option allows GCC to generate instructions that
5291 are found only in the POWER architecture and to use the MQ register.
5292 Specifying @samp{-mpower2} implies @samp{-power} and also allows GCC
5293 to generate instructions that are present in the POWER2 architecture but
5294 not the original POWER architecture.
5296 The @samp{-mpowerpc} option allows GCC to generate instructions that
5297 are found only in the 32-bit subset of the PowerPC architecture.
5298 Specifying @samp{-mpowerpc-gpopt} implies @samp{-mpowerpc} and also allows
5299 GCC to use the optional PowerPC architecture instructions in the
5300 General Purpose group, including floating-point square root. Specifying
5301 @samp{-mpowerpc-gfxopt} implies @samp{-mpowerpc} and also allows GCC to
5302 use the optional PowerPC architecture instructions in the Graphics
5303 group, including floating-point select.
5305 The @samp{-mpowerpc64} option allows GCC to generate the additional
5306 64-bit instructions that are found in the full PowerPC64 architecture
5307 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
5308 @samp{-mno-powerpc64}.
5310 If you specify both @samp{-mno-power} and @samp{-mno-powerpc}, GCC
5311 will use only the instructions in the common subset of both
5312 architectures plus some special AIX common-mode calls, and will not use
5313 the MQ register. Specifying both @samp{-mpower} and @samp{-mpowerpc}
5314 permits GCC to use any instruction from either architecture and to
5315 allow use of the MQ register; specify this for the Motorola MPC601.
5317 @item -mnew-mnemonics
5318 @itemx -mold-mnemonics
5319 @kindex -mnew-mnemonics
5320 @kindex -mold-mnemonics
5321 Select which mnemonics to use in the generated assembler code.
5322 @samp{-mnew-mnemonics} requests output that uses the assembler mnemonics
5323 defined for the PowerPC architecture, while @samp{-mold-mnemonics}
5324 requests the assembler mnemonics defined for the POWER architecture.
5325 Instructions defined in only one architecture have only one mnemonic;
5326 GCC uses that mnemonic irrespective of which of these options is
5329 GCC defaults to the mnemonics appropriate for the architecture in
5330 use. Specifying @samp{-mcpu=@var{cpu_type}} sometimes overrides the
5331 value of these option. Unless you are building a cross-compiler, you
5332 should normally not specify either @samp{-mnew-mnemonics} or
5333 @samp{-mold-mnemonics}, but should instead accept the default.
5335 @item -mcpu=@var{cpu_type}
5337 Set architecture type, register usage, choice of mnemonics, and
5338 instruction scheduling parameters for machine type @var{cpu_type}.
5339 Supported values for @var{cpu_type} are @samp{rios}, @samp{rios1},
5340 @samp{rsc}, @samp{rios2}, @samp{rs64a}, @samp{601}, @samp{602},
5341 @samp{603}, @samp{603e}, @samp{604}, @samp{604e}, @samp{620},
5342 @samp{630}, @samp{740}, @samp{750}, @samp{power}, @samp{power2},
5343 @samp{powerpc}, @samp{403}, @samp{505}, @samp{801}, @samp{821},
5344 @samp{823}, and @samp{860} and @samp{common}. @samp{-mcpu=power},
5345 @samp{-mcpu=power2}, @samp{-mcpu=powerpc}, and @samp{-mcpu=powerpc64}
5346 specify generic POWER, POWER2, pure 32-bit PowerPC (i.e., not MPC601),
5347 and 64-bit PowerPC architecture machine types, with an appropriate,
5348 generic processor model assumed for scheduling purposes.@refill
5350 Specifying any of the following options:
5351 @samp{-mcpu=rios1}, @samp{-mcpu=rios2}, @samp{-mcpu=rsc},
5352 @samp{-mcpu=power}, or @samp{-mcpu=power2}
5353 enables the @samp{-mpower} option and disables the @samp{-mpowerpc} option;
5354 @samp{-mcpu=601} enables both the @samp{-mpower} and @samp{-mpowerpc} options.
5355 All of @samp{-mcpu=rs64a}, @samp{-mcpu=602}, @samp{-mcpu=603},
5356 @samp{-mcpu=603e}, @samp{-mcpu=604}, @samp{-mcpu=620}, @samp{-mcpu=630},
5357 @samp{-mcpu=740}, and @samp{-mcpu=750}
5358 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
5359 Exactly similarly, all of @samp{-mcpu=403},
5360 @samp{-mcpu=505}, @samp{-mcpu=821}, @samp{-mcpu=860} and @samp{-mcpu=powerpc}
5361 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
5362 @samp{-mcpu=common} disables both the
5363 @samp{-mpower} and @samp{-mpowerpc} options.@refill
5365 AIX versions 4 or greater selects @samp{-mcpu=common} by default, so
5366 that code will operate on all members of the RS/6000 POWER and PowerPC
5367 families. In that case, GCC will use only the instructions in the
5368 common subset of both architectures plus some special AIX common-mode
5369 calls, and will not use the MQ register. GCC assumes a generic
5370 processor model for scheduling purposes.
5372 Specifying any of the options @samp{-mcpu=rios1}, @samp{-mcpu=rios2},
5373 @samp{-mcpu=rsc}, @samp{-mcpu=power}, or @samp{-mcpu=power2} also
5374 disables the @samp{new-mnemonics} option. Specifying @samp{-mcpu=601},
5375 @samp{-mcpu=602}, @samp{-mcpu=603}, @samp{-mcpu=603e}, @samp{-mcpu=604},
5376 @samp{-mcpu=620}, @samp{-mcpu=630}, @samp{-mcpu=403}, @samp{-mcpu=505},
5377 @samp{-mcpu=821}, @samp{-mcpu=860} or @samp{-mcpu=powerpc} also enables
5378 the @samp{new-mnemonics} option.@refill
5380 Specifying @samp{-mcpu=403}, @samp{-mcpu=821}, or @samp{-mcpu=860} also
5381 enables the @samp{-msoft-float} option.
5383 @item -mtune=@var{cpu_type}
5384 Set the instruction scheduling parameters for machine type
5385 @var{cpu_type}, but do not set the architecture type, register usage,
5386 choice of mnemonics like @samp{-mcpu=}@var{cpu_type} would. The same
5387 values for @var{cpu_type} are used for @samp{-mtune=}@var{cpu_type} as
5388 for @samp{-mcpu=}@var{cpu_type}. The @samp{-mtune=}@var{cpu_type}
5389 option overrides the @samp{-mcpu=}@var{cpu_type} option in terms of
5390 instruction scheduling parameters.
5393 @itemx -mno-fp-in-toc
5394 @itemx -mno-sum-in-toc
5395 @itemx -mminimal-toc
5396 @kindex -mminimal-toc
5397 Modify generation of the TOC (Table Of Contents), which is created for
5398 every executable file. The @samp{-mfull-toc} option is selected by
5399 default. In that case, GCC will allocate at least one TOC entry for
5400 each unique non-automatic variable reference in your program. GCC
5401 will also place floating-point constants in the TOC. However, only
5402 16,384 entries are available in the TOC.
5404 If you receive a linker error message that saying you have overflowed
5405 the available TOC space, you can reduce the amount of TOC space used
5406 with the @samp{-mno-fp-in-toc} and @samp{-mno-sum-in-toc} options.
5407 @samp{-mno-fp-in-toc} prevents GCC from putting floating-point
5408 constants in the TOC and @samp{-mno-sum-in-toc} forces GCC to
5409 generate code to calculate the sum of an address and a constant at
5410 run-time instead of putting that sum into the TOC. You may specify one
5411 or both of these options. Each causes GCC to produce very slightly
5412 slower and larger code at the expense of conserving TOC space.
5414 If you still run out of space in the TOC even when you specify both of
5415 these options, specify @samp{-mminimal-toc} instead. This option causes
5416 GCC to make only one TOC entry for every file. When you specify this
5417 option, GCC will produce code that is slower and larger but which
5418 uses extremely little TOC space. You may wish to use this option
5419 only on files that contain less frequently executed code. @refill
5425 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
5426 @code{long} type, and the infrastructure needed to support them.
5427 Specifying @samp{-maix64} implies @samp{-mpowerpc64} and
5428 @samp{-mpowerpc}, while @samp{-maix32} disables the 64-bit ABI and
5429 implies @samp{-mno-powerpc64}. GCC defaults to @samp{-maix32}.
5434 On AIX, pass floating-point arguments to prototyped functions beyond the
5435 register save area (RSA) on the stack in addition to argument FPRs. The
5436 AIX calling convention was extended but not initially documented to
5437 handle an obscure K&R C case of calling a function that takes the
5438 address of its arguments with fewer arguments than declared. AIX XL
5439 compilers access floating point arguments which do not fit in the
5440 RSA from the stack when a subroutine is compiled without
5441 optimization. Because always storing floating-point arguments on the
5442 stack is inefficient and rarely needed, this option is not enabled by
5443 default and only is necessary when calling subroutines compiled by AIX
5444 XL compilers without optimization.
5448 Support @dfn{AIX Threads}. Link an application written to use
5449 @dfn{pthreads} with special libraries and startup code to enable the
5454 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE). Link an
5455 application written to use message passing with special startup code to
5456 enable the application to run. The system must have PE installed in the
5457 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
5458 must be overridden with the @samp{-specs=} option to specify the
5459 appropriate directory location. The Parallel Environment does not
5460 support threads, so the @samp{-mpe} option and the @samp{-mthreads}
5461 option are incompatible.
5465 @kindex -msoft-float
5466 Generate code that does not use (uses) the floating-point register set.
5467 Software floating point emulation is provided if you use the
5468 @samp{-msoft-float} option, and pass the option to GCC when linking.
5471 @itemx -mno-multiple
5472 Generate code that uses (does not use) the load multiple word
5473 instructions and the store multiple word instructions. These
5474 instructions are generated by default on POWER systems, and not
5475 generated on PowerPC systems. Do not use @samp{-mmultiple} on little
5476 endian PowerPC systems, since those instructions do not work when the
5477 processor is in little endian mode. The exceptions are PPC740 and
5478 PPC750 which permit the instructions usage in little endian mode.
5483 Generate code that uses (does not use) the load string instructions
5484 and the store string word instructions to save multiple registers and
5485 do small block moves. These instructions are generated by default on
5486 POWER systems, and not generated on PowerPC systems. Do not use
5487 @samp{-mstring} on little endian PowerPC systems, since those
5488 instructions do not work when the processor is in little endian mode.
5489 The exceptions are PPC740 and PPC750 which permit the instructions
5490 usage in little endian mode.
5495 Generate code that uses (does not use) the load or store instructions
5496 that update the base register to the address of the calculated memory
5497 location. These instructions are generated by default. If you use
5498 @samp{-mno-update}, there is a small window between the time that the
5499 stack pointer is updated and the address of the previous frame is
5500 stored, which means code that walks the stack frame across interrupts or
5501 signals may get corrupted data.
5504 @itemx -mno-fused-madd
5505 @kindex -mfused-madd
5506 Generate code that uses (does not use) the floating point multiply and
5507 accumulate instructions. These instructions are generated by default if
5508 hardware floating is used.
5510 @item -mno-bit-align
5513 On System V.4 and embedded PowerPC systems do not (do) force structures
5514 and unions that contain bit fields to be aligned to the base type of the
5517 For example, by default a structure containing nothing but 8
5518 @code{unsigned} bitfields of length 1 would be aligned to a 4 byte
5519 boundary and have a size of 4 bytes. By using @samp{-mno-bit-align},
5520 the structure would be aligned to a 1 byte boundary and be one byte in
5523 @item -mno-strict-align
5524 @itemx -mstrict-align
5525 @kindex -mstrict-align
5526 On System V.4 and embedded PowerPC systems do not (do) assume that
5527 unaligned memory references will be handled by the system.
