1 @c Copyright (C) 1988,89,92-97,1998 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 GNU CC Command Options
7 @cindex GNU CC command options
8 @cindex command options
9 @cindex options, GNU CC command
11 When you invoke GNU CC, 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 GNU CC are useful
24 for C programs; when an option is only useful with another language
25 (usually C++), the explanation says so explicitly. If the description
26 for a particular option does not mention a source language, you can use
27 that option with all supported languages.
29 @cindex C++ compilation options
30 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
31 options for compiling C++ programs.
33 @cindex grouping options
34 @cindex options, grouping
35 The @code{gcc} program accepts options and file names as operands. Many
36 options have multiletter names; therefore multiple single-letter options
37 may @emph{not} be grouped: @samp{-dr} is very different from @w{@samp{-d
40 @cindex order of options
41 @cindex options, order
42 You can mix options and other arguments. For the most part, the order
43 you use doesn't matter. Order does matter when you use several options
44 of the same kind; for example, if you specify @samp{-L} more than once,
45 the directories are searched in the order specified.
47 Many options have long names starting with @samp{-f} or with
48 @samp{-W}---for example, @samp{-fforce-mem},
49 @samp{-fstrength-reduce}, @samp{-Wformat} and so on. Most of
50 these have both positive and negative forms; the negative form of
51 @samp{-ffoo} would be @samp{-fno-foo}. This manual documents
52 only one of these two forms, whichever one is not the default.
55 * Option Summary:: Brief list of all options, without explanations.
56 * Overall Options:: Controlling the kind of output:
57 an executable, object files, assembler files,
58 or preprocessed source.
59 * Invoking G++:: Compiling C++ programs.
60 * C Dialect Options:: Controlling the variant of C language compiled.
61 * C++ Dialect Options:: Variations on C++.
62 * Warning Options:: How picky should the compiler be?
63 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
64 * Optimize Options:: How much optimization?
65 * Preprocessor Options:: Controlling header files and macro definitions.
66 Also, getting dependency information for Make.
67 * Assembler Options:: Passing options to the assembler.
68 * Link Options:: Specifying libraries and so on.
69 * Directory Options:: Where to find header files and libraries.
70 Where to find the compiler executable files.
71 * Target Options:: Running a cross-compiler, or an old version of GNU CC.
72 * Submodel Options:: Specifying minor hardware or convention variations,
73 such as 68010 vs 68020.
74 * Code Gen Options:: Specifying conventions for function calls, data layout
76 * Environment Variables:: Env vars that affect GNU CC.
77 * Running Protoize:: Automatically adding or removing function prototypes.
81 @section Option Summary
83 Here is a summary of all the options, grouped by type. Explanations are
84 in the following sections.
88 @xref{Overall Options,,Options Controlling the Kind of Output}.
90 -c -S -E -o @var{file} -pipe -v -x @var{language}
93 @item C Language Options
94 @xref{C Dialect Options,,Options Controlling C Dialect}.
96 -ansi -fallow-single-precision -fcond-mismatch -fno-asm
97 -fno-builtin -ffreestanding -fhosted -fsigned-bitfields -fsigned-char
98 -funsigned-bitfields -funsigned-char -fwritable-strings
99 -traditional -traditional-cpp -trigraphs
102 @item C++ Language Options
103 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
105 -fall-virtual -fdollars-in-identifiers -felide-constructors
106 -fenum-int-equiv -fexternal-templates -ffor-scope -fno-for-scope
107 -fhandle-signatures -fmemoize-lookups -fname-mangling-version-@var{n}
108 -fno-default-inline -fno-gnu-keywords -fnonnull-objects -fguiding-decls
109 -foperator-names -fstrict-prototype -fthis-is-variable
110 -ftemplate-depth-@var{n} -nostdinc++ -traditional +e@var{n}
113 @item Warning Options
114 @xref{Warning Options,,Options to Request or Suppress Warnings}.
116 -fsyntax-only -pedantic -pedantic-errors
117 -w -W -Wall -Waggregate-return -Wbad-function-cast
118 -Wcast-align -Wcast-qual -Wchar-subscript -Wcomment
119 -Wconversion -Werror -Wformat
120 -Wid-clash-@var{len} -Wimplicit -Wimplicit-int
121 -Wimplicit-function-declarations -Wimport -Winline
122 -Wlarger-than-@var{len} -Wmain -Wmissing-declarations
123 -Wmissing-prototypes -Wnested-externs
124 -Wno-import -Wold-style-cast -Woverloaded-virtual -Wparentheses
125 -Wpointer-arith -Wredundant-decls -Wreorder -Wreturn-type -Wshadow
126 -Wsign-compare -Wstrict-prototypes -Wswitch -Wsynth
127 -Wtemplate-debugging -Wtraditional -Wtrigraphs
128 -Wundef -Wuninitialized -Wunused -Wwrite-strings
132 @item Debugging Options
133 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
135 -a -ax -d@var{letters} -fpretend-float
136 -fprofile-arcs -ftest-coverage
137 -g -g@var{level} -gcoff -gdwarf -gdwarf-1 -gdwarf-1+ -gdwarf-2
138 -ggdb -gstabs -gstabs+ -gxcoff -gxcoff+
139 -p -pg -print-file-name=@var{library} -print-libgcc-file-name
140 -print-prog-name=@var{program} -print-search-dirs -save-temps
143 @item Optimization Options
144 @xref{Optimize Options,,Options that Control Optimization}.
146 -fbranch-probabilities
147 -fcaller-saves -fcse-follow-jumps -fcse-skip-blocks
148 -fdelayed-branch -fexpensive-optimizations
149 -ffast-math -ffloat-store -fforce-addr -fforce-mem
150 -ffunction-sections -finline-functions
151 -fkeep-inline-functions -fno-default-inline
152 -fno-defer-pop -fno-function-cse
153 -fno-inline -fno-peephole -fomit-frame-pointer -fregmove
154 -frerun-cse-after-loop -frerun-loop-opt -fschedule-insns
155 -fschedule-insns2 -fstrength-reduce -fthread-jumps
156 -funroll-all-loops -funroll-loops
157 -fmove-all-movables -freduce-all-givs
158 -O -O0 -O1 -O2 -O3 -Os
161 @item Preprocessor Options
162 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
164 -A@var{question}(@var{answer}) -C -dD -dM -dN
165 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H
167 -include @var{file} -imacros @var{file}
168 -iprefix @var{file} -iwithprefix @var{dir}
169 -iwithprefixbefore @var{dir} -isystem @var{dir}
170 -M -MD -MM -MMD -MG -nostdinc -P -trigraphs
171 -undef -U@var{macro} -Wp,@var{option}
174 @item Assembler Option
175 @xref{Assembler Options,,Passing Options to the Assembler}.
181 @xref{Link Options,,Options for Linking}.
183 @var{object-file-name} -l@var{library}
184 -nostartfiles -nodefaultlibs -nostdlib
185 -s -static -shared -symbolic
186 -Wl,@var{option} -Xlinker @var{option}
190 @item Directory Options
191 @xref{Directory Options,,Options for Directory Search}.
193 -B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}
197 @c I wrote this xref this way to avoid overfull hbox. -- rms
198 @xref{Target Options}.
200 -b @var{machine} -V @var{version}
203 @item Machine Dependent Options
204 @xref{Submodel Options,,Hardware Models and Configurations}.
206 @emph{M680x0 Options}
207 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040
208 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 -mfpa
209 -mnobitfield -mrtd -mshort -msoft-float -malign-int
216 -mtune=@var{cpu type}
217 -mcmodel=@var{code model}
218 -malign-jumps=@var{num} -malign-loops=@var{num}
219 -malign-functions=@var{num}
221 -mapp-regs -mbroken-saverestore -mcypress -mepilogue
222 -mflat -mfpu -mhard-float -mhard-quad-float
223 -mimpure-text -mlive-g0 -mno-app-regs -mno-epilogue
224 -mno-flat -mno-fpu -mno-impure-text
225 -mno-stack-bias -mno-unaligned-doubles
226 -msoft-float -msoft-quad-float -msparclite -mstack-bias
227 -msupersparc -munaligned-doubles -mv8
229 @emph{Convex Options}
230 -mc1 -mc2 -mc32 -mc34 -mc38
231 -margcount -mnoargcount
233 -mvolatile-cache -mvolatile-nocache
235 @emph{AMD29K Options}
236 -m29000 -m29050 -mbw -mnbw -mdw -mndw
237 -mlarge -mnormal -msmall
238 -mkernel-registers -mno-reuse-arg-regs
239 -mno-stack-check -mno-storem-bug
240 -mreuse-arg-regs -msoft-float -mstack-check
241 -mstorem-bug -muser-registers
244 -mapcs-frame -mapcs-26 -mapcs-32
245 -mlittle-endian -mbig-endian -mwords-little-endian
246 -mshort-load-bytes -mno-short-load-bytes
247 -msoft-float -mhard-float
248 -mbsd -mxopen -mno-symrename
250 @emph{MN10300 Options}
254 @emph{M32R/D Options}
255 -mcode-model=@var{model type} -msdata=@var{sdata type}
259 -m88000 -m88100 -m88110 -mbig-pic
260 -mcheck-zero-division -mhandle-large-shift
261 -midentify-revision -mno-check-zero-division
262 -mno-ocs-debug-info -mno-ocs-frame-position
263 -mno-optimize-arg-area -mno-serialize-volatile
264 -mno-underscores -mocs-debug-info
265 -mocs-frame-position -moptimize-arg-area
266 -mserialize-volatile -mshort-data-@var{num} -msvr3
267 -msvr4 -mtrap-large-shift -muse-div-instruction
268 -mversion-03.00 -mwarn-passed-structs
270 @emph{RS/6000 and PowerPC Options}
272 -mtune=@var{cpu type}
273 -mpower -mno-power -mpower2 -mno-power2
274 -mpowerpc -mno-powerpc
275 -mpowerpc-gpopt -mno-powerpc-gpopt
276 -mpowerpc-gfxopt -mno-powerpc-gfxopt
277 -mnew-mnemonics -mno-new-mnemonics
278 -mfull-toc -mminimal-toc -mno-fop-in-toc -mno-sum-in-toc
279 -mxl-call -mno-xl-call -mthreads -mpe
280 -msoft-float -mhard-float -mmultiple -mno-multiple
281 -mstring -mno-string -mupdate -mno-update
282 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align
283 -mstrict-align -mno-strict-align -mrelocatable
284 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib
285 -mtoc -mno-toc -mtraceback -mno-traceback
286 -mlittle -mlittle-endian -mbig -mbig-endian
287 -mcall-aix -mcall-sysv -mprototype -mno-prototype
288 -msim -mmvme -mads -myellowknife -memb
289 -msdata -msdata=@var{opt} -G @var{num}
292 -mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs
293 -mfull-fp-blocks -mhc-struct-return -min-line-mul
294 -mminimum-fp-blocks -mnohc-struct-return
297 -mabicalls -mcpu=@var{cpu type} -membedded-data
298 -membedded-pic -mfp32 -mfp64 -mgas -mgp32 -mgp64
299 -mgpopt -mhalf-pic -mhard-float -mint64 -mips1
300 -mips2 -mips3 -mlong64 -mlong-calls -mmemcpy
301 -mmips-as -mmips-tfile -mno-abicalls
302 -mno-embedded-data -mno-embedded-pic
303 -mno-gpopt -mno-long-calls
304 -mno-memcpy -mno-mips-tfile -mno-rnames -mno-stats
305 -mrnames -msoft-float
306 -m4650 -msingle-float -mmad
307 -mstats -EL -EB -G @var{num} -nocpp
311 -march=@var{cpu type}
312 -mieee-fp -mno-fancy-math-387
313 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib
314 -mno-wide-multiply -mrtd -malign-double
315 -mreg-alloc=@var{list} -mregparm=@var{num}
316 -malign-jumps=@var{num} -malign-loops=@var{num}
317 -malign-functions=@var{num}
320 -mbig-switch -mdisable-fpregs -mdisable-indexing -mfast-indirect-calls
321 -mgas -mjump-in-delay -mlong-load-store -mno-big-switch -mno-disable-fpregs
322 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas
325 -mno-portable-runtime -mno-soft-float -mno-space -mno-space-regs
327 -mpa-risc-1-0 -mpa-risc-1-1 -mportable-runtime
328 -mschedule=@var{list} -mspace -mspace-regs
330 @emph{Intel 960 Options}
331 -m@var{cpu type} -masm-compat -mclean-linkage
332 -mcode-align -mcomplex-addr -mleaf-procedures
333 -mic-compat -mic2.0-compat -mic3.0-compat
334 -mintel-asm -mno-clean-linkage -mno-code-align
335 -mno-complex-addr -mno-leaf-procedures
336 -mno-old-align -mno-strict-align -mno-tail-call
337 -mnumerics -mold-align -msoft-float -mstrict-align
340 @emph{DEC Alpha Options}
341 -mfp-regs -mno-fp-regs -mno-soft-float -msoft-float
343 -mieee -mieee-with-inexact -mieee-conformant
344 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode}
345 -mtrap-precision=@var{mode} -mbuild-constants
347 -mbwx -mno-bwx -mcix -mno-cix -mmax -mno-max
348 -mmemory-latency=@var{time}
350 @emph{Clipper Options}
353 @emph{H8/300 Options}
354 -mrelax -mh -ms -mint32 -malign-300
357 -m1 -m2 -m3 -m3e -mb -ml -mrelax
359 @emph{System V Options}
360 -Qy -Qn -YP,@var{paths} -Ym,@var{dir}
363 -mlong-calls -mno-long-calls -mep -mno-ep
364 -mprolog-function -mno-prolog-function -mspace
365 -mtda=@var{n} -msda=@var{n} -mzda=@var{n}
369 @item Code Generation Options
370 @xref{Code Gen Options,,Options for Code Generation Conventions}.
372 -fcall-saved-@var{reg} -fcall-used-@var{reg}
373 -fexceptions -ffixed-@var{reg} -finhibit-size-directive
374 -fcheck-memory-usage -fprefix-function-name
375 -fno-common -fno-ident -fno-gnu-linker
376 -fpcc-struct-return -fpic -fPIC
377 -freg-struct-return -fshared-data -fshort-enums
378 -fshort-double -fvolatile -fvolatile-global
379 -fverbose-asm -fpack-struct -fstack-check +e0 +e1
380 -fargument-alias -fargument-noalias
381 -fargument-noalias-global
386 * Overall Options:: Controlling the kind of output:
387 an executable, object files, assembler files,
388 or preprocessed source.
389 * C Dialect Options:: Controlling the variant of C language compiled.
390 * C++ Dialect Options:: Variations on C++.
391 * Warning Options:: How picky should the compiler be?
392 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
393 * Optimize Options:: How much optimization?
394 * Preprocessor Options:: Controlling header files and macro definitions.
395 Also, getting dependency information for Make.
396 * Assembler Options:: Passing options to the assembler.
397 * Link Options:: Specifying libraries and so on.
398 * Directory Options:: Where to find header files and libraries.
399 Where to find the compiler executable files.
400 * Target Options:: Running a cross-compiler, or an old version of GNU CC.
403 @node Overall Options
404 @section Options Controlling the Kind of Output
406 Compilation can involve up to four stages: preprocessing, compilation
407 proper, assembly and linking, always in that order. The first three
408 stages apply to an individual source file, and end by producing an
409 object file; linking combines all the object files (those newly
410 compiled, and those specified as input) into an executable file.
412 @cindex file name suffix
413 For any given input file, the file name suffix determines what kind of
418 C source code which must be preprocessed.
421 C source code which should not be preprocessed.
424 C++ source code which should not be preprocessed.
427 Objective-C source code. Note that you must link with the library
428 @file{libobjc.a} to make an Objective-C program work.
431 C header file (not to be compiled or linked).
434 @itemx @var{file}.cxx
435 @itemx @var{file}.cpp
437 C++ source code which must be preprocessed. Note that in @samp{.cxx},
438 the last two letters must both be literally @samp{x}. Likewise,
439 @samp{.C} refers to a literal capital C.
445 Assembler code which must be preprocessed.
448 An object file to be fed straight into linking.
449 Any file name with no recognized suffix is treated this way.
452 You can specify the input language explicitly with the @samp{-x} option:
455 @item -x @var{language}
456 Specify explicitly the @var{language} for the following input files
457 (rather than letting the compiler choose a default based on the file
458 name suffix). This option applies to all following input files until
459 the next @samp{-x} option. Possible values for @var{language} are:
462 c-header cpp-output c++-cpp-output
463 assembler assembler-with-cpp
467 Turn off any specification of a language, so that subsequent files are
468 handled according to their file name suffixes (as they are if @samp{-x}
469 has not been used at all).
472 If you only want some of the stages of compilation, you can use
473 @samp{-x} (or filename suffixes) to tell @code{gcc} where to start, and
474 one of the options @samp{-c}, @samp{-S}, or @samp{-E} to say where
475 @code{gcc} is to stop. Note that some combinations (for example,
476 @samp{-x cpp-output -E} instruct @code{gcc} to do nothing at all.
480 Compile or assemble the source files, but do not link. The linking
481 stage simply is not done. The ultimate output is in the form of an
482 object file for each source file.
484 By default, the object file name for a source file is made by replacing
485 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
487 Unrecognized input files, not requiring compilation or assembly, are
491 Stop after the stage of compilation proper; do not assemble. The output
492 is in the form of an assembler code file for each non-assembler input
495 By default, the assembler file name for a source file is made by
496 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
498 Input files that don't require compilation are ignored.
501 Stop after the preprocessing stage; do not run the compiler proper. The
502 output is in the form of preprocessed source code, which is sent to the
505 Input files which don't require preprocessing are ignored.
507 @cindex output file option
509 Place output in file @var{file}. This applies regardless to whatever
510 sort of output is being produced, whether it be an executable file,
511 an object file, an assembler file or preprocessed C code.
513 Since only one output file can be specified, it does not make sense to
514 use @samp{-o} when compiling more than one input file, unless you are
515 producing an executable file as output.
517 If @samp{-o} is not specified, the default is to put an executable file
518 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
519 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
520 all preprocessed C source on standard output.@refill
523 Print (on standard error output) the commands executed to run the stages
524 of compilation. Also print the version number of the compiler driver
525 program and of the preprocessor and the compiler proper.
528 Use pipes rather than temporary files for communication between the
529 various stages of compilation. This fails to work on some systems where
530 the assembler is unable to read from a pipe; but the GNU assembler has
535 @section Compiling C++ Programs
537 @cindex suffixes for C++ source
538 @cindex C++ source file suffixes
539 C++ source files conventionally use one of the suffixes @samp{.C},
540 @samp{.cc}, @samp{cpp}, or @samp{.cxx}; preprocessed C++ files use the
541 suffix @samp{.ii}. GNU CC recognizes files with these names and
542 compiles them as C++ programs even if you call the compiler the same way
543 as for compiling C programs (usually with the name @code{gcc}).
547 However, C++ programs often require class libraries as well as a
548 compiler that understands the C++ language---and under some
549 circumstances, you might want to compile programs from standard input,
550 or otherwise without a suffix that flags them as C++ programs.
551 @code{g++} is a program that calls GNU CC with the default language
552 set to C++, and automatically specifies linking against the GNU class
554 @cindex @code{g++ 1.@var{xx}}
555 @cindex @code{g++}, separate compiler
556 @cindex @code{g++} older version
557 @footnote{Prior to release 2 of the compiler,
558 there was a separate @code{g++} compiler. That version was based on GNU
559 CC, but not integrated with it. Versions of @code{g++} with a
560 @samp{1.@var{xx}} version number---for example, @code{g++} version 1.37
561 or 1.42---are much less reliable than the versions integrated with GCC
562 2. Moreover, combining G++ @samp{1.@var{xx}} with a version 2 GCC will
563 simply not work.} On many systems, the script @code{g++} is also
564 installed with the name @code{c++}.
566 @cindex invoking @code{g++}
567 When you compile C++ programs, you may specify many of the same
568 command-line options that you use for compiling programs in any
569 language; or command-line options meaningful for C and related
570 languages; or options that are meaningful only for C++ programs.
571 @xref{C Dialect Options,,Options Controlling C Dialect}, for
572 explanations of options for languages related to C.
573 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
574 explanations of options that are meaningful only for C++ programs.
576 @node C Dialect Options
577 @section Options Controlling C Dialect
578 @cindex dialect options
579 @cindex language dialect options
580 @cindex options, dialect
582 The following options control the dialect of C (or languages derived
583 from C, such as C++ and Objective C) that the compiler accepts:
588 Support all ANSI standard C programs.
590 This turns off certain features of GNU C that are incompatible with ANSI
591 C, such as the @code{asm}, @code{inline} and @code{typeof} keywords, and
592 predefined macros such as @code{unix} and @code{vax} that identify the
593 type of system you are using. It also enables the undesirable and
594 rarely used ANSI trigraph feature, and it disables recognition of C++
595 style @samp{//} comments.
597 The alternate keywords @code{__asm__}, @code{__extension__},
598 @code{__inline__} and @code{__typeof__} continue to work despite
599 @samp{-ansi}. You would not want to use them in an ANSI C program, of
600 course, but it is useful to put them in header files that might be included
601 in compilations done with @samp{-ansi}. Alternate predefined macros
602 such as @code{__unix__} and @code{__vax__} are also available, with or
603 without @samp{-ansi}.
605 The @samp{-ansi} option does not cause non-ANSI programs to be
606 rejected gratuitously. For that, @samp{-pedantic} is required in
607 addition to @samp{-ansi}. @xref{Warning Options}.
609 The macro @code{__STRICT_ANSI__} is predefined when the @samp{-ansi}
610 option is used. Some header files may notice this macro and refrain
611 from declaring certain functions or defining certain macros that the
612 ANSI standard doesn't call for; this is to avoid interfering with any
613 programs that might use these names for other things.
615 The functions @code{alloca}, @code{abort}, @code{exit}, and
616 @code{_exit} are not builtin functions when @samp{-ansi} is used.
619 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
620 keyword, so that code can use these words as identifiers. You can use
621 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
622 instead. @samp{-ansi} implies @samp{-fno-asm}.
624 In C++, this switch only affects the @code{typeof} keyword, since
625 @code{asm} and @code{inline} are standard keywords. You may want to
626 use the @samp{-fno-gnu-keywords} flag instead, as it also disables the
627 other, C++-specific, extension keywords such as @code{headof}.
630 @cindex builtin functions
646 Don't recognize builtin functions that do not begin with two leading
647 underscores. Currently, the functions affected include @code{abort},
648 @code{abs}, @code{alloca}, @code{cos}, @code{exit}, @code{fabs},
649 @code{ffs}, @code{labs}, @code{memcmp}, @code{memcpy}, @code{sin},
650 @code{sqrt}, @code{strcmp}, @code{strcpy}, and @code{strlen}.
652 GCC normally generates special code to handle certain builtin functions
653 more efficiently; for instance, calls to @code{alloca} may become single
654 instructions that adjust the stack directly, and calls to @code{memcpy}
655 may become inline copy loops. The resulting code is often both smaller
656 and faster, but since the function calls no longer appear as such, you
657 cannot set a breakpoint on those calls, nor can you change the behavior
658 of the functions by linking with a different library.
660 The @samp{-ansi} option prevents @code{alloca} and @code{ffs} from being
661 builtin functions, since these functions do not have an ANSI standard
665 @cindex hosted environment
667 Assert that compilation takes place in a hosted environment. This implies
668 @samp{-fbuiltin}. A hosted environment is one in which the
669 entire standard library is available, and in which @code{main} has a return
670 type of @code{int}. Examples are nearly everything except a kernel.
671 This is equivalent to @samp{-fno-freestanding}.
674 @cindex hosted environment
676 Assert that compilation takes place in a freestanding environment. This
677 implies @samp{-fno-builtin}. A freestanding environment
678 is one in which the standard library may not exist, and program startup may
679 not necessarily be at @code{main}. The most obvious example is an OS kernel.
680 This is equivalent to @samp{-fno-hosted}.
683 Support ANSI C trigraphs. You don't want to know about this
684 brain-damage. The @samp{-ansi} option implies @samp{-trigraphs}.
686 @cindex traditional C language
687 @cindex C language, traditional
689 Attempt to support some aspects of traditional C compilers.
694 All @code{extern} declarations take effect globally even if they
695 are written inside of a function definition. This includes implicit
696 declarations of functions.
699 The newer keywords @code{typeof}, @code{inline}, @code{signed}, @code{const}
700 and @code{volatile} are not recognized. (You can still use the
701 alternative keywords such as @code{__typeof__}, @code{__inline__}, and
705 Comparisons between pointers and integers are always allowed.
708 Integer types @code{unsigned short} and @code{unsigned char} promote
709 to @code{unsigned int}.
712 Out-of-range floating point literals are not an error.
715 Certain constructs which ANSI regards as a single invalid preprocessing
716 number, such as @samp{0xe-0xd}, are treated as expressions instead.
719 String ``constants'' are not necessarily constant; they are stored in
720 writable space, and identical looking constants are allocated
721 separately. (This is the same as the effect of
722 @samp{-fwritable-strings}.)
724 @cindex @code{longjmp} and automatic variables
726 All automatic variables not declared @code{register} are preserved by
727 @code{longjmp}. Ordinarily, GNU C follows ANSI C: automatic variables
728 not declared @code{volatile} may be clobbered.
733 @cindex escape sequences, traditional
734 The character escape sequences @samp{\x} and @samp{\a} evaluate as the
735 literal characters @samp{x} and @samp{a} respectively. Without
736 @w{@samp{-traditional}}, @samp{\x} is a prefix for the hexadecimal
737 representation of a character, and @samp{\a} produces a bell.
