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 -x @var{language}
94 -v --target-help --help
97 @item C Language Options
98 @xref{C Dialect Options,,Options Controlling C Dialect}.
100 -ansi -std=@var{standard} -fno-asm -fno-builtin
101 -fhosted -ffreestanding
102 -trigraphs -traditional -traditional-cpp
103 -fallow-single-precision -fcond-mismatch
104 -fsigned-bitfields -fsigned-char
105 -funsigned-bitfields -funsigned-char
106 -fwritable-strings -fshort-wchar
109 @item C++ Language Options
110 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
112 -fno-access-control -fcheck-new -fconserve-space
113 -fdollars-in-identifiers -fno-elide-constructors
114 -fno-enforce-eh-specs -fexternal-templates
115 -falt-external-templates
116 -ffor-scope -fno-for-scope -fno-gnu-keywords -fhonor-std
117 -fhuge-objects -fno-implicit-templates
118 -fno-implicit-inline-templates -finit-priority
119 -fno-implement-inlines -fms-extensions
120 -fname-mangling-version-@var{n} -fno-operator-names
121 -fno-optional-diags -fpermissive
122 -frepo -fno-rtti -fsquangle -ftemplate-depth-@var{n}
123 -fuse-cxa-atexit -fvtable-thunks -nostdinc++
124 -fno-default-inline -Wctor-dtor-privacy
125 -Wnon-virtual-dtor -Wreorder
126 -Weffc++ -Wno-deprecated
127 -Wno-non-template-friend -Wold-style-cast
128 -Woverloaded-virtual -Wno-pmf-conversions
132 @item Language Independent Options
133 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
135 -fmessage-length=@var{n}
136 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}
139 @item Warning Options
140 @xref{Warning Options,,Options to Request or Suppress Warnings}.
142 -fsyntax-only -pedantic -pedantic-errors
143 -w -W -Wall -Waggregate-return
144 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment
145 -Wconversion -Wdisabled-optimization -Werror
146 -Wfloat-equal -Wformat
147 -Wid-clash-@var{len} -Wimplicit -Wimplicit-int
148 -Wimplicit-function-declaration
149 -Werror-implicit-function-declaration
151 -Wlarger-than-@var{len} -Wlong-long
152 -Wmain -Wmissing-declarations
153 -Wmissing-format-attribute -Wmissing-noreturn
154 -Wmultichar -Wno-import -Wpacked -Wpadded
155 -Wparentheses -Wpointer-arith -Wredundant-decls
156 -Wreturn-type -Wsequence-point -Wshadow
157 -Wsign-compare -Wswitch -Wsystem-headers
158 -Wtrigraphs -Wundef -Wuninitialized
159 -Wunknown-pragmas -Wunreachable-code
160 -Wunused -Wunused-function -Wunused-label -Wunused-parameter
161 -Wunused-value -Wunused-variable -Wwrite-strings
164 @item C-only Warning Options
166 -Wbad-function-cast -Wmissing-prototypes -Wnested-externs
167 -Wstrict-prototypes -Wtraditional
170 @item Debugging Options
171 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
173 -a -ax -d@var{letters} -fdump-unnumbered -fdump-translation-unit-@var{file}
174 -fpretend-float -fprofile-arcs -ftest-coverage
175 -g -g@var{level} -gcoff -gdwarf -gdwarf-1 -gdwarf-1+ -gdwarf-2
176 -ggdb -gstabs -gstabs+ -gxcoff -gxcoff+
177 -p -pg -print-file-name=@var{library} -print-libgcc-file-name
178 -print-prog-name=@var{program} -print-search-dirs -Q
182 @item Optimization Options
183 @xref{Optimize Options,,Options that Control Optimization}.
185 -falign-functions=@var{n} -falign-jumps=@var{n}
186 -falign-labels=@var{n} -falign-loops=@var{n}
187 -fbranch-probabilities -fcaller-saves
188 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections -fdce
189 -fdelayed-branch -fdelete-null-pointer-checks
190 -fexpensive-optimizations -ffast-math -ffloat-store
191 -fforce-addr -fforce-mem -ffunction-sections -fgcse
192 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions
193 -fkeep-static-consts -fmove-all-movables
194 -fno-default-inline -fno-defer-pop
195 -fno-function-cse -fno-inline -fno-math-errno -fno-peephole
196 -fomit-frame-pointer -foptimize-register-move
197 -foptimize-sibling-calls -freduce-all-givs
198 -fregmove -frename-registers
199 -frerun-cse-after-loop -frerun-loop-opt
200 -fschedule-insns -fschedule-insns2
201 -fsingle-precision-constant -fssa
202 -fstrength-reduce -fstrict-aliasing -fthread-jumps -ftrapv
203 -funroll-all-loops -funroll-loops
204 -O -O0 -O1 -O2 -O3 -Os
207 @item Preprocessor Options
208 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
210 -A@var{question}(@var{answer}) -C -dD -dM -dN
211 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H
213 -include @var{file} -imacros @var{file}
214 -iprefix @var{file} -iwithprefix @var{dir}
215 -iwithprefixbefore @var{dir} -isystem @var{dir} -isystem-c++ @var{dir}
216 -M -MD -MM -MMD -MG -nostdinc -P -trigraphs
217 -undef -U@var{macro} -Wp,@var{option}
220 @item Assembler Option
221 @xref{Assembler Options,,Passing Options to the Assembler}.
227 @xref{Link Options,,Options for Linking}.
229 @var{object-file-name} -l@var{library}
230 -nostartfiles -nodefaultlibs -nostdlib
231 -s -static -shared -symbolic
232 -Wl,@var{option} -Xlinker @var{option}
236 @item Directory Options
237 @xref{Directory Options,,Options for Directory Search}.
239 -B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}
243 @c I wrote this xref this way to avoid overfull hbox. -- rms
244 @xref{Target Options}.
246 -b @var{machine} -V @var{version}
249 @item Machine Dependent Options
250 @xref{Submodel Options,,Hardware Models and Configurations}.
252 @emph{M680x0 Options}
253 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040
254 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020
255 -mfpa -mnobitfield -mrtd -mshort -msoft-float -mpcrel
256 -malign-int -mstrict-align
258 @emph{M68hc1x Options}
259 -m6811 -m6812 -m68hc11 -m68hc12
260 -mauto-incdec -mshort -msoft-reg-count=@var{count}
267 -mtune=@var{cpu type}
268 -mcmodel=@var{code model}
270 -mapp-regs -mbroken-saverestore -mcypress
271 -mepilogue -mfaster-structs -mflat
272 -mfpu -mhard-float -mhard-quad-float
273 -mimpure-text -mlive-g0 -mno-app-regs
274 -mno-epilogue -mno-faster-structs -mno-flat -mno-fpu
275 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles
276 -msoft-float -msoft-quad-float -msparclite -mstack-bias
277 -msupersparc -munaligned-doubles -mv8
279 @emph{Convex Options}
280 -mc1 -mc2 -mc32 -mc34 -mc38
281 -margcount -mnoargcount
283 -mvolatile-cache -mvolatile-nocache
285 @emph{AMD29K Options}
286 -m29000 -m29050 -mbw -mnbw -mdw -mndw
287 -mlarge -mnormal -msmall
288 -mkernel-registers -mno-reuse-arg-regs
289 -mno-stack-check -mno-storem-bug
290 -mreuse-arg-regs -msoft-float -mstack-check
291 -mstorem-bug -muser-registers
294 -mapcs-frame -mno-apcs-frame
296 -mapcs-stack-check -mno-apcs-stack-check
297 -mapcs-float -mno-apcs-float
298 -mapcs-reentrant -mno-apcs-reentrant
299 -msched-prolog -mno-sched-prolog
300 -mlittle-endian -mbig-endian -mwords-little-endian
301 -malignment-traps -mno-alignment-traps
302 -msoft-float -mhard-float -mfpe
303 -mthumb-interwork -mno-thumb-interwork
304 -mcpu= -march= -mfpe=
305 -mstructure-size-boundary=
306 -mbsd -mxopen -mno-symrename
308 -mlong-calls -mno-long-calls
309 -mnop-fun-dllimport -mno-nop-fun-dllimport
310 -msingle-pic-base -mno-single-pic-base
314 -mtpcs-frame -mno-tpcs-frame
315 -mtpcs-leaf-frame -mno-tpcs-leaf-frame
316 -mlittle-endian -mbig-endian
317 -mthumb-interwork -mno-thumb-interwork
318 -mstructure-size-boundary=
319 -mnop-fun-dllimport -mno-nop-fun-dllimport
320 -mcallee-super-interworking -mno-callee-super-interworking
321 -mcaller-super-interworking -mno-caller-super-interworking
322 -msingle-pic-base -mno-single-pic-base
325 @emph{MN10200 Options}
328 @emph{MN10300 Options}
335 @emph{M32R/D Options}
336 -mcode-model=@var{model type} -msdata=@var{sdata type}
340 -m88000 -m88100 -m88110 -mbig-pic
341 -mcheck-zero-division -mhandle-large-shift
342 -midentify-revision -mno-check-zero-division
343 -mno-ocs-debug-info -mno-ocs-frame-position
344 -mno-optimize-arg-area -mno-serialize-volatile
345 -mno-underscores -mocs-debug-info
346 -mocs-frame-position -moptimize-arg-area
347 -mserialize-volatile -mshort-data-@var{num} -msvr3
348 -msvr4 -mtrap-large-shift -muse-div-instruction
349 -mversion-03.00 -mwarn-passed-structs
351 @emph{RS/6000 and PowerPC Options}
353 -mtune=@var{cpu type}
354 -mpower -mno-power -mpower2 -mno-power2
355 -mpowerpc -mpowerpc64 -mno-powerpc
356 -mpowerpc-gpopt -mno-powerpc-gpopt
357 -mpowerpc-gfxopt -mno-powerpc-gfxopt
358 -mnew-mnemonics -mold-mnemonics
359 -mfull-toc -mminimal-toc -mno-fop-in-toc -mno-sum-in-toc
360 -m64 -m32 -mxl-call -mno-xl-call -mthreads -mpe
361 -msoft-float -mhard-float -mmultiple -mno-multiple
362 -mstring -mno-string -mupdate -mno-update
363 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align
364 -mstrict-align -mno-strict-align -mrelocatable
365 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib
366 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian
367 -mcall-aix -mcall-sysv -mprototype -mno-prototype
368 -msim -mmvme -mads -myellowknife -memb -msdata
369 -msdata=@var{opt} -mvxworks -G @var{num}
372 -mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs
373 -mfull-fp-blocks -mhc-struct-return -min-line-mul
374 -mminimum-fp-blocks -mnohc-struct-return
377 -mabicalls -mcpu=@var{cpu type}
378 -membedded-data -muninit-const-in-rodata
379 -membedded-pic -mfp32 -mfp64 -mgas -mgp32 -mgp64
380 -mgpopt -mhalf-pic -mhard-float -mint64 -mips1
381 -mips2 -mips3 -mips4 -mlong64 -mlong32 -mlong-calls -mmemcpy
382 -mmips-as -mmips-tfile -mno-abicalls
383 -mno-embedded-data -mno-uninit-const-in-rodata -mno-embedded-pic
384 -mno-gpopt -mno-long-calls
385 -mno-memcpy -mno-mips-tfile -mno-rnames -mno-stats
386 -mrnames -msoft-float
387 -m4650 -msingle-float -mmad
388 -mstats -EL -EB -G @var{num} -nocpp
389 -mabi=32 -mabi=n32 -mabi=64 -mabi=eabi
393 -mcpu=@var{cpu type} -march=@var{cpu type}
394 -mintel-syntax -mieee-fp -mno-fancy-math-387
395 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib
396 -mno-wide-multiply -mrtd -malign-double
397 -mreg-alloc=@var{list} -mregparm=@var{num}
398 -malign-jumps=@var{num} -malign-loops=@var{num}
399 -malign-functions=@var{num} -mpreferred-stack-boundary=@var{num}
400 -mthreads -mno-align-stringops -minline-all-stringops
401 -mpush-args -maccumulate-outgoing-args
404 -march=@var{architecture type}
405 -mbig-switch -mdisable-fpregs -mdisable-indexing
406 -mfast-indirect-calls -mgas -mjump-in-delay
407 -mlong-load-store -mno-big-switch -mno-disable-fpregs
408 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas
409 -mno-jump-in-delay -mno-long-load-store
410 -mno-portable-runtime -mno-soft-float
411 -mno-space-regs -msoft-float -mpa-risc-1-0
412 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime
413 -mschedule=@var{cpu type} -mspace-regs
415 @emph{Intel 960 Options}
416 -m@var{cpu type} -masm-compat -mclean-linkage
417 -mcode-align -mcomplex-addr -mleaf-procedures
418 -mic-compat -mic2.0-compat -mic3.0-compat
419 -mintel-asm -mno-clean-linkage -mno-code-align
420 -mno-complex-addr -mno-leaf-procedures
421 -mno-old-align -mno-strict-align -mno-tail-call
422 -mnumerics -mold-align -msoft-float -mstrict-align
425 @emph{DEC Alpha Options}
426 -mfp-regs -mno-fp-regs -mno-soft-float -msoft-float
428 -mieee -mieee-with-inexact -mieee-conformant
429 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode}
430 -mtrap-precision=@var{mode} -mbuild-constants
432 -mbwx -mno-bwx -mcix -mno-cix -mmax -mno-max
433 -mmemory-latency=@var{time}
435 @emph{Clipper Options}
438 @emph{H8/300 Options}
439 -mrelax -mh -ms -mint32 -malign-300
443 -m4-nofpu -m4-single-only -m4-single -m4
444 -mb -ml -mdalign -mrelax
445 -mbigtable -mfmovd -mhitachi -mnomacsave
446 -misize -mpadstruct -mspace
450 @emph{System V Options}
451 -Qy -Qn -YP,@var{paths} -Ym,@var{dir}
455 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text section}
456 -mdata=@var{data section} -mrodata=@var{readonly data section}
458 @emph{TMS320C3x/C4x Options}
459 -mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm
460 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload
461 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned
462 -mparallel-insns -mparallel-mpy -mpreserve-float
465 -mlong-calls -mno-long-calls -mep -mno-ep
466 -mprolog-function -mno-prolog-function -mspace
467 -mtda=@var{n} -msda=@var{n} -mzda=@var{n}
471 -m32032 -m32332 -m32532 -m32081 -m32381 -mmult-add -mnomult-add
472 -msoft-float -mrtd -mnortd -mregparam -mnoregparam -msb -mnosb
473 -mbitfield -mnobitfield -mhimem -mnohimem
476 -mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts
480 -mhardlit, -mno-hardlit -mdiv -mno-div -mrelax-immediates
481 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields
482 -m4byte-functions -mno-4byte-functions -mcallgraph-data
483 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim
484 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment
487 @item Code Generation Options
488 @xref{Code Gen Options,,Options for Code Generation Conventions}.
490 -fcall-saved-@var{reg} -fcall-used-@var{reg}
491 -fexceptions -funwind-tables -ffixed-@var{reg}
492 -finhibit-size-directive -finstrument-functions
493 -fcheck-memory-usage -fprefix-function-name
494 -fno-common -fno-ident -fno-gnu-linker
495 -fpcc-struct-return -fpic -fPIC
496 -freg-struct-return -fshared-data -fshort-enums
497 -fshort-double -fvolatile -fvolatile-global -fvolatile-static
498 -fverbose-asm -fpack-struct -fstack-check
499 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym}
500 -fargument-alias -fargument-noalias
501 -fargument-noalias-global
507 * Overall Options:: Controlling the kind of output:
508 an executable, object files, assembler files,
509 or preprocessed source.
510 * C Dialect Options:: Controlling the variant of C language compiled.
511 * C++ Dialect Options:: Variations on C++.
512 * Language Independent Options:: Controlling how diagnostics should be
514 * Warning Options:: How picky should the compiler be?
515 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
516 * Optimize Options:: How much optimization?
517 * Preprocessor Options:: Controlling header files and macro definitions.
518 Also, getting dependency information for Make.
519 * Assembler Options:: Passing options to the assembler.
520 * Link Options:: Specifying libraries and so on.
521 * Directory Options:: Where to find header files and libraries.
522 Where to find the compiler executable files.
523 * Spec Files:: How to pass switches to sub-processes.
524 * Target Options:: Running a cross-compiler, or an old version of GCC.
527 @node Overall Options
528 @section Options Controlling the Kind of Output
530 Compilation can involve up to four stages: preprocessing, compilation
531 proper, assembly and linking, always in that order. The first three
532 stages apply to an individual source file, and end by producing an
533 object file; linking combines all the object files (those newly
534 compiled, and those specified as input) into an executable file.
536 @cindex file name suffix
537 For any given input file, the file name suffix determines what kind of
542 C source code which must be preprocessed.
545 C source code which should not be preprocessed.
548 C++ source code which should not be preprocessed.
551 Objective-C source code. Note that you must link with the library
552 @file{libobjc.a} to make an Objective-C program work.
555 Objective-C source code which should not be preprocessed.
558 C header file (not to be compiled or linked).
562 @itemx @var{file}.cxx
563 @itemx @var{file}.cpp
564 @itemx @var{file}.c++
566 C++ source code which must be preprocessed. Note that in @samp{.cxx},
567 the last two letters must both be literally @samp{x}. Likewise,
568 @samp{.C} refers to a literal capital C.
571 @itemx @var{file}.for
572 @itemx @var{file}.FOR
573 Fortran source code which should not be preprocessed.
576 @itemx @var{file}.fpp
577 @itemx @var{file}.FPP
578 Fortran source code which must be preprocessed (with the traditional
582 Fortran source code which must be preprocessed with a RATFOR
583 preprocessor (not included with GCC).
585 @xref{Overall Options,,Options Controlling the Kind of Output, g77,
586 Using and Porting GNU Fortran}, for more details of the handling of
589 @c FIXME: Descriptions of Java file types.
595 @c GCC also knows about some suffixes for languages not yet included:
605 @itemx @var{file}.chi
606 CHILL source code (preprocessed with the traditional preprocessor).
612 Assembler code which must be preprocessed.
615 An object file to be fed straight into linking.
616 Any file name with no recognized suffix is treated this way.
619 You can specify the input language explicitly with the @samp{-x} option:
622 @item -x @var{language}
623 Specify explicitly the @var{language} for the following input files
624 (rather than letting the compiler choose a default based on the file
625 name suffix). This option applies to all following input files until
626 the next @samp{-x} option. Possible values for @var{language} are:
628 c c-header cpp-output
630 objective-c objc-cpp-output
631 assembler assembler-with-cpp
632 f77 f77-cpp-input ratfor
635 @c Also f77-version, for internal use only.
638 Turn off any specification of a language, so that subsequent files are
639 handled according to their file name suffixes (as they are if @samp{-x}
640 has not been used at all).
642 @item -pass-exit-codes
643 Normally the @code{gcc} program will exit with the code of 1 if any
644 phase of the compiler returns a non-success return code. If you specify
645 @samp{-pass-exit-codes}, the @code{gcc} program will instead return with
646 numerically highest error produced by any phase that returned an error
650 If you only want some of the stages of compilation, you can use
651 @samp{-x} (or filename suffixes) to tell @code{gcc} where to start, and
652 one of the options @samp{-c}, @samp{-S}, or @samp{-E} to say where
653 @code{gcc} is to stop. Note that some combinations (for example,
654 @samp{-x cpp-output -E} instruct @code{gcc} to do nothing at all.
658 Compile or assemble the source files, but do not link. The linking
659 stage simply is not done. The ultimate output is in the form of an
660 object file for each source file.
662 By default, the object file name for a source file is made by replacing
663 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
665 Unrecognized input files, not requiring compilation or assembly, are
669 Stop after the stage of compilation proper; do not assemble. The output
670 is in the form of an assembler code file for each non-assembler input
673 By default, the assembler file name for a source file is made by
674 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
676 Input files that don't require compilation are ignored.
679 Stop after the preprocessing stage; do not run the compiler proper. The
680 output is in the form of preprocessed source code, which is sent to the
683 Input files which don't require preprocessing are ignored.
685 @cindex output file option
687 Place output in file @var{file}. This applies regardless to whatever
688 sort of output is being produced, whether it be an executable file,
689 an object file, an assembler file or preprocessed C code.
691 Since only one output file can be specified, it does not make sense to
692 use @samp{-o} when compiling more than one input file, unless you are
693 producing an executable file as output.
695 If @samp{-o} is not specified, the default is to put an executable file
696 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
697 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
698 all preprocessed C source on standard output.@refill
701 Print (on standard error output) the commands executed to run the stages
702 of compilation. Also print the version number of the compiler driver
703 program and of the preprocessor and the compiler proper.
706 Use pipes rather than temporary files for communication between the
707 various stages of compilation. This fails to work on some systems where
708 the assembler is unable to read from a pipe; but the GNU assembler has
712 Print (on the standard output) a description of the command line options
713 understood by @code{gcc}. If the @code{-v} option is also specified
714 then @code{--help} will also be passed on to the various processes
715 invoked by @code{gcc}, so that they can display the command line options
716 they accept. If the @code{-W} option is also specified then command
717 line options which have no documentation associated with them will also
721 Print (on the standard output) a description of target specific command
722 line options for each tool.
726 @section Compiling C++ Programs
728 @cindex suffixes for C++ source
729 @cindex C++ source file suffixes
730 C++ source files conventionally use one of the suffixes @samp{.C},
731 @samp{.cc}, @samp{.cpp}, @samp{.c++}, @samp{.cp}, or @samp{.cxx};
732 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
733 files with these names and compiles them as C++ programs even if you
734 call the compiler the same way as for compiling C programs (usually with
735 the name @code{gcc}).
739 However, C++ programs often require class libraries as well as a
740 compiler that understands the C++ language---and under some
741 circumstances, you might want to compile programs from standard input,
742 or otherwise without a suffix that flags them as C++ programs.
743 @code{g++} is a program that calls GCC with the default language
744 set to C++, and automatically specifies linking against the C++
745 library. On many systems, @code{g++} is also
746 installed with the name @code{c++}.
748 @cindex invoking @code{g++}
749 When you compile C++ programs, you may specify many of the same
750 command-line options that you use for compiling programs in any
751 language; or command-line options meaningful for C and related
752 languages; or options that are meaningful only for C++ programs.
753 @xref{C Dialect Options,,Options Controlling C Dialect}, for
754 explanations of options for languages related to C.
755 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
756 explanations of options that are meaningful only for C++ programs.
758 @node C Dialect Options
759 @section Options Controlling C Dialect
760 @cindex dialect options
761 @cindex language dialect options
762 @cindex options, dialect
764 The following options control the dialect of C (or languages derived
765 from C, such as C++ and Objective C) that the compiler accepts:
771 In C mode, support all ISO C89 programs. In C++ mode,
772 remove GNU extensions that conflict with ISO C++.
774 This turns off certain features of GCC that are incompatible with ISO
775 C (when compiling C code), or of standard C++ (when compiling C++ code),
776 such as the @code{asm} and @code{typeof} keywords, and
777 predefined macros such as @code{unix} and @code{vax} that identify the
778 type of system you are using. It also enables the undesirable and
779 rarely used ISO trigraph feature. For the C compiler,
780 it disables recognition of C++ style @samp{//} comments as well as
781 the @code{inline} keyword.
783 The alternate keywords @code{__asm__}, @code{__extension__},
784 @code{__inline__} and @code{__typeof__} continue to work despite
785 @samp{-ansi}. You would not want to use them in an ISO C program, of
786 course, but it is useful to put them in header files that might be included
787 in compilations done with @samp{-ansi}. Alternate predefined macros
788 such as @code{__unix__} and @code{__vax__} are also available, with or
789 without @samp{-ansi}.
791 The @samp{-ansi} option does not cause non-ISO programs to be
792 rejected gratuitously. For that, @samp{-pedantic} is required in
793 addition to @samp{-ansi}. @xref{Warning Options}.
795 The macro @code{__STRICT_ANSI__} is predefined when the @samp{-ansi}
796 option is used. Some header files may notice this macro and refrain
797 from declaring certain functions or defining certain macros that the
798 ISO standard doesn't call for; this is to avoid interfering with any
799 programs that might use these names for other things.
801 Functions which would normally be builtin but do not have semantics
802 defined by ISO C (such as @code{alloca} and @code{ffs}) are not builtin
803 functions with @samp{-ansi} is used. @xref{Other Builtins,,Other
804 built-in functions provided by GNU CC}, for details of the functions
808 Determine the language standard. A value for this option must be provided;
816 ISO C as modified in amend. 1
819 ISO C99. Note that this standard is not yet fully supported; see
820 @uref{http://gcc.gnu.org/c99status.html} for more information.
823 same as -std=iso9899:1990
826 same as -std=iso9899:1999
829 default, iso9899:1990 + gnu extensions
832 iso9899:1999 + gnu extensions
835 same as -std=iso9899:1999, deprecated
838 same as -std=iso9899:1999, deprecated
841 same as -std=gnu99, deprecated
845 Even when this option is not specified, you can still use some of the
846 features of newer standards in so far as they do not conflict with
847 previous C standards. For example, you may use @code{__restrict__} even
848 when -std=c99 is not specified.
850 @xref{Standards,,Language Standards Supported by GCC}, for details of
851 these standard versions.
854 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
855 keyword, so that code can use these words as identifiers. You can use
856 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
857 instead. @samp{-ansi} implies @samp{-fno-asm}.
859 In C++, this switch only affects the @code{typeof} keyword, since
860 @code{asm} and @code{inline} are standard keywords. You may want to
861 use the @samp{-fno-gnu-keywords} flag instead, which has the same effect.
864 @cindex builtin functions
865 Don't recognize builtin functions that do not begin with
866 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
867 functions provided by GNU CC}, for details of the functions affected,
868 including those which are not builtin functions when @samp{-ansi} is
869 used because they do not have an ISO standard meaning.
871 GCC normally generates special code to handle certain builtin functions
872 more efficiently; for instance, calls to @code{alloca} may become single
873 instructions that adjust the stack directly, and calls to @code{memcpy}
874 may become inline copy loops. The resulting code is often both smaller
875 and faster, but since the function calls no longer appear as such, you
876 cannot set a breakpoint on those calls, nor can you change the behavior
877 of the functions by linking with a different library.
880 @cindex hosted environment
882 Assert that compilation takes place in a hosted environment. This implies
883 @samp{-fbuiltin}. A hosted environment is one in which the
884 entire standard library is available, and in which @code{main} has a return
885 type of @code{int}. Examples are nearly everything except a kernel.
886 This is equivalent to @samp{-fno-freestanding}.
889 @cindex hosted environment
891 Assert that compilation takes place in a freestanding environment. This
892 implies @samp{-fno-builtin}. A freestanding environment
893 is one in which the standard library may not exist, and program startup may
894 not necessarily be at @code{main}. The most obvious example is an OS kernel.
895 This is equivalent to @samp{-fno-hosted}.
897 @xref{Standards,,Language Standards Supported by GCC}, for details of
898 freestanding and hosted environments.
901 Support ISO C trigraphs. You don't want to know about this
902 brain-damage. The @samp{-ansi} option implies @samp{-trigraphs}.
904 @cindex traditional C language
905 @cindex C language, traditional
907 Attempt to support some aspects of traditional C compilers.
912 All @code{extern} declarations take effect globally even if they
913 are written inside of a function definition. This includes implicit
914 declarations of functions.
917 The newer keywords @code{typeof}, @code{inline}, @code{signed}, @code{const}
918 and @code{volatile} are not recognized. (You can still use the
919 alternative keywords such as @code{__typeof__}, @code{__inline__}, and
923 Comparisons between pointers and integers are always allowed.
926 Integer types @code{unsigned short} and @code{unsigned char} promote
927 to @code{unsigned int}.
930 Out-of-range floating point literals are not an error.
933 Certain constructs which ISO regards as a single invalid preprocessing
934 number, such as @samp{0xe-0xd}, are treated as expressions instead.
937 String ``constants'' are not necessarily constant; they are stored in
938 writable space, and identical looking constants are allocated
939 separately. (This is the same as the effect of
940 @samp{-fwritable-strings}.)
942 @cindex @code{longjmp} and automatic variables
944 All automatic variables not declared @code{register} are preserved by
945 @code{longjmp}. Ordinarily, GNU C follows ISO C: automatic variables
946 not declared @code{volatile} may be clobbered.
951 @cindex escape sequences, traditional
952 The character escape sequences @samp{\x} and @samp{\a} evaluate as the
953 literal characters @samp{x} and @samp{a} respectively. Without
954 @w{@samp{-traditional}}, @samp{\x} is a prefix for the hexadecimal
955 representation of a character, and @samp{\a} produces a bell.
958 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
959 if your program uses names that are normally GNU C builtin functions for
960 other purposes of its own.
962 You cannot use @samp{-traditional} if you include any header files that
963 rely on ISO C features. Some vendors are starting to ship systems with
964 ISO C header files and you cannot use @samp{-traditional} on such
965 systems to compile files that include any system headers.
967 The @samp{-traditional} option also enables @samp{-traditional-cpp},
968 which is described next.
970 @item -traditional-cpp
971 Attempt to support some aspects of traditional C preprocessors.
976 Comments convert to nothing at all, rather than to a space. This allows
977 traditional token concatenation.
980 In a preprocessing directive, the @samp{#} symbol must appear as the first
984 Macro arguments are recognized within string constants in a macro
985 definition (and their values are stringified, though without additional
986 quote marks, when they appear in such a context). The preprocessor
987 always considers a string constant to end at a newline.
990 @cindex detecting @w{@samp{-traditional}}
991 The predefined macro @code{__STDC__} is not defined when you use
992 @samp{-traditional}, but @code{__GNUC__} is (since the GNU extensions
993 which @code{__GNUC__} indicates are not affected by
994 @samp{-traditional}). If you need to write header files that work
995 differently depending on whether @samp{-traditional} is in use, by
996 testing both of these predefined macros you can distinguish four
997 situations: GNU C, traditional GNU C, other ISO C compilers, and other
998 old C compilers. The predefined macro @code{__STDC_VERSION__} is also
999 not defined when you use @samp{-traditional}. @xref{Standard
1000 Predefined,,Standard Predefined Macros,cpp.info,The C Preprocessor},
1001 for more discussion of these and other predefined macros.
1004 @cindex string constants vs newline
1005 @cindex newline vs string constants
1006 The preprocessor considers a string constant to end at a newline (unless
1007 the newline is escaped with @samp{\}). (Without @w{@samp{-traditional}},
1008 string constants can contain the newline character as typed.)
1011 @item -fcond-mismatch
1012 Allow conditional expressions with mismatched types in the second and
1013 third arguments. The value of such an expression is void.
1015 @item -funsigned-char
1016 Let the type @code{char} be unsigned, like @code{unsigned char}.
1018 Each kind of machine has a default for what @code{char} should
1019 be. It is either like @code{unsigned char} by default or like
1020 @code{signed char} by default.
1022 Ideally, a portable program should always use @code{signed char} or
1023 @code{unsigned char} when it depends on the signedness of an object.
1024 But many programs have been written to use plain @code{char} and
1025 expect it to be signed, or expect it to be unsigned, depending on the
1026 machines they were written for. This option, and its inverse, let you
1027 make such a program work with the opposite default.
1029 The type @code{char} is always a distinct type from each of
1030 @code{signed char} or @code{unsigned char}, even though its behavior
1031 is always just like one of those two.
1034 Let the type @code{char} be signed, like @code{signed char}.
1036 Note that this is equivalent to @samp{-fno-unsigned-char}, which is
1037 the negative form of @samp{-funsigned-char}. Likewise, the option
1038 @samp{-fno-signed-char} is equivalent to @samp{-funsigned-char}.
1040 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
1041 if your program uses names that are normally GNU C builtin functions for
1042 other purposes of its own.
1044 You cannot use @samp{-traditional} if you include any header files that
1045 rely on ISO C features. Some vendors are starting to ship systems with
1046 ISO C header files and you cannot use @samp{-traditional} on such
1047 systems to compile files that include any system headers.
1049 @item -fsigned-bitfields
1050 @itemx -funsigned-bitfields
1051 @itemx -fno-signed-bitfields
1052 @itemx -fno-unsigned-bitfields
1053 These options control whether a bitfield is signed or unsigned, when the
1054 declaration does not use either @code{signed} or @code{unsigned}. By
1055 default, such a bitfield is signed, because this is consistent: the
1056 basic integer types such as @code{int} are signed types.
1058 However, when @samp{-traditional} is used, bitfields are all unsigned
1061 @item -fwritable-strings
1062 Store string constants in the writable data segment and don't uniquize
1063 them. This is for compatibility with old programs which assume they can
1064 write into string constants. The option @samp{-traditional} also has
1067 Writing into string constants is a very bad idea; ``constants'' should
1070 @item -fallow-single-precision
1071 Do not promote single precision math operations to double precision,
1072 even when compiling with @samp{-traditional}.
1074 Traditional K&R C promotes all floating point operations to double
1075 precision, regardless of the sizes of the operands. On the
1076 architecture for which you are compiling, single precision may be faster
1077 than double precision. If you must use @samp{-traditional}, but want
1078 to use single precision operations when the operands are single
1079 precision, use this option. This option has no effect when compiling
1080 with ISO or GNU C conventions (the default).
1083 Override the underlying type for @samp{wchar_t} to be @samp{short
1084 unsigned int} instead of the default for the target. This option is
1085 useful for building programs to run under WINE.
1088 @node C++ Dialect Options
1089 @section Options Controlling C++ Dialect
1091 @cindex compiler options, C++
1092 @cindex C++ options, command line
1093 @cindex options, C++
1094 This section describes the command-line options that are only meaningful
1095 for C++ programs; but you can also use most of the GNU compiler options
1096 regardless of what language your program is in. For example, you
1097 might compile a file @code{firstClass.C} like this:
1100 g++ -g -frepo -O -c firstClass.C
1104 In this example, only @samp{-frepo} is an option meant
1105 only for C++ programs; you can use the other options with any
1106 language supported by GCC.
1108 Here is a list of options that are @emph{only} for compiling C++ programs:
1111 @item -fno-access-control
1112 Turn off all access checking. This switch is mainly useful for working
1113 around bugs in the access control code.
1116 Check that the pointer returned by @code{operator new} is non-null
1117 before attempting to modify the storage allocated. The current Working
1118 Paper requires that @code{operator new} never return a null pointer, so
1119 this check is normally unnecessary.
1121 An alternative to using this option is to specify that your
1122 @code{operator new} does not throw any exceptions; if you declare it
1123 @samp{throw()}, g++ will check the return value. See also @samp{new
1126 @item -fconserve-space
1127 Put uninitialized or runtime-initialized global variables into the
1128 common segment, as C does. This saves space in the executable at the
1129 cost of not diagnosing duplicate definitions. If you compile with this
1130 flag and your program mysteriously crashes after @code{main()} has
1131 completed, you may have an object that is being destroyed twice because
1132 two definitions were merged.
