1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
8 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997,
9 1998, 1999, 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
11 Permission is granted to copy, distribute and/or modify this document
12 under the terms of the GNU Free Documentation License, Version 1.2 or
13 any later version published by the Free Software Foundation; with the
14 Invariant Sections being ``GNU General Public License'' and ``Funding
15 Free Software'', the Front-Cover texts being (a) (see below), and with
16 the Back-Cover Texts being (b) (see below). A copy of the license is
17 included in the gfdl(7) man page.
19 (a) The FSF's Front-Cover Text is:
23 (b) The FSF's Back-Cover Text is:
25 You have freedom to copy and modify this GNU Manual, like GNU
26 software. Copies published by the Free Software Foundation raise
27 funds for GNU development.
29 @c Set file name and title for the man page.
31 @settitle GNU project C and C++ compiler
33 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
34 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
35 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
36 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
37 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
38 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
39 [@option{-o} @var{outfile}] @var{infile}@dots{}
41 Only the most useful options are listed here; see below for the
42 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
45 gpl(7), gfdl(7), fsf-funding(7),
46 cpp(1), gcov(1), g77(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
47 and the Info entries for @file{gcc}, @file{cpp}, @file{g77}, @file{as},
48 @file{ld}, @file{binutils} and @file{gdb}.
51 For instructions on reporting bugs, see
52 @w{@uref{http://gcc.gnu.org/bugs.html}}. Use of the @command{gccbug}
53 script to report bugs is recommended.
56 See the Info entry for @command{gcc}, or
57 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
58 for contributors to GCC@.
63 @chapter GCC Command Options
64 @cindex GCC command options
65 @cindex command options
66 @cindex options, GCC command
68 @c man begin DESCRIPTION
69 When you invoke GCC, it normally does preprocessing, compilation,
70 assembly and linking. The ``overall options'' allow you to stop this
71 process at an intermediate stage. For example, the @option{-c} option
72 says not to run the linker. Then the output consists of object files
73 output by the assembler.
75 Other options are passed on to one stage of processing. Some options
76 control the preprocessor and others the compiler itself. Yet other
77 options control the assembler and linker; most of these are not
78 documented here, since you rarely need to use any of them.
80 @cindex C compilation options
81 Most of the command line options that you can use with GCC are useful
82 for C programs; when an option is only useful with another language
83 (usually C++), the explanation says so explicitly. If the description
84 for a particular option does not mention a source language, you can use
85 that option with all supported languages.
87 @cindex C++ compilation options
88 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
89 options for compiling C++ programs.
91 @cindex grouping options
92 @cindex options, grouping
93 The @command{gcc} program accepts options and file names as operands. Many
94 options have multi-letter names; therefore multiple single-letter options
95 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
98 @cindex order of options
99 @cindex options, order
100 You can mix options and other arguments. For the most part, the order
101 you use doesn't matter. Order does matter when you use several options
102 of the same kind; for example, if you specify @option{-L} more than once,
103 the directories are searched in the order specified.
105 Many options have long names starting with @samp{-f} or with
106 @samp{-W}---for example, @option{-fforce-mem},
107 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
108 these have both positive and negative forms; the negative form of
109 @option{-ffoo} would be @option{-fno-foo}. This manual documents
110 only one of these two forms, whichever one is not the default.
114 @xref{Option Index}, for an index to GCC's options.
117 * Option Summary:: Brief list of all options, without explanations.
118 * Overall Options:: Controlling the kind of output:
119 an executable, object files, assembler files,
120 or preprocessed source.
121 * Invoking G++:: Compiling C++ programs.
122 * C Dialect Options:: Controlling the variant of C language compiled.
123 * C++ Dialect Options:: Variations on C++.
124 * Objective-C Dialect Options:: Variations on Objective-C.
125 * Language Independent Options:: Controlling how diagnostics should be
127 * Warning Options:: How picky should the compiler be?
128 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
129 * Optimize Options:: How much optimization?
130 * Preprocessor Options:: Controlling header files and macro definitions.
131 Also, getting dependency information for Make.
132 * Assembler Options:: Passing options to the assembler.
133 * Link Options:: Specifying libraries and so on.
134 * Directory Options:: Where to find header files and libraries.
135 Where to find the compiler executable files.
136 * Spec Files:: How to pass switches to sub-processes.
137 * Target Options:: Running a cross-compiler, or an old version of GCC.
138 * Submodel Options:: Specifying minor hardware or convention variations,
139 such as 68010 vs 68020.
140 * Code Gen Options:: Specifying conventions for function calls, data layout
142 * Environment Variables:: Env vars that affect GCC.
143 * Precompiled Headers:: Compiling a header once, and using it many times.
144 * Running Protoize:: Automatically adding or removing function prototypes.
150 @section Option Summary
152 Here is a summary of all the options, grouped by type. Explanations are
153 in the following sections.
156 @item Overall Options
157 @xref{Overall Options,,Options Controlling the Kind of Output}.
158 @gccoptlist{-c -S -E -o @var{file} -pipe -pass-exit-codes @gol
159 -x @var{language} -v -### --help --target-help --version}
161 @item C Language Options
162 @xref{C Dialect Options,,Options Controlling C Dialect}.
163 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
164 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
165 -fhosted -ffreestanding -fms-extensions @gol
166 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
167 -fallow-single-precision -fcond-mismatch @gol
168 -fsigned-bitfields -fsigned-char @gol
169 -funsigned-bitfields -funsigned-char @gol
172 @item C++ Language Options
173 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
174 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
175 -fconserve-space -fno-const-strings @gol
176 -fno-elide-constructors @gol
177 -fno-enforce-eh-specs @gol
178 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
179 -fno-implicit-templates @gol
180 -fno-implicit-inline-templates @gol
181 -fno-implement-inlines -fms-extensions @gol
182 -fno-nonansi-builtins -fno-operator-names @gol
183 -fno-optional-diags -fpermissive @gol
184 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
185 -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
186 -fno-default-inline -Wabi -Wctor-dtor-privacy @gol
187 -Wnon-virtual-dtor -Wreorder @gol
188 -Weffc++ -Wno-deprecated @gol
189 -Wno-non-template-friend -Wold-style-cast @gol
190 -Woverloaded-virtual -Wno-pmf-conversions @gol
191 -Wsign-promo -Wsynth}
193 @item Objective-C Language Options
194 @xref{Objective-C Dialect Options,,Options Controlling Objective-C Dialect}.
196 -fconstant-string-class=@var{class-name} @gol
197 -fgnu-runtime -fnext-runtime @gol
198 -fno-nil-receivers @gol
199 -fobjc-exceptions @gol
200 -freplace-objc-classes @gol
203 -Wno-protocol -Wselector -Wundeclared-selector}
205 @item Language Independent Options
206 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
207 @gccoptlist{-fmessage-length=@var{n} @gol
208 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
210 @item Warning Options
211 @xref{Warning Options,,Options to Request or Suppress Warnings}.