740 In C++ programs, assignment to @code{this} is permitted with
741 @samp{-traditional}. (The option @samp{-fthis-is-variable} also has
745 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
746 if your program uses names that are normally GNU C builtin functions for
747 other purposes of its own.
749 You cannot use @samp{-traditional} if you include any header files that
750 rely on ANSI C features. Some vendors are starting to ship systems with
751 ANSI C header files and you cannot use @samp{-traditional} on such
752 systems to compile files that include any system headers.
754 The @samp{-traditional} option also enables the @samp{-traditional-cpp}
755 option, which is described next.
757 @item -traditional-cpp
758 Attempt to support some aspects of traditional C preprocessors.
763 Comments convert to nothing at all, rather than to a space. This allows
764 traditional token concatenation.
767 In a preprocessing directive, the @samp{#} symbol must appear as the first
771 Macro arguments are recognized within string constants in a macro
772 definition (and their values are stringified, though without additional
773 quote marks, when they appear in such a context). The preprocessor
774 always considers a string constant to end at a newline.
777 @cindex detecting @w{@samp{-traditional}}
778 The predefined macro @code{__STDC__} is not defined when you use
779 @samp{-traditional}, but @code{__GNUC__} is (since the GNU extensions
780 which @code{__GNUC__} indicates are not affected by
781 @samp{-traditional}). If you need to write header files that work
782 differently depending on whether @samp{-traditional} is in use, by
783 testing both of these predefined macros you can distinguish four
784 situations: GNU C, traditional GNU C, other ANSI C compilers, and other
785 old C compilers. The predefined macro @code{__STDC_VERSION__} is also
786 not defined when you use @samp{-traditional}. @xref{Standard
787 Predefined,,Standard Predefined Macros,cpp.info,The C Preprocessor},
788 for more discussion of these and other predefined macros.
791 @cindex string constants vs newline
792 @cindex newline vs string constants
793 The preprocessor considers a string constant to end at a newline (unless
794 the newline is escaped with @samp{\}). (Without @w{@samp{-traditional}},
795 string constants can contain the newline character as typed.)
798 @item -fcond-mismatch
799 Allow conditional expressions with mismatched types in the second and
800 third arguments. The value of such an expression is void.
802 @item -funsigned-char
803 Let the type @code{char} be unsigned, like @code{unsigned char}.
805 Each kind of machine has a default for what @code{char} should
806 be. It is either like @code{unsigned char} by default or like
807 @code{signed char} by default.
809 Ideally, a portable program should always use @code{signed char} or
810 @code{unsigned char} when it depends on the signedness of an object.
811 But many programs have been written to use plain @code{char} and
812 expect it to be signed, or expect it to be unsigned, depending on the
813 machines they were written for. This option, and its inverse, let you
814 make such a program work with the opposite default.
816 The type @code{char} is always a distinct type from each of
817 @code{signed char} or @code{unsigned char}, even though its behavior
818 is always just like one of those two.
821 Let the type @code{char} be signed, like @code{signed char}.
823 Note that this is equivalent to @samp{-fno-unsigned-char}, which is
824 the negative form of @samp{-funsigned-char}. Likewise, the option
825 @samp{-fno-signed-char} is equivalent to @samp{-funsigned-char}.
827 You may wish to use @samp{-fno-builtin} as well as @samp{-traditional}
828 if your program uses names that are normally GNU C builtin functions for
829 other purposes of its own.
831 You cannot use @samp{-traditional} if you include any header files that
832 rely on ANSI C features. Some vendors are starting to ship systems with
833 ANSI C header files and you cannot use @samp{-traditional} on such
834 systems to compile files that include any system headers.
836 @item -fsigned-bitfields
837 @itemx -funsigned-bitfields
838 @itemx -fno-signed-bitfields
839 @itemx -fno-unsigned-bitfields
840 These options control whether a bitfield is signed or unsigned, when the
841 declaration does not use either @code{signed} or @code{unsigned}. By
842 default, such a bitfield is signed, because this is consistent: the
843 basic integer types such as @code{int} are signed types.
845 However, when @samp{-traditional} is used, bitfields are all unsigned
848 @item -fwritable-strings
849 Store string constants in the writable data segment and don't uniquize
850 them. This is for compatibility with old programs which assume they can
851 write into string constants. The option @samp{-traditional} also has
854 Writing into string constants is a very bad idea; ``constants'' should
857 @item -fallow-single-precision
858 Do not promote single precision math operations to double precision,
859 even when compiling with @samp{-traditional}.
861 Traditional K&R C promotes all floating point operations to double
862 precision, regardless of the sizes of the operands. On the
863 architecture for which you are compiling, single precision may be faster
864 than double precision. If you must use @samp{-traditional}, but want
865 to use single precision operations when the operands are single
866 precision, use this option. This option has no effect when compiling
867 with ANSI or GNU C conventions (the default).
871 @node C++ Dialect Options
872 @section Options Controlling C++ Dialect
874 @cindex compiler options, C++
875 @cindex C++ options, command line
877 This section describes the command-line options that are only meaningful
878 for C++ programs; but you can also use most of the GNU compiler options
879 regardless of what language your program is in. For example, you
880 might compile a file @code{firstClass.C} like this:
883 g++ -g -felide-constructors -O -c firstClass.C
887 In this example, only @samp{-felide-constructors} is an option meant
888 only for C++ programs; you can use the other options with any
889 language supported by GNU CC.
891 Here is a list of options that are @emph{only} for compiling C++ programs:
894 @item -fno-access-control
895 Turn off all access checking. This switch is mainly useful for working
896 around bugs in the access control code.
899 Treat all possible member functions as virtual, implicitly.
900 All member functions (except for constructor functions and @code{new} or
901 @code{delete} member operators) are treated as virtual functions of the
902 class where they appear.
904 This does not mean that all calls to these member functions will be made
905 through the internal table of virtual functions. Under some
906 circumstances, the compiler can determine that a call to a given virtual
907 function can be made directly; in these cases the calls are direct in
911 Check that the pointer returned by @code{operator new} is non-null
912 before attempting to modify the storage allocated. The current Working
913 Paper requires that @code{operator new} never return a null pointer, so
914 this check is normally unnecessary.
916 @item -fconserve-space
917 Put uninitialized or runtime-initialized global variables into the
918 common segment, as C does. This saves space in the executable at the
919 cost of not diagnosing duplicate definitions. If you compile with this
920 flag and your program mysteriously crashes after @code{main()} has
921 completed, you may have an object that is being destroyed twice because
922 two definitions were merged.
924 @item -fdollars-in-identifiers
925 Accept @samp{$} in identifiers. You can also explicitly prohibit use of
926 @samp{$} with the option @samp{-fno-dollars-in-identifiers}. (GNU C allows
927 @samp{$} by default on most target systems, but there are a few exceptions.)
928 Traditional C allowed the character @samp{$} to form part of
929 identifiers. However, ANSI C and C++ forbid @samp{$} in identifiers.
931 @item -fenum-int-equiv
932 Anachronistically permit implicit conversion of @code{int} to
933 enumeration types. Current C++ allows conversion of @code{enum} to
934 @code{int}, but not the other way around.
936 @item -fexternal-templates
937 Cause template instantiations to obey @samp{#pragma interface} and
938 @samp{implementation}; template instances are emitted or not according
939 to the location of the template definition. @xref{Template
940 Instantiation}, for more information.
942 This option is deprecated.
944 @item -falt-external-templates
945 Similar to -fexternal-templates, but template instances are emitted or
946 not according to the place where they are first instantiated.
947 @xref{Template Instantiation}, for more information.
949 This option is deprecated.
952 @itemx -fno-for-scope
953 If -ffor-scope is specified, the scope of variables declared in
954 a @i{for-init-statement} is limited to the @samp{for} loop itself,
955 as specified by the draft C++ standard.
956 If -fno-for-scope is specified, the scope of variables declared in
957 a @i{for-init-statement} extends to the end of the enclosing scope,
958 as was the case in old versions of gcc, and other (traditional)
959 implementations of C++.
961 The default if neither flag is given to follow the standard,
962 but to allow and give a warning for old-style code that would
963 otherwise be invalid, or have different behavior.
965 @item -fno-gnu-keywords
966 Do not recognize @code{classof}, @code{headof}, @code{signature},
967 @code{sigof} or @code{typeof} as a keyword, so that code can use these
968 words as identifiers. You can use the keywords @code{__classof__},
969 @code{__headof__}, @code{__signature__}, @code{__sigof__}, and
970 @code{__typeof__} instead. @samp{-ansi} implies
971 @samp{-fno-gnu-keywords}.
973 @item -fguiding-decls
974 Treat a function declaration with the same type as a potential function
975 template instantiation as though it declares that instantiation, not a
976 normal function. If a definition is given for the function later in the
977 translation unit (or another translation unit if the target supports
978 weak symbols), that definition will be used; otherwise the template will
979 be instantiated. This behavior reflects the C++ language prior to
980 September 1996, when guiding declarations were removed.
982 This option implies @samp{-fname-mangling-version-0}, and will not work
983 with other name mangling versions.
985 @item -fno-implicit-templates
986 Never emit code for templates which are instantiated implicitly (i.e. by
987 use); only emit code for explicit instantiations. @xref{Template
988 Instantiation}, for more information.
990 @item -fhandle-signatures
991 Recognize the @code{signature} and @code{sigof} keywords for specifying
992 abstract types. The default (@samp{-fno-handle-signatures}) is not to
993 recognize them. @xref{C++ Signatures, Type Abstraction using
997 Support virtual function calls for objects that exceed the size
998 representable by a @samp{short int}. Users should not use this flag by
999 default; if you need to use it, the compiler will tell you so. If you
1000 compile any of your code with this flag, you must compile @emph{all} of
1001 your code with this flag (including libg++, if you use it).
1003 This flag is not useful when compiling with -fvtable-thunks.
1005 @item -fno-implement-inlines
1006 To save space, do not emit out-of-line copies of inline functions
1007 controlled by @samp{#pragma implementation}. This will cause linker
1008 errors if these functions are not inlined everywhere they are called.
1010 @item -fmemoize-lookups
1011 @itemx -fsave-memoized
1012 Use heuristics to compile faster. These heuristics are not enabled by
1013 default, since they are only effective for certain input files. Other
1014 input files compile more slowly.
1016 The first time the compiler must build a call to a member function (or
1017 reference to a data member), it must (1) determine whether the class
1018 implements member functions of that name; (2) resolve which member
1019 function to call (which involves figuring out what sorts of type
1020 conversions need to be made); and (3) check the visibility of the member
1021 function to the caller. All of this adds up to slower compilation.
1022 Normally, the second time a call is made to that member function (or
1023 reference to that data member), it must go through the same lengthy
1024 process again. This means that code like this:
1027 cout << "This " << p << " has " << n << " legs.\n";
1031 makes six passes through all three steps. By using a software cache, a
1032 ``hit'' significantly reduces this cost. Unfortunately, using the cache
1033 introduces another layer of mechanisms which must be implemented, and so
1034 incurs its own overhead. @samp{-fmemoize-lookups} enables the software
1037 Because access privileges (visibility) to members and member functions
1038 may differ from one function context to the next, G++ may need to flush
1039 the cache. With the @samp{-fmemoize-lookups} flag, the cache is flushed
1040 after every function that is compiled. The @samp{-fsave-memoized} flag
1041 enables the same software cache, but when the compiler determines that
1042 the context of the last function compiled would yield the same access
1043 privileges of the next function to compile, it preserves the cache.
1044 This is most helpful when defining many member functions for the same
1045 class: with the exception of member functions which are friends of other
1046 classes, each member function has exactly the same access privileges as
1047 every other, and the cache need not be flushed.
1049 The code that implements these flags has rotted; you should probably
1052 @item -fstrict-prototype
1053 Within an @samp{extern "C"} linkage specification, treat a function
1054 declaration with no arguments, such as @samp{int foo ();}, as declaring
1055 the function to take no arguments. Normally, such a declaration means
1056 that the function @code{foo} can take any combination of arguments, as
1057 in C. @samp{-pedantic} implies @samp{-fstrict-prototype} unless
1058 overridden with @samp{-fno-strict-prototype}.
1060 This flag no longer affects declarations with C++ linkage.
1062 @item -fname-mangling-version-@var{n}
1063 Control the way in which names are mangled. Version 0 is compatible
1064 with versions of g++ before 2.8. Version 1 is the default. Version 1
1065 will allow correct mangling of function templates. For example,
1066 version 0 mangling does not mangle foo<int, double> and foo<int, char>
1067 given this declaration:
1070 template <class T, class U> void foo(T t);
1073 @item -fno-nonnull-objects
1074 Don't assume that a reference is initialized to refer to a valid object.
1075 Although the current C++ Working Paper prohibits null references, some
1076 old code may rely on them, and you can use @samp{-fno-nonnull-objects}
1077 to turn on checking.
1079 At the moment, the compiler only does this checking for conversions to
1080 virtual base classes.
1082 @item -foperator-names
1083 Recognize the operator name keywords @code{and}, @code{bitand},
1084 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1085 synonyms for the symbols they refer to. @samp{-ansi} implies
1086 @samp{-foperator-names}.
1089 Enable automatic template instantiation. This option also implies
1090 @samp{-fno-implicit-templates}. @xref{Template Instantiation}, for more
1093 @item -fthis-is-variable
1094 Permit assignment to @code{this}. The incorporation of user-defined
1095 free store management into C++ has made assignment to @samp{this} an
1096 anachronism. Therefore, by default it is invalid to assign to
1097 @code{this} within a class member function; that is, GNU C++ treats
1098 @samp{this} in a member function of class @code{X} as a non-lvalue of
1099 type @samp{X *}. However, for backwards compatibility, you can make it
1100 valid with @samp{-fthis-is-variable}.
1102 @item -fvtable-thunks
1103 Use @samp{thunks} to implement the virtual function dispatch table
1104 (@samp{vtable}). The traditional (cfront-style) approach to
1105 implementing vtables was to store a pointer to the function and two
1106 offsets for adjusting the @samp{this} pointer at the call site. Newer
1107 implementations store a single pointer to a @samp{thunk} function which
1108 does any necessary adjustment and then calls the target function.
1110 This option also enables a heuristic for controlling emission of
1111 vtables; if a class has any non-inline virtual functions, the vtable
1112 will be emitted in the translation unit containing the first one of
1115 @item -ftemplate-depth-@var{n}
1116 Set the maximum instantiation depth for template classes to @var{n}.
1117 A limit on the template instantiation depth is needed to detect
1118 endless recursions during template class instantiation. ANSI/ISO C++
1119 conforming programs must not rely on a maximum depth greater than 17.
1122 Do not search for header files in the standard directories specific to
1123 C++, but do still search the other standard directories. (This option
1124 is used when building libg++.)
1127 For C++ programs (in addition to the effects that apply to both C and
1128 C++), this has the same effect as @samp{-fthis-is-variable}.
1129 @xref{C Dialect Options,, Options Controlling C Dialect}.
1132 In addition, these optimization, warning, and code generation options
1133 have meanings only for C++ programs:
1136 @item -fno-default-inline
1137 Do not assume @samp{inline} for functions defined inside a class scope.
1138 @xref{Optimize Options,,Options That Control Optimization}.
1140 @item -Wold-style-cast
1141 @itemx -Woverloaded-virtual
1142 @itemx -Wtemplate-debugging
1143 Warnings that apply only to C++ programs. @xref{Warning
1144 Options,,Options to Request or Suppress Warnings}.
1147 Warn about violation of some style rules from Effective C++ by Scott Myers.
1150 Control how virtual function definitions are used, in a fashion
1151 compatible with @code{cfront} 1.x. @xref{Code Gen Options,,Options for
1152 Code Generation Conventions}.
1155 @node Warning Options
1156 @section Options to Request or Suppress Warnings
1157 @cindex options to control warnings
1158 @cindex warning messages
1159 @cindex messages, warning
1160 @cindex suppressing warnings
1162 Warnings are diagnostic messages that report constructions which
1163 are not inherently erroneous but which are risky or suggest there
1164 may have been an error.
1166 You can request many specific warnings with options beginning @samp{-W},
1167 for example @samp{-Wimplicit} to request warnings on implicit
1168 declarations. Each of these specific warning options also has a
1169 negative form beginning @samp{-Wno-} to turn off warnings;
1170 for example, @samp{-Wno-implicit}. This manual lists only one of the
1171 two forms, whichever is not the default.
1173 These options control the amount and kinds of warnings produced by GNU
1177 @cindex syntax checking
1179 Check the code for syntax errors, but don't do anything beyond that.
1182 Issue all the warnings demanded by strict ANSI standard C; reject
1183 all programs that use forbidden extensions.
1185 Valid ANSI standard C programs should compile properly with or without
1186 this option (though a rare few will require @samp{-ansi}). However,
1187 without this option, certain GNU extensions and traditional C features
1188 are supported as well. With this option, they are rejected.
1190 @samp{-pedantic} does not cause warning messages for use of the
1191 alternate keywords whose names begin and end with @samp{__}. Pedantic
1192 warnings are also disabled in the expression that follows
1193 @code{__extension__}. However, only system header files should use
1194 these escape routes; application programs should avoid them.
1195 @xref{Alternate Keywords}.
1197 This option is not intended to be @i{useful}; it exists only to satisfy
1198 pedants who would otherwise claim that GNU CC fails to support the ANSI
1201 Some users try to use @samp{-pedantic} to check programs for strict ANSI
1202 C conformance. They soon find that it does not do quite what they want:
1203 it finds some non-ANSI practices, but not all---only those for which
1204 ANSI C @emph{requires} a diagnostic.
1206 A feature to report any failure to conform to ANSI C might be useful in
1207 some instances, but would require considerable additional work and would
1208 be quite different from @samp{-pedantic}. We recommend, rather, that
1209 users take advantage of the extensions of GNU C and disregard the
1210 limitations of other compilers. Aside from certain supercomputers and
1211 obsolete small machines, there is less and less reason ever to use any
1212 other C compiler other than for bootstrapping GNU CC.
1214 @item -pedantic-errors
1215 Like @samp{-pedantic}, except that errors are produced rather than
1219 Inhibit all warning messages.
1222 Inhibit warning messages about the use of @samp{#import}.
1224 @item -Wchar-subscripts
1225 Warn if an array subscript has type @code{char}. This is a common cause
1226 of error, as programmers often forget that this type is signed on some
1230 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
1231 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
1234 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
1235 the arguments supplied have types appropriate to the format string
1238 @item -Wimplicit-int
1239 Warn when a declaration does not specify a type.
1241 @item -Wimplicit-function-declarations
1242 Warn whenever a function is used before being declared.
1245 Same as @samp{-Wimplicit-int} @samp{-Wimplicit-function-declaration}.
1248 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
1249 function with external linkage, returning int, taking either zero
1250 arguments, two, or three arguments of appropriate types.
1253 Warn if parentheses are omitted in certain contexts, such
1254 as when there is an assignment in a context where a truth value
1255 is expected, or when operators are nested whose precedence people
1256 often get confused about.
1258 Also warn about constructions where there may be confusion to which
1259 @code{if} statement an @code{else} branch belongs. Here is an example of
1272 In C, every @code{else} branch belongs to the innermost possible @code{if}
1273 statement, which in this example is @code{if (b)}. This is often not
1274 what the programmer expected, as illustrated in the above example by
1275 indentation the programmer chose. When there is the potential for this
1276 confusion, GNU C will issue a warning when this flag is specified.
1277 To eliminate the warning, add explicit braces around the innermost
1278 @code{if} statement so there is no way the @code{else} could belong to
1279 the enclosing @code{if}. The resulting code would look like this:
1294 Warn whenever a function is defined with a return-type that defaults
1295 to @code{int}. Also warn about any @code{return} statement with no
1296 return-value in a function whose return-type is not @code{void}.
1299 Warn whenever a @code{switch} statement has an index of enumeral type
1300 and lacks a @code{case} for one or more of the named codes of that
1301 enumeration. (The presence of a @code{default} label prevents this
1302 warning.) @code{case} labels outside the enumeration range also
1303 provoke warnings when this option is used.
1306 Warn if any trigraphs are encountered (assuming they are enabled).
1309 Warn whenever a variable is unused aside from its declaration,
1310 whenever a function is declared static but never defined, whenever a
1311 label is declared but not used, and whenever a statement computes a
1312 result that is explicitly not used.
1314 In order to get a warning about an unused function parameter, you must
1315 specify both @samp{-W} and @samp{-Wunused}.
1317 To suppress this warning for an expression, simply cast it to void. For
1318 unused variables and parameters, use the @samp{unused} attribute
1319 (@pxref{Variable Attributes}).
1321 @item -Wuninitialized
1322 An automatic variable is used without first being initialized.
1324 These warnings are possible only in optimizing compilation,
1325 because they require data flow information that is computed only
1326 when optimizing. If you don't specify @samp{-O}, you simply won't
1329 These warnings occur only for variables that are candidates for
1330 register allocation. Therefore, they do not occur for a variable that
1331 is declared @code{volatile}, or whose address is taken, or whose size
1332 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
1333 structures, unions or arrays, even when they are in registers.
1335 Note that there may be no warning about a variable that is used only
1336 to compute a value that itself is never used, because such
1337 computations may be deleted by data flow analysis before the warnings
1340 These warnings are made optional because GNU CC is not smart
1341 enough to see all the reasons why the code might be correct
1342 despite appearing to have an error. Here is one example of how
1361 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
1362 always initialized, but GNU CC doesn't know this. Here is
1363 another common case:
1368 if (change_y) save_y = y, y = new_y;
1370 if (change_y) y = save_y;
1375 This has no bug because @code{save_y} is used only if it is set.
1377 Some spurious warnings can be avoided if you declare all the functions
1378 you use that never return as @code{noreturn}. @xref{Function
1381 @item -Wreorder (C++ only)
1382 @cindex reordering, warning
1383 @cindex warning for reordering of member initializers
1384 Warn when the order of member initializers given in the code does not
1385 match the order in which they must be executed. For instance:
1391 A(): j (0), i (1) @{ @}
1395 Here the compiler will warn that the member initializers for @samp{i}
1396 and @samp{j} will be rearranged to match the declaration order of the
1399 @item -Wtemplate-debugging
1400 @cindex template debugging
1401 When using templates in a C++ program, warn if debugging is not yet
1402 fully available (C++ only).
1404 @item -Wunknown-pragmas
1405 @cindex warning for unknown pragmas
1406 @cindex unknown pragmas, warning
1407 @cindex pragmas, warning of unknown
1408 Warn when a #pragma directive is encountered which is not understood by
1409 GCC. If this command line option is used, warnings will even be issued
1410 for unknown pragmas in system header files. This is not the case if
1411 the warnings were only enabled by the @samp{-Wall} command line option.
1414 All of the above @samp{-W} options combined. This enables all the
1415 warnings about constructions that some users consider questionable, and
1416 that are easy to avoid (or modify to prevent the warning), even in
1417 conjunction with macros.
1420 The following @samp{-W@dots{}} options are not implied by @samp{-Wall}.
1421 Some of them warn about constructions that users generally do not
1422 consider questionable, but which occasionally you might wish to check
1423 for; others warn about constructions that are necessary or hard to avoid
1424 in some cases, and there is no simple way to modify the code to suppress
1429 Print extra warning messages for these events:
1432 @cindex @code{longjmp} warnings
1434 A nonvolatile automatic variable might be changed by a call to
1435 @code{longjmp}. These warnings as well are possible only in
1436 optimizing compilation.
1438 The compiler sees only the calls to @code{setjmp}. It cannot know
1439 where @code{longjmp} will be called; in fact, a signal handler could
1440 call it at any point in the code. As a result, you may get a warning
1441 even when there is in fact no problem because @code{longjmp} cannot
1442 in fact be called at the place which would cause a problem.
1445 A function can return either with or without a value. (Falling
1446 off the end of the function body is considered returning without
1447 a value.) For example, this function would evoke such a
1461 An expression-statement or the left-hand side of a comma expression
1462 contains no side effects.
1463 To suppress the warning, cast the unused expression to void.
1464 For example, an expression such as @samp{x[i,j]} will cause a warning,
1465 but @samp{x[(void)i,j]} will not.
1468 An unsigned value is compared against zero with @samp{<} or @samp{<=}.
1471 A comparison like @samp{x<=y<=z} appears; this is equivalent to
1472 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
1473 that of ordinary mathematical notation.
1476 Storage-class specifiers like @code{static} are not the first things in
1477 a declaration. According to the C Standard, this usage is obsolescent.
1480 If @samp{-Wall} or @samp{-Wunused} is also specified, warn about unused
1484 A comparison between signed and unsigned values could produce an
1485 incorrect result when the signed value is converted to unsigned.
1486 (But do not warn if @samp{-Wno-sign-compare} is also specified.)
1489 An aggregate has a partly bracketed initializer.
1490 For example, the following code would evoke such a warning,
1491 because braces are missing around the initializer for @code{x.h}:
1494 struct s @{ int f, g; @};
1495 struct t @{ struct s h; int i; @};
1496 struct t x = @{ 1, 2, 3 @};
1501 Warn about certain constructs that behave differently in traditional and
1506 Macro arguments occurring within string constants in the macro body.
1507 These would substitute the argument in traditional C, but are part of
1508 the constant in ANSI C.
1511 A function declared external in one block and then used after the end of
1515 A @code{switch} statement has an operand of type @code{long}.
1519 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
1522 Warn whenever a local variable shadows another local variable.
1524 @item -Wid-clash-@var{len}
1525 Warn whenever two distinct identifiers match in the first @var{len}
1526 characters. This may help you prepare a program that will compile
1527 with certain obsolete, brain-damaged compilers.