1134 This option is no longer useful on most targets, now that support has
1135 been added for putting variables into BSS without making them common.
1137 @item -fdollars-in-identifiers
1138 Accept @samp{$} in identifiers. You can also explicitly prohibit use of
1139 @samp{$} with the option @samp{-fno-dollars-in-identifiers}. (GNU C allows
1140 @samp{$} by default on most target systems, but there are a few exceptions.)
1141 Traditional C allowed the character @samp{$} to form part of
1142 identifiers. However, ISO C and C++ forbid @samp{$} in identifiers.
1144 @item -fno-elide-constructors
1145 The C++ standard allows an implementation to omit creating a temporary
1146 which is only used to initialize another object of the same type.
1147 Specifying this option disables that optimization, and forces g++ to
1148 call the copy constructor in all cases.
1150 @item -fno-enforce-eh-specs
1151 Don't check for violation of exception specifications at runtime. This
1152 option violates the C++ standard, but may be useful for reducing code
1153 size in production builds, much like defining @samp{NDEBUG}. The compiler
1154 will still optimize based on the exception specifications.
1156 @item -fexternal-templates
1157 Cause template instantiations to obey @samp{#pragma interface} and
1158 @samp{implementation}; template instances are emitted or not according
1159 to the location of the template definition. @xref{Template
1160 Instantiation}, for more information.
1162 This option is deprecated.
1164 @item -falt-external-templates
1165 Similar to -fexternal-templates, but template instances are emitted or
1166 not according to the place where they are first instantiated.
1167 @xref{Template Instantiation}, for more information.
1169 This option is deprecated.
1172 @itemx -fno-for-scope
1173 If -ffor-scope is specified, the scope of variables declared in
1174 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1175 as specified by the C++ standard.
1176 If -fno-for-scope is specified, the scope of variables declared in
1177 a @i{for-init-statement} extends to the end of the enclosing scope,
1178 as was the case in old versions of gcc, and other (traditional)
1179 implementations of C++.
1181 The default if neither flag is given to follow the standard,
1182 but to allow and give a warning for old-style code that would
1183 otherwise be invalid, or have different behavior.
1185 @item -fno-gnu-keywords
1186 Do not recognize @code{typeof} as a keyword, so that code can use this
1187 word as an identifier. You can use the keyword @code{__typeof__} instead.
1188 @samp{-ansi} implies @samp{-fno-gnu-keywords}.
1191 Treat the @code{namespace std} as a namespace, instead of ignoring
1192 it. For compatibility with earlier versions of g++, the compiler will,
1193 by default, ignore @code{namespace-declarations},
1194 @code{using-declarations}, @code{using-directives}, and
1195 @code{namespace-names}, if they involve @code{std}.
1197 @item -fhuge-objects
1198 Support virtual function calls for objects that exceed the size
1199 representable by a @samp{short int}. Users should not use this flag by
1200 default; if you need to use it, the compiler will tell you so.
1202 This flag is not useful when compiling with -fvtable-thunks.
1204 Like all options that change the ABI, all C++ code, @emph{including
1205 libgcc} must be built with the same setting of this option.
1207 @item -fno-implicit-templates
1208 Never emit code for non-inline templates which are instantiated
1209 implicitly (i.e. by use); only emit code for explicit instantiations.
1210 @xref{Template Instantiation}, for more information.
1212 @item -fno-implicit-inline-templates
1213 Don't emit code for implicit instantiations of inline templates, either.
1214 The default is to handle inlines differently so that compiles with and
1215 without optimization will need the same set of explicit instantiations.
1217 @item -finit-priority
1218 Support @samp{__attribute__ ((init_priority (n)))} for controlling the
1219 order of initialization of file-scope objects. On ELF targets, this
1220 requires GNU ld 2.10 or later.
1222 @item -fno-implement-inlines
1223 To save space, do not emit out-of-line copies of inline functions
1224 controlled by @samp{#pragma implementation}. This will cause linker
1225 errors if these functions are not inlined everywhere they are called.
1227 @item -fms-extensions
1228 Disable pedwarns about constructs used in MFC, such as implicit int and
1229 getting a pointer to member function via non-standard syntax.
1231 @item -fname-mangling-version-@var{n}
1232 Control the way in which names are mangled. Version 0 is compatible
1233 with versions of g++ before 2.8. Version 1 is the default. Version 1
1234 will allow correct mangling of function templates. For example,
1235 version 0 mangling does not mangle foo<int, double> and foo<int, char>
1236 given this declaration:
1239 template <class T, class U> void foo(T t);
1242 Like all options that change the ABI, all C++ code, @emph{including
1243 libgcc} must be built with the same setting of this option.
1245 @item -fno-operator-names
1246 Do not treat the operator name keywords @code{and}, @code{bitand},
1247 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1248 synonyms as keywords.
1250 @item -fno-optional-diags
1251 Disable diagnostics that the standard says a compiler does not need to
1252 issue. Currently, the only such diagnostic issued by g++ is the one for
1253 a name having multiple meanings within a class.
1256 Downgrade messages about nonconformant code from errors to warnings. By
1257 default, g++ effectively sets @samp{-pedantic-errors} without
1258 @samp{-pedantic}; this option reverses that. This behavior and this
1259 option are superseded by @samp{-pedantic}, which works as it does for GNU C.
1262 Enable automatic template instantiation. This option also implies
1263 @samp{-fno-implicit-templates}. @xref{Template Instantiation}, for more
1267 Disable generation of information about every class with virtual
1268 functions for use by the C++ runtime type identification features
1269 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1270 of the language, you can save some space by using this flag. Note that
1271 exception handling uses the same information, but it will generate it as
1275 @itemx -fno-squangle
1276 @samp{-fsquangle} will enable a compressed form of name mangling for
1277 identifiers. In particular, it helps to shorten very long names by recognizing
1278 types and class names which occur more than once, replacing them with special
1279 short ID codes. This option also requires any C++ libraries being used to
1280 be compiled with this option as well. The compiler has this disabled (the
1281 equivalent of @samp{-fno-squangle}) by default.
1283 Like all options that change the ABI, all C++ code, @emph{including
1284 libgcc.a} must be built with the same setting of this option.
1286 @item -ftemplate-depth-@var{n}
1287 Set the maximum instantiation depth for template classes to @var{n}.
1288 A limit on the template instantiation depth is needed to detect
1289 endless recursions during template class instantiation. ANSI/ISO C++
1290 conforming programs must not rely on a maximum depth greater than 17.
1292 @item -fuse-cxa-atexit
1293 Register destructors for objects with static storage duration with the
1294 @code{__cxa_atexit} function rather than the @code{atexit} function.
1295 This option is required for fully standards-compliant handling of static
1296 destructors, but will only work if your C library supports
1297 @code{__cxa_atexit}.
1299 @item -fvtable-thunks
1300 Use @samp{thunks} to implement the virtual function dispatch table
1301 (@samp{vtable}). The traditional (cfront-style) approach to
1302 implementing vtables was to store a pointer to the function and two
1303 offsets for adjusting the @samp{this} pointer at the call site. Newer
1304 implementations store a single pointer to a @samp{thunk} function which
1305 does any necessary adjustment and then calls the target function.
1307 This option also enables a heuristic for controlling emission of
1308 vtables; if a class has any non-inline virtual functions, the vtable
1309 will be emitted in the translation unit containing the first one of
1312 Like all options that change the ABI, all C++ code, @emph{including
1313 libgcc.a} must be built with the same setting of this option.
1316 Do not search for header files in the standard directories specific to
1317 C++, but do still search the other standard directories. (This option
1318 is used when building the C++ library.)
1321 In addition, these optimization, warning, and code generation options
1322 have meanings only for C++ programs:
1325 @item -fno-default-inline
1326 Do not assume @samp{inline} for functions defined inside a class scope.
1327 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1328 functions will have linkage like inline functions; they just won't be
1331 @item -Wctor-dtor-privacy (C++ only)
1332 Warn when a class seems unusable, because all the constructors or
1333 destructors in a class are private and the class has no friends or
1334 public static member functions.
1336 @item -Wnon-virtual-dtor (C++ only)
1337 Warn when a class declares a non-virtual destructor that should probably
1338 be virtual, because it looks like the class will be used polymorphically.
1340 @item -Wreorder (C++ only)
1341 @cindex reordering, warning
1342 @cindex warning for reordering of member initializers
1343 Warn when the order of member initializers given in the code does not
1344 match the order in which they must be executed. For instance:
1350 A(): j (0), i (1) @{ @}
1354 Here the compiler will warn that the member initializers for @samp{i}
1355 and @samp{j} will be rearranged to match the declaration order of the
1359 The following @samp{-W@dots{}} options are not affected by @samp{-Wall}.
1362 @item -Weffc++ (C++ only)
1363 Warn about violations of various style guidelines from Scott Meyers'
1364 @cite{Effective C++} books. If you use this option, you should be aware
1365 that the standard library headers do not obey all of these guidelines;
1366 you can use @samp{grep -v} to filter out those warnings.
1368 @item -Wno-deprecated (C++ only)
1369 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1371 @item -Wno-non-template-friend (C++ only)
1372 Disable warnings when non-templatized friend functions are declared
1373 within a template. With the advent of explicit template specification
1374 support in g++, if the name of the friend is an unqualified-id (ie,
1375 @samp{friend foo(int)}), the C++ language specification demands that the
1376 friend declare or define an ordinary, nontemplate function. (Section
1377 14.5.3). Before g++ implemented explicit specification, unqualified-ids
1378 could be interpreted as a particular specialization of a templatized
1379 function. Because this non-conforming behavior is no longer the default
1380 behavior for g++, @samp{-Wnon-template-friend} allows the compiler to
1381 check existing code for potential trouble spots, and is on by default.
1382 This new compiler behavior can be turned off with
1383 @samp{-Wno-non-template-friend} which keeps the conformant compiler code
1384 but disables the helpful warning.
1386 @item -Wold-style-cast (C++ only)
1387 Warn if an old-style (C-style) cast is used within a C++ program. The
1388 new-style casts (@samp{static_cast}, @samp{reinterpret_cast}, and
1389 @samp{const_cast}) are less vulnerable to unintended effects.
1391 @item -Woverloaded-virtual (C++ only)
1392 @cindex overloaded virtual fn, warning
1393 @cindex warning for overloaded virtual fn
1394 Warn when a derived class function declaration may be an error in
1395 defining a virtual function. In a derived class, the
1396 definitions of virtual functions must match the type signature of a
1397 virtual function declared in the base class. With this option, the
1398 compiler warns when you define a function with the same name as a
1399 virtual function, but with a type signature that does not match any
1400 declarations from the base class.
1402 @item -Wno-pmf-conversions (C++ only)
1403 Disable the diagnostic for converting a bound pointer to member function
1406 @item -Wsign-promo (C++ only)
1407 Warn when overload resolution chooses a promotion from unsigned or
1408 enumeral type to a signed type over a conversion to an unsigned type of
1409 the same size. Previous versions of g++ would try to preserve
1410 unsignedness, but the standard mandates the current behavior.
1412 @item -Wsynth (C++ only)
1413 @cindex warning for synthesized methods
1414 @cindex synthesized methods, warning
1415 Warn when g++'s synthesis behavior does not match that of cfront. For
1421 A& operator = (int);
1431 In this example, g++ will synthesize a default @samp{A& operator =
1432 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1435 @node Language Independent Options
1436 @section Options to Control Diagnostic Messages Formatting
1437 @cindex options to control diagnostics formatting
1438 @cindex diagnostic messages
1439 @cindex message formatting
1441 Traditionally, diagnostic messages have been formatted irrespective of
1442 the output device's aspect (e.g. its width, ...). The options described
1443 below can be used to control the diagnostic messages formatting
1444 algorithm, e.g. how many characters per line, how often source location
1445 information should be reported. Right now, only the C++ front-end can
1446 honor these options. However it is expected, in the near future, that
1447 the remaining front-ends would be able to digest them correctly.
1450 @item -fmessage-length=@var{n}
1451 Try to format error messages so that they fit on lines of about @var{n}
1452 characters. The default is 72 characters for g++ and 0 for the rest of
1453 the front-ends supported by GCC. If @var{n} is zero, then no
1454 line-wrapping will be done; each error message will appear on a single
1457 @item -fdiagnostics-show-location=once
1458 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1459 reporter to emit @emph{once} source location information; that is, in
1460 case the message is too long to fit on a single physical line and has to
1461 be wrapped, the source location won't be emitted (as prefix) again,
1462 over and over, in subsequent continuation lines. This is the default
1465 @item -fdiagnostics-show-location=every-line
1466 Only meaningful in line-wrapping mode. Instructs the diagnostic
1467 messages reporter to emit the same source location information (as
1468 prefix) for physical lines that result from the process of breaking a
1469 a message which is too long to fit on a single line.
1473 @node Warning Options
1474 @section Options to Request or Suppress Warnings
1475 @cindex options to control warnings
1476 @cindex warning messages
1477 @cindex messages, warning
1478 @cindex suppressing warnings
1480 Warnings are diagnostic messages that report constructions which
1481 are not inherently erroneous but which are risky or suggest there
1482 may have been an error.
1484 You can request many specific warnings with options beginning @samp{-W},
1485 for example @samp{-Wimplicit} to request warnings on implicit
1486 declarations. Each of these specific warning options also has a
1487 negative form beginning @samp{-Wno-} to turn off warnings;
1488 for example, @samp{-Wno-implicit}. This manual lists only one of the
1489 two forms, whichever is not the default.
1491 These options control the amount and kinds of warnings produced by GCC:
1494 @cindex syntax checking
1496 Check the code for syntax errors, but don't do anything beyond that.
1499 Issue all the warnings demanded by strict ISO C and ISO C++;
1500 reject all programs that use forbidden extensions, and some other
1501 programs that do not follow ISO C and ISO C++. For ISO C, follows the
1502 version of the ISO C standard specified by any @samp{-std} option used.
1504 Valid ISO C and ISO C++ programs should compile properly with or without
1505 this option (though a rare few will require @samp{-ansi}). However,
1506 without this option, certain GNU extensions and traditional C and C++
1507 features are supported as well. With this option, they are rejected.
1509 @samp{-pedantic} does not cause warning messages for use of the
1510 alternate keywords whose names begin and end with @samp{__}. Pedantic
1511 warnings are also disabled in the expression that follows
1512 @code{__extension__}. However, only system header files should use
1513 these escape routes; application programs should avoid them.
1514 @xref{Alternate Keywords}.
1516 Some users try to use @samp{-pedantic} to check programs for strict ISO
1517 C conformance. They soon find that it does not do quite what they want:
1518 it finds some non-ISO practices, but not all---only those for which
1519 ISO C @emph{requires} a diagnostic, and some others for which
1520 diagnostics have been added.
1522 A feature to report any failure to conform to ISO C might be useful in
1523 some instances, but would require considerable additional work and would
1524 be quite different from @samp{-pedantic}. We don't have plans to
1525 support such a feature in the near future.
1527 @item -pedantic-errors
1528 Like @samp{-pedantic}, except that errors are produced rather than
1532 Inhibit all warning messages.
1535 Inhibit warning messages about the use of @samp{#import}.
1537 @item -Wchar-subscripts
1538 Warn if an array subscript has type @code{char}. This is a common cause
1539 of error, as programmers often forget that this type is signed on some
1543 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
1544 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
1547 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
1548 the arguments supplied have types appropriate to the format string
1551 @item -Wimplicit-int
1552 Warn when a declaration does not specify a type.
1554 @item -Wimplicit-function-declaration
1555 @itemx -Werror-implicit-function-declaration
1556 Give a warning (or error) whenever a function is used before being
1560 Same as @samp{-Wimplicit-int} and @samp{-Wimplicit-function-}@*
1564 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
1565 function with external linkage, returning int, taking either zero
1566 arguments, two, or three arguments of appropriate types.
1569 Warn if a multicharacter constant (@samp{'FOOF'}) is used. Usually they
1570 indicate a typo in the user's code, as they have implementation-defined
1571 values, and should not be used in portable code.
1574 Warn if parentheses are omitted in certain contexts, such
1575 as when there is an assignment in a context where a truth value
1576 is expected, or when operators are nested whose precedence people
1577 often get confused about.
1579 Also warn about constructions where there may be confusion to which
1580 @code{if} statement an @code{else} branch belongs. Here is an example of
1593 In C, every @code{else} branch belongs to the innermost possible @code{if}
1594 statement, which in this example is @code{if (b)}. This is often not
1595 what the programmer expected, as illustrated in the above example by
1596 indentation the programmer chose. When there is the potential for this
1597 confusion, GNU C will issue a warning when this flag is specified.
1598 To eliminate the warning, add explicit braces around the innermost
1599 @code{if} statement so there is no way the @code{else} could belong to
1600 the enclosing @code{if}. The resulting code would look like this:
1614 @item -Wsequence-point
1615 Warn about code that may have undefined semantics because of violations
1616 of sequence point rules in the C standard.
1618 The C standard defines the order in which expressions in a C program are
1619 evaluated in terms of @dfn{sequence points}, which represent a partial
1620 ordering between the execution of parts of the program: those executed
1621 before the sequence point, and those executed after it. These occur
1622 after the evaluation of a full expression (one which is not part of a
1623 larger expression), after the evaluation of the first operand of a
1624 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
1625 function is called (but after the evaluation of its arguments and the
1626 expression denoting the called function), and in certain other places.
1627 Other than as expressed by the sequence point rules, the order of
1628 evaluation of subexpressions of an expression is not specified. All
1629 these rules describe only a partial order rather than a total order,
1630 since, for example, if two functions are called within one expression
1631 with no sequence point between them, the order in which the functions
1632 are called is not specified. However, the standards committee have
1633 ruled that function calls do not overlap.
1635 It is not specified when between sequence points modifications to the
1636 values of objects take effect. Programs whose behavior depends on this
1637 have undefined behavior; the C standard specifies that ``Between the
1638 previous and next sequence point an object shall have its stored value
1639 modified at most once by the evaluation of an expression. Furthermore,
1640 the prior value shall be read only to determine the value to be
1641 stored.''. If a program breaks these rules, the results on any
1642 particular implementation are entirely unpredictable.
1644 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
1645 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
1646 diagnosed by this option, and it may give an occasional false positive
1647 result, but in general it has been found fairly effective at detecting
1648 this sort of problem in programs.
1650 The present implementation of this option only works for C programs. A
1651 future implementation may also work for C++ programs.
1653 There is some controversy over the precise meaning of the sequence point
1654 rules in subtle cases. Alternative formal definitions may be found in
1655 Clive Feather's ``Annex S''
1656 @uref{http://wwwold.dkuug.dk/JTC1/SC22/WG14/www/docs/n908.htm} and in
1657 Michael Norrish's thesis
1658 @uref{http://www.cl.cam.ac.uk/users/mn200/PhD/thesis-report.ps.gz}.
1661 Warn whenever a function is defined with a return-type that defaults to
1662 @code{int}. Also warn about any @code{return} statement with no
1663 return-value in a function whose return-type is not @code{void}.
1665 For C++, a function without return type always produces a diagnostic
1666 message, even when @samp{-Wno-return-type} is specified. The only
1667 exceptions are @samp{main} and functions defined in system headers.
1670 Warn whenever a @code{switch} statement has an index of enumeral type
1671 and lacks a @code{case} for one or more of the named codes of that
1672 enumeration. (The presence of a @code{default} label prevents this
1673 warning.) @code{case} labels outside the enumeration range also
1674 provoke warnings when this option is used.
1677 Warn if any trigraphs are encountered that might change the meaning of
1678 the program (trigraphs within comments are not warned about).
1680 @item -Wunused-function
1681 Warn whenever a static function is declared but not defined or a
1682 non\-inline static function is unused.
1684 @item -Wunused-label
1685 Warn whenever a label is declared but not used.
1687 To suppress this warning use the @samp{unused} attribute
1688 (@pxref{Variable Attributes}).
1690 @item -Wunused-parameter
1691 Warn whenever a function parameter is unused aside from its declaration.
1693 To suppress this warning use the @samp{unused} attribute
1694 (@pxref{Variable Attributes}).
1696 @item -Wunused-variable
1697 Warn whenever a local variable or non-constant static variable is unused
1698 aside from its declaration
1700 To suppress this warning use the @samp{unused} attribute
1701 (@pxref{Variable Attributes}).
1703 @item -Wunused-value
1704 Warn whenever a statement computes a result that is explicitly not used.
1706 To suppress this warning cast the expression to @samp{void}.
1709 All all the above @samp{-Wunused} options combined.
1711 In order to get a warning about an unused function parameter, you must
1712 either specify @samp{-W -Wunused} or separately specify
1713 @samp{-Wunused-parameter}.
1715 @item -Wuninitialized
1716 Warn if an automatic variable is used without first being initialized or
1717 if a variable may be clobbered by a @code{setjmp} call.
1719 These warnings are possible only in optimizing compilation,
1720 because they require data flow information that is computed only
1721 when optimizing. If you don't specify @samp{-O}, you simply won't
1724 These warnings occur only for variables that are candidates for
1725 register allocation. Therefore, they do not occur for a variable that
1726 is declared @code{volatile}, or whose address is taken, or whose size
1727 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
1728 structures, unions or arrays, even when they are in registers.
1730 Note that there may be no warning about a variable that is used only
1731 to compute a value that itself is never used, because such
1732 computations may be deleted by data flow analysis before the warnings
1735 These warnings are made optional because GCC is not smart
1736 enough to see all the reasons why the code might be correct
1737 despite appearing to have an error. Here is one example of how
1756 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
1757 always initialized, but GCC doesn't know this. Here is
1758 another common case:
1763 if (change_y) save_y = y, y = new_y;
1765 if (change_y) y = save_y;
1770 This has no bug because @code{save_y} is used only if it is set.
1772 @cindex @code{longjmp} warnings
1773 This option also warns when a non-volatile automatic variable might be
1774 changed by a call to @code{longjmp}. These warnings as well are possible
1775 only in optimizing compilation.
1777 The compiler sees only the calls to @code{setjmp}. It cannot know
1778 where @code{longjmp} will be called; in fact, a signal handler could
1779 call it at any point in the code. As a result, you may get a warning
1780 even when there is in fact no problem because @code{longjmp} cannot
1781 in fact be called at the place which would cause a problem.
1783 Some spurious warnings can be avoided if you declare all the functions
1784 you use that never return as @code{noreturn}. @xref{Function
1787 @item -Wreorder (C++ only)
1788 @cindex reordering, warning
1789 @cindex warning for reordering of member initializers
1790 Warn when the order of member initializers given in the code does not
1791 match the order in which they must be executed. For instance:
1793 @item -Wunknown-pragmas
1794 @cindex warning for unknown pragmas
1795 @cindex unknown pragmas, warning
1796 @cindex pragmas, warning of unknown
1797 Warn when a #pragma directive is encountered which is not understood by
1798 GCC. If this command line option is used, warnings will even be issued
1799 for unknown pragmas in system header files. This is not the case if
1800 the warnings were only enabled by the @samp{-Wall} command line option.
1803 All of the above @samp{-W} options combined. This enables all the
1804 warnings about constructions that some users consider questionable, and
1805 that are easy to avoid (or modify to prevent the warning), even in
1806 conjunction with macros.
1808 @item -Wsystem-headers
1809 @cindex warnings from system headers
1810 @cindex system headers, warnings from
1811 Print warning messages for constructs found in system header files.
1812 Warnings from system headers are normally suppressed, on the assumption
1813 that they usually do not indicate real problems and would only make the
1814 compiler output harder to read. Using this command line option tells
1815 GCC to emit warnings from system headers as if they occurred in user
1816 code. However, note that using @samp{-Wall} in conjunction with this
1817 option will @emph{not} warn about unknown pragmas in system
1818 headers---for that, @samp{-Wunknown-pragmas} must also be used.
1821 The following @samp{-W@dots{}} options are not implied by @samp{-Wall}.
1822 Some of them warn about constructions that users generally do not
1823 consider questionable, but which occasionally you might wish to check
1824 for; others warn about constructions that are necessary or hard to avoid
1825 in some cases, and there is no simple way to modify the code to suppress
1830 Print extra warning messages for these events:
1834 A function can return either with or without a value. (Falling
1835 off the end of the function body is considered returning without
1836 a value.) For example, this function would evoke such a
1850 An expression-statement or the left-hand side of a comma expression
1851 contains no side effects.
1852 To suppress the warning, cast the unused expression to void.
1853 For example, an expression such as @samp{x[i,j]} will cause a warning,
1854 but @samp{x[(void)i,j]} will not.
1857 An unsigned value is compared against zero with @samp{<} or @samp{<=}.
1860 A comparison like @samp{x<=y<=z} appears; this is equivalent to
1861 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
1862 that of ordinary mathematical notation.
1865 Storage-class specifiers like @code{static} are not the first things in
1866 a declaration. According to the C Standard, this usage is obsolescent.
1869 The return type of a function has a type qualifier such as @code{const}.
1870 Such a type qualifier has no effect, since the value returned by a
1871 function is not an lvalue. (But don't warn about the GNU extension of
1872 @code{volatile void} return types. That extension will be warned about
1873 if @samp{-pedantic} is specified.)
1876 If @samp{-Wall} or @samp{-Wunused} is also specified, warn about unused
1880 A comparison between signed and unsigned values could produce an
1881 incorrect result when the signed value is converted to unsigned.
1882 (But don't warn if @samp{-Wno-sign-compare} is also specified.)
1885 An aggregate has a partly bracketed initializer.
1886 For example, the following code would evoke such a warning,
1887 because braces are missing around the initializer for @code{x.h}:
1890 struct s @{ int f, g; @};
1891 struct t @{ struct s h; int i; @};
1892 struct t x = @{ 1, 2, 3 @};
1896 An aggregate has an initializer which does not initialize all members.
1897 For example, the following code would cause such a warning, because
1898 @code{x.h} would be implicitly initialized to zero:
1901 struct s @{ int f, g, h; @};
1902 struct s x = @{ 3, 4 @};
1907 Warn if floating point values are used in equality comparisons.
1909 The idea behind this is that sometimes it is convenient (for the
1910 programmer) to consider floating-point values as approximations to
1911 infinitely precise real numbers. If you are doing this, then you need
1912 to compute (by analysing the code, or in some other way) the maximum or
1913 likely maximum error that the computation introduces, and allow for it
1914 when performing comparisons (and when producing output, but that's a
1915 different problem). In particular, instead of testing for equality, you
1916 would check to see whether the two values have ranges that overlap; and
1917 this is done with the relational operators, so equality comparisons are
1920 @item -Wtraditional (C only)
1921 Warn about certain constructs that behave differently in traditional and
1926 Macro arguments occurring within string constants in the macro body.
1927 These would substitute the argument in traditional C, but are part of
1928 the constant in ISO C.
1931 A function declared external in one block and then used after the end of
1935 A @code{switch} statement has an operand of type @code{long}.
1938 A non-@code{static} function declaration follows a @code{static} one.
1939 This construct is not accepted by some traditional C compilers.
1942 The ISO type of an integer constant has a different width or
1943 signedness from its traditional type. This warning is only issued if
1944 the base of the constant is ten. I.e. hexadecimal or octal values, which
1945 typically represent bit patterns, are not warned about.
1948 Usage of ISO string concatenation is detected.
1951 A function macro appears without arguments.
1954 The unary plus operator.
1957 Initialization of automatic aggregates.
1960 Identifier conflicts with labels. Traditional C lacks a separate
1961 namespace for labels.
1964 Initialization of unions. If the initializer is zero, the warning is
1965 omitted. This is done under the assumption that the zero initializer in
1966 user code appears conditioned on e.g. @code{__STDC__} to avoid missing
1967 initializer warnings and relies on default initialization to zero in the
1971 The `U' integer constant suffix, or the `F' or `L' floating point
1972 constant suffixes. (Traditonal C does support the `L' suffix on integer
1973 constants.) Note, these suffixes appear in macros defined in the system
1974 headers of most modern systems, e.g. the _MIN/_MAX macros in limits.h.
1975 Use of these macros can lead to spurious warnings as they do not
1976 necessarily reflect whether the code in question is any less portable to
1977 traditional C given that suitable backup definitions are provided.
1981 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
1984 Warn whenever a local variable shadows another local variable.
1986 @item -Wid-clash-@var{len}
1987 Warn whenever two distinct identifiers match in the first @var{len}
1988 characters. This may help you prepare a program that will compile
1989 with certain obsolete, brain-damaged compilers.
1991 @item -Wlarger-than-@var{len}
1992 Warn whenever an object of larger than @var{len} bytes is defined.
1994 @item -Wpointer-arith
1995 Warn about anything that depends on the ``size of'' a function type or
1996 of @code{void}. GNU C assigns these types a size of 1, for
1997 convenience in calculations with @code{void *} pointers and pointers
2000 @item -Wbad-function-cast (C only)
2001 Warn whenever a function call is cast to a non-matching type.
2002 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2005 Warn whenever a pointer is cast so as to remove a type qualifier from
2006 the target type. For example, warn if a @code{const char *} is cast
2007 to an ordinary @code{char *}.
2010 Warn whenever a pointer is cast such that the required alignment of the
2011 target is increased. For example, warn if a @code{char *} is cast to
2012 an @code{int *} on machines where integers can only be accessed at
2013 two- or four-byte boundaries.
2015 @item -Wwrite-strings
2016 Give string constants the type @code{const char[@var{length}]} so that
2017 copying the address of one into a non-@code{const} @code{char *}
2018 pointer will get a warning. These warnings will help you find at
2019 compile time code that can try to write into a string constant, but
2020 only if you have been very careful about using @code{const} in
2021 declarations and prototypes. Otherwise, it will just be a nuisance;
2022 this is why we did not make @samp{-Wall} request these warnings.
2025 Warn if a prototype causes a type conversion that is different from what
2026 would happen to the same argument in the absence of a prototype. This
2027 includes conversions of fixed point to floating and vice versa, and
2028 conversions changing the width or signedness of a fixed point argument
2029 except when the same as the default promotion.
2031 Also, warn if a negative integer constant expression is implicitly
2032 converted to an unsigned type. For example, warn about the assignment
2033 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2034 casts like @code{(unsigned) -1}.
2036 @item -Wsign-compare
2037 @cindex warning for comparison of signed and unsigned values
2038 @cindex comparison of signed and unsigned values, warning
2039 @cindex signed and unsigned values, comparison warning
2040 Warn when a comparison between signed and unsigned values could produce
2041 an incorrect result when the signed value is converted to unsigned.
2042 This warning is also enabled by @samp{-W}; to get the other warnings
2043 of @samp{-W} without this warning, use @samp{-W -Wno-sign-compare}.
2045 @item -Waggregate-return
2046 Warn if any functions that return structures or unions are defined or
2047 called. (In languages where you can return an array, this also elicits
2050 @item -Wstrict-prototypes (C only)
2051 Warn if a function is declared or defined without specifying the
2052 argument types. (An old-style function definition is permitted without
2053 a warning if preceded by a declaration which specifies the argument
2056 @item -Wmissing-prototypes (C only)
2057 Warn if a global function is defined without a previous prototype
2058 declaration. This warning is issued even if the definition itself
2059 provides a prototype. The aim is to detect global functions that fail
2060 to be declared in header files.
2062 @item -Wmissing-declarations
2063 Warn if a global function is defined without a previous declaration.
2064 Do so even if the definition itself provides a prototype.
2065 Use this option to detect global functions that are not declared in
2068 @item -Wmissing-noreturn
2069 Warn about functions which might be candidates for attribute @code{noreturn}.
2070 Note these are only possible candidates, not absolute ones. Care should
2071 be taken to manually verify functions actually do not ever return before
2072 adding the @code{noreturn} attribute, otherwise subtle code generation
2073 bugs could be introduced.
2075 @item -Wmissing-format-attribute
2076 If @samp{-Wformat} is enabled, also warn about functions which might be
2077 candidates for @code{format} attributes. Note these are only possible
2078 candidates, not absolute ones. GCC will guess that @code{format}
2079 attributes might be appropriate for any function that calls a function
2080 like @code{vprintf} or @code{vscanf}, but this might not always be the
2081 case, and some functions for which @code{format} attributes are
2082 appropriate may not be detected. This option has no effect unless
2083 @samp{-Wformat} is enabled (possibly by @samp{-Wall}).
2086 Warn if a structure is given the packed attribute, but the packed
2087 attribute has no effect on the layout or size of the structure.
2088 Such structures may be mis-aligned for little benefit. For
2089 instance, in this code, the variable @code{f.x} in @code{struct bar}
2090 will be misaligned even though @code{struct bar} does not itself
2091 have the packed attribute:
2098 @} __attribute__((packed));
2107 Warn if padding is included in a structure, either to align an element
2108 of the structure or to align the whole structure. Sometimes when this
2109 happens it is possible to rearrange the fields of the structure to
2110 reduce the padding and so make the structure smaller.
2112 @item -Wredundant-decls
2113 Warn if anything is declared more than once in the same scope, even in
2114 cases where multiple declaration is valid and changes nothing.
2116 @item -Wnested-externs (C only)
2117 Warn if an @code{extern} declaration is encountered within a function.
2119 @item -Wunreachable-code
2120 Warn if the compiler detects that code will never be executed.
2122 This option is intended to warn when the compiler detects that at
2123 least a whole line of source code will never be executed, because
2124 some condition is never satisfied or because it is after a
2125 procedure that never returns.
2127 It is possible for this option to produce a warning even though there
2128 are circumstances under which part of the affected line can be executed,
2129 so care should be taken when removing apparently-unreachable code.
2131 For instance, when a function is inlined, a warning may mean that the
2132 line is unreachable in only one inlined copy of the function.
2134 This option is not made part of @samp{-Wall} because in a debugging
2135 version of a program there is often substantial code which checks
2136 correct functioning of the program and is, hopefully, unreachable
2137 because the program does work. Another common use of unreachable
2138 code is to provide behaviour which is selectable at compile-time.
2141 Warn if a function can not be inlined and it was declared as inline.
2144 Warn if @samp{long long} type is used. This is default. To inhibit
2145 the warning messages, use @samp{-Wno-long-long}. Flags
2146 @samp{-Wlong-long} and @samp{-Wno-long-long} are taken into account
2147 only when @samp{-pedantic} flag is used.
2149 @item -Wdisabled-optimization
2150 Warn if a requested optimization pass is disabled. This warning does
2151 not generally indicate that there is anything wrong with your code; it
2152 merely indicates that GCC's optimizers were unable to handle the code
2153 effectively. Often, the problem is that your code is too big or too
2154 complex; GCC will refuse to optimize programs when the optimization
2155 itself is likely to take inordinate amounts of time.