5530 @itemx -mno-relocatable
5531 @kindex -mrelocatable
5532 On embedded PowerPC systems generate code that allows (does not allow)
5533 the program to be relocated to a different address at runtime. If you
5534 use @samp{-mrelocatable} on any module, all objects linked together must
5535 be compiled with @samp{-mrelocatable} or @samp{-mrelocatable-lib}.
5537 @item -mrelocatable-lib
5538 @itemx -mno-relocatable-lib
5539 On embedded PowerPC systems generate code that allows (does not allow)
5540 the program to be relocated to a different address at runtime. Modules
5541 compiled with @samp{-mrelocatable-lib} can be linked with either modules
5542 compiled without @samp{-mrelocatable} and @samp{-mrelocatable-lib} or
5543 with modules compiled with the @samp{-mrelocatable} options.
5547 On System V.4 and embedded PowerPC systems do not (do) assume that
5548 register 2 contains a pointer to a global area pointing to the addresses
5549 used in the program.
5552 @itemx -mlittle-endian
5553 On System V.4 and embedded PowerPC systems compile code for the
5554 processor in little endian mode. The @samp{-mlittle-endian} option is
5555 the same as @samp{-mlittle}.
5559 On System V.4 and embedded PowerPC systems compile code for the
5560 processor in big endian mode. The @samp{-mbig-endian} option is
5561 the same as @samp{-mbig}.
5564 On System V.4 and embedded PowerPC systems compile code using calling
5565 conventions that adheres to the March 1995 draft of the System V
5566 Application Binary Interface, PowerPC processor supplement. This is the
5567 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
5569 @item -mcall-sysv-eabi
5570 Specify both @samp{-mcall-sysv} and @samp{-meabi} options.
5572 @item -mcall-sysv-noeabi
5573 Specify both @samp{-mcall-sysv} and @samp{-mno-eabi} options.
5576 On System V.4 and embedded PowerPC systems compile code using calling
5577 conventions that are similar to those used on AIX. This is the
5578 default if you configured GCC using @samp{powerpc-*-eabiaix}.
5580 @item -mcall-solaris
5581 On System V.4 and embedded PowerPC systems compile code for the Solaris
5585 On System V.4 and embedded PowerPC systems compile code for the
5586 Linux-based GNU system.
5589 @itemx -mno-prototype
5590 On System V.4 and embedded PowerPC systems assume that all calls to
5591 variable argument functions are properly prototyped. Otherwise, the
5592 compiler must insert an instruction before every non prototyped call to
5593 set or clear bit 6 of the condition code register (@var{CR}) to
5594 indicate whether floating point values were passed in the floating point
5595 registers in case the function takes a variable arguments. With
5596 @samp{-mprototype}, only calls to prototyped variable argument functions
5597 will set or clear the bit.
5600 On embedded PowerPC systems, assume that the startup module is called
5601 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
5602 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
5606 On embedded PowerPC systems, assume that the startup module is called
5607 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
5611 On embedded PowerPC systems, assume that the startup module is called
5612 @file{crt0.o} and the standard C libraries are @file{libads.a} and
5616 On embedded PowerPC systems, assume that the startup module is called
5617 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
5621 On System V.4 and embedded PowerPC systems, specify that you are
5622 compiling for a VxWorks system.
5625 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
5626 header to indicate that @samp{eabi} extended relocations are used.
5630 On System V.4 and embedded PowerPC systems do (do not) adhere to the
5631 Embedded Applications Binary Interface (eabi) which is a set of
5632 modifications to the System V.4 specifications. Selecting @code{-meabi}
5633 means that the stack is aligned to an 8 byte boundary, a function
5634 @code{__eabi} is called to from @code{main} to set up the eabi
5635 environment, and the @samp{-msdata} option can use both @code{r2} and
5636 @code{r13} to point to two separate small data areas. Selecting
5637 @code{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
5638 do not call an initialization function from @code{main}, and the
5639 @samp{-msdata} option will only use @code{r13} to point to a single
5640 small data area. The @samp{-meabi} option is on by default if you
5641 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
5644 On System V.4 and embedded PowerPC systems, put small initialized
5645 @code{const} global and static data in the @samp{.sdata2} section, which
5646 is pointed to by register @code{r2}. Put small initialized
5647 non-@code{const} global and static data in the @samp{.sdata} section,
5648 which is pointed to by register @code{r13}. Put small uninitialized
5649 global and static data in the @samp{.sbss} section, which is adjacent to
5650 the @samp{.sdata} section. The @samp{-msdata=eabi} option is
5651 incompatible with the @samp{-mrelocatable} option. The
5652 @samp{-msdata=eabi} option also sets the @samp{-memb} option.
5655 On System V.4 and embedded PowerPC systems, put small global and static
5656 data in the @samp{.sdata} section, which is pointed to by register
5657 @code{r13}. Put small uninitialized global and static data in the
5658 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
5659 The @samp{-msdata=sysv} option is incompatible with the
5660 @samp{-mrelocatable} option.
5662 @item -msdata=default
5664 On System V.4 and embedded PowerPC systems, if @samp{-meabi} is used,
5665 compile code the same as @samp{-msdata=eabi}, otherwise compile code the
5666 same as @samp{-msdata=sysv}.
5669 On System V.4 and embedded PowerPC systems, put small global and static
5670 data in the @samp{.sdata} section. Put small uninitialized global and
5671 static data in the @samp{.sbss} section. Do not use register @code{r13}
5672 to address small data however. This is the default behavior unless
5673 other @samp{-msdata} options are used.
5677 On embedded PowerPC systems, put all initialized global and static data
5678 in the @samp{.data} section, and all uninitialized data in the
5679 @samp{.bss} section.
5682 @cindex smaller data references (PowerPC)
5683 @cindex .sdata/.sdata2 references (PowerPC)
5684 On embedded PowerPC systems, put global and static items less than or
5685 equal to @var{num} bytes into the small data or bss sections instead of
5686 the normal data or bss section. By default, @var{num} is 8. The
5687 @samp{-G @var{num}} switch is also passed to the linker.
5688 All modules should be compiled with the same @samp{-G @var{num}} value.
5691 @itemx -mno-regnames
5692 On System V.4 and embedded PowerPC systems do (do not) emit register
5693 names in the assembly language output using symbolic forms.
5698 @subsection IBM RT Options
5700 @cindex IBM RT options
5702 These @samp{-m} options are defined for the IBM RT PC:
5706 Use an in-line code sequence for integer multiplies. This is the
5709 @item -mcall-lib-mul
5710 Call @code{lmul$$} for integer multiples.
5712 @item -mfull-fp-blocks
5713 Generate full-size floating point data blocks, including the minimum
5714 amount of scratch space recommended by IBM. This is the default.
5716 @item -mminimum-fp-blocks
5717 Do not include extra scratch space in floating point data blocks. This
5718 results in smaller code, but slower execution, since scratch space must
5719 be allocated dynamically.
5721 @cindex @file{varargs.h} and RT PC
5722 @cindex @file{stdarg.h} and RT PC
5723 @item -mfp-arg-in-fpregs
5724 Use a calling sequence incompatible with the IBM calling convention in
5725 which floating point arguments are passed in floating point registers.
5726 Note that @code{varargs.h} and @code{stdargs.h} will not work with
5727 floating point operands if this option is specified.
5729 @item -mfp-arg-in-gregs
5730 Use the normal calling convention for floating point arguments. This is
5733 @item -mhc-struct-return
5734 Return structures of more than one word in memory, rather than in a
5735 register. This provides compatibility with the MetaWare HighC (hc)
5736 compiler. Use the option @samp{-fpcc-struct-return} for compatibility
5737 with the Portable C Compiler (pcc).
5739 @item -mnohc-struct-return
5740 Return some structures of more than one word in registers, when
5741 convenient. This is the default. For compatibility with the
5742 IBM-supplied compilers, use the option @samp{-fpcc-struct-return} or the
5743 option @samp{-mhc-struct-return}.
5747 @subsection MIPS Options
5748 @cindex MIPS options
5750 These @samp{-m} options are defined for the MIPS family of computers:
5753 @item -mcpu=@var{cpu type}
5754 Assume the defaults for the machine type @var{cpu type} when scheduling
5755 instructions. The choices for @var{cpu type} are @samp{r2000}, @samp{r3000},
5756 @samp{r3900}, @samp{r4000}, @samp{r4100}, @samp{r4300}, @samp{r4400},
5757 @samp{r4600}, @samp{r4650}, @samp{r5000}, @samp{r6000}, @samp{r8000},
5758 and @samp{orion}. Additionally, the @samp{r2000}, @samp{r3000},
5759 @samp{r4000}, @samp{r5000}, and @samp{r6000} can be abbreviated as
5760 @samp{r2k} (or @samp{r2K}), @samp{r3k}, etc. While picking a specific
5761 @var{cpu type} will schedule things appropriately for that particular
5762 chip, the compiler will not generate any code that does not meet level 1
5763 of the MIPS ISA (instruction set architecture) without a @samp{-mipsX}
5764 or @samp{-mabi} switch being used.
5767 Issue instructions from level 1 of the MIPS ISA. This is the default.
5768 @samp{r3000} is the default @var{cpu type} at this ISA level.
5771 Issue instructions from level 2 of the MIPS ISA (branch likely, square
5772 root instructions). @samp{r6000} is the default @var{cpu type} at this
5776 Issue instructions from level 3 of the MIPS ISA (64 bit instructions).
5777 @samp{r4000} is the default @var{cpu type} at this ISA level.
5780 Issue instructions from level 4 of the MIPS ISA (conditional move,
5781 prefetch, enhanced FPU instructions). @samp{r8000} is the default
5782 @var{cpu type} at this ISA level.
5785 Assume that 32 32-bit floating point registers are available. This is
5789 Assume that 32 64-bit floating point registers are available. This is
5790 the default when the @samp{-mips3} option is used.
5793 Assume that 32 32-bit general purpose registers are available. This is
5797 Assume that 32 64-bit general purpose registers are available. This is
5798 the default when the @samp{-mips3} option is used.
5801 Force int and long types to be 64 bits wide. See @samp{-mlong32} for an
5802 explanation of the default, and the width of pointers.
5805 Force long types to be 64 bits wide. See @samp{-mlong32} for an
5806 explanation of the default, and the width of pointers.
5809 Force long, int, and pointer types to be 32 bits wide.
5811 If none of @samp{-mlong32}, @samp{-mlong64}, or @samp{-mint64} are set,
5812 the size of ints, longs, and pointers depends on the ABI and ISA chosen.
5813 For @samp{-mabi=32}, and @samp{-mabi=n32}, ints and longs are 32 bits
5814 wide. For @samp{-mabi=64}, ints are 32 bits, and longs are 64 bits wide.
5815 For @samp{-mabi=eabi} and either @samp{-mips1} or @samp{-mips2}, ints
5816 and longs are 32 bits wide. For @samp{-mabi=eabi} and higher ISAs, ints
5817 are 32 bits, and longs are 64 bits wide. The width of pointer types is
5818 the smaller of the width of longs or the width of general purpose
5819 registers (which in turn depends on the ISA).
5826 Generate code for the indicated ABI. The default instruction level is
5827 @samp{-mips1} for @samp{32}, @samp{-mips3} for @samp{n32}, and
5828 @samp{-mips4} otherwise. Conversely, with @samp{-mips1} or
5829 @samp{-mips2}, the default ABI is @samp{32}; otherwise, the default ABI
5833 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
5834 add normal debug information. This is the default for all
5835 platforms except for the OSF/1 reference platform, using the OSF/rose
5836 object format. If the either of the @samp{-gstabs} or @samp{-gstabs+}
5837 switches are used, the @file{mips-tfile} program will encapsulate the
5838 stabs within MIPS ECOFF.
5841 Generate code for the GNU assembler. This is the default on the OSF/1
5842 reference platform, using the OSF/rose object format. Also, this is
5843 the default if the configure option @samp{--with-gnu-as} is used.