1529 @item -Wlarger-than-@var{len}
1530 Warn whenever an object of larger than @var{len} bytes is defined.
1532 @item -Wpointer-arith
1533 Warn about anything that depends on the ``size of'' a function type or
1534 of @code{void}. GNU C assigns these types a size of 1, for
1535 convenience in calculations with @code{void *} pointers and pointers
1538 @item -Wbad-function-cast
1539 Warn whenever a function call is cast to a non-matching type.
1540 For example, warn if @code{int malloc()} is cast to @code{anything *}.
1543 Warn whenever a pointer is cast so as to remove a type qualifier from
1544 the target type. For example, warn if a @code{const char *} is cast
1545 to an ordinary @code{char *}.
1548 Warn whenever a pointer is cast such that the required alignment of the
1549 target is increased. For example, warn if a @code{char *} is cast to
1550 an @code{int *} on machines where integers can only be accessed at
1551 two- or four-byte boundaries.
1553 @item -Wwrite-strings
1554 Give string constants the type @code{const char[@var{length}]} so that
1555 copying the address of one into a non-@code{const} @code{char *}
1556 pointer will get a warning. These warnings will help you find at
1557 compile time code that can try to write into a string constant, but
1558 only if you have been very careful about using @code{const} in
1559 declarations and prototypes. Otherwise, it will just be a nuisance;
1560 this is why we did not make @samp{-Wall} request these warnings.
1563 Warn if a prototype causes a type conversion that is different from what
1564 would happen to the same argument in the absence of a prototype. This
1565 includes conversions of fixed point to floating and vice versa, and
1566 conversions changing the width or signedness of a fixed point argument
1567 except when the same as the default promotion.
1569 Also, warn if a negative integer constant expression is implicitly
1570 converted to an unsigned type. For example, warn about the assignment
1571 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
1572 casts like @code{(unsigned) -1}.
1574 @item -Wsign-compare
1575 @cindex warning for comparison of signed and unsigned values
1576 @cindex comparison of signed and unsigned values, warning
1577 @cindex signed and unsigned values, comparison warning
1578 Warn when a comparison between signed and unsigned values could produce
1579 an incorrect result when the signed value is converted to unsigned.
1580 This warning is also enabled by @samp{-W}; to get the other warnings
1581 of @samp{-W} without this warning, use @samp{-W -Wno-sign-compare}.
1583 @item -Waggregate-return
1584 Warn if any functions that return structures or unions are defined or
1585 called. (In languages where you can return an array, this also elicits
1588 @item -Wstrict-prototypes
1589 Warn if a function is declared or defined without specifying the
1590 argument types. (An old-style function definition is permitted without
1591 a warning if preceded by a declaration which specifies the argument
1594 @item -Wmissing-prototypes
1595 Warn if a global function is defined without a previous prototype
1596 declaration. This warning is issued even if the definition itself
1597 provides a prototype. The aim is to detect global functions that fail
1598 to be declared in header files.
1600 @item -Wmissing-declarations
1601 Warn if a global function is defined without a previous declaration.
1602 Do so even if the definition itself provides a prototype.
1603 Use this option to detect global functions that are not declared in
1606 @item -Wredundant-decls
1607 Warn if anything is declared more than once in the same scope, even in
1608 cases where multiple declaration is valid and changes nothing.
1610 @item -Wnested-externs
1611 Warn if an @code{extern} declaration is encountered within an function.
1614 Warn if a function can not be inlined, and either it was declared as inline,
1615 or else the @samp{-finline-functions} option was given.
1617 @item -Wold-style-cast
1618 Warn if an old-style (C-style) cast is used within a program.
1620 @item -Woverloaded-virtual
1621 @cindex overloaded virtual fn, warning
1622 @cindex warning for overloaded virtual fn
1623 Warn when a derived class function declaration may be an error in
1624 defining a virtual function (C++ only). In a derived class, the
1625 definitions of virtual functions must match the type signature of a
1626 virtual function declared in the base class. With this option, the
1627 compiler warns when you define a function with the same name as a
1628 virtual function, but with a type signature that does not match any
1629 declarations from the base class.
1631 @item -Wsynth (C++ only)
1632 @cindex warning for synthesized methods
1633 @cindex synthesized methods, warning
1634 Warn when g++'s synthesis behavior does not match that of cfront. For
1640 A& operator = (int);
1650 In this example, g++ will synthesize a default @samp{A& operator =
1651 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1654 Make all warnings into errors.
1657 @node Debugging Options
1658 @section Options for Debugging Your Program or GNU CC
1659 @cindex options, debugging
1660 @cindex debugging information options
1662 GNU CC has various special options that are used for debugging
1663 either your program or GCC:
1667 Produce debugging information in the operating system's native format
1668 (stabs, COFF, XCOFF, or DWARF). GDB can work with this debugging
1671 On most systems that use stabs format, @samp{-g} enables use of extra
1672 debugging information that only GDB can use; this extra information
1673 makes debugging work better in GDB but will probably make other debuggers
1675 refuse to read the program. If you want to control for certain whether
1676 to generate the extra information, use @samp{-gstabs+}, @samp{-gstabs},
1677 @samp{-gxcoff+}, @samp{-gxcoff}, @samp{-gdwarf-1+}, or @samp{-gdwarf-1}
1680 Unlike most other C compilers, GNU CC allows you to use @samp{-g} with
1681 @samp{-O}. The shortcuts taken by optimized code may occasionally
1682 produce surprising results: some variables you declared may not exist
1683 at all; flow of control may briefly move where you did not expect it;
1684 some statements may not be executed because they compute constant
1685 results or their values were already at hand; some statements may
1686 execute in different places because they were moved out of loops.
1688 Nevertheless it proves possible to debug optimized output. This makes
1689 it reasonable to use the optimizer for programs that might have bugs.
1691 The following options are useful when GNU CC is generated with the
1692 capability for more than one debugging format.
1695 Produce debugging information for use by GDB. This means to use the
1696 most expressive format available (DWARF 2, stabs, or the native format
1697 if neither of those are supported), including GDB extensions if at all
1701 Produce debugging information in stabs format (if that is supported),
1702 without GDB extensions. This is the format used by DBX on most BSD
1703 systems. On MIPS, Alpha and System V Release 4 systems this option
1704 produces stabs debugging output which is not understood by DBX or SDB.
1705 On System V Release 4 systems this option requires the GNU assembler.
1708 Produce debugging information in stabs format (if that is supported),
1709 using GNU extensions understood only by the GNU debugger (GDB). The
1710 use of these extensions is likely to make other debuggers crash or
1711 refuse to read the program.
1714 Produce debugging information in COFF format (if that is supported).
1715 This is the format used by SDB on most System V systems prior to
1719 Produce debugging information in XCOFF format (if that is supported).
1720 This is the format used by the DBX debugger on IBM RS/6000 systems.
1723 Produce debugging information in XCOFF format (if that is supported),
1724 using GNU extensions understood only by the GNU debugger (GDB). The
1725 use of these extensions is likely to make other debuggers crash or
1726 refuse to read the program, and may cause assemblers other than the GNU
1727 assembler (GAS) to fail with an error.
1730 Produce debugging information in DWARF version 1 format (if that is
1731 supported). This is the format used by SDB on most System V Release 4
1735 Produce debugging information in DWARF version 1 format (if that is
1736 supported), using GNU extensions understood only by the GNU debugger
1737 (GDB). The use of these extensions is likely to make other debuggers
1738 crash or refuse to read the program.
1741 Produce debugging information in DWARF version 2 format (if that is
1742 supported). This is the format used by DBX on IRIX 6.
1745 @itemx -ggdb@var{level}
1746 @itemx -gstabs@var{level}
1747 @itemx -gcoff@var{level}
1748 @itemx -gxcoff@var{level}
1749 @itemx -gdwarf@var{level}
1750 @itemx -gdwarf-2@var{level}
1751 Request debugging information and also use @var{level} to specify how
1752 much information. The default level is 2.
1754 Level 1 produces minimal information, enough for making backtraces in
1755 parts of the program that you don't plan to debug. This includes
1756 descriptions of functions and external variables, but no information
1757 about local variables and no line numbers.
1759 Level 3 includes extra information, such as all the macro definitions
1760 present in the program. Some debuggers support macro expansion when
1765 Generate extra code to write profile information suitable for the
1766 analysis program @code{prof}. You must use this option when compiling
1767 the source files you want data about, and you must also use it when
1770 @cindex @code{gprof}
1772 Generate extra code to write profile information suitable for the
1773 analysis program @code{gprof}. You must use this option when compiling
1774 the source files you want data about, and you must also use it when
1779 Generate extra code to write profile information for basic blocks, which will
1780 record the number of times each basic block is executed, the basic block start
1781 address, and the function name containing the basic block. If @samp{-g} is
1782 used, the line number and filename of the start of the basic block will also be
1783 recorded. If not overridden by the machine description, the default action is
1784 to append to the text file @file{bb.out}.
1786 This data could be analyzed by a program like @code{tcov}. Note,
1787 however, that the format of the data is not what @code{tcov} expects.
1788 Eventually GNU @code{gprof} should be extended to process this data.
1791 Makes the compiler print out each function name as it is compiled, and
1792 print some statistics about each pass when it finishes.
1795 Generate extra code to profile basic blocks. Your executable will
1796 produce output that is a superset of that produced when @samp{-a} is
1797 used. Additional output is the source and target address of the basic
1798 blocks where a jump takes place, the number of times a jump is executed,
1799 and (optionally) the complete sequence of basic blocks being executed.
1800 The output is appended to file @file{bb.out}.
1802 You can examine different profiling aspects without recompilation. Your
1803 executable will read a list of function names from file @file{bb.in}.
1804 Profiling starts when a function on the list is entered and stops when
1805 that invocation is exited. To exclude a function from profiling, prefix
1806 its name with `-'. If a function name is not unique, you can
1807 disambiguate it by writing it in the form
1808 @samp{/path/filename.d:functionname}. Your executable will write the
1809 available paths and filenames in file @file{bb.out}.
1811 Several function names have a special meaning:
1814 Write source, target and frequency of jumps to file @file{bb.out}.
1815 @item __bb_hidecall__
1816 Exclude function calls from frequency count.
1817 @item __bb_showret__
1818 Include function returns in frequency count.
1820 Write the sequence of basic blocks executed to file @file{bbtrace.gz}.
1821 The file will be compressed using the program @samp{gzip}, which must
1822 exist in your @code{PATH}. On systems without the @samp{popen}
1823 function, the file will be named @file{bbtrace} and will not be
1824 compressed. @strong{Profiling for even a few seconds on these systems
1825 will produce a very large file.} Note: @code{__bb_hidecall__} and
1826 @code{__bb_showret__} will not affect the sequence written to
1830 Here's a short example using different profiling parameters
1831 in file @file{bb.in}. Assume function @code{foo} consists of basic blocks
1832 1 and 2 and is called twice from block 3 of function @code{main}. After
1833 the calls, block 3 transfers control to block 4 of @code{main}.
1835 With @code{__bb_trace__} and @code{main} contained in file @file{bb.in},
1836 the following sequence of blocks is written to file @file{bbtrace.gz}:
1837 0 3 1 2 1 2 4. The return from block 2 to block 3 is not shown, because
1838 the return is to a point inside the block and not to the top. The
1839 block address 0 always indicates, that control is transferred
1840 to the trace from somewhere outside the observed functions. With
1841 @samp{-foo} added to @file{bb.in}, the blocks of function
1842 @code{foo} are removed from the trace, so only 0 3 4 remains.
1844 With @code{__bb_jumps__} and @code{main} contained in file @file{bb.in},
1845 jump frequencies will be written to file @file{bb.out}. The
1846 frequencies are obtained by constructing a trace of blocks
1847 and incrementing a counter for every neighbouring pair of blocks
1848 in the trace. The trace 0 3 1 2 1 2 4 displays the following
1852 Jump from block 0x0 to block 0x3 executed 1 time(s)
1853 Jump from block 0x3 to block 0x1 executed 1 time(s)
1854 Jump from block 0x1 to block 0x2 executed 2 time(s)
1855 Jump from block 0x2 to block 0x1 executed 1 time(s)
1856 Jump from block 0x2 to block 0x4 executed 1 time(s)
1859 With @code{__bb_hidecall__}, control transfer due to call instructions
1860 is removed from the trace, that is the trace is cut into three parts: 0
1861 3 4, 0 1 2 and 0 1 2. With @code{__bb_showret__}, control transfer due
1862 to return instructions is added to the trace. The trace becomes: 0 3 1
1863 2 3 1 2 3 4. Note, that this trace is not the same, as the sequence
1864 written to @file{bbtrace.gz}. It is solely used for counting jump
1867 @item -fprofile-arcs
1868 Instrument @dfn{arcs} during compilation. For each function of your
1869 program, GNU CC creates a program flow graph, then finds a spanning tree
1870 for the graph. Only arcs that are not on the spanning tree have to be
1871 instrumented: the compiler adds code to count the number of times that these
1872 arcs are executed. When an arc is the only exit or only entrance to a
1873 block, the instrumentation code can be added to the block; otherwise, a
1874 new basic block must be created to hold the instrumentation code.
1876 Since not every arc in the program must be instrumented, programs
1877 compiled with this option run faster than programs compiled with
1878 @samp{-a}, which adds instrumentation code to every basic block in the
1879 program. The tradeoff: since @code{gcov} does not have
1880 execution counts for all branches, it must start with the execution
1881 counts for the instrumented branches, and then iterate over the program
1882 flow graph until the entire graph has been solved. Hence, @code{gcov}
1883 runs a little more slowly than a program which uses information from
1886 @samp{-fprofile-arcs} also makes it possible to estimate branch
1887 probabilities, and to calculate basic block execution counts. In
1888 general, basic block execution counts do not give enough information to
1889 estimate all branch probabilities. When the compiled program exits, it
1890 saves the arc execution counts to a file called
1891 @file{@var{sourcename}.da}. Use the compiler option
1892 @samp{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
1893 Control Optimization}) when recompiling, to optimize using estimated
1894 branch probabilities.
1897 @item -ftest-coverage
1898 Create data files for the @code{gcov} code-coverage utility
1899 (@pxref{Gcov,, @code{gcov}: a GNU CC Test Coverage Program}).
1900 The data file names begin with the name of your source file:
1903 @item @var{sourcename}.bb
1904 A mapping from basic blocks to line numbers, which @code{gcov} uses to
1905 associate basic block execution counts with line numbers.
1907 @item @var{sourcename}.bbg
1908 A list of all arcs in the program flow graph. This allows @code{gcov}
1909 to reconstruct the program flow graph, so that it can compute all basic
1910 block and arc execution counts from the information in the
1911 @code{@var{sourcename}.da} file (this last file is the output from
1912 @samp{-fprofile-arcs}).
1916 Makes the compiler print out each function name as it is compiled, and
1917 print some statistics about each pass when it finishes.
1919 @item -d@var{letters}
1920 Says to make debugging dumps during compilation at times specified by
1921 @var{letters}. This is used for debugging the compiler. The file names
1922 for most of the dumps are made by appending a word to the source file
1923 name (e.g. @file{foo.c.rtl} or @file{foo.c.jump}). Here are the
1924 possible letters for use in @var{letters}, and their meanings:
1928 Dump after computing branch probabilities, to @file{@var{file}.bp}.
1930 Dump after instruction combination, to the file @file{@var{file}.combine}.
1932 Dump after delayed branch scheduling, to @file{@var{file}.dbr}.
1934 Dump after purging ADDRESSOF, to @file{@var{file}.addressof}.
1936 Dump after flow analysis, to @file{@var{file}.flow}.
1938 Dump after global register allocation, to @file{@var{file}.greg}.
1940 Dump after first jump optimization, to @file{@var{file}.jump}.
1942 Dump after last jump optimization, to @file{@var{file}.jump2}.
1944 Dump after conversion from registers to stack, to @file{@var{file}.stack}.
1946 Dump after local register allocation, to @file{@var{file}.lreg}.
1948 Dump after loop optimization, to @file{@var{file}.loop}.
1950 Dump after performing the machine dependent reorganisation pass, to
1951 @file{@var{file}.mach}.
1953 Dump after the register move pass, to @file{@var{file}.regmove}.
1955 Dump after RTL generation, to @file{@var{file}.rtl}.
1957 Dump after the second instruction scheduling pass, to @file{@var{file}.sched2}.
1959 Dump after CSE (including the jump optimization that sometimes follows
1960 CSE), to @file{@var{file}.cse}.
1962 Dump after the first instruction scheduling pass, to @file{@var{file}.sched}.
1964 Dump after the second CSE pass (including the jump optimization that
1965 sometimes follows CSE), to @file{@var{file}.cse2}.
1967 Just generate RTL for a function instead of compiling it. Usually used
1970 Produce all the dumps listed above.
1972 Print statistics on memory usage, at the end of the run, to
1975 Annotate the assembler output with a comment indicating which
1976 pattern and alternative was used.
1978 Dump debugging information during parsing, to standard error.
1980 Annotate the assembler output with miscellaneous debugging information.
1983 @item -fpretend-float
1984 When running a cross-compiler, pretend that the target machine uses the
1985 same floating point format as the host machine. This causes incorrect
1986 output of the actual floating constants, but the actual instruction
1987 sequence will probably be the same as GNU CC would make when running on
1991 Store the usual ``temporary'' intermediate files permanently; place them
1992 in the current directory and name them based on the source file. Thus,
1993 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
1994 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}.
1996 @item -print-file-name=@var{library}
1997 Print the full absolute name of the library file @var{library} that
1998 would be used when linking---and don't do anything else. With this
1999 option, GNU CC does not compile or link anything; it just prints the
2002 @item -print-prog-name=@var{program}
2003 Like @samp{-print-file-name}, but searches for a program such as @samp{cpp}.
2005 @item -print-libgcc-file-name
2006 Same as @samp{-print-file-name=libgcc.a}.
2008 This is useful when you use @samp{-nostdlib} or @samp{-nodefaultlibs}
2009 but you do want to link with @file{libgcc.a}. You can do
2012 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
2015 @item -print-search-dirs
2016 Print the name of the configured installation directory and a list of
2017 program and library directories gcc will search---and don't do anything else.
2019 This is useful when gcc prints the error message
2020 @samp{installation problem, cannot exec cpp: No such file or directory}.
2021 To resolve this you either need to put @file{cpp} and the other compiler
2022 components where gcc expects to find them, or you can set the environment
2023 variable @code{GCC_EXEC_PREFIX} to the directory where you installed them.
2024 Don't forget the trailing '/'.
2025 @xref{Environment Variables}.
2028 @node Optimize Options
2029 @section Options That Control Optimization
2030 @cindex optimize options
2031 @cindex options, optimization
2033 These options control various sorts of optimizations:
2038 Optimize. Optimizing compilation takes somewhat more time, and a lot
2039 more memory for a large function.
2041 Without @samp{-O}, the compiler's goal is to reduce the cost of
2042 compilation and to make debugging produce the expected results.
2043 Statements are independent: if you stop the program with a breakpoint
2044 between statements, you can then assign a new value to any variable or
2045 change the program counter to any other statement in the function and
2046 get exactly the results you would expect from the source code.
2048 Without @samp{-O}, the compiler only allocates variables declared
2049 @code{register} in registers. The resulting compiled code is a little
2050 worse than produced by PCC without @samp{-O}.
2052 With @samp{-O}, the compiler tries to reduce code size and execution
2055 When you specify @samp{-O}, the compiler turns on @samp{-fthread-jumps}
2056 and @samp{-fdefer-pop} on all machines. The compiler turns on
2057 @samp{-fdelayed-branch} on machines that have delay slots, and
2058 @samp{-fomit-frame-pointer} on machines that can support debugging even
2059 without a frame pointer. On some machines the compiler also turns
2060 on other flags.@refill
2063 Optimize even more. GNU CC performs nearly all supported optimizations
2064 that do not involve a space-speed tradeoff. The compiler does not
2065 perform loop unrolling or function inlining when you specify @samp{-O2}.
2066 As compared to @samp{-O}, this option increases both compilation time
2067 and the performance of the generated code.
2069 @samp{-O2} turns on all optional optimizations except for loop unrolling
2070 and function inlining. It also turns on the @samp{-fforce-mem} option
2071 on all machines and frame pointer elimination on machines where doing so
2072 does not interfere with debugging.
2075 Optimize yet more. @samp{-O3} turns on all optimizations specified by
2076 @samp{-O2} and also turns on the @samp{inline-functions} option.
2082 Optimize for size. @samp{-Os} enables all @samp{-O2} optimizations that
2083 do not typically increase code size. It also performs further
2084 optimizations designed to reduce code size.
2086 If you use multiple @samp{-O} options, with or without level numbers,
2087 the last such option is the one that is effective.
2090 Options of the form @samp{-f@var{flag}} specify machine-independent
2091 flags. Most flags have both positive and negative forms; the negative
2092 form of @samp{-ffoo} would be @samp{-fno-foo}. In the table below,
2093 only one of the forms is listed---the one which is not the default.
2094 You can figure out the other form by either removing @samp{no-} or
2099 Do not store floating point variables in registers, and inhibit other
2100 options that might change whether a floating point value is taken from a
2103 @cindex floating point precision
2104 This option prevents undesirable excess precision on machines such as
2105 the 68000 where the floating registers (of the 68881) keep more
2106 precision than a @code{double} is supposed to have. Similarly for the
2107 x86 architecture. For most programs, the excess precision does only
2108 good, but a few programs rely on the precise definition of IEEE floating
2109 point. Use @samp{-ffloat-store} for such programs.
2111 @item -fno-default-inline
2112 Do not make member functions inline by default merely because they are
2113 defined inside the class scope (C++ only). Otherwise, when you specify
2114 @w{@samp{-O}}, member functions defined inside class scope are compiled
2115 inline by default; i.e., you don't need to add @samp{inline} in front of
2116 the member function name.
2118 @item -fno-defer-pop
2119 Always pop the arguments to each function call as soon as that function
2120 returns. For machines which must pop arguments after a function call,
2121 the compiler normally lets arguments accumulate on the stack for several
2122 function calls and pops them all at once.
2125 Force memory operands to be copied into registers before doing
2126 arithmetic on them. This produces better code by making all memory
2127 references potential common subexpressions. When they are not common
2128 subexpressions, instruction combination should eliminate the separate
2129 register-load. The @samp{-O2} option turns on this option.
2132 Force memory address constants to be copied into registers before
2133 doing arithmetic on them. This may produce better code just as
2134 @samp{-fforce-mem} may.
2136 @item -fomit-frame-pointer
2137 Don't keep the frame pointer in a register for functions that
2138 don't need one. This avoids the instructions to save, set up and
2139 restore frame pointers; it also makes an extra register available
2140 in many functions. @strong{It also makes debugging impossible on
2144 On some machines, such as the Vax, this flag has no effect, because
2145 the standard calling sequence automatically handles the frame pointer
2146 and nothing is saved by pretending it doesn't exist. The
2147 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2148 whether a target machine supports this flag. @xref{Registers}.@refill
2151 On some machines, such as the Vax, this flag has no effect, because
2152 the standard calling sequence automatically handles the frame pointer
2153 and nothing is saved by pretending it doesn't exist. The
2154 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
2155 whether a target machine supports this flag. @xref{Registers,,Register
2156 Usage, gcc.info, Using and Porting GCC}.@refill
2160 Don't pay attention to the @code{inline} keyword. Normally this option
2161 is used to keep the compiler from expanding any functions inline.
2162 Note that if you are not optimizing, no functions can be expanded inline.
2164 @item -finline-functions
2165 Integrate all simple functions into their callers. The compiler
2166 heuristically decides which functions are simple enough to be worth
2167 integrating in this way.
2169 If all calls to a given function are integrated, and the function is
2170 declared @code{static}, then the function is normally not output as
2171 assembler code in its own right.
2173 @item -fkeep-inline-functions
2174 Even if all calls to a given function are integrated, and the function
2175 is declared @code{static}, nevertheless output a separate run-time
2176 callable version of the function. This switch does not affect
2177 @code{extern inline} functions.
2179 @item -fkeep-static-consts
2180 Emit variables declared @code{static const} when optimization isn't turned
2181 on, even if the variables aren't referenced.
2183 GNU CC enables this option by default. If you want to force the compiler to
2184 check if the variable was referenced, regardless of whether or not
2185 optimization is turned on, use the @samp{-fno-keep-static-consts} option.
2187 @item -fno-function-cse
2188 Do not put function addresses in registers; make each instruction that
2189 calls a constant function contain the function's address explicitly.
2191 This option results in less efficient code, but some strange hacks
2192 that alter the assembler output may be confused by the optimizations
2193 performed when this option is not used.
2196 This option allows GCC to violate some ANSI or IEEE rules and/or
2197 specifications in the interest of optimizing code for speed. For
2198 example, it allows the compiler to assume arguments to the @code{sqrt}
2199 function are non-negative numbers and that no floating-point values
2202 This option should never be turned on by any @samp{-O} option since
2203 it can result in incorrect output for programs which depend on
2204 an exact implementation of IEEE or ANSI rules/specifications for
2208 @c following causes underfulls.. they don't look great, but we deal.
2210 The following options control specific optimizations. The @samp{-O2}
2211 option turns on all of these optimizations except @samp{-funroll-loops}
2212 and @samp{-funroll-all-loops}. On most machines, the @samp{-O} option
2213 turns on the @samp{-fthread-jumps} and @samp{-fdelayed-branch} options,
2214 but specific machines may handle it differently.
2216 You can use the following flags in the rare cases when ``fine-tuning''
2217 of optimizations to be performed is desired.
2220 @item -fstrength-reduce
2221 Perform the optimizations of loop strength reduction and
2222 elimination of iteration variables.