2158 Make all warnings into errors.
2161 @node Debugging Options
2162 @section Options for Debugging Your Program or GCC
2163 @cindex options, debugging
2164 @cindex debugging information options
2166 GCC has various special options that are used for debugging
2167 either your program or GCC:
2171 Produce debugging information in the operating system's native format
2172 (stabs, COFF, XCOFF, or DWARF). GDB can work with this debugging
2175 On most systems that use stabs format, @samp{-g} enables use of extra
2176 debugging information that only GDB can use; this extra information
2177 makes debugging work better in GDB but will probably make other debuggers
2179 refuse to read the program. If you want to control for certain whether
2180 to generate the extra information, use @samp{-gstabs+}, @samp{-gstabs},
2181 @samp{-gxcoff+}, @samp{-gxcoff}, @samp{-gdwarf-1+}, or @samp{-gdwarf-1}
2184 Unlike most other C compilers, GCC allows you to use @samp{-g} with
2185 @samp{-O}. The shortcuts taken by optimized code may occasionally
2186 produce surprising results: some variables you declared may not exist
2187 at all; flow of control may briefly move where you did not expect it;
2188 some statements may not be executed because they compute constant
2189 results or their values were already at hand; some statements may
2190 execute in different places because they were moved out of loops.
2192 Nevertheless it proves possible to debug optimized output. This makes
2193 it reasonable to use the optimizer for programs that might have bugs.
2195 The following options are useful when GCC is generated with the
2196 capability for more than one debugging format.
2199 Produce debugging information for use by GDB. This means to use the
2200 most expressive format available (DWARF 2, stabs, or the native format
2201 if neither of those are supported), including GDB extensions if at all
2205 Produce debugging information in stabs format (if that is supported),
2206 without GDB extensions. This is the format used by DBX on most BSD
2207 systems. On MIPS, Alpha and System V Release 4 systems this option
2208 produces stabs debugging output which is not understood by DBX or SDB.
2209 On System V Release 4 systems this option requires the GNU assembler.
2212 Produce debugging information in stabs format (if that is supported),
2213 using GNU extensions understood only by the GNU debugger (GDB). The
2214 use of these extensions is likely to make other debuggers crash or
2215 refuse to read the program.
2218 Produce debugging information in COFF format (if that is supported).
2219 This is the format used by SDB on most System V systems prior to
2223 Produce debugging information in XCOFF format (if that is supported).
2224 This is the format used by the DBX debugger on IBM RS/6000 systems.
2227 Produce debugging information in XCOFF format (if that is supported),
2228 using GNU extensions understood only by the GNU debugger (GDB). The
2229 use of these extensions is likely to make other debuggers crash or
2230 refuse to read the program, and may cause assemblers other than the GNU
2231 assembler (GAS) to fail with an error.
2234 Produce debugging information in DWARF version 1 format (if that is
2235 supported). This is the format used by SDB on most System V Release 4
2239 Produce debugging information in DWARF version 1 format (if that is
2240 supported), using GNU extensions understood only by the GNU debugger
2241 (GDB). The use of these extensions is likely to make other debuggers
2242 crash or refuse to read the program.
2245 Produce debugging information in DWARF version 2 format (if that is
2246 supported). This is the format used by DBX on IRIX 6.
2249 @itemx -ggdb@var{level}
2250 @itemx -gstabs@var{level}
2251 @itemx -gcoff@var{level}
2252 @itemx -gxcoff@var{level}
2253 @itemx -gdwarf@var{level}
2254 @itemx -gdwarf-2@var{level}
2255 Request debugging information and also use @var{level} to specify how
2256 much information. The default level is 2.
2258 Level 1 produces minimal information, enough for making backtraces in
2259 parts of the program that you don't plan to debug. This includes
2260 descriptions of functions and external variables, but no information
2261 about local variables and no line numbers.
2263 Level 3 includes extra information, such as all the macro definitions
2264 present in the program. Some debuggers support macro expansion when
2269 Generate extra code to write profile information suitable for the
2270 analysis program @code{prof}. You must use this option when compiling
2271 the source files you want data about, and you must also use it when
2274 @cindex @code{gprof}
2276 Generate extra code to write profile information suitable for the
2277 analysis program @code{gprof}. You must use this option when compiling
2278 the source files you want data about, and you must also use it when
2283 Generate extra code to write profile information for basic blocks, which will
2284 record the number of times each basic block is executed, the basic block start
2285 address, and the function name containing the basic block. If @samp{-g} is
2286 used, the line number and filename of the start of the basic block will also be
2287 recorded. If not overridden by the machine description, the default action is
2288 to append to the text file @file{bb.out}.
2290 This data could be analyzed by a program like @code{tcov}. Note,
2291 however, that the format of the data is not what @code{tcov} expects.
2292 Eventually GNU @code{gprof} should be extended to process this data.
2295 Makes the compiler print out each function name as it is compiled, and
2296 print some statistics about each pass when it finishes.
2299 Generate extra code to profile basic blocks. Your executable will
2300 produce output that is a superset of that produced when @samp{-a} is
2301 used. Additional output is the source and target address of the basic
2302 blocks where a jump takes place, the number of times a jump is executed,
2303 and (optionally) the complete sequence of basic blocks being executed.
2304 The output is appended to file @file{bb.out}.
2306 You can examine different profiling aspects without recompilation. Your
2307 executable will read a list of function names from file @file{bb.in}.
2308 Profiling starts when a function on the list is entered and stops when
2309 that invocation is exited. To exclude a function from profiling, prefix
2310 its name with `-'. If a function name is not unique, you can
2311 disambiguate it by writing it in the form
2312 @samp{/path/filename.d:functionname}. Your executable will write the
2313 available paths and filenames in file @file{bb.out}.
2315 Several function names have a special meaning:
2318 Write source, target and frequency of jumps to file @file{bb.out}.
2319 @item __bb_hidecall__
2320 Exclude function calls from frequency count.
2321 @item __bb_showret__
2322 Include function returns in frequency count.
2324 Write the sequence of basic blocks executed to file @file{bbtrace.gz}.
2325 The file will be compressed using the program @samp{gzip}, which must
2326 exist in your @code{PATH}. On systems without the @samp{popen}
2327 function, the file will be named @file{bbtrace} and will not be
2328 compressed. @strong{Profiling for even a few seconds on these systems
2329 will produce a very large file.} Note: @code{__bb_hidecall__} and
2330 @code{__bb_showret__} will not affect the sequence written to
2334 Here's a short example using different profiling parameters
2335 in file @file{bb.in}. Assume function @code{foo} consists of basic blocks
2336 1 and 2 and is called twice from block 3 of function @code{main}. After
2337 the calls, block 3 transfers control to block 4 of @code{main}.
2339 With @code{__bb_trace__} and @code{main} contained in file @file{bb.in},
2340 the following sequence of blocks is written to file @file{bbtrace.gz}:
2341 0 3 1 2 1 2 4. The return from block 2 to block 3 is not shown, because
2342 the return is to a point inside the block and not to the top. The
2343 block address 0 always indicates, that control is transferred
2344 to the trace from somewhere outside the observed functions. With
2345 @samp{-foo} added to @file{bb.in}, the blocks of function
2346 @code{foo} are removed from the trace, so only 0 3 4 remains.
2348 With @code{__bb_jumps__} and @code{main} contained in file @file{bb.in},
2349 jump frequencies will be written to file @file{bb.out}. The
2350 frequencies are obtained by constructing a trace of blocks
2351 and incrementing a counter for every neighbouring pair of blocks
2352 in the trace. The trace 0 3 1 2 1 2 4 displays the following
2356 Jump from block 0x0 to block 0x3 executed 1 time(s)
2357 Jump from block 0x3 to block 0x1 executed 1 time(s)
2358 Jump from block 0x1 to block 0x2 executed 2 time(s)
2359 Jump from block 0x2 to block 0x1 executed 1 time(s)
2360 Jump from block 0x2 to block 0x4 executed 1 time(s)
2363 With @code{__bb_hidecall__}, control transfer due to call instructions
2364 is removed from the trace, that is the trace is cut into three parts: 0
2365 3 4, 0 1 2 and 0 1 2. With @code{__bb_showret__}, control transfer due
2366 to return instructions is added to the trace. The trace becomes: 0 3 1
2367 2 3 1 2 3 4. Note, that this trace is not the same, as the sequence
2368 written to @file{bbtrace.gz}. It is solely used for counting jump
2371 @item -fprofile-arcs
2372 Instrument @dfn{arcs} during compilation. For each function of your
2373 program, GCC creates a program flow graph, then finds a spanning tree
2374 for the graph. Only arcs that are not on the spanning tree have to be
2375 instrumented: the compiler adds code to count the number of times that these
2376 arcs are executed. When an arc is the only exit or only entrance to a
2377 block, the instrumentation code can be added to the block; otherwise, a
2378 new basic block must be created to hold the instrumentation code.
2380 Since not every arc in the program must be instrumented, programs
2381 compiled with this option run faster than programs compiled with
2382 @samp{-a}, which adds instrumentation code to every basic block in the
2383 program. The tradeoff: since @code{gcov} does not have
2384 execution counts for all branches, it must start with the execution
2385 counts for the instrumented branches, and then iterate over the program
2386 flow graph until the entire graph has been solved. Hence, @code{gcov}
2387 runs a little more slowly than a program which uses information from
2390 @samp{-fprofile-arcs} also makes it possible to estimate branch
2391 probabilities, and to calculate basic block execution counts. In
2392 general, basic block execution counts do not give enough information to
2393 estimate all branch probabilities. When the compiled program exits, it
2394 saves the arc execution counts to a file called
2395 @file{@var{sourcename}.da}. Use the compiler option
2396 @samp{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
2397 Control Optimization}) when recompiling, to optimize using estimated
2398 branch probabilities.
2401 @item -ftest-coverage
2402 Create data files for the @code{gcov} code-coverage utility
2403 (@pxref{Gcov,, @code{gcov}: a GCC Test Coverage Program}).
2404 The data file names begin with the name of your source file:
2407 @item @var{sourcename}.bb
2408 A mapping from basic blocks to line numbers, which @code{gcov} uses to
2409 associate basic block execution counts with line numbers.
2411 @item @var{sourcename}.bbg
2412 A list of all arcs in the program flow graph. This allows @code{gcov}
2413 to reconstruct the program flow graph, so that it can compute all basic
2414 block and arc execution counts from the information in the
2415 @code{@var{sourcename}.da} file (this last file is the output from
2416 @samp{-fprofile-arcs}).
2419 @item -d@var{letters}
2420 Says to make debugging dumps during compilation at times specified by
2421 @var{letters}. This is used for debugging the compiler. The file names
2422 for most of the dumps are made by appending a pass number and a word to
2423 the source file name (e.g. @file{foo.c.00.rtl} or @file{foo.c.01.sibling}).
2424 Here are the possible letters for use in @var{letters}, and their meanings:
2428 Annotate the assembler output with miscellaneous debugging information.
2430 Dump after computing branch probabilities, to @file{@var{file}.11.bp}.
2432 Dump after block reordering, to @file{@var{file}.26.bbro}.
2434 Dump after instruction combination, to the file @file{@var{file}.14.combine}.
2436 Dump after the first if conversion, to the file @file{@var{file}.15.ce}.
2438 Dump after delayed branch scheduling, to @file{@var{file}.29.dbr}.
2440 Dump all macro definitions, at the end of preprocessing, in addition to
2443 Dump after SSA optimizations, to @file{@var{file}.05.ssa} and
2444 @file{@var{file}.06.ussa}.
2446 Dump after the second if conversion, to @file{@var{file}.24.ce2}.
2448 Dump after life analysis, to @file{@var{file}.13.life}.
2450 Dump after purging @code{ADDRESSOF} codes, to @file{@var{file}.04.addressof}.
2452 Dump after global register allocation, to @file{@var{file}.19.greg}.
2454 Dump after post-reload CSE and other optimizations, to @file{@var{file}.20.postreload}.
2456 Dump after GCSE, to @file{@var{file}.08.gcse}.
2458 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
2460 Dump after the first jump optimization, to @file{@var{file}.02.jump}.
2462 Dump after the last jump optimization, to @file{@var{file}.27.jump2}.
2464 Dump after conversion from registers to stack, to @file{@var{file}.29.stack}.
2466 Dump after local register allocation, to @file{@var{file}.18.lreg}.
2468 Dump after loop optimization, to @file{@var{file}.09.loop}.
2470 Dump after performing the machine dependent reorganisation pass, to
2471 @file{@var{file}.28.mach}.
2473 Dump after register renumbering, to @file{@var{file}.23.rnreg}.
2475 Dump after the register move pass, to @file{@var{file}.16.regmove}.
2477 Dump after RTL generation, to @file{@var{file}.00.rtl}.
2479 Dump after the second instruction scheduling pass, to
2480 @file{@var{file}.25.sched2}.
2482 Dump after CSE (including the jump optimization that sometimes follows
2483 CSE), to @file{@var{file}.03.cse}.
2485 Dump after the first instruction scheduling pass, to
2486 @file{@var{file}.17.sched}.
2488 Dump after the second CSE pass (including the jump optimization that
2489 sometimes follows CSE), to @file{@var{file}.10.cse2}.
2491 Dump after the second flow pass, to @file{@var{file}.21.flow2}.
2493 Dump after dead code elimination, to @file{@var{file}.06.dce}.
2495 Dump after the peephole pass, to @file{@var{file}.22.peephole2}.
2497 Produce all the dumps listed above.
2499 Print statistics on memory usage, at the end of the run, to
2502 Annotate the assembler output with a comment indicating which
2503 pattern and alternative was used. The length of each instruction is
2506 Dump the RTL in the assembler output as a comment before each instruction.
2507 Also turns on @samp{-dp} annotation.
2509 For each of the other indicated dump files (except for
2510 @file{@var{file}.00.rtl}), dump a representation of the control flow graph
2511 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
2513 Just generate RTL for a function instead of compiling it. Usually used
2516 Dump debugging information during parsing, to standard error.
2519 @item -fdump-unnumbered
2520 When doing debugging dumps (see -d option above), suppress instruction
2521 numbers and line number note output. This makes it more feasible to
2522 use diff on debugging dumps for compiler invocations with different
2523 options, in particular with and without -g.
2525 @item -fdump-translation-unit-@var{file} (C and C++ only)
2526 Dump a representation of the tree structure for the entire translation
2529 @item -fpretend-float
2530 When running a cross-compiler, pretend that the target machine uses the
2531 same floating point format as the host machine. This causes incorrect
2532 output of the actual floating constants, but the actual instruction
2533 sequence will probably be the same as GCC would make when running on
2537 Store the usual ``temporary'' intermediate files permanently; place them
2538 in the current directory and name them based on the source file. Thus,
2539 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
2540 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
2541 preprocessed @file{foo.i} output file even though the compiler now
2542 normally uses an integrated preprocessor.
2545 Report the CPU time taken by each subprocess in the compilation
2546 sequence. For C source files, this is the compiler proper and assembler
2547 (plus the linker if linking is done). The output looks like this:
2554 The first number on each line is the ``user time,'' that is time spent
2555 executing the program itself. The second number is ``system time,''
2556 time spent executing operating system routines on behalf of the program.
2557 Both numbers are in seconds.
2559 @item -print-file-name=@var{library}
2560 Print the full absolute name of the library file @var{library} that
2561 would be used when linking---and don't do anything else. With this
2562 option, GCC does not compile or link anything; it just prints the
2565 @item -print-prog-name=@var{program}
2566 Like @samp{-print-file-name}, but searches for a program such as @samp{cpp}.
2568 @item -print-libgcc-file-name
2569 Same as @samp{-print-file-name=libgcc.a}.
2571 This is useful when you use @samp{-nostdlib} or @samp{-nodefaultlibs}
2572 but you do want to link with @file{libgcc.a}. You can do
2575 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
2578 @item -print-search-dirs
2579 Print the name of the configured installation directory and a list of
2580 program and library directories gcc will search---and don't do anything else.
2582 This is useful when gcc prints the error message
2583 @samp{installation problem, cannot exec cpp: No such file or directory}.
2584 To resolve this you either need to put @file{cpp} and the other compiler
2585 components where gcc expects to find them, or you can set the environment
2586 variable @code{GCC_EXEC_PREFIX} to the directory where you installed them.
2587 Don't forget the trailing '/'.
2588 @xref{Environment Variables}.
2591 @node Optimize Options
2592 @section Options That Control Optimization
2593 @cindex optimize options
2594 @cindex options, optimization
2596 These options control various sorts of optimizations:
2601 Optimize. Optimizing compilation takes somewhat more time, and a lot
2602 more memory for a large function.
2604 Without @samp{-O}, the compiler's goal is to reduce the cost of
2605 compilation and to make debugging produce the expected results.
2606 Statements are independent: if you stop the program with a breakpoint
2607 between statements, you can then assign a new value to any variable or
2608 change the program counter to any other statement in the function and
2609 get exactly the results you would expect from the source code.
2611 Without @samp{-O}, the compiler only allocates variables declared
2612 @code{register} in registers. The resulting compiled code is a little
2613 worse than produced by PCC without @samp{-O}.
2615 With @samp{-O}, the compiler tries to reduce code size and execution
2618 When you specify @samp{-O}, the compiler turns on @samp{-fthread-jumps}
2619 and @samp{-fdefer-pop} on all machines. The compiler turns on
2620 @samp{-fdelayed-branch} on machines that have delay slots, and
2621 @samp{-fomit-frame-pointer} on machines that can support debugging even
2622 without a frame pointer. On some machines the compiler also turns
2623 on other flags.@refill
2626 Optimize even more. GCC performs nearly all supported optimizations
2627 that do not involve a space-speed tradeoff. The compiler does not
2628 perform loop unrolling or function inlining when you specify @samp{-O2}.
2629 As compared to @samp{-O}, this option increases both compilation time
2630 and the performance of the generated code.
2632 @samp{-O2} turns on all optional optimizations except for loop unrolling,
2633 function inlining, and register renaming. It also turns on the
2634 @samp{-fforce-mem} option on all machines and frame pointer elimination
2635 on machines where doing so does not interfere with debugging.
2638 Optimize yet more. @samp{-O3} turns on all optimizations specified by
2639 @samp{-O2} and also turns on the @samp{-finline-functions} and
2640 @samp{-frename-registers} options.
2646 Optimize for size. @samp{-Os} enables all @samp{-O2} optimizations that
2647 do not typically increase code size. It also performs further
2648 optimizations designed to reduce code size.
2650 If you use multiple @samp{-O} options, with or without level numbers,
2651 the last such option is the one that is effective.
2654 Options of the form @samp{-f@var{flag}} specify machine-independent
2655 flags. Most flags have both positive and negative forms; the negative
2656 form of @samp{-ffoo} would be @samp{-fno-foo}. In the table below,
2657 only one of the forms is listed---the one which is not the default.
2658 You can figure out the other form by either removing @samp{no-} or
2663 Do not store floating point variables in registers, and inhibit other
2664 options that might change whether a floating point value is taken from a
2667 @cindex floating point precision
2668 This option prevents undesirable excess precision on machines such as
2669 the 68000 where the floating registers (of the 68881) keep more
2670 precision than a @code{double} is supposed to have. Similarly for the
2671 x86 architecture. For most programs, the excess precision does only
2672 good, but a few programs rely on the precise definition of IEEE floating
2673 point. Use @samp{-ffloat-store} for such programs, after modifying
2674 them to store all pertinent intermediate computations into variables.
2676 @item -fno-default-inline
2677 Do not make member functions inline by default merely because they are
2678 defined inside the class scope (C++ only). Otherwise, when you specify
2679 @w{@samp{-O}}, member functions defined inside class scope are compiled
2680 inline by default; i.e., you don't need to add @samp{inline} in front of
2681 the member function name.
2683 @item -fno-defer-pop
2684 Always pop the arguments to each function call as soon as that function
2685 returns. For machines which must pop arguments after a function call,
2686 the compiler normally lets arguments accumulate on the stack for several
2687 function calls and pops them all at once.
2690 Force memory operands to be copied into registers before doing
2691 arithmetic on them. This produces better code by making all memory
2692 references potential common subexpressions. When they are not common
2693 subexpressions, instruction combination should eliminate the separate
2694 register-load. The @samp{-O2} option turns on this option.
2697 Force memory address constants to be copied into registers before
2698 doing arithmetic on them. This may produce better code just as
2699 @samp{-fforce-mem} may.
2701 @item -fomit-frame-pointer
2702 Don't keep the frame pointer in a register for functions that
2703 don't need one. This avoids the instructions to save, set up and
2704 restore frame pointers; it also makes an extra register available
2705 in many functions. @strong{It also makes debugging impossible on
2709 On some machines, such as the Vax, this flag has no effect, because
2710 the standard calling sequence automatically handles the frame pointer
2711 and nothing is saved by pretending it doesn't exist. The
2712 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2713 whether a target machine supports this flag. @xref{Registers}.@refill
2716 On some machines, such as the Vax, this flag has no effect, because
2717 the standard calling sequence automatically handles the frame pointer
2718 and nothing is saved by pretending it doesn't exist. The
2719 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2720 whether a target machine supports this flag. @xref{Registers,,Register
2721 Usage, gcc.info, Using and Porting GCC}.@refill
2724 @item -foptimize-sibling-calls
2725 Optimize sibling and tail recursive calls.
2728 This option generates traps for signed overflow on addition, subtraction,
2729 multiplication operations.
2732 Don't pay attention to the @code{inline} keyword. Normally this option
2733 is used to keep the compiler from expanding any functions inline.
2734 Note that if you are not optimizing, no functions can be expanded inline.
2736 @item -finline-functions
2737 Integrate all simple functions into their callers. The compiler
2738 heuristically decides which functions are simple enough to be worth
2739 integrating in this way.
2741 If all calls to a given function are integrated, and the function is
2742 declared @code{static}, then the function is normally not output as
2743 assembler code in its own right.
2745 @item -finline-limit=@var{n}
2746 By default, gcc limits the size of functions that can be inlined. This flag
2747 allows the control of this limit for functions that are explicitly marked as
2748 inline (ie marked with the inline keyword or defined within the class
2749 definition in c++). @var{n} is the size of functions that can be inlined in
2750 number of pseudo instructions (not counting parameter handling). The default
2751 value of n is 10000. Increasing this value can result in more inlined code at
2752 the cost of compilation time and memory consumption. Decreasing usually makes
2753 the compilation faster and less code will be inlined (which presumably
2754 means slower programs). This option is particularly useful for programs that
2755 use inlining heavily such as those based on recursive templates with c++.
2757 @emph{Note:} pseudo instruction represents, in this particular context, an
2758 abstract measurement of function's size. In no way, it represents a count
2759 of assembly instructions and as such its exact meaning might change from one
2760 release to an another.
2762 @item -fkeep-inline-functions
2763 Even if all calls to a given function are integrated, and the function
2764 is declared @code{static}, nevertheless output a separate run-time
2765 callable version of the function. This switch does not affect
2766 @code{extern inline} functions.
2768 @item -fkeep-static-consts
2769 Emit variables declared @code{static const} when optimization isn't turned
2770 on, even if the variables aren't referenced.
2772 GCC enables this option by default. If you want to force the compiler to
2773 check if the variable was referenced, regardless of whether or not
2774 optimization is turned on, use the @samp{-fno-keep-static-consts} option.
2776 @item -fno-function-cse
2777 Do not put function addresses in registers; make each instruction that
2778 calls a constant function contain the function's address explicitly.
2780 This option results in less efficient code, but some strange hacks
2781 that alter the assembler output may be confused by the optimizations
2782 performed when this option is not used.
2785 This option allows GCC to violate some ISO or IEEE rules and/or
2786 specifications in the interest of optimizing code for speed. For
2787 example, it allows the compiler to assume arguments to the @code{sqrt}
2788 function are non-negative numbers and that no floating-point values
2791 This option should never be turned on by any @samp{-O} option since
2792 it can result in incorrect output for programs which depend on
2793 an exact implementation of IEEE or ISO rules/specifications for
2796 @item -fno-math-errno
2797 Do not set ERRNO after calling math functions that are executed
2798 with a single instruction, e.g., sqrt. A program that relies on
2799 IEEE exceptions for math error handling may want to use this flag
2800 for speed while maintaining IEEE arithmetic compatibility.
2802 The default is @samp{-fmath-errno}. The @samp{-ffast-math} option
2803 sets @samp{-fno-math-errno}.
2806 @c following causes underfulls.. they don't look great, but we deal.
2808 The following options control specific optimizations. The @samp{-O2}
2809 option turns on all of these optimizations except @samp{-funroll-loops}
2810 and @samp{-funroll-all-loops}. On most machines, the @samp{-O} option
2811 turns on the @samp{-fthread-jumps} and @samp{-fdelayed-branch} options,
2812 but specific machines may handle it differently.
2814 You can use the following flags in the rare cases when ``fine-tuning''
2815 of optimizations to be performed is desired.
2818 @item -fstrength-reduce
2819 Perform the optimizations of loop strength reduction and
2820 elimination of iteration variables.
2822 @item -fthread-jumps
2823 Perform optimizations where we check to see if a jump branches to a
2824 location where another comparison subsumed by the first is found. If
2825 so, the first branch is redirected to either the destination of the
2826 second branch or a point immediately following it, depending on whether
2827 the condition is known to be true or false.
2829 @item -fcse-follow-jumps
2830 In common subexpression elimination, scan through jump instructions
2831 when the target of the jump is not reached by any other path. For
2832 example, when CSE encounters an @code{if} statement with an
2833 @code{else} clause, CSE will follow the jump when the condition
2836 @item -fcse-skip-blocks
2837 This is similar to @samp{-fcse-follow-jumps}, but causes CSE to
2838 follow jumps which conditionally skip over blocks. When CSE
2839 encounters a simple @code{if} statement with no else clause,
2840 @samp{-fcse-skip-blocks} causes CSE to follow the jump around the
2841 body of the @code{if}.
2843 @item -frerun-cse-after-loop
2844 Re-run common subexpression elimination after loop optimizations has been
2847 @item -frerun-loop-opt
2848 Run the loop optimizer twice.
2851 Perform a global common subexpression elimination pass.
2852 This pass also performs global constant and copy propagation.
2854 @item -fdelete-null-pointer-checks
2855 Use global dataflow analysis to identify and eliminate useless null
2856 pointer checks. Programs which rely on NULL pointer dereferences @emph{not}
2857 halting the program may not work properly with this option. Use
2858 -fno-delete-null-pointer-checks to disable this optimizing for programs
2859 which depend on that behavior.
2861 @item -fexpensive-optimizations
2862 Perform a number of minor optimizations that are relatively expensive.
2864 @item -foptimize-register-move
2866 Attempt to reassign register numbers in move instructions and as
2867 operands of other simple instructions in order to maximize the amount of
2868 register tying. This is especially helpful on machines with two-operand
2869 instructions. GCC enables this optimization by default with @samp{-O2}
2872 Note @code{-fregmove} and @code{-foptimize-register-move} are the same
2875 @item -fdelayed-branch
2876 If supported for the target machine, attempt to reorder instructions
2877 to exploit instruction slots available after delayed branch
2880 @item -fschedule-insns
2881 If supported for the target machine, attempt to reorder instructions to
2882 eliminate execution stalls due to required data being unavailable. This
2883 helps machines that have slow floating point or memory load instructions
2884 by allowing other instructions to be issued until the result of the load
2885 or floating point instruction is required.
2887 @item -fschedule-insns2
2888 Similar to @samp{-fschedule-insns}, but requests an additional pass of
2889 instruction scheduling after register allocation has been done. This is
2890 especially useful on machines with a relatively small number of
2891 registers and where memory load instructions take more than one cycle.
2893 @item -ffunction-sections
2894 @itemx -fdata-sections
2895 Place each function or data item into its own section in the output
2896 file if the target supports arbitrary sections. The name of the
2897 function or the name of the data item determines the section's name
2900 Use these options on systems where the linker can perform optimizations
2901 to improve locality of reference in the instruction space. HPPA
2902 processors running HP-UX and Sparc processors running Solaris 2 have
2903 linkers with such optimizations. Other systems using the ELF object format
2904 as well as AIX may have these optimizations in the future.
2906 Only use these options when there are significant benefits from doing
2907 so. When you specify these options, the assembler and linker will
2908 create larger object and executable files and will also be slower.
2909 You will not be able to use @code{gprof} on all systems if you
2910 specify this option and you may have problems with debugging if
2911 you specify both this option and @samp{-g}.
2913 @item -fcaller-saves
2914 Enable values to be allocated in registers that will be clobbered by
2915 function calls, by emitting extra instructions to save and restore the
2916 registers around such calls. Such allocation is done only when it
2917 seems to result in better code than would otherwise be produced.
2919 This option is always enabled by default on certain machines, usually
2920 those which have no call-preserved registers to use instead.
2922 For all machines, optimization level 2 and higher enables this flag by
2925 @item -funroll-loops
2926 Perform the optimization of loop unrolling. This is only done for loops
2927 whose number of iterations can be determined at compile time or run time.
2928 @samp{-funroll-loops} implies both @samp{-fstrength-reduce} and
2929 @samp{-frerun-cse-after-loop}.
2931 @item -funroll-all-loops
2932 Perform the optimization of loop unrolling. This is done for all loops
2933 and usually makes programs run more slowly. @samp{-funroll-all-loops}
2934 implies @samp{-fstrength-reduce} as well as @samp{-frerun-cse-after-loop}.
2936 @item -fmove-all-movables
2937 Forces all invariant computations in loops to be moved
2940 @item -freduce-all-givs
2941 Forces all general-induction variables in loops to be
2944 @emph{Note:} When compiling programs written in Fortran,
2945 @samp{-fmove-all-movables} and @samp{-freduce-all-givs} are enabled
2946 by default when you use the optimizer.
2948 These options may generate better or worse code; results are highly
2949 dependent on the structure of loops within the source code.
2951 These two options are intended to be removed someday, once
2952 they have helped determine the efficacy of various
2953 approaches to improving loop optimizations.
2955 Please let us (@email{gcc@@gcc.gnu.org} and @email{fortran@@gnu.org})
2956 know how use of these options affects
2957 the performance of your production code.
2958 We're very interested in code that runs @emph{slower}
2959 when these options are @emph{enabled}.
2962 Disable any machine-specific peephole optimizations.
2964 @item -fbranch-probabilities
2965 After running a program compiled with @samp{-fprofile-arcs}
2966 (@pxref{Debugging Options,, Options for Debugging Your Program or
2967 @code{gcc}}), you can compile it a second time using
2968 @samp{-fbranch-probabilities}, to improve optimizations based on
2969 guessing the path a branch might take.
2972 With @samp{-fbranch-probabilities}, GCC puts a @samp{REG_EXEC_COUNT}
2973 note on the first instruction of each basic block, and a
2974 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
2975 These can be used to improve optimization. Currently, they are only
2976 used in one place: in @file{reorg.c}, instead of guessing which path a
2977 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
2978 exactly determine which path is taken more often.
2981 @item -fstrict-aliasing
2982 Allows the compiler to assume the strictest aliasing rules applicable to
2983 the language being compiled. For C (and C++), this activates
2984 optimizations based on the type of expressions. In particular, an
2985 object of one type is assumed never to reside at the same address as an
2986 object of a different type, unless the types are almost the same. For
2987 example, an @code{unsigned int} can alias an @code{int}, but not a
2988 @code{void*} or a @code{double}. A character type may alias any other
2991 Pay special attention to code like this:
3004 The practice of reading from a different union member than the one most
3005 recently written to (called ``type-punning'') is common. Even with
3006 @samp{-fstrict-aliasing}, type-punning is allowed, provided the memory
3007 is accessed through the union type. So, the code above will work as
3008 expected. However, this code might not:
3020 Every language that wishes to perform language-specific alias analysis
3021 should define a function that computes, given an @code{tree}
3022 node, an alias set for the node. Nodes in different alias sets are not
3023 allowed to alias. For an example, see the C front-end function
3024 @code{c_get_alias_set}.
3027 @item -falign-functions
3028 @itemx -falign-functions=@var{n}
3029 Align the start of functions to the next power-of-two greater than
3030 @var{n}, skipping up to @var{n} bytes. For instance,
3031 @samp{-falign-functions=32} aligns functions to the next 32-byte
3032 boundary, but @samp{-falign-functions=24} would align to the next
3033 32-byte boundary only if this can be done by skipping 23 bytes or less.
3035 @samp{-fno-align-functions} and @samp{-falign-functions=1} are
3036 equivalent and mean that functions will not be aligned.
3038 Some assemblers only support this flag when @var{n} is a power of two;
3039 in that case, it is rounded up.
3041 If @var{n} is not specified, use a machine-dependent default.
3043 @item -falign-labels
3044 @itemx -falign-labels=@var{n}
3045 Align all branch targets to a power-of-two boundary, skipping up to
3046 @var{n} bytes like @samp{-falign-functions}. This option can easily
3047 make code slower, because it must insert dummy operations for when the
3048 branch target is reached in the usual flow of the code.
3050 If @samp{-falign-loops} or @samp{-falign-jumps} are applicable and
3051 are greater than this value, then their values are used instead.
3053 If @var{n} is not specified, use a machine-dependent default which is
3054 very likely to be @samp{1}, meaning no alignment.
3057 @itemx -falign-loops=@var{n}
3058 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
3059 like @samp{-falign-functions}. The hope is that the loop will be
3060 executed many times, which will make up for any execution of the dummy
3063 If @var{n} is not specified, use a machine-dependent default.
3066 @itemx -falign-jumps=@var{n}
3067 Align branch targets to a power-of-two boundary, for branch targets
3068 where the targets can only be reached by jumping, skipping up to @var{n}
3069 bytes like @samp{-falign-functions}. In this case, no dummy operations
3072 If @var{n} is not specified, use a machine-dependent default.
3075 Perform optimizations in static single assignment form. Each function's
3076 flow graph is translated into SSA form, optimizations are performed, and
3077 the flow graph is translated back from SSA form. User's should not
3078 specify this option, since it is not yet ready for production use.
3081 Perform dead-code elimination in SSA form. Requires @samp{-fssa}. Like
3082 @samp{-fssa}, this is an experimental feature.
3084 @item -fsingle-precision-constant
3085 Treat floating point constant as single precision constant instead of
3086 implicitly converting it to double precision constant.
3088 @item -frename-registers
3089 Attempt to avoid false dependancies in scheduled code by making use
3090 of registers left over after register allocation. This optimization
3091 will most benefit processors with lots of registers. It can, however,
3092 make debugging impossible, since variables will no longer stay in
3093 a ``home register''.
3096 @node Preprocessor Options
3097 @section Options Controlling the Preprocessor
3098 @cindex preprocessor options
3099 @cindex options, preprocessor
3101 These options control the C preprocessor, which is run on each C source
3102 file before actual compilation.