212 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
213 -w -Wextra -Wall -Waggregate-return @gol
214 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
215 -Wconversion -Wno-deprecated-declarations @gol
216 -Wdisabled-optimization -Wno-div-by-zero -Wendif-labels @gol
217 -Werror -Werror-implicit-function-declaration @gol
218 -Wfloat-equal -Wformat -Wformat=2 @gol
219 -Wno-format-extra-args -Wformat-nonliteral @gol
220 -Wformat-security -Wformat-y2k @gol
221 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
222 -Wimport -Wno-import -Winit-self -Winline @gol
223 -Wno-invalid-offsetof -Winvalid-pch @gol
224 -Wlarger-than-@var{len} -Wlong-long @gol
225 -Wmain -Wmissing-braces @gol
226 -Wmissing-format-attribute -Wmissing-noreturn @gol
227 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
228 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
229 -Wreturn-type -Wsequence-point -Wshadow @gol
230 -Wsign-compare -Wstrict-aliasing @gol
231 -Wswitch -Wswitch-default -Wswitch-enum @gol
232 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
233 -Wunknown-pragmas -Wunreachable-code @gol
234 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
235 -Wunused-value -Wunused-variable -Wwrite-strings}
237 @item C-only Warning Options
238 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
239 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
240 -Wstrict-prototypes -Wtraditional @gol
241 -Wdeclaration-after-statement}
243 @item Debugging Options
244 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
245 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
246 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
247 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
248 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
249 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
250 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
251 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
252 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs @gol
253 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
254 -ftest-coverage -ftime-report @gol
255 -g -g@var{level} -gcoff -gdwarf-2 @gol
256 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
257 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
258 -print-multi-directory -print-multi-lib @gol
259 -print-prog-name=@var{program} -print-search-dirs -Q @gol
262 @item Optimization Options
263 @xref{Optimize Options,,Options that Control Optimization}.
264 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
265 -falign-labels=@var{n} -falign-loops=@var{n} @gol
266 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
267 -fbranch-target-load-optimize2 -fcaller-saves -fcprop-registers @gol
268 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections @gol
269 -fdelayed-branch -fdelete-null-pointer-checks @gol
270 -fexpensive-optimizations -ffast-math -ffloat-store @gol
271 -fforce-addr -fforce-mem -ffunction-sections @gol
272 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -floop-optimize @gol
273 -fcrossjumping -fif-conversion -fif-conversion2 @gol
274 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
275 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
276 -fmove-all-movables -fnew-ra -fno-branch-count-reg @gol
277 -fno-default-inline -fno-defer-pop @gol
278 -fno-function-cse -fno-guess-branch-probability @gol
279 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
280 -funsafe-math-optimizations -ffinite-math-only @gol
281 -fno-trapping-math -fno-zero-initialized-in-bss @gol
282 -fomit-frame-pointer -foptimize-register-move @gol
283 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
284 -fprofile-generate -fprofile-use @gol
285 -freduce-all-givs -fregmove -frename-registers @gol
286 -freorder-blocks -freorder-functions @gol
287 -frerun-cse-after-loop -frerun-loop-opt @gol
288 -frounding-math -fschedule-insns -fschedule-insns2 @gol
289 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
290 -fsched-spec-load-dangerous @gol
291 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
292 -fsched2-use-superblocks @gol
293 -fsched2-use-traces -fsignaling-nans @gol
294 -fsingle-precision-constant @gol
295 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
296 -funroll-all-loops -funroll-loops -fpeel-loops @gol
297 -funswitch-loops -fold-unroll-loops -fold-unroll-all-loops @gol
298 --param @var{name}=@var{value}
299 -O -O0 -O1 -O2 -O3 -Os}
301 @item Preprocessor Options
302 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
303 @gccoptlist{-A@var{question}=@var{answer} @gol
304 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
305 -C -dD -dI -dM -dN @gol
306 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
307 -idirafter @var{dir} @gol
308 -include @var{file} -imacros @var{file} @gol
309 -iprefix @var{file} -iwithprefix @var{dir} @gol
310 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
311 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
312 -P -fworking-directory -remap @gol
313 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
314 -Xpreprocessor @var{option}}
316 @item Assembler Option
317 @xref{Assembler Options,,Passing Options to the Assembler}.
318 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
321 @xref{Link Options,,Options for Linking}.
322 @gccoptlist{@var{object-file-name} -l@var{library} @gol
323 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
324 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
325 -Wl,@var{option} -Xlinker @var{option} @gol
328 @item Directory Options
329 @xref{Directory Options,,Options for Directory Search}.
330 @gccoptlist{-B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}}
333 @c I wrote this xref this way to avoid overfull hbox. -- rms
334 @xref{Target Options}.
335 @gccoptlist{-V @var{version} -b @var{machine}}
337 @item Machine Dependent Options
338 @xref{Submodel Options,,Hardware Models and Configurations}.
340 @emph{M680x0 Options}
341 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
342 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
343 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
344 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
345 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
347 @emph{M68hc1x Options}
348 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
349 -mauto-incdec -minmax -mlong-calls -mshort @gol
350 -msoft-reg-count=@var{count}}
353 @gccoptlist{-mg -mgnu -munix}
356 @gccoptlist{-mcpu=@var{cpu-type} @gol
357 -mtune=@var{cpu-type} @gol
358 -mcmodel=@var{code-model} @gol
360 -mapp-regs -mbroken-saverestore -mcypress @gol
361 -mfaster-structs -mflat @gol
362 -mfpu -mhard-float -mhard-quad-float @gol
363 -mimpure-text -mlittle-endian -mlive-g0 -mno-app-regs @gol
364 -mno-faster-structs -mno-flat -mno-fpu @gol
365 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles @gol
366 -msoft-float -msoft-quad-float -msparclite -mstack-bias @gol
367 -msupersparc -munaligned-doubles -mv8}
370 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
371 -mapcs-26 -mapcs-32 @gol
372 -mapcs-stack-check -mno-apcs-stack-check @gol
373 -mapcs-float -mno-apcs-float @gol
374 -mapcs-reentrant -mno-apcs-reentrant @gol
375 -msched-prolog -mno-sched-prolog @gol
376 -mlittle-endian -mbig-endian -mwords-little-endian @gol
377 -malignment-traps -mno-alignment-traps @gol
378 -msoft-float -mhard-float -mfpe @gol
379 -mthumb-interwork -mno-thumb-interwork @gol
380 -mcpu=@var{name} -march=@var{name} -mfpe=@var{name} @gol
381 -mstructure-size-boundary=@var{n} @gol
382 -mabort-on-noreturn @gol
383 -mlong-calls -mno-long-calls @gol
384 -msingle-pic-base -mno-single-pic-base @gol
385 -mpic-register=@var{reg} @gol
386 -mnop-fun-dllimport @gol
387 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
388 -mpoke-function-name @gol
390 -mtpcs-frame -mtpcs-leaf-frame @gol
391 -mcaller-super-interworking -mcallee-super-interworking}
393 @emph{MN10300 Options}
394 @gccoptlist{-mmult-bug -mno-mult-bug @gol
395 -mam33 -mno-am33 @gol
396 -mam33-2 -mno-am33-2 @gol
399 @emph{M32R/D Options}
400 @gccoptlist{-m32r2 -m32rx -m32r @gol
402 -malign-loops -mno-align-loops @gol
403 -missue-rate=@var{number} @gol
404 -mbranch-cost=@var{number} @gol
405 -mmodel=@var{code-size-model-type} @gol
406 -msdata=@var{sdata-type} @gol
407 -mno-flush-func -mflush-func=@var{name} @gol
408 -mno-flush-trap -mflush-trap=@var{number} @gol
411 @emph{RS/6000 and PowerPC Options}
412 @gccoptlist{-mcpu=@var{cpu-type} @gol
413 -mtune=@var{cpu-type} @gol
414 -mpower -mno-power -mpower2 -mno-power2 @gol
415 -mpowerpc -mpowerpc64 -mno-powerpc @gol
416 -maltivec -mno-altivec @gol
417 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
418 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
419 -mnew-mnemonics -mold-mnemonics @gol
420 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
421 -m64 -m32 -mxl-call -mno-xl-call -mpe @gol
422 -malign-power -malign-natural @gol
423 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
424 -mstring -mno-string -mupdate -mno-update @gol
425 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
426 -mstrict-align -mno-strict-align -mrelocatable @gol
427 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
428 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