5845 @item -msplit-addresses
5846 @itemx -mno-split-addresses
5847 Generate code to load the high and low parts of address constants separately.
5848 This allows @code{gcc} to optimize away redundant loads of the high order
5849 bits of addresses. This optimization requires GNU as and GNU ld.
5850 This optimization is enabled by default for some embedded targets where
5851 GNU as and GNU ld are standard.
5855 The @samp{-mrnames} switch says to output code using the MIPS software
5856 names for the registers, instead of the hardware names (ie, @var{a0}
5857 instead of @var{$4}). The only known assembler that supports this option
5858 is the Algorithmics assembler.
5862 The @samp{-mgpopt} switch says to write all of the data declarations
5863 before the instructions in the text section, this allows the MIPS
5864 assembler to generate one word memory references instead of using two
5865 words for short global or static data items. This is on by default if
5866 optimization is selected.
5870 For each non-inline function processed, the @samp{-mstats} switch
5871 causes the compiler to emit one line to the standard error file to
5872 print statistics about the program (number of registers saved, stack
5877 The @samp{-mmemcpy} switch makes all block moves call the appropriate
5878 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
5879 generating inline code.
5882 @itemx -mno-mips-tfile
5883 The @samp{-mno-mips-tfile} switch causes the compiler not
5884 postprocess the object file with the @file{mips-tfile} program,
5885 after the MIPS assembler has generated it to add debug support. If
5886 @file{mips-tfile} is not run, then no local variables will be
5887 available to the debugger. In addition, @file{stage2} and
5888 @file{stage3} objects will have the temporary file names passed to the
5889 assembler embedded in the object file, which means the objects will
5890 not compare the same. The @samp{-mno-mips-tfile} switch should only
5891 be used when there are bugs in the @file{mips-tfile} program that
5892 prevents compilation.
5895 Generate output containing library calls for floating point.
5896 @strong{Warning:} the requisite libraries are not part of GCC.
5897 Normally the facilities of the machine's usual C compiler are used, but
5898 this can't be done directly in cross-compilation. You must make your
5899 own arrangements to provide suitable library functions for
5903 Generate output containing floating point instructions. This is the
5904 default if you use the unmodified sources.
5907 @itemx -mno-abicalls
5908 Emit (or do not emit) the pseudo operations @samp{.abicalls},
5909 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
5910 position independent code.
5913 @itemx -mno-long-calls
5914 Do all calls with the @samp{JALR} instruction, which requires
5915 loading up a function's address into a register before the call.
5916 You need to use this switch, if you call outside of the current
5917 512 megabyte segment to functions that are not through pointers.
5920 @itemx -mno-half-pic
5921 Put pointers to extern references into the data section and load them
5922 up, rather than put the references in the text section.
5924 @item -membedded-pic
5925 @itemx -mno-embedded-pic
5926 Generate PIC code suitable for some embedded systems. All calls are
5927 made using PC relative address, and all data is addressed using the $gp
5928 register. No more than 65536 bytes of global data may be used. This
5929 requires GNU as and GNU ld which do most of the work. This currently
5930 only works on targets which use ECOFF; it does not work with ELF.
5932 @item -membedded-data
5933 @itemx -mno-embedded-data
5934 Allocate variables to the read-only data section first if possible, then
5935 next in the small data section if possible, otherwise in data. This gives
5936 slightly slower code than the default, but reduces the amount of RAM required
5937 when executing, and thus may be preferred for some embedded systems.
5939 @item -muninit-const-in-rodata
5940 @itemx -mno-uninit-const-in-rodata
5941 When used together with -membedded-data, it will always store uninitialized
5942 const variables in the read-only data section.
5944 @item -msingle-float
5945 @itemx -mdouble-float
5946 The @samp{-msingle-float} switch tells gcc to assume that the floating
5947 point coprocessor only supports single precision operations, as on the
5948 @samp{r4650} chip. The @samp{-mdouble-float} switch permits gcc to use
5949 double precision operations. This is the default.
5953 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
5954 as on the @samp{r4650} chip.
5957 Turns on @samp{-msingle-float}, @samp{-mmad}, and, at least for now,
5962 Enable 16-bit instructions.
5965 Use the entry and exit pseudo ops. This option can only be used with
5969 Compile code for the processor in little endian mode.
5970 The requisite libraries are assumed to exist.
5973 Compile code for the processor in big endian mode.
5974 The requisite libraries are assumed to exist.
5977 @cindex smaller data references (MIPS)
5978 @cindex gp-relative references (MIPS)
5979 Put global and static items less than or equal to @var{num} bytes into
5980 the small data or bss sections instead of the normal data or bss
5981 section. This allows the assembler to emit one word memory reference
5982 instructions based on the global pointer (@var{gp} or @var{$28}),
5983 instead of the normal two words used. By default, @var{num} is 8 when
5984 the MIPS assembler is used, and 0 when the GNU assembler is used. The
5985 @samp{-G @var{num}} switch is also passed to the assembler and linker.
5986 All modules should be compiled with the same @samp{-G @var{num}}
5990 Tell the MIPS assembler to not run its preprocessor over user
5991 assembler files (with a @samp{.s} suffix) when assembling them.
5994 Pass an option to gas which will cause nops to be inserted if
5995 the read of the destination register of an mfhi or mflo instruction
5996 occurs in the following two instructions.
5999 Do not include the default crt0.
6003 These options are defined by the macro
6004 @code{TARGET_SWITCHES} in the machine description. The default for the
6005 options is also defined by that macro, which enables you to change the
6010 @subsection Intel 386 Options
6011 @cindex i386 Options
6012 @cindex Intel 386 Options
6014 These @samp{-m} options are defined for the i386 family of computers:
6017 @item -mcpu=@var{cpu type}
6018 Assume the defaults for the machine type @var{cpu type} when scheduling
6019 instructions. The choices for @var{cpu type} are:
6021 @multitable @columnfractions .20 .20 .20 .20
6022 @item @samp{i386} @tab @samp{i486} @tab @samp{i586} @tab @samp{i686}
6023 @item @samp{pentium} @tab @samp{pentiumpro} @tab @samp{k6}
6026 While picking a specific @var{cpu type} will schedule things appropriately
6027 for that particular chip, the compiler will not generate any code that
6028 does not run on the i386 without the @samp{-march=@var{cpu type}} option
6029 being used. @samp{i586} is equivalent to @samp{pentium} and @samp{i686}
6030 is equivalent to @samp{pentiumpro}. @samp{k6} is the AMD chip as
6031 opposed to the Intel ones.
6033 @item -march=@var{cpu type}
6034 Generate instructions for the machine type @var{cpu type}. The choices
6035 for @var{cpu type} are the same as for @samp{-mcpu}. Moreover,
6036 specifying @samp{-march=@var{cpu type}} implies @samp{-mcpu=@var{cpu type}}.
6042 Synonyms for -mcpu=i386, -mcpu=i486, -mcpu=pentium, and -mcpu=pentiumpro
6043 respectively. These synonyms are deprecated.
6047 Control whether or not the compiler uses IEEE floating point
6048 comparisons. These handle correctly the case where the result of a
6049 comparison is unordered.
6052 Generate output containing library calls for floating point.
6053 @strong{Warning:} the requisite libraries are not part of GCC.
6054 Normally the facilities of the machine's usual C compiler are used, but
6055 this can't be done directly in cross-compilation. You must make your
6056 own arrangements to provide suitable library functions for
6059 On machines where a function returns floating point results in the 80387
6060 register stack, some floating point opcodes may be emitted even if
6061 @samp{-msoft-float} is used.
6063 @item -mno-fp-ret-in-387
6064 Do not use the FPU registers for return values of functions.
6066 The usual calling convention has functions return values of types
6067 @code{float} and @code{double} in an FPU register, even if there
6068 is no FPU. The idea is that the operating system should emulate
6071 The option @samp{-mno-fp-ret-in-387} causes such values to be returned
6072 in ordinary CPU registers instead.
6074 @item -mno-fancy-math-387
6075 Some 387 emulators do not support the @code{sin}, @code{cos} and
6076 @code{sqrt} instructions for the 387. Specify this option to avoid
6077 generating those instructions. This option is the default on FreeBSD.
6078 As of revision 2.6.1, these instructions are not generated unless you
6079 also use the @samp{-ffast-math} switch.
6081 @item -malign-double
6082 @itemx -mno-align-double
6083 Control whether GCC aligns @code{double}, @code{long double}, and
6084 @code{long long} variables on a two word boundary or a one word
6085 boundary. Aligning @code{double} variables on a two word boundary will
6086 produce code that runs somewhat faster on a @samp{Pentium} at the
6087 expense of more memory.
6089 @strong{Warning:} if you use the @samp{-malign-double} switch,
6090 structures containing the above types will be aligned differently than
6091 the published application binary interface specifications for the 386.
6094 @itemx -mno-svr3-shlib
6095 Control whether GCC places uninitialized locals into @code{bss} or
6096 @code{data}. @samp{-msvr3-shlib} places these locals into @code{bss}.
6097 These options are meaningful only on System V Release 3.
6099 @item -mno-wide-multiply
6100 @itemx -mwide-multiply
6101 Control whether GCC uses the @code{mul} and @code{imul} that produce
6102 64 bit results in @code{eax:edx} from 32 bit operands to do @code{long
6103 long} multiplies and 32-bit division by constants.
6106 Use a different function-calling convention, in which functions that
6107 take a fixed number of arguments return with the @code{ret} @var{num}
6108 instruction, which pops their arguments while returning. This saves one
6109 instruction in the caller since there is no need to pop the arguments
6112 You can specify that an individual function is called with this calling
6113 sequence with the function attribute @samp{stdcall}. You can also
6114 override the @samp{-mrtd} option by using the function attribute
6115 @samp{cdecl}. @xref{Function Attributes}.
6117 @strong{Warning:} this calling convention is incompatible with the one
6118 normally used on Unix, so you cannot use it if you need to call
6119 libraries compiled with the Unix compiler.
6121 Also, you must provide function prototypes for all functions that
6122 take variable numbers of arguments (including @code{printf});
6123 otherwise incorrect code will be generated for calls to those
6126 In addition, seriously incorrect code will result if you call a
6127 function with too many arguments. (Normally, extra arguments are
6128 harmlessly ignored.)
6130 @item -mreg-alloc=@var{regs}
6131 Control the default allocation order of integer registers. The
6132 string @var{regs} is a series of letters specifying a register. The
6133 supported letters are: @code{a} allocate EAX; @code{b} allocate EBX;
6134 @code{c} allocate ECX; @code{d} allocate EDX; @code{S} allocate ESI;
6135 @code{D} allocate EDI; @code{B} allocate EBP.
6137 @item -mregparm=@var{num}
6138 Control how many registers are used to pass integer arguments. By
6139 default, no registers are used to pass arguments, and at most 3
6140 registers can be used. You can control this behavior for a specific
6141 function by using the function attribute @samp{regparm}.
6142 @xref{Function Attributes}.
6144 @strong{Warning:} if you use this switch, and
6145 @var{num} is nonzero, then you must build all modules with the same
6146 value, including any libraries. This includes the system libraries and
6149 @item -malign-loops=@var{num}
6150 Align loops to a 2 raised to a @var{num} byte boundary. If
6151 @samp{-malign-loops} is not specified, the default is 2 unless
6152 gas 2.8 (or later) is being used in which case the default is
6153 to align the loop on a 16 byte boundary if it is less than 8
6156 @item -malign-jumps=@var{num}
6157 Align instructions that are only jumped to to a 2 raised to a @var{num}
6158 byte boundary. If @samp{-malign-jumps} is not specified, the default is
6159 2 if optimizing for a 386, and 4 if optimizing for a 486 unless
6160 gas 2.8 (or later) is being used in which case the default is
6161 to align the instruction on a 16 byte boundary if it is less
6164 @item -malign-functions=@var{num}
6165 Align the start of functions to a 2 raised to @var{num} byte boundary.
6166 If @samp{-malign-functions} is not specified, the default is 2 if optimizing
6167 for a 386, and 4 if optimizing for a 486.