3104 If you use the @samp{-E} option, nothing is done except preprocessing.
3105 Some of these options make sense only together with @samp{-E} because
3106 they cause the preprocessor output to be unsuitable for actual
3110 @item -include @var{file}
3111 Process @var{file} as input before processing the regular input file.
3112 In effect, the contents of @var{file} are compiled first. Any @samp{-D}
3113 and @samp{-U} options on the command line are always processed before
3114 @samp{-include @var{file}}, regardless of the order in which they are
3115 written. All the @samp{-include} and @samp{-imacros} options are
3116 processed in the order in which they are written.
3118 @item -imacros @var{file}
3119 Process @var{file} as input, discarding the resulting output, before
3120 processing the regular input file. Because the output generated from
3121 @var{file} is discarded, the only effect of @samp{-imacros @var{file}}
3122 is to make the macros defined in @var{file} available for use in the
3125 Any @samp{-D} and @samp{-U} options on the command line are always
3126 processed before @samp{-imacros @var{file}}, regardless of the order in
3127 which they are written. All the @samp{-include} and @samp{-imacros}
3128 options are processed in the order in which they are written.
3130 @item -idirafter @var{dir}
3131 @cindex second include path
3132 Add the directory @var{dir} to the second include path. The directories
3133 on the second include path are searched when a header file is not found
3134 in any of the directories in the main include path (the one that
3137 @item -iprefix @var{prefix}
3138 Specify @var{prefix} as the prefix for subsequent @samp{-iwithprefix}
3141 @item -iwithprefix @var{dir}
3142 Add a directory to the second include path. The directory's name is
3143 made by concatenating @var{prefix} and @var{dir}, where @var{prefix} was
3144 specified previously with @samp{-iprefix}. If you have not specified a
3145 prefix yet, the directory containing the installed passes of the
3146 compiler is used as the default.
3148 @item -iwithprefixbefore @var{dir}
3149 Add a directory to the main include path. The directory's name is made
3150 by concatenating @var{prefix} and @var{dir}, as in the case of
3151 @samp{-iwithprefix}.
3153 @item -isystem @var{dir}
3154 Add a directory to the beginning of the second include path, marking it
3155 as a system directory, so that it gets the same special treatment as
3156 is applied to the standard system directories.
3159 Do not search the standard system directories for header files. Only
3160 the directories you have specified with @samp{-I} options (and the
3161 current directory, if appropriate) are searched. @xref{Directory
3162 Options}, for information on @samp{-I}.
3164 By using both @samp{-nostdinc} and @samp{-I-}, you can limit the include-file
3165 search path to only those directories you specify explicitly.
3168 Do not predefine any nonstandard macros. (Including architecture flags).
3171 Run only the C preprocessor. Preprocess all the C source files
3172 specified and output the results to standard output or to the
3173 specified output file.
3176 Tell the preprocessor not to discard comments. Used with the
3180 Tell the preprocessor not to generate @samp{#line} directives.
3181 Used with the @samp{-E} option.
3184 @cindex dependencies, make
3186 Tell the preprocessor to output a rule suitable for @code{make}
3187 describing the dependencies of each object file. For each source file,
3188 the preprocessor outputs one @code{make}-rule whose target is the object
3189 file name for that source file and whose dependencies are all the
3190 @code{#include} header files it uses. This rule may be a single line or
3191 may be continued with @samp{\}-newline if it is long. The list of rules
3192 is printed on standard output instead of the preprocessed C program.
3194 @samp{-M} implies @samp{-E}.
3196 Another way to specify output of a @code{make} rule is by setting
3197 the environment variable @code{DEPENDENCIES_OUTPUT} (@pxref{Environment
3201 Like @samp{-M} but the output mentions only the user header files
3202 included with @samp{#include "@var{file}"}. System header files
3203 included with @samp{#include <@var{file}>} are omitted.
3206 Like @samp{-M} but the dependency information is written to a file made by
3207 replacing ".c" with ".d" at the end of the input file names.
3208 This is in addition to compiling the file as specified---@samp{-MD} does
3209 not inhibit ordinary compilation the way @samp{-M} does.
3211 In Mach, you can use the utility @code{md} to merge multiple dependency
3212 files into a single dependency file suitable for using with the @samp{make}
3216 Like @samp{-MD} except mention only user header files, not system
3220 Treat missing header files as generated files and assume they live in the
3221 same directory as the source file. If you specify @samp{-MG}, you
3222 must also specify either @samp{-M} or @samp{-MM}. @samp{-MG} is not
3223 supported with @samp{-MD} or @samp{-MMD}.
3226 Print the name of each header file used, in addition to other normal
3229 @item -A@var{question}(@var{answer})
3230 Assert the answer @var{answer} for @var{question}, in case it is tested
3231 with a preprocessing conditional such as @samp{#if
3232 #@var{question}(@var{answer})}. @samp{-A-} disables the standard
3233 assertions that normally describe the target machine.
3236 Define macro @var{macro} with the string @samp{1} as its definition.
3238 @item -D@var{macro}=@var{defn}
3239 Define macro @var{macro} as @var{defn}. All instances of @samp{-D} on
3240 the command line are processed before any @samp{-U} options.
3243 Undefine macro @var{macro}. @samp{-U} options are evaluated after all
3244 @samp{-D} options, but before any @samp{-include} and @samp{-imacros}
3248 Tell the preprocessor to output only a list of the macro definitions
3249 that are in effect at the end of preprocessing. Used with the @samp{-E}
3253 Tell the preprocessing to pass all macro definitions into the output, in
3254 their proper sequence in the rest of the output.
3257 Like @samp{-dD} except that the macro arguments and contents are omitted.
3258 Only @samp{#define @var{name}} is included in the output.
3261 Support ISO C trigraphs. The @samp{-ansi} option also has this effect.
3263 @item -Wp,@var{option}
3264 Pass @var{option} as an option to the preprocessor. If @var{option}
3265 contains commas, it is split into multiple options at the commas.
3268 @node Assembler Options
3269 @section Passing Options to the Assembler
3271 @c prevent bad page break with this line
3272 You can pass options to the assembler.
3275 @item -Wa,@var{option}
3276 Pass @var{option} as an option to the assembler. If @var{option}
3277 contains commas, it is split into multiple options at the commas.
3281 @section Options for Linking
3282 @cindex link options
3283 @cindex options, linking
3285 These options come into play when the compiler links object files into
3286 an executable output file. They are meaningless if the compiler is
3287 not doing a link step.
3291 @item @var{object-file-name}
3292 A file name that does not end in a special recognized suffix is
3293 considered to name an object file or library. (Object files are
3294 distinguished from libraries by the linker according to the file
3295 contents.) If linking is done, these object files are used as input
3301 If any of these options is used, then the linker is not run, and
3302 object file names should not be used as arguments. @xref{Overall
3306 @item -l@var{library}
3307 Search the library named @var{library} when linking.
3309 It makes a difference where in the command you write this option; the
3310 linker searches processes libraries and object files in the order they
3311 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
3312 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
3313 to functions in @samp{z}, those functions may not be loaded.
3315 The linker searches a standard list of directories for the library,
3316 which is actually a file named @file{lib@var{library}.a}. The linker
3317 then uses this file as if it had been specified precisely by name.
3319 The directories searched include several standard system directories
3320 plus any that you specify with @samp{-L}.
3322 Normally the files found this way are library files---archive files
3323 whose members are object files. The linker handles an archive file by
3324 scanning through it for members which define symbols that have so far
3325 been referenced but not defined. But if the file that is found is an
3326 ordinary object file, it is linked in the usual fashion. The only
3327 difference between using an @samp{-l} option and specifying a file name
3328 is that @samp{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
3329 and searches several directories.
3332 You need this special case of the @samp{-l} option in order to
3333 link an Objective C program.
3336 Do not use the standard system startup files when linking.
3337 The standard system libraries are used normally, unless @code{-nostdlib}
3338 or @code{-nodefaultlibs} is used.
3340 @item -nodefaultlibs
3341 Do not use the standard system libraries when linking.
3342 Only the libraries you specify will be passed to the linker.
3343 The standard startup files are used normally, unless @code{-nostartfiles}
3344 is used. The compiler may generate calls to memcmp, memset, and memcpy
3345 for System V (and ISO C) environments or to bcopy and bzero for
3346 BSD environments. These entries are usually resolved by entries in
3347 libc. These entry points should be supplied through some other
3348 mechanism when this option is specified.
3351 Do not use the standard system startup files or libraries when linking.
3352 No startup files and only the libraries you specify will be passed to
3353 the linker. The compiler may generate calls to memcmp, memset, and memcpy
3354 for System V (and ISO C) environments or to bcopy and bzero for
3355 BSD environments. These entries are usually resolved by entries in
3356 libc. These entry points should be supplied through some other
3357 mechanism when this option is specified.
3359 @cindex @code{-lgcc}, use with @code{-nostdlib}
3360 @cindex @code{-nostdlib} and unresolved references
3361 @cindex unresolved references and @code{-nostdlib}
3362 @cindex @code{-lgcc}, use with @code{-nodefaultlibs}
3363 @cindex @code{-nodefaultlibs} and unresolved references
3364 @cindex unresolved references and @code{-nodefaultlibs}
3365 One of the standard libraries bypassed by @samp{-nostdlib} and
3366 @samp{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
3367 that GCC uses to overcome shortcomings of particular machines, or special
3368 needs for some languages.
3370 (@xref{Interface,,Interfacing to GCC Output}, for more discussion of
3374 (@xref{Interface,,Interfacing to GCC Output,gcc.info,Porting GCC},
3375 for more discussion of @file{libgcc.a}.)
3377 In most cases, you need @file{libgcc.a} even when you want to avoid
3378 other standard libraries. In other words, when you specify @samp{-nostdlib}
3379 or @samp{-nodefaultlibs} you should usually specify @samp{-lgcc} as well.
3380 This ensures that you have no unresolved references to internal GCC
3381 library subroutines. (For example, @samp{__main}, used to ensure C++
3382 constructors will be called; @pxref{Collect2,,@code{collect2}}.)
3385 Remove all symbol table and relocation information from the executable.
3388 On systems that support dynamic linking, this prevents linking with the shared
3389 libraries. On other systems, this option has no effect.
3392 Produce a shared object which can then be linked with other objects to
3393 form an executable. Not all systems support this option. For predictable
3394 results, you must also specify the same set of options that were used to
3395 generate code (@samp{-fpic}, @samp{-fPIC}, or model suboptions)
3396 when you specify this option.@footnote{On some systems, @code{gcc -shared}
3397 needs to build supplementary stub code for constructors to work. On
3398 multi-libbed systems, @code{gcc -shared} must select the correct support
3399 libraries to link against. Failing to supply the correct flags may lead
3400 to subtle defects. Supplying them in cases where they are not necessary
3404 Bind references to global symbols when building a shared object. Warn
3405 about any unresolved references (unless overridden by the link editor
3406 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
3409 @item -Xlinker @var{option}
3410 Pass @var{option} as an option to the linker. You can use this to
3411 supply system-specific linker options which GCC does not know how to
3414 If you want to pass an option that takes an argument, you must use
3415 @samp{-Xlinker} twice, once for the option and once for the argument.
3416 For example, to pass @samp{-assert definitions}, you must write
3417 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
3418 @samp{-Xlinker "-assert definitions"}, because this passes the entire
3419 string as a single argument, which is not what the linker expects.
3421 @item -Wl,@var{option}
3422 Pass @var{option} as an option to the linker. If @var{option} contains
3423 commas, it is split into multiple options at the commas.
3425 @item -u @var{symbol}
3426 Pretend the symbol @var{symbol} is undefined, to force linking of
3427 library modules to define it. You can use @samp{-u} multiple times with
3428 different symbols to force loading of additional library modules.
3431 @node Directory Options
3432 @section Options for Directory Search
3433 @cindex directory options
3434 @cindex options, directory search
3437 These options specify directories to search for header files, for
3438 libraries and for parts of the compiler:
3442 Add the directory @var{dir} to the head of the list of directories to be
3443 searched for header files. This can be used to override a system header
3444 file, substituting your own version, since these directories are
3445 searched before the system header file directories. If you use more
3446 than one @samp{-I} option, the directories are scanned in left-to-right
3447 order; the standard system directories come after.
3450 Any directories you specify with @samp{-I} options before the @samp{-I-}
3451 option are searched only for the case of @samp{#include "@var{file}"};
3452 they are not searched for @samp{#include <@var{file}>}.
3454 If additional directories are specified with @samp{-I} options after
3455 the @samp{-I-}, these directories are searched for all @samp{#include}
3456 directives. (Ordinarily @emph{all} @samp{-I} directories are used
3459 In addition, the @samp{-I-} option inhibits the use of the current
3460 directory (where the current input file came from) as the first search
3461 directory for @samp{#include "@var{file}"}. There is no way to
3462 override this effect of @samp{-I-}. With @samp{-I.} you can specify
3463 searching the directory which was current when the compiler was
3464 invoked. That is not exactly the same as what the preprocessor does
3465 by default, but it is often satisfactory.
3467 @samp{-I-} does not inhibit the use of the standard system directories
3468 for header files. Thus, @samp{-I-} and @samp{-nostdinc} are
3472 Add directory @var{dir} to the list of directories to be searched
3475 @item -B@var{prefix}
3476 This option specifies where to find the executables, libraries,
3477 include files, and data files of the compiler itself.
3479 The compiler driver program runs one or more of the subprograms
3480 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
3481 @var{prefix} as a prefix for each program it tries to run, both with and
3482 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
3484 For each subprogram to be run, the compiler driver first tries the
3485 @samp{-B} prefix, if any. If that name is not found, or if @samp{-B}
3486 was not specified, the driver tries two standard prefixes, which are
3487 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc-lib/}. If neither of
3488 those results in a file name that is found, the unmodified program
3489 name is searched for using the directories specified in your
3490 @samp{PATH} environment variable.
3492 @samp{-B} prefixes that effectively specify directory names also apply
3493 to libraries in the linker, because the compiler translates these
3494 options into @samp{-L} options for the linker. They also apply to
3495 includes files in the preprocessor, because the compiler translates these
3496 options into @samp{-isystem} options for the preprocessor. In this case,
3497 the compiler appends @samp{include} to the prefix.
3499 The run-time support file @file{libgcc.a} can also be searched for using
3500 the @samp{-B} prefix, if needed. If it is not found there, the two
3501 standard prefixes above are tried, and that is all. The file is left
3502 out of the link if it is not found by those means.
3504 Another way to specify a prefix much like the @samp{-B} prefix is to use
3505 the environment variable @code{GCC_EXEC_PREFIX}. @xref{Environment
3508 @item -specs=@var{file}
3509 Process @var{file} after the compiler reads in the standard @file{specs}
3510 file, in order to override the defaults that the @file{gcc} driver
3511 program uses when determining what switches to pass to @file{cc1},
3512 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
3513 @samp{-specs=}@var{file} can be specified on the command line, and they
3514 are processed in order, from left to right.
3518 @section Specifying subprocesses and the switches to pass to them
3520 @code{GCC} is a driver program. It performs its job by invoking a
3521 sequence of other programs to do the work of compiling, assembling and
3522 linking. GCC interprets its command-line parameters and uses these to
3523 deduce which programs it should invoke, and which command-line options
3524 it ought to place on their command lines. This behaviour is controlled
3525 by @dfn{spec strings}. In most cases there is one spec string for each
3526 program that GCC can invoke, but a few programs have multiple spec
3527 strings to control their behaviour. The spec strings built into GCC can
3528 be overridden by using the @samp{-specs=} command-line switch to specify
3531 @dfn{Spec files} are plaintext files that are used to construct spec
3532 strings. They consist of a sequence of directives separated by blank
3533 lines. The type of directive is determined by the first non-whitespace
3534 character on the line and it can be one of the following:
3537 @item %@var{command}
3538 Issues a @var{command} to the spec file processor. The commands that can
3542 @item %include <@var{file}>
3544 Search for @var{file} and insert its text at the current point in the
3547 @item %include_noerr <@var{file}>
3548 @cindex %include_noerr
3549 Just like @samp{%include}, but do not generate an error message if the include
3550 file cannot be found.
3552 @item %rename @var{old_name} @var{new_name}
3554 Rename the spec string @var{old_name} to @var{new_name}.
3558 @item *[@var{spec_name}]:
3559 This tells the compiler to create, override or delete the named spec
3560 string. All lines after this directive up to the next directive or
3561 blank line are considered to be the text for the spec string. If this
3562 results in an empty string then the spec will be deleted. (Or, if the
3563 spec did not exist, then nothing will happened.) Otherwise, if the spec
3564 does not currently exist a new spec will be created. If the spec does
3565 exist then its contents will be overridden by the text of this
3566 directive, unless the first character of that text is the @samp{+}
3567 character, in which case the text will be appended to the spec.
3569 @item [@var{suffix}]:
3570 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
3571 and up to the next directive or blank line are considered to make up the
3572 spec string for the indicated suffix. When the compiler encounters an
3573 input file with the named suffix, it will processes the spec string in
3574 order to work out how to compile that file. For example:
3581 This says that any input file whose name ends in @samp{.ZZ} should be
3582 passed to the program @samp{z-compile}, which should be invoked with the
3583 command-line switch @samp{-input} and with the result of performing the
3584 @samp{%i} substitution. (See below.)
3586 As an alternative to providing a spec string, the text that follows a
3587 suffix directive can be one of the following:
3590 @item @@@var{language}
3591 This says that the suffix is an alias for a known @var{language}. This is
3592 similar to using the @code{-x} command-line switch to GCC to specify a
3593 language explicitly. For example:
3600 Says that .ZZ files are, in fact, C++ source files.
3603 This causes an error messages saying:
3606 @var{name} compiler not installed on this system.
3610 GCC already has an extensive list of suffixes built into it.
3611 This directive will add an entry to the end of the list of suffixes, but
3612 since the list is searched from the end backwards, it is effectively
3613 possible to override earlier entries using this technique.
3617 GCC has the following spec strings built into it. Spec files can
3618 override these strings or create their own. Note that individual
3619 targets can also add their own spec strings to this list.
3622 asm Options to pass to the assembler
3623 asm_final Options to pass to the assembler post-processor
3624 cpp Options to pass to the C preprocessor
3625 cc1 Options to pass to the C compiler
3626 cc1plus Options to pass to the C++ compiler
3627 endfile Object files to include at the end of the link
3628 link Options to pass to the linker
3629 lib Libraries to include on the command line to the linker
3630 libgcc Decides which GCC support library to pass to the linker
3631 linker Sets the name of the linker
3632 predefines Defines to be passed to the C preprocessor
3633 signed_char Defines to pass to CPP to say whether @code{char} is signed by default
3634 startfile Object files to include at the start of the link
3637 Here is a small example of a spec file:
3643 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
3646 This example renames the spec called @samp{lib} to @samp{old_lib} and
3647 then overrides the previous definition of @samp{lib} with a new one.
3648 The new definition adds in some extra command-line options before
3649 including the text of the old definition.
3651 @dfn{Spec strings} are a list of command-line options to be passed to their
3652 corresponding program. In addition, the spec strings can contain
3653 @samp{%}-prefixed sequences to substitute variable text or to
3654 conditionally insert text into the command line. Using these constructs
3655 it is possible to generate quite complex command lines.
3657 Here is a table of all defined @samp{%}-sequences for spec
3658 strings. Note that spaces are not generated automatically around the
3659 results of expanding these sequences. Therefore you can concatenate them
3660 together or combine them with constant text in a single argument.
3664 Substitute one @samp{%} into the program name or argument.
3667 Substitute the name of the input file being processed.
3670 Substitute the basename of the input file being processed.
3671 This is the substring up to (and not including) the last period
3672 and not including the directory.
3675 Marks the argument containing or following the @samp{%d} as a
3676 temporary file name, so that that file will be deleted if GCC exits
3677 successfully. Unlike @samp{%g}, this contributes no text to the
3680 @item %g@var{suffix}
3681 Substitute a file name that has suffix @var{suffix} and is chosen
3682 once per compilation, and mark the argument in the same way as
3683 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
3684 name is now chosen in a way that is hard to predict even when previously
3685 chosen file names are known. For example, @samp{%g.s ... %g.o ... %g.s}
3686 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
3687 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
3688 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
3689 was simply substituted with a file name chosen once per compilation,
3690 without regard to any appended suffix (which was therefore treated
3691 just like ordinary text), making such attacks more likely to succeed.
3693 @item %u@var{suffix}
3694 Like @samp{%g}, but generates a new temporary file name even if
3695 @samp{%u@var{suffix}} was already seen.
3697 @item %U@var{suffix}
3698 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
3699 new one if there is no such last file name. In the absence of any
3700 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
3701 the same suffix @emph{space}, so @samp{%g.s ... %U.s ... %g.s ... %U.s}
3702 would involve the generation of two distinct file names, one
3703 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
3704 simply substituted with a file name chosen for the previous @samp{%u},
3705 without regard to any appended suffix.
3708 Marks the argument containing or following the @samp{%w} as the
3709 designated output file of this compilation. This puts the argument
3710 into the sequence of arguments that @samp{%o} will substitute later.
3713 Substitutes the names of all the output files, with spaces
3714 automatically placed around them. You should write spaces
3715 around the @samp{%o} as well or the results are undefined.
3716 @samp{%o} is for use in the specs for running the linker.
3717 Input files whose names have no recognized suffix are not compiled
3718 at all, but they are included among the output files, so they will
3722 Substitutes the suffix for object files. Note that this is
3723 handled specially when it immediately follows @samp{%g, %u, or %U},
3724 because of the need for those to form complete file names. The
3725 handling is such that @samp{%O} is treated exactly as if it had already
3726 been substituted, except that @samp{%g, %u, and %U} do not currently
3727 support additional @var{suffix} characters following @samp{%O} as they would
3728 following, for example, @samp{.o}.
3731 Substitutes the standard macro predefinitions for the
3732 current target machine. Use this when running @code{cpp}.
3735 Like @samp{%p}, but puts @samp{__} before and after the name of each
3736 predefined macro, except for macros that start with @samp{__} or with
3737 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
3741 Substitute a @samp{-iprefix} option made from GCC_EXEC_PREFIX.
3744 Current argument is the name of a library or startup file of some sort.
3745 Search for that file in a standard list of directories and substitute
3746 the full name found.
3749 Print @var{str} as an error message. @var{str} is terminated by a newline.
3750 Use this when inconsistent options are detected.
3753 Output @samp{-} if the input for the current command is coming from a pipe.
3756 Substitute the contents of spec string @var{name} at this point.
3759 Like @samp{%(...)} but put @samp{__} around @samp{-D} arguments.
3761 @item %x@{@var{option}@}
3762 Accumulate an option for @samp{%X}.
3765 Output the accumulated linker options specified by @samp{-Wl} or a @samp{%x}
3769 Output the accumulated assembler options specified by @samp{-Wa}.
3772 Output the accumulated preprocessor options specified by @samp{-Wp}.
3775 Substitute the major version number of GCC.
3776 (For version 2.9.5, this is 2.)
3779 Substitute the minor version number of GCC.
3780 (For version 2.9.5, this is 9.)
3783 Process the @code{asm} spec. This is used to compute the
3784 switches to be passed to the assembler.
3787 Process the @code{asm_final} spec. This is a spec string for
3788 passing switches to an assembler post-processor, if such a program is
3792 Process the @code{link} spec. This is the spec for computing the
3793 command line passed to the linker. Typically it will make use of the
3794 @samp{%L %G %S %D and %E} sequences.
3797 Dump out a @samp{-L} option for each directory that GCC believes might
3798 contain startup files. If the target supports multilibs then the
3799 current multilib directory will be prepended to each of these paths.
3802 Process the @code{lib} spec. This is a spec string for deciding which
3803 libraries should be included on the command line to the linker.
3806 Process the @code{libgcc} spec. This is a spec string for deciding
3807 which GCC support library should be included on the command line to the linker.
3810 Process the @code{startfile} spec. This is a spec for deciding which
3811 object files should be the first ones passed to the linker. Typically
3812 this might be a file named @file{crt0.o}.
3815 Process the @code{endfile} spec. This is a spec string that specifies
3816 the last object files that will be passed to the linker.
3819 Process the @code{cpp} spec. This is used to construct the arguments
3820 to be passed to the C preprocessor.
3823 Process the @code{signed_char} spec. This is intended to be used
3824 to tell cpp whether a char is signed. It typically has the definition:
3826 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
3830 Process the @code{cc1} spec. This is used to construct the options to be
3831 passed to the actual C compiler (@samp{cc1}).
3834 Process the @code{cc1plus} spec. This is used to construct the options to be
3835 passed to the actual C++ compiler (@samp{cc1plus}).
3838 Substitute the variable part of a matched option. See below.
3839 Note that each comma in the substituted string is replaced by
3843 Substitutes the @code{-S} switch, if that switch was given to GCC.
3844 If that switch was not specified, this substitutes nothing. Note that
3845 the leading dash is omitted when specifying this option, and it is
3846 automatically inserted if the substitution is performed. Thus the spec
3847 string @samp{%@{foo@}} would match the command-line option @samp{-foo}
3848 and would output the command line option @samp{-foo}.
3850 @item %W@{@code{S}@}
3851 Like %@{@code{S}@} but mark last argument supplied within as a file to be
3854 @item %@{@code{S}*@}
3855 Substitutes all the switches specified to GCC whose names start
3856 with @code{-S}, but which also take an argument. This is used for
3857 switches like @samp{-o, -D, -I}, etc. GCC considers @samp{-o foo} as being
3858 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
3859 text, including the space. Thus two arguments would be generated.
3861 @item %@{^@code{S}*@}
3862 Like %@{@code{S}*@}, but don't put a blank between a switch and its
3863 argument. Thus %@{^o*@} would only generate one argument, not two.
3865 @item %@{<@code{S}@}
3866 Remove all occurrences of @code{-S} from the command line. Note - this
3867 command is position dependent. @samp{%} commands in the spec string
3868 before this option will see @code{-S}, @samp{%} commands in the spec
3869 string after this option will not.
3871 @item %@{@code{S}*:@code{X}@}
3872 Substitutes @code{X} if one or more switches whose names start with
3873 @code{-S} are specified to GCC. Note that the tail part of the
3874 @code{-S} option (i.e. the part matched by the @samp{*}) will be substituted
3875 for each occurrence of @samp{%*} within @code{X}.
3877 @item %@{@code{S}:@code{X}@}
3878 Substitutes @code{X}, but only if the @samp{-S} switch was given to GCC.
3880 @item %@{!@code{S}:@code{X}@}
3881 Substitutes @code{X}, but only if the @samp{-S} switch was @emph{not} given to GCC.
3883 @item %@{|@code{S}:@code{X}@}
3884 Like %@{@code{S}:@code{X}@}, but if no @code{S} switch, substitute @samp{-}.
3886 @item %@{|!@code{S}:@code{X}@}
3887 Like %@{!@code{S}:@code{X}@}, but if there is an @code{S} switch, substitute @samp{-}.
3889 @item %@{.@code{S}:@code{X}@}
3890 Substitutes @code{X}, but only if processing a file with suffix @code{S}.
3892 @item %@{!.@code{S}:@code{X}@}
3893 Substitutes @code{X}, but only if @emph{not} processing a file with suffix @code{S}.
3895 @item %@{@code{S}|@code{P}:@code{X}@}
3896 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC. This may be
3897 combined with @samp{!} and @samp{.} sequences as well, although they
3898 have a stronger binding than the @samp{|}. For example a spec string
3902 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
3905 will output the following command-line options from the following input
3906 command-line options:
3911 -d fred.c -foo -baz -boggle
3912 -d jim.d -bar -baz -boggle
3917 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or
3918 %@{!@code{S}:@code{X}@} construct may contain other nested @samp{%} constructs
3919 or spaces, or even newlines. They are processed as usual, as described
3922 The @samp{-O, -f, -m, and -W} switches are handled specifically in these
3923 constructs. If another value of @samp{-O} or the negated form of a @samp{-f, -m, or
3924 -W} switch is found later in the command line, the earlier switch
3925 value is ignored, except with @{@code{S}*@} where @code{S} is just one
3926 letter, which passes all matching options.
3928 The character @samp{|} at the beginning of the predicate text is used to indicate
3929 that a command should be piped to the following command, but only if @samp{-pipe}
3932 It is built into GCC which switches take arguments and which do not.
3933 (You might think it would be useful to generalize this to allow each
3934 compiler's spec to say which switches take arguments. But this cannot
3935 be done in a consistent fashion. GCC cannot even decide which input
3936 files have been specified without knowing which switches take arguments,
3937 and it must know which input files to compile in order to tell which
3940 GCC also knows implicitly that arguments starting in @samp{-l} are to be
3941 treated as compiler output files, and passed to the linker in their
3942 proper position among the other output files.
3944 @node Target Options
3945 @section Specifying Target Machine and Compiler Version
3946 @cindex target options
3947 @cindex cross compiling
3948 @cindex specifying machine version
3949 @cindex specifying compiler version and target machine
3950 @cindex compiler version, specifying
3951 @cindex target machine, specifying
3953 By default, GCC compiles code for the same type of machine that you
3954 are using. However, it can also be installed as a cross-compiler, to
3955 compile for some other type of machine. In fact, several different
3956 configurations of GCC, for different target machines, can be
3957 installed side by side. Then you specify which one to use with the
3960 In addition, older and newer versions of GCC can be installed side
3961 by side. One of them (probably the newest) will be the default, but
3962 you may sometimes wish to use another.
3965 @item -b @var{machine}
3966 The argument @var{machine} specifies the target machine for compilation.
3967 This is useful when you have installed GCC as a cross-compiler.
3969 The value to use for @var{machine} is the same as was specified as the
3970 machine type when configuring GCC as a cross-compiler. For
3971 example, if a cross-compiler was configured with @samp{configure
3972 i386v}, meaning to compile for an 80386 running System V, then you
3973 would specify @samp{-b i386v} to run that cross compiler.
3975 When you do not specify @samp{-b}, it normally means to compile for
3976 the same type of machine that you are using.
3978 @item -V @var{version}
3979 The argument @var{version} specifies which version of GCC to run.
3980 This is useful when multiple versions are installed. For example,
3981 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
3983 The default version, when you do not specify @samp{-V}, is the last
3984 version of GCC that you installed.
3987 The @samp{-b} and @samp{-V} options actually work by controlling part of
3988 the file name used for the executable files and libraries used for
3989 compilation. A given version of GCC, for a given target machine, is
3990 normally kept in the directory @file{/usr/local/lib/gcc-lib/@var{machine}/@var{version}}.@refill
3992 Thus, sites can customize the effect of @samp{-b} or @samp{-V} either by
3993 changing the names of these directories or adding alternate names (or
3994 symbolic links). If in directory @file{/usr/local/lib/gcc-lib/} the
3995 file @file{80386} is a link to the file @file{i386v}, then @samp{-b
3996 80386} becomes an alias for @samp{-b i386v}.
3998 In one respect, the @samp{-b} or @samp{-V} do not completely change
3999 to a different compiler: the top-level driver program @code{gcc}
4000 that you originally invoked continues to run and invoke the other
4001 executables (preprocessor, compiler per se, assembler and linker)
4002 that do the real work. However, since no real work is done in the
4003 driver program, it usually does not matter that the driver program
4004 in use is not the one for the specified target. It is common for the
4005 interface to the other executables to change incompatibly between
4006 compiler versions, so unless the version specified is very close to that
4007 of the driver (for example, @samp{-V 3.0} with a driver program from GCC
4008 version 3.0.1), use of @samp{-V} may not work; for example, using
4009 @samp{-V 2.95.2} will not work with a driver program from GCC 3.0.
4011 The only way that the driver program depends on the target machine is
4012 in the parsing and handling of special machine-specific options.
4013 However, this is controlled by a file which is found, along with the
4014 other executables, in the directory for the specified version and
4015 target machine. As a result, a single installed driver program adapts
4016 to any specified target machine, and sufficiently similar compiler
4019 The driver program executable does control one significant thing,
4020 however: the default version and target machine. Therefore, you can
4021 install different instances of the driver program, compiled for
4022 different targets or versions, under different names.
4024 For example, if the driver for version 2.0 is installed as @code{ogcc}
4025 and that for version 2.1 is installed as @code{gcc}, then the command
4026 @code{gcc} will use version 2.1 by default, while @code{ogcc} will use
4027 2.0 by default. However, you can choose either version with either
4028 command with the @samp{-V} option.
4030 @node Submodel Options
4031 @section Hardware Models and Configurations
4032 @cindex submodel options
4033 @cindex specifying hardware config
4034 @cindex hardware models and configurations, specifying
4035 @cindex machine dependent options
4037 Earlier we discussed the standard option @samp{-b} which chooses among
4038 different installed compilers for completely different target
4039 machines, such as Vax vs. 68000 vs. 80386.
4041 In addition, each of these target machine types can have its own
4042 special options, starting with @samp{-m}, to choose among various
4043 hardware models or configurations---for example, 68010 vs 68020,
4044 floating coprocessor or none. A single installed version of the
4045 compiler can compile for any model or configuration, according to the
4048 Some configurations of the compiler also support additional special
4049 options, usually for compatibility with other compilers on the same
4053 These options are defined by the macro @code{TARGET_SWITCHES} in the
4054 machine description. The default for the options is also defined by
4055 that macro, which enables you to change the defaults.
4071 * RS/6000 and PowerPC Options::
4076 * Intel 960 Options::
4077 * DEC Alpha Options::
4081 * System V Options::
4082 * TMS320C3x/C4x Options::
4091 @node M680x0 Options
4092 @subsection M680x0 Options
4093 @cindex M680x0 options
4095 These are the @samp{-m} options defined for the 68000 series. The default
4096 values for these options depends on which style of 68000 was selected when
4097 the compiler was configured; the defaults for the most common choices are
4103 Generate output for a 68000. This is the default
4104 when the compiler is configured for 68000-based systems.
4106 Use this option for microcontrollers with a 68000 or EC000 core,
4107 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
4111 Generate output for a 68020. This is the default
4112 when the compiler is configured for 68020-based systems.
4115 Generate output containing 68881 instructions for floating point.
4116 This is the default for most 68020 systems unless @samp{-nfp} was
4117 specified when the compiler was configured.
4120 Generate output for a 68030. This is the default when the compiler is
4121 configured for 68030-based systems.
4124 Generate output for a 68040. This is the default when the compiler is
4125 configured for 68040-based systems.
4127 This option inhibits the use of 68881/68882 instructions that have to be
4128 emulated by software on the 68040. Use this option if your 68040 does not
4129 have code to emulate those instructions.
4132 Generate output for a 68060. This is the default when the compiler is
4133 configured for 68060-based systems.
4135 This option inhibits the use of 68020 and 68881/68882 instructions that
4136 have to be emulated by software on the 68060. Use this option if your 68060
4137 does not have code to emulate those instructions.
4140 Generate output for a CPU32. This is the default
4141 when the compiler is configured for CPU32-based systems.
4143 Use this option for microcontrollers with a
4144 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
4145 68336, 68340, 68341, 68349 and 68360.