2224 @item -fthread-jumps
2225 Perform optimizations where we check to see if a jump branches to a
2226 location where another comparison subsumed by the first is found. If
2227 so, the first branch is redirected to either the destination of the
2228 second branch or a point immediately following it, depending on whether
2229 the condition is known to be true or false.
2231 @item -fcse-follow-jumps
2232 In common subexpression elimination, scan through jump instructions
2233 when the target of the jump is not reached by any other path. For
2234 example, when CSE encounters an @code{if} statement with an
2235 @code{else} clause, CSE will follow the jump when the condition
2238 @item -fcse-skip-blocks
2239 This is similar to @samp{-fcse-follow-jumps}, but causes CSE to
2240 follow jumps which conditionally skip over blocks. When CSE
2241 encounters a simple @code{if} statement with no else clause,
2242 @samp{-fcse-skip-blocks} causes CSE to follow the jump around the
2243 body of the @code{if}.
2245 @item -frerun-cse-after-loop
2246 Re-run common subexpression elimination after loop optimizations has been
2249 @item -frerun-loop-opt
2250 Run the loop optimizer twice.
2252 @item -fexpensive-optimizations
2253 Perform a number of minor optimizations that are relatively expensive.
2255 @item -fdelayed-branch
2256 If supported for the target machine, attempt to reorder instructions
2257 to exploit instruction slots available after delayed branch
2260 @item -fschedule-insns
2261 If supported for the target machine, attempt to reorder instructions to
2262 eliminate execution stalls due to required data being unavailable. This
2263 helps machines that have slow floating point or memory load instructions
2264 by allowing other instructions to be issued until the result of the load
2265 or floating point instruction is required.
2267 @item -fschedule-insns2
2268 Similar to @samp{-fschedule-insns}, but requests an additional pass of
2269 instruction scheduling after register allocation has been done. This is
2270 especially useful on machines with a relatively small number of
2271 registers and where memory load instructions take more than one cycle.
2273 @item -ffunction-sections
2274 Place each function into its own section in the output file if the
2275 target supports arbitrary sections. The function's name determines
2276 the section's name in the output file.
2278 Use this option on systems where the linker can perform optimizations
2279 to improve locality of reference in the instruction space. HPPA
2280 processors running HP-UX and Sparc processors running Solaris 2 have
2281 linkers with such optimizations. Other systems using the ELF object format
2282 as well as AIX may have these optimizations in the future.
2284 Only use this option when there are significant benefits from doing
2285 so. When you specify this option, the assembler and linker will
2286 create larger object and executable files and will also be slower.
2287 You will not be able to use @code{gprof} on all systems if you
2288 specify this option and you may have problems with debugging if
2289 you specify both this option and @samp{-g}.
2291 @item -fcaller-saves
2292 Enable values to be allocated in registers that will be clobbered by
2293 function calls, by emitting extra instructions to save and restore the
2294 registers around such calls. Such allocation is done only when it
2295 seems to result in better code than would otherwise be produced.
2297 This option is enabled by default on certain machines, usually those
2298 which have no call-preserved registers to use instead.
2300 @item -funroll-loops
2301 Perform the optimization of loop unrolling. This is only done for loops
2302 whose number of iterations can be determined at compile time or run time.
2303 @samp{-funroll-loop} implies both @samp{-fstrength-reduce} and
2304 @samp{-frerun-cse-after-loop}.
2306 @item -funroll-all-loops
2307 Perform the optimization of loop unrolling. This is done for all loops
2308 and usually makes programs run more slowly. @samp{-funroll-all-loops}
2309 implies @samp{-fstrength-reduce} as well as @samp{-frerun-cse-after-loop}.
2311 @item -fmove-all-movables
2312 Forces all invariant computations in loops to be moved
2315 @item -freduce-all-givs
2316 Forces all general-induction variables in loops to be
2319 @emph{Note:} When compiling programs written in Fortran,
2320 @samp{-fmove-all-moveables} and @samp{-freduce-all-givs} are enabled
2321 by default when you use the optimizer.
2323 These options may generate better or worse code; results are highly
2324 dependent on the structure of loops within the source code.
2326 These two options are intended to be removed someday, once
2327 they have helped determine the efficacy of various
2328 approaches to improving loop optimizations.
2330 Please let us (@code{egcs@@cygnus.com} and @code{fortran@@gnu.org})
2331 know how use of these options affects
2332 the performance of your production code.
2333 We're very interested in code that runs @emph{slower}
2334 when these options are @emph{enabled}.
2337 Disable any machine-specific peephole optimizations.
2339 @item -fbranch-probabilities
2340 After running a program compiled with @samp{-fprofile-arcs}
2341 (@pxref{Debugging Options,, Options for Debugging Your Program or
2342 @code{gcc}}), you can compile it a second time using
2343 @samp{-fbranch-probabilities}, to improve optimizations based on
2344 guessing the path a branch might take.
2347 With @samp{-fbranch-probabilities}, GNU CC puts a @samp{REG_EXEC_COUNT}
2348 note on the first instruction of each basic block, and a
2349 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
2350 These can be used to improve optimization. Currently, they are only
2351 used in one place: in @file{reorg.c}, instead of guessing which path a
2352 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
2353 exactly determine which path is taken more often.
2357 Some machines only support 2 operands per instruction. On such
2358 machines, GNU CC might have to do extra copies. The @samp{-fregmove}
2359 option overrides the default for the machine to do the copy before
2360 register allocation.
2363 @node Preprocessor Options
2364 @section Options Controlling the Preprocessor
2365 @cindex preprocessor options
2366 @cindex options, preprocessor
2368 These options control the C preprocessor, which is run on each C source
2369 file before actual compilation.
2371 If you use the @samp{-E} option, nothing is done except preprocessing.
2372 Some of these options make sense only together with @samp{-E} because
2373 they cause the preprocessor output to be unsuitable for actual
2377 @item -include @var{file}
2378 Process @var{file} as input before processing the regular input file.
2379 In effect, the contents of @var{file} are compiled first. Any @samp{-D}
2380 and @samp{-U} options on the command line are always processed before
2381 @samp{-include @var{file}}, regardless of the order in which they are
2382 written. All the @samp{-include} and @samp{-imacros} options are
2383 processed in the order in which they are written.
2385 @item -imacros @var{file}
2386 Process @var{file} as input, discarding the resulting output, before
2387 processing the regular input file. Because the output generated from
2388 @var{file} is discarded, the only effect of @samp{-imacros @var{file}}
2389 is to make the macros defined in @var{file} available for use in the
2392 Any @samp{-D} and @samp{-U} options on the command line are always
2393 processed before @samp{-imacros @var{file}}, regardless of the order in
2394 which they are written. All the @samp{-include} and @samp{-imacros}
2395 options are processed in the order in which they are written.
2397 @item -idirafter @var{dir}
2398 @cindex second include path
2399 Add the directory @var{dir} to the second include path. The directories
2400 on the second include path are searched when a header file is not found
2401 in any of the directories in the main include path (the one that
2404 @item -iprefix @var{prefix}
2405 Specify @var{prefix} as the prefix for subsequent @samp{-iwithprefix}
2408 @item -iwithprefix @var{dir}
2409 Add a directory to the second include path. The directory's name is
2410 made by concatenating @var{prefix} and @var{dir}, where @var{prefix} was
2411 specified previously with @samp{-iprefix}. If you have not specified a
2412 prefix yet, the directory containing the installed passes of the
2413 compiler is used as the default.
2415 @item -iwithprefixbefore @var{dir}
2416 Add a directory to the main include path. The directory's name is made
2417 by concatenating @var{prefix} and @var{dir}, as in the case of
2418 @samp{-iwithprefix}.
2420 @item -isystem @var{dir}
2421 Add a directory to the beginning of the second include path, marking it
2422 as a system directory, so that it gets the same special treatment as
2423 is applied to the standard system directories.
2426 Do not search the standard system directories for header files. Only
2427 the directories you have specified with @samp{-I} options (and the
2428 current directory, if appropriate) are searched. @xref{Directory
2429 Options}, for information on @samp{-I}.
2431 By using both @samp{-nostdinc} and @samp{-I-}, you can limit the include-file
2432 search path to only those directories you specify explicitly.
2435 Do not predefine any nonstandard macros. (Including architecture flags).
2438 Run only the C preprocessor. Preprocess all the C source files
2439 specified and output the results to standard output or to the
2440 specified output file.
2443 Tell the preprocessor not to discard comments. Used with the
2447 Tell the preprocessor not to generate @samp{#line} directives.
2448 Used with the @samp{-E} option.
2451 @cindex dependencies, make
2453 Tell the preprocessor to output a rule suitable for @code{make}
2454 describing the dependencies of each object file. For each source file,
2455 the preprocessor outputs one @code{make}-rule whose target is the object
2456 file name for that source file and whose dependencies are all the
2457 @code{#include} header files it uses. This rule may be a single line or
2458 may be continued with @samp{\}-newline if it is long. The list of rules
2459 is printed on standard output instead of the preprocessed C program.
2461 @samp{-M} implies @samp{-E}.
2463 Another way to specify output of a @code{make} rule is by setting
2464 the environment variable @code{DEPENDENCIES_OUTPUT} (@pxref{Environment
2468 Like @samp{-M} but the output mentions only the user header files
2469 included with @samp{#include "@var{file}"}. System header files
2470 included with @samp{#include <@var{file}>} are omitted.
2473 Like @samp{-M} but the dependency information is written to a file made by
2474 replacing ".c" with ".d" at the end of the input file names.
2475 This is in addition to compiling the file as specified---@samp{-MD} does
2476 not inhibit ordinary compilation the way @samp{-M} does.
2478 In Mach, you can use the utility @code{md} to merge multiple dependency
2479 files into a single dependency file suitable for using with the @samp{make}
2483 Like @samp{-MD} except mention only user header files, not system
2487 Treat missing header files as generated files and assume they live in the
2488 same directory as the source file. If you specify @samp{-MG}, you
2489 must also specify either @samp{-M} or @samp{-MM}. @samp{-MG} is not
2490 supported with @samp{-MD} or @samp{-MMD}.
2493 Print the name of each header file used, in addition to other normal
2496 @item -A@var{question}(@var{answer})
2497 Assert the answer @var{answer} for @var{question}, in case it is tested
2498 with a preprocessing conditional such as @samp{#if
2499 #@var{question}(@var{answer})}. @samp{-A-} disables the standard
2500 assertions that normally describe the target machine.
2503 Define macro @var{macro} with the string @samp{1} as its definition.
2505 @item -D@var{macro}=@var{defn}
2506 Define macro @var{macro} as @var{defn}. All instances of @samp{-D} on
2507 the command line are processed before any @samp{-U} options.
2510 Undefine macro @var{macro}. @samp{-U} options are evaluated after all
2511 @samp{-D} options, but before any @samp{-include} and @samp{-imacros}
2515 Tell the preprocessor to output only a list of the macro definitions
2516 that are in effect at the end of preprocessing. Used with the @samp{-E}
2520 Tell the preprocessing to pass all macro definitions into the output, in
2521 their proper sequence in the rest of the output.
2524 Like @samp{-dD} except that the macro arguments and contents are omitted.
2525 Only @samp{#define @var{name}} is included in the output.
2528 Support ANSI C trigraphs. The @samp{-ansi} option also has this effect.
2530 @item -Wp,@var{option}
2531 Pass @var{option} as an option to the preprocessor. If @var{option}
2532 contains commas, it is split into multiple options at the commas.
2535 @node Assembler Options
2536 @section Passing Options to the Assembler
2538 @c prevent bad page break with this line
2539 You can pass options to the assembler.
2542 @item -Wa,@var{option}
2543 Pass @var{option} as an option to the assembler. If @var{option}
2544 contains commas, it is split into multiple options at the commas.
2548 @section Options for Linking
2549 @cindex link options
2550 @cindex options, linking
2552 These options come into play when the compiler links object files into
2553 an executable output file. They are meaningless if the compiler is
2554 not doing a link step.
2558 @item @var{object-file-name}
2559 A file name that does not end in a special recognized suffix is
2560 considered to name an object file or library. (Object files are
2561 distinguished from libraries by the linker according to the file
2562 contents.) If linking is done, these object files are used as input
2568 If any of these options is used, then the linker is not run, and
2569 object file names should not be used as arguments. @xref{Overall
2573 @item -l@var{library}
2574 Search the library named @var{library} when linking.
2576 It makes a difference where in the command you write this option; the
2577 linker searches processes libraries and object files in the order they
2578 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
2579 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
2580 to functions in @samp{z}, those functions may not be loaded.
2582 The linker searches a standard list of directories for the library,
2583 which is actually a file named @file{lib@var{library}.a}. The linker
2584 then uses this file as if it had been specified precisely by name.
2586 The directories searched include several standard system directories
2587 plus any that you specify with @samp{-L}.
2589 Normally the files found this way are library files---archive files
2590 whose members are object files. The linker handles an archive file by
2591 scanning through it for members which define symbols that have so far
2592 been referenced but not defined. But if the file that is found is an
2593 ordinary object file, it is linked in the usual fashion. The only
2594 difference between using an @samp{-l} option and specifying a file name
2595 is that @samp{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
2596 and searches several directories.
2599 You need this special case of the @samp{-l} option in order to
2600 link an Objective C program.
2603 Do not use the standard system startup files when linking.
2604 The standard system libraries are used normally, unless @code{-nostdlib}
2605 or @code{-nodefaultlibs} is used.
2607 @item -nodefaultlibs
2608 Do not use the standard system libraries when linking.
2609 Only the libraries you specify will be passed to the linker.
2610 The standard startup files are used normally, unless @code{-nostartfiles}
2614 Do not use the standard system startup files or libraries when linking.
2615 No startup files and only the libraries you specify will be passed to
2618 @cindex @code{-lgcc}, use with @code{-nostdlib}
2619 @cindex @code{-nostdlib} and unresolved references
2620 @cindex unresolved references and @code{-nostdlib}
2621 @cindex @code{-lgcc}, use with @code{-nodefaultlibs}
2622 @cindex @code{-nodefaultlibs} and unresolved references
2623 @cindex unresolved references and @code{-nodefaultlibs}
2624 One of the standard libraries bypassed by @samp{-nostdlib} and
2625 @samp{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
2626 that GNU CC uses to overcome shortcomings of particular machines, or special
2627 needs for some languages.
2629 (@xref{Interface,,Interfacing to GNU CC Output}, for more discussion of
2633 (@xref{Interface,,Interfacing to GNU CC Output,gcc.info,Porting GNU CC},
2634 for more discussion of @file{libgcc.a}.)
2636 In most cases, you need @file{libgcc.a} even when you want to avoid
2637 other standard libraries. In other words, when you specify @samp{-nostdlib}
2638 or @samp{-nodefaultlibs} you should usually specify @samp{-lgcc} as well.
2639 This ensures that you have no unresolved references to internal GNU CC
2640 library subroutines. (For example, @samp{__main}, used to ensure C++
2641 constructors will be called; @pxref{Collect2,,@code{collect2}}.)
2644 Remove all symbol table and relocation information from the executable.
2647 On systems that support dynamic linking, this prevents linking with the shared
2648 libraries. On other systems, this option has no effect.
2651 Produce a shared object which can then be linked with other objects to
2652 form an executable. Not all systems support this option. You must
2653 also specify @samp{-fpic} or @samp{-fPIC} on some systems when
2654 you specify this option.
2657 Bind references to global symbols when building a shared object. Warn
2658 about any unresolved references (unless overridden by the link editor
2659 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
2662 @item -Xlinker @var{option}
2663 Pass @var{option} as an option to the linker. You can use this to
2664 supply system-specific linker options which GNU CC does not know how to
2667 If you want to pass an option that takes an argument, you must use
2668 @samp{-Xlinker} twice, once for the option and once for the argument.
2669 For example, to pass @samp{-assert definitions}, you must write
2670 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
2671 @samp{-Xlinker "-assert definitions"}, because this passes the entire
2672 string as a single argument, which is not what the linker expects.
2674 @item -Wl,@var{option}
2675 Pass @var{option} as an option to the linker. If @var{option} contains
2676 commas, it is split into multiple options at the commas.
2678 @item -u @var{symbol}
2679 Pretend the symbol @var{symbol} is undefined, to force linking of
2680 library modules to define it. You can use @samp{-u} multiple times with
2681 different symbols to force loading of additional library modules.
2684 @node Directory Options
2685 @section Options for Directory Search
2686 @cindex directory options
2687 @cindex options, directory search
2690 These options specify directories to search for header files, for
2691 libraries and for parts of the compiler:
2695 Add the directory @var{dir} to the head of the list of directories to be
2696 searched for header files. This can be used to override a system header
2697 file, substituting your own version, since these directories are
2698 searched before the system header file directories. If you use more
2699 than one @samp{-I} option, the directories are scanned in left-to-right
2700 order; the standard system directories come after.
2703 Any directories you specify with @samp{-I} options before the @samp{-I-}
2704 option are searched only for the case of @samp{#include "@var{file}"};
2705 they are not searched for @samp{#include <@var{file}>}.
2707 If additional directories are specified with @samp{-I} options after
2708 the @samp{-I-}, these directories are searched for all @samp{#include}
2709 directives. (Ordinarily @emph{all} @samp{-I} directories are used
2712 In addition, the @samp{-I-} option inhibits the use of the current
2713 directory (where the current input file came from) as the first search
2714 directory for @samp{#include "@var{file}"}. There is no way to
2715 override this effect of @samp{-I-}. With @samp{-I.} you can specify
2716 searching the directory which was current when the compiler was
2717 invoked. That is not exactly the same as what the preprocessor does
2718 by default, but it is often satisfactory.
2720 @samp{-I-} does not inhibit the use of the standard system directories
2721 for header files. Thus, @samp{-I-} and @samp{-nostdinc} are
2725 Add directory @var{dir} to the list of directories to be searched
2728 @item -B@var{prefix}
2729 This option specifies where to find the executables, libraries,
2730 include files, and data files of the compiler itself.
2732 The compiler driver program runs one or more of the subprograms
2733 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
2734 @var{prefix} as a prefix for each program it tries to run, both with and
2735 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
2737 For each subprogram to be run, the compiler driver first tries the
2738 @samp{-B} prefix, if any. If that name is not found, or if @samp{-B}
2739 was not specified, the driver tries two standard prefixes, which are
2740 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc-lib/}. If neither of
2741 those results in a file name that is found, the unmodified program
2742 name is searched for using the directories specified in your
2743 @samp{PATH} environment variable.
2745 @samp{-B} prefixes that effectively specify directory names also apply
2746 to libraries in the linker, because the compiler translates these
2747 options into @samp{-L} options for the linker. They also apply to
2748 includes files in the preprocessor, because the compiler translates these
2749 options into @samp{-isystem} options for the preprocessor. In this case,
2750 the compiler appends @samp{include} to the prefix.
2752 The run-time support file @file{libgcc.a} can also be searched for using
2753 the @samp{-B} prefix, if needed. If it is not found there, the two
2754 standard prefixes above are tried, and that is all. The file is left
2755 out of the link if it is not found by those means.
2757 Another way to specify a prefix much like the @samp{-B} prefix is to use
2758 the environment variable @code{GCC_EXEC_PREFIX}. @xref{Environment
2761 @item -specs=@var{file}
2762 Process @var{file} after the compiler reads in the standard @file{specs}
2763 file, in order to override the defaults that the @file{gcc} driver
2764 program uses when determining what switches to pass to @file{cc1},
2765 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
2766 @samp{-specs=}@var{file} can be specified on the command line, and they
2767 are processed in order, from left to right.
2770 @node Target Options
2771 @section Specifying Target Machine and Compiler Version
2772 @cindex target options
2773 @cindex cross compiling
2774 @cindex specifying machine version
2775 @cindex specifying compiler version and target machine
2776 @cindex compiler version, specifying
2777 @cindex target machine, specifying
2779 By default, GNU CC compiles code for the same type of machine that you
2780 are using. However, it can also be installed as a cross-compiler, to
2781 compile for some other type of machine. In fact, several different
2782 configurations of GNU CC, for different target machines, can be
2783 installed side by side. Then you specify which one to use with the
2786 In addition, older and newer versions of GNU CC can be installed side
2787 by side. One of them (probably the newest) will be the default, but
2788 you may sometimes wish to use another.
2791 @item -b @var{machine}
2792 The argument @var{machine} specifies the target machine for compilation.
2793 This is useful when you have installed GNU CC as a cross-compiler.
2795 The value to use for @var{machine} is the same as was specified as the
2796 machine type when configuring GNU CC as a cross-compiler. For
2797 example, if a cross-compiler was configured with @samp{configure
2798 i386v}, meaning to compile for an 80386 running System V, then you
2799 would specify @samp{-b i386v} to run that cross compiler.
2801 When you do not specify @samp{-b}, it normally means to compile for
2802 the same type of machine that you are using.
2804 @item -V @var{version}
2805 The argument @var{version} specifies which version of GNU CC to run.
2806 This is useful when multiple versions are installed. For example,
2807 @var{version} might be @samp{2.0}, meaning to run GNU CC version 2.0.
2809 The default version, when you do not specify @samp{-V}, is the last
2810 version of GNU CC that you installed.
2813 The @samp{-b} and @samp{-V} options actually work by controlling part of
2814 the file name used for the executable files and libraries used for
2815 compilation. A given version of GNU CC, for a given target machine, is
2816 normally kept in the directory @file{/usr/local/lib/gcc-lib/@var{machine}/@var{version}}.@refill
2818 Thus, sites can customize the effect of @samp{-b} or @samp{-V} either by
2819 changing the names of these directories or adding alternate names (or
2820 symbolic links). If in directory @file{/usr/local/lib/gcc-lib/} the
2821 file @file{80386} is a link to the file @file{i386v}, then @samp{-b
2822 80386} becomes an alias for @samp{-b i386v}.
2824 In one respect, the @samp{-b} or @samp{-V} do not completely change
2825 to a different compiler: the top-level driver program @code{gcc}
2826 that you originally invoked continues to run and invoke the other
2827 executables (preprocessor, compiler per se, assembler and linker)
2828 that do the real work. However, since no real work is done in the
2829 driver program, it usually does not matter that the driver program
2830 in use is not the one for the specified target and version.
2832 The only way that the driver program depends on the target machine is
2833 in the parsing and handling of special machine-specific options.
2834 However, this is controlled by a file which is found, along with the
2835 other executables, in the directory for the specified version and
2836 target machine. As a result, a single installed driver program adapts
2837 to any specified target machine and compiler version.
2839 The driver program executable does control one significant thing,
2840 however: the default version and target machine. Therefore, you can
2841 install different instances of the driver program, compiled for
2842 different targets or versions, under different names.
2844 For example, if the driver for version 2.0 is installed as @code{ogcc}
2845 and that for version 2.1 is installed as @code{gcc}, then the command
2846 @code{gcc} will use version 2.1 by default, while @code{ogcc} will use
2847 2.0 by default. However, you can choose either version with either
2848 command with the @samp{-V} option.
2850 @node Submodel Options
2851 @section Hardware Models and Configurations
2852 @cindex submodel options
2853 @cindex specifying hardware config
2854 @cindex hardware models and configurations, specifying
2855 @cindex machine dependent options
2857 Earlier we discussed the standard option @samp{-b} which chooses among
2858 different installed compilers for completely different target
2859 machines, such as Vax vs. 68000 vs. 80386.
2861 In addition, each of these target machine types can have its own
2862 special options, starting with @samp{-m}, to choose among various
2863 hardware models or configurations---for example, 68010 vs 68020,
2864 floating coprocessor or none. A single installed version of the
2865 compiler can compile for any model or configuration, according to the
2868 Some configurations of the compiler also support additional special
2869 options, usually for compatibility with other compilers on the same
2873 These options are defined by the macro @code{TARGET_SWITCHES} in the
2874 machine description. The default for the options is also defined by
2875 that macro, which enables you to change the defaults.
2888 * RS/6000 and PowerPC Options::
2893 * Intel 960 Options::
2894 * DEC Alpha Options::
2898 * System V Options::
2902 @node M680x0 Options
2903 @subsection M680x0 Options
2904 @cindex M680x0 options
2906 These are the @samp{-m} options defined for the 68000 series. The default
2907 values for these options depends on which style of 68000 was selected when
2908 the compiler was configured; the defaults for the most common choices are
2914 Generate output for a 68000. This is the default
2915 when the compiler is configured for 68000-based systems.
2917 Use this option for microcontrollers with a 68000 or EC000 core,
2918 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
2922 Generate output for a 68020. This is the default
2923 when the compiler is configured for 68020-based systems.
2926 Generate output containing 68881 instructions for floating point.
2927 This is the default for most 68020 systems unless @samp{-nfp} was
2928 specified when the compiler was configured.
2931 Generate output for a 68030. This is the default when the compiler is
2932 configured for 68030-based systems.
2935 Generate output for a 68040. This is the default when the compiler is
2936 configured for 68040-based systems.
2938 This option inhibits the use of 68881/68882 instructions that have to be
2939 emulated by software on the 68040. Use this option if your 68040 does not
2940 have code to emulate those instructions.
2943 Generate output for a 68060. This is the default when the compiler is
2944 configured for 68060-based systems.
2946 This option inhibits the use of 68020 and 68881/68882 instructions that
2947 have to be emulated by software on the 68060. Use this option if your 68060
2948 does not have code to emulate those instructions.
2951 Generate output for a CPU32. This is the default
2952 when the compiler is configured for CPU32-based systems.
2954 Use this option for microcontrollers with a
2955 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
2956 68336, 68340, 68341, 68349 and 68360.
2959 Generate output for a 520X "coldfire" family cpu. This is the default
2960 when the compiler is configured for 520X-based systems.