6169 @item -mpreferred-stack-boundary=@var{num}
6170 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
6171 byte boundary. If @samp{-mpreferred-stack-boundary} is not specified,
6172 the default is 4 (16 bytes or 128 bits).
6174 The stack is required to be aligned on a 4 byte boundary. On Pentium
6175 and PentiumPro, @code{double} and @code{long double} values should be
6176 aligned to an 8 byte boundary (see @samp{-malign-double}) or suffer
6177 significant run time performance penalties. On Pentium III, the
6178 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
6179 penalties if it is not 16 byte aligned.
6181 To ensure proper alignment of this values on the stack, the stack boundary
6182 must be as aligned as that required by any value stored on the stack.
6183 Further, every function must be generated such that it keeps the stack
6184 aligned. Thus calling a function compiled with a higher preferred
6185 stack boundary from a function compiled with a lower preferred stack
6186 boundary will most likely misalign the stack. It is recommended that
6187 libraries that use callbacks always use the default setting.
6189 This extra alignment does consume extra stack space. Code that is sensitive
6190 to stack space usage, such as embedded systems and operating system kernels,
6191 may want to reduce the preferred alignment to
6192 @samp{-mpreferred-stack-boundary=2}.
6196 Use PUSH operations to store outgoing parameters. This method is shorter
6197 and usually equally fast as method using SUB/MOV operations and is enabled
6198 by default. In some cases disabling it may improve performance because of
6199 improved scheduling and reduced dependencies.
6201 @item -maccumulate-outgoing-args
6202 @kindex -maccumulate-outgoing-args
6203 If enabled, the maximum amount of space required for outgoing arguments will be
6204 computed in the function prologue. This in faster on most modern CPUs
6205 because of reduced dependencies, improved scheduling and reduced stack usage
6206 when preferred stack boundary is not equal to 2. The drawback is a notable
6207 increase in code size. This switch implies -mno-push-args.
6211 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
6212 on thread-safe exception handling must compile and link all code with the
6213 @samp{-mthreads} option. When compiling, @samp{-mthreads} defines
6214 @samp{-D_MT}; when linking, it links in a special thread helper library
6215 @samp{-lmingwthrd} which cleans up per thread exception handling data.
6217 @item -mno-align-stringops
6218 @kindex -mno-align-stringops
6219 Do not align destination of inlined string operations. This switch reduces
6220 code size and improves performance in case the destination is already aligned,
6221 but gcc don't know about it.
6223 @item -minline-all-stringops
6224 @kindex -minline-all-stringops
6225 By default GCC inlines string operations only when destination is known to be
6226 aligned at least to 4 byte boundary. This enables more inlining, increase code
6227 size, but may improve performance of code that depends on fast memcpy, strlen
6228 and memset for short lengths.
6232 @subsection HPPA Options
6233 @cindex HPPA Options
6235 These @samp{-m} options are defined for the HPPA family of computers:
6238 @item -march=@var{architecture type}
6239 Generate code for the specified architecture. The choices for
6240 @var{architecture type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
6241 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
6242 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
6243 architecture option for your machine. Code compiled for lower numbered
6244 architectures will run on higher numbered architectures, but not the
6247 PA 2.0 support currently requires gas snapshot 19990413 or later. The
6248 next release of binutils (current is 2.9.1) will probably contain PA 2.0
6252 @itemx -mpa-risc-1-1
6253 @itemx -mpa-risc-2-0
6254 Synonyms for -march=1.0, -march=1.1, and -march=2.0 respectively.
6257 Generate code suitable for big switch tables. Use this option only if
6258 the assembler/linker complain about out of range branches within a switch
6261 @item -mjump-in-delay
6262 Fill delay slots of function calls with unconditional jump instructions
6263 by modifying the return pointer for the function call to be the target
6264 of the conditional jump.
6266 @item -mdisable-fpregs
6267 Prevent floating point registers from being used in any manner. This is
6268 necessary for compiling kernels which perform lazy context switching of
6269 floating point registers. If you use this option and attempt to perform
6270 floating point operations, the compiler will abort.
6272 @item -mdisable-indexing
6273 Prevent the compiler from using indexing address modes. This avoids some
6274 rather obscure problems when compiling MIG generated code under MACH.
6276 @item -mno-space-regs
6277 Generate code that assumes the target has no space registers. This allows
6278 GCC to generate faster indirect calls and use unscaled index address modes.
6280 Such code is suitable for level 0 PA systems and kernels.
6282 @item -mfast-indirect-calls
6283 Generate code that assumes calls never cross space boundaries. This
6284 allows GCC to emit code which performs faster indirect calls.
6286 This option will not work in the presence of shared libraries or nested
6289 @item -mlong-load-store
6290 Generate 3-instruction load and store sequences as sometimes required by
6291 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
6294 @item -mportable-runtime
6295 Use the portable calling conventions proposed by HP for ELF systems.
6298 Enable the use of assembler directives only GAS understands.
6300 @item -mschedule=@var{cpu type}
6301 Schedule code according to the constraints for the machine type
6302 @var{cpu type}. The choices for @var{cpu type} are @samp{700}
6303 @samp{7100}, @samp{7100LC}, @samp{7200}, and @samp{8000}. Refer to
6304 @file{/usr/lib/sched.models} on an HP-UX system to determine the
6305 proper scheduling option for your machine.
6308 Enable the optimization pass in the HPUX linker. Note this makes symbolic
6309 debugging impossible. It also triggers a bug in the HPUX 8 and HPUX 9 linkers
6310 in which they give bogus error messages when linking some programs.
6313 Generate output containing library calls for floating point.
6314 @strong{Warning:} the requisite libraries are not available for all HPPA
6315 targets. Normally the facilities of the machine's usual C compiler are
6316 used, but this cannot be done directly in cross-compilation. You must make
6317 your own arrangements to provide suitable library functions for
6318 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
6319 does provide software floating point support.
6321 @samp{-msoft-float} changes the calling convention in the output file;
6322 therefore, it is only useful if you compile @emph{all} of a program with
6323 this option. In particular, you need to compile @file{libgcc.a}, the
6324 library that comes with GCC, with @samp{-msoft-float} in order for
6328 @node Intel 960 Options
6329 @subsection Intel 960 Options
6331 These @samp{-m} options are defined for the Intel 960 implementations:
6334 @item -m@var{cpu type}
6335 Assume the defaults for the machine type @var{cpu type} for some of
6336 the other options, including instruction scheduling, floating point
6337 support, and addressing modes. The choices for @var{cpu type} are
6338 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
6339 @samp{sa}, and @samp{sb}.
6345 The @samp{-mnumerics} option indicates that the processor does support
6346 floating-point instructions. The @samp{-msoft-float} option indicates
6347 that floating-point support should not be assumed.
6349 @item -mleaf-procedures
6350 @itemx -mno-leaf-procedures
6351 Do (or do not) attempt to alter leaf procedures to be callable with the
6352 @code{bal} instruction as well as @code{call}. This will result in more
6353 efficient code for explicit calls when the @code{bal} instruction can be
6354 substituted by the assembler or linker, but less efficient code in other
6355 cases, such as calls via function pointers, or using a linker that doesn't
6356 support this optimization.
6359 @itemx -mno-tail-call
6360 Do (or do not) make additional attempts (beyond those of the
6361 machine-independent portions of the compiler) to optimize tail-recursive
6362 calls into branches. You may not want to do this because the detection of
6363 cases where this is not valid is not totally complete. The default is
6364 @samp{-mno-tail-call}.
6366 @item -mcomplex-addr
6367 @itemx -mno-complex-addr
6368 Assume (or do not assume) that the use of a complex addressing mode is a
6369 win on this implementation of the i960. Complex addressing modes may not
6370 be worthwhile on the K-series, but they definitely are on the C-series.
6371 The default is currently @samp{-mcomplex-addr} for all processors except
6375 @itemx -mno-code-align
6376 Align code to 8-byte boundaries for faster fetching (or don't bother).
6377 Currently turned on by default for C-series implementations only.
6380 @item -mclean-linkage
6381 @itemx -mno-clean-linkage
6382 These options are not fully implemented.
6386 @itemx -mic2.0-compat
6387 @itemx -mic3.0-compat
6388 Enable compatibility with iC960 v2.0 or v3.0.
6392 Enable compatibility with the iC960 assembler.
6394 @item -mstrict-align
6395 @itemx -mno-strict-align
6396 Do not permit (do permit) unaligned accesses.
6399 Enable structure-alignment compatibility with Intel's gcc release version
6400 1.3 (based on gcc 1.37). This option implies @samp{-mstrict-align}.
6402 @item -mlong-double-64
6403 Implement type @samp{long double} as 64-bit floating point numbers.
6404 Without the option @samp{long double} is implemented by 80-bit
6405 floating point numbers. The only reason we have it because there is
6406 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
6407 is only useful for people using soft-float targets. Otherwise, we
6408 should recommend against use of it.
6412 @node DEC Alpha Options
6413 @subsection DEC Alpha Options
6415 These @samp{-m} options are defined for the DEC Alpha implementations:
6418 @item -mno-soft-float
6420 Use (do not use) the hardware floating-point instructions for
6421 floating-point operations. When @code{-msoft-float} is specified,
6422 functions in @file{libgcc1.c} will be used to perform floating-point
6423 operations. Unless they are replaced by routines that emulate the
6424 floating-point operations, or compiled in such a way as to call such
6425 emulations routines, these routines will issue floating-point
6426 operations. If you are compiling for an Alpha without floating-point
6427 operations, you must ensure that the library is built so as not to call
6430 Note that Alpha implementations without floating-point operations are
6431 required to have floating-point registers.
6435 Generate code that uses (does not use) the floating-point register set.
6436 @code{-mno-fp-regs} implies @code{-msoft-float}. If the floating-point
6437 register set is not used, floating point operands are passed in integer
6438 registers as if they were integers and floating-point results are passed
6439 in $0 instead of $f0. This is a non-standard calling sequence, so any
6440 function with a floating-point argument or return value called by code
6441 compiled with @code{-mno-fp-regs} must also be compiled with that
6444 A typical use of this option is building a kernel that does not use,
6445 and hence need not save and restore, any floating-point registers.
6448 The Alpha architecture implements floating-point hardware optimized for
6449 maximum performance. It is mostly compliant with the IEEE floating
6450 point standard. However, for full compliance, software assistance is
6451 required. This option generates code fully IEEE compliant code
6452 @emph{except} that the @var{inexact flag} is not maintained (see below).
6453 If this option is turned on, the CPP macro @code{_IEEE_FP} is defined
6454 during compilation. The option is a shorthand for: @samp{-D_IEEE_FP
6455 -mfp-trap-mode=su -mtrap-precision=i -mieee-conformant}. The resulting
6456 code is less efficient but is able to correctly support denormalized
6457 numbers and exceptional IEEE values such as not-a-number and plus/minus
6458 infinity. Other Alpha compilers call this option
6459 @code{-ieee_with_no_inexact}.
6461 @item -mieee-with-inexact
6462 @c overfull hbox here --bob 22 jul96
6463 @c original text between ignore ... end ignore
6465 This is like @samp{-mieee} except the generated code also maintains the
6466 IEEE @var{inexact flag}. Turning on this option causes the generated
6467 code to implement fully-compliant IEEE math. The option is a shorthand
6468 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus @samp{-mieee-conformant},
6469 @samp{-mfp-trap-mode=sui}, and @samp{-mtrap-precision=i}. On some Alpha
6470 implementations the resulting code may execute significantly slower than
6471 the code generated by default. Since there is very little code that
6472 depends on the @var{inexact flag}, you should normally not specify this
6473 option. Other Alpha compilers call this option
6474 @samp{-ieee_with_inexact}.
6476 @c changed paragraph
6477 This is like @samp{-mieee} except the generated code also maintains the
6478 IEEE @var{inexact flag}. Turning on this option causes the generated
6479 code to implement fully-compliant IEEE math. The option is a shorthand
6480 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus the three following:
6481 @samp{-mieee-conformant},
6482 @samp{-mfp-trap-mode=sui},
6483 and @samp{-mtrap-precision=i}.