4148 Generate output for a 520X "coldfire" family cpu. This is the default
4149 when the compiler is configured for 520X-based systems.
4151 Use this option for microcontroller with a 5200 core, including
4152 the MCF5202, MCF5203, MCF5204 and MCF5202.
4156 Generate output for a 68040, without using any of the new instructions.
4157 This results in code which can run relatively efficiently on either a
4158 68020/68881 or a 68030 or a 68040. The generated code does use the
4159 68881 instructions that are emulated on the 68040.
4162 Generate output for a 68060, without using any of the new instructions.
4163 This results in code which can run relatively efficiently on either a
4164 68020/68881 or a 68030 or a 68040. The generated code does use the
4165 68881 instructions that are emulated on the 68060.
4168 Generate output containing Sun FPA instructions for floating point.
4171 Generate output containing library calls for floating point.
4172 @strong{Warning:} the requisite libraries are not available for all m68k
4173 targets. Normally the facilities of the machine's usual C compiler are
4174 used, but this can't be done directly in cross-compilation. You must
4175 make your own arrangements to provide suitable library functions for
4176 cross-compilation. The embedded targets @samp{m68k-*-aout} and
4177 @samp{m68k-*-coff} do provide software floating point support.
4180 Consider type @code{int} to be 16 bits wide, like @code{short int}.
4183 Do not use the bit-field instructions. The @samp{-m68000}, @samp{-mcpu32}
4184 and @samp{-m5200} options imply @w{@samp{-mnobitfield}}.
4187 Do use the bit-field instructions. The @samp{-m68020} option implies
4188 @samp{-mbitfield}. This is the default if you use a configuration
4189 designed for a 68020.
4192 Use a different function-calling convention, in which functions
4193 that take a fixed number of arguments return with the @code{rtd}
4194 instruction, which pops their arguments while returning. This
4195 saves one instruction in the caller since there is no need to pop
4196 the arguments there.
4198 This calling convention is incompatible with the one normally
4199 used on Unix, so you cannot use it if you need to call libraries
4200 compiled with the Unix compiler.
4202 Also, you must provide function prototypes for all functions that
4203 take variable numbers of arguments (including @code{printf});
4204 otherwise incorrect code will be generated for calls to those
4207 In addition, seriously incorrect code will result if you call a
4208 function with too many arguments. (Normally, extra arguments are
4209 harmlessly ignored.)
4211 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
4212 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
4215 @itemx -mno-align-int
4216 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
4217 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
4218 boundary (@samp{-malign-int}) or a 16-bit boundary (@samp{-mno-align-int}).
4219 Aligning variables on 32-bit boundaries produces code that runs somewhat
4220 faster on processors with 32-bit busses at the expense of more memory.
4222 @strong{Warning:} if you use the @samp{-malign-int} switch, GCC will
4223 align structures containing the above types differently than
4224 most published application binary interface specifications for the m68k.
4227 Use the pc-relative addressing mode of the 68000 directly, instead of
4228 using a global offset table. At present, this option implies -fpic,
4229 allowing at most a 16-bit offset for pc-relative addressing. -fPIC is
4230 not presently supported with -mpcrel, though this could be supported for
4231 68020 and higher processors.
4233 @item -mno-strict-align
4234 @itemx -mstrict-align
4235 @kindex -mstrict-align
4236 Do not (do) assume that unaligned memory references will be handled by
4241 @node M68hc1x Options
4242 @subsection M68hc1x Options
4243 @cindex M68hc1x options
4245 These are the @samp{-m} options defined for the 68hc11 and 68hc12
4246 microcontrollers. The default values for these options depends on
4247 which style of microcontroller was selected when the compiler was configured;
4248 the defaults for the most common choices are given below.
4253 Generate output for a 68HC11. This is the default
4254 when the compiler is configured for 68HC11-based systems.
4258 Generate output for a 68HC12. This is the default
4259 when the compiler is configured for 68HC12-based systems.
4262 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
4266 Consider type @code{int} to be 16 bits wide, like @code{short int}.
4268 @item -msoft-reg-count=@var{count}
4269 Specify the number of pseudo-soft registers which are used for the
4270 code generation. The maximum number is 32. Using more pseudo-soft
4271 register may or may not result in better code depending on the program.
4272 The default is 4 for 68HC11 and 2 for 68HC12.
4277 @subsection VAX Options
4280 These @samp{-m} options are defined for the Vax:
4284 Do not output certain jump instructions (@code{aobleq} and so on)
4285 that the Unix assembler for the Vax cannot handle across long
4289 Do output those jump instructions, on the assumption that you
4290 will assemble with the GNU assembler.
4293 Output code for g-format floating point numbers instead of d-format.
4297 @subsection SPARC Options
4298 @cindex SPARC options
4300 These @samp{-m} switches are supported on the SPARC:
4305 Specify @samp{-mapp-regs} to generate output using the global registers
4306 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
4309 To be fully SVR4 ABI compliant at the cost of some performance loss,
4310 specify @samp{-mno-app-regs}. You should compile libraries and system
4311 software with this option.
4315 Generate output containing floating point instructions. This is the
4320 Generate output containing library calls for floating point.
4321 @strong{Warning:} the requisite libraries are not available for all SPARC
4322 targets. Normally the facilities of the machine's usual C compiler are
4323 used, but this cannot be done directly in cross-compilation. You must make
4324 your own arrangements to provide suitable library functions for
4325 cross-compilation. The embedded targets @samp{sparc-*-aout} and
4326 @samp{sparclite-*-*} do provide software floating point support.
4328 @samp{-msoft-float} changes the calling convention in the output file;
4329 therefore, it is only useful if you compile @emph{all} of a program with
4330 this option. In particular, you need to compile @file{libgcc.a}, the
4331 library that comes with GCC, with @samp{-msoft-float} in order for
4334 @item -mhard-quad-float
4335 Generate output containing quad-word (long double) floating point
4338 @item -msoft-quad-float
4339 Generate output containing library calls for quad-word (long double)
4340 floating point instructions. The functions called are those specified
4341 in the SPARC ABI. This is the default.
4343 As of this writing, there are no sparc implementations that have hardware
4344 support for the quad-word floating point instructions. They all invoke
4345 a trap handler for one of these instructions, and then the trap handler
4346 emulates the effect of the instruction. Because of the trap handler overhead,
4347 this is much slower than calling the ABI library routines. Thus the
4348 @samp{-msoft-quad-float} option is the default.
4352 With @samp{-mepilogue} (the default), the compiler always emits code for
4353 function exit at the end of each function. Any function exit in
4354 the middle of the function (such as a return statement in C) will
4355 generate a jump to the exit code at the end of the function.
4357 With @samp{-mno-epilogue}, the compiler tries to emit exit code inline
4358 at every function exit.
4362 With @samp{-mflat}, the compiler does not generate save/restore instructions
4363 and will use a "flat" or single register window calling convention.
4364 This model uses %i7 as the frame pointer and is compatible with the normal
4365 register window model. Code from either may be intermixed.
4366 The local registers and the input registers (0-5) are still treated as
4367 "call saved" registers and will be saved on the stack as necessary.
4369 With @samp{-mno-flat} (the default), the compiler emits save/restore
4370 instructions (except for leaf functions) and is the normal mode of operation.
4372 @item -mno-unaligned-doubles
4373 @itemx -munaligned-doubles
4374 Assume that doubles have 8 byte alignment. This is the default.
4376 With @samp{-munaligned-doubles}, GCC assumes that doubles have 8 byte
4377 alignment only if they are contained in another type, or if they have an
4378 absolute address. Otherwise, it assumes they have 4 byte alignment.
4379 Specifying this option avoids some rare compatibility problems with code
4380 generated by other compilers. It is not the default because it results
4381 in a performance loss, especially for floating point code.
4383 @item -mno-faster-structs
4384 @itemx -mfaster-structs
4385 With @samp{-mfaster-structs}, the compiler assumes that structures
4386 should have 8 byte alignment. This enables the use of pairs of
4387 @code{ldd} and @code{std} instructions for copies in structure
4388 assignment, in place of twice as many @code{ld} and @code{st} pairs.
4389 However, the use of this changed alignment directly violates the Sparc
4390 ABI. Thus, it's intended only for use on targets where the developer
4391 acknowledges that their resulting code will not be directly in line with
4392 the rules of the ABI.
4396 These two options select variations on the SPARC architecture.
4398 By default (unless specifically configured for the Fujitsu SPARClite),
4399 GCC generates code for the v7 variant of the SPARC architecture.
4401 @samp{-mv8} will give you SPARC v8 code. The only difference from v7
4402 code is that the compiler emits the integer multiply and integer
4403 divide instructions which exist in SPARC v8 but not in SPARC v7.
4405 @samp{-msparclite} will give you SPARClite code. This adds the integer
4406 multiply, integer divide step and scan (@code{ffs}) instructions which
4407 exist in SPARClite but not in SPARC v7.
4409 These options are deprecated and will be deleted in a future GCC release.
4410 They have been replaced with @samp{-mcpu=xxx}.
4414 These two options select the processor for which the code is optimised.
4416 With @samp{-mcypress} (the default), the compiler optimizes code for the
4417 Cypress CY7C602 chip, as used in the SparcStation/SparcServer 3xx series.
4418 This is also appropriate for the older SparcStation 1, 2, IPX etc.
4420 With @samp{-msupersparc} the compiler optimizes code for the SuperSparc cpu, as
4421 used in the SparcStation 10, 1000 and 2000 series. This flag also enables use
4422 of the full SPARC v8 instruction set.
4424 These options are deprecated and will be deleted in a future GCC release.
4425 They have been replaced with @samp{-mcpu=xxx}.
4427 @item -mcpu=@var{cpu_type}
4428 Set the instruction set, register set, and instruction scheduling parameters
4429 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
4430 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
4431 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
4432 @samp{sparclet}, @samp{tsc701}, @samp{v9}, and @samp{ultrasparc}.
4434 Default instruction scheduling parameters are used for values that select
4435 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
4436 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
4438 Here is a list of each supported architecture and their supported
4443 v8: supersparc, hypersparc
4444 sparclite: f930, f934, sparclite86x
4449 @item -mtune=@var{cpu_type}
4450 Set the instruction scheduling parameters for machine type
4451 @var{cpu_type}, but do not set the instruction set or register set that the
4452 option @samp{-mcpu=}@var{cpu_type} would.
4454 The same values for @samp{-mcpu=}@var{cpu_type} are used for
4455 @samp{-mtune=}@*@var{cpu_type}, though the only useful values are those that
4456 select a particular cpu implementation: @samp{cypress}, @samp{supersparc},
4457 @samp{hypersparc}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
4458 @samp{tsc701}, @samp{ultrasparc}.
4462 These @samp{-m} switches are supported in addition to the above
4463 on the SPARCLET processor.
4466 @item -mlittle-endian
4467 Generate code for a processor running in little-endian mode.
4470 Treat register @code{%g0} as a normal register.
4471 GCC will continue to clobber it as necessary but will not assume
4472 it always reads as 0.
4474 @item -mbroken-saverestore
4475 Generate code that does not use non-trivial forms of the @code{save} and
4476 @code{restore} instructions. Early versions of the SPARCLET processor do
4477 not correctly handle @code{save} and @code{restore} instructions used with
4478 arguments. They correctly handle them used without arguments. A @code{save}
4479 instruction used without arguments increments the current window pointer
4480 but does not allocate a new stack frame. It is assumed that the window
4481 overflow trap handler will properly handle this case as will interrupt
4485 These @samp{-m} switches are supported in addition to the above
4486 on SPARC V9 processors in 64 bit environments.
4489 @item -mlittle-endian
4490 Generate code for a processor running in little-endian mode.
4494 Generate code for a 32 bit or 64 bit environment.
4495 The 32 bit environment sets int, long and pointer to 32 bits.
4496 The 64 bit environment sets int to 32 bits and long and pointer
4499 @item -mcmodel=medlow
4500 Generate code for the Medium/Low code model: the program must be linked
4501 in the low 32 bits of the address space. Pointers are 64 bits.
4502 Programs can be statically or dynamically linked.
4504 @item -mcmodel=medmid
4505 Generate code for the Medium/Middle code model: the program must be linked
4506 in the low 44 bits of the address space, the text segment must be less than
4507 2G bytes, and data segment must be within 2G of the text segment.
4508 Pointers are 64 bits.
4510 @item -mcmodel=medany
4511 Generate code for the Medium/Anywhere code model: the program may be linked
4512 anywhere in the address space, the text segment must be less than
4513 2G bytes, and data segment must be within 2G of the text segment.
4514 Pointers are 64 bits.
4516 @item -mcmodel=embmedany
4517 Generate code for the Medium/Anywhere code model for embedded systems:
4518 assume a 32 bit text and a 32 bit data segment, both starting anywhere
4519 (determined at link time). Register %g4 points to the base of the
4520 data segment. Pointers still 64 bits.
4521 Programs are statically linked, PIC is not supported.
4524 @itemx -mno-stack-bias
4525 With @samp{-mstack-bias}, GCC assumes that the stack pointer, and
4526 frame pointer if present, are offset by -2047 which must be added back
4527 when making stack frame references.
4528 Otherwise, assume no such offset is present.
4531 @node Convex Options
4532 @subsection Convex Options
4533 @cindex Convex options
4535 These @samp{-m} options are defined for Convex:
4539 Generate output for C1. The code will run on any Convex machine.
4540 The preprocessor symbol @code{__convex__c1__} is defined.
4543 Generate output for C2. Uses instructions not available on C1.
4544 Scheduling and other optimizations are chosen for max performance on C2.
4545 The preprocessor symbol @code{__convex_c2__} is defined.
4548 Generate output for C32xx. Uses instructions not available on C1.
4549 Scheduling and other optimizations are chosen for max performance on C32.
4550 The preprocessor symbol @code{__convex_c32__} is defined.
4553 Generate output for C34xx. Uses instructions not available on C1.
4554 Scheduling and other optimizations are chosen for max performance on C34.
4555 The preprocessor symbol @code{__convex_c34__} is defined.
4558 Generate output for C38xx. Uses instructions not available on C1.
4559 Scheduling and other optimizations are chosen for max performance on C38.
4560 The preprocessor symbol @code{__convex_c38__} is defined.
4563 Generate code which puts an argument count in the word preceding each
4564 argument list. This is compatible with regular CC, and a few programs
4565 may need the argument count word. GDB and other source-level debuggers
4566 do not need it; this info is in the symbol table.
4569 Omit the argument count word. This is the default.
4571 @item -mvolatile-cache
4572 Allow volatile references to be cached. This is the default.
4574 @item -mvolatile-nocache
4575 Volatile references bypass the data cache, going all the way to memory.
4576 This is only needed for multi-processor code that does not use standard
4577 synchronization instructions. Making non-volatile references to volatile
4578 locations will not necessarily work.
4581 Type long is 32 bits, the same as type int. This is the default.
4584 Type long is 64 bits, the same as type long long. This option is useless,
4585 because no library support exists for it.
4588 @node AMD29K Options
4589 @subsection AMD29K Options
4590 @cindex AMD29K options
4592 These @samp{-m} options are defined for the AMD Am29000:
4597 @cindex DW bit (29k)
4598 Generate code that assumes the @code{DW} bit is set, i.e., that byte and
4599 halfword operations are directly supported by the hardware. This is the
4604 Generate code that assumes the @code{DW} bit is not set.
4608 @cindex byte writes (29k)
4609 Generate code that assumes the system supports byte and halfword write
4610 operations. This is the default.
4614 Generate code that assumes the systems does not support byte and
4615 halfword write operations. @samp{-mnbw} implies @samp{-mndw}.
4619 @cindex memory model (29k)
4620 Use a small memory model that assumes that all function addresses are
4621 either within a single 256 KB segment or at an absolute address of less
4622 than 256k. This allows the @code{call} instruction to be used instead
4623 of a @code{const}, @code{consth}, @code{calli} sequence.
4627 Use the normal memory model: Generate @code{call} instructions only when
4628 calling functions in the same file and @code{calli} instructions
4629 otherwise. This works if each file occupies less than 256 KB but allows
4630 the entire executable to be larger than 256 KB. This is the default.
4633 Always use @code{calli} instructions. Specify this option if you expect
4634 a single file to compile into more than 256 KB of code.
4638 @cindex processor selection (29k)
4639 Generate code for the Am29050.
4643 Generate code for the Am29000. This is the default.
4645 @item -mkernel-registers
4646 @kindex -mkernel-registers
4647 @cindex kernel and user registers (29k)
4648 Generate references to registers @code{gr64-gr95} instead of to
4649 registers @code{gr96-gr127}. This option can be used when compiling
4650 kernel code that wants a set of global registers disjoint from that used
4653 Note that when this option is used, register names in @samp{-f} flags
4654 must use the normal, user-mode, names.
4656 @item -muser-registers
4657 @kindex -muser-registers
4658 Use the normal set of global registers, @code{gr96-gr127}. This is the
4662 @itemx -mno-stack-check
4663 @kindex -mstack-check
4664 @cindex stack checks (29k)
4665 Insert (or do not insert) a call to @code{__msp_check} after each stack
4666 adjustment. This is often used for kernel code.
4669 @itemx -mno-storem-bug
4670 @kindex -mstorem-bug
4671 @cindex storem bug (29k)
4672 @samp{-mstorem-bug} handles 29k processors which cannot handle the
4673 separation of a mtsrim insn and a storem instruction (most 29000 chips
4674 to date, but not the 29050).
4676 @item -mno-reuse-arg-regs
4677 @itemx -mreuse-arg-regs
4678 @kindex -mreuse-arg-regs
4679 @samp{-mno-reuse-arg-regs} tells the compiler to only use incoming argument
4680 registers for copying out arguments. This helps detect calling a function
4681 with fewer arguments than it was declared with.
4683 @item -mno-impure-text
4684 @itemx -mimpure-text
4685 @kindex -mimpure-text
4686 @samp{-mimpure-text}, used in addition to @samp{-shared}, tells the compiler to
4687 not pass @samp{-assert pure-text} to the linker when linking a shared object.
4690 @kindex -msoft-float
4691 Generate output containing library calls for floating point.
4692 @strong{Warning:} the requisite libraries are not part of GCC.
4693 Normally the facilities of the machine's usual C compiler are used, but
4694 this can't be done directly in cross-compilation. You must make your
4695 own arrangements to provide suitable library functions for
4700 Do not generate multm or multmu instructions. This is useful for some embedded
4701 systems which do not have trap handlers for these instructions.
4705 @subsection ARM Options
4708 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
4713 @kindex -mapcs-frame
4714 Generate a stack frame that is compliant with the ARM Procedure Call
4715 Standard for all functions, even if this is not strictly necessary for
4716 correct execution of the code. Specifying @samp{-fomit-frame-pointer}
4717 with this option will cause the stack frames not to be generated for
4718 leaf functions. The default is @samp{-mno-apcs-frame}.
4722 This is a synonym for @samp{-mapcs-frame}.
4726 Generate code for a processor running with a 26-bit program counter,
4727 and conforming to the function calling standards for the APCS 26-bit
4728 option. This option replaces the @samp{-m2} and @samp{-m3} options
4729 of previous releases of the compiler.
4733 Generate code for a processor running with a 32-bit program counter,
4734 and conforming to the function calling standards for the APCS 32-bit
4735 option. This option replaces the @samp{-m6} option of previous releases
4738 @item -mapcs-stack-check
4739 @kindex -mapcs-stack-check
4740 @kindex -mno-apcs-stack-check
4741 Generate code to check the amount of stack space available upon entry to
4742 every function (that actually uses some stack space). If there is
4743 insufficient space available then either the function
4744 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
4745 called, depending upon the amount of stack space required. The run time
4746 system is required to provide these functions. The default is
4747 @samp{-mno-apcs-stack-check}, since this produces smaller code.
4750 @kindex -mapcs-float
4751 @kindex -mno-apcs-float
4752 Pass floating point arguments using the float point registers. This is
4753 one of the variants of the APCS. This option is recommended if the
4754 target hardware has a floating point unit or if a lot of floating point
4755 arithmetic is going to be performed by the code. The default is
4756 @samp{-mno-apcs-float}, since integer only code is slightly increased in
4757 size if @samp{-mapcs-float} is used.
4759 @item -mapcs-reentrant
4760 @kindex -mapcs-reentrant
4761 @kindex -mno-apcs-reentrant
4762 Generate reentrant, position independent code. This is the equivalent
4763 to specifying the @samp{-fpic} option. The default is
4764 @samp{-mno-apcs-reentrant}.
4766 @item -mthumb-interwork
4767 @kindex -mthumb-interwork
4768 @kindex -mno-thumb-interwork
4769 Generate code which supports calling between the ARM and THUMB
4770 instruction sets. Without this option the two instruction sets cannot
4771 be reliably used inside one program. The default is
4772 @samp{-mno-thumb-interwork}, since slightly larger code is generated
4773 when @samp{-mthumb-interwork} is specified.
4775 @item -mno-sched-prolog
4776 @kindex -mno-sched-prolog
4777 @kindex -msched-prolog
4778 Prevent the reordering of instructions in the function prolog, or the
4779 merging of those instruction with the instructions in the function's
4780 body. This means that all functions will start with a recognizable set
4781 of instructions (or in fact one of a choice from a small set of
4782 different function prologues), and this information can be used to
4783 locate the start if functions inside an executable piece of code. The
4784 default is @samp{-msched-prolog}.
4787 Generate output containing floating point instructions. This is the
4791 Generate output containing library calls for floating point.
4792 @strong{Warning:} the requisite libraries are not available for all ARM
4793 targets. Normally the facilities of the machine's usual C compiler are
4794 used, but this cannot be done directly in cross-compilation. You must make
4795 your own arrangements to provide suitable library functions for
4798 @samp{-msoft-float} changes the calling convention in the output file;
4799 therefore, it is only useful if you compile @emph{all} of a program with
4800 this option. In particular, you need to compile @file{libgcc.a}, the
4801 library that comes with GCC, with @samp{-msoft-float} in order for
4804 @item -mlittle-endian
4805 Generate code for a processor running in little-endian mode. This is
4806 the default for all standard configurations.
4809 Generate code for a processor running in big-endian mode; the default is
4810 to compile code for a little-endian processor.
4812 @item -mwords-little-endian
4813 This option only applies when generating code for big-endian processors.
4814 Generate code for a little-endian word order but a big-endian byte
4815 order. That is, a byte order of the form @samp{32107654}. Note: this
4816 option should only be used if you require compatibility with code for
4817 big-endian ARM processors generated by versions of the compiler prior to
4820 @item -malignment-traps
4821 @kindex -malignment-traps
4822 Generate code that will not trap if the MMU has alignment traps enabled.
4823 On ARM architectures prior to ARMv4, there were no instructions to
4824 access half-word objects stored in memory. However, when reading from
4825 memory a feature of the ARM architecture allows a word load to be used,
4826 even if the address is unaligned, and the processor core will rotate the
4827 data as it is being loaded. This option tells the compiler that such
4828 misaligned accesses will cause a MMU trap and that it should instead
4829 synthesise the access as a series of byte accesses. The compiler can
4830 still use word accesses to load half-word data if it knows that the
4831 address is aligned to a word boundary.
4833 This option is ignored when compiling for ARM architecture 4 or later,
4834 since these processors have instructions to directly access half-word
4837 @item -mno-alignment-traps
4838 @kindex -mno-alignment-traps
4839 Generate code that assumes that the MMU will not trap unaligned
4840 accesses. This produces better code when the target instruction set
4841 does not have half-word memory operations (implementations prior to
4844 Note that you cannot use this option to access unaligned word objects,
4845 since the processor will only fetch one 32-bit aligned object from
4848 The default setting for most targets is -mno-alignment-traps, since
4849 this produces better code when there are no half-word memory
4850 instructions available.
4852 @item -mshort-load-bytes
4853 @kindex -mshort-load-bytes
4854 This is a depreciated alias for @samp{-malignment-traps}.
4856 @item -mno-short-load-bytes
4857 @kindex -mno-short-load-bytes
4858 This is a depreciated alias for @samp{-mno-alignment-traps}.
4860 @item -mshort-load-words
4861 @kindex -mshort-load-words
4862 This is a depreciated alias for @samp{-mno-alignment-traps}.
4864 @item -mno-short-load-words
4865 @kindex -mno-short-load-words
4866 This is a depreciated alias for @samp{-malignment-traps}.
4870 This option only applies to RISC iX. Emulate the native BSD-mode
4871 compiler. This is the default if @samp{-ansi} is not specified.
4875 This option only applies to RISC iX. Emulate the native X/Open-mode
4878 @item -mno-symrename
4879 @kindex -mno-symrename
4880 This option only applies to RISC iX. Do not run the assembler
4881 post-processor, @samp{symrename}, after code has been assembled.
4882 Normally it is necessary to modify some of the standard symbols in
4883 preparation for linking with the RISC iX C library; this option
4884 suppresses this pass. The post-processor is never run when the
4885 compiler is built for cross-compilation.
4889 This specifies the name of the target ARM processor. GCC uses this name
4890 to determine what kind of instructions it can use when generating
4891 assembly code. Permissible names are: arm2, arm250, arm3, arm6, arm60,
4892 arm600, arm610, arm620, arm7, arm7m, arm7d, arm7dm, arm7di, arm7dmi,
4893 arm70, arm700, arm700i, arm710, arm710c, arm7100, arm7500, arm7500fe,
4894 arm7tdmi, arm8, strongarm, strongarm110, strongarm1100, arm8, arm810,
4895 arm9, arm920, arm920t, arm9tdmi.
4897 @itemx -mtune=<name>
4899 This option is very similar to the @samp{-mcpu=} option, except that
4900 instead of specifying the actual target processor type, and hence
4901 restricting which instructions can be used, it specifies that GCC should
4902 tune the performance of the code as if the target were of the type
4903 specified in this option, but still choosing the instructions that it
4904 will generate based on the cpu specified by a @samp{-mcpu=} option.
4905 For some arm implementations better performance can be obtained by using
4910 This specifies the name of the target ARM architecture. GCC uses this
4911 name to determine what kind of instructions it can use when generating
4912 assembly code. This option can be used in conjunction with or instead
4913 of the @samp{-mcpu=} option. Permissible names are: armv2, armv2a,
4914 armv3, armv3m, armv4, armv4t, armv5.
4916 @item -mfpe=<number>
4917 @itemx -mfp=<number>
4920 This specifies the version of the floating point emulation available on
4921 the target. Permissible values are 2 and 3. @samp{-mfp=} is a synonym
4922 for @samp{-mfpe=} to support older versions of GCC.
4924 @item -mstructure-size-boundary=<n>
4925 @kindex -mstructure-size-boundary
4926 The size of all structures and unions will be rounded up to a multiple
4927 of the number of bits set by this option. Permissible values are 8 and
4928 32. The default value varies for different toolchains. For the COFF
4929 targeted toolchain the default value is 8. Specifying the larger number
4930 can produce faster, more efficient code, but can also increase the size
4931 of the program. The two values are potentially incompatible. Code
4932 compiled with one value cannot necessarily expect to work with code or
4933 libraries compiled with the other value, if they exchange information
4934 using structures or unions. Programmers are encouraged to use the 32
4935 value as future versions of the toolchain may default to this value.
4937 @item -mabort-on-noreturn
4938 @kindex -mabort-on-noreturn
4939 @kindex -mnoabort-on-noreturn
4940 Generate a call to the function abort at the end of a noreturn function.
4941 It will be executed if the function tries to return.
4944 @itemx -mno-long-calls
4945 Tells the compiler to perform function calls by first loading the
4946 address of the function into a register and then performing a subroutine
4947 call on this register. This switch is needed if the target function
4948 will lie outside of the 64 megabyte addressing range of the offset based
4949 version of subroutine call instruction.
4951 Even if this switch is enabled, not all function calls will be turned
4952 into long calls. The heuristic is that static functions, functions
4953 which have the @samp{short-call} attribute, functions that are inside
4954 the scope of a @samp{#pragma no_long_calls} directive and functions whose
4955 definitions have already been compiled within the current compilation
4956 unit, will not be turned into long calls. The exception to this rule is
4957 that weak function definitions, functions with the @samp{long-call}
4958 attribute or the @samp{section} attribute, and functions that are within
4959 the scope of a @samp{#pragma long_calls} directive, will always be
4960 turned into long calls.
4962 This feature is not enabled by default. Specifying
4963 @samp{--no-long-calls} will restore the default behaviour, as will
4964 placing the function calls within the scope of a @samp{#pragma
4965 long_calls_off} directive. Note these switches have no effect on how
4966 the compiler generates code to handle function calls via function
4969 @item -mnop-fun-dllimport
4970 @kindex -mnop-fun-dllimport
4971 Disable the support for the @emph{dllimport} attribute.
4973 @item -msingle-pic-base
4974 @kindex -msingle-pic-base
4975 Treat the register used for PIC addressing as read-only, rather than
4976 loading it in the prologue for each function. The run-time system is
4977 responsible for initialising this register with an appropriate value
4978 before execution begins.
4980 @item -mpic-register=<reg>
4981 @kindex -mpic-register=
4982 Specify the register to be used for PIC addressing. The default is R10
4983 unless stack-checking is enabled, when R9 is used.
4988 @subsection Thumb Options
4989 @cindex Thumb Options
4993 @item -mthumb-interwork
4994 @kindex -mthumb-interwork
4995 @kindex -mno-thumb-interwork
4996 Generate code which supports calling between the THUMB and ARM
4997 instruction sets. Without this option the two instruction sets cannot
4998 be reliably used inside one program. The default is
4999 @samp{-mno-thumb-interwork}, since slightly smaller code is generated
5003 @kindex -mtpcs-frame
5004 @kindex -mno-tpcs-frame
5005 Generate a stack frame that is compliant with the Thumb Procedure Call
5006 Standard for all non-leaf functions. (A leaf function is one that does
5007 not call any other functions). The default is @samp{-mno-apcs-frame}.
5009 @item -mtpcs-leaf-frame
5010 @kindex -mtpcs-leaf-frame
5011 @kindex -mno-tpcs-leaf-frame
5012 Generate a stack frame that is compliant with the Thumb Procedure Call
5013 Standard for all leaf functions. (A leaf function is one that does
5014 not call any other functions). The default is @samp{-mno-apcs-leaf-frame}.
5016 @item -mlittle-endian
5017 @kindex -mlittle-endian
5018 Generate code for a processor running in little-endian mode. This is
5019 the default for all standard configurations.
5022 @kindex -mbig-endian
5023 Generate code for a processor running in big-endian mode.
5025 @item -mstructure-size-boundary=<n>
5026 @kindex -mstructure-size-boundary
5027 The size of all structures and unions will be rounded up to a multiple
5028 of the number of bits set by this option. Permissible values are 8 and
5029 32. The default value varies for different toolchains. For the COFF
5030 targeted toolchain the default value is 8. Specifying the larger number
5031 can produced faster, more efficient code, but can also increase the size
5032 of the program. The two values are potentially incompatible. Code
5033 compiled with one value cannot necessarily expect to work with code or
5034 libraries compiled with the other value, if they exchange information
5035 using structures or unions. Programmers are encouraged to use the 32
5036 value as future versions of the toolchain may default to this value.
5038 @item -mnop-fun-dllimport
5039 @kindex -mnop-fun-dllimport
5040 Disable the support for the @emph{dllimport} attribute.
5042 @item -mcallee-super-interworking
5043 @kindex -mcallee-super-interworking
5044 Gives all externally visible functions in the file being compiled an ARM
5045 instruction set header which switches to Thumb mode before executing the
5046 rest of the function. This allows these functions to be called from
5047 non-interworking code.
5049 @item -mcaller-super-interworking
5050 @kindex -mcaller-super-interworking
5051 Allows calls via function pointers (including virtual functions) to
5052 execute correctly regardless of whether the target code has been
5053 compiled for interworking or not. There is a small overhead in the cost
5054 of executing a function pointer if this option is enabled.
5056 @item -msingle-pic-base
5057 @kindex -msingle-pic-base
5058 Treat the register used for PIC addressing as read-only, rather than
5059 loading it in the prologue for each function. The run-time system is
5060 responsible for initialising this register with an appropriate value
5061 before execution begins.
5063 @item -mpic-register=<reg>
5064 @kindex -mpic-register=
5065 Specify the register to be used for PIC addressing. The default is R10.
5069 @node MN10200 Options
5070 @subsection MN10200 Options
5071 @cindex MN10200 options
5072 These @samp{-m} options are defined for Matsushita MN10200 architectures:
5076 Indicate to the linker that it should perform a relaxation optimization pass
5077 to shorten branches, calls and absolute memory addresses. This option only
5078 has an effect when used on the command line for the final link step.
5080 This option makes symbolic debugging impossible.
5083 @node MN10300 Options
5084 @subsection MN10300 Options
5085 @cindex MN10300 options
5086 These @samp{-m} options are defined for Matsushita MN10300 architectures:
5090 Generate code to avoid bugs in the multiply instructions for the MN10300
5091 processors. This is the default.
5094 Do not generate code to avoid bugs in the multiply instructions for the
5098 Generate code which uses features specific to the AM33 processor.
5101 Do not generate code which uses features specific to the AM33 processor. This
5105 Indicate to the linker that it should perform a relaxation optimization pass
5106 to shorten branches, calls and absolute memory addresses. This option only
5107 has an effect when used on the command line for the final link step.
5109 This option makes symbolic debugging impossible.
5113 @node M32R/D Options
5114 @subsection M32R/D Options
5115 @cindex M32R/D options
5117 These @samp{-m} options are defined for Mitsubishi M32R/D architectures:
5120 @item -mcode-model=small
5121 Assume all objects live in the lower 16MB of memory (so that their addresses
5122 can be loaded with the @code{ld24} instruction), and assume all subroutines
5123 are reachable with the @code{bl} instruction.
5124 This is the default.
5126 The addressability of a particular object can be set with the
5127 @code{model} attribute.
5129 @item -mcode-model=medium
5130 Assume objects may be anywhere in the 32 bit address space (the compiler
5131 will generate @code{seth/add3} instructions to load their addresses), and
5132 assume all subroutines are reachable with the @code{bl} instruction.
5134 @item -mcode-model=large
5135 Assume objects may be anywhere in the 32 bit address space (the compiler
5136 will generate @code{seth/add3} instructions to load their addresses), and
5137 assume subroutines may not be reachable with the @code{bl} instruction
5138 (the compiler will generate the much slower @code{seth/add3/jl}
5139 instruction sequence).
5142 Disable use of the small data area. Variables will be put into
5143 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
5144 @code{section} attribute has been specified).
5145 This is the default.
5147 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
5148 Objects may be explicitly put in the small data area with the
5149 @code{section} attribute using one of these sections.
5152 Put small global and static data in the small data area, but do not
5153 generate special code to reference them.
5156 Put small global and static data in the small data area, and generate
5157 special instructions to reference them.