2962 Use this option for microcontroller with a 5200 core, including
2963 the MCF5202, MCF5203, MCF5204 and MCF5202.
2967 Generate output for a 68040, without using any of the new instructions.
2968 This results in code which can run relatively efficiently on either a
2969 68020/68881 or a 68030 or a 68040. The generated code does use the
2970 68881 instructions that are emulated on the 68040.
2973 Generate output for a 68060, without using any of the new instructions.
2974 This results in code which can run relatively efficiently on either a
2975 68020/68881 or a 68030 or a 68040. The generated code does use the
2976 68881 instructions that are emulated on the 68060.
2979 Generate output containing Sun FPA instructions for floating point.
2982 Generate output containing library calls for floating point.
2983 @strong{Warning:} the requisite libraries are not available for all m68k
2984 targets. Normally the facilities of the machine's usual C compiler are
2985 used, but this can't be done directly in cross-compilation. You must
2986 make your own arrangements to provide suitable library functions for
2987 cross-compilation. The embedded targets @samp{m68k-*-aout} and
2988 @samp{m68k-*-coff} do provide software floating point support.
2991 Consider type @code{int} to be 16 bits wide, like @code{short int}.
2994 Do not use the bit-field instructions. The @samp{-m68000}, @samp{-mcpu32}
2995 and @samp{-m5200} options imply @w{@samp{-mnobitfield}}.
2998 Do use the bit-field instructions. The @samp{-m68020} option implies
2999 @samp{-mbitfield}. This is the default if you use a configuration
3000 designed for a 68020.
3003 Use a different function-calling convention, in which functions
3004 that take a fixed number of arguments return with the @code{rtd}
3005 instruction, which pops their arguments while returning. This
3006 saves one instruction in the caller since there is no need to pop
3007 the arguments there.
3009 This calling convention is incompatible with the one normally
3010 used on Unix, so you cannot use it if you need to call libraries
3011 compiled with the Unix compiler.
3013 Also, you must provide function prototypes for all functions that
3014 take variable numbers of arguments (including @code{printf});
3015 otherwise incorrect code will be generated for calls to those
3018 In addition, seriously incorrect code will result if you call a
3019 function with too many arguments. (Normally, extra arguments are
3020 harmlessly ignored.)
3022 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
3023 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
3026 @itemx -mno-align-int
3027 Control whether GNU CC aligns @code{int}, @code{long}, @code{long long},
3028 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
3029 boundary (@samp{-malign-int}) or a 16-bit boundary (@samp{-mno-align-int}).
3030 Aligning variables on 32-bit boundaries produces code that runs somewhat
3031 faster on processors with 32-bit busses at the expense of more memory.
3033 @strong{Warning:} if you use the @samp{-malign-int} switch, GNU CC will
3034 align structures containing the above types differently than
3035 most published application binary interface specifications for the m68k.
3040 @subsection VAX Options
3043 These @samp{-m} options are defined for the Vax:
3047 Do not output certain jump instructions (@code{aobleq} and so on)
3048 that the Unix assembler for the Vax cannot handle across long
3052 Do output those jump instructions, on the assumption that you
3053 will assemble with the GNU assembler.
3056 Output code for g-format floating point numbers instead of d-format.
3060 @subsection SPARC Options
3061 @cindex SPARC options
3063 These @samp{-m} switches are supported on the SPARC:
3068 Specify @samp{-mapp-regs} to generate output using the global registers
3069 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
3072 To be fully SVR4 ABI compliant at the cost of some performance loss,
3073 specify @samp{-mno-app-regs}. You should compile libraries and system
3074 software with this option.
3078 Generate output containing floating point instructions. This is the
3083 Generate output containing library calls for floating point.
3084 @strong{Warning:} the requisite libraries are not available for all SPARC
3085 targets. Normally the facilities of the machine's usual C compiler are
3086 used, but this cannot be done directly in cross-compilation. You must make
3087 your own arrangements to provide suitable library functions for
3088 cross-compilation. The embedded targets @samp{sparc-*-aout} and
3089 @samp{sparclite-*-*} do provide software floating point support.
3091 @samp{-msoft-float} changes the calling convention in the output file;
3092 therefore, it is only useful if you compile @emph{all} of a program with
3093 this option. In particular, you need to compile @file{libgcc.a}, the
3094 library that comes with GNU CC, with @samp{-msoft-float} in order for
3097 @item -mhard-quad-float
3098 Generate output containing quad-word (long double) floating point
3101 @item -msoft-quad-float
3102 Generate output containing library calls for quad-word (long double)
3103 floating point instructions. The functions called are those specified
3104 in the SPARC ABI. This is the default.
3106 As of this writing, there are no sparc implementations that have hardware
3107 support for the quad-word floating point instructions. They all invoke
3108 a trap handler for one of these instructions, and then the trap handler
3109 emulates the effect of the instruction. Because of the trap handler overhead,
3110 this is much slower than calling the ABI library routines. Thus the
3111 @samp{-msoft-quad-float} option is the default.
3115 With @samp{-mepilogue} (the default), the compiler always emits code for
3116 function exit at the end of each function. Any function exit in
3117 the middle of the function (such as a return statement in C) will
3118 generate a jump to the exit code at the end of the function.
3120 With @samp{-mno-epilogue}, the compiler tries to emit exit code inline
3121 at every function exit.
3125 With @samp{-mflat}, the compiler does not generate save/restore instructions
3126 and will use a "flat" or single register window calling convention.
3127 This model uses %i7 as the frame pointer and is compatible with the normal
3128 register window model. Code from either may be intermixed.
3129 The local registers and the input registers (0-5) are still treated as
3130 "call saved" registers and will be saved on the stack as necessary.
3132 With @samp{-mno-flat} (the default), the compiler emits save/restore
3133 instructions (except for leaf functions) and is the normal mode of operation.
3135 @item -mno-unaligned-doubles
3136 @itemx -munaligned-doubles
3137 Assume that doubles have 8 byte alignment. This is the default.
3139 With @samp{-munaligned-doubles}, GNU CC assumes that doubles have 8 byte
3140 alignment only if they are contained in another type, or if they have an
3141 absolute address. Otherwise, it assumes they have 4 byte alignment.
3142 Specifying this option avoids some rare compatibility problems with code
3143 generated by other compilers. It is not the default because it results
3144 in a performance loss, especially for floating point code.
3148 These two options select variations on the SPARC architecture.
3150 By default (unless specifically configured for the Fujitsu SPARClite),
3151 GCC generates code for the v7 variant of the SPARC architecture.
3153 @samp{-mv8} will give you SPARC v8 code. The only difference from v7
3154 code is that the compiler emits the integer multiply and integer
3155 divide instructions which exist in SPARC v8 but not in SPARC v7.
3157 @samp{-msparclite} will give you SPARClite code. This adds the integer
3158 multiply, integer divide step and scan (@code{ffs}) instructions which
3159 exist in SPARClite but not in SPARC v7.
3161 These options are deprecated and will be deleted in GNU CC 2.9.
3162 They have been replaced with @samp{-mcpu=xxx}.
3166 These two options select the processor for which the code is optimised.
3168 With @samp{-mcypress} (the default), the compiler optimizes code for the
3169 Cypress CY7C602 chip, as used in the SparcStation/SparcServer 3xx series.
3170 This is also appropriate for the older SparcStation 1, 2, IPX etc.
3172 With @samp{-msupersparc} the compiler optimizes code for the SuperSparc cpu, as
3173 used in the SparcStation 10, 1000 and 2000 series. This flag also enables use
3174 of the full SPARC v8 instruction set.
3176 These options are deprecated and will be deleted in GNU CC 2.9.
3177 They have been replaced with @samp{-mcpu=xxx}.
3179 @item -mcpu=@var{cpu_type}
3180 Set the instruction set, register set, and instruction scheduling parameters
3181 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
3182 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
3183 @samp{f930}, @samp{f934}, @samp{sparclet}, @samp{tsc701}, @samp{v9}, and
3186 Default instruction scheduling parameters are used for values that select
3187 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
3188 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
3190 Here is a list of each supported architecture and their supported
3196 sparclite: f930, f934
3201 @item -mtune=@var{cpu_type}
3202 Set the instruction scheduling parameters for machine type
3203 @var{cpu_type}, but do not set the instruction set or register set that the
3204 option @samp{-mcpu=}@var{cpu_type} would.
3206 The same values for @samp{-mcpu=}@var{cpu_type} are used for
3207 @samp{-mtune=}@var{cpu_type}, though the only useful values are those that
3208 select a particular cpu implementation: @samp{cypress}, @samp{supersparc},
3209 @samp{f930}, @samp{f934}, @samp{tsc701}, @samp{ultrasparc}.
3211 @item -malign-loops=@var{num}
3212 Align loops to a 2 raised to a @var{num} byte boundary. If
3213 @samp{-malign-loops} is not specified, the default is 2.
3215 @item -malign-jumps=@var{num}
3216 Align instructions that are only jumped to to a 2 raised to a @var{num}
3217 byte boundary. If @samp{-malign-jumps} is not specified, the default is 2.
3219 @item -malign-functions=@var{num}
3220 Align the start of functions to a 2 raised to @var{num} byte boundary.
3221 If @samp{-malign-functions} is not specified, the default is 2 if compiling
3222 for 32 bit sparc, and 5 if compiling for 64 bit sparc.
3226 These @samp{-m} switches are supported in addition to the above
3227 on the SPARCLET processor.
3230 @item -mlittle-endian
3231 Generate code for a processor running in little-endian mode.
3234 Treat register @code{%g0} as a normal register.
3235 GCC will continue to clobber it as necessary but will not assume
3236 it always reads as 0.
3238 @item -mbroken-saverestore
3239 Generate code that does not use non-trivial forms of the @code{save} and
3240 @code{restore} instructions. Early versions of the SPARCLET processor do
3241 not correctly handle @code{save} and @code{restore} instructions used with
3242 arguments. They correctly handle them used without arguments. A @code{save}
3243 instruction used without arguments increments the current window pointer
3244 but does not allocate a new stack frame. It is assumed that the window
3245 overflow trap handler will properly handle this case as will interrupt
3249 These @samp{-m} switches are supported in addition to the above
3250 on SPARC V9 processors in 64 bit environments.
3253 @item -mlittle-endian
3254 Generate code for a processor running in little-endian mode.
3258 Generate code for a 32 bit or 64 bit environment.
3259 The 32 bit environment sets int, long and pointer to 32 bits.
3260 The 64 bit environment sets int to 32 bits and long and pointer
3263 @item -mcmodel=medlow
3264 Generate code for the Medium/Low code model: the program must be linked
3265 in the low 32 bits of the address space. Pointers are 64 bits.
3266 Programs can be statically or dynamically linked.
3268 @item -mcmodel=medmid
3269 Generate code for the Medium/Middle code model: the program must be linked
3270 in the low 44 bits of the address space, the text segment must be less than
3271 2G bytes, and data segment must be within 2G of the text segment.
3272 Pointers are 64 bits.
3274 @item -mcmodel=medany
3275 Generate code for the Medium/Anywhere code model: the program may be linked
3276 anywhere in the address space, the text segment must be less than
3277 2G bytes, and data segment must be within 2G of the text segment.
3278 Pointers are 64 bits.
3280 @item -mcmodel=embmedany
3281 Generate code for the Medium/Anywhere code model for embedded systems:
3282 assume a 32 bit text and a 32 bit data segment, both starting anywhere
3283 (determined at link time). Register %g4 points to the base of the
3284 data segment. Pointers still 64 bits.
3285 Programs are statically linked, PIC is not supported.
3288 @itemx -mno-stack-bias
3289 With @samp{-mstack-bias}, GNU CC assumes that the stack pointer, and
3290 frame pointer if present, are offset by -2047 which must be added back
3291 when making stack frame references.
3292 Otherwise, assume no such offset is present.
3295 @node Convex Options
3296 @subsection Convex Options
3297 @cindex Convex options
3299 These @samp{-m} options are defined for Convex:
3303 Generate output for C1. The code will run on any Convex machine.
3304 The preprocessor symbol @code{__convex__c1__} is defined.
3307 Generate output for C2. Uses instructions not available on C1.
3308 Scheduling and other optimizations are chosen for max performance on C2.
3309 The preprocessor symbol @code{__convex_c2__} is defined.
3312 Generate output for C32xx. Uses instructions not available on C1.
3313 Scheduling and other optimizations are chosen for max performance on C32.
3314 The preprocessor symbol @code{__convex_c32__} is defined.
3317 Generate output for C34xx. Uses instructions not available on C1.
3318 Scheduling and other optimizations are chosen for max performance on C34.
3319 The preprocessor symbol @code{__convex_c34__} is defined.
3322 Generate output for C38xx. Uses instructions not available on C1.
3323 Scheduling and other optimizations are chosen for max performance on C38.
3324 The preprocessor symbol @code{__convex_c38__} is defined.
3327 Generate code which puts an argument count in the word preceding each
3328 argument list. This is compatible with regular CC, and a few programs
3329 may need the argument count word. GDB and other source-level debuggers
3330 do not need it; this info is in the symbol table.
3333 Omit the argument count word. This is the default.
3335 @item -mvolatile-cache
3336 Allow volatile references to be cached. This is the default.
3338 @item -mvolatile-nocache
3339 Volatile references bypass the data cache, going all the way to memory.
3340 This is only needed for multi-processor code that does not use standard
3341 synchronization instructions. Making non-volatile references to volatile
3342 locations will not necessarily work.
3345 Type long is 32 bits, the same as type int. This is the default.
3348 Type long is 64 bits, the same as type long long. This option is useless,
3349 because no library support exists for it.
3352 @node AMD29K Options
3353 @subsection AMD29K Options
3354 @cindex AMD29K options
3356 These @samp{-m} options are defined for the AMD Am29000:
3361 @cindex DW bit (29k)
3362 Generate code that assumes the @code{DW} bit is set, i.e., that byte and
3363 halfword operations are directly supported by the hardware. This is the
3368 Generate code that assumes the @code{DW} bit is not set.
3372 @cindex byte writes (29k)
3373 Generate code that assumes the system supports byte and halfword write
3374 operations. This is the default.
3378 Generate code that assumes the systems does not support byte and
3379 halfword write operations. @samp{-mnbw} implies @samp{-mndw}.
3383 @cindex memory model (29k)
3384 Use a small memory model that assumes that all function addresses are
3385 either within a single 256 KB segment or at an absolute address of less
3386 than 256k. This allows the @code{call} instruction to be used instead
3387 of a @code{const}, @code{consth}, @code{calli} sequence.
3391 Use the normal memory model: Generate @code{call} instructions only when
3392 calling functions in the same file and @code{calli} instructions
3393 otherwise. This works if each file occupies less than 256 KB but allows
3394 the entire executable to be larger than 256 KB. This is the default.
3397 Always use @code{calli} instructions. Specify this option if you expect
3398 a single file to compile into more than 256 KB of code.
3402 @cindex processor selection (29k)
3403 Generate code for the Am29050.
3407 Generate code for the Am29000. This is the default.
3409 @item -mkernel-registers
3410 @kindex -mkernel-registers
3411 @cindex kernel and user registers (29k)
3412 Generate references to registers @code{gr64-gr95} instead of to
3413 registers @code{gr96-gr127}. This option can be used when compiling
3414 kernel code that wants a set of global registers disjoint from that used
3417 Note that when this option is used, register names in @samp{-f} flags
3418 must use the normal, user-mode, names.
3420 @item -muser-registers
3421 @kindex -muser-registers
3422 Use the normal set of global registers, @code{gr96-gr127}. This is the
3426 @itemx -mno-stack-check
3427 @kindex -mstack-check
3428 @cindex stack checks (29k)
3429 Insert (or do not insert) a call to @code{__msp_check} after each stack
3430 adjustment. This is often used for kernel code.
3433 @itemx -mno-storem-bug
3434 @kindex -mstorem-bug
3435 @cindex storem bug (29k)
3436 @samp{-mstorem-bug} handles 29k processors which cannot handle the
3437 separation of a mtsrim insn and a storem instruction (most 29000 chips
3438 to date, but not the 29050).
3440 @item -mno-reuse-arg-regs
3441 @itemx -mreuse-arg-regs
3442 @kindex -mreuse-arg-regs
3443 @samp{-mno-reuse-arg-regs} tells the compiler to only use incoming argument
3444 registers for copying out arguments. This helps detect calling a function
3445 with fewer arguments than it was declared with.
3447 @item -mno-impure-text
3448 @itemx -mimpure-text
3449 @kindex -mimpure-text
3450 @samp{-mimpure-text}, used in addition to @samp{-shared}, tells the compiler to
3451 not pass @samp{-assert pure-text} to the linker when linking a shared object.
3454 @kindex -msoft-float
3455 Generate output containing library calls for floating point.
3456 @strong{Warning:} the requisite libraries are not part of GNU CC.
3457 Normally the facilities of the machine's usual C compiler are used, but
3458 this can't be done directly in cross-compilation. You must make your
3459 own arrangements to provide suitable library functions for
3464 @subsection ARM Options
3467 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
3472 @kindex -mapcs-frame
3473 Generate a stack frame that is compliant with the ARM Procedure Call
3474 Standard for all functions, even if this is not strictly necessary for
3475 correct execution of the code.
3479 Generate code for a processor running with a 26-bit program counter,
3480 and conforming to the function calling standards for the APCS 26-bit
3481 option. This option replaces the @samp{-m2} and @samp{-m3} options
3482 of previous releases of the compiler.
3486 Generate code for a processor running with a 32-bit program counter,
3487 and conforming to the function calling standards for the APCS 32-bit
3488 option. This option replaces the @samp{-m6} option of previous releases
3492 Generate output containing floating point instructions. This is the
3496 Generate output containing library calls for floating point.
3497 @strong{Warning:} the requisite libraries are not available for all ARM
3498 targets. Normally the facilities of the machine's usual C compiler are
3499 used, but this cannot be done directly in cross-compilation. You must make
3500 your own arrangements to provide suitable library functions for
3503 @samp{-msoft-float} changes the calling convention in the output file;
3504 therefore, it is only useful if you compile @emph{all} of a program with
3505 this option. In particular, you need to compile @file{libgcc.a}, the
3506 library that comes with GNU CC, with @samp{-msoft-float} in order for
3509 @item -mlittle-endian
3510 Generate code for a processor running in little-endian mode. This is
3511 the default for all standard configurations.
3514 Generate code for a processor running in big-endian mode; the default is
3515 to compile code for a little-endian processor.
3517 @item -mwords-little-endian
3518 This option only applies when generating code for big-endian processors.
3519 Generate code for a little-endian word order but a big-endian byte
3520 order. That is, a byte order of the form @samp{32107654}. Note: this
3521 option should only be used if you require compatibility with code for
3522 big-endian ARM processors generated by versions of the compiler prior to
3525 @item -mshort-load-bytes
3526 @kindex -mshort-load-bytes
3527 Do not try to load half-words (eg @samp{short}s) by loading a word from
3528 an unaligned address. For some targets the MMU is configured to trap
3529 unaligned loads; use this option to generate code that is safe in these
3532 @item -mno-short-load-bytes
3533 @kindex -mno-short-load-bytes
3534 Use unaligned word loads to load half-words (eg @samp{short}s). This
3535 option produces more efficient code, but the MMU is sometimes configured
3536 to trap these instructions.
3540 This option only applies to RISC iX. Emulate the native BSD-mode
3541 compiler. This is the default if @samp{-ansi} is not specified.
3545 This option only applies to RISC iX. Emulate the native X/Open-mode
3548 @item -mno-symrename
3549 @kindex -mno-symrename
3550 This option only applies to RISC iX. Do not run the assembler
3551 post-processor, @samp{symrename}, after code has been assembled.
3552 Normally it is necessary to modify some of the standard symbols in
3553 preparation for linking with the RISC iX C library; this option
3554 suppresses this pass. The post-processor is never run when the
3555 compiler is built for cross-compilation.
3558 @node MN10300 Options
3559 @subsection MN10300 Options
3560 @cindex MN10300 options
3561 These @samp{-m} options are defined for Matsushita MN10300 architectures:
3565 Generate code to avoid bugs in the multiply instructions for the MN10300
3566 processors. This is the default.
3569 Do not generate code to avoid bugs in the multiply instructions for the
3573 @node M32R/D Options
3574 @subsection M32R/D Options
3575 @cindex M32R/D options
3577 These @samp{-m} options are defined for Mitsubishi M32R/D architectures:
3580 @item -mcode-model=small
3581 Assume all objects live in the lower 16MB of memory (so that their addresses
3582 can be loaded with the @code{ld24} instruction), and assume all subroutines
3583 are reachable with the @code{bl} instruction.
3584 This is the default.
3586 The addressability of a particular object can be set with the
3587 @code{model} attribute.
3589 @item -mcode-model=medium
3590 Assume objects may be anywhere in the 32 bit address space (the compiler
3591 will generate @code{seth/add3} instructions to load their addresses), and
3592 assume all subroutines are reachable with the @code{bl} instruction.
3594 @item -mcode-model=large
3595 Assume objects may be anywhere in the 32 bit address space (the compiler
3596 will generate @code{seth/add3} instructions to load their addresses), and
3597 assume subroutines may not be reachable with the @code{bl} instruction
3598 (the compiler will generate the much slower @code{seth/add3/jl}
3599 instruction sequence).
3602 Disable use of the small data area. Variables will be put into
3603 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
3604 @code{section} attribute has been specified).
3605 This is the default.
3607 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
3608 Objects may be explicitly put in the small data area with the
3609 @code{section} attribute using one of these sections.
3612 Put small global and static data in the small data area, but do not
3613 generate special code to reference them.
3616 Put small global and static data in the small data area, and generate
3617 special instructions to reference them.
3620 @cindex smaller data references
3621 Put global and static objects less than or equal to @var{num} bytes
3622 into the small data or bss sections instead of the normal data or bss
3623 sections. The default value of @var{num} is 8.
3624 The @samp{-msdata} option must be set to one of @samp{sdata} or @samp{use}
3625 for this option to have any effect.
3627 All modules should be compiled with the same @samp{-G @var{num}} value.
3628 Compiling with different values of @var{num} may or may not work; if it
3629 doesn't the linker will give an error message - incorrect code will not be
3635 @subsection M88K Options
3636 @cindex M88k options
3638 These @samp{-m} options are defined for Motorola 88k architectures:
3643 Generate code that works well on both the m88100 and the
3648 Generate code that works best for the m88100, but that also
3653 Generate code that works best for the m88110, and may not run
3658 Obsolete option to be removed from the next revision.
3661 @item -midentify-revision
3662 @kindex -midentify-revision
3664 @cindex identifying source, compiler (88k)
3665 Include an @code{ident} directive in the assembler output recording the
3666 source file name, compiler name and version, timestamp, and compilation
3669 @item -mno-underscores
3670 @kindex -mno-underscores
3671 @cindex underscores, avoiding (88k)
3672 In assembler output, emit symbol names without adding an underscore
3673 character at the beginning of each name. The default is to use an
3674 underscore as prefix on each name.
3676 @item -mocs-debug-info
3677 @itemx -mno-ocs-debug-info
3678 @kindex -mocs-debug-info
3679 @kindex -mno-ocs-debug-info
3681 @cindex debugging, 88k OCS
3682 Include (or omit) additional debugging information (about registers used
3683 in each stack frame) as specified in the 88open Object Compatibility
3684 Standard, ``OCS''. This extra information allows debugging of code that
3685 has had the frame pointer eliminated. The default for DG/UX, SVr4, and
3686 Delta 88 SVr3.2 is to include this information; other 88k configurations
3687 omit this information by default.
3689 @item -mocs-frame-position
3690 @kindex -mocs-frame-position
3691 @cindex register positions in frame (88k)
3692 When emitting COFF debugging information for automatic variables and
3693 parameters stored on the stack, use the offset from the canonical frame
3694 address, which is the stack pointer (register 31) on entry to the
3695 function. The DG/UX, SVr4, Delta88 SVr3.2, and BCS configurations use
3696 @samp{-mocs-frame-position}; other 88k configurations have the default
3697 @samp{-mno-ocs-frame-position}.
3699 @item -mno-ocs-frame-position
3700 @kindex -mno-ocs-frame-position
3701 @cindex register positions in frame (88k)
3702 When emitting COFF debugging information for automatic variables and
3703 parameters stored on the stack, use the offset from the frame pointer
3704 register (register 30). When this option is in effect, the frame
3705 pointer is not eliminated when debugging information is selected by the
3708 @item -moptimize-arg-area
3709 @itemx -mno-optimize-arg-area
3710 @kindex -moptimize-arg-area
3711 @kindex -mno-optimize-arg-area
3712 @cindex arguments in frame (88k)
3713 Control how function arguments are stored in stack frames.
3714 @samp{-moptimize-arg-area} saves space by optimizing them, but this
3715 conflicts with the 88open specifications. The opposite alternative,
3716 @samp{-mno-optimize-arg-area}, agrees with 88open standards. By default
3717 GNU CC does not optimize the argument area.
3719 @item -mshort-data-@var{num}
3720 @kindex -mshort-data-@var{num}
3721 @cindex smaller data references (88k)
3722 @cindex r0-relative references (88k)
3723 Generate smaller data references by making them relative to @code{r0},
3724 which allows loading a value using a single instruction (rather than the
3725 usual two). You control which data references are affected by
3726 specifying @var{num} with this option. For example, if you specify
3727 @samp{-mshort-data-512}, then the data references affected are those
3728 involving displacements of less than 512 bytes.