429 -mdynamic-no-pic @gol
430 -mprioritize-restricted-insns=@var{priority} @gol
431 -msched-costly-dep=@var{dependence_type} @gol
432 -minsert-sched-nops=@var{scheme} @gol
433 -mcall-sysv -mcall-netbsd @gol
434 -maix-struct-return -msvr4-struct-return @gol
435 -mabi=altivec -mabi=no-altivec @gol
436 -mabi=spe -mabi=no-spe @gol
437 -misel=yes -misel=no @gol
438 -mspe=yes -mspe=no @gol
439 -mfloat-gprs=yes -mfloat-gprs=no @gol
440 -mprototype -mno-prototype @gol
441 -msim -mmvme -mads -myellowknife -memb -msdata @gol
442 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
444 @emph{Darwin Options}
445 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
446 -arch_only -bind_at_load -bundle -bundle_loader @gol
447 -client_name -compatibility_version -current_version @gol
448 -dependency-file -dylib_file -dylinker_install_name @gol
449 -dynamic -dynamiclib -exported_symbols_list @gol
450 -filelist -flat_namespace -force_cpusubtype_ALL @gol
451 -force_flat_namespace -headerpad_max_install_names @gol
452 -image_base -init -install_name -keep_private_externs @gol
453 -multi_module -multiply_defined -multiply_defined_unused @gol
454 -noall_load -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
455 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
456 -private_bundle -read_only_relocs -sectalign @gol
457 -sectobjectsymbols -whyload -seg1addr @gol
458 -sectcreate -sectobjectsymbols -sectorder @gol
459 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
460 -segprot -segs_read_only_addr -segs_read_write_addr @gol
461 -single_module -static -sub_library -sub_umbrella @gol
462 -twolevel_namespace -umbrella -undefined @gol
463 -unexported_symbols_list -weak_reference_mismatches @gol
467 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
468 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
469 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
470 -mxgot -mno-xgot -membedded-pic -mno-embedded-pic @gol
471 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
472 -msingle-float -mdouble-float -mint64 -mlong64 -mlong32 @gol
473 -G@var{num} -membedded-data -mno-embedded-data @gol
474 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
475 -msplit-addresses -mno-split-addresses @gol
476 -mexplicit-relocs -mno-explicit-relocs @gol
477 -mrnames -mno-rnames @gol
478 -mcheck-zero-division -mno-check-zero-division @gol
479 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
480 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
481 -mfix-sb1 -mno-fix-sb1 -mflush-func=@var{func} @gol
482 -mno-flush-func -mbranch-likely -mno-branch-likely}
484 @emph{i386 and x86-64 Options}
485 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
486 -mfpmath=@var{unit} @gol
487 -masm=@var{dialect} -mno-fancy-math-387 @gol
488 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
489 -mno-wide-multiply -mrtd -malign-double @gol
490 -mpreferred-stack-boundary=@var{num} @gol
491 -mmmx -msse -msse2 -mpni -m3dnow @gol
492 -mthreads -mno-align-stringops -minline-all-stringops @gol
493 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
494 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
495 -mno-red-zone -mno-tls-direct-seg-refs @gol
496 -mcmodel=@var{code-model} @gol
500 @gccoptlist{-march=@var{architecture-type} @gol
501 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
502 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
503 -mjump-in-delay -mlinker-opt -mlong-calls @gol
504 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
505 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
506 -mno-jump-in-delay -mno-long-load-store @gol
507 -mno-portable-runtime -mno-soft-float @gol
508 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
509 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
510 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
511 -nolibdld -static -threads}
513 @emph{Intel 960 Options}
514 @gccoptlist{-m@var{cpu-type} -masm-compat -mclean-linkage @gol
515 -mcode-align -mcomplex-addr -mleaf-procedures @gol
516 -mic-compat -mic2.0-compat -mic3.0-compat @gol
517 -mintel-asm -mno-clean-linkage -mno-code-align @gol
518 -mno-complex-addr -mno-leaf-procedures @gol
519 -mno-old-align -mno-strict-align -mno-tail-call @gol
520 -mnumerics -mold-align -msoft-float -mstrict-align @gol
523 @emph{DEC Alpha Options}
524 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
525 -mieee -mieee-with-inexact -mieee-conformant @gol
526 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
527 -mtrap-precision=@var{mode} -mbuild-constants @gol
528 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
529 -mbwx -mmax -mfix -mcix @gol
530 -mfloat-vax -mfloat-ieee @gol
531 -mexplicit-relocs -msmall-data -mlarge-data @gol
532 -msmall-text -mlarge-text @gol
533 -mmemory-latency=@var{time}}
535 @emph{DEC Alpha/VMS Options}
536 @gccoptlist{-mvms-return-codes}
538 @emph{H8/300 Options}
539 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
542 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
543 -m4-nofpu -m4-single-only -m4-single -m4 @gol
544 -m5-64media -m5-64media-nofpu @gol
545 -m5-32media -m5-32media-nofpu @gol
546 -m5-compact -m5-compact-nofpu @gol
547 -mb -ml -mdalign -mrelax @gol
548 -mbigtable -mfmovd -mhitachi -mnomacsave @gol
549 -mieee -misize -mpadstruct -mspace @gol
550 -mprefergot -musermode}
552 @emph{System V Options}
553 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
556 @gccoptlist{-EB -EL @gol
557 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
558 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
560 @emph{TMS320C3x/C4x Options}
561 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
562 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
563 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
564 -mparallel-insns -mparallel-mpy -mpreserve-float}
567 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
568 -mprolog-function -mno-prolog-function -mspace @gol
569 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
570 -mapp-regs -mno-app-regs @gol
571 -mdisable-callt -mno-disable-callt @gol
577 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
578 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
579 -mregparam -mnoregparam -msb -mnosb @gol
580 -mbitfield -mnobitfield -mhimem -mnohimem}
583 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
584 -mcall-prologues -mno-tablejump -mtiny-stack}
587 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
588 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
589 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
590 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
591 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
594 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
595 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
596 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
597 -mno-base-addresses -msingle-exit -mno-single-exit}
600 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
601 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
602 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
603 -minline-float-divide-max-throughput @gol
604 -minline-int-divide-min-latency @gol
605 -minline-int-divide-max-throughput -mno-dwarf2-asm @gol
606 -mfixed-range=@var{register-range}}
609 @gccoptlist{-mextmem -mextmemory -monchip -mno-asm-optimize @gol
610 -masm-optimize -mbranch-cost=@var{n} -mcond-exec=@var{n}}
612 @emph{S/390 and zSeries Options}
613 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
614 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
615 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
616 -m64 -m31 -mdebug -mno-debug -mesa -mzarch -mfused-madd -mno-fused-madd}
619 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
620 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
621 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
622 -mstack-align -mdata-align -mconst-align @gol
623 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
624 -melf -maout -melinux -mlinux -sim -sim2}
626 @emph{PDP-11 Options}
627 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
628 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
629 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
630 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
631 -mbranch-expensive -mbranch-cheap @gol
632 -msplit -mno-split -munix-asm -mdec-asm}
634 @emph{Xstormy16 Options}
637 @emph{Xtensa Options}
638 @gccoptlist{-mconst16 -mno-const16 @gol
639 -mfused-madd -mno-fused-madd @gol
640 -mtext-section-literals -mno-text-section-literals @gol
641 -mtarget-align -mno-target-align @gol
642 -mlongcalls -mno-longcalls}
645 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
646 -mhard-float -msoft-float @gol
647 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
648 -mdouble -mno-double @gol
649 -mmedia -mno-media -mmuladd -mno-muladd @gol
650 -mlibrary-pic -macc-4 -macc-8 @gol
651 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
652 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
653 -mvliw-branch -mno-vliw-branch @gol
654 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
655 -mno-nested-cond-exec -mtomcat-stats @gol
658 @item Code Generation Options
659 @xref{Code Gen Options,,Options for Code Generation Conventions}.