6484 On some Alpha implementations the resulting code may execute
6485 significantly slower than the code generated by default. Since there
6486 is very little code that depends on the @var{inexact flag}, you should
6487 normally not specify this option. Other Alpha compilers call this
6488 option @samp{-ieee_with_inexact}.
6489 @c end changes to prevent overfull hboxes
6491 @item -mfp-trap-mode=@var{trap mode}
6492 This option controls what floating-point related traps are enabled.
6493 Other Alpha compilers call this option @samp{-fptm }@var{trap mode}.
6494 The trap mode can be set to one of four values:
6498 This is the default (normal) setting. The only traps that are enabled
6499 are the ones that cannot be disabled in software (e.g., division by zero
6503 In addition to the traps enabled by @samp{n}, underflow traps are enabled
6507 Like @samp{su}, but the instructions are marked to be safe for software
6508 completion (see Alpha architecture manual for details).
6511 Like @samp{su}, but inexact traps are enabled as well.
6514 @item -mfp-rounding-mode=@var{rounding mode}
6515 Selects the IEEE rounding mode. Other Alpha compilers call this option
6516 @samp{-fprm }@var{rounding mode}. The @var{rounding mode} can be one
6521 Normal IEEE rounding mode. Floating point numbers are rounded towards
6522 the nearest machine number or towards the even machine number in case
6526 Round towards minus infinity.
6529 Chopped rounding mode. Floating point numbers are rounded towards zero.
6532 Dynamic rounding mode. A field in the floating point control register
6533 (@var{fpcr}, see Alpha architecture reference manual) controls the
6534 rounding mode in effect. The C library initializes this register for
6535 rounding towards plus infinity. Thus, unless your program modifies the
6536 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
6539 @item -mtrap-precision=@var{trap precision}
6540 In the Alpha architecture, floating point traps are imprecise. This
6541 means without software assistance it is impossible to recover from a
6542 floating trap and program execution normally needs to be terminated.
6543 GCC can generate code that can assist operating system trap handlers
6544 in determining the exact location that caused a floating point trap.
6545 Depending on the requirements of an application, different levels of
6546 precisions can be selected:
6550 Program precision. This option is the default and means a trap handler
6551 can only identify which program caused a floating point exception.
6554 Function precision. The trap handler can determine the function that
6555 caused a floating point exception.
6558 Instruction precision. The trap handler can determine the exact
6559 instruction that caused a floating point exception.
6562 Other Alpha compilers provide the equivalent options called
6563 @samp{-scope_safe} and @samp{-resumption_safe}.
6565 @item -mieee-conformant
6566 This option marks the generated code as IEEE conformant. You must not
6567 use this option unless you also specify @samp{-mtrap-precision=i} and either
6568 @samp{-mfp-trap-mode=su} or @samp{-mfp-trap-mode=sui}. Its only effect
6569 is to emit the line @samp{.eflag 48} in the function prologue of the
6570 generated assembly file. Under DEC Unix, this has the effect that
6571 IEEE-conformant math library routines will be linked in.
6573 @item -mbuild-constants
6574 Normally GCC examines a 32- or 64-bit integer constant to
6575 see if it can construct it from smaller constants in two or three
6576 instructions. If it cannot, it will output the constant as a literal and
6577 generate code to load it from the data segment at runtime.
6579 Use this option to require GCC to construct @emph{all} integer constants
6580 using code, even if it takes more instructions (the maximum is six).
6582 You would typically use this option to build a shared library dynamic
6583 loader. Itself a shared library, it must relocate itself in memory
6584 before it can find the variables and constants in its own data segment.
6588 Select whether to generate code to be assembled by the vendor-supplied
6589 assembler (@samp{-malpha-as}) or by the GNU assembler @samp{-mgas}.
6597 Indicate whether GCC should generate code to use the optional BWX,
6598 CIX, and MAX instruction sets. The default is to use the instruction sets
6599 supported by the CPU type specified via @samp{-mcpu=} option or that
6600 of the CPU on which GCC was built if none was specified.
6602 @item -mcpu=@var{cpu_type}
6603 Set the instruction set, register set, and instruction scheduling
6604 parameters for machine type @var{cpu_type}. You can specify either the
6605 @samp{EV} style name or the corresponding chip number. GCC
6606 supports scheduling parameters for the EV4 and EV5 family of processors
6607 and will choose the default values for the instruction set from
6608 the processor you specify. If you do not specify a processor type,
6609 GCC will default to the processor on which the compiler was built.
6611 Supported values for @var{cpu_type} are
6616 Schedules as an EV4 and has no instruction set extensions.
6620 Schedules as an EV5 and has no instruction set extensions.
6624 Schedules as an EV5 and supports the BWX extension.
6629 Schedules as an EV5 and supports the BWX and MAX extensions.
6633 Schedules as an EV5 (until Digital releases the scheduling parameters
6634 for the EV6) and supports the BWX, CIX, and MAX extensions.
6637 @item -mmemory-latency=@var{time}
6638 Sets the latency the scheduler should assume for typical memory
6639 references as seen by the application. This number is highly
6640 dependent on the memory access patterns used by the application
6641 and the size of the external cache on the machine.
6643 Valid options for @var{time} are
6647 A decimal number representing clock cycles.
6653 The compiler contains estimates of the number of clock cycles for
6654 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
6655 (also called Dcache, Scache, and Bcache), as well as to main memory.
6656 Note that L3 is only valid for EV5.
6661 @node Clipper Options
6662 @subsection Clipper Options
6664 These @samp{-m} options are defined for the Clipper implementations:
6668 Produce code for a C300 Clipper processor. This is the default.
6671 Produce code for a C400 Clipper processor i.e. use floating point
6675 @node H8/300 Options
6676 @subsection H8/300 Options
6678 These @samp{-m} options are defined for the H8/300 implementations:
6682 Shorten some address references at link time, when possible; uses the
6683 linker option @samp{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
6684 ld.info, Using ld}, for a fuller description.
6687 Generate code for the H8/300H.
6690 Generate code for the H8/S.
6693 Generate code for the H8/S2600. This switch must be used with -ms.
6696 Make @code{int} data 32 bits by default.
6699 On the H8/300H and H8/S, use the same alignment rules as for the H8/300.
6700 The default for the H8/300H and H8/S is to align longs and floats on 4
6702 @samp{-malign-300} causes them to be aligned on 2 byte boundaries.
6703 This option has no effect on the H8/300.
6707 @subsection SH Options
6709 These @samp{-m} options are defined for the SH implementations:
6713 Generate code for the SH1.
6716 Generate code for the SH2.
6719 Generate code for the SH3.
6722 Generate code for the SH3e.
6725 Generate code for the SH4 without a floating-point unit.
6727 @item -m4-single-only
6728 Generate code for the SH4 with a floating-point unit that only
6729 supports single-precision arithmentic.
6732 Generate code for the SH4 assuming the floating-point unit is in
6733 single-precision mode by default.
6736 Generate code for the SH4.
6739 Compile code for the processor in big endian mode.
6742 Compile code for the processor in little endian mode.
6745 Align doubles at 64 bit boundaries. Note that this changes the calling
6746 conventions, and thus some functions from the standard C library will
6747 not work unless you recompile it first with -mdalign.
6750 Shorten some address references at link time, when possible; uses the
6751 linker option @samp{-relax}.
6754 Use 32-bit offsets in @code{switch} tables. The default is to use
6758 Enable the use of the instruction @code{fmovd}.
6761 Comply with the calling conventions defined by Hitachi.
6764 Mark the @code{MAC} register as call-clobbered, even if
6765 @code{-mhitachi} is given.
6768 Dump instruction size and location in the assembly code.
6771 This option is deprecated. It pads structures to multiple of 4 bytes,
6772 which is incompatible with the SH ABI.
6775 Optimize for space instead of speed. Implied by @code{-Os}.
6778 When generating position-independent code, emit function calls using
6779 the Global Offset Table instead of the Procedure Linkage Table.
6782 Generate a library function call to invalidate instruction cache
6783 entries, after fixing up a trampoline. This library function call
6784 doesn't assume it can write to the whole memory address space. This
6785 is the default when the target is @code{sh-*-linux*}.
6788 @node System V Options
6789 @subsection Options for System V
6791 These additional options are available on System V Release 4 for
6792 compatibility with other compilers on those systems:
6796 Create a shared object.
6797 It is recommended that @samp{-symbolic} or @samp{-shared} be used instead.
6800 Identify the versions of each tool used by the compiler, in a
6801 @code{.ident} assembler directive in the output.
6804 Refrain from adding @code{.ident} directives to the output file (this is
6807 @item -YP,@var{dirs}
6808 Search the directories @var{dirs}, and no others, for libraries
6809 specified with @samp{-l}.
6812 Look in the directory @var{dir} to find the M4 preprocessor.
6813 The assembler uses this option.
6814 @c This is supposed to go with a -Yd for predefined M4 macro files, but
6815 @c the generic assembler that comes with Solaris takes just -Ym.
6818 @node TMS320C3x/C4x Options
6819 @subsection TMS320C3x/C4x Options
6820 @cindex TMS320C3x/C4x Options
6822 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
6826 @item -mcpu=@var{cpu_type}
6827 Set the instruction set, register set, and instruction scheduling
6828 parameters for machine type @var{cpu_type}. Supported values for
6829 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
6830 @samp{c44}. The default is @samp{c40} to generate code for the
6835 @itemx -msmall-memory
6837 Generates code for the big or small memory model. The small memory
6838 model assumed that all data fits into one 64K word page. At run-time
6839 the data page (DP) register must be set to point to the 64K page
6840 containing the .bss and .data program sections. The big memory model is
6841 the default and requires reloading of the DP register for every direct
6846 Allow (disallow) allocation of general integer operands into the block
6851 Enable (disable) generation of code using decrement and branch,
6852 DBcond(D), instructions. This is enabled by default for the C4x. To be
6853 on the safe side, this is disabled for the C3x, since the maximum
6854 iteration count on the C3x is 2^23 + 1 (but who iterates loops more than
6855 2^23 times on the C3x?). Note that GCC will try to reverse a loop so
6856 that it can utilise the decrement and branch instruction, but will give
6857 up if there is more than one memory reference in the loop. Thus a loop
6858 where the loop counter is decremented can generate slightly more
6859 efficient code, in cases where the RPTB instruction cannot be utilised.
6861 @item -mdp-isr-reload
6863 Force the DP register to be saved on entry to an interrupt service
6864 routine (ISR), reloaded to point to the data section, and restored on
6865 exit from the ISR. This should not be required unless someone has
6866 violated the small memory model by modifying the DP register, say within
6871 For the C3x use the 24-bit MPYI instruction for integer multiplies
6872 instead of a library call to guarantee 32-bit results. Note that if one
6873 of the operands is a constant, then the multiplication will be performed
6874 using shifts and adds. If the -mmpyi option is not specified for the C3x,
6875 then squaring operations are performed inline instead of a library call.
6878 @itemx -mno-fast-fix
6879 The C3x/C4x FIX instruction to convert a floating point value to an
6880 integer value chooses the nearest integer less than or equal to the
6881 floating point value rather than to the nearest integer. Thus if the
6882 floating point number is negative, the result will be incorrectly
6883 truncated an additional code is necessary to detect and correct this
6884 case. This option can be used to disable generation of the additional
6885 code required to correct the result.
6889 Enable (disable) generation of repeat block sequences using the RPTB
6890 instruction for zero overhead looping. The RPTB construct is only used
6891 for innermost loops that do not call functions or jump across the loop
6892 boundaries. There is no advantage having nested RPTB loops due to the
6893 overhead required to save and restore the RC, RS, and RE registers.