5160 @cindex smaller data references
5161 Put global and static objects less than or equal to @var{num} bytes
5162 into the small data or bss sections instead of the normal data or bss
5163 sections. The default value of @var{num} is 8.
5164 The @samp{-msdata} option must be set to one of @samp{sdata} or @samp{use}
5165 for this option to have any effect.
5167 All modules should be compiled with the same @samp{-G @var{num}} value.
5168 Compiling with different values of @var{num} may or may not work; if it
5169 doesn't the linker will give an error message - incorrect code will not be
5175 @subsection M88K Options
5176 @cindex M88k options
5178 These @samp{-m} options are defined for Motorola 88k architectures:
5183 Generate code that works well on both the m88100 and the
5188 Generate code that works best for the m88100, but that also
5193 Generate code that works best for the m88110, and may not run
5198 Obsolete option to be removed from the next revision.
5201 @item -midentify-revision
5202 @kindex -midentify-revision
5204 @cindex identifying source, compiler (88k)
5205 Include an @code{ident} directive in the assembler output recording the
5206 source file name, compiler name and version, timestamp, and compilation
5209 @item -mno-underscores
5210 @kindex -mno-underscores
5211 @cindex underscores, avoiding (88k)
5212 In assembler output, emit symbol names without adding an underscore
5213 character at the beginning of each name. The default is to use an
5214 underscore as prefix on each name.
5216 @item -mocs-debug-info
5217 @itemx -mno-ocs-debug-info
5218 @kindex -mocs-debug-info
5219 @kindex -mno-ocs-debug-info
5221 @cindex debugging, 88k OCS
5222 Include (or omit) additional debugging information (about registers used
5223 in each stack frame) as specified in the 88open Object Compatibility
5224 Standard, ``OCS''. This extra information allows debugging of code that
5225 has had the frame pointer eliminated. The default for DG/UX, SVr4, and
5226 Delta 88 SVr3.2 is to include this information; other 88k configurations
5227 omit this information by default.
5229 @item -mocs-frame-position
5230 @kindex -mocs-frame-position
5231 @cindex register positions in frame (88k)
5232 When emitting COFF debugging information for automatic variables and
5233 parameters stored on the stack, use the offset from the canonical frame
5234 address, which is the stack pointer (register 31) on entry to the
5235 function. The DG/UX, SVr4, Delta88 SVr3.2, and BCS configurations use
5236 @samp{-mocs-frame-position}; other 88k configurations have the default
5237 @samp{-mno-ocs-frame-position}.
5239 @item -mno-ocs-frame-position
5240 @kindex -mno-ocs-frame-position
5241 @cindex register positions in frame (88k)
5242 When emitting COFF debugging information for automatic variables and
5243 parameters stored on the stack, use the offset from the frame pointer
5244 register (register 30). When this option is in effect, the frame
5245 pointer is not eliminated when debugging information is selected by the
5248 @item -moptimize-arg-area
5249 @itemx -mno-optimize-arg-area
5250 @kindex -moptimize-arg-area
5251 @kindex -mno-optimize-arg-area
5252 @cindex arguments in frame (88k)
5253 Control how function arguments are stored in stack frames.
5254 @samp{-moptimize-arg-area} saves space by optimizing them, but this
5255 conflicts with the 88open specifications. The opposite alternative,
5256 @samp{-mno-optimize-arg-area}, agrees with 88open standards. By default
5257 GCC does not optimize the argument area.
5259 @item -mshort-data-@var{num}
5260 @kindex -mshort-data-@var{num}
5261 @cindex smaller data references (88k)
5262 @cindex r0-relative references (88k)
5263 Generate smaller data references by making them relative to @code{r0},
5264 which allows loading a value using a single instruction (rather than the
5265 usual two). You control which data references are affected by
5266 specifying @var{num} with this option. For example, if you specify
5267 @samp{-mshort-data-512}, then the data references affected are those
5268 involving displacements of less than 512 bytes.
5269 @samp{-mshort-data-@var{num}} is not effective for @var{num} greater
5272 @item -mserialize-volatile
5273 @kindex -mserialize-volatile
5274 @itemx -mno-serialize-volatile
5275 @kindex -mno-serialize-volatile
5276 @cindex sequential consistency on 88k
5277 Do, or don't, generate code to guarantee sequential consistency
5278 of volatile memory references. By default, consistency is
5281 The order of memory references made by the MC88110 processor does
5282 not always match the order of the instructions requesting those
5283 references. In particular, a load instruction may execute before
5284 a preceding store instruction. Such reordering violates
5285 sequential consistency of volatile memory references, when there
5286 are multiple processors. When consistency must be guaranteed,
5287 GNU C generates special instructions, as needed, to force
5288 execution in the proper order.
5290 The MC88100 processor does not reorder memory references and so
5291 always provides sequential consistency. However, by default, GNU
5292 C generates the special instructions to guarantee consistency
5293 even when you use @samp{-m88100}, so that the code may be run on an
5294 MC88110 processor. If you intend to run your code only on the
5295 MC88100 processor, you may use @samp{-mno-serialize-volatile}.
5297 The extra code generated to guarantee consistency may affect the
5298 performance of your application. If you know that you can safely
5299 forgo this guarantee, you may use @samp{-mno-serialize-volatile}.
5305 @cindex assembler syntax, 88k
5307 Turn on (@samp{-msvr4}) or off (@samp{-msvr3}) compiler extensions
5308 related to System V release 4 (SVr4). This controls the following:
5312 Which variant of the assembler syntax to emit.
5314 @samp{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
5315 that is used on System V release 4.
5317 @samp{-msvr4} makes GCC issue additional declaration directives used in
5321 @samp{-msvr4} is the default for the m88k-motorola-sysv4 and
5322 m88k-dg-dgux m88k configurations. @samp{-msvr3} is the default for all
5323 other m88k configurations.
5325 @item -mversion-03.00
5326 @kindex -mversion-03.00
5327 This option is obsolete, and is ignored.
5328 @c ??? which asm syntax better for GAS? option there too?
5330 @item -mno-check-zero-division
5331 @itemx -mcheck-zero-division
5332 @kindex -mno-check-zero-division
5333 @kindex -mcheck-zero-division
5334 @cindex zero division on 88k
5335 Do, or don't, generate code to guarantee that integer division by
5336 zero will be detected. By default, detection is guaranteed.
5338 Some models of the MC88100 processor fail to trap upon integer
5339 division by zero under certain conditions. By default, when
5340 compiling code that might be run on such a processor, GNU C
5341 generates code that explicitly checks for zero-valued divisors
5342 and traps with exception number 503 when one is detected. Use of
5343 mno-check-zero-division suppresses such checking for code
5344 generated to run on an MC88100 processor.
5346 GNU C assumes that the MC88110 processor correctly detects all
5347 instances of integer division by zero. When @samp{-m88110} is
5348 specified, both @samp{-mcheck-zero-division} and
5349 @samp{-mno-check-zero-division} are ignored, and no explicit checks for
5350 zero-valued divisors are generated.
5352 @item -muse-div-instruction
5353 @kindex -muse-div-instruction
5354 @cindex divide instruction, 88k
5355 Use the div instruction for signed integer division on the
5356 MC88100 processor. By default, the div instruction is not used.
5358 On the MC88100 processor the signed integer division instruction
5359 div) traps to the operating system on a negative operand. The
5360 operating system transparently completes the operation, but at a
5361 large cost in execution time. By default, when compiling code
5362 that might be run on an MC88100 processor, GNU C emulates signed
5363 integer division using the unsigned integer division instruction
5364 divu), thereby avoiding the large penalty of a trap to the
5365 operating system. Such emulation has its own, smaller, execution
5366 cost in both time and space. To the extent that your code's
5367 important signed integer division operations are performed on two
5368 nonnegative operands, it may be desirable to use the div
5369 instruction directly.
5371 On the MC88110 processor the div instruction (also known as the
5372 divs instruction) processes negative operands without trapping to
5373 the operating system. When @samp{-m88110} is specified,
5374 @samp{-muse-div-instruction} is ignored, and the div instruction is used
5375 for signed integer division.
5377 Note that the result of dividing INT_MIN by -1 is undefined. In
5378 particular, the behavior of such a division with and without
5379 @samp{-muse-div-instruction} may differ.
5381 @item -mtrap-large-shift
5382 @itemx -mhandle-large-shift
5383 @kindex -mtrap-large-shift
5384 @kindex -mhandle-large-shift
5385 @cindex bit shift overflow (88k)
5386 @cindex large bit shifts (88k)
5387 Include code to detect bit-shifts of more than 31 bits; respectively,
5388 trap such shifts or emit code to handle them properly. By default GCC
5389 makes no special provision for large bit shifts.
5391 @item -mwarn-passed-structs
5392 @kindex -mwarn-passed-structs
5393 @cindex structure passing (88k)
5394 Warn when a function passes a struct as an argument or result.
5395 Structure-passing conventions have changed during the evolution of the C
5396 language, and are often the source of portability problems. By default,
5397 GCC issues no such warning.
5400 @node RS/6000 and PowerPC Options
5401 @subsection IBM RS/6000 and PowerPC Options
5402 @cindex RS/6000 and PowerPC Options
5403 @cindex IBM RS/6000 and PowerPC Options
5405 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
5413 @itemx -mpowerpc-gpopt
5414 @itemx -mno-powerpc-gpopt
5415 @itemx -mpowerpc-gfxopt
5416 @itemx -mno-powerpc-gfxopt
5418 @itemx -mno-powerpc64
5422 @kindex -mpowerpc-gpopt
5423 @kindex -mpowerpc-gfxopt
5425 GCC supports two related instruction set architectures for the
5426 RS/6000 and PowerPC. The @dfn{POWER} instruction set are those
5427 instructions supported by the @samp{rios} chip set used in the original
5428 RS/6000 systems and the @dfn{PowerPC} instruction set is the
5429 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
5430 the IBM 4xx microprocessors.
5432 Neither architecture is a subset of the other. However there is a
5433 large common subset of instructions supported by both. An MQ
5434 register is included in processors supporting the POWER architecture.
5436 You use these options to specify which instructions are available on the
5437 processor you are using. The default value of these options is
5438 determined when configuring GCC. Specifying the
5439 @samp{-mcpu=@var{cpu_type}} overrides the specification of these
5440 options. We recommend you use the @samp{-mcpu=@var{cpu_type}} option
5441 rather than the options listed above.
5443 The @samp{-mpower} option allows GCC to generate instructions that
5444 are found only in the POWER architecture and to use the MQ register.
5445 Specifying @samp{-mpower2} implies @samp{-power} and also allows GCC
5446 to generate instructions that are present in the POWER2 architecture but
5447 not the original POWER architecture.
5449 The @samp{-mpowerpc} option allows GCC to generate instructions that
5450 are found only in the 32-bit subset of the PowerPC architecture.
5451 Specifying @samp{-mpowerpc-gpopt} implies @samp{-mpowerpc} and also allows
5452 GCC to use the optional PowerPC architecture instructions in the
5453 General Purpose group, including floating-point square root. Specifying
5454 @samp{-mpowerpc-gfxopt} implies @samp{-mpowerpc} and also allows GCC to
5455 use the optional PowerPC architecture instructions in the Graphics
5456 group, including floating-point select.
5458 The @samp{-mpowerpc64} option allows GCC to generate the additional
5459 64-bit instructions that are found in the full PowerPC64 architecture
5460 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
5461 @samp{-mno-powerpc64}.
5463 If you specify both @samp{-mno-power} and @samp{-mno-powerpc}, GCC
5464 will use only the instructions in the common subset of both
5465 architectures plus some special AIX common-mode calls, and will not use
5466 the MQ register. Specifying both @samp{-mpower} and @samp{-mpowerpc}
5467 permits GCC to use any instruction from either architecture and to
5468 allow use of the MQ register; specify this for the Motorola MPC601.
5470 @item -mnew-mnemonics
5471 @itemx -mold-mnemonics
5472 @kindex -mnew-mnemonics
5473 @kindex -mold-mnemonics
5474 Select which mnemonics to use in the generated assembler code.
5475 @samp{-mnew-mnemonics} requests output that uses the assembler mnemonics
5476 defined for the PowerPC architecture, while @samp{-mold-mnemonics}
5477 requests the assembler mnemonics defined for the POWER architecture.
5478 Instructions defined in only one architecture have only one mnemonic;
5479 GCC uses that mnemonic irrespective of which of these options is
5482 GCC defaults to the mnemonics appropriate for the architecture in
5483 use. Specifying @samp{-mcpu=@var{cpu_type}} sometimes overrides the
5484 value of these option. Unless you are building a cross-compiler, you
5485 should normally not specify either @samp{-mnew-mnemonics} or
5486 @samp{-mold-mnemonics}, but should instead accept the default.
5488 @item -mcpu=@var{cpu_type}
5490 Set architecture type, register usage, choice of mnemonics, and
5491 instruction scheduling parameters for machine type @var{cpu_type}.
5492 Supported values for @var{cpu_type} are @samp{rios}, @samp{rios1},
5493 @samp{rsc}, @samp{rios2}, @samp{rs64a}, @samp{601}, @samp{602},
5494 @samp{603}, @samp{603e}, @samp{604}, @samp{604e}, @samp{620},
5495 @samp{630}, @samp{740}, @samp{750}, @samp{power}, @samp{power2},
5496 @samp{powerpc}, @samp{403}, @samp{505}, @samp{801}, @samp{821},
5497 @samp{823}, and @samp{860} and @samp{common}. @samp{-mcpu=power},
5498 @samp{-mcpu=power2}, @samp{-mcpu=powerpc}, and @samp{-mcpu=powerpc64}
5499 specify generic POWER, POWER2, pure 32-bit PowerPC (i.e., not MPC601),
5500 and 64-bit PowerPC architecture machine types, with an appropriate,
5501 generic processor model assumed for scheduling purposes.@refill
5503 Specifying any of the following options:
5504 @samp{-mcpu=rios1}, @samp{-mcpu=rios2}, @samp{-mcpu=rsc},
5505 @samp{-mcpu=power}, or @samp{-mcpu=power2}
5506 enables the @samp{-mpower} option and disables the @samp{-mpowerpc} option;
5507 @samp{-mcpu=601} enables both the @samp{-mpower} and @samp{-mpowerpc} options.
5508 All of @samp{-mcpu=rs64a}, @samp{-mcpu=602}, @samp{-mcpu=603},
5509 @samp{-mcpu=603e}, @samp{-mcpu=604}, @samp{-mcpu=620}, @samp{-mcpu=630},
5510 @samp{-mcpu=740}, and @samp{-mcpu=750}
5511 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
5512 Exactly similarly, all of @samp{-mcpu=403},
5513 @samp{-mcpu=505}, @samp{-mcpu=821}, @samp{-mcpu=860} and @samp{-mcpu=powerpc}
5514 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
5515 @samp{-mcpu=common} disables both the
5516 @samp{-mpower} and @samp{-mpowerpc} options.@refill
5518 AIX versions 4 or greater selects @samp{-mcpu=common} by default, so
5519 that code will operate on all members of the RS/6000 POWER and PowerPC
5520 families. In that case, GCC will use only the instructions in the
5521 common subset of both architectures plus some special AIX common-mode
5522 calls, and will not use the MQ register. GCC assumes a generic
5523 processor model for scheduling purposes.
5525 Specifying any of the options @samp{-mcpu=rios1}, @samp{-mcpu=rios2},
5526 @samp{-mcpu=rsc}, @samp{-mcpu=power}, or @samp{-mcpu=power2} also
5527 disables the @samp{new-mnemonics} option. Specifying @samp{-mcpu=601},
5528 @samp{-mcpu=602}, @samp{-mcpu=603}, @samp{-mcpu=603e}, @samp{-mcpu=604},
5529 @samp{-mcpu=620}, @samp{-mcpu=630}, @samp{-mcpu=403}, @samp{-mcpu=505},
5530 @samp{-mcpu=821}, @samp{-mcpu=860} or @samp{-mcpu=powerpc} also enables
5531 the @samp{new-mnemonics} option.@refill
5533 Specifying @samp{-mcpu=403}, @samp{-mcpu=821}, or @samp{-mcpu=860} also
5534 enables the @samp{-msoft-float} option.
5536 @item -mtune=@var{cpu_type}
5537 Set the instruction scheduling parameters for machine type
5538 @var{cpu_type}, but do not set the architecture type, register usage,
5539 choice of mnemonics like @samp{-mcpu=}@var{cpu_type} would. The same
5540 values for @var{cpu_type} are used for @samp{-mtune=}@var{cpu_type} as
5541 for @samp{-mcpu=}@var{cpu_type}. The @samp{-mtune=}@var{cpu_type}
5542 option overrides the @samp{-mcpu=}@var{cpu_type} option in terms of
5543 instruction scheduling parameters.
5546 @itemx -mno-fp-in-toc
5547 @itemx -mno-sum-in-toc
5548 @itemx -mminimal-toc
5549 @kindex -mminimal-toc
5550 Modify generation of the TOC (Table Of Contents), which is created for
5551 every executable file. The @samp{-mfull-toc} option is selected by
5552 default. In that case, GCC will allocate at least one TOC entry for
5553 each unique non-automatic variable reference in your program. GCC
5554 will also place floating-point constants in the TOC. However, only
5555 16,384 entries are available in the TOC.
5557 If you receive a linker error message that saying you have overflowed
5558 the available TOC space, you can reduce the amount of TOC space used
5559 with the @samp{-mno-fp-in-toc} and @samp{-mno-sum-in-toc} options.
5560 @samp{-mno-fp-in-toc} prevents GCC from putting floating-point
5561 constants in the TOC and @samp{-mno-sum-in-toc} forces GCC to
5562 generate code to calculate the sum of an address and a constant at
5563 run-time instead of putting that sum into the TOC. You may specify one
5564 or both of these options. Each causes GCC to produce very slightly
5565 slower and larger code at the expense of conserving TOC space.
5567 If you still run out of space in the TOC even when you specify both of
5568 these options, specify @samp{-mminimal-toc} instead. This option causes
5569 GCC to make only one TOC entry for every file. When you specify this
5570 option, GCC will produce code that is slower and larger but which
5571 uses extremely little TOC space. You may wish to use this option
5572 only on files that contain less frequently executed code. @refill
5578 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
5579 @code{long} type, and the infrastructure needed to support them.
5580 Specifying @samp{-maix64} implies @samp{-mpowerpc64} and
5581 @samp{-mpowerpc}, while @samp{-maix32} disables the 64-bit ABI and
5582 implies @samp{-mno-powerpc64}. GCC defaults to @samp{-maix32}.
5587 On AIX, pass floating-point arguments to prototyped functions beyond the
5588 register save area (RSA) on the stack in addition to argument FPRs. The
5589 AIX calling convention was extended but not initially documented to
5590 handle an obscure K&R C case of calling a function that takes the
5591 address of its arguments with fewer arguments than declared. AIX XL
5592 compilers access floating point arguments which do not fit in the
5593 RSA from the stack when a subroutine is compiled without
5594 optimization. Because always storing floating-point arguments on the
5595 stack is inefficient and rarely needed, this option is not enabled by
5596 default and only is necessary when calling subroutines compiled by AIX
5597 XL compilers without optimization.
5601 Support @dfn{AIX Threads}. Link an application written to use
5602 @dfn{pthreads} with special libraries and startup code to enable the
5607 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE). Link an
5608 application written to use message passing with special startup code to
5609 enable the application to run. The system must have PE installed in the
5610 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
5611 must be overridden with the @samp{-specs=} option to specify the
5612 appropriate directory location. The Parallel Environment does not
5613 support threads, so the @samp{-mpe} option and the @samp{-mthreads}
5614 option are incompatible.
5618 @kindex -msoft-float
5619 Generate code that does not use (uses) the floating-point register set.
5620 Software floating point emulation is provided if you use the
5621 @samp{-msoft-float} option, and pass the option to GCC when linking.
5624 @itemx -mno-multiple
5625 Generate code that uses (does not use) the load multiple word
5626 instructions and the store multiple word instructions. These
5627 instructions are generated by default on POWER systems, and not
5628 generated on PowerPC systems. Do not use @samp{-mmultiple} on little
5629 endian PowerPC systems, since those instructions do not work when the
5630 processor is in little endian mode. The exceptions are PPC740 and
5631 PPC750 which permit the instructions usage in little endian mode.
5636 Generate code that uses (does not use) the load string instructions
5637 and the store string word instructions to save multiple registers and
5638 do small block moves. These instructions are generated by default on
5639 POWER systems, and not generated on PowerPC systems. Do not use
5640 @samp{-mstring} on little endian PowerPC systems, since those
5641 instructions do not work when the processor is in little endian mode.
5642 The exceptions are PPC740 and PPC750 which permit the instructions
5643 usage in little endian mode.
5648 Generate code that uses (does not use) the load or store instructions
5649 that update the base register to the address of the calculated memory
5650 location. These instructions are generated by default. If you use
5651 @samp{-mno-update}, there is a small window between the time that the
5652 stack pointer is updated and the address of the previous frame is
5653 stored, which means code that walks the stack frame across interrupts or
5654 signals may get corrupted data.
5657 @itemx -mno-fused-madd
5658 @kindex -mfused-madd
5659 Generate code that uses (does not use) the floating point multiply and
5660 accumulate instructions. These instructions are generated by default if
5661 hardware floating is used.
5663 @item -mno-bit-align
5666 On System V.4 and embedded PowerPC systems do not (do) force structures
5667 and unions that contain bit fields to be aligned to the base type of the
5670 For example, by default a structure containing nothing but 8
5671 @code{unsigned} bitfields of length 1 would be aligned to a 4 byte
5672 boundary and have a size of 4 bytes. By using @samp{-mno-bit-align},
5673 the structure would be aligned to a 1 byte boundary and be one byte in
5676 @item -mno-strict-align
5677 @itemx -mstrict-align
5678 @kindex -mstrict-align
5679 On System V.4 and embedded PowerPC systems do not (do) assume that
5680 unaligned memory references will be handled by the system.
5683 @itemx -mno-relocatable
5684 @kindex -mrelocatable
5685 On embedded PowerPC systems generate code that allows (does not allow)
5686 the program to be relocated to a different address at runtime. If you
5687 use @samp{-mrelocatable} on any module, all objects linked together must
5688 be compiled with @samp{-mrelocatable} or @samp{-mrelocatable-lib}.
5690 @item -mrelocatable-lib
5691 @itemx -mno-relocatable-lib
5692 On embedded PowerPC systems generate code that allows (does not allow)
5693 the program to be relocated to a different address at runtime. Modules
5694 compiled with @samp{-mrelocatable-lib} can be linked with either modules
5695 compiled without @samp{-mrelocatable} and @samp{-mrelocatable-lib} or
5696 with modules compiled with the @samp{-mrelocatable} options.
5700 On System V.4 and embedded PowerPC systems do not (do) assume that
5701 register 2 contains a pointer to a global area pointing to the addresses
5702 used in the program.
5705 @itemx -mlittle-endian
5706 On System V.4 and embedded PowerPC systems compile code for the
5707 processor in little endian mode. The @samp{-mlittle-endian} option is
5708 the same as @samp{-mlittle}.
5712 On System V.4 and embedded PowerPC systems compile code for the
5713 processor in big endian mode. The @samp{-mbig-endian} option is
5714 the same as @samp{-mbig}.
5717 On System V.4 and embedded PowerPC systems compile code using calling
5718 conventions that adheres to the March 1995 draft of the System V
5719 Application Binary Interface, PowerPC processor supplement. This is the
5720 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
5722 @item -mcall-sysv-eabi
5723 Specify both @samp{-mcall-sysv} and @samp{-meabi} options.
5725 @item -mcall-sysv-noeabi
5726 Specify both @samp{-mcall-sysv} and @samp{-mno-eabi} options.
5729 On System V.4 and embedded PowerPC systems compile code using calling
5730 conventions that are similar to those used on AIX. This is the
5731 default if you configured GCC using @samp{powerpc-*-eabiaix}.
5733 @item -mcall-solaris
5734 On System V.4 and embedded PowerPC systems compile code for the Solaris
5738 On System V.4 and embedded PowerPC systems compile code for the
5739 Linux-based GNU system.
5742 @itemx -mno-prototype
5743 On System V.4 and embedded PowerPC systems assume that all calls to
5744 variable argument functions are properly prototyped. Otherwise, the
5745 compiler must insert an instruction before every non prototyped call to
5746 set or clear bit 6 of the condition code register (@var{CR}) to
5747 indicate whether floating point values were passed in the floating point
5748 registers in case the function takes a variable arguments. With
5749 @samp{-mprototype}, only calls to prototyped variable argument functions
5750 will set or clear the bit.
5753 On embedded PowerPC systems, assume that the startup module is called
5754 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
5755 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
5759 On embedded PowerPC systems, assume that the startup module is called
5760 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
5764 On embedded PowerPC systems, assume that the startup module is called
5765 @file{crt0.o} and the standard C libraries are @file{libads.a} and
5769 On embedded PowerPC systems, assume that the startup module is called
5770 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
5774 On System V.4 and embedded PowerPC systems, specify that you are
5775 compiling for a VxWorks system.
5778 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
5779 header to indicate that @samp{eabi} extended relocations are used.
5783 On System V.4 and embedded PowerPC systems do (do not) adhere to the
5784 Embedded Applications Binary Interface (eabi) which is a set of
5785 modifications to the System V.4 specifications. Selecting @code{-meabi}
5786 means that the stack is aligned to an 8 byte boundary, a function
5787 @code{__eabi} is called to from @code{main} to set up the eabi
5788 environment, and the @samp{-msdata} option can use both @code{r2} and
5789 @code{r13} to point to two separate small data areas. Selecting
5790 @code{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
5791 do not call an initialization function from @code{main}, and the
5792 @samp{-msdata} option will only use @code{r13} to point to a single
5793 small data area. The @samp{-meabi} option is on by default if you
5794 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
5797 On System V.4 and embedded PowerPC systems, put small initialized
5798 @code{const} global and static data in the @samp{.sdata2} section, which
5799 is pointed to by register @code{r2}. Put small initialized
5800 non-@code{const} global and static data in the @samp{.sdata} section,
5801 which is pointed to by register @code{r13}. Put small uninitialized
5802 global and static data in the @samp{.sbss} section, which is adjacent to
5803 the @samp{.sdata} section. The @samp{-msdata=eabi} option is
5804 incompatible with the @samp{-mrelocatable} option. The
5805 @samp{-msdata=eabi} option also sets the @samp{-memb} option.
5808 On System V.4 and embedded PowerPC systems, put small global and static
5809 data in the @samp{.sdata} section, which is pointed to by register
5810 @code{r13}. Put small uninitialized global and static data in the
5811 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
5812 The @samp{-msdata=sysv} option is incompatible with the
5813 @samp{-mrelocatable} option.
5815 @item -msdata=default
5817 On System V.4 and embedded PowerPC systems, if @samp{-meabi} is used,
5818 compile code the same as @samp{-msdata=eabi}, otherwise compile code the
5819 same as @samp{-msdata=sysv}.
5822 On System V.4 and embedded PowerPC systems, put small global and static
5823 data in the @samp{.sdata} section. Put small uninitialized global and
5824 static data in the @samp{.sbss} section. Do not use register @code{r13}
5825 to address small data however. This is the default behavior unless
5826 other @samp{-msdata} options are used.
5830 On embedded PowerPC systems, put all initialized global and static data
5831 in the @samp{.data} section, and all uninitialized data in the
5832 @samp{.bss} section.
5835 @cindex smaller data references (PowerPC)
5836 @cindex .sdata/.sdata2 references (PowerPC)
5837 On embedded PowerPC systems, put global and static items less than or
5838 equal to @var{num} bytes into the small data or bss sections instead of
5839 the normal data or bss section. By default, @var{num} is 8. The
5840 @samp{-G @var{num}} switch is also passed to the linker.
5841 All modules should be compiled with the same @samp{-G @var{num}} value.
5844 @itemx -mno-regnames
5845 On System V.4 and embedded PowerPC systems do (do not) emit register
5846 names in the assembly language output using symbolic forms.
5851 @subsection IBM RT Options
5853 @cindex IBM RT options
5855 These @samp{-m} options are defined for the IBM RT PC:
5859 Use an in-line code sequence for integer multiplies. This is the
5862 @item -mcall-lib-mul
5863 Call @code{lmul$$} for integer multiples.
5865 @item -mfull-fp-blocks
5866 Generate full-size floating point data blocks, including the minimum
5867 amount of scratch space recommended by IBM. This is the default.
5869 @item -mminimum-fp-blocks
5870 Do not include extra scratch space in floating point data blocks. This
5871 results in smaller code, but slower execution, since scratch space must
5872 be allocated dynamically.
5874 @cindex @file{varargs.h} and RT PC
5875 @cindex @file{stdarg.h} and RT PC
5876 @item -mfp-arg-in-fpregs
5877 Use a calling sequence incompatible with the IBM calling convention in
5878 which floating point arguments are passed in floating point registers.
5879 Note that @code{varargs.h} and @code{stdargs.h} will not work with
5880 floating point operands if this option is specified.
5882 @item -mfp-arg-in-gregs
5883 Use the normal calling convention for floating point arguments. This is
5886 @item -mhc-struct-return
5887 Return structures of more than one word in memory, rather than in a
5888 register. This provides compatibility with the MetaWare HighC (hc)
5889 compiler. Use the option @samp{-fpcc-struct-return} for compatibility
5890 with the Portable C Compiler (pcc).
5892 @item -mnohc-struct-return
5893 Return some structures of more than one word in registers, when
5894 convenient. This is the default. For compatibility with the
5895 IBM-supplied compilers, use the option @samp{-fpcc-struct-return} or the
5896 option @samp{-mhc-struct-return}.
5900 @subsection MIPS Options
5901 @cindex MIPS options
5903 These @samp{-m} options are defined for the MIPS family of computers:
5906 @item -mcpu=@var{cpu type}
5907 Assume the defaults for the machine type @var{cpu type} when scheduling
5908 instructions. The choices for @var{cpu type} are @samp{r2000}, @samp{r3000},
5909 @samp{r3900}, @samp{r4000}, @samp{r4100}, @samp{r4300}, @samp{r4400},
5910 @samp{r4600}, @samp{r4650}, @samp{r5000}, @samp{r6000}, @samp{r8000},
5911 and @samp{orion}. Additionally, the @samp{r2000}, @samp{r3000},
5912 @samp{r4000}, @samp{r5000}, and @samp{r6000} can be abbreviated as
5913 @samp{r2k} (or @samp{r2K}), @samp{r3k}, etc. While picking a specific
5914 @var{cpu type} will schedule things appropriately for that particular
5915 chip, the compiler will not generate any code that does not meet level 1
5916 of the MIPS ISA (instruction set architecture) without a @samp{-mipsX}
5917 or @samp{-mabi} switch being used.
5920 Issue instructions from level 1 of the MIPS ISA. This is the default.
5921 @samp{r3000} is the default @var{cpu type} at this ISA level.
5924 Issue instructions from level 2 of the MIPS ISA (branch likely, square
5925 root instructions). @samp{r6000} is the default @var{cpu type} at this
5929 Issue instructions from level 3 of the MIPS ISA (64 bit instructions).
5930 @samp{r4000} is the default @var{cpu type} at this ISA level.
5933 Issue instructions from level 4 of the MIPS ISA (conditional move,
5934 prefetch, enhanced FPU instructions). @samp{r8000} is the default
5935 @var{cpu type} at this ISA level.
5938 Assume that 32 32-bit floating point registers are available. This is
5942 Assume that 32 64-bit floating point registers are available. This is
5943 the default when the @samp{-mips3} option is used.
5946 Assume that 32 32-bit general purpose registers are available. This is
5950 Assume that 32 64-bit general purpose registers are available. This is
5951 the default when the @samp{-mips3} option is used.
5954 Force int and long types to be 64 bits wide. See @samp{-mlong32} for an
5955 explanation of the default, and the width of pointers.
5958 Force long types to be 64 bits wide. See @samp{-mlong32} for an
5959 explanation of the default, and the width of pointers.
5962 Force long, int, and pointer types to be 32 bits wide.
5964 If none of @samp{-mlong32}, @samp{-mlong64}, or @samp{-mint64} are set,
5965 the size of ints, longs, and pointers depends on the ABI and ISA chosen.
5966 For @samp{-mabi=32}, and @samp{-mabi=n32}, ints and longs are 32 bits
5967 wide. For @samp{-mabi=64}, ints are 32 bits, and longs are 64 bits wide.
5968 For @samp{-mabi=eabi} and either @samp{-mips1} or @samp{-mips2}, ints
5969 and longs are 32 bits wide. For @samp{-mabi=eabi} and higher ISAs, ints
5970 are 32 bits, and longs are 64 bits wide. The width of pointer types is
5971 the smaller of the width of longs or the width of general purpose
5972 registers (which in turn depends on the ISA).
5979 Generate code for the indicated ABI. The default instruction level is
5980 @samp{-mips1} for @samp{32}, @samp{-mips3} for @samp{n32}, and
5981 @samp{-mips4} otherwise. Conversely, with @samp{-mips1} or
5982 @samp{-mips2}, the default ABI is @samp{32}; otherwise, the default ABI
5986 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
5987 add normal debug information. This is the default for all
5988 platforms except for the OSF/1 reference platform, using the OSF/rose
5989 object format. If the either of the @samp{-gstabs} or @samp{-gstabs+}
5990 switches are used, the @file{mips-tfile} program will encapsulate the
5991 stabs within MIPS ECOFF.
5994 Generate code for the GNU assembler. This is the default on the OSF/1
5995 reference platform, using the OSF/rose object format. Also, this is
5996 the default if the configure option @samp{--with-gnu-as} is used.
5998 @item -msplit-addresses
5999 @itemx -mno-split-addresses
6000 Generate code to load the high and low parts of address constants separately.
6001 This allows @code{gcc} to optimize away redundant loads of the high order
6002 bits of addresses. This optimization requires GNU as and GNU ld.
6003 This optimization is enabled by default for some embedded targets where
6004 GNU as and GNU ld are standard.
6008 The @samp{-mrnames} switch says to output code using the MIPS software
6009 names for the registers, instead of the hardware names (ie, @var{a0}
6010 instead of @var{$4}). The only known assembler that supports this option
6011 is the Algorithmics assembler.
6015 The @samp{-mgpopt} switch says to write all of the data declarations
6016 before the instructions in the text section, this allows the MIPS
6017 assembler to generate one word memory references instead of using two
6018 words for short global or static data items. This is on by default if
6019 optimization is selected.
6023 For each non-inline function processed, the @samp{-mstats} switch
6024 causes the compiler to emit one line to the standard error file to
6025 print statistics about the program (number of registers saved, stack
6030 The @samp{-mmemcpy} switch makes all block moves call the appropriate
6031 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
6032 generating inline code.