660 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
661 -ffixed-@var{reg} -fexceptions @gol
662 -fnon-call-exceptions -funwind-tables @gol
663 -fasynchronous-unwind-tables @gol
664 -finhibit-size-directive -finstrument-functions @gol
665 -fno-common -fno-ident @gol
666 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
667 -freg-struct-return -fshared-data -fshort-enums @gol
668 -fshort-double -fshort-wchar @gol
669 -fverbose-asm -fpack-struct -fstack-check @gol
670 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
671 -fargument-alias -fargument-noalias @gol
672 -fargument-noalias-global -fleading-underscore @gol
673 -ftls-model=@var{model} @gol
674 -ftrapv -fwrapv -fbounds-check}
678 * Overall Options:: Controlling the kind of output:
679 an executable, object files, assembler files,
680 or preprocessed source.
681 * C Dialect Options:: Controlling the variant of C language compiled.
682 * C++ Dialect Options:: Variations on C++.
683 * Objective-C Dialect Options:: Variations on Objective-C.
684 * Language Independent Options:: Controlling how diagnostics should be
686 * Warning Options:: How picky should the compiler be?
687 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
688 * Optimize Options:: How much optimization?
689 * Preprocessor Options:: Controlling header files and macro definitions.
690 Also, getting dependency information for Make.
691 * Assembler Options:: Passing options to the assembler.
692 * Link Options:: Specifying libraries and so on.
693 * Directory Options:: Where to find header files and libraries.
694 Where to find the compiler executable files.
695 * Spec Files:: How to pass switches to sub-processes.
696 * Target Options:: Running a cross-compiler, or an old version of GCC.
699 @node Overall Options
700 @section Options Controlling the Kind of Output
702 Compilation can involve up to four stages: preprocessing, compilation
703 proper, assembly and linking, always in that order. GCC is capable of
704 preprocessing and compiling several files either into several
705 assembler input files, or into one assembler input file; then each
706 assembler input file produces an object file, and linking combines all
707 the object files (those newly compiled, and those specified as input)
708 into an executable file.
710 @cindex file name suffix
711 For any given input file, the file name suffix determines what kind of
716 C source code which must be preprocessed.
719 C source code which should not be preprocessed.
722 C++ source code which should not be preprocessed.
725 Objective-C source code. Note that you must link with the library
726 @file{libobjc.a} to make an Objective-C program work.
729 Objective-C source code which should not be preprocessed.
732 C or C++ header file to be turned into a precompiled header.
736 @itemx @var{file}.cxx
737 @itemx @var{file}.cpp
738 @itemx @var{file}.CPP
739 @itemx @var{file}.c++
741 C++ source code which must be preprocessed. Note that in @samp{.cxx},
742 the last two letters must both be literally @samp{x}. Likewise,
743 @samp{.C} refers to a literal capital C@.
747 C++ header file to be turned into a precompiled header.
750 @itemx @var{file}.for
751 @itemx @var{file}.FOR
752 Fortran source code which should not be preprocessed.
755 @itemx @var{file}.fpp
756 @itemx @var{file}.FPP
757 Fortran source code which must be preprocessed (with the traditional
761 Fortran source code which must be preprocessed with a RATFOR
762 preprocessor (not included with GCC)@.
764 @xref{Overall Options,,Options Controlling the Kind of Output, g77,
765 Using and Porting GNU Fortran}, for more details of the handling of
768 @c FIXME: Descriptions of Java file types.
775 Ada source code file which contains a library unit declaration (a
776 declaration of a package, subprogram, or generic, or a generic
777 instantiation), or a library unit renaming declaration (a package,
778 generic, or subprogram renaming declaration). Such files are also
781 @itemx @var{file}.adb
782 Ada source code file containing a library unit body (a subprogram or
783 package body). Such files are also called @dfn{bodies}.
785 @c GCC also knows about some suffixes for languages not yet included:
794 Assembler code which must be preprocessed.
797 An object file to be fed straight into linking.
798 Any file name with no recognized suffix is treated this way.
802 You can specify the input language explicitly with the @option{-x} option:
805 @item -x @var{language}
806 Specify explicitly the @var{language} for the following input files
807 (rather than letting the compiler choose a default based on the file
808 name suffix). This option applies to all following input files until
809 the next @option{-x} option. Possible values for @var{language} are:
811 c c-header cpp-output
812 c++ c++-header c++-cpp-output
813 objective-c objective-c-header objc-cpp-output
814 assembler assembler-with-cpp
816 f77 f77-cpp-input ratfor
822 Turn off any specification of a language, so that subsequent files are
823 handled according to their file name suffixes (as they are if @option{-x}
824 has not been used at all).
826 @item -pass-exit-codes
827 @opindex pass-exit-codes
828 Normally the @command{gcc} program will exit with the code of 1 if any
829 phase of the compiler returns a non-success return code. If you specify
830 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
831 numerically highest error produced by any phase that returned an error
835 If you only want some of the stages of compilation, you can use
836 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
837 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
838 @command{gcc} is to stop. Note that some combinations (for example,
839 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
844 Compile or assemble the source files, but do not link. The linking
845 stage simply is not done. The ultimate output is in the form of an
846 object file for each source file.
848 By default, the object file name for a source file is made by replacing
849 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
851 Unrecognized input files, not requiring compilation or assembly, are
856 Stop after the stage of compilation proper; do not assemble. The output
857 is in the form of an assembler code file for each non-assembler input
860 By default, the assembler file name for a source file is made by
861 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
863 Input files that don't require compilation are ignored.
867 Stop after the preprocessing stage; do not run the compiler proper. The
868 output is in the form of preprocessed source code, which is sent to the
871 Input files which don't require preprocessing are ignored.
873 @cindex output file option
876 Place output in file @var{file}. This applies regardless to whatever
877 sort of output is being produced, whether it be an executable file,
878 an object file, an assembler file or preprocessed C code.
880 If you specify @option{-o} when compiling more than one input file, or
881 you are producing an executable file as output, all the source files
882 on the command line will be compiled at once.
884 If @option{-o} is not specified, the default is to put an executable file
885 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
886 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
887 all preprocessed C source on standard output.
891 Print (on standard error output) the commands executed to run the stages
892 of compilation. Also print the version number of the compiler driver
893 program and of the preprocessor and the compiler proper.
897 Like @option{-v} except the commands are not executed and all command
898 arguments are quoted. This is useful for shell scripts to capture the
899 driver-generated command lines.
903 Use pipes rather than temporary files for communication between the
904 various stages of compilation. This fails to work on some systems where
905 the assembler is unable to read from a pipe; but the GNU assembler has
910 Print (on the standard output) a description of the command line options
911 understood by @command{gcc}. If the @option{-v} option is also specified
912 then @option{--help} will also be passed on to the various processes
913 invoked by @command{gcc}, so that they can display the command line options
914 they accept. If the @option{-Wextra} option is also specified then command
915 line options which have no documentation associated with them will also
920 Print (on the standard output) a description of target specific command
921 line options for each tool.