3729 @samp{-mshort-data-@var{num}} is not effective for @var{num} greater
3732 @item -mserialize-volatile
3733 @kindex -mserialize-volatile
3734 @itemx -mno-serialize-volatile
3735 @kindex -mno-serialize-volatile
3736 @cindex sequential consistency on 88k
3737 Do, or don't, generate code to guarantee sequential consistency
3738 of volatile memory references. By default, consistency is
3741 The order of memory references made by the MC88110 processor does
3742 not always match the order of the instructions requesting those
3743 references. In particular, a load instruction may execute before
3744 a preceding store instruction. Such reordering violates
3745 sequential consistency of volatile memory references, when there
3746 are multiple processors. When consistency must be guaranteed,
3747 GNU C generates special instructions, as needed, to force
3748 execution in the proper order.
3750 The MC88100 processor does not reorder memory references and so
3751 always provides sequential consistency. However, by default, GNU
3752 C generates the special instructions to guarantee consistency
3753 even when you use @samp{-m88100}, so that the code may be run on an
3754 MC88110 processor. If you intend to run your code only on the
3755 MC88100 processor, you may use @samp{-mno-serialize-volatile}.
3757 The extra code generated to guarantee consistency may affect the
3758 performance of your application. If you know that you can safely
3759 forgo this guarantee, you may use @samp{-mno-serialize-volatile}.
3765 @cindex assembler syntax, 88k
3767 Turn on (@samp{-msvr4}) or off (@samp{-msvr3}) compiler extensions
3768 related to System V release 4 (SVr4). This controls the following:
3772 Which variant of the assembler syntax to emit.
3774 @samp{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
3775 that is used on System V release 4.
3777 @samp{-msvr4} makes GNU CC issue additional declaration directives used in
3781 @samp{-msvr4} is the default for the m88k-motorola-sysv4 and
3782 m88k-dg-dgux m88k configurations. @samp{-msvr3} is the default for all
3783 other m88k configurations.
3785 @item -mversion-03.00
3786 @kindex -mversion-03.00
3787 This option is obsolete, and is ignored.
3788 @c ??? which asm syntax better for GAS? option there too?
3790 @item -mno-check-zero-division
3791 @itemx -mcheck-zero-division
3792 @kindex -mno-check-zero-division
3793 @kindex -mcheck-zero-division
3794 @cindex zero division on 88k
3795 Do, or don't, generate code to guarantee that integer division by
3796 zero will be detected. By default, detection is guaranteed.
3798 Some models of the MC88100 processor fail to trap upon integer
3799 division by zero under certain conditions. By default, when
3800 compiling code that might be run on such a processor, GNU C
3801 generates code that explicitly checks for zero-valued divisors
3802 and traps with exception number 503 when one is detected. Use of
3803 mno-check-zero-division suppresses such checking for code
3804 generated to run on an MC88100 processor.
3806 GNU C assumes that the MC88110 processor correctly detects all
3807 instances of integer division by zero. When @samp{-m88110} is
3808 specified, both @samp{-mcheck-zero-division} and
3809 @samp{-mno-check-zero-division} are ignored, and no explicit checks for
3810 zero-valued divisors are generated.
3812 @item -muse-div-instruction
3813 @kindex -muse-div-instruction
3814 @cindex divide instruction, 88k
3815 Use the div instruction for signed integer division on the
3816 MC88100 processor. By default, the div instruction is not used.
3818 On the MC88100 processor the signed integer division instruction
3819 div) traps to the operating system on a negative operand. The
3820 operating system transparently completes the operation, but at a
3821 large cost in execution time. By default, when compiling code
3822 that might be run on an MC88100 processor, GNU C emulates signed
3823 integer division using the unsigned integer division instruction
3824 divu), thereby avoiding the large penalty of a trap to the
3825 operating system. Such emulation has its own, smaller, execution
3826 cost in both time and space. To the extent that your code's
3827 important signed integer division operations are performed on two
3828 nonnegative operands, it may be desirable to use the div
3829 instruction directly.
3831 On the MC88110 processor the div instruction (also known as the
3832 divs instruction) processes negative operands without trapping to
3833 the operating system. When @samp{-m88110} is specified,
3834 @samp{-muse-div-instruction} is ignored, and the div instruction is used
3835 for signed integer division.
3837 Note that the result of dividing INT_MIN by -1 is undefined. In
3838 particular, the behavior of such a division with and without
3839 @samp{-muse-div-instruction} may differ.
3841 @item -mtrap-large-shift
3842 @itemx -mhandle-large-shift
3843 @kindex -mtrap-large-shift
3844 @kindex -mhandle-large-shift
3845 @cindex bit shift overflow (88k)
3846 @cindex large bit shifts (88k)
3847 Include code to detect bit-shifts of more than 31 bits; respectively,
3848 trap such shifts or emit code to handle them properly. By default GNU CC
3849 makes no special provision for large bit shifts.
3851 @item -mwarn-passed-structs
3852 @kindex -mwarn-passed-structs
3853 @cindex structure passing (88k)
3854 Warn when a function passes a struct as an argument or result.
3855 Structure-passing conventions have changed during the evolution of the C
3856 language, and are often the source of portability problems. By default,
3857 GNU CC issues no such warning.
3860 @node RS/6000 and PowerPC Options
3861 @subsection IBM RS/6000 and PowerPC Options
3862 @cindex RS/6000 and PowerPC Options
3863 @cindex IBM RS/6000 and PowerPC Options
3865 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
3873 @itemx -mpowerpc-gpopt
3874 @itemx -mno-powerpc-gpopt
3875 @itemx -mpowerpc-gfxopt
3876 @itemx -mno-powerpc-gfxopt
3880 @kindex -mpowerpc-gpopt
3881 @kindex -mpowerpc-gfxopt
3882 GNU CC supports two related instruction set architectures for the
3883 RS/6000 and PowerPC. The @dfn{POWER} instruction set are those
3884 instructions supported by the @samp{rios} chip set used in the original
3885 RS/6000 systems and the @dfn{PowerPC} instruction set is the
3886 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
3887 the IBM 4xx microprocessors.
3889 Neither architecture is a subset of the other. However there is a
3890 large common subset of instructions supported by both. An MQ
3891 register is included in processors supporting the POWER architecture.
3893 You use these options to specify which instructions are available on the
3894 processor you are using. The default value of these options is
3895 determined when configuring GNU CC. Specifying the
3896 @samp{-mcpu=@var{cpu_type}} overrides the specification of these
3897 options. We recommend you use the @samp{-mcpu=@var{cpu_type}} option
3898 rather than the options listed above.
3900 The @samp{-mpower} option allows GNU CC to generate instructions that
3901 are found only in the POWER architecture and to use the MQ register.
3902 Specifying @samp{-mpower2} implies @samp{-power} and also allows GNU CC
3903 to generate instructions that are present in the POWER2 architecture but
3904 not the original POWER architecture.
3906 The @samp{-mpowerpc} option allows GNU CC to generate instructions that
3907 are found only in the 32-bit subset of the PowerPC architecture.
3908 Specifying @samp{-mpowerpc-gpopt} implies @samp{-mpowerpc} and also allows
3909 GNU CC to use the optional PowerPC architecture instructions in the
3910 General Purpose group, including floating-point square root. Specifying
3911 @samp{-mpowerpc-gfxopt} implies @samp{-mpowerpc} and also allows GNU CC to
3912 use the optional PowerPC architecture instructions in the Graphics
3913 group, including floating-point select.
3915 If you specify both @samp{-mno-power} and @samp{-mno-powerpc}, GNU CC
3916 will use only the instructions in the common subset of both
3917 architectures plus some special AIX common-mode calls, and will not use
3918 the MQ register. Specifying both @samp{-mpower} and @samp{-mpowerpc}
3919 permits GNU CC to use any instruction from either architecture and to
3920 allow use of the MQ register; specify this for the Motorola MPC601.
3922 @item -mnew-mnemonics
3923 @itemx -mold-mnemonics
3924 @kindex -mnew-mnemonics
3925 @kindex -mold-mnemonics
3926 Select which mnemonics to use in the generated assembler code.
3927 @samp{-mnew-mnemonics} requests output that uses the assembler mnemonics
3928 defined for the PowerPC architecture, while @samp{-mold-mnemonics}
3929 requests the assembler mnemonics defined for the POWER architecture.
3930 Instructions defined in only one architecture have only one mnemonic;
3931 GNU CC uses that mnemonic irrespective of which of these options is
3934 PowerPC assemblers support both the old and new mnemonics, as will later
3935 POWER assemblers. Current POWER assemblers only support the old
3936 mnemonics. Specify @samp{-mnew-mnemonics} if you have an assembler that
3937 supports them, otherwise specify @samp{-mold-mnemonics}.
3939 The default value of these options depends on how GNU CC was configured.
3940 Specifying @samp{-mcpu=@var{cpu_type}} sometimes overrides the value of
3941 these option. Unless you are building a cross-compiler, you should
3942 normally not specify either @samp{-mnew-mnemonics} or
3943 @samp{-mold-mnemonics}, but should instead accept the default.
3945 @item -mcpu=@var{cpu_type}
3946 Set architecture type, register usage, choice of mnemonics, and
3947 instruction scheduling parameters for machine type @var{cpu_type}.
3948 Supported values for @var{cpu_type} are @samp{rs6000}, @samp{rios1},
3949 @samp{rios2}, @samp{rsc}, @samp{601}, @samp{602}, @samp{603},
3950 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{power},
3951 @samp{power2}, @samp{powerpc}, @samp{403}, @samp{505}, @samp{801},
3952 @samp{821}, @samp{823}, and @samp{860} and @samp{common}.
3953 @samp{-mcpu=power}, @samp{-mcpu=power2}, and @samp{-mcpu=powerpc}
3954 specify generic POWER, POWER2 and pure PowerPC (i.e., not MPC601)
3955 architecture machine types, with an appropriate, generic processor model
3956 assumed for scheduling purposes.@refill
3958 @c overfull hbox here --bob 22 jul96
3959 @c original text between ignore ... end ignore
3961 Specifying any of the @samp{-mcpu=rios1}, @samp{-mcpu=rios2},
3962 @samp{-mcpu=rsc}, @samp{-mcpu=power}, or @samp{-mcpu=power2} options
3963 enables the @samp{-mpower} option and disables the @samp{-mpowerpc}
3964 option; @samp{-mcpu=601} enables both the @samp{-mpower} and
3965 @samp{-mpowerpc} options; all of @samp{-mcpu=602}, @samp{-mcpu=603},
3966 @samp{-mcpu=603e}, @samp{-mcpu=604}, @samp{-mcpu=604e},
3967 @samp{-mcpu=620}, @samp{-mcpu=403}, @samp{-mcpu=505}, @samp{-mcpu=801},
3968 @samp{-mcpu=821}, @samp{-mcpu=823}, @samp{-mcpu=860} and
3969 @samp{-mcpu=powerpc} enable the @samp{-mpowerpc} option and disable the
3970 @samp{-mpower} option; @samp{-mcpu=common} disables both the
3971 @samp{-mpower} and @samp{-mpowerpc} options.@refill
3973 @c changed paragraph
3974 Specifying any of the following options:
3975 @samp{-mcpu=rios1}, @samp{-mcpu=rios2}, @samp{-mcpu=rsc},
3976 @samp{-mcpu=power}, or @samp{-mcpu=power2}
3977 enables the @samp{-mpower} option and disables the @samp{-mpowerpc} option;
3978 @samp{-mcpu=601} enables both the @samp{-mpower} and @samp{-mpowerpc} options.
3979 All of @samp{-mcpu=602}, @samp{-mcpu=603}, @samp{-mcpu=603e},
3980 @samp{-mcpu=604}, @samp{-mcpu=620},
3981 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
3982 Exactly similarly, all of @samp{-mcpu=403},
3983 @samp{-mcpu=505}, @samp{-mcpu=821}, @samp{-mcpu=860} and @samp{-mcpu=powerpc}
3984 enable the @samp{-mpowerpc} option and disable the @samp{-mpower} option.
3985 @samp{-mcpu=common} disables both the
3986 @samp{-mpower} and @samp{-mpowerpc} options.@refill
3987 @c end changes to prevent overfull hboxes
3989 AIX versions 4 or greater selects @samp{-mcpu=common} by default, so
3990 that code will operate on all members of the RS/6000 and PowerPC
3991 families. In that case, GNU CC will use only the instructions in the
3992 common subset of both architectures plus some special AIX common-mode
3993 calls, and will not use the MQ register. GNU CC assumes a generic
3994 processor model for scheduling purposes.
3996 Specifying any of the options @samp{-mcpu=rios1}, @samp{-mcpu=rios2},
3997 @samp{-mcpu=rsc}, @samp{-mcpu=power}, or @samp{-mcpu=power2} also
3998 disables the @samp{new-mnemonics} option. Specifying @samp{-mcpu=601},
3999 @samp{-mcpu=602}, @samp{-mcpu=603}, @samp{-mcpu=603e}, @samp{-mcpu=604},
4000 @samp{620}, @samp{403}, or @samp{-mcpu=powerpc} also enables the
4001 @samp{new-mnemonics} option.@refill
4003 Specifying @samp{-mcpu=403}, @samp{-mcpu=821}, or @samp{-mcpu=860} also
4004 enables the @samp{-msoft-float} option.
4006 @item -mtune=@var{cpu_type}
4007 Set the instruction scheduling parameters for machine type
4008 @var{cpu_type}, but do not set the architecture type, register usage,
4009 choice of mnemonics like @samp{-mcpu=}@var{cpu_type} would. The same
4010 values for @var{cpu_type} are used for @samp{-mtune=}@var{cpu_type} as
4011 for @samp{-mcpu=}@var{cpu_type}. The @samp{-mtune=}@var{cpu_type}
4012 option overrides the @samp{-mcpu=}@var{cpu_type} option in terms of
4013 instruction scheduling parameters.
4016 @itemx -mno-fp-in-toc
4017 @itemx -mno-sum-in-toc
4018 @itemx -mminimal-toc
4019 Modify generation of the TOC (Table Of Contents), which is created for
4020 every executable file. The @samp{-mfull-toc} option is selected by
4021 default. In that case, GNU CC will allocate at least one TOC entry for
4022 each unique non-automatic variable reference in your program. GNU CC
4023 will also place floating-point constants in the TOC. However, only
4024 16,384 entries are available in the TOC.
4026 If you receive a linker error message that saying you have overflowed
4027 the available TOC space, you can reduce the amount of TOC space used
4028 with the @samp{-mno-fp-in-toc} and @samp{-mno-sum-in-toc} options.
4029 @samp{-mno-fp-in-toc} prevents GNU CC from putting floating-point
4030 constants in the TOC and @samp{-mno-sum-in-toc} forces GNU CC to
4031 generate code to calculate the sum of an address and a constant at
4032 run-time instead of putting that sum into the TOC. You may specify one
4033 or both of these options. Each causes GNU CC to produce very slightly
4034 slower and larger code at the expense of conserving TOC space.
4036 If you still run out of space in the TOC even when you specify both of
4037 these options, specify @samp{-mminimal-toc} instead. This option causes
4038 GNU CC to make only one TOC entry for every file. When you specify this
4039 option, GNU CC will produce code that is slower and larger but which
4040 uses extremely little TOC space. You may wish to use this option
4041 only on files that contain less frequently executed code. @refill
4045 On AIX, pass floating-point arguments to prototyped functions beyond the
4046 register save area (RSA) on the stack in addition to argument FPRs. The
4047 AIX calling convention was extended but not initially documented to
4048 handle an obscure K&R C case of calling a function that takes the
4049 address of its arguments with fewer arguments than declared. AIX XL
4050 compilers assume that floating point arguments which do not fit in the
4051 RSA are on the stack when they compile a subroutine without
4052 optimization. Because always storing floating-point arguments on the
4053 stack is inefficient and rarely needed, this option is not enabled by
4054 default and only is necessary when calling subroutines compiled by AIX
4055 XL compilers without optimization.
4058 Support @dfn{AIX Threads}. Link an application written to use
4059 @dfn{pthreads} with special libraries and startup code to enable the
4063 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE). Link an
4064 application written to use message passing with special startup code to
4065 enable the application to run. The system must have PE installed in the
4066 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
4067 must be overridden with the @samp{-specs=} option to specify the
4068 appropriate directory location. The Parallel Environment does not
4069 support threads, so the @samp{-mpe} option and the @samp{-mthreads}
4070 option are incompatible.
4074 Generate code that does not use (uses) the floating-point register set.
4075 Software floating point emulation is provided if you use the
4076 @samp{-msoft-float} option, and pass the option to GNU CC when linking.
4079 @itemx -mno-multiple
4080 Generate code that uses (does not use) the load multiple word
4081 instructions and the store multiple word instructions. These
4082 instructions are generated by default on POWER systems, and not
4083 generated on PowerPC systems. Do not use @samp{-mmultiple} on little
4084 endian PowerPC systems, since those instructions do not work when the
4085 processor is in little endian mode.
4089 Generate code that uses (does not use) the load string instructions and the
4090 store string word instructions to save multiple registers and do small block
4091 moves. These instructions are generated by default on POWER systems, and not
4092 generated on PowerPC systems. Do not use @samp{-mstring} on little endian
4093 PowerPC systems, since those instructions do not work when the processor is in
4098 Generate code that uses (does not use) the load or store instructions
4099 that update the base register to the address of the calculated memory
4100 location. These instructions are generated by default. If you use
4101 @samp{-mno-update}, there is a small window between the time that the
4102 stack pointer is updated and the address of the previous frame is
4103 stored, which means code that walks the stack frame across interrupts or
4104 signals may get corrupted data.
4107 @itemx -mno-fused-madd
4108 Generate code that uses (does not use) the floating point multiply and
4109 accumulate instructions. These instructions are generated by default if
4110 hardware floating is used.
4112 @item -mno-bit-align
4114 On System V.4 and embedded PowerPC systems do not (do) force structures
4115 and unions that contain bit fields to be aligned to the base type of the
4118 For example, by default a structure containing nothing but 8
4119 @code{unsigned} bitfields of length 1 would be aligned to a 4 byte
4120 boundary and have a size of 4 bytes. By using @samp{-mno-bit-align},
4121 the structure would be aligned to a 1 byte boundary and be one byte in
4124 @item -mno-strict-align
4125 @itemx -mstrict-align
4126 On System V.4 and embedded PowerPC systems do not (do) assume that
4127 unaligned memory references will be handled by the system.
4130 @itemx -mno-relocatable
4131 On embedded PowerPC systems generate code that allows (does not allow)
4132 the program to be relocated to a different address at runtime. If you
4133 use @samp{-mrelocatable} on any module, all objects linked together must
4134 be compiled with @samp{-mrelocatable} or @samp{-mrelocatable-lib}.
4136 @item -mrelocatable-lib
4137 @itemx -mno-relocatable-lib
4138 On embedded PowerPC systems generate code that allows (does not allow)
4139 the program to be relocated to a different address at runtime. Modules
4140 compiled with @samp{-mrelocatable-lib} can be linked with either modules
4141 compiled without @samp{-mrelocatable} and @samp{-mrelocatable-lib} or
4142 with modules compiled with the @samp{-mrelocatable} options.
4146 On System V.4 and embedded PowerPC systems do not (do) assume that
4147 register 2 contains a pointer to a global area pointing to the addresses
4148 used in the program.
4150 @item -mno-traceback
4152 On embedded PowerPC systems do not (do) generate a traceback tag before
4153 the start of the function. This tag can be used by the debugger to
4154 identify where the start of a function is.
4157 @itemx -mlittle-endian
4158 On System V.4 and embedded PowerPC systems compile code for the
4159 processor in little endian mode. The @samp{-mlittle-endian} option is
4160 the same as @samp{-mlittle}.
4164 On System V.4 and embedded PowerPC systems compile code for the
4165 processor in big endian mode. The @samp{-mbig-endian} option is
4166 the same as @samp{-mbig}.
4169 On System V.4 and embedded PowerPC systems compile code using calling
4170 conventions that adheres to the March 1995 draft of the System V
4171 Application Binary Interface, PowerPC processor supplement. This is the
4172 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
4174 @item -mcall-sysv-eabi
4175 Specify both @samp{-mcall-sysv} and @samp{-meabi} options.
4177 @item -mcall-sysv-noeabi
4178 Specify both @samp{-mcall-sysv} and @samp{-mno-eabi} options.
4181 On System V.4 and embedded PowerPC systems compile code using calling
4182 conventions that are similar to those used on AIX. This is the
4183 default if you configured GCC using @samp{powerpc-*-eabiaix}.
4185 @item -mcall-solaris
4186 On System V.4 and embedded PowerPC systems compile code for the Solaris
4190 On System V.4 and embedded PowerPC systems compile code for the
4191 Linux-based GNU system.
4194 @itemx -mno-prototype
4195 On System V.4 and embedded PowerPC systems assume that all calls to
4196 variable argument functions are properly prototyped. Otherwise, the
4197 compiler must insert an instruction before every non prototyped call to
4198 set or clear bit 6 of the condition code register (@var{CR}) to
4199 indicate whether floating point values were passed in the floating point
4200 registers in case the function takes a variable arguments. With
4201 @samp{-mprototype}, only calls to prototyped variable argument functions
4202 will set or clear the bit.
4205 On embedded PowerPC systems, assume that the startup module is called
4206 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
4207 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
4211 On embedded PowerPC systems, assume that the startup module is called
4212 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
4216 On embedded PowerPC systems, assume that the startup module is called
4217 @file{crt0.o} and the standard C libraries are @file{libads.a} and
4221 On embedded PowerPC systems, assume that the startup module is called
4222 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
4226 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
4227 header to indicate that @samp{eabi} extended relocations are used.
4231 On System V.4 and embedded PowerPC systems do (do not) adhere to the
4232 Embedded Applications Binary Interface (eabi) which is a set of
4233 modifications to the System V.4 specifications. Selecting @code{-meabi}
4234 means that the stack is aligned to an 8 byte boundary, a function
4235 @code{__eabi} is called to from @code{main} to set up the eabi
4236 environment, and the @samp{-msdata} option can use both @code{r2} and
4237 @code{r13} to point to two separate small data areas. Selecting
4238 @code{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
4239 do not call an initialization function from @code{main}, and the
4240 @samp{-msdata} option will only use @code{r13} to point to a single
4241 small data area. The @samp{-meabi} option is on by default if you
4242 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
4245 On System V.4 and embedded PowerPC systems, put small initialized
4246 @code{const} global and static data in the @samp{.sdata2} section, which
4247 is pointed to by register @code{r2}. Put small initialized
4248 non-@code{const} global and static data in the @samp{.sdata} section,
4249 which is pointed to by register @code{r13}. Put small uninitialized
4250 global and static data in the @samp{.sbss} section, which is adjacent to
4251 the @samp{.sdata} section. The @samp{-msdata=eabi} option is
4252 incompatible with the @samp{-mrelocatable} option. The
4253 @samp{-msdata=eabi} option also sets the @samp{-memb} option.
4256 On System V.4 and embedded PowerPC systems, put small global and static
4257 data in the @samp{.sdata} section, which is pointed to by register
4258 @code{r13}. Put small uninitialized global and static data in the
4259 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
4260 The @samp{-msdata=sysv} option is incompatible with the
4261 @samp{-mrelocatable} option.
4263 @item -msdata=default
4265 On System V.4 and embedded PowerPC systems, if @samp{-meabi} is used,
4266 compile code the same as @samp{-msdata=eabi}, otherwise compile code the
4267 same as @samp{-msdata=sysv}.
4270 On System V.4 and embedded PowerPC systems, put small global and static
4271 data in the @samp{.sdata} section. Put small uninitialized global and
4272 static data in the @samp{.sbss} section. Do not use register @code{r13}
4273 to address small data however. This is the default behavior unless
4274 other @samp{-msdata} options are used.
4278 On embedded PowerPC systems, put all initialized global and static data
4279 in the @samp{.data} section, and all uninitialized data in the
4280 @samp{.bss} section.
4283 @cindex smaller data references (PowerPC)
4284 @cindex .sdata/.sdata2 references (PowerPC)
4285 On embedded PowerPC systems, put global and static items less than or
4286 equal to @var{num} bytes into the small data or bss sections instead of
4287 the normal data or bss section. By default, @var{num} is 8. The
4288 @samp{-G @var{num}} switch is also passed to the linker.
4289 All modules should be compiled with the same @samp{-G @var{num}} value.
4292 @itemx -mno-regnames
4293 On System V.4 and embedded PowerPC systems do (do not) emit register
4294 names in the assembly language output using symbolic forms.
4297 @subsection IBM RT Options
4299 @cindex IBM RT options
4301 These @samp{-m} options are defined for the IBM RT PC:
4305 Use an in-line code sequence for integer multiplies. This is the
4308 @item -mcall-lib-mul
4309 Call @code{lmul$$} for integer multiples.
4311 @item -mfull-fp-blocks
4312 Generate full-size floating point data blocks, including the minimum
4313 amount of scratch space recommended by IBM. This is the default.
4315 @item -mminimum-fp-blocks
4316 Do not include extra scratch space in floating point data blocks. This
4317 results in smaller code, but slower execution, since scratch space must
4318 be allocated dynamically.
4320 @cindex @file{varargs.h} and RT PC
4321 @cindex @file{stdarg.h} and RT PC
4322 @item -mfp-arg-in-fpregs
4323 Use a calling sequence incompatible with the IBM calling convention in
4324 which floating point arguments are passed in floating point registers.
4325 Note that @code{varargs.h} and @code{stdargs.h} will not work with
4326 floating point operands if this option is specified.
4328 @item -mfp-arg-in-gregs
4329 Use the normal calling convention for floating point arguments. This is
4332 @item -mhc-struct-return
4333 Return structures of more than one word in memory, rather than in a
4334 register. This provides compatibility with the MetaWare HighC (hc)
4335 compiler. Use the option @samp{-fpcc-struct-return} for compatibility
4336 with the Portable C Compiler (pcc).