6894 This is enabled by default with -O2.
6896 @item -mrpts=@var{count}
6898 Enable (disable) the use of the single instruction repeat instruction
6899 RPTS. If a repeat block contains a single instruction, and the loop
6900 count can be guaranteed to be less than the value @var{count}, GCC will
6901 emit a RPTS instruction instead of a RPTB. If no value is specified,
6902 then a RPTS will be emitted even if the loop count cannot be determined
6903 at compile time. Note that the repeated instruction following RPTS does
6904 not have to be reloaded from memory each iteration, thus freeing up the
6905 CPU buses for operands. However, since interrupts are blocked by this
6906 instruction, it is disabled by default.
6908 @item -mloop-unsigned
6909 @itemx -mno-loop-unsigned
6910 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
6911 is 2^31 + 1 since these instructions test if the iteration count is
6912 negative to terminate the loop. If the iteration count is unsigned
6913 there is a possibility than the 2^31 + 1 maximum iteration count may be
6914 exceeded. This switch allows an unsigned iteration count.
6917 Try to emit an assembler syntax that the TI assembler (asm30) is happy
6918 with. This also enforces compatibility with the API employed by the TI
6919 C3x C compiler. For example, long doubles are passed as structures
6920 rather than in floating point registers.
6924 Generate code that uses registers (stack) for passing arguments to functions.
6925 By default, arguments are passed in registers where possible rather
6926 than by pushing arguments on to the stack.
6928 @item -mparallel-insns
6929 @itemx -mno-parallel-insns
6930 Allow the generation of parallel instructions. This is enabled by
6933 @item -mparallel-mpy
6934 @itemx -mno-parallel-mpy
6935 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
6936 provided -mparallel-insns is also specified. These instructions have
6937 tight register constraints which can pessimize the code generation
6943 @subsection V850 Options
6944 @cindex V850 Options
6946 These @samp{-m} options are defined for V850 implementations:
6950 @itemx -mno-long-calls
6951 Treat all calls as being far away (near). If calls are assumed to be
6952 far away, the compiler will always load the functions address up into a
6953 register, and call indirect through the pointer.
6957 Do not optimize (do optimize) basic blocks that use the same index
6958 pointer 4 or more times to copy pointer into the @code{ep} register, and
6959 use the shorter @code{sld} and @code{sst} instructions. The @samp{-mep}
6960 option is on by default if you optimize.
6962 @item -mno-prolog-function
6963 @itemx -mprolog-function
6964 Do not use (do use) external functions to save and restore registers at
6965 the prolog and epilog of a function. The external functions are slower,
6966 but use less code space if more than one function saves the same number
6967 of registers. The @samp{-mprolog-function} option is on by default if
6971 Try to make the code as small as possible. At present, this just turns
6972 on the @samp{-mep} and @samp{-mprolog-function} options.
6975 Put static or global variables whose size is @var{n} bytes or less into
6976 the tiny data area that register @code{ep} points to. The tiny data
6977 area can hold up to 256 bytes in total (128 bytes for byte references).
6980 Put static or global variables whose size is @var{n} bytes or less into
6981 the small data area that register @code{gp} points to. The small data
6982 area can hold up to 64 kilobytes.
6985 Put static or global variables whose size is @var{n} bytes or less into
6986 the first 32 kilobytes of memory.
6989 Specify that the target processor is the V850.
6992 Generate code suitable for big switch tables. Use this option only if
6993 the assembler/linker complain about out of range branches within a switch
6998 @subsection ARC Options
7001 These options are defined for ARC implementations:
7005 Compile code for little endian mode. This is the default.
7008 Compile code for big endian mode.
7011 Prepend the name of the cpu to all public symbol names.
7012 In multiple-processor systems, there are many ARC variants with different
7013 instruction and register set characteristics. This flag prevents code
7014 compiled for one cpu to be linked with code compiled for another.
7015 No facility exists for handling variants that are "almost identical".
7016 This is an all or nothing option.
7018 @item -mcpu=@var{cpu}
7019 Compile code for ARC variant @var{cpu}.
7020 Which variants are supported depend on the configuration.
7021 All variants support @samp{-mcpu=base}, this is the default.
7023 @item -mtext=@var{text section}
7024 @itemx -mdata=@var{data section}
7025 @itemx -mrodata=@var{readonly data section}
7026 Put functions, data, and readonly data in @var{text section},
7027 @var{data section}, and @var{readonly data section} respectively
7028 by default. This can be overridden with the @code{section} attribute.
7029 @xref{Variable Attributes}.
7034 @subsection NS32K Options
7035 @cindex NS32K options
7037 These are the @samp{-m} options defined for the 32000 series. The default
7038 values for these options depends on which style of 32000 was selected when
7039 the compiler was configured; the defaults for the most common choices are
7045 Generate output for a 32032. This is the default
7046 when the compiler is configured for 32032 and 32016 based systems.
7050 Generate output for a 32332. This is the default
7051 when the compiler is configured for 32332-based systems.
7055 Generate output for a 32532. This is the default
7056 when the compiler is configured for 32532-based systems.
7059 Generate output containing 32081 instructions for floating point.
7060 This is the default for all systems.
7063 Generate output containing 32381 instructions for floating point. This
7064 also implies @samp{-m32081}. The 32381 is only compatible with the 32332
7065 and 32532 cpus. This is the default for the pc532-netbsd configuration.
7068 Try and generate multiply-add floating point instructions @code{polyF}
7069 and @code{dotF}. This option is only available if the @samp{-m32381}
7070 option is in effect. Using these instructions requires changes to to
7071 register allocation which generally has a negative impact on
7072 performance. This option should only be enabled when compiling code
7073 particularly likely to make heavy use of multiply-add instructions.
7076 Do not try and generate multiply-add floating point instructions
7077 @code{polyF} and @code{dotF}. This is the default on all platforms.
7080 Generate output containing library calls for floating point.
7081 @strong{Warning:} the requisite libraries may not be available.
7084 Do not use the bit-field instructions. On some machines it is faster to
7085 use shifting and masking operations. This is the default for the pc532.
7088 Do use the bit-field instructions. This is the default for all platforms
7092 Use a different function-calling convention, in which functions
7093 that take a fixed number of arguments return pop their
7094 arguments on return with the @code{ret} instruction.
7096 This calling convention is incompatible with the one normally
7097 used on Unix, so you cannot use it if you need to call libraries
7098 compiled with the Unix compiler.
7100 Also, you must provide function prototypes for all functions that
7101 take variable numbers of arguments (including @code{printf});
7102 otherwise incorrect code will be generated for calls to those
7105 In addition, seriously incorrect code will result if you call a
7106 function with too many arguments. (Normally, extra arguments are
7107 harmlessly ignored.)
7109 This option takes its name from the 680x0 @code{rtd} instruction.
7113 Use a different function-calling convention where the first two arguments
7114 are passed in registers.
7116 This calling convention is incompatible with the one normally
7117 used on Unix, so you cannot use it if you need to call libraries
7118 compiled with the Unix compiler.
7121 Do not pass any arguments in registers. This is the default for all
7125 It is OK to use the sb as an index register which is always loaded with
7126 zero. This is the default for the pc532-netbsd target.
7129 The sb register is not available for use or has not been initialized to
7130 zero by the run time system. This is the default for all targets except
7131 the pc532-netbsd. It is also implied whenever @samp{-mhimem} or
7132 @samp{-fpic} is set.
7135 Many ns32000 series addressing modes use displacements of up to 512MB.
7136 If an address is above 512MB then displacements from zero can not be used.
7137 This option causes code to be generated which can be loaded above 512MB.
7138 This may be useful for operating systems or ROM code.
7141 Assume code will be loaded in the first 512MB of virtual address space.
7142 This is the default for all platforms.
7148 @subsection AVR Options
7151 These options are defined for AVR implementations:
7154 @item -mmcu=@var{mcu}
7155 Specify ATMEL AVR mcu (at90s23xx,attiny22,at90s44xx,at90s85xx,atmega603,
7159 Output instruction size's to the asm file
7161 @item -minit-stack=@var{N}
7162 Specify the initial stack address
7164 @item -mno-interrupts
7165 Generated code is not compatible with hardware interrupts.
7166 Code size will be smaller.
7168 @item -mcall-prologues
7169 Functions prologues/epilogues expanded as call to appropriate
7170 subroutines. Code size will be smaller.
7174 @subsection MCore Options
7175 @cindex MCore options
7177 These are the @samp{-m} options defined for the Motorola M*Core
7185 Inline constants into the code stream if it can be done in two
7186 instructions or less.
7191 Use the divide instruction. (Enabled by default).
7193 @item -mrelax-immediate
7194 @itemx -mrelax-immediate
7195 @itemx -mno-relax-immediate
7196 Allow arbitrary sized immediates in bit operations.
7198 @item -mwide-bitfields
7199 @itemx -mwide-bitfields
7200 @itemx -mno-wide-bitfields
7201 Always treat bitfields as int-sized.
7203 @item -m4byte-functions
7204 @itemx -m4byte-functions
7205 @itemx -mno-4byte-functions
7206 Force all functions to be aligned to a four byte boundary.
7208 @item -mcallgraph-data
7209 @itemx -mcallgraph-data
7210 @itemx -mno-callgraph-data
7211 Emit callgraph information.
7215 @itemx -mno-slow-bytes
7216 Prefer word access when reading byte quantities.
7218 @item -mlittle-endian
7219 @itemx -mlittle-endian
7221 Generate code for a little endian target.
7226 Generate code for the 210 processor.
7230 @subsection D30V Options
7231 @cindex D30V Options
7233 These @samp{-m} options are defined for D30V implementations:
7237 Link the @samp{.text}, @samp{.data}, @samp{.bss}, @samp{.strings},
7238 @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections into external
7239 memory, which starts at location @code{0x80000000}.
7242 Same as the @samp{-mextmem} switch.
7245 Link the @samp{.text} section into onchip text memory, which starts at
7246 location @code{0x0}. Also link @samp{.data}, @samp{.bss},
7247 @samp{.strings}, @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections
7248 into onchip data memory, which starts at location @code{0x20000000}.
7250 @item -mno-asm-optimize
7251 @itemx -masm-optimize
7252 Disable (enable) passing @samp{-O} to the assembler when optimizing.
7253 The assembler uses the @samp{-O} option to automatically parallelize
7254 adjacent short instructions where possible.
7256 @item -mbranch-cost=@var{n}
7257 Increase the internal costs of branches to @var{n}. Higher costs means
7258 that the compiler will issue more instructions to avoid doing a branch.
7261 @item -mcond-exec=@var{n}
7262 Specify the maximum number of conditionally executed instructions that
7263 replace a branch. The default is 4.
7266 @node Code Gen Options
7267 @section Options for Code Generation Conventions
7268 @cindex code generation conventions
7269 @cindex options, code generation
7270 @cindex run-time options
7272 These machine-independent options control the interface conventions
7273 used in code generation.
7275 Most of them have both positive and negative forms; the negative form
7276 of @samp{-ffoo} would be @samp{-fno-foo}. In the table below, only
7277 one of the forms is listed---the one which is not the default. You
7278 can figure out the other form by either removing @samp{no-} or adding
7283 Enable exception handling. Generates extra code needed to propagate
7284 exceptions. For some targets, this implies GNU CC will generate frame
7285 unwind information for all functions, which can produce significant data
7286 size overhead, although it does not affect execution. If you do not
7287 specify this option, GNU CC will enable it by default for languages like
7288 C++ which normally require exception handling, and disable itfor
7289 languages like C that do not normally require it. However, you may need
7290 to enable this option when compiling C code that needs to interoperate
7291 properly with exception handlers written in C++. You may also wish to
7292 disable this option if you are compiling older C++ programs that don't
7293 use exception handling.
7295 @item -funwind-tables
7296 Similar to @code{-fexceptions}, except that it will just generate any needed
7297 static data, but will not affect the generated code in any other way.
7298 You will normally not enable this option; instead, a language processor
7299 that needs this handling would enable it on your behalf.