925 Display the version number and copyrights of the invoked GCC.
929 @section Compiling C++ Programs
931 @cindex suffixes for C++ source
932 @cindex C++ source file suffixes
933 C++ source files conventionally use one of the suffixes @samp{.C},
934 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
935 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
936 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
937 files with these names and compiles them as C++ programs even if you
938 call the compiler the same way as for compiling C programs (usually
939 with the name @command{gcc}).
943 However, C++ programs often require class libraries as well as a
944 compiler that understands the C++ language---and under some
945 circumstances, you might want to compile programs or header files from
946 standard input, or otherwise without a suffix that flags them as C++
947 programs. You might also like to precompile a C header file with a
948 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
949 program that calls GCC with the default language set to C++, and
950 automatically specifies linking against the C++ library. On many
951 systems, @command{g++} is also installed with the name @command{c++}.
953 @cindex invoking @command{g++}
954 When you compile C++ programs, you may specify many of the same
955 command-line options that you use for compiling programs in any
956 language; or command-line options meaningful for C and related
957 languages; or options that are meaningful only for C++ programs.
958 @xref{C Dialect Options,,Options Controlling C Dialect}, for
959 explanations of options for languages related to C@.
960 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
961 explanations of options that are meaningful only for C++ programs.
963 @node C Dialect Options
964 @section Options Controlling C Dialect
965 @cindex dialect options
966 @cindex language dialect options
967 @cindex options, dialect
969 The following options control the dialect of C (or languages derived
970 from C, such as C++ and Objective-C) that the compiler accepts:
977 In C mode, support all ISO C90 programs. In C++ mode,
978 remove GNU extensions that conflict with ISO C++.
980 This turns off certain features of GCC that are incompatible with ISO
981 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
982 such as the @code{asm} and @code{typeof} keywords, and
983 predefined macros such as @code{unix} and @code{vax} that identify the
984 type of system you are using. It also enables the undesirable and
985 rarely used ISO trigraph feature. For the C compiler,
986 it disables recognition of C++ style @samp{//} comments as well as
987 the @code{inline} keyword.
989 The alternate keywords @code{__asm__}, @code{__extension__},
990 @code{__inline__} and @code{__typeof__} continue to work despite
991 @option{-ansi}. You would not want to use them in an ISO C program, of
992 course, but it is useful to put them in header files that might be included
993 in compilations done with @option{-ansi}. Alternate predefined macros
994 such as @code{__unix__} and @code{__vax__} are also available, with or
995 without @option{-ansi}.
997 The @option{-ansi} option does not cause non-ISO programs to be
998 rejected gratuitously. For that, @option{-pedantic} is required in
999 addition to @option{-ansi}. @xref{Warning Options}.
1001 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1002 option is used. Some header files may notice this macro and refrain
1003 from declaring certain functions or defining certain macros that the
1004 ISO standard doesn't call for; this is to avoid interfering with any
1005 programs that might use these names for other things.
1007 Functions which would normally be built in but do not have semantics
1008 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1009 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1010 built-in functions provided by GCC}, for details of the functions
1015 Determine the language standard. This option is currently only
1016 supported when compiling C or C++. A value for this option must be
1017 provided; possible values are
1022 ISO C90 (same as @option{-ansi}).
1024 @item iso9899:199409
1025 ISO C90 as modified in amendment 1.
1031 ISO C99. Note that this standard is not yet fully supported; see
1032 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1033 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1036 Default, ISO C90 plus GNU extensions (including some C99 features).
1040 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1041 this will become the default. The name @samp{gnu9x} is deprecated.
1044 The 1998 ISO C++ standard plus amendments.
1047 The same as @option{-std=c++98} plus GNU extensions. This is the
1048 default for C++ code.
1051 Even when this option is not specified, you can still use some of the
1052 features of newer standards in so far as they do not conflict with
1053 previous C standards. For example, you may use @code{__restrict__} even
1054 when @option{-std=c99} is not specified.
1056 The @option{-std} options specifying some version of ISO C have the same
1057 effects as @option{-ansi}, except that features that were not in ISO C90
1058 but are in the specified version (for example, @samp{//} comments and
1059 the @code{inline} keyword in ISO C99) are not disabled.
1061 @xref{Standards,,Language Standards Supported by GCC}, for details of
1062 these standard versions.
1064 @item -aux-info @var{filename}
1066 Output to the given filename prototyped declarations for all functions
1067 declared and/or defined in a translation unit, including those in header
1068 files. This option is silently ignored in any language other than C@.
1070 Besides declarations, the file indicates, in comments, the origin of
1071 each declaration (source file and line), whether the declaration was
1072 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1073 @samp{O} for old, respectively, in the first character after the line
1074 number and the colon), and whether it came from a declaration or a
1075 definition (@samp{C} or @samp{F}, respectively, in the following
1076 character). In the case of function definitions, a K&R-style list of
1077 arguments followed by their declarations is also provided, inside
1078 comments, after the declaration.
1082 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1083 keyword, so that code can use these words as identifiers. You can use
1084 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1085 instead. @option{-ansi} implies @option{-fno-asm}.
1087 In C++, this switch only affects the @code{typeof} keyword, since
1088 @code{asm} and @code{inline} are standard keywords. You may want to
1089 use the @option{-fno-gnu-keywords} flag instead, which has the same
1090 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1091 switch only affects the @code{asm} and @code{typeof} keywords, since
1092 @code{inline} is a standard keyword in ISO C99.
1095 @itemx -fno-builtin-@var{function}
1096 @opindex fno-builtin
1097 @cindex built-in functions
1098 Don't recognize built-in functions that do not begin with
1099 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1100 functions provided by GCC}, for details of the functions affected,
1101 including those which are not built-in functions when @option{-ansi} or
1102 @option{-std} options for strict ISO C conformance are used because they
1103 do not have an ISO standard meaning.
1105 GCC normally generates special code to handle certain built-in functions
1106 more efficiently; for instance, calls to @code{alloca} may become single
1107 instructions that adjust the stack directly, and calls to @code{memcpy}
1108 may become inline copy loops. The resulting code is often both smaller
1109 and faster, but since the function calls no longer appear as such, you
1110 cannot set a breakpoint on those calls, nor can you change the behavior
1111 of the functions by linking with a different library.
1113 With the @option{-fno-builtin-@var{function}} option
1114 only the built-in function @var{function} is
1115 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1116 function is named this is not built-in in this version of GCC, this
1117 option is ignored. There is no corresponding
1118 @option{-fbuiltin-@var{function}} option; if you wish to enable
1119 built-in functions selectively when using @option{-fno-builtin} or
1120 @option{-ffreestanding}, you may define macros such as:
1123 #define abs(n) __builtin_abs ((n))
1124 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1129 @cindex hosted environment
1131 Assert that compilation takes place in a hosted environment. This implies
1132 @option{-fbuiltin}. A hosted environment is one in which the
1133 entire standard library is available, and in which @code{main} has a return
1134 type of @code{int}. Examples are nearly everything except a kernel.
1135 This is equivalent to @option{-fno-freestanding}.
1137 @item -ffreestanding
1138 @opindex ffreestanding
1139 @cindex hosted environment
1141 Assert that compilation takes place in a freestanding environment. This
1142 implies @option{-fno-builtin}. A freestanding environment
1143 is one in which the standard library may not exist, and program startup may
1144 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1145 This is equivalent to @option{-fno-hosted}.