4338 @item -mnohc-struct-return
4339 Return some structures of more than one word in registers, when
4340 convenient. This is the default. For compatibility with the
4341 IBM-supplied compilers, use the option @samp{-fpcc-struct-return} or the
4342 option @samp{-mhc-struct-return}.
4346 @subsection MIPS Options
4347 @cindex MIPS options
4349 These @samp{-m} options are defined for the MIPS family of computers:
4352 @item -mcpu=@var{cpu type}
4353 Assume the defaults for the machine type @var{cpu type} when scheduling
4354 instructions. The choices for @var{cpu type} are @samp{r2000}, @samp{r3000},
4355 @samp{r4000}, @samp{r4400}, @samp{r4600}, and @samp{r6000}. While picking a
4356 specific @var{cpu type} will schedule things appropriately for that
4357 particular chip, the compiler will not generate any code that does not
4358 meet level 1 of the MIPS ISA (instruction set architecture) without
4359 the @samp{-mips2} or @samp{-mips3} switches being used.
4362 Issue instructions from level 1 of the MIPS ISA. This is the default.
4363 @samp{r3000} is the default @var{cpu type} at this ISA level.
4366 Issue instructions from level 2 of the MIPS ISA (branch likely, square
4367 root instructions). @samp{r6000} is the default @var{cpu type} at this
4371 Issue instructions from level 3 of the MIPS ISA (64 bit instructions).
4372 @samp{r4000} is the default @var{cpu type} at this ISA level.
4373 This option does not change the sizes of any of the C data types.
4376 Assume that 32 32-bit floating point registers are available. This is
4380 Assume that 32 64-bit floating point registers are available. This is
4381 the default when the @samp{-mips3} option is used.
4384 Assume that 32 32-bit general purpose registers are available. This is
4388 Assume that 32 64-bit general purpose registers are available. This is
4389 the default when the @samp{-mips3} option is used.
4392 Types long, int, and pointer are 64 bits. This works only if @samp{-mips3}
4396 Types long and pointer are 64 bits, and type int is 32 bits.
4397 This works only if @samp{-mips3} is also specified.
4400 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
4401 add normal debug information. This is the default for all
4402 platforms except for the OSF/1 reference platform, using the OSF/rose
4403 object format. If the either of the @samp{-gstabs} or @samp{-gstabs+}
4404 switches are used, the @file{mips-tfile} program will encapsulate the
4405 stabs within MIPS ECOFF.
4408 Generate code for the GNU assembler. This is the default on the OSF/1
4409 reference platform, using the OSF/rose object format. Also, this is
4410 the default if the configure option @samp{--with-gnu-as} is used.
4412 @item -msplit-addresses
4413 @itemx -mno-split-addresses
4414 Generate code to load the high and low parts of address constants separately.
4415 This allows @code{gcc} to optimize away redundant loads of the high order
4416 bits of addresses. This optimization requires GNU as and GNU ld.
4417 This optimization is enabled by default for some embedded targets where
4418 GNU as and GNU ld are standard.
4422 The @samp{-mrnames} switch says to output code using the MIPS software
4423 names for the registers, instead of the hardware names (ie, @var{a0}
4424 instead of @var{$4}). The only known assembler that supports this option
4425 is the Algorithmics assembler.
4429 The @samp{-mgpopt} switch says to write all of the data declarations
4430 before the instructions in the text section, this allows the MIPS
4431 assembler to generate one word memory references instead of using two
4432 words for short global or static data items. This is on by default if
4433 optimization is selected.
4437 For each non-inline function processed, the @samp{-mstats} switch
4438 causes the compiler to emit one line to the standard error file to
4439 print statistics about the program (number of registers saved, stack
4444 The @samp{-mmemcpy} switch makes all block moves call the appropriate
4445 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
4446 generating inline code.
4449 @itemx -mno-mips-tfile
4450 The @samp{-mno-mips-tfile} switch causes the compiler not
4451 postprocess the object file with the @file{mips-tfile} program,
4452 after the MIPS assembler has generated it to add debug support. If
4453 @file{mips-tfile} is not run, then no local variables will be
4454 available to the debugger. In addition, @file{stage2} and
4455 @file{stage3} objects will have the temporary file names passed to the
4456 assembler embedded in the object file, which means the objects will
4457 not compare the same. The @samp{-mno-mips-tfile} switch should only
4458 be used when there are bugs in the @file{mips-tfile} program that
4459 prevents compilation.
4462 Generate output containing library calls for floating point.
4463 @strong{Warning:} the requisite libraries are not part of GNU CC.
4464 Normally the facilities of the machine's usual C compiler are used, but
4465 this can't be done directly in cross-compilation. You must make your
4466 own arrangements to provide suitable library functions for
4470 Generate output containing floating point instructions. This is the
4471 default if you use the unmodified sources.
4474 @itemx -mno-abicalls
4475 Emit (or do not emit) the pseudo operations @samp{.abicalls},
4476 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
4477 position independent code.
4480 @itemx -mno-long-calls
4481 Do all calls with the @samp{JALR} instruction, which requires
4482 loading up a function's address into a register before the call.
4483 You need to use this switch, if you call outside of the current
4484 512 megabyte segment to functions that are not through pointers.
4487 @itemx -mno-half-pic
4488 Put pointers to extern references into the data section and load them
4489 up, rather than put the references in the text section.
4491 @item -membedded-pic
4492 @itemx -mno-embedded-pic
4493 Generate PIC code suitable for some embedded systems. All calls are made
4494 using PC relative address, and all data is addressed using the $gp register.
4495 This requires GNU as and GNU ld which do most of the work.
4497 @item -membedded-data
4498 @itemx -mno-embedded-data
4499 Allocate variables to the read-only data section first if possible, then
4500 next in the small data section if possible, otherwise in data. This gives
4501 slightly slower code than the default, but reduces the amount of RAM required
4502 when executing, and thus may be preferred for some embedded systems.
4504 @item -msingle-float
4505 @itemx -mdouble-float
4506 The @samp{-msingle-float} switch tells gcc to assume that the floating
4507 point coprocessor only supports single precision operations, as on the
4508 @samp{r4650} chip. The @samp{-mdouble-float} switch permits gcc to use
4509 double precision operations. This is the default.
4513 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
4514 as on the @samp{r4650} chip.
4517 Turns on @samp{-msingle-float}, @samp{-mmad}, and, at least for now,
4521 Compile code for the processor in little endian mode.
4522 The requisite libraries are assumed to exist.
4525 Compile code for the processor in big endian mode.
4526 The requisite libraries are assumed to exist.
4529 @cindex smaller data references (MIPS)
4530 @cindex gp-relative references (MIPS)
4531 Put global and static items less than or equal to @var{num} bytes into
4532 the small data or bss sections instead of the normal data or bss
4533 section. This allows the assembler to emit one word memory reference
4534 instructions based on the global pointer (@var{gp} or @var{$28}),
4535 instead of the normal two words used. By default, @var{num} is 8 when
4536 the MIPS assembler is used, and 0 when the GNU assembler is used. The
4537 @samp{-G @var{num}} switch is also passed to the assembler and linker.
4538 All modules should be compiled with the same @samp{-G @var{num}}
4542 Tell the MIPS assembler to not run it's preprocessor over user
4543 assembler files (with a @samp{.s} suffix) when assembling them.
4547 These options are defined by the macro
4548 @code{TARGET_SWITCHES} in the machine description. The default for the
4549 options is also defined by that macro, which enables you to change the
4554 @subsection Intel 386 Options
4555 @cindex i386 Options
4556 @cindex Intel 386 Options
4558 These @samp{-m} options are defined for the i386 family of computers:
4561 @item -mcpu=@var{cpu type}
4562 Assume the defaults for the machine type @var{cpu type} when scheduling
4563 instructions. The choices for @var{cpu type} are: @samp{i386},
4564 @samp{i486}, @samp{i586} (@samp{pentium}), @samp{pentium}, @samp{i686}
4565 (@samp{pentiumpro}) and @samp{pentiumpro}. While picking a specific
4566 @var{cpu type} will schedule things appropriately for that particular
4567 chip, the compiler will not generate any code that does not run on the
4568 i386 without the @samp{-march=@var{cpu type}} option being used.
4570 @item -march=@var{cpu type}
4571 Generate instructions for the machine type @var{cpu type}. The choices
4572 for @var{cpu type} are: @samp{i386}, @samp{i486}, @samp{pentium}, and
4573 @samp{pentiumpro}. Specifying @samp{-march=@var{cpu type}} implies
4574 @samp{-mcpu=@var{cpu type}}.
4580 Synonyms for -mcpu=i386, -mcpu=i486, -mcpu=pentium, and -mcpu=pentiumpro
4585 Control whether or not the compiler uses IEEE floating point
4586 comparisons. These handle correctly the case where the result of a
4587 comparison is unordered.
4590 Generate output containing library calls for floating point.
4591 @strong{Warning:} the requisite libraries are not part of GNU CC.
4592 Normally the facilities of the machine's usual C compiler are used, but
4593 this can't be done directly in cross-compilation. You must make your
4594 own arrangements to provide suitable library functions for
4597 On machines where a function returns floating point results in the 80387
4598 register stack, some floating point opcodes may be emitted even if
4599 @samp{-msoft-float} is used.
4601 @item -mno-fp-ret-in-387
4602 Do not use the FPU registers for return values of functions.
4604 The usual calling convention has functions return values of types
4605 @code{float} and @code{double} in an FPU register, even if there
4606 is no FPU. The idea is that the operating system should emulate
4609 The option @samp{-mno-fp-ret-in-387} causes such values to be returned
4610 in ordinary CPU registers instead.
4612 @item -mno-fancy-math-387
4613 Some 387 emulators do not support the @code{sin}, @code{cos} and
4614 @code{sqrt} instructions for the 387. Specify this option to avoid
4615 generating those instructions. This option is the default on FreeBSD.
4616 As of revision 2.6.1, these instructions are not generated unless you
4617 also use the @samp{-ffast-math} switch.
4619 @item -malign-double
4620 @itemx -mno-align-double
4621 Control whether GNU CC aligns @code{double}, @code{long double}, and
4622 @code{long long} variables on a two word boundary or a one word
4623 boundary. Aligning @code{double} variables on a two word boundary will
4624 produce code that runs somewhat faster on a @samp{Pentium} at the
4625 expense of more memory.
4627 @strong{Warning:} if you use the @samp{-malign-double} switch,
4628 structures containing the above types will be aligned differently than
4629 the published application binary interface specifications for the 386.
4632 @itemx -mno-svr3-shlib
4633 Control whether GNU CC places uninitialized locals into @code{bss} or
4634 @code{data}. @samp{-msvr3-shlib} places these locals into @code{bss}.
4635 These options are meaningful only on System V Release 3.
4637 @item -mno-wide-multiply
4638 @itemx -mwide-multiply
4639 Control whether GNU CC uses the @code{mul} and @code{imul} that produce
4640 64 bit results in @code{eax:edx} from 32 bit operands to do @code{long
4641 long} multiplies and 32-bit division by constants.
4644 Use a different function-calling convention, in which functions that
4645 take a fixed number of arguments return with the @code{ret} @var{num}
4646 instruction, which pops their arguments while returning. This saves one
4647 instruction in the caller since there is no need to pop the arguments
4650 You can specify that an individual function is called with this calling
4651 sequence with the function attribute @samp{stdcall}. You can also
4652 override the @samp{-mrtd} option by using the function attribute
4653 @samp{cdecl}. @xref{Function Attributes}
4655 @strong{Warning:} this calling convention is incompatible with the one
4656 normally used on Unix, so you cannot use it if you need to call
4657 libraries compiled with the Unix compiler.
4659 Also, you must provide function prototypes for all functions that
4660 take variable numbers of arguments (including @code{printf});
4661 otherwise incorrect code will be generated for calls to those
4664 In addition, seriously incorrect code will result if you call a
4665 function with too many arguments. (Normally, extra arguments are
4666 harmlessly ignored.)
4668 @item -mreg-alloc=@var{regs}
4669 Control the default allocation order of integer registers. The
4670 string @var{regs} is a series of letters specifying a register. The
4671 supported letters are: @code{a} allocate EAX; @code{b} allocate EBX;
4672 @code{c} allocate ECX; @code{d} allocate EDX; @code{S} allocate ESI;
4673 @code{D} allocate EDI; @code{B} allocate EBP.
4675 @item -mregparm=@var{num}
4676 Control how many registers are used to pass integer arguments. By
4677 default, no registers are used to pass arguments, and at most 3
4678 registers can be used. You can control this behavior for a specific
4679 function by using the function attribute @samp{regparm}. @xref{Function Attributes}
4681 @strong{Warning:} if you use this switch, and
4682 @var{num} is nonzero, then you must build all modules with the same
4683 value, including any libraries. This includes the system libraries and
4686 @item -malign-loops=@var{num}
4687 Align loops to a 2 raised to a @var{num} byte boundary. If
4688 @samp{-malign-loops} is not specified, the default is 2.
4690 @item -malign-jumps=@var{num}
4691 Align instructions that are only jumped to to a 2 raised to a @var{num}
4692 byte boundary. If @samp{-malign-jumps} is not specified, the default is
4693 2 if optimizing for a 386, and 4 if optimizing for a 486.
4695 @item -malign-functions=@var{num}
4696 Align the start of functions to a 2 raised to @var{num} byte boundary.
4697 If @samp{-malign-functions} is not specified, the default is 2 if optimizing
4698 for a 386, and 4 if optimizing for a 486.
4702 @subsection HPPA Options
4703 @cindex HPPA Options
4705 These @samp{-m} options are defined for the HPPA family of computers:
4709 Generate code for a PA 1.0 processor.
4712 Generate code for a PA 1.1 processor.
4715 Generate code suitable for big switch tables. Use this option only if
4716 the assembler/linker complain about out of range branches within a switch
4719 @item -mjump-in-delay
4720 Fill delay slots of function calls with unconditional jump instructions
4721 by modifying the return pointer for the function call to be the target
4722 of the conditional jump.
4724 @item -mdisable-fpregs
4725 Prevent floating point registers from being used in any manner. This is
4726 necessary for compiling kernels which perform lazy context switching of
4727 floating point registers. If you use this option and attempt to perform
4728 floating point operations, the compiler will abort.
4730 @item -mdisable-indexing
4731 Prevent the compiler from using indexing address modes. This avoids some
4732 rather obscure problems when compiling MIG generated code under MACH.
4734 @item -mno-space-regs
4735 Generate code that assumes the target has no space registers. This allows
4736 GCC to generate faster indirect calls and use unscaled index address modes.
4738 Such code is suitable for level 0 PA systems and kernels.
4740 @item -mfast-indirect-calls
4741 Generate code that assumes calls never cross space boundaries. This
4742 allows GCC to emit code which performs faster indirect calls.
4744 This option will not work in the presense of shared libraries or nested
4748 Optimize for space rather than execution time. Currently this only
4749 enables out of line function prologues and epilogues. This option is
4750 incompatible with PIC code generation and profiling.
4752 @item -mlong-load-store
4753 Generate 3-instruction load and store sequences as sometimes required by
4754 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
4757 @item -mportable-runtime
4758 Use the portable calling conventions proposed by HP for ELF systems.
4761 Enable the use of assembler directives only GAS understands.
4763 @item -mschedule=@var{cpu type}
4764 Schedule code according to the constraints for the machine type
4765 @var{cpu type}. The choices for @var{cpu type} are @samp{700} for
4766 7@var{n}0 machines, @samp{7100} for 7@var{n}5 machines, and @samp{7100}
4767 for 7@var{n}2 machines. @samp{7100} is the default for @var{cpu type}.
4769 Note the @samp{7100LC} scheduling information is incomplete and using
4770 @samp{7100LC} often leads to bad schedules. For now it's probably best
4771 to use @samp{7100} instead of @samp{7100LC} for the 7@var{n}2 machines.
4774 Enable the optimization pass in the HPUX linker. Note this makes symbolic
4775 debugging impossible. It also triggers a bug in the HPUX 8 and HPUX 9 linkers
4776 in which they give bogus error messages when linking some programs.
4779 Generate output containing library calls for floating point.
4780 @strong{Warning:} the requisite libraries are not available for all HPPA
4781 targets. Normally the facilities of the machine's usual C compiler are
4782 used, but this cannot be done directly in cross-compilation. You must make
4783 your own arrangements to provide suitable library functions for
4784 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
4785 does provide software floating point support.
4787 @samp{-msoft-float} changes the calling convention in the output file;
4788 therefore, it is only useful if you compile @emph{all} of a program with
4789 this option. In particular, you need to compile @file{libgcc.a}, the
4790 library that comes with GNU CC, with @samp{-msoft-float} in order for
4794 @node Intel 960 Options
4795 @subsection Intel 960 Options
4797 These @samp{-m} options are defined for the Intel 960 implementations:
4800 @item -m@var{cpu type}
4801 Assume the defaults for the machine type @var{cpu type} for some of
4802 the other options, including instruction scheduling, floating point
4803 support, and addressing modes. The choices for @var{cpu type} are
4804 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
4805 @samp{sa}, and @samp{sb}.
4811 The @samp{-mnumerics} option indicates that the processor does support
4812 floating-point instructions. The @samp{-msoft-float} option indicates
4813 that floating-point support should not be assumed.
4815 @item -mleaf-procedures
4816 @itemx -mno-leaf-procedures
4817 Do (or do not) attempt to alter leaf procedures to be callable with the
4818 @code{bal} instruction as well as @code{call}. This will result in more
4819 efficient code for explicit calls when the @code{bal} instruction can be
4820 substituted by the assembler or linker, but less efficient code in other
4821 cases, such as calls via function pointers, or using a linker that doesn't
4822 support this optimization.
4825 @itemx -mno-tail-call
4826 Do (or do not) make additional attempts (beyond those of the
4827 machine-independent portions of the compiler) to optimize tail-recursive
4828 calls into branches. You may not want to do this because the detection of
4829 cases where this is not valid is not totally complete. The default is
4830 @samp{-mno-tail-call}.
4832 @item -mcomplex-addr
4833 @itemx -mno-complex-addr
4834 Assume (or do not assume) that the use of a complex addressing mode is a
4835 win on this implementation of the i960. Complex addressing modes may not
4836 be worthwhile on the K-series, but they definitely are on the C-series.
4837 The default is currently @samp{-mcomplex-addr} for all processors except
4841 @itemx -mno-code-align
4842 Align code to 8-byte boundaries for faster fetching (or don't bother).
4843 Currently turned on by default for C-series implementations only.
4846 @item -mclean-linkage
4847 @itemx -mno-clean-linkage
4848 These options are not fully implemented.
4852 @itemx -mic2.0-compat
4853 @itemx -mic3.0-compat
4854 Enable compatibility with iC960 v2.0 or v3.0.
4858 Enable compatibility with the iC960 assembler.
4860 @item -mstrict-align
4861 @itemx -mno-strict-align
4862 Do not permit (do permit) unaligned accesses.
4865 Enable structure-alignment compatibility with Intel's gcc release version
4866 1.3 (based on gcc 1.37). This option implies @samp{-mstrict-align}.
4869 @node DEC Alpha Options
4870 @subsection DEC Alpha Options
4872 These @samp{-m} options are defined for the DEC Alpha implementations:
4875 @item -mno-soft-float
4877 Use (do not use) the hardware floating-point instructions for
4878 floating-point operations. When @code{-msoft-float} is specified,
4879 functions in @file{libgcc1.c} will be used to perform floating-point
4880 operations. Unless they are replaced by routines that emulate the
4881 floating-point operations, or compiled in such a way as to call such
4882 emulations routines, these routines will issue floating-point
4883 operations. If you are compiling for an Alpha without floating-point
4884 operations, you must ensure that the library is built so as not to call
4887 Note that Alpha implementations without floating-point operations are
4888 required to have floating-point registers.
4892 Generate code that uses (does not use) the floating-point register set.
4893 @code{-mno-fp-regs} implies @code{-msoft-float}. If the floating-point
4894 register set is not used, floating point operands are passed in integer
4895 registers as if they were integers and floating-point results are passed
4896 in $0 instead of $f0. This is a non-standard calling sequence, so any
4897 function with a floating-point argument or return value called by code
4898 compiled with @code{-mno-fp-regs} must also be compiled with that
4901 A typical use of this option is building a kernel that does not use,
4902 and hence need not save and restore, any floating-point registers.
4905 The Alpha architecture implements floating-point hardware optimized for
4906 maximum performance. It is mostly compliant with the IEEE floating
4907 point standard. However, for full compliance, software assistance is
4908 required. This option generates code fully IEEE compliant code
4909 @emph{except} that the @var{inexact flag} is not maintained (see below).
4910 If this option is turned on, the CPP macro @code{_IEEE_FP} is defined
4911 during compilation. The option is a shorthand for: @samp{-D_IEEE_FP
4912 -mfp-trap-mode=su -mtrap-precision=i -mieee-conformant}. The resulting
4913 code is less efficient but is able to correctly support denormalized
4914 numbers and exceptional IEEE values such as not-a-number and plus/minus
4915 infinity. Other Alpha compilers call this option
4916 @code{-ieee_with_no_inexact}.
4918 @item -mieee-with-inexact
4919 @c overfull hbox here --bob 22 jul96
4920 @c original text between ignore ... end ignore
4922 This is like @samp{-mieee} except the generated code also maintains the
4923 IEEE @var{inexact flag}. Turning on this option causes the generated
4924 code to implement fully-compliant IEEE math. The option is a shorthand
4925 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus @samp{-mieee-conformant},
4926 @samp{-mfp-trap-mode=sui}, and @samp{-mtrap-precision=i}. On some Alpha
4927 implementations the resulting code may execute significantly slower than
4928 the code generated by default. Since there is very little code that
4929 depends on the @var{inexact flag}, you should normally not specify this
4930 option. Other Alpha compilers call this option
4931 @samp{-ieee_with_inexact}.
4933 @c changed paragraph
4934 This is like @samp{-mieee} except the generated code also maintains the
4935 IEEE @var{inexact flag}. Turning on this option causes the generated
4936 code to implement fully-compliant IEEE math. The option is a shorthand
4937 for @samp{-D_IEEE_FP -D_IEEE_FP_INEXACT} plus the three following:
4938 @samp{-mieee-conformant},
4939 @samp{-mfp-trap-mode=sui},
4940 and @samp{-mtrap-precision=i}.
4941 On some Alpha implementations the resulting code may execute
4942 significantly slower than the code generated by default. Since there
4943 is very little code that depends on the @var{inexact flag}, you should
4944 normally not specify this option. Other Alpha compilers call this
4945 option @samp{-ieee_with_inexact}.
4946 @c end changes to prevent overfull hboxes
4948 @item -mfp-trap-mode=@var{trap mode}
4949 This option controls what floating-point related traps are enabled.
4950 Other Alpha compilers call this option @samp{-fptm }@var{trap mode}.
4951 The trap mode can be set to one of four values:
4955 This is the default (normal) setting. The only traps that are enabled
4956 are the ones that cannot be disabled in software (e.g., division by zero
4960 In addition to the traps enabled by @samp{n}, underflow traps are enabled
4964 Like @samp{su}, but the instructions are marked to be safe for software
4965 completion (see Alpha architecture manual for details).
4968 Like @samp{su}, but inexact traps are enabled as well.
4971 @item -mfp-rounding-mode=@var{rounding mode}
4972 Selects the IEEE rounding mode. Other Alpha compilers call this option
4973 @samp{-fprm }@var{rounding mode}. The @var{rounding mode} can be one
4978 Normal IEEE rounding mode. Floating point numbers are rounded towards
4979 the nearest machine number or towards the even machine number in case
4983 Round towards minus infinity.
4986 Chopped rounding mode. Floating point numbers are rounded towards zero.
4989 Dynamic rounding mode. A field in the floating point control register
4990 (@var{fpcr}, see Alpha architecture reference manual) controls the
4991 rounding mode in effect. The C library initializes this register for
4992 rounding towards plus infinity. Thus, unless your program modifies the
4993 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.@end table
4995 @item -mtrap-precision=@var{trap precision}
4996 In the Alpha architecture, floating point traps are imprecise. This
4997 means without software assistance it is impossible to recover from a
4998 floating trap and program execution normally needs to be terminated.
4999 GNU CC can generate code that can assist operating system trap handlers
5000 in determining the exact location that caused a floating point trap.
5001 Depending on the requirements of an application, different levels of
5002 precisions can be selected:
5006 Program precision. This option is the default and means a trap handler
5007 can only identify which program caused a floating point exception.
5010 Function precision. The trap handler can determine the function that
5011 caused a floating point exception.
5014 Instruction precision. The trap handler can determine the exact
5015 instruction that caused a floating point exception.
5018 Other Alpha compilers provide the equivalent options called
5019 @samp{-scope_safe} and @samp{-resumption_safe}.
5021 @item -mieee-conformant
5022 This option marks the generated code as IEEE conformant. You must not
5023 use this option unless you also specify @samp{-mtrap-precision=i} and either
5024 @samp{-mfp-trap-mode=su} or @samp{-mfp-trap-mode=sui}. Its only effect
5025 is to emit the line @samp{.eflag 48} in the function prologue of the
5026 generated assembly file. Under DEC Unix, this has the effect that
5027 IEEE-conformant math library routines will be linked in.
5029 @item -mbuild-constants
5030 Normally GNU CC examines a 32- or 64-bit integer constant to
5031 see if it can construct it from smaller constants in two or three
5032 instructions. If it cannot, it will output the constant as a literal and
5033 generate code to load it from the data segment at runtime.