7301 @item -fpcc-struct-return
7302 Return ``short'' @code{struct} and @code{union} values in memory like
7303 longer ones, rather than in registers. This convention is less
7304 efficient, but it has the advantage of allowing intercallability between
7305 GCC-compiled files and files compiled with other compilers.
7307 The precise convention for returning structures in memory depends
7308 on the target configuration macros.
7310 Short structures and unions are those whose size and alignment match
7311 that of some integer type.
7313 @item -freg-struct-return
7314 Use the convention that @code{struct} and @code{union} values are
7315 returned in registers when possible. This is more efficient for small
7316 structures than @samp{-fpcc-struct-return}.
7318 If you specify neither @samp{-fpcc-struct-return} nor its contrary
7319 @samp{-freg-struct-return}, GCC defaults to whichever convention is
7320 standard for the target. If there is no standard convention, GCC
7321 defaults to @samp{-fpcc-struct-return}, except on targets where GCC
7322 is the principal compiler. In those cases, we can choose the standard,
7323 and we chose the more efficient register return alternative.
7326 Allocate to an @code{enum} type only as many bytes as it needs for the
7327 declared range of possible values. Specifically, the @code{enum} type
7328 will be equivalent to the smallest integer type which has enough room.
7330 @item -fshort-double
7331 Use the same size for @code{double} as for @code{float}.
7334 Requests that the data and non-@code{const} variables of this
7335 compilation be shared data rather than private data. The distinction
7336 makes sense only on certain operating systems, where shared data is
7337 shared between processes running the same program, while private data
7338 exists in one copy per process.
7341 Allocate even uninitialized global variables in the data section of the
7342 object file, rather than generating them as common blocks. This has the
7343 effect that if the same variable is declared (without @code{extern}) in
7344 two different compilations, you will get an error when you link them.
7345 The only reason this might be useful is if you wish to verify that the
7346 program will work on other systems which always work this way.
7349 Ignore the @samp{#ident} directive.
7351 @item -fno-gnu-linker
7352 Do not output global initializations (such as C++ constructors and
7353 destructors) in the form used by the GNU linker (on systems where the GNU
7354 linker is the standard method of handling them). Use this option when
7355 you want to use a non-GNU linker, which also requires using the
7356 @code{collect2} program to make sure the system linker includes
7357 constructors and destructors. (@code{collect2} is included in the GCC
7358 distribution.) For systems which @emph{must} use @code{collect2}, the
7359 compiler driver @code{gcc} is configured to do this automatically.
7361 @item -finhibit-size-directive
7362 Don't output a @code{.size} assembler directive, or anything else that
7363 would cause trouble if the function is split in the middle, and the
7364 two halves are placed at locations far apart in memory. This option is
7365 used when compiling @file{crtstuff.c}; you should not need to use it
7369 Put extra commentary information in the generated assembly code to
7370 make it more readable. This option is generally only of use to those
7371 who actually need to read the generated assembly code (perhaps while
7372 debugging the compiler itself).
7374 @samp{-fno-verbose-asm}, the default, causes the
7375 extra information to be omitted and is useful when comparing two assembler
7379 Consider all memory references through pointers to be volatile.
7381 @item -fvolatile-global
7382 Consider all memory references to extern and global data items to
7383 be volatile. GCC does not consider static data items to be volatile
7384 because of this switch.
7386 @item -fvolatile-static
7387 Consider all memory references to static data to be volatile.
7390 @cindex global offset table
7392 Generate position-independent code (PIC) suitable for use in a shared
7393 library, if supported for the target machine. Such code accesses all
7394 constant addresses through a global offset table (GOT). The dynamic
7395 loader resolves the GOT entries when the program starts (the dynamic
7396 loader is not part of GCC; it is part of the operating system). If
7397 the GOT size for the linked executable exceeds a machine-specific
7398 maximum size, you get an error message from the linker indicating that
7399 @samp{-fpic} does not work; in that case, recompile with @samp{-fPIC}
7400 instead. (These maximums are 16k on the m88k, 8k on the Sparc, and 32k
7401 on the m68k and RS/6000. The 386 has no such limit.)
7403 Position-independent code requires special support, and therefore works
7404 only on certain machines. For the 386, GCC supports PIC for System V
7405 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
7406 position-independent.
7409 If supported for the target machine, emit position-independent code,
7410 suitable for dynamic linking and avoiding any limit on the size of the
7411 global offset table. This option makes a difference on the m68k, m88k,
7414 Position-independent code requires special support, and therefore works
7415 only on certain machines.
7417 @item -ffixed-@var{reg}
7418 Treat the register named @var{reg} as a fixed register; generated code
7419 should never refer to it (except perhaps as a stack pointer, frame
7420 pointer or in some other fixed role).
7422 @var{reg} must be the name of a register. The register names accepted
7423 are machine-specific and are defined in the @code{REGISTER_NAMES}
7424 macro in the machine description macro file.
7426 This flag does not have a negative form, because it specifies a
7429 @item -fcall-used-@var{reg}
7430 Treat the register named @var{reg} as an allocable register that is
7431 clobbered by function calls. It may be allocated for temporaries or
7432 variables that do not live across a call. Functions compiled this way
7433 will not save and restore the register @var{reg}.
7435 It is an error to used this flag with the frame pointer or stack pointer.
7436 Use of this flag for other registers that have fixed pervasive roles in
7437 the machine's execution model will produce disastrous results.
7439 This flag does not have a negative form, because it specifies a
7442 @item -fcall-saved-@var{reg}
7443 Treat the register named @var{reg} as an allocable register saved by
7444 functions. It may be allocated even for temporaries or variables that
7445 live across a call. Functions compiled this way will save and restore
7446 the register @var{reg} if they use it.
7448 It is an error to used this flag with the frame pointer or stack pointer.
7449 Use of this flag for other registers that have fixed pervasive roles in
7450 the machine's execution model will produce disastrous results.
7452 A different sort of disaster will result from the use of this flag for
7453 a register in which function values may be returned.
7455 This flag does not have a negative form, because it specifies a
7459 Pack all structure members together without holes. Usually you would
7460 not want to use this option, since it makes the code suboptimal, and
7461 the offsets of structure members won't agree with system libraries.
7463 @item -fcheck-memory-usage
7464 Generate extra code to check each memory access. GCC will generate
7465 code that is suitable for a detector of bad memory accesses such as
7468 Normally, you should compile all, or none, of your code with this option.
7470 If you do mix code compiled with and without this option,
7471 you must ensure that all code that has side effects
7472 and that is called by code compiled with this option
7473 is, itself, compiled with this option.
7474 If you do not, you might get erroneous messages from the detector.
7476 If you use functions from a library that have side-effects (such as
7477 @code{read}), you might not be able to recompile the library and
7478 specify this option. In that case, you can enable the
7479 @samp{-fprefix-function-name} option, which requests GCC to encapsulate
7480 your code and make other functions look as if they were compiled with
7481 @samp{-fcheck-memory-usage}. This is done by calling ``stubs'',
7482 which are provided by the detector. If you cannot find or build
7483 stubs for every function you call, you might have to specify
7484 @samp{-fcheck-memory-usage} without @samp{-fprefix-function-name}.
7486 If you specify this option, you can not use the @code{asm} or
7487 @code{__asm__} keywords in functions with memory checking enabled. GNU
7488 CC cannot understand what the @code{asm} statement may do, and therefore
7489 cannot generate the appropriate code, so it will reject it. However, if
7490 you specify the function attribute @code{no_check_memory_usage} (see
7491 @pxref{Function Attributes}, GNU CC will disable memory checking within a
7492 function; you may use @code{asm} statements inside such functions. You
7493 may have an inline expansion of a non-checked function within a checked
7494 function; in that case GNU CC will not generate checks for the inlined
7495 function's memory accesses.
7497 If you move your @code{asm} statements to non-checked inline functions
7498 and they do access memory, you can add calls to the support code in your
7499 inline function, to indicate any reads, writes, or copies being done.
7500 These calls would be similar to those done in the stubs described above.
7502 @item -fprefix-function-name
7503 Request GCC to add a prefix to the symbols generated for function names.
7504 GCC adds a prefix to the names of functions defined as well as
7505 functions called. Code compiled with this option and code compiled
7506 without the option can't be linked together, unless stubs are used.
7508 If you compile the following code with @samp{-fprefix-function-name}
7510 extern void bar (int);
7519 GCC will compile the code as if it was written:
7521 extern void prefix_bar (int);
7525 return prefix_bar (a + 5);
7528 This option is designed to be used with @samp{-fcheck-memory-usage}.
7530 @item -finstrument-functions
7531 Generate instrumentation calls for entry and exit to functions. Just
7532 after function entry and just before function exit, the following
7533 profiling functions will be called with the address of the current
7534 function and its call site. (On some platforms,
7535 @code{__builtin_return_address} does not work beyond the current
7536 function, so the call site information may not be available to the
7537 profiling functions otherwise.)
7540 void __cyg_profile_func_enter (void *this_fn, void *call_site);
7541 void __cyg_profile_func_exit (void *this_fn, void *call_site);
7544 The first argument is the address of the start of the current function,
7545 which may be looked up exactly in the symbol table.
7547 This instrumentation is also done for functions expanded inline in other
7548 functions. The profiling calls will indicate where, conceptually, the
7549 inline function is entered and exited. This means that addressable
7550 versions of such functions must be available. If all your uses of a
7551 function are expanded inline, this may mean an additional expansion of
7552 code size. If you use @samp{extern inline} in your C code, an
7553 addressable version of such functions must be provided. (This is
7554 normally the case anyways, but if you get lucky and the optimizer always
7555 expands the functions inline, you might have gotten away without
7556 providing static copies.)
7558 A function may be given the attribute @code{no_instrument_function}, in
7559 which case this instrumentation will not be done. This can be used, for
7560 example, for the profiling functions listed above, high-priority
7561 interrupt routines, and any functions from which the profiling functions
7562 cannot safely be called (perhaps signal handlers, if the profiling
7563 routines generate output or allocate memory).
7566 Generate code to verify that you do not go beyond the boundary of the
7567 stack. You should specify this flag if you are running in an
7568 environment with multiple threads, but only rarely need to specify it in
7569 a single-threaded environment since stack overflow is automatically
7570 detected on nearly all systems if there is only one stack.
7572 Note that this switch does not actually cause checking to be done; the
7573 operating system must do that. The switch causes generation of code
7574 to ensure that the operating system sees the stack being extended.
7576 @item -fstack-limit-register=@var{reg}
7577 @itemx -fstack-limit-symbol=@var{sym}
7578 @itemx -fno-stack-limit
7579 Generate code to ensure that the stack does not grow beyond a certain value,
7580 either the value of a register or the address of a symbol. If the stack
7581 would grow beyond the value, a signal is raised. For most targets,
7582 the signal is raised before the stack overruns the boundary, so
7583 it is possible to catch the signal without taking special precautions.
7585 For instance, if the stack starts at address @samp{0x80000000} and grows
7586 downwards you can use the flags
7587 @samp{-fstack-limit-symbol=__stack_limit}
7588 @samp{-Wl,--defsym,__stack_limit=0x7ffe0000} which will enforce a stack
7591 @cindex aliasing of parameters
7592 @cindex parameters, aliased
7593 @item -fargument-alias
7594 @itemx -fargument-noalias
7595 @itemx -fargument-noalias-global
7596 Specify the possible relationships among parameters and between
7597 parameters and global data.
7599 @samp{-fargument-alias} specifies that arguments (parameters) may
7600 alias each other and may alias global storage.
7601 @samp{-fargument-noalias} specifies that arguments do not alias
7602 each other, but may alias global storage.
7603 @samp{-fargument-noalias-global} specifies that arguments do not
7604 alias each other and do not alias global storage.
7606 Each language will automatically use whatever option is required by
7607 the language standard. You should not need to use these options yourself.
7609 @item -fleading-underscore
7610 This option and its counterpart, -fno-leading-underscore, forcibly
7611 change the way C symbols are represented in the object file. One use
7612 is to help link with legacy assembly code.