1147 @xref{Standards,,Language Standards Supported by GCC}, for details of
1148 freestanding and hosted environments.
1150 @item -fms-extensions
1151 @opindex fms-extensions
1152 Accept some non-standard constructs used in Microsoft header files.
1156 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1157 options for strict ISO C conformance) implies @option{-trigraphs}.
1159 @item -no-integrated-cpp
1160 @opindex no-integrated-cpp
1161 Performs a compilation in two passes: preprocessing and compiling. This
1162 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1163 @option{-B} option. The user supplied compilation step can then add in
1164 an additional preprocessing step after normal preprocessing but before
1165 compiling. The default is to use the integrated cpp (internal cpp)
1167 The semantics of this option will change if "cc1", "cc1plus", and
1168 "cc1obj" are merged.
1170 @cindex traditional C language
1171 @cindex C language, traditional
1173 @itemx -traditional-cpp
1174 @opindex traditional-cpp
1175 @opindex traditional
1176 Formerly, these options caused GCC to attempt to emulate a pre-standard
1177 C compiler. They are now only supported with the @option{-E} switch.
1178 The preprocessor continues to support a pre-standard mode. See the GNU
1179 CPP manual for details.
1181 @item -fcond-mismatch
1182 @opindex fcond-mismatch
1183 Allow conditional expressions with mismatched types in the second and
1184 third arguments. The value of such an expression is void. This option
1185 is not supported for C++.
1187 @item -funsigned-char
1188 @opindex funsigned-char
1189 Let the type @code{char} be unsigned, like @code{unsigned char}.
1191 Each kind of machine has a default for what @code{char} should
1192 be. It is either like @code{unsigned char} by default or like
1193 @code{signed char} by default.
1195 Ideally, a portable program should always use @code{signed char} or
1196 @code{unsigned char} when it depends on the signedness of an object.
1197 But many programs have been written to use plain @code{char} and
1198 expect it to be signed, or expect it to be unsigned, depending on the
1199 machines they were written for. This option, and its inverse, let you
1200 make such a program work with the opposite default.
1202 The type @code{char} is always a distinct type from each of
1203 @code{signed char} or @code{unsigned char}, even though its behavior
1204 is always just like one of those two.
1207 @opindex fsigned-char
1208 Let the type @code{char} be signed, like @code{signed char}.
1210 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1211 the negative form of @option{-funsigned-char}. Likewise, the option
1212 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1214 @item -fsigned-bitfields
1215 @itemx -funsigned-bitfields
1216 @itemx -fno-signed-bitfields
1217 @itemx -fno-unsigned-bitfields
1218 @opindex fsigned-bitfields
1219 @opindex funsigned-bitfields
1220 @opindex fno-signed-bitfields
1221 @opindex fno-unsigned-bitfields
1222 These options control whether a bit-field is signed or unsigned, when the
1223 declaration does not use either @code{signed} or @code{unsigned}. By
1224 default, such a bit-field is signed, because this is consistent: the
1225 basic integer types such as @code{int} are signed types.
1227 @item -fwritable-strings
1228 @opindex fwritable-strings
1229 Store string constants in the writable data segment and don't uniquize
1230 them. This is for compatibility with old programs which assume they can
1231 write into string constants.
1233 Writing into string constants is a very bad idea; ``constants'' should
1236 This option is deprecated.
1239 @node C++ Dialect Options
1240 @section Options Controlling C++ Dialect
1242 @cindex compiler options, C++
1243 @cindex C++ options, command line
1244 @cindex options, C++
1245 This section describes the command-line options that are only meaningful
1246 for C++ programs; but you can also use most of the GNU compiler options
1247 regardless of what language your program is in. For example, you
1248 might compile a file @code{firstClass.C} like this:
1251 g++ -g -frepo -O -c firstClass.C
1255 In this example, only @option{-frepo} is an option meant
1256 only for C++ programs; you can use the other options with any
1257 language supported by GCC@.
1259 Here is a list of options that are @emph{only} for compiling C++ programs:
1263 @item -fabi-version=@var{n}
1264 @opindex fabi-version
1265 Use version @var{n} of the C++ ABI. Version 2 is the version of the
1266 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1267 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1268 the version that conforms most closely to the C++ ABI specification.
1269 Therefore, the ABI obtained using version 0 will change as ABI bugs
1272 The default is version 1.
1274 @item -fno-access-control
1275 @opindex fno-access-control
1276 Turn off all access checking. This switch is mainly useful for working
1277 around bugs in the access control code.
1281 Check that the pointer returned by @code{operator new} is non-null
1282 before attempting to modify the storage allocated. This check is
1283 normally unnecessary because the C++ standard specifies that
1284 @code{operator new} will only return @code{0} if it is declared
1285 @samp{throw()}, in which case the compiler will always check the
1286 return value even without this option. In all other cases, when
1287 @code{operator new} has a non-empty exception specification, memory
1288 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1289 @samp{new (nothrow)}.
1291 @item -fconserve-space
1292 @opindex fconserve-space
1293 Put uninitialized or runtime-initialized global variables into the
1294 common segment, as C does. This saves space in the executable at the
1295 cost of not diagnosing duplicate definitions. If you compile with this
1296 flag and your program mysteriously crashes after @code{main()} has
1297 completed, you may have an object that is being destroyed twice because
1298 two definitions were merged.
1300 This option is no longer useful on most targets, now that support has
1301 been added for putting variables into BSS without making them common.
1303 @item -fno-const-strings
1304 @opindex fno-const-strings
1305 Give string constants type @code{char *} instead of type @code{const
1306 char *}. By default, G++ uses type @code{const char *} as required by
1307 the standard. Even if you use @option{-fno-const-strings}, you cannot
1308 actually modify the value of a string constant, unless you also use
1309 @option{-fwritable-strings}.
1311 This option might be removed in a future release of G++. For maximum
1312 portability, you should structure your code so that it works with
1313 string constants that have type @code{const char *}.
1315 @item -fno-elide-constructors
1316 @opindex fno-elide-constructors
1317 The C++ standard allows an implementation to omit creating a temporary
1318 which is only used to initialize another object of the same type.
1319 Specifying this option disables that optimization, and forces G++ to
1320 call the copy constructor in all cases.
1322 @item -fno-enforce-eh-specs
1323 @opindex fno-enforce-eh-specs
1324 Don't check for violation of exception specifications at runtime. This
1325 option violates the C++ standard, but may be useful for reducing code
1326 size in production builds, much like defining @samp{NDEBUG}. The compiler
1327 will still optimize based on the exception specifications.
1330 @itemx -fno-for-scope
1332 @opindex fno-for-scope
1333 If @option{-ffor-scope} is specified, the scope of variables declared in
1334 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1335 as specified by the C++ standard.
1336 If @option{-fno-for-scope} is specified, the scope of variables declared in
1337 a @i{for-init-statement} extends to the end of the enclosing scope,
1338 as was the case in old versions of G++, and other (traditional)
1339 implementations of C++.
1341 The default if neither flag is given to follow the standard,
1342 but to allow and give a warning for old-style code that would
1343 otherwise be invalid, or have different behavior.
1345 @item -fno-gnu-keywords
1346 @opindex fno-gnu-keywords
1347 Do not recognize @code{typeof} as a keyword, so that code can use this
1348 word as an identifier. You can use the keyword @code{__typeof__} instead.
1349 @option{-ansi} implies @option{-fno-gnu-keywords}.