5035 Use this option to require GNU CC to construct @emph{all} integer constants
5036 using code, even if it takes more instructions (the maximum is six).
5038 You would typically use this option to build a shared library dynamic
5039 loader. Itself a shared library, it must relocate itself in memory
5040 before it can find the variables and constants in its own data segment.
5044 Select whether to generate code to be assembled by the vendor-supplied
5045 assembler (@samp{-malpha-as}) or by the GNU assembler @samp{-mgas}.
5053 Indicate whether GNU CC should generate code to use the optional BWX,
5054 CIX, and MAX instruction sets. The default is to use the instruction sets
5055 supported by the CPU type specified via @samp{-mcpu=} option or that
5056 of the CPU on which GNU CC was built if none was specified.
5058 @item -mcpu=@var{cpu_type}
5059 Set the instruction set, register set, and instruction scheduling
5060 parameters for machine type @var{cpu_type}. You can specify either the
5061 @samp{EV} style name or the corresponding chip number. GNU CC
5062 supports scheduling parameters for the EV4 and EV5 family of processors
5063 and will choose the default values for the instruction set from
5064 the processor you specify. If you do not specify a processor type,
5065 GNU CC will default to the processor on which the compiler was built.
5067 Supported values for @var{cpu_type} are
5072 Schedules as an EV4 and has no instruction set extensions.
5076 Schedules as an EV5 and has no instruction set extensions.
5080 Schedules as an EV5 and supports the BWX extension.
5085 Schedules as an EV5 and supports the BWX and MAX extensions.
5089 Schedules as an EV5 (until Digital releases the scheduling parameters
5090 for the EV6) and supports the BWX, CIX, and MAX extensions.
5093 @item -mmemory-latency=@var{time}
5094 Sets the latency the scheduler should assume for typical memory
5095 references as seen by the application. This number is highly
5096 dependant on the memory access patterns used by the application
5097 and the size of the external cache on the machine.
5099 Valid options for @var{time} are
5103 A decimal number representing clock cycles.
5109 The compiler contains estimates of the number of clock cycles for
5110 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
5111 (also called Dcache, Scache, and Bcache), as well as to main memory.
5112 Note that L3 is only valid for EV5.
5117 @node Clipper Options
5118 @subsection Clipper Options
5120 These @samp{-m} options are defined for the Clipper implementations:
5124 Produce code for a C300 Clipper processor. This is the default.
5127 Produce code for a C400 Clipper processor i.e. use floating point
5131 @node H8/300 Options
5132 @subsection H8/300 Options
5134 These @samp{-m} options are defined for the H8/300 implementations:
5138 Shorten some address references at link time, when possible; uses the
5139 linker option @samp{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
5140 ld.info, Using ld}, for a fuller description.
5143 Generate code for the H8/300H.
5146 Generate code for the H8/S.
5149 Make @code{int} data 32 bits by default.
5152 On the h8/300h, use the same alignment rules as for the h8/300.
5153 The default for the h8/300h is to align longs and floats on 4 byte boundaries.
5154 @samp{-malign-300} causes them to be aligned on 2 byte boundaries.
5155 This option has no effect on the h8/300.
5159 @subsection SH Options
5161 These @samp{-m} options are defined for the SH implementations:
5165 Generate code for the SH1.
5168 Generate code for the SH2.
5171 Generate code for the SH3.
5174 Generate code for the SH3e.
5177 Compile code for the processor in big endian mode.
5180 Compile code for the processor in little endian mode.
5183 Shorten some address references at link time, when possible; uses the
5184 linker option @samp{-relax}.
5187 @node System V Options
5188 @subsection Options for System V
5190 These additional options are available on System V Release 4 for
5191 compatibility with other compilers on those systems:
5195 Create a shared object.
5196 It is recommended that @samp{-symbolic} or @samp{-shared} be used instead.
5199 Identify the versions of each tool used by the compiler, in a
5200 @code{.ident} assembler directive in the output.
5203 Refrain from adding @code{.ident} directives to the output file (this is
5206 @item -YP,@var{dirs}
5207 Search the directories @var{dirs}, and no others, for libraries
5208 specified with @samp{-l}.
5211 Look in the directory @var{dir} to find the M4 preprocessor.
5212 The assembler uses this option.
5213 @c This is supposed to go with a -Yd for predefined M4 macro files, but
5214 @c the generic assembler that comes with Solaris takes just -Ym.
5218 @subsection V850 Options
5219 @cindex V850 Options
5221 These @samp{-m} options are defined for V850 implementations:
5225 @itemx -mno-long-calls
5226 Treat all calls as being far away (near). If calls are assumed to be
5227 far away, the compiler will always load the functions address up into a
5228 register, and call indirect through the pointer.
5232 Do not optimize (do optimize) basic blocks that use the same index
5233 pointer 4 or more times to copy pointer into the @code{ep} register, and
5234 use the shorter @code{sld} and @code{sst} instructions. The @samp{-mep}
5235 option is on by default if you optimize.
5237 @item -mno-prolog-function
5238 @itemx -mprolog-function
5239 Do not use (do use) external functions to save and restore registers at
5240 the prolog and epilog of a function. The external functions are slower,
5241 but use less code space if more than one function saves the same number
5242 of registers. The @samp{-mprolog-function} option is on by default if
5246 Try to make the code as small as possible. At present, this just turns
5247 on the @samp{-mep} and @samp{-mprolog-function} options.
5250 Put static or global variables whose size is @var{n} bytes or less into
5251 the tiny data area that register @code{ep} points to. The tiny data
5252 area can hold up to 256 bytes in total (128 bytes for byte references).
5255 Put static or global variables whose size is @var{n} bytes or less into
5256 the small data area that register @code{gp} points to. The small data
5257 area can hold up to 64 kilobytes.
5260 Put static or global variables whose size is @var{n} bytes or less into
5261 the first 32 kilobytes of memory.
5264 Specify that the target processor is the V850.
5267 Generate code suitable for big switch tables. Use this option only if
5268 the assembler/linker complain about out of range branches within a switch
5272 @node Code Gen Options
5273 @section Options for Code Generation Conventions
5274 @cindex code generation conventions
5275 @cindex options, code generation
5276 @cindex run-time options
5278 These machine-independent options control the interface conventions
5279 used in code generation.
5281 Most of them have both positive and negative forms; the negative form
5282 of @samp{-ffoo} would be @samp{-fno-foo}. In the table below, only
5283 one of the forms is listed---the one which is not the default. You
5284 can figure out the other form by either removing @samp{no-} or adding
5289 Enable exception handling, and generate extra code needed to propagate
5290 exceptions. If you do not specify this option, GNU CC enables it by
5291 default for languages like C++ that normally require exception handling,
5292 and disabled for languages like C that do not normally require it.
5293 However, when compiling C code that needs to interoperate properly with
5294 exception handlers written in C++, you may need to enable this option.
5295 You may also wish to disable this option is you are compiling older C++
5296 programs that don't use exception handling.
5298 @item -fpcc-struct-return
5299 Return ``short'' @code{struct} and @code{union} values in memory like
5300 longer ones, rather than in registers. This convention is less
5301 efficient, but it has the advantage of allowing intercallability between
5302 GNU CC-compiled files and files compiled with other compilers.
5304 The precise convention for returning structures in memory depends
5305 on the target configuration macros.
5307 Short structures and unions are those whose size and alignment match
5308 that of some integer type.
5310 @item -freg-struct-return
5311 Use the convention that @code{struct} and @code{union} values are
5312 returned in registers when possible. This is more efficient for small
5313 structures than @samp{-fpcc-struct-return}.
5315 If you specify neither @samp{-fpcc-struct-return} nor its contrary
5316 @samp{-freg-struct-return}, GNU CC defaults to whichever convention is
5317 standard for the target. If there is no standard convention, GNU CC
5318 defaults to @samp{-fpcc-struct-return}, except on targets where GNU CC
5319 is the principal compiler. In those cases, we can choose the standard,
5320 and we chose the more efficient register return alternative.
5323 Allocate to an @code{enum} type only as many bytes as it needs for the
5324 declared range of possible values. Specifically, the @code{enum} type
5325 will be equivalent to the smallest integer type which has enough room.
5327 @item -fshort-double
5328 Use the same size for @code{double} as for @code{float}.
5331 Requests that the data and non-@code{const} variables of this
5332 compilation be shared data rather than private data. The distinction
5333 makes sense only on certain operating systems, where shared data is
5334 shared between processes running the same program, while private data
5335 exists in one copy per process.
5338 Allocate even uninitialized global variables in the bss section of the
5339 object file, rather than generating them as common blocks. This has the
5340 effect that if the same variable is declared (without @code{extern}) in
5341 two different compilations, you will get an error when you link them.
5342 The only reason this might be useful is if you wish to verify that the
5343 program will work on other systems which always work this way.
5346 Ignore the @samp{#ident} directive.
5348 @item -fno-gnu-linker
5349 Do not output global initializations (such as C++ constructors and
5350 destructors) in the form used by the GNU linker (on systems where the GNU
5351 linker is the standard method of handling them). Use this option when
5352 you want to use a non-GNU linker, which also requires using the
5353 @code{collect2} program to make sure the system linker includes
5354 constructors and destructors. (@code{collect2} is included in the GNU CC
5355 distribution.) For systems which @emph{must} use @code{collect2}, the
5356 compiler driver @code{gcc} is configured to do this automatically.
5358 @item -finhibit-size-directive
5359 Don't output a @code{.size} assembler directive, or anything else that
5360 would cause trouble if the function is split in the middle, and the
5361 two halves are placed at locations far apart in memory. This option is
5362 used when compiling @file{crtstuff.c}; you should not need to use it
5366 Put extra commentary information in the generated assembly code to
5367 make it more readable. This option is generally only of use to those
5368 who actually need to read the generated assembly code (perhaps while
5369 debugging the compiler itself).
5371 @samp{-fno-verbose-asm}, the default, causes the
5372 extra information to be omitted and is useful when comparing two assembler
5376 Consider all memory references through pointers to be volatile.
5378 @item -fvolatile-global
5379 Consider all memory references to extern and global data items to
5383 @cindex global offset table
5385 Generate position-independent code (PIC) suitable for use in a shared
5386 library, if supported for the target machine. Such code accesses all
5387 constant addresses through a global offset table (GOT). The dynamic
5388 loader resolves the GOT entries when the program starts (the dynamic
5389 loader is not part of GNU CC; it is part of the operating system). If
5390 the GOT size for the linked executable exceeds a machine-specific
5391 maximum size, you get an error message from the linker indicating that
5392 @samp{-fpic} does not work; in that case, recompile with @samp{-fPIC}
5393 instead. (These maximums are 16k on the m88k, 8k on the Sparc, and 32k
5394 on the m68k and RS/6000. The 386 has no such limit.)
5396 Position-independent code requires special support, and therefore works
5397 only on certain machines. For the 386, GNU CC supports PIC for System V
5398 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
5399 position-independent.
5402 If supported for the target machine, emit position-independent code,
5403 suitable for dynamic linking and avoiding any limit on the size of the
5404 global offset table. This option makes a difference on the m68k, m88k,
5407 Position-independent code requires special support, and therefore works
5408 only on certain machines.
5410 @item -ffixed-@var{reg}
5411 Treat the register named @var{reg} as a fixed register; generated code
5412 should never refer to it (except perhaps as a stack pointer, frame
5413 pointer or in some other fixed role).
5415 @var{reg} must be the name of a register. The register names accepted
5416 are machine-specific and are defined in the @code{REGISTER_NAMES}
5417 macro in the machine description macro file.
5419 This flag does not have a negative form, because it specifies a
5422 @item -fcall-used-@var{reg}
5423 Treat the register named @var{reg} as an allocable register that is
5424 clobbered by function calls. It may be allocated for temporaries or
5425 variables that do not live across a call. Functions compiled this way
5426 will not save and restore the register @var{reg}.
5428 Use of this flag for a register that has a fixed pervasive role in the
5429 machine's execution model, such as the stack pointer or frame pointer,
5430 will produce disastrous results.
5432 This flag does not have a negative form, because it specifies a
5435 @item -fcall-saved-@var{reg}
5436 Treat the register named @var{reg} as an allocable register saved by
5437 functions. It may be allocated even for temporaries or variables that
5438 live across a call. Functions compiled this way will save and restore
5439 the register @var{reg} if they use it.
5441 Use of this flag for a register that has a fixed pervasive role in the
5442 machine's execution model, such as the stack pointer or frame pointer,
5443 will produce disastrous results.
5445 A different sort of disaster will result from the use of this flag for
5446 a register in which function values may be returned.
5448 This flag does not have a negative form, because it specifies a
5452 Pack all structure members together without holes. Usually you would
5453 not want to use this option, since it makes the code suboptimal, and
5454 the offsets of structure members won't agree with system libraries.
5456 @item -fcheck-memory-usage
5457 Generate extra code to check each memory access. GNU CC will generate
5458 code that is suitable for a detector of bad memory accesses such as
5459 @file{Checker}. If you specify this option, you can not use the
5460 @code{asm} or @code{__asm__} keywords.
5462 You must also specify this option when you compile functions you call that
5463 have side effects. If you do not, you may get erroneous messages from
5464 the detector. Normally, you should compile all your code with this option.
5465 If you use functions from a library that have side-effects (such as
5466 @code{read}), you may not be able to recompile the library and
5467 specify this option. In that case, you can enable the
5468 @samp{-fprefix-function-name} option, which requests GNU CC to encapsulate
5469 your code and make other functions look as if they were compiled with
5470 @samp{-fcheck-memory-usage}. This is done by calling ``stubs'',
5471 which are provided by the detector. If you cannot find or build
5472 stubs for every function you call, you may have to specify
5473 @samp{-fcheck-memory-usage} without @samp{-fprefix-function-name}.
5475 @item -fprefix-function-name
5476 Request GNU CC to add a prefix to the symbols generated for function names.
5477 GNU CC adds a prefix to the names of functions defined as well as
5478 functions called. Code compiled with this option and code compiled
5479 without the option can't be linked together, unless or stubs are used.
5481 If you compile the following code with @samp{-fprefix-function-name}
5483 extern void bar (int);
5493 GNU CC will compile the code as if it was written:
5495 extern void prefix_bar (int);
5499 return prefix_bar (a + 5);
5502 This option is designed to be used with @samp{-fcheck-memory-usage}.
5505 Generate code to verify that you do not go beyond the boundary of the
5506 stack. You should specify this flag if you are running in an
5507 environment with multiple threads, but only rarely need to specify it in
5508 a single-threaded environment since stack overflow is automatically
5509 detected on nearly all systems if there is only one stack.
5512 Enable exception handling. For some targets, this implies
5513 generation of frame unwind information for all functions, which can produce
5514 significant data size overhead, though it does not affect execution.
5516 This option is on by default for languages that support exception
5517 handling (such as C++), and off for those that don't (such as C).
5521 Control whether virtual function definitions in classes are used to
5522 generate code, or only to define interfaces for their callers. (C++
5525 These options are provided for compatibility with @code{cfront} 1.x
5526 usage; the recommended alternative GNU C++ usage is in flux. @xref{C++
5527 Interface,,Declarations and Definitions in One Header}.
5529 With @samp{+e0}, virtual function definitions in classes are declared
5530 @code{extern}; the declaration is used only as an interface
5531 specification, not to generate code for the virtual functions (in this
5534 With @samp{+e1}, G++ actually generates the code implementing virtual
5535 functions defined in the code, and makes them publicly visible.
5537 @cindex aliasing of parameters
5538 @cindex parameters, aliased
5539 @item -fargument-alias
5540 @itemx -fargument-noalias
5541 @itemx -fargument-noalias-global
5542 Specify the possible relationships among parameters and between
5543 parameters and global data.
5545 @samp{-fargument-alias} specifies that arguments (parameters) may
5546 alias each other and may alias global storage.
5547 @samp{-fargument-noalias} specifies that arguments do not alias
5548 each other, but may alias global storage.
5549 @samp{-fargument-noalias-global} specifies that arguments do not
5550 alias each other and do not alias global storage.
5552 Each language will automatically use whatever option is required by
5553 the language standard. You should not need to use these options yourself.
5556 @node Environment Variables
5557 @section Environment Variables Affecting GNU CC
5558 @cindex environment variables
5560 This section describes several environment variables that affect how GNU
5561 CC operates. They work by specifying directories or prefixes to use
5562 when searching for various kinds of files.
5565 Note that you can also specify places to search using options such as
5566 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
5567 take precedence over places specified using environment variables, which
5568 in turn take precedence over those specified by the configuration of GNU
5572 Note that you can also specify places to search using options such as
5573 @samp{-B}, @samp{-I} and @samp{-L} (@pxref{Directory Options}). These
5574 take precedence over places specified using environment variables, which
5575 in turn take precedence over those specified by the configuration of GNU
5582 If @code{TMPDIR} is set, it specifies the directory to use for temporary
5583 files. GNU CC uses temporary files to hold the output of one stage of
5584 compilation which is to be used as input to the next stage: for example,
5585 the output of the preprocessor, which is the input to the compiler
5588 @item GCC_EXEC_PREFIX
5589 @findex GCC_EXEC_PREFIX
5590 If @code{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
5591 names of the subprograms executed by the compiler. No slash is added
5592 when this prefix is combined with the name of a subprogram, but you can
5593 specify a prefix that ends with a slash if you wish.
5595 If GNU CC cannot find the subprogram using the specified prefix, it
5596 tries looking in the usual places for the subprogram.
5598 The default value of @code{GCC_EXEC_PREFIX} is
5599 @file{@var{prefix}/lib/gcc-lib/} where @var{prefix} is the value
5600 of @code{prefix} when you ran the @file{configure} script.
5602 Other prefixes specified with @samp{-B} take precedence over this prefix.
5604 This prefix is also used for finding files such as @file{crt0.o} that are
5607 In addition, the prefix is used in an unusual way in finding the
5608 directories to search for header files. For each of the standard
5609 directories whose name normally begins with @samp{/usr/local/lib/gcc-lib}
5610 (more precisely, with the value of @code{GCC_INCLUDE_DIR}), GNU CC tries
5611 replacing that beginning with the specified prefix to produce an
5612 alternate directory name. Thus, with @samp{-Bfoo/}, GNU CC will search
5613 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
5614 These alternate directories are searched first; the standard directories
5618 @findex COMPILER_PATH
5619 The value of @code{COMPILER_PATH} is a colon-separated list of
5620 directories, much like @code{PATH}. GNU CC tries the directories thus
5621 specified when searching for subprograms, if it can't find the
5622 subprograms using @code{GCC_EXEC_PREFIX}.
5625 @findex LIBRARY_PATH
5626 The value of @code{LIBRARY_PATH} is a colon-separated list of
5627 directories, much like @code{PATH}. When configured as a native compiler,
5628 GNU CC tries the directories thus specified when searching for special
5629 linker files, if it can't find them using @code{GCC_EXEC_PREFIX}. Linking
5630 using GNU CC also uses these directories when searching for ordinary
5631 libraries for the @samp{-l} option (but directories specified with
5632 @samp{-L} come first).
5634 @item C_INCLUDE_PATH
5635 @itemx CPLUS_INCLUDE_PATH
5636 @itemx OBJC_INCLUDE_PATH
5637 @findex C_INCLUDE_PATH
5638 @findex CPLUS_INCLUDE_PATH
5639 @findex OBJC_INCLUDE_PATH
5640 @c @itemx OBJCPLUS_INCLUDE_PATH
5641 These environment variables pertain to particular languages. Each
5642 variable's value is a colon-separated list of directories, much like
5643 @code{PATH}. When GNU CC searches for header files, it tries the
5644 directories listed in the variable for the language you are using, after
5645 the directories specified with @samp{-I} but before the standard header
5648 @item DEPENDENCIES_OUTPUT
5649 @findex DEPENDENCIES_OUTPUT
5650 @cindex dependencies for make as output
5651 If this variable is set, its value specifies how to output dependencies
5652 for Make based on the header files processed by the compiler. This
5653 output looks much like the output from the @samp{-M} option
5654 (@pxref{Preprocessor Options}), but it goes to a separate file, and is
5655 in addition to the usual results of compilation.
5657 The value of @code{DEPENDENCIES_OUTPUT} can be just a file name, in
5658 which case the Make rules are written to that file, guessing the target
5659 name from the source file name. Or the value can have the form
5660 @samp{@var{file} @var{target}}, in which case the rules are written to
5661 file @var{file} using @var{target} as the target name.
5664 @node Running Protoize
5665 @section Running Protoize
5667 The program @code{protoize} is an optional part of GNU C. You can use
5668 it to add prototypes to a program, thus converting the program to ANSI
5669 C in one respect. The companion program @code{unprotoize} does the
5670 reverse: it removes argument types from any prototypes that are found.
5672 When you run these programs, you must specify a set of source files as
5673 command line arguments. The conversion programs start out by compiling
5674 these files to see what functions they define. The information gathered
5675 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
5677 After scanning comes actual conversion. The specified files are all
5678 eligible to be converted; any files they include (whether sources or
5679 just headers) are eligible as well.
5681 But not all the eligible files are converted. By default,
5682 @code{protoize} and @code{unprotoize} convert only source and header
5683 files in the current directory. You can specify additional directories
5684 whose files should be converted with the @samp{-d @var{directory}}
5685 option. You can also specify particular files to exclude with the
5686 @samp{-x @var{file}} option. A file is converted if it is eligible, its
5687 directory name matches one of the specified directory names, and its
5688 name within the directory has not been excluded.
5690 Basic conversion with @code{protoize} consists of rewriting most
5691 function definitions and function declarations to specify the types of
5692 the arguments. The only ones not rewritten are those for varargs
5695 @code{protoize} optionally inserts prototype declarations at the
5696 beginning of the source file, to make them available for any calls that
5697 precede the function's definition. Or it can insert prototype
5698 declarations with block scope in the blocks where undeclared functions
5701 Basic conversion with @code{unprotoize} consists of rewriting most
5702 function declarations to remove any argument types, and rewriting
5703 function definitions to the old-style pre-ANSI form.
5705 Both conversion programs print a warning for any function declaration or
5706 definition that they can't convert. You can suppress these warnings
5709 The output from @code{protoize} or @code{unprotoize} replaces the
5710 original source file. The original file is renamed to a name ending
5711 with @samp{.save}. If the @samp{.save} file already exists, then
5712 the source file is simply discarded.
5714 @code{protoize} and @code{unprotoize} both depend on GNU CC itself to
5715 scan the program and collect information about the functions it uses.
5716 So neither of these programs will work until GNU CC is installed.
5718 Here is a table of the options you can use with @code{protoize} and
5719 @code{unprotoize}. Each option works with both programs unless
5723 @item -B @var{directory}
5724 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
5725 usual directory (normally @file{/usr/local/lib}). This file contains
5726 prototype information about standard system functions. This option
5727 applies only to @code{protoize}.
5729 @item -c @var{compilation-options}
5730 Use @var{compilation-options} as the options when running @code{gcc} to
5731 produce the @samp{.X} files. The special option @samp{-aux-info} is
5732 always passed in addition, to tell @code{gcc} to write a @samp{.X} file.
5734 Note that the compilation options must be given as a single argument to
5735 @code{protoize} or @code{unprotoize}. If you want to specify several
5736 @code{gcc} options, you must quote the entire set of compilation options
5737 to make them a single word in the shell.
5739 There are certain @code{gcc} arguments that you cannot use, because they
5740 would produce the wrong kind of output. These include @samp{-g},
5741 @samp{-O}, @samp{-c}, @samp{-S}, and @samp{-o} If you include these in
5742 the @var{compilation-options}, they are ignored.
5745 Rename files to end in @samp{.C} instead of @samp{.c}.
5746 This is convenient if you are converting a C program to C++.
5747 This option applies only to @code{protoize}.
5750 Add explicit global declarations. This means inserting explicit
5751 declarations at the beginning of each source file for each function
5752 that is called in the file and was not declared. These declarations
5753 precede the first function definition that contains a call to an
5754 undeclared function. This option applies only to @code{protoize}.
5756 @item -i @var{string}
5757 Indent old-style parameter declarations with the string @var{string}.
5758 This option applies only to @code{protoize}.
5760 @code{unprotoize} converts prototyped function definitions to old-style
5761 function definitions, where the arguments are declared between the
5762 argument list and the initial @samp{@{}. By default, @code{unprotoize}
5763 uses five spaces as the indentation. If you want to indent with just
5764 one space instead, use @samp{-i " "}.
5767 Keep the @samp{.X} files. Normally, they are deleted after conversion
5771 Add explicit local declarations. @code{protoize} with @samp{-l} inserts
5772 a prototype declaration for each function in each block which calls the
5773 function without any declaration. This option applies only to
5777 Make no real changes. This mode just prints information about the conversions
5778 that would have been done without @samp{-n}.
5781 Make no @samp{.save} files. The original files are simply deleted.
5782 Use this option with caution.
5784 @item -p @var{program}
5785 Use the program @var{program} as the compiler. Normally, the name
5789 Work quietly. Most warnings are suppressed.
5792 Print the version number, just like @samp{-v} for @code{gcc}.
5795 If you need special compiler options to compile one of your program's
5796 source files, then you should generate that file's @samp{.X} file
5797 specially, by running @code{gcc} on that source file with the
5798 appropriate options and the option @samp{-aux-info}. Then run
5799 @code{protoize} on the entire set of files. @code{protoize} will use
5800 the existing @samp{.X} file because it is newer than the source file.
5804 gcc -Dfoo=bar file1.c -aux-info
5809 You need to include the special files along with the rest in the
5810 @code{protoize} command, even though their @samp{.X} files already
5811 exist, because otherwise they won't get converted.
5813 @xref{Protoize Caveats}, for more information on how to use
5814 @code{protoize} successfully.