7614 Be warned that you should know what you are doing when invoking this
7615 option, and that not all targets provide complete support for it.
7618 @node Environment Variables
7619 @section Environment Variables Affecting GCC
7620 @cindex environment variables
7622 This section describes several environment variables that affect how GCC
7623 operates. Some of them work by specifying directories or prefixes to use
7624 when searching for various kinds of files. Some are used to specify other
7625 aspects of the compilation environment.
7628 Note that you can also specify places to search using options such as
7629 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
7630 take precedence over places specified using environment variables, which
7631 in turn take precedence over those specified by the configuration of GCC.
7635 Note that you can also specify places to search using options such as
7636 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
7637 take precedence over places specified using environment variables, which
7638 in turn take precedence over those specified by the configuration of GCC.
7645 @c @itemx LC_COLLATE
7647 @c @itemx LC_MONETARY
7648 @c @itemx LC_NUMERIC
7653 @c @findex LC_COLLATE
7655 @c @findex LC_MONETARY
7656 @c @findex LC_NUMERIC
7660 These environment variables control the way that GCC uses
7661 localization information that allow GCC to work with different
7662 national conventions. GCC inspects the locale categories
7663 @code{LC_CTYPE} and @code{LC_MESSAGES} if it has been configured to do
7664 so. These locale categories can be set to any value supported by your
7665 installation. A typical value is @samp{en_UK} for English in the United
7668 The @code{LC_CTYPE} environment variable specifies character
7669 classification. GCC uses it to determine the character boundaries in
7670 a string; this is needed for some multibyte encodings that contain quote
7671 and escape characters that would otherwise be interpreted as a string
7674 The @code{LC_MESSAGES} environment variable specifies the language to
7675 use in diagnostic messages.
7677 If the @code{LC_ALL} environment variable is set, it overrides the value
7678 of @code{LC_CTYPE} and @code{LC_MESSAGES}; otherwise, @code{LC_CTYPE}
7679 and @code{LC_MESSAGES} default to the value of the @code{LANG}
7680 environment variable. If none of these variables are set, GCC
7681 defaults to traditional C English behavior.
7685 If @code{TMPDIR} is set, it specifies the directory to use for temporary
7686 files. GCC uses temporary files to hold the output of one stage of
7687 compilation which is to be used as input to the next stage: for example,
7688 the output of the preprocessor, which is the input to the compiler
7691 @item GCC_EXEC_PREFIX
7692 @findex GCC_EXEC_PREFIX
7693 If @code{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
7694 names of the subprograms executed by the compiler. No slash is added
7695 when this prefix is combined with the name of a subprogram, but you can
7696 specify a prefix that ends with a slash if you wish.
7698 If @code{GCC_EXEC_PREFIX} is not set, GNU CC will attempt to figure out
7699 an appropriate prefix to use based on the pathname it was invoked with.
7701 If GCC cannot find the subprogram using the specified prefix, it
7702 tries looking in the usual places for the subprogram.
7704 The default value of @code{GCC_EXEC_PREFIX} is
7705 @file{@var{prefix}/lib/gcc-lib/} where @var{prefix} is the value
7706 of @code{prefix} when you ran the @file{configure} script.
7708 Other prefixes specified with @samp{-B} take precedence over this prefix.
7710 This prefix is also used for finding files such as @file{crt0.o} that are
7713 In addition, the prefix is used in an unusual way in finding the
7714 directories to search for header files. For each of the standard
7715 directories whose name normally begins with @samp{/usr/local/lib/gcc-lib}
7716 (more precisely, with the value of @code{GCC_INCLUDE_DIR}), GCC tries
7717 replacing that beginning with the specified prefix to produce an
7718 alternate directory name. Thus, with @samp{-Bfoo/}, GCC will search
7719 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
7720 These alternate directories are searched first; the standard directories
7724 @findex COMPILER_PATH
7725 The value of @code{COMPILER_PATH} is a colon-separated list of
7726 directories, much like @code{PATH}. GCC tries the directories thus
7727 specified when searching for subprograms, if it can't find the
7728 subprograms using @code{GCC_EXEC_PREFIX}.
7731 @findex LIBRARY_PATH
7732 The value of @code{LIBRARY_PATH} is a colon-separated list of
7733 directories, much like @code{PATH}. When configured as a native compiler,
7734 GCC tries the directories thus specified when searching for special
7735 linker files, if it can't find them using @code{GCC_EXEC_PREFIX}. Linking
7736 using GCC also uses these directories when searching for ordinary
7737 libraries for the @samp{-l} option (but directories specified with
7738 @samp{-L} come first).
7740 @item C_INCLUDE_PATH
7741 @itemx CPLUS_INCLUDE_PATH
7742 @itemx OBJC_INCLUDE_PATH
7743 @findex C_INCLUDE_PATH
7744 @findex CPLUS_INCLUDE_PATH
7745 @findex OBJC_INCLUDE_PATH
7746 @c @itemx OBJCPLUS_INCLUDE_PATH
7747 These environment variables pertain to particular languages. Each
7748 variable's value is a colon-separated list of directories, much like
7749 @code{PATH}. When GCC searches for header files, it tries the
7750 directories listed in the variable for the language you are using, after
7751 the directories specified with @samp{-I} but before the standard header
7754 @item DEPENDENCIES_OUTPUT
7755 @findex DEPENDENCIES_OUTPUT
7756 @cindex dependencies for make as output
7757 If this variable is set, its value specifies how to output dependencies
7758 for Make based on the header files processed by the compiler. This
7759 output looks much like the output from the @samp{-M} option
7760 (@pxref{Preprocessor Options}), but it goes to a separate file, and is
7761 in addition to the usual results of compilation.
7763 The value of @code{DEPENDENCIES_OUTPUT} can be just a file name, in
7764 which case the Make rules are written to that file, guessing the target
7765 name from the source file name. Or the value can have the form
7766 @samp{@var{file} @var{target}}, in which case the rules are written to
7767 file @var{file} using @var{target} as the target name.
7771 @cindex locale definition
7772 This variable is used to pass locale information to the compiler. One way in
7773 which this information is used is to determine the character set to be used
7774 when character literals, string literals and comments are parsed in C and C++.
7775 When the compiler is configured to allow multibyte characters,
7776 the following values for @code{LANG} are recognized:
7780 Recognize JIS characters.
7782 Recognize SJIS characters.
7784 Recognize EUCJP characters.
7787 If @code{LANG} is not defined, or if it has some other value, then the
7788 compiler will use mblen and mbtowc as defined by the default locale to
7789 recognize and translate multibyte characters.
7792 @node Running Protoize
7793 @section Running Protoize
7795 The program @code{protoize} is an optional part of GNU C. You can use
7796 it to add prototypes to a program, thus converting the program to ANSI
7797 C in one respect. The companion program @code{unprotoize} does the
7798 reverse: it removes argument types from any prototypes that are found.
7800 When you run these programs, you must specify a set of source files as
7801 command line arguments. The conversion programs start out by compiling
7802 these files to see what functions they define. The information gathered
7803 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
7805 After scanning comes actual conversion. The specified files are all
7806 eligible to be converted; any files they include (whether sources or
7807 just headers) are eligible as well.
7809 But not all the eligible files are converted. By default,
7810 @code{protoize} and @code{unprotoize} convert only source and header
7811 files in the current directory. You can specify additional directories
7812 whose files should be converted with the @samp{-d @var{directory}}
7813 option. You can also specify particular files to exclude with the
7814 @samp{-x @var{file}} option. A file is converted if it is eligible, its
7815 directory name matches one of the specified directory names, and its
7816 name within the directory has not been excluded.
7818 Basic conversion with @code{protoize} consists of rewriting most
7819 function definitions and function declarations to specify the types of
7820 the arguments. The only ones not rewritten are those for varargs
7823 @code{protoize} optionally inserts prototype declarations at the
7824 beginning of the source file, to make them available for any calls that
7825 precede the function's definition. Or it can insert prototype
7826 declarations with block scope in the blocks where undeclared functions
7829 Basic conversion with @code{unprotoize} consists of rewriting most
7830 function declarations to remove any argument types, and rewriting
7831 function definitions to the old-style pre-ANSI form.
7833 Both conversion programs print a warning for any function declaration or
7834 definition that they can't convert. You can suppress these warnings
7837 The output from @code{protoize} or @code{unprotoize} replaces the
7838 original source file. The original file is renamed to a name ending
7839 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
7840 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
7841 for DOS) file already exists, then the source file is simply discarded.
7843 @code{protoize} and @code{unprotoize} both depend on GCC itself to
7844 scan the program and collect information about the functions it uses.
7845 So neither of these programs will work until GCC is installed.
7847 Here is a table of the options you can use with @code{protoize} and
7848 @code{unprotoize}. Each option works with both programs unless
7852 @item -B @var{directory}
7853 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
7854 usual directory (normally @file{/usr/local/lib}). This file contains
7855 prototype information about standard system functions. This option
7856 applies only to @code{protoize}.
7858 @item -c @var{compilation-options}
7859 Use @var{compilation-options} as the options when running @code{gcc} to
7860 produce the @samp{.X} files. The special option @samp{-aux-info} is
7861 always passed in addition, to tell @code{gcc} to write a @samp{.X} file.
7863 Note that the compilation options must be given as a single argument to
7864 @code{protoize} or @code{unprotoize}. If you want to specify several
7865 @code{gcc} options, you must quote the entire set of compilation options
7866 to make them a single word in the shell.
7868 There are certain @code{gcc} arguments that you cannot use, because they
7869 would produce the wrong kind of output. These include @samp{-g},
7870 @samp{-O}, @samp{-c}, @samp{-S}, and @samp{-o} If you include these in
7871 the @var{compilation-options}, they are ignored.
7874 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
7875 systems) instead of @samp{.c}. This is convenient if you are converting
7876 a C program to C++. This option applies only to @code{protoize}.
7879 Add explicit global declarations. This means inserting explicit
7880 declarations at the beginning of each source file for each function
7881 that is called in the file and was not declared. These declarations
7882 precede the first function definition that contains a call to an
7883 undeclared function. This option applies only to @code{protoize}.
7885 @item -i @var{string}
7886 Indent old-style parameter declarations with the string @var{string}.
7887 This option applies only to @code{protoize}.
7889 @code{unprotoize} converts prototyped function definitions to old-style
7890 function definitions, where the arguments are declared between the
7891 argument list and the initial @samp{@{}. By default, @code{unprotoize}
7892 uses five spaces as the indentation. If you want to indent with just
7893 one space instead, use @samp{-i " "}.
7896 Keep the @samp{.X} files. Normally, they are deleted after conversion
7900 Add explicit local declarations. @code{protoize} with @samp{-l} inserts
7901 a prototype declaration for each function in each block which calls the
7902 function without any declaration. This option applies only to
7906 Make no real changes. This mode just prints information about the conversions
7907 that would have been done without @samp{-n}.
7910 Make no @samp{.save} files. The original files are simply deleted.
7911 Use this option with caution.
7913 @item -p @var{program}
7914 Use the program @var{program} as the compiler. Normally, the name
7918 Work quietly. Most warnings are suppressed.
7921 Print the version number, just like @samp{-v} for @code{gcc}.
7924 If you need special compiler options to compile one of your program's
7925 source files, then you should generate that file's @samp{.X} file
7926 specially, by running @code{gcc} on that source file with the
7927 appropriate options and the option @samp{-aux-info}. Then run
7928 @code{protoize} on the entire set of files. @code{protoize} will use
7929 the existing @samp{.X} file because it is newer than the source file.
7933 gcc -Dfoo=bar file1.c -aux-info
7938 You need to include the special files along with the rest in the
7939 @code{protoize} command, even though their @samp{.X} files already
7940 exist, because otherwise they won't get converted.
7942 @xref{Protoize Caveats}, for more information on how to use
7943 @code{protoize} successfully.