6035 @itemx -mno-mips-tfile
6036 The @samp{-mno-mips-tfile} switch causes the compiler not
6037 postprocess the object file with the @file{mips-tfile} program,
6038 after the MIPS assembler has generated it to add debug support. If
6039 @file{mips-tfile} is not run, then no local variables will be
6040 available to the debugger. In addition, @file{stage2} and
6041 @file{stage3} objects will have the temporary file names passed to the
6042 assembler embedded in the object file, which means the objects will
6043 not compare the same. The @samp{-mno-mips-tfile} switch should only
6044 be used when there are bugs in the @file{mips-tfile} program that
6045 prevents compilation.
6048 Generate output containing library calls for floating point.
6049 @strong{Warning:} the requisite libraries are not part of GCC.
6050 Normally the facilities of the machine's usual C compiler are used, but
6051 this can't be done directly in cross-compilation. You must make your
6052 own arrangements to provide suitable library functions for
6056 Generate output containing floating point instructions. This is the
6057 default if you use the unmodified sources.
6060 @itemx -mno-abicalls
6061 Emit (or do not emit) the pseudo operations @samp{.abicalls},
6062 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
6063 position independent code.
6066 @itemx -mno-long-calls
6067 Do all calls with the @samp{JALR} instruction, which requires
6068 loading up a function's address into a register before the call.
6069 You need to use this switch, if you call outside of the current
6070 512 megabyte segment to functions that are not through pointers.
6073 @itemx -mno-half-pic
6074 Put pointers to extern references into the data section and load them
6075 up, rather than put the references in the text section.
6077 @item -membedded-pic
6078 @itemx -mno-embedded-pic
6079 Generate PIC code suitable for some embedded systems. All calls are
6080 made using PC relative address, and all data is addressed using the $gp
6081 register. No more than 65536 bytes of global data may be used. This
6082 requires GNU as and GNU ld which do most of the work. This currently
6083 only works on targets which use ECOFF; it does not work with ELF.
6085 @item -membedded-data
6086 @itemx -mno-embedded-data
6087 Allocate variables to the read-only data section first if possible, then
6088 next in the small data section if possible, otherwise in data. This gives
6089 slightly slower code than the default, but reduces the amount of RAM required
6090 when executing, and thus may be preferred for some embedded systems.
6092 @item -muninit-const-in-rodata
6093 @itemx -mno-uninit-const-in-rodata
6094 When used together with -membedded-data, it will always store uninitialized
6095 const variables in the read-only data section.
6097 @item -msingle-float
6098 @itemx -mdouble-float
6099 The @samp{-msingle-float} switch tells gcc to assume that the floating
6100 point coprocessor only supports single precision operations, as on the
6101 @samp{r4650} chip. The @samp{-mdouble-float} switch permits gcc to use
6102 double precision operations. This is the default.
6106 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
6107 as on the @samp{r4650} chip.
6110 Turns on @samp{-msingle-float}, @samp{-mmad}, and, at least for now,
6115 Enable 16-bit instructions.
6118 Use the entry and exit pseudo ops. This option can only be used with
6122 Compile code for the processor in little endian mode.
6123 The requisite libraries are assumed to exist.
6126 Compile code for the processor in big endian mode.
6127 The requisite libraries are assumed to exist.
6130 @cindex smaller data references (MIPS)
6131 @cindex gp-relative references (MIPS)
6132 Put global and static items less than or equal to @var{num} bytes into
6133 the small data or bss sections instead of the normal data or bss
6134 section. This allows the assembler to emit one word memory reference
6135 instructions based on the global pointer (@var{gp} or @var{$28}),
6136 instead of the normal two words used. By default, @var{num} is 8 when
6137 the MIPS assembler is used, and 0 when the GNU assembler is used. The
6138 @samp{-G @var{num}} switch is also passed to the assembler and linker.
6139 All modules should be compiled with the same @samp{-G @var{num}}
6143 Tell the MIPS assembler to not run its preprocessor over user
6144 assembler files (with a @samp{.s} suffix) when assembling them.
6147 Pass an option to gas which will cause nops to be inserted if
6148 the read of the destination register of an mfhi or mflo instruction
6149 occurs in the following two instructions.
6152 Do not include the default crt0.
6156 These options are defined by the macro
6157 @code{TARGET_SWITCHES} in the machine description. The default for the
6158 options is also defined by that macro, which enables you to change the
6163 @subsection Intel 386 Options
6164 @cindex i386 Options
6165 @cindex Intel 386 Options
6167 These @samp{-m} options are defined for the i386 family of computers:
6170 @item -mcpu=@var{cpu type}
6171 Assume the defaults for the machine type @var{cpu type} when scheduling
6172 instructions. The choices for @var{cpu type} are:
6174 @multitable @columnfractions .20 .20 .20 .20
6175 @item @samp{i386} @tab @samp{i486} @tab @samp{i586} @tab @samp{i686}
6176 @item @samp{pentium} @tab @samp{pentiumpro} @tab @samp{k6} @tab @samp{athlon}
6179 While picking a specific @var{cpu type} will schedule things appropriately
6180 for that particular chip, the compiler will not generate any code that
6181 does not run on the i386 without the @samp{-march=@var{cpu type}} option
6182 being used. @samp{i586} is equivalent to @samp{pentium} and @samp{i686}
6183 is equivalent to @samp{pentiumpro}. @samp{k6} is the AMD chip as
6184 opposed to the Intel ones.
6186 @item -march=@var{cpu type}
6187 Generate instructions for the machine type @var{cpu type}. The choices
6188 for @var{cpu type} are the same as for @samp{-mcpu}. Moreover,
6189 specifying @samp{-march=@var{cpu type}} implies @samp{-mcpu=@var{cpu type}}.
6195 Synonyms for -mcpu=i386, -mcpu=i486, -mcpu=pentium, and -mcpu=pentiumpro
6196 respectively. These synonyms are deprecated.
6198 @item -mintel-syntax
6199 Emit assembly using Intel syntax opcodes instead of AT&T syntax.
6203 Control whether or not the compiler uses IEEE floating point
6204 comparisons. These handle correctly the case where the result of a
6205 comparison is unordered.
6208 Generate output containing library calls for floating point.
6209 @strong{Warning:} the requisite libraries are not part of GCC.
6210 Normally the facilities of the machine's usual C compiler are used, but
6211 this can't be done directly in cross-compilation. You must make your
6212 own arrangements to provide suitable library functions for
6215 On machines where a function returns floating point results in the 80387
6216 register stack, some floating point opcodes may be emitted even if
6217 @samp{-msoft-float} is used.
6219 @item -mno-fp-ret-in-387
6220 Do not use the FPU registers for return values of functions.
6222 The usual calling convention has functions return values of types
6223 @code{float} and @code{double} in an FPU register, even if there
6224 is no FPU. The idea is that the operating system should emulate
6227 The option @samp{-mno-fp-ret-in-387} causes such values to be returned
6228 in ordinary CPU registers instead.
6230 @item -mno-fancy-math-387
6231 Some 387 emulators do not support the @code{sin}, @code{cos} and
6232 @code{sqrt} instructions for the 387. Specify this option to avoid
6233 generating those instructions. This option is the default on FreeBSD.
6234 As of revision 2.6.1, these instructions are not generated unless you
6235 also use the @samp{-ffast-math} switch.
6237 @item -malign-double
6238 @itemx -mno-align-double
6239 Control whether GCC aligns @code{double}, @code{long double}, and
6240 @code{long long} variables on a two word boundary or a one word
6241 boundary. Aligning @code{double} variables on a two word boundary will
6242 produce code that runs somewhat faster on a @samp{Pentium} at the
6243 expense of more memory.
6245 @strong{Warning:} if you use the @samp{-malign-double} switch,
6246 structures containing the above types will be aligned differently than
6247 the published application binary interface specifications for the 386.
6250 @itemx -mno-svr3-shlib
6251 Control whether GCC places uninitialized locals into @code{bss} or
6252 @code{data}. @samp{-msvr3-shlib} places these locals into @code{bss}.
6253 These options are meaningful only on System V Release 3.
6255 @item -mno-wide-multiply
6256 @itemx -mwide-multiply
6257 Control whether GCC uses the @code{mul} and @code{imul} that produce
6258 64 bit results in @code{eax:edx} from 32 bit operands to do @code{long
6259 long} multiplies and 32-bit division by constants.
6262 Use a different function-calling convention, in which functions that
6263 take a fixed number of arguments return with the @code{ret} @var{num}
6264 instruction, which pops their arguments while returning. This saves one
6265 instruction in the caller since there is no need to pop the arguments
6268 You can specify that an individual function is called with this calling
6269 sequence with the function attribute @samp{stdcall}. You can also
6270 override the @samp{-mrtd} option by using the function attribute
6271 @samp{cdecl}. @xref{Function Attributes}.
6273 @strong{Warning:} this calling convention is incompatible with the one
6274 normally used on Unix, so you cannot use it if you need to call
6275 libraries compiled with the Unix compiler.
6277 Also, you must provide function prototypes for all functions that
6278 take variable numbers of arguments (including @code{printf});
6279 otherwise incorrect code will be generated for calls to those
6282 In addition, seriously incorrect code will result if you call a
6283 function with too many arguments. (Normally, extra arguments are
6284 harmlessly ignored.)
6286 @item -mreg-alloc=@var{regs}
6287 Control the default allocation order of integer registers. The
6288 string @var{regs} is a series of letters specifying a register. The
6289 supported letters are: @code{a} allocate EAX; @code{b} allocate EBX;
6290 @code{c} allocate ECX; @code{d} allocate EDX; @code{S} allocate ESI;
6291 @code{D} allocate EDI; @code{B} allocate EBP.
6293 @item -mregparm=@var{num}
6294 Control how many registers are used to pass integer arguments. By
6295 default, no registers are used to pass arguments, and at most 3
6296 registers can be used. You can control this behavior for a specific
6297 function by using the function attribute @samp{regparm}.
6298 @xref{Function Attributes}.
6300 @strong{Warning:} if you use this switch, and
6301 @var{num} is nonzero, then you must build all modules with the same
6302 value, including any libraries. This includes the system libraries and
6305 @item -malign-loops=@var{num}
6306 Align loops to a 2 raised to a @var{num} byte boundary. If
6307 @samp{-malign-loops} is not specified, the default is 2 unless
6308 gas 2.8 (or later) is being used in which case the default is
6309 to align the loop on a 16 byte boundary if it is less than 8
6312 @item -malign-jumps=@var{num}
6313 Align instructions that are only jumped to to a 2 raised to a @var{num}
6314 byte boundary. If @samp{-malign-jumps} is not specified, the default is
6315 2 if optimizing for a 386, and 4 if optimizing for a 486 unless
6316 gas 2.8 (or later) is being used in which case the default is
6317 to align the instruction on a 16 byte boundary if it is less
6320 @item -malign-functions=@var{num}
6321 Align the start of functions to a 2 raised to @var{num} byte boundary.
6322 If @samp{-malign-functions} is not specified, the default is 2 if optimizing
6323 for a 386, and 4 if optimizing for a 486.
6325 @item -mpreferred-stack-boundary=@var{num}
6326 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
6327 byte boundary. If @samp{-mpreferred-stack-boundary} is not specified,
6328 the default is 4 (16 bytes or 128 bits).
6330 The stack is required to be aligned on a 4 byte boundary. On Pentium
6331 and PentiumPro, @code{double} and @code{long double} values should be
6332 aligned to an 8 byte boundary (see @samp{-malign-double}) or suffer
6333 significant run time performance penalties. On Pentium III, the
6334 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
6335 penalties if it is not 16 byte aligned.
6337 To ensure proper alignment of this values on the stack, the stack boundary
6338 must be as aligned as that required by any value stored on the stack.
6339 Further, every function must be generated such that it keeps the stack
6340 aligned. Thus calling a function compiled with a higher preferred
6341 stack boundary from a function compiled with a lower preferred stack
6342 boundary will most likely misalign the stack. It is recommended that
6343 libraries that use callbacks always use the default setting.
6345 This extra alignment does consume extra stack space. Code that is sensitive
6346 to stack space usage, such as embedded systems and operating system kernels,
6347 may want to reduce the preferred alignment to
6348 @samp{-mpreferred-stack-boundary=2}.
6352 Use PUSH operations to store outgoing parameters. This method is shorter
6353 and usually equally fast as method using SUB/MOV operations and is enabled
6354 by default. In some cases disabling it may improve performance because of
6355 improved scheduling and reduced dependencies.
6357 @item -maccumulate-outgoing-args
6358 @kindex -maccumulate-outgoing-args
6359 If enabled, the maximum amount of space required for outgoing arguments will be
6360 computed in the function prologue. This in faster on most modern CPUs
6361 because of reduced dependencies, improved scheduling and reduced stack usage
6362 when preferred stack boundary is not equal to 2. The drawback is a notable
6363 increase in code size. This switch implies -mno-push-args.
6367 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
6368 on thread-safe exception handling must compile and link all code with the
6369 @samp{-mthreads} option. When compiling, @samp{-mthreads} defines
6370 @samp{-D_MT}; when linking, it links in a special thread helper library
6371 @samp{-lmingwthrd} which cleans up per thread exception handling data.
6373 @item -mno-align-stringops
6374 @kindex -mno-align-stringops
6375 Do not align destination of inlined string operations. This switch reduces
6376 code size and improves performance in case the destination is already aligned,
6377 but gcc don't know about it.
6379 @item -minline-all-stringops
6380 @kindex -minline-all-stringops
6381 By default GCC inlines string operations only when destination is known to be
6382 aligned at least to 4 byte boundary. This enables more inlining, increase code
6383 size, but may improve performance of code that depends on fast memcpy, strlen
6384 and memset for short lengths.
6388 @subsection HPPA Options
6389 @cindex HPPA Options
6391 These @samp{-m} options are defined for the HPPA family of computers:
6394 @item -march=@var{architecture type}
6395 Generate code for the specified architecture. The choices for
6396 @var{architecture type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
6397 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
6398 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
6399 architecture option for your machine. Code compiled for lower numbered
6400 architectures will run on higher numbered architectures, but not the
6403 PA 2.0 support currently requires gas snapshot 19990413 or later. The
6404 next release of binutils (current is 2.9.1) will probably contain PA 2.0
6408 @itemx -mpa-risc-1-1
6409 @itemx -mpa-risc-2-0
6410 Synonyms for -march=1.0, -march=1.1, and -march=2.0 respectively.
6413 Generate code suitable for big switch tables. Use this option only if
6414 the assembler/linker complain about out of range branches within a switch
6417 @item -mjump-in-delay
6418 Fill delay slots of function calls with unconditional jump instructions
6419 by modifying the return pointer for the function call to be the target
6420 of the conditional jump.
6422 @item -mdisable-fpregs
6423 Prevent floating point registers from being used in any manner. This is
6424 necessary for compiling kernels which perform lazy context switching of
6425 floating point registers. If you use this option and attempt to perform
6426 floating point operations, the compiler will abort.
6428 @item -mdisable-indexing
6429 Prevent the compiler from using indexing address modes. This avoids some
6430 rather obscure problems when compiling MIG generated code under MACH.
6432 @item -mno-space-regs
6433 Generate code that assumes the target has no space registers. This allows
6434 GCC to generate faster indirect calls and use unscaled index address modes.
6436 Such code is suitable for level 0 PA systems and kernels.
6438 @item -mfast-indirect-calls
6439 Generate code that assumes calls never cross space boundaries. This
6440 allows GCC to emit code which performs faster indirect calls.
6442 This option will not work in the presence of shared libraries or nested
6445 @item -mlong-load-store
6446 Generate 3-instruction load and store sequences as sometimes required by
6447 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
6450 @item -mportable-runtime
6451 Use the portable calling conventions proposed by HP for ELF systems.
6454 Enable the use of assembler directives only GAS understands.
6456 @item -mschedule=@var{cpu type}
6457 Schedule code according to the constraints for the machine type
6458 @var{cpu type}. The choices for @var{cpu type} are @samp{700}
6459 @samp{7100}, @samp{7100LC}, @samp{7200}, and @samp{8000}. Refer to
6460 @file{/usr/lib/sched.models} on an HP-UX system to determine the
6461 proper scheduling option for your machine.
6464 Enable the optimization pass in the HPUX linker. Note this makes symbolic
6465 debugging impossible. It also triggers a bug in the HPUX 8 and HPUX 9 linkers
6466 in which they give bogus error messages when linking some programs.
6469 Generate output containing library calls for floating point.
6470 @strong{Warning:} the requisite libraries are not available for all HPPA
6471 targets. Normally the facilities of the machine's usual C compiler are
6472 used, but this cannot be done directly in cross-compilation. You must make
6473 your own arrangements to provide suitable library functions for
6474 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
6475 does provide software floating point support.
6477 @samp{-msoft-float} changes the calling convention in the output file;
6478 therefore, it is only useful if you compile @emph{all} of a program with
6479 this option. In particular, you need to compile @file{libgcc.a}, the
6480 library that comes with GCC, with @samp{-msoft-float} in order for
6484 @node Intel 960 Options
6485 @subsection Intel 960 Options
6487 These @samp{-m} options are defined for the Intel 960 implementations:
6490 @item -m@var{cpu type}
6491 Assume the defaults for the machine type @var{cpu type} for some of
6492 the other options, including instruction scheduling, floating point
6493 support, and addressing modes. The choices for @var{cpu type} are
6494 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
6495 @samp{sa}, and @samp{sb}.
6501 The @samp{-mnumerics} option indicates that the processor does support
6502 floating-point instructions. The @samp{-msoft-float} option indicates
6503 that floating-point support should not be assumed.
6505 @item -mleaf-procedures
6506 @itemx -mno-leaf-procedures
6507 Do (or do not) attempt to alter leaf procedures to be callable with the
6508 @code{bal} instruction as well as @code{call}. This will result in more
6509 efficient code for explicit calls when the @code{bal} instruction can be
6510 substituted by the assembler or linker, but less efficient code in other
6511 cases, such as calls via function pointers, or using a linker that doesn't
6512 support this optimization.
6515 @itemx -mno-tail-call
6516 Do (or do not) make additional attempts (beyond those of the
6517 machine-independent portions of the compiler) to optimize tail-recursive
6518 calls into branches. You may not want to do this because the detection of
6519 cases where this is not valid is not totally complete. The default is
6520 @samp{-mno-tail-call}.
6522 @item -mcomplex-addr
6523 @itemx -mno-complex-addr
6524 Assume (or do not assume) that the use of a complex addressing mode is a
6525 win on this implementation of the i960. Complex addressing modes may not
6526 be worthwhile on the K-series, but they definitely are on the C-series.
6527 The default is currently @samp{-mcomplex-addr} for all processors except
6531 @itemx -mno-code-align
6532 Align code to 8-byte boundaries for faster fetching (or don't bother).
6533 Currently turned on by default for C-series implementations only.
6536 @item -mclean-linkage
6537 @itemx -mno-clean-linkage
6538 These options are not fully implemented.
6542 @itemx -mic2.0-compat
6543 @itemx -mic3.0-compat
6544 Enable compatibility with iC960 v2.0 or v3.0.
6548 Enable compatibility with the iC960 assembler.
6550 @item -mstrict-align
6551 @itemx -mno-strict-align
6552 Do not permit (do permit) unaligned accesses.
6555 Enable structure-alignment compatibility with Intel's gcc release version
6556 1.3 (based on gcc 1.37). This option implies @samp{-mstrict-align}.
6558 @item -mlong-double-64
6559 Implement type @samp{long double} as 64-bit floating point numbers.
6560 Without the option @samp{long double} is implemented by 80-bit
6561 floating point numbers. The only reason we have it because there is
6562 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
6563 is only useful for people using soft-float targets. Otherwise, we
6564 should recommend against use of it.
6568 @node DEC Alpha Options
6569 @subsection DEC Alpha Options
6571 These @samp{-m} options are defined for the DEC Alpha implementations:
6574 @item -mno-soft-float
6576 Use (do not use) the hardware floating-point instructions for
6577 floating-point operations. When @code{-msoft-float} is specified,
6578 functions in @file{libgcc1.c} will be used to perform floating-point
6579 operations. Unless they are replaced by routines that emulate the
6580 floating-point operations, or compiled in such a way as to call such
6581 emulations routines, these routines will issue floating-point
6582 operations. If you are compiling for an Alpha without floating-point
6583 operations, you must ensure that the library is built so as not to call
6586 Note that Alpha implementations without floating-point operations are
6587 required to have floating-point registers.
6591 Generate code that uses (does not use) the floating-point register set.
6592 @code{-mno-fp-regs} implies @code{-msoft-float}. If the floating-point
6593 register set is not used, floating point operands are passed in integer
6594 registers as if they were integers and floating-point results are passed
6595 in $0 instead of $f0. This is a non-standard calling sequence, so any
6596 function with a floating-point argument or return value called by code
6597 compiled with @code{-mno-fp-regs} must also be compiled with that
6600 A typical use of this option is building a kernel that does not use,
6601 and hence need not save and restore, any floating-point registers.
6604 The Alpha architecture implements floating-point hardware optimized for
6605 maximum performance. It is mostly compliant with the IEEE floating
6606 point standard. However, for full compliance, software assistance is
6607 required. This option generates code fully IEEE compliant code
6608 @emph{except} that the @var{inexact flag} is not maintained (see below).
6609 If this option is turned on, the CPP macro @code{_IEEE_FP} is defined
6610 during compilation. The option is a shorthand for: @samp{-D_IEEE_FP
6611 -mfp-trap-mode=su -mtrap-precision=i -mieee-conformant}. The resulting
6612 code is less efficient but is able to correctly support denormalized
6613 numbers and exceptional IEEE values such as not-a-number and plus/minus
6614 infinity. Other Alpha compilers call this option
6615 @code{-ieee_with_no_inexact}.
6617 @item -mieee-with-inexact
6618 @c overfull hbox here --bob 22 jul96
6619 @c original text between ignore ... end ignore
6621 This is like @samp{-mieee} except the generated code also maintains the
6622 IEEE @var{inexact flag}. Turning on this option causes the generated
6623 code to implement fully-compliant IEEE math. The option is a shorthand
6624 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus @samp{-mieee-conformant},
6625 @samp{-mfp-trap-mode=sui}, and @samp{-mtrap-precision=i}. On some Alpha
6626 implementations the resulting code may execute significantly slower than
6627 the code generated by default. Since there is very little code that
6628 depends on the @var{inexact flag}, you should normally not specify this
6629 option. Other Alpha compilers call this option
6630 @samp{-ieee_with_inexact}.
6632 @c changed paragraph
6633 This is like @samp{-mieee} except the generated code also maintains the
6634 IEEE @var{inexact flag}. Turning on this option causes the generated
6635 code to implement fully-compliant IEEE math. The option is a shorthand
6636 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus the three following:
6637 @samp{-mieee-conformant},
6638 @samp{-mfp-trap-mode=sui},
6639 and @samp{-mtrap-precision=i}.
6640 On some Alpha implementations the resulting code may execute
6641 significantly slower than the code generated by default. Since there
6642 is very little code that depends on the @var{inexact flag}, you should
6643 normally not specify this option. Other Alpha compilers call this
6644 option @samp{-ieee_with_inexact}.
6645 @c end changes to prevent overfull hboxes
6647 @item -mfp-trap-mode=@var{trap mode}
6648 This option controls what floating-point related traps are enabled.
6649 Other Alpha compilers call this option @samp{-fptm }@var{trap mode}.
6650 The trap mode can be set to one of four values:
6654 This is the default (normal) setting. The only traps that are enabled
6655 are the ones that cannot be disabled in software (e.g., division by zero
6659 In addition to the traps enabled by @samp{n}, underflow traps are enabled
6663 Like @samp{su}, but the instructions are marked to be safe for software
6664 completion (see Alpha architecture manual for details).
6667 Like @samp{su}, but inexact traps are enabled as well.
6670 @item -mfp-rounding-mode=@var{rounding mode}
6671 Selects the IEEE rounding mode. Other Alpha compilers call this option
6672 @samp{-fprm }@var{rounding mode}. The @var{rounding mode} can be one
6677 Normal IEEE rounding mode. Floating point numbers are rounded towards
6678 the nearest machine number or towards the even machine number in case
6682 Round towards minus infinity.
6685 Chopped rounding mode. Floating point numbers are rounded towards zero.
6688 Dynamic rounding mode. A field in the floating point control register
6689 (@var{fpcr}, see Alpha architecture reference manual) controls the
6690 rounding mode in effect. The C library initializes this register for
6691 rounding towards plus infinity. Thus, unless your program modifies the
6692 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
6695 @item -mtrap-precision=@var{trap precision}
6696 In the Alpha architecture, floating point traps are imprecise. This
6697 means without software assistance it is impossible to recover from a
6698 floating trap and program execution normally needs to be terminated.
6699 GCC can generate code that can assist operating system trap handlers
6700 in determining the exact location that caused a floating point trap.
6701 Depending on the requirements of an application, different levels of
6702 precisions can be selected:
6706 Program precision. This option is the default and means a trap handler
6707 can only identify which program caused a floating point exception.
6710 Function precision. The trap handler can determine the function that
6711 caused a floating point exception.
6714 Instruction precision. The trap handler can determine the exact
6715 instruction that caused a floating point exception.
6718 Other Alpha compilers provide the equivalent options called
6719 @samp{-scope_safe} and @samp{-resumption_safe}.
6721 @item -mieee-conformant
6722 This option marks the generated code as IEEE conformant. You must not
6723 use this option unless you also specify @samp{-mtrap-precision=i} and either
6724 @samp{-mfp-trap-mode=su} or @samp{-mfp-trap-mode=sui}. Its only effect
6725 is to emit the line @samp{.eflag 48} in the function prologue of the
6726 generated assembly file. Under DEC Unix, this has the effect that
6727 IEEE-conformant math library routines will be linked in.
6729 @item -mbuild-constants
6730 Normally GCC examines a 32- or 64-bit integer constant to
6731 see if it can construct it from smaller constants in two or three
6732 instructions. If it cannot, it will output the constant as a literal and
6733 generate code to load it from the data segment at runtime.
6735 Use this option to require GCC to construct @emph{all} integer constants
6736 using code, even if it takes more instructions (the maximum is six).
6738 You would typically use this option to build a shared library dynamic
6739 loader. Itself a shared library, it must relocate itself in memory
6740 before it can find the variables and constants in its own data segment.
6744 Select whether to generate code to be assembled by the vendor-supplied
6745 assembler (@samp{-malpha-as}) or by the GNU assembler @samp{-mgas}.
6753 Indicate whether GCC should generate code to use the optional BWX,
6754 CIX, and MAX instruction sets. The default is to use the instruction sets
6755 supported by the CPU type specified via @samp{-mcpu=} option or that
6756 of the CPU on which GCC was built if none was specified.
6758 @item -mcpu=@var{cpu_type}
6759 Set the instruction set, register set, and instruction scheduling
6760 parameters for machine type @var{cpu_type}. You can specify either the
6761 @samp{EV} style name or the corresponding chip number. GCC
6762 supports scheduling parameters for the EV4 and EV5 family of processors
6763 and will choose the default values for the instruction set from
6764 the processor you specify. If you do not specify a processor type,
6765 GCC will default to the processor on which the compiler was built.
6767 Supported values for @var{cpu_type} are
6772 Schedules as an EV4 and has no instruction set extensions.
6776 Schedules as an EV5 and has no instruction set extensions.
6780 Schedules as an EV5 and supports the BWX extension.
6785 Schedules as an EV5 and supports the BWX and MAX extensions.
6789 Schedules as an EV5 (until Digital releases the scheduling parameters
6790 for the EV6) and supports the BWX, CIX, and MAX extensions.
6793 @item -mmemory-latency=@var{time}
6794 Sets the latency the scheduler should assume for typical memory
6795 references as seen by the application. This number is highly
6796 dependent on the memory access patterns used by the application
6797 and the size of the external cache on the machine.
6799 Valid options for @var{time} are
6803 A decimal number representing clock cycles.
6809 The compiler contains estimates of the number of clock cycles for
6810 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
6811 (also called Dcache, Scache, and Bcache), as well as to main memory.
6812 Note that L3 is only valid for EV5.
6817 @node Clipper Options
6818 @subsection Clipper Options
6820 These @samp{-m} options are defined for the Clipper implementations:
6824 Produce code for a C300 Clipper processor. This is the default.
6827 Produce code for a C400 Clipper processor i.e. use floating point
6831 @node H8/300 Options
6832 @subsection H8/300 Options
6834 These @samp{-m} options are defined for the H8/300 implementations:
6838 Shorten some address references at link time, when possible; uses the
6839 linker option @samp{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
6840 ld.info, Using ld}, for a fuller description.
6843 Generate code for the H8/300H.
6846 Generate code for the H8/S.
6849 Generate code for the H8/S2600. This switch must be used with -ms.
6852 Make @code{int} data 32 bits by default.
6855 On the H8/300H and H8/S, use the same alignment rules as for the H8/300.
6856 The default for the H8/300H and H8/S is to align longs and floats on 4
6858 @samp{-malign-300} causes them to be aligned on 2 byte boundaries.
6859 This option has no effect on the H8/300.
6863 @subsection SH Options
6865 These @samp{-m} options are defined for the SH implementations:
6869 Generate code for the SH1.
6872 Generate code for the SH2.
6875 Generate code for the SH3.
6878 Generate code for the SH3e.
6881 Generate code for the SH4 without a floating-point unit.
6883 @item -m4-single-only
6884 Generate code for the SH4 with a floating-point unit that only
6885 supports single-precision arithmentic.
6888 Generate code for the SH4 assuming the floating-point unit is in
6889 single-precision mode by default.
6892 Generate code for the SH4.
6895 Compile code for the processor in big endian mode.
6898 Compile code for the processor in little endian mode.
6901 Align doubles at 64 bit boundaries. Note that this changes the calling
6902 conventions, and thus some functions from the standard C library will
6903 not work unless you recompile it first with -mdalign.
6906 Shorten some address references at link time, when possible; uses the
6907 linker option @samp{-relax}.
6910 Use 32-bit offsets in @code{switch} tables. The default is to use
6914 Enable the use of the instruction @code{fmovd}.
6917 Comply with the calling conventions defined by Hitachi.
6920 Mark the @code{MAC} register as call-clobbered, even if
6921 @code{-mhitachi} is given.
6924 Dump instruction size and location in the assembly code.
6927 This option is deprecated. It pads structures to multiple of 4 bytes,
6928 which is incompatible with the SH ABI.
6931 Optimize for space instead of speed. Implied by @code{-Os}.
6934 When generating position-independent code, emit function calls using
6935 the Global Offset Table instead of the Procedure Linkage Table.
6938 Generate a library function call to invalidate instruction cache
6939 entries, after fixing up a trampoline. This library function call
6940 doesn't assume it can write to the whole memory address space. This
6941 is the default when the target is @code{sh-*-linux*}.
6944 @node System V Options
6945 @subsection Options for System V
6947 These additional options are available on System V Release 4 for
6948 compatibility with other compilers on those systems:
6952 Create a shared object.
6953 It is recommended that @samp{-symbolic} or @samp{-shared} be used instead.
6956 Identify the versions of each tool used by the compiler, in a
6957 @code{.ident} assembler directive in the output.
6960 Refrain from adding @code{.ident} directives to the output file (this is
6963 @item -YP,@var{dirs}
6964 Search the directories @var{dirs}, and no others, for libraries
6965 specified with @samp{-l}.
6968 Look in the directory @var{dir} to find the M4 preprocessor.
6969 The assembler uses this option.
6970 @c This is supposed to go with a -Yd for predefined M4 macro files, but
6971 @c the generic assembler that comes with Solaris takes just -Ym.
6974 @node TMS320C3x/C4x Options
6975 @subsection TMS320C3x/C4x Options
6976 @cindex TMS320C3x/C4x Options
6978 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
6982 @item -mcpu=@var{cpu_type}
6983 Set the instruction set, register set, and instruction scheduling
6984 parameters for machine type @var{cpu_type}. Supported values for
6985 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
6986 @samp{c44}. The default is @samp{c40} to generate code for the
6991 @itemx -msmall-memory
6993 Generates code for the big or small memory model. The small memory
6994 model assumed that all data fits into one 64K word page. At run-time
6995 the data page (DP) register must be set to point to the 64K page
6996 containing the .bss and .data program sections. The big memory model is
6997 the default and requires reloading of the DP register for every direct
7002 Allow (disallow) allocation of general integer operands into the block
7007 Enable (disable) generation of code using decrement and branch,
7008 DBcond(D), instructions. This is enabled by default for the C4x. To be
7009 on the safe side, this is disabled for the C3x, since the maximum
7010 iteration count on the C3x is 2^23 + 1 (but who iterates loops more than
7011 2^23 times on the C3x?). Note that GCC will try to reverse a loop so
7012 that it can utilise the decrement and branch instruction, but will give
7013 up if there is more than one memory reference in the loop. Thus a loop
7014 where the loop counter is decremented can generate slightly more
7015 efficient code, in cases where the RPTB instruction cannot be utilised.
7017 @item -mdp-isr-reload
7019 Force the DP register to be saved on entry to an interrupt service
7020 routine (ISR), reloaded to point to the data section, and restored on
7021 exit from the ISR. This should not be required unless someone has
7022 violated the small memory model by modifying the DP register, say within
7027 For the C3x use the 24-bit MPYI instruction for integer multiplies
7028 instead of a library call to guarantee 32-bit results. Note that if one
7029 of the operands is a constant, then the multiplication will be performed
7030 using shifts and adds. If the -mmpyi option is not specified for the C3x,
7031 then squaring operations are performed inline instead of a library call.
7034 @itemx -mno-fast-fix
7035 The C3x/C4x FIX instruction to convert a floating point value to an
7036 integer value chooses the nearest integer less than or equal to the
7037 floating point value rather than to the nearest integer. Thus if the
7038 floating point number is negative, the result will be incorrectly
7039 truncated an additional code is necessary to detect and correct this
7040 case. This option can be used to disable generation of the additional
7041 code required to correct the result.
7045 Enable (disable) generation of repeat block sequences using the RPTB
7046 instruction for zero overhead looping. The RPTB construct is only used
7047 for innermost loops that do not call functions or jump across the loop
7048 boundaries. There is no advantage having nested RPTB loops due to the
7049 overhead required to save and restore the RC, RS, and RE registers.
7050 This is enabled by default with -O2.
7052 @item -mrpts=@var{count}
7054 Enable (disable) the use of the single instruction repeat instruction
7055 RPTS. If a repeat block contains a single instruction, and the loop
7056 count can be guaranteed to be less than the value @var{count}, GCC will
7057 emit a RPTS instruction instead of a RPTB. If no value is specified,
7058 then a RPTS will be emitted even if the loop count cannot be determined
7059 at compile time. Note that the repeated instruction following RPTS does
7060 not have to be reloaded from memory each iteration, thus freeing up the
7061 CPU buses for operands. However, since interrupts are blocked by this
7062 instruction, it is disabled by default.