1351 @item -fno-implicit-templates
1352 @opindex fno-implicit-templates
1353 Never emit code for non-inline templates which are instantiated
1354 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1355 @xref{Template Instantiation}, for more information.
1357 @item -fno-implicit-inline-templates
1358 @opindex fno-implicit-inline-templates
1359 Don't emit code for implicit instantiations of inline templates, either.
1360 The default is to handle inlines differently so that compiles with and
1361 without optimization will need the same set of explicit instantiations.
1363 @item -fno-implement-inlines
1364 @opindex fno-implement-inlines
1365 To save space, do not emit out-of-line copies of inline functions
1366 controlled by @samp{#pragma implementation}. This will cause linker
1367 errors if these functions are not inlined everywhere they are called.
1369 @item -fms-extensions
1370 @opindex fms-extensions
1371 Disable pedantic warnings about constructs used in MFC, such as implicit
1372 int and getting a pointer to member function via non-standard syntax.
1374 @item -fno-nonansi-builtins
1375 @opindex fno-nonansi-builtins
1376 Disable built-in declarations of functions that are not mandated by
1377 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1378 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1380 @item -fno-operator-names
1381 @opindex fno-operator-names
1382 Do not treat the operator name keywords @code{and}, @code{bitand},
1383 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1384 synonyms as keywords.
1386 @item -fno-optional-diags
1387 @opindex fno-optional-diags
1388 Disable diagnostics that the standard says a compiler does not need to
1389 issue. Currently, the only such diagnostic issued by G++ is the one for
1390 a name having multiple meanings within a class.
1393 @opindex fpermissive
1394 Downgrade some diagnostics about nonconformant code from errors to
1395 warnings. Thus, using @option{-fpermissive} will allow some
1396 nonconforming code to compile.
1400 Enable automatic template instantiation at link time. This option also
1401 implies @option{-fno-implicit-templates}. @xref{Template
1402 Instantiation}, for more information.
1406 Disable generation of information about every class with virtual
1407 functions for use by the C++ runtime type identification features
1408 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1409 of the language, you can save some space by using this flag. Note that
1410 exception handling uses the same information, but it will generate it as
1415 Emit statistics about front-end processing at the end of the compilation.
1416 This information is generally only useful to the G++ development team.
1418 @item -ftemplate-depth-@var{n}
1419 @opindex ftemplate-depth
1420 Set the maximum instantiation depth for template classes to @var{n}.
1421 A limit on the template instantiation depth is needed to detect
1422 endless recursions during template class instantiation. ANSI/ISO C++
1423 conforming programs must not rely on a maximum depth greater than 17.
1425 @item -fuse-cxa-atexit
1426 @opindex fuse-cxa-atexit
1427 Register destructors for objects with static storage duration with the
1428 @code{__cxa_atexit} function rather than the @code{atexit} function.
1429 This option is required for fully standards-compliant handling of static
1430 destructors, but will only work if your C library supports
1431 @code{__cxa_atexit}.
1435 Do not use weak symbol support, even if it is provided by the linker.
1436 By default, G++ will use weak symbols if they are available. This
1437 option exists only for testing, and should not be used by end-users;
1438 it will result in inferior code and has no benefits. This option may
1439 be removed in a future release of G++.
1443 Do not search for header files in the standard directories specific to
1444 C++, but do still search the other standard directories. (This option
1445 is used when building the C++ library.)
1448 In addition, these optimization, warning, and code generation options
1449 have meanings only for C++ programs:
1452 @item -fno-default-inline
1453 @opindex fno-default-inline
1454 Do not assume @samp{inline} for functions defined inside a class scope.
1455 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1456 functions will have linkage like inline functions; they just won't be
1459 @item -Wabi @r{(C++ only)}
1461 Warn when G++ generates code that is probably not compatible with the
1462 vendor-neutral C++ ABI. Although an effort has been made to warn about
1463 all such cases, there are probably some cases that are not warned about,
1464 even though G++ is generating incompatible code. There may also be
1465 cases where warnings are emitted even though the code that is generated
1468 You should rewrite your code to avoid these warnings if you are
1469 concerned about the fact that code generated by G++ may not be binary
1470 compatible with code generated by other compilers.
1472 The known incompatibilities at this point include:
1477 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1478 pack data into the same byte as a base class. For example:
1481 struct A @{ virtual void f(); int f1 : 1; @};
1482 struct B : public A @{ int f2 : 1; @};
1486 In this case, G++ will place @code{B::f2} into the same byte
1487 as@code{A::f1}; other compilers will not. You can avoid this problem
1488 by explicitly padding @code{A} so that its size is a multiple of the
1489 byte size on your platform; that will cause G++ and other compilers to
1490 layout @code{B} identically.
1493 Incorrect handling of tail-padding for virtual bases. G++ does not use
1494 tail padding when laying out virtual bases. For example:
1497 struct A @{ virtual void f(); char c1; @};
1498 struct B @{ B(); char c2; @};
1499 struct C : public A, public virtual B @{@};
1503 In this case, G++ will not place @code{B} into the tail-padding for
1504 @code{A}; other compilers will. You can avoid this problem by
1505 explicitly padding @code{A} so that its size is a multiple of its
1506 alignment (ignoring virtual base classes); that will cause G++ and other
1507 compilers to layout @code{C} identically.
1510 Incorrect handling of bit-fields with declared widths greater than that
1511 of their underlying types, when the bit-fields appear in a union. For
1515 union U @{ int i : 4096; @};
1519 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1520 union too small by the number of bits in an @code{int}.
1523 Empty classes can be placed at incorrect offsets. For example:
1533 struct C : public B, public A @{@};
1537 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1538 it should be placed at offset zero. G++ mistakenly believes that the
1539 @code{A} data member of @code{B} is already at offset zero.
1542 Names of template functions whose types involve @code{typename} or
1543 template template parameters can be mangled incorrectly.
1546 template <typename Q>
1547 void f(typename Q::X) @{@}
1549 template <template <typename> class Q>
1550 void f(typename Q<int>::X) @{@}
1554 Instantiations of these templates may be mangled incorrectly.
1558 @item -Wctor-dtor-privacy @r{(C++ only)}
1559 @opindex Wctor-dtor-privacy
1560 Warn when a class seems unusable because all the constructors or
1561 destructors in that class are private, and it has neither friends nor
1562 public static member functions.
1564 @item -Wnon-virtual-dtor @r{(C++ only)}
1565 @opindex Wnon-virtual-dtor
1566 Warn when a class appears to be polymorphic, thereby requiring a virtual
1567 destructor, yet it declares a non-virtual one.
1568 This warning is enabled by @option{-Wall}.
1570 @item -Wreorder @r{(C++ only)}
1572 @cindex reordering, warning
1573 @cindex warning for reordering of member initializers
1574 Warn when the order of member initializers given in the code does not
1575 match the order in which they must be executed. For instance:
1581 A(): j (0), i (1) @{ @}
1585 The compiler will rearrange the member initializers for @samp{i}
1586 and @samp{j} to match the declaration order of the members, emitting
1587 a warning to that effect. This warning is enabled by @option{-Wall}.
1590 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1593 @item -Weffc++ @r{(C++ only)}
1595 Warn about violations of the following style guidelines from Scott Meyers'
1596 @cite{Effective C++} book:
1600 Item 11: Define a copy constructor and an assignment operator for classes
1601 with dynamically allocated memory.