7064 @item -mloop-unsigned
7065 @itemx -mno-loop-unsigned
7066 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
7067 is 2^31 + 1 since these instructions test if the iteration count is
7068 negative to terminate the loop. If the iteration count is unsigned
7069 there is a possibility than the 2^31 + 1 maximum iteration count may be
7070 exceeded. This switch allows an unsigned iteration count.
7073 Try to emit an assembler syntax that the TI assembler (asm30) is happy
7074 with. This also enforces compatibility with the API employed by the TI
7075 C3x C compiler. For example, long doubles are passed as structures
7076 rather than in floating point registers.
7080 Generate code that uses registers (stack) for passing arguments to functions.
7081 By default, arguments are passed in registers where possible rather
7082 than by pushing arguments on to the stack.
7084 @item -mparallel-insns
7085 @itemx -mno-parallel-insns
7086 Allow the generation of parallel instructions. This is enabled by
7089 @item -mparallel-mpy
7090 @itemx -mno-parallel-mpy
7091 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
7092 provided -mparallel-insns is also specified. These instructions have
7093 tight register constraints which can pessimize the code generation
7099 @subsection V850 Options
7100 @cindex V850 Options
7102 These @samp{-m} options are defined for V850 implementations:
7106 @itemx -mno-long-calls
7107 Treat all calls as being far away (near). If calls are assumed to be
7108 far away, the compiler will always load the functions address up into a
7109 register, and call indirect through the pointer.
7113 Do not optimize (do optimize) basic blocks that use the same index
7114 pointer 4 or more times to copy pointer into the @code{ep} register, and
7115 use the shorter @code{sld} and @code{sst} instructions. The @samp{-mep}
7116 option is on by default if you optimize.
7118 @item -mno-prolog-function
7119 @itemx -mprolog-function
7120 Do not use (do use) external functions to save and restore registers at
7121 the prolog and epilog of a function. The external functions are slower,
7122 but use less code space if more than one function saves the same number
7123 of registers. The @samp{-mprolog-function} option is on by default if
7127 Try to make the code as small as possible. At present, this just turns
7128 on the @samp{-mep} and @samp{-mprolog-function} options.
7131 Put static or global variables whose size is @var{n} bytes or less into
7132 the tiny data area that register @code{ep} points to. The tiny data
7133 area can hold up to 256 bytes in total (128 bytes for byte references).
7136 Put static or global variables whose size is @var{n} bytes or less into
7137 the small data area that register @code{gp} points to. The small data
7138 area can hold up to 64 kilobytes.
7141 Put static or global variables whose size is @var{n} bytes or less into
7142 the first 32 kilobytes of memory.
7145 Specify that the target processor is the V850.
7148 Generate code suitable for big switch tables. Use this option only if
7149 the assembler/linker complain about out of range branches within a switch
7154 @subsection ARC Options
7157 These options are defined for ARC implementations:
7161 Compile code for little endian mode. This is the default.
7164 Compile code for big endian mode.
7167 Prepend the name of the cpu to all public symbol names.
7168 In multiple-processor systems, there are many ARC variants with different
7169 instruction and register set characteristics. This flag prevents code
7170 compiled for one cpu to be linked with code compiled for another.
7171 No facility exists for handling variants that are "almost identical".
7172 This is an all or nothing option.
7174 @item -mcpu=@var{cpu}
7175 Compile code for ARC variant @var{cpu}.
7176 Which variants are supported depend on the configuration.
7177 All variants support @samp{-mcpu=base}, this is the default.
7179 @item -mtext=@var{text section}
7180 @itemx -mdata=@var{data section}
7181 @itemx -mrodata=@var{readonly data section}
7182 Put functions, data, and readonly data in @var{text section},
7183 @var{data section}, and @var{readonly data section} respectively
7184 by default. This can be overridden with the @code{section} attribute.
7185 @xref{Variable Attributes}.
7190 @subsection NS32K Options
7191 @cindex NS32K options
7193 These are the @samp{-m} options defined for the 32000 series. The default
7194 values for these options depends on which style of 32000 was selected when
7195 the compiler was configured; the defaults for the most common choices are
7201 Generate output for a 32032. This is the default
7202 when the compiler is configured for 32032 and 32016 based systems.
7206 Generate output for a 32332. This is the default
7207 when the compiler is configured for 32332-based systems.
7211 Generate output for a 32532. This is the default
7212 when the compiler is configured for 32532-based systems.
7215 Generate output containing 32081 instructions for floating point.
7216 This is the default for all systems.
7219 Generate output containing 32381 instructions for floating point. This
7220 also implies @samp{-m32081}. The 32381 is only compatible with the 32332
7221 and 32532 cpus. This is the default for the pc532-netbsd configuration.
7224 Try and generate multiply-add floating point instructions @code{polyF}
7225 and @code{dotF}. This option is only available if the @samp{-m32381}
7226 option is in effect. Using these instructions requires changes to to
7227 register allocation which generally has a negative impact on
7228 performance. This option should only be enabled when compiling code
7229 particularly likely to make heavy use of multiply-add instructions.
7232 Do not try and generate multiply-add floating point instructions
7233 @code{polyF} and @code{dotF}. This is the default on all platforms.
7236 Generate output containing library calls for floating point.
7237 @strong{Warning:} the requisite libraries may not be available.
7240 Do not use the bit-field instructions. On some machines it is faster to
7241 use shifting and masking operations. This is the default for the pc532.
7244 Do use the bit-field instructions. This is the default for all platforms
7248 Use a different function-calling convention, in which functions
7249 that take a fixed number of arguments return pop their
7250 arguments on return with the @code{ret} instruction.
7252 This calling convention is incompatible with the one normally
7253 used on Unix, so you cannot use it if you need to call libraries
7254 compiled with the Unix compiler.
7256 Also, you must provide function prototypes for all functions that
7257 take variable numbers of arguments (including @code{printf});
7258 otherwise incorrect code will be generated for calls to those
7261 In addition, seriously incorrect code will result if you call a
7262 function with too many arguments. (Normally, extra arguments are
7263 harmlessly ignored.)
7265 This option takes its name from the 680x0 @code{rtd} instruction.
7269 Use a different function-calling convention where the first two arguments
7270 are passed in registers.
7272 This calling convention is incompatible with the one normally
7273 used on Unix, so you cannot use it if you need to call libraries
7274 compiled with the Unix compiler.
7277 Do not pass any arguments in registers. This is the default for all
7281 It is OK to use the sb as an index register which is always loaded with
7282 zero. This is the default for the pc532-netbsd target.
7285 The sb register is not available for use or has not been initialized to
7286 zero by the run time system. This is the default for all targets except
7287 the pc532-netbsd. It is also implied whenever @samp{-mhimem} or
7288 @samp{-fpic} is set.
7291 Many ns32000 series addressing modes use displacements of up to 512MB.
7292 If an address is above 512MB then displacements from zero can not be used.
7293 This option causes code to be generated which can be loaded above 512MB.
7294 This may be useful for operating systems or ROM code.
7297 Assume code will be loaded in the first 512MB of virtual address space.
7298 This is the default for all platforms.
7304 @subsection AVR Options
7307 These options are defined for AVR implementations:
7310 @item -mmcu=@var{mcu}
7311 Specify ATMEL AVR mcu (at90s23xx,attiny22,at90s44xx,at90s85xx,atmega603,
7315 Output instruction size's to the asm file
7317 @item -minit-stack=@var{N}
7318 Specify the initial stack address
7320 @item -mno-interrupts
7321 Generated code is not compatible with hardware interrupts.
7322 Code size will be smaller.
7324 @item -mcall-prologues
7325 Functions prologues/epilogues expanded as call to appropriate
7326 subroutines. Code size will be smaller.
7330 @subsection MCore Options
7331 @cindex MCore options
7333 These are the @samp{-m} options defined for the Motorola M*Core
7341 Inline constants into the code stream if it can be done in two
7342 instructions or less.
7347 Use the divide instruction. (Enabled by default).
7349 @item -mrelax-immediate
7350 @itemx -mrelax-immediate
7351 @itemx -mno-relax-immediate
7352 Allow arbitrary sized immediates in bit operations.
7354 @item -mwide-bitfields
7355 @itemx -mwide-bitfields
7356 @itemx -mno-wide-bitfields
7357 Always treat bitfields as int-sized.
7359 @item -m4byte-functions
7360 @itemx -m4byte-functions
7361 @itemx -mno-4byte-functions
7362 Force all functions to be aligned to a four byte boundary.
7364 @item -mcallgraph-data
7365 @itemx -mcallgraph-data
7366 @itemx -mno-callgraph-data
7367 Emit callgraph information.
7371 @itemx -mno-slow-bytes
7372 Prefer word access when reading byte quantities.
7374 @item -mlittle-endian
7375 @itemx -mlittle-endian
7377 Generate code for a little endian target.
7382 Generate code for the 210 processor.
7386 @subsection D30V Options
7387 @cindex D30V Options
7389 These @samp{-m} options are defined for D30V implementations:
7393 Link the @samp{.text}, @samp{.data}, @samp{.bss}, @samp{.strings},
7394 @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections into external
7395 memory, which starts at location @code{0x80000000}.
7398 Same as the @samp{-mextmem} switch.
7401 Link the @samp{.text} section into onchip text memory, which starts at
7402 location @code{0x0}. Also link @samp{.data}, @samp{.bss},
7403 @samp{.strings}, @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections
7404 into onchip data memory, which starts at location @code{0x20000000}.
7406 @item -mno-asm-optimize
7407 @itemx -masm-optimize
7408 Disable (enable) passing @samp{-O} to the assembler when optimizing.
7409 The assembler uses the @samp{-O} option to automatically parallelize
7410 adjacent short instructions where possible.
7412 @item -mbranch-cost=@var{n}
7413 Increase the internal costs of branches to @var{n}. Higher costs means
7414 that the compiler will issue more instructions to avoid doing a branch.
7417 @item -mcond-exec=@var{n}
7418 Specify the maximum number of conditionally executed instructions that
7419 replace a branch. The default is 4.
7422 @node Code Gen Options
7423 @section Options for Code Generation Conventions
7424 @cindex code generation conventions
7425 @cindex options, code generation
7426 @cindex run-time options
7428 These machine-independent options control the interface conventions
7429 used in code generation.
7431 Most of them have both positive and negative forms; the negative form
7432 of @samp{-ffoo} would be @samp{-fno-foo}. In the table below, only
7433 one of the forms is listed---the one which is not the default. You
7434 can figure out the other form by either removing @samp{no-} or adding
7439 Enable exception handling. Generates extra code needed to propagate
7440 exceptions. For some targets, this implies GNU CC will generate frame
7441 unwind information for all functions, which can produce significant data
7442 size overhead, although it does not affect execution. If you do not
7443 specify this option, GNU CC will enable it by default for languages like
7444 C++ which normally require exception handling, and disable itfor
7445 languages like C that do not normally require it. However, you may need
7446 to enable this option when compiling C code that needs to interoperate
7447 properly with exception handlers written in C++. You may also wish to
7448 disable this option if you are compiling older C++ programs that don't
7449 use exception handling.
7451 @item -funwind-tables
7452 Similar to @code{-fexceptions}, except that it will just generate any needed
7453 static data, but will not affect the generated code in any other way.
7454 You will normally not enable this option; instead, a language processor
7455 that needs this handling would enable it on your behalf.
7457 @item -fpcc-struct-return
7458 Return ``short'' @code{struct} and @code{union} values in memory like
7459 longer ones, rather than in registers. This convention is less
7460 efficient, but it has the advantage of allowing intercallability between
7461 GCC-compiled files and files compiled with other compilers.
7463 The precise convention for returning structures in memory depends
7464 on the target configuration macros.
7466 Short structures and unions are those whose size and alignment match
7467 that of some integer type.
7469 @item -freg-struct-return
7470 Use the convention that @code{struct} and @code{union} values are
7471 returned in registers when possible. This is more efficient for small
7472 structures than @samp{-fpcc-struct-return}.
7474 If you specify neither @samp{-fpcc-struct-return} nor its contrary
7475 @samp{-freg-struct-return}, GCC defaults to whichever convention is
7476 standard for the target. If there is no standard convention, GCC
7477 defaults to @samp{-fpcc-struct-return}, except on targets where GCC
7478 is the principal compiler. In those cases, we can choose the standard,
7479 and we chose the more efficient register return alternative.
7482 Allocate to an @code{enum} type only as many bytes as it needs for the
7483 declared range of possible values. Specifically, the @code{enum} type
7484 will be equivalent to the smallest integer type which has enough room.
7486 @item -fshort-double
7487 Use the same size for @code{double} as for @code{float}.
7490 Requests that the data and non-@code{const} variables of this
7491 compilation be shared data rather than private data. The distinction
7492 makes sense only on certain operating systems, where shared data is
7493 shared between processes running the same program, while private data
7494 exists in one copy per process.
7497 Allocate even uninitialized global variables in the data section of the
7498 object file, rather than generating them as common blocks. This has the
7499 effect that if the same variable is declared (without @code{extern}) in
7500 two different compilations, you will get an error when you link them.
7501 The only reason this might be useful is if you wish to verify that the
7502 program will work on other systems which always work this way.
7505 Ignore the @samp{#ident} directive.
7507 @item -fno-gnu-linker
7508 Do not output global initializations (such as C++ constructors and
7509 destructors) in the form used by the GNU linker (on systems where the GNU
7510 linker is the standard method of handling them). Use this option when
7511 you want to use a non-GNU linker, which also requires using the
7512 @code{collect2} program to make sure the system linker includes
7513 constructors and destructors. (@code{collect2} is included in the GCC
7514 distribution.) For systems which @emph{must} use @code{collect2}, the
7515 compiler driver @code{gcc} is configured to do this automatically.
7517 @item -finhibit-size-directive
7518 Don't output a @code{.size} assembler directive, or anything else that
7519 would cause trouble if the function is split in the middle, and the
7520 two halves are placed at locations far apart in memory. This option is
7521 used when compiling @file{crtstuff.c}; you should not need to use it
7525 Put extra commentary information in the generated assembly code to
7526 make it more readable. This option is generally only of use to those
7527 who actually need to read the generated assembly code (perhaps while
7528 debugging the compiler itself).
7530 @samp{-fno-verbose-asm}, the default, causes the
7531 extra information to be omitted and is useful when comparing two assembler
7535 Consider all memory references through pointers to be volatile.
7537 @item -fvolatile-global
7538 Consider all memory references to extern and global data items to
7539 be volatile. GCC does not consider static data items to be volatile
7540 because of this switch.
7542 @item -fvolatile-static
7543 Consider all memory references to static data to be volatile.
7546 @cindex global offset table
7548 Generate position-independent code (PIC) suitable for use in a shared
7549 library, if supported for the target machine. Such code accesses all
7550 constant addresses through a global offset table (GOT). The dynamic
7551 loader resolves the GOT entries when the program starts (the dynamic
7552 loader is not part of GCC; it is part of the operating system). If
7553 the GOT size for the linked executable exceeds a machine-specific
7554 maximum size, you get an error message from the linker indicating that
7555 @samp{-fpic} does not work; in that case, recompile with @samp{-fPIC}
7556 instead. (These maximums are 16k on the m88k, 8k on the Sparc, and 32k
7557 on the m68k and RS/6000. The 386 has no such limit.)
7559 Position-independent code requires special support, and therefore works
7560 only on certain machines. For the 386, GCC supports PIC for System V
7561 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
7562 position-independent.
7565 If supported for the target machine, emit position-independent code,
7566 suitable for dynamic linking and avoiding any limit on the size of the
7567 global offset table. This option makes a difference on the m68k, m88k,
7570 Position-independent code requires special support, and therefore works
7571 only on certain machines.
7573 @item -ffixed-@var{reg}
7574 Treat the register named @var{reg} as a fixed register; generated code
7575 should never refer to it (except perhaps as a stack pointer, frame
7576 pointer or in some other fixed role).
7578 @var{reg} must be the name of a register. The register names accepted
7579 are machine-specific and are defined in the @code{REGISTER_NAMES}
7580 macro in the machine description macro file.
7582 This flag does not have a negative form, because it specifies a
7585 @item -fcall-used-@var{reg}
7586 Treat the register named @var{reg} as an allocable register that is
7587 clobbered by function calls. It may be allocated for temporaries or
7588 variables that do not live across a call. Functions compiled this way
7589 will not save and restore the register @var{reg}.
7591 It is an error to used this flag with the frame pointer or stack pointer.
7592 Use of this flag for other registers that have fixed pervasive roles in
7593 the machine's execution model will produce disastrous results.
7595 This flag does not have a negative form, because it specifies a
7598 @item -fcall-saved-@var{reg}
7599 Treat the register named @var{reg} as an allocable register saved by
7600 functions. It may be allocated even for temporaries or variables that
7601 live across a call. Functions compiled this way will save and restore
7602 the register @var{reg} if they use it.
7604 It is an error to used this flag with the frame pointer or stack pointer.
7605 Use of this flag for other registers that have fixed pervasive roles in
7606 the machine's execution model will produce disastrous results.
7608 A different sort of disaster will result from the use of this flag for
7609 a register in which function values may be returned.
7611 This flag does not have a negative form, because it specifies a
7615 Pack all structure members together without holes. Usually you would
7616 not want to use this option, since it makes the code suboptimal, and
7617 the offsets of structure members won't agree with system libraries.
7619 @item -fcheck-memory-usage
7620 Generate extra code to check each memory access. GCC will generate
7621 code that is suitable for a detector of bad memory accesses such as
7624 Normally, you should compile all, or none, of your code with this option.
7626 If you do mix code compiled with and without this option,
7627 you must ensure that all code that has side effects
7628 and that is called by code compiled with this option
7629 is, itself, compiled with this option.
7630 If you do not, you might get erroneous messages from the detector.
7632 If you use functions from a library that have side-effects (such as
7633 @code{read}), you might not be able to recompile the library and
7634 specify this option. In that case, you can enable the
7635 @samp{-fprefix-function-name} option, which requests GCC to encapsulate
7636 your code and make other functions look as if they were compiled with
7637 @samp{-fcheck-memory-usage}. This is done by calling ``stubs'',
7638 which are provided by the detector. If you cannot find or build
7639 stubs for every function you call, you might have to specify
7640 @samp{-fcheck-memory-usage} without @samp{-fprefix-function-name}.
7642 If you specify this option, you can not use the @code{asm} or
7643 @code{__asm__} keywords in functions with memory checking enabled. GNU
7644 CC cannot understand what the @code{asm} statement may do, and therefore
7645 cannot generate the appropriate code, so it will reject it. However, if
7646 you specify the function attribute @code{no_check_memory_usage} (see
7647 @pxref{Function Attributes}, GNU CC will disable memory checking within a
7648 function; you may use @code{asm} statements inside such functions. You
7649 may have an inline expansion of a non-checked function within a checked
7650 function; in that case GNU CC will not generate checks for the inlined
7651 function's memory accesses.
7653 If you move your @code{asm} statements to non-checked inline functions
7654 and they do access memory, you can add calls to the support code in your
7655 inline function, to indicate any reads, writes, or copies being done.
7656 These calls would be similar to those done in the stubs described above.
7658 @item -fprefix-function-name
7659 Request GCC to add a prefix to the symbols generated for function names.
7660 GCC adds a prefix to the names of functions defined as well as
7661 functions called. Code compiled with this option and code compiled
7662 without the option can't be linked together, unless stubs are used.
7664 If you compile the following code with @samp{-fprefix-function-name}
7666 extern void bar (int);
7675 GCC will compile the code as if it was written:
7677 extern void prefix_bar (int);
7681 return prefix_bar (a + 5);
7684 This option is designed to be used with @samp{-fcheck-memory-usage}.
7686 @item -finstrument-functions
7687 Generate instrumentation calls for entry and exit to functions. Just
7688 after function entry and just before function exit, the following
7689 profiling functions will be called with the address of the current
7690 function and its call site. (On some platforms,
7691 @code{__builtin_return_address} does not work beyond the current
7692 function, so the call site information may not be available to the
7693 profiling functions otherwise.)
7696 void __cyg_profile_func_enter (void *this_fn, void *call_site);
7697 void __cyg_profile_func_exit (void *this_fn, void *call_site);
7700 The first argument is the address of the start of the current function,
7701 which may be looked up exactly in the symbol table.
7703 This instrumentation is also done for functions expanded inline in other
7704 functions. The profiling calls will indicate where, conceptually, the
7705 inline function is entered and exited. This means that addressable
7706 versions of such functions must be available. If all your uses of a
7707 function are expanded inline, this may mean an additional expansion of
7708 code size. If you use @samp{extern inline} in your C code, an
7709 addressable version of such functions must be provided. (This is
7710 normally the case anyways, but if you get lucky and the optimizer always
7711 expands the functions inline, you might have gotten away without
7712 providing static copies.)
7714 A function may be given the attribute @code{no_instrument_function}, in
7715 which case this instrumentation will not be done. This can be used, for
7716 example, for the profiling functions listed above, high-priority
7717 interrupt routines, and any functions from which the profiling functions
7718 cannot safely be called (perhaps signal handlers, if the profiling
7719 routines generate output or allocate memory).
7722 Generate code to verify that you do not go beyond the boundary of the
7723 stack. You should specify this flag if you are running in an
7724 environment with multiple threads, but only rarely need to specify it in
7725 a single-threaded environment since stack overflow is automatically
7726 detected on nearly all systems if there is only one stack.
7728 Note that this switch does not actually cause checking to be done; the
7729 operating system must do that. The switch causes generation of code
7730 to ensure that the operating system sees the stack being extended.
7732 @item -fstack-limit-register=@var{reg}
7733 @itemx -fstack-limit-symbol=@var{sym}
7734 @itemx -fno-stack-limit
7735 Generate code to ensure that the stack does not grow beyond a certain value,
7736 either the value of a register or the address of a symbol. If the stack
7737 would grow beyond the value, a signal is raised. For most targets,
7738 the signal is raised before the stack overruns the boundary, so
7739 it is possible to catch the signal without taking special precautions.
7741 For instance, if the stack starts at address @samp{0x80000000} and grows
7742 downwards you can use the flags
7743 @samp{-fstack-limit-symbol=__stack_limit}
7744 @samp{-Wl,--defsym,__stack_limit=0x7ffe0000} which will enforce a stack
7747 @cindex aliasing of parameters
7748 @cindex parameters, aliased
7749 @item -fargument-alias
7750 @itemx -fargument-noalias
7751 @itemx -fargument-noalias-global
7752 Specify the possible relationships among parameters and between
7753 parameters and global data.
7755 @samp{-fargument-alias} specifies that arguments (parameters) may
7756 alias each other and may alias global storage.
7757 @samp{-fargument-noalias} specifies that arguments do not alias
7758 each other, but may alias global storage.
7759 @samp{-fargument-noalias-global} specifies that arguments do not
7760 alias each other and do not alias global storage.
7762 Each language will automatically use whatever option is required by
7763 the language standard. You should not need to use these options yourself.
7765 @item -fleading-underscore
7766 This option and its counterpart, -fno-leading-underscore, forcibly
7767 change the way C symbols are represented in the object file. One use
7768 is to help link with legacy assembly code.
7770 Be warned that you should know what you are doing when invoking this
7771 option, and that not all targets provide complete support for it.
7774 @node Environment Variables
7775 @section Environment Variables Affecting GCC
7776 @cindex environment variables
7778 This section describes several environment variables that affect how GCC
7779 operates. Some of them work by specifying directories or prefixes to use
7780 when searching for various kinds of files. Some are used to specify other
7781 aspects of the compilation environment.
7784 Note that you can also specify places to search using options such as
7785 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
7786 take precedence over places specified using environment variables, which
7787 in turn take precedence over those specified by the configuration of GCC.
7791 Note that you can also specify places to search using options such as
7792 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
7793 take precedence over places specified using environment variables, which
7794 in turn take precedence over those specified by the configuration of GCC.
7801 @c @itemx LC_COLLATE
7803 @c @itemx LC_MONETARY
7804 @c @itemx LC_NUMERIC
7809 @c @findex LC_COLLATE
7811 @c @findex LC_MONETARY
7812 @c @findex LC_NUMERIC
7816 These environment variables control the way that GCC uses
7817 localization information that allow GCC to work with different
7818 national conventions. GCC inspects the locale categories
7819 @code{LC_CTYPE} and @code{LC_MESSAGES} if it has been configured to do
7820 so. These locale categories can be set to any value supported by your
7821 installation. A typical value is @samp{en_UK} for English in the United
7824 The @code{LC_CTYPE} environment variable specifies character
7825 classification. GCC uses it to determine the character boundaries in
7826 a string; this is needed for some multibyte encodings that contain quote
7827 and escape characters that would otherwise be interpreted as a string
7830 The @code{LC_MESSAGES} environment variable specifies the language to
7831 use in diagnostic messages.
7833 If the @code{LC_ALL} environment variable is set, it overrides the value
7834 of @code{LC_CTYPE} and @code{LC_MESSAGES}; otherwise, @code{LC_CTYPE}
7835 and @code{LC_MESSAGES} default to the value of the @code{LANG}
7836 environment variable. If none of these variables are set, GCC
7837 defaults to traditional C English behavior.
7841 If @code{TMPDIR} is set, it specifies the directory to use for temporary
7842 files. GCC uses temporary files to hold the output of one stage of
7843 compilation which is to be used as input to the next stage: for example,
7844 the output of the preprocessor, which is the input to the compiler
7847 @item GCC_EXEC_PREFIX
7848 @findex GCC_EXEC_PREFIX
7849 If @code{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
7850 names of the subprograms executed by the compiler. No slash is added
7851 when this prefix is combined with the name of a subprogram, but you can
7852 specify a prefix that ends with a slash if you wish.
7854 If @code{GCC_EXEC_PREFIX} is not set, GNU CC will attempt to figure out
7855 an appropriate prefix to use based on the pathname it was invoked with.
7857 If GCC cannot find the subprogram using the specified prefix, it
7858 tries looking in the usual places for the subprogram.
7860 The default value of @code{GCC_EXEC_PREFIX} is
7861 @file{@var{prefix}/lib/gcc-lib/} where @var{prefix} is the value
7862 of @code{prefix} when you ran the @file{configure} script.
7864 Other prefixes specified with @samp{-B} take precedence over this prefix.
7866 This prefix is also used for finding files such as @file{crt0.o} that are
7869 In addition, the prefix is used in an unusual way in finding the
7870 directories to search for header files. For each of the standard
7871 directories whose name normally begins with @samp{/usr/local/lib/gcc-lib}
7872 (more precisely, with the value of @code{GCC_INCLUDE_DIR}), GCC tries
7873 replacing that beginning with the specified prefix to produce an
7874 alternate directory name. Thus, with @samp{-Bfoo/}, GCC will search
7875 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
7876 These alternate directories are searched first; the standard directories
7880 @findex COMPILER_PATH
7881 The value of @code{COMPILER_PATH} is a colon-separated list of
7882 directories, much like @code{PATH}. GCC tries the directories thus
7883 specified when searching for subprograms, if it can't find the
7884 subprograms using @code{GCC_EXEC_PREFIX}.
7887 @findex LIBRARY_PATH
7888 The value of @code{LIBRARY_PATH} is a colon-separated list of
7889 directories, much like @code{PATH}. When configured as a native compiler,
7890 GCC tries the directories thus specified when searching for special
7891 linker files, if it can't find them using @code{GCC_EXEC_PREFIX}. Linking
7892 using GCC also uses these directories when searching for ordinary
7893 libraries for the @samp{-l} option (but directories specified with
7894 @samp{-L} come first).
7896 @item C_INCLUDE_PATH
7897 @itemx CPLUS_INCLUDE_PATH
7898 @itemx OBJC_INCLUDE_PATH
7899 @findex C_INCLUDE_PATH
7900 @findex CPLUS_INCLUDE_PATH
7901 @findex OBJC_INCLUDE_PATH
7902 @c @itemx OBJCPLUS_INCLUDE_PATH
7903 These environment variables pertain to particular languages. Each
7904 variable's value is a colon-separated list of directories, much like
7905 @code{PATH}. When GCC searches for header files, it tries the
7906 directories listed in the variable for the language you are using, after
7907 the directories specified with @samp{-I} but before the standard header
7910 @item DEPENDENCIES_OUTPUT
7911 @findex DEPENDENCIES_OUTPUT
7912 @cindex dependencies for make as output
7913 If this variable is set, its value specifies how to output dependencies
7914 for Make based on the header files processed by the compiler. This
7915 output looks much like the output from the @samp{-M} option
7916 (@pxref{Preprocessor Options}), but it goes to a separate file, and is
7917 in addition to the usual results of compilation.
7919 The value of @code{DEPENDENCIES_OUTPUT} can be just a file name, in
7920 which case the Make rules are written to that file, guessing the target
7921 name from the source file name. Or the value can have the form
7922 @samp{@var{file} @var{target}}, in which case the rules are written to
7923 file @var{file} using @var{target} as the target name.
7927 @cindex locale definition
7928 This variable is used to pass locale information to the compiler. One way in
7929 which this information is used is to determine the character set to be used
7930 when character literals, string literals and comments are parsed in C and C++.
7931 When the compiler is configured to allow multibyte characters,
7932 the following values for @code{LANG} are recognized:
7936 Recognize JIS characters.
7938 Recognize SJIS characters.
7940 Recognize EUCJP characters.
7943 If @code{LANG} is not defined, or if it has some other value, then the
7944 compiler will use mblen and mbtowc as defined by the default locale to
7945 recognize and translate multibyte characters.
7948 @node Running Protoize
7949 @section Running Protoize
7951 The program @code{protoize} is an optional part of GNU C. You can use
7952 it to add prototypes to a program, thus converting the program to ISO
7953 C in one respect. The companion program @code{unprotoize} does the
7954 reverse: it removes argument types from any prototypes that are found.
7956 When you run these programs, you must specify a set of source files as
7957 command line arguments. The conversion programs start out by compiling
7958 these files to see what functions they define. The information gathered
7959 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
7961 After scanning comes actual conversion. The specified files are all
7962 eligible to be converted; any files they include (whether sources or
7963 just headers) are eligible as well.
7965 But not all the eligible files are converted. By default,
7966 @code{protoize} and @code{unprotoize} convert only source and header
7967 files in the current directory. You can specify additional directories
7968 whose files should be converted with the @samp{-d @var{directory}}
7969 option. You can also specify particular files to exclude with the
7970 @samp{-x @var{file}} option. A file is converted if it is eligible, its
7971 directory name matches one of the specified directory names, and its
7972 name within the directory has not been excluded.
7974 Basic conversion with @code{protoize} consists of rewriting most
7975 function definitions and function declarations to specify the types of
7976 the arguments. The only ones not rewritten are those for varargs
7979 @code{protoize} optionally inserts prototype declarations at the
7980 beginning of the source file, to make them available for any calls that
7981 precede the function's definition. Or it can insert prototype
7982 declarations with block scope in the blocks where undeclared functions
7985 Basic conversion with @code{unprotoize} consists of rewriting most
7986 function declarations to remove any argument types, and rewriting
7987 function definitions to the old-style pre-ISO form.
7989 Both conversion programs print a warning for any function declaration or
7990 definition that they can't convert. You can suppress these warnings
7993 The output from @code{protoize} or @code{unprotoize} replaces the
7994 original source file. The original file is renamed to a name ending
7995 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
7996 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
7997 for DOS) file already exists, then the source file is simply discarded.
7999 @code{protoize} and @code{unprotoize} both depend on GCC itself to
8000 scan the program and collect information about the functions it uses.
8001 So neither of these programs will work until GCC is installed.
8003 Here is a table of the options you can use with @code{protoize} and
8004 @code{unprotoize}. Each option works with both programs unless
8008 @item -B @var{directory}
8009 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
8010 usual directory (normally @file{/usr/local/lib}). This file contains
8011 prototype information about standard system functions. This option
8012 applies only to @code{protoize}.
8014 @item -c @var{compilation-options}
8015 Use @var{compilation-options} as the options when running @code{gcc} to
8016 produce the @samp{.X} files. The special option @samp{-aux-info} is
8017 always passed in addition, to tell @code{gcc} to write a @samp{.X} file.
8019 Note that the compilation options must be given as a single argument to
8020 @code{protoize} or @code{unprotoize}. If you want to specify several
8021 @code{gcc} options, you must quote the entire set of compilation options
8022 to make them a single word in the shell.
8024 There are certain @code{gcc} arguments that you cannot use, because they
8025 would produce the wrong kind of output. These include @samp{-g},
8026 @samp{-O}, @samp{-c}, @samp{-S}, and @samp{-o} If you include these in
8027 the @var{compilation-options}, they are ignored.
8030 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
8031 systems) instead of @samp{.c}. This is convenient if you are converting
8032 a C program to C++. This option applies only to @code{protoize}.
8035 Add explicit global declarations. This means inserting explicit
8036 declarations at the beginning of each source file for each function
8037 that is called in the file and was not declared. These declarations
8038 precede the first function definition that contains a call to an
8039 undeclared function. This option applies only to @code{protoize}.
8041 @item -i @var{string}
8042 Indent old-style parameter declarations with the string @var{string}.
8043 This option applies only to @code{protoize}.
8045 @code{unprotoize} converts prototyped function definitions to old-style
8046 function definitions, where the arguments are declared between the
8047 argument list and the initial @samp{@{}. By default, @code{unprotoize}
8048 uses five spaces as the indentation. If you want to indent with just
8049 one space instead, use @samp{-i " "}.
8052 Keep the @samp{.X} files. Normally, they are deleted after conversion
8056 Add explicit local declarations. @code{protoize} with @samp{-l} inserts
8057 a prototype declaration for each function in each block which calls the
8058 function without any declaration. This option applies only to
8062 Make no real changes. This mode just prints information about the conversions
8063 that would have been done without @samp{-n}.
8066 Make no @samp{.save} files. The original files are simply deleted.
8067 Use this option with caution.
8069 @item -p @var{program}
8070 Use the program @var{program} as the compiler. Normally, the name
8074 Work quietly. Most warnings are suppressed.
8077 Print the version number, just like @samp{-v} for @code{gcc}.
8080 If you need special compiler options to compile one of your program's
8081 source files, then you should generate that file's @samp{.X} file
8082 specially, by running @code{gcc} on that source file with the
8083 appropriate options and the option @samp{-aux-info}. Then run
8084 @code{protoize} on the entire set of files. @code{protoize} will use
8085 the existing @samp{.X} file because it is newer than the source file.
8089 gcc -Dfoo=bar file1.c -aux-info
8094 You need to include the special files along with the rest in the
8095 @code{protoize} command, even though their @samp{.X} files already
8096 exist, because otherwise they won't get converted.
8098 @xref{Protoize Caveats}, for more information on how to use
8099 @code{protoize} successfully.