1604 Item 12: Prefer initialization to assignment in constructors.
1607 Item 14: Make destructors virtual in base classes.
1610 Item 15: Have @code{operator=} return a reference to @code{*this}.
1613 Item 23: Don't try to return a reference when you must return an object.
1617 Also warn about violations of the following style guidelines from
1618 Scott Meyers' @cite{More Effective C++} book:
1622 Item 6: Distinguish between prefix and postfix forms of increment and
1623 decrement operators.
1626 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1630 When selecting this option, be aware that the standard library
1631 headers do not obey all of these guidelines; use @samp{grep -v}
1632 to filter out those warnings.
1634 @item -Wno-deprecated @r{(C++ only)}
1635 @opindex Wno-deprecated
1636 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1638 @item -Wno-non-template-friend @r{(C++ only)}
1639 @opindex Wno-non-template-friend
1640 Disable warnings when non-templatized friend functions are declared
1641 within a template. Since the advent of explicit template specification
1642 support in G++, if the name of the friend is an unqualified-id (i.e.,
1643 @samp{friend foo(int)}), the C++ language specification demands that the
1644 friend declare or define an ordinary, nontemplate function. (Section
1645 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1646 could be interpreted as a particular specialization of a templatized
1647 function. Because this non-conforming behavior is no longer the default
1648 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1649 check existing code for potential trouble spots and is on by default.
1650 This new compiler behavior can be turned off with
1651 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1652 but disables the helpful warning.
1654 @item -Wold-style-cast @r{(C++ only)}
1655 @opindex Wold-style-cast
1656 Warn if an old-style (C-style) cast to a non-void type is used within
1657 a C++ program. The new-style casts (@samp{static_cast},
1658 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1659 unintended effects and much easier to search for.
1661 @item -Woverloaded-virtual @r{(C++ only)}
1662 @opindex Woverloaded-virtual
1663 @cindex overloaded virtual fn, warning
1664 @cindex warning for overloaded virtual fn
1665 Warn when a function declaration hides virtual functions from a
1666 base class. For example, in:
1673 struct B: public A @{
1678 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1686 will fail to compile.
1688 @item -Wno-pmf-conversions @r{(C++ only)}
1689 @opindex Wno-pmf-conversions
1690 Disable the diagnostic for converting a bound pointer to member function
1693 @item -Wsign-promo @r{(C++ only)}
1694 @opindex Wsign-promo
1695 Warn when overload resolution chooses a promotion from unsigned or
1696 enumeral type to a signed type, over a conversion to an unsigned type of
1697 the same size. Previous versions of G++ would try to preserve
1698 unsignedness, but the standard mandates the current behavior.
1700 @item -Wsynth @r{(C++ only)}
1702 @cindex warning for synthesized methods
1703 @cindex synthesized methods, warning
1704 Warn when G++'s synthesis behavior does not match that of cfront. For
1710 A& operator = (int);
1720 In this example, G++ will synthesize a default @samp{A& operator =
1721 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1724 @node Objective-C Dialect Options
1725 @section Options Controlling Objective-C Dialect
1727 @cindex compiler options, Objective-C
1728 @cindex Objective-C options, command line
1729 @cindex options, Objective-C
1730 (NOTE: This manual does not describe the Objective-C language itself. See
1731 @w{@uref{http://gcc.gnu.org/readings.html}} for references.)
1733 This section describes the command-line options that are only meaningful
1734 for Objective-C programs, but you can also use most of the GNU compiler
1735 options regardless of what language your program is in. For example,
1736 you might compile a file @code{some_class.m} like this:
1739 gcc -g -fgnu-runtime -O -c some_class.m
1743 In this example, @option{-fgnu-runtime} is an option meant only for
1744 Objective-C programs; you can use the other options with any language
1747 Here is a list of options that are @emph{only} for compiling Objective-C
1751 @item -fconstant-string-class=@var{class-name}
1752 @opindex fconstant-string-class
1753 Use @var{class-name} as the name of the class to instantiate for each
1754 literal string specified with the syntax @code{@@"@dots{}"}. The default
1755 class name is @code{NXConstantString} if the GNU runtime is being used, and
1756 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1757 @option{-fconstant-cfstrings} option, if also present, will override the
1758 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1759 to be laid out as constant CoreFoundation strings.
1762 @opindex fgnu-runtime
1763 Generate object code compatible with the standard GNU Objective-C
1764 runtime. This is the default for most types of systems.
1766 @item -fnext-runtime
1767 @opindex fnext-runtime
1768 Generate output compatible with the NeXT runtime. This is the default
1769 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1770 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1773 @item -fno-nil-receivers
1774 @opindex -fno-nil-receivers
1775 Assume that all Objective-C message dispatches (e.g.,
1776 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1777 is not @code{nil}. This allows for more efficient entry points in the runtime to be
1778 used. Currently, this option is only available in conjunction with
1779 the NeXT runtime on Mac OS X 10.3 and later.
1781 @item -fobjc-exceptions
1782 @opindex -fobjc-exceptions
1783 Enable syntactic support for structured exception handling in Objective-C,
1784 similar to what is offered by C++ and Java. Currently, this option is only
1785 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1793 @@catch (AnObjCClass *exc) @{
1800 @@catch (AnotherClass *exc) @{
1803 @@catch (id allOthers) @{
1813 The @code{@@throw} statement may appear anywhere in an Objective-C or
1814 Objective-C++ program; when used inside of a @code{@@catch} block, the
1815 @code{@@throw} may appear without an argument (as shown above), in which case
1816 the object caught by the @code{@@catch} will be rethrown.
1818 Note that only (pointers to) Objective-C objects may be thrown and
1819 caught using this scheme. When an object is thrown, it will be caught
1820 by the nearest @code{@@catch} clause capable of handling objects of that type,
1821 analogously to how @code{catch} blocks work in C++ and Java. A
1822 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1823 any and all Objective-C exceptions not caught by previous @code{@@catch}
1826 The @code{@@finally} clause, if present, will be executed upon exit from the
1827 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1828 regardless of whether any exceptions are thrown, caught or rethrown
1829 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1830 of the @code{finally} clause in Java.
1832 There are several caveats to using the new exception mechanism:
1836 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1837 idioms provided by the @code{NSException} class, the new
1838 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1839 systems, due to additional functionality needed in the (NeXT) Objective-C
1843 As mentioned above, the new exceptions do not support handling
1844 types other than Objective-C objects. Furthermore, when used from
1845 Objective-C++, the Objective-C exception model does not interoperate with C++
1846 exceptions at this time. This means you cannot @code{@@throw} an exception
1847 from Objective-C and @code{catch} it in C++, or vice versa
1848 (i.e., @code{throw @dots{} @@catch}).
1851 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1852 blocks for thread-safe execution:
1855 @@synchronized (ObjCClass *guard) @{
1860 Upon entering the @code{@@synchronized} block, a thread of execution shall
1861 first check whether a lock has been placed on the corresponding @code{guard}
1862 object by another thread. If it has, the current thread shall wait until
1863 the other thread relinquishes its lock. Once @code{guard} becomes available,
1864 the current thread will place its own lock on it, execute the code contained in
1865 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1866 making @code{guard} available to other threads).
1868 Unlike Java, Objective-C does not allow for entire methods to be marked
1869 @code{@@synchronized}. Note that throwing exceptions out of
1870 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1871 to be unlocked properly.
1873 @item -freplace-objc-classes
1874 @opindex -freplace-objc-classes
1875 Emit a special marker instructing @command{ld(1)} not to statically link in
1876 the resulting object file, and allow @command{dyld(1)} to load it in at
1877 run time instead. This is used in conjunction with the Fix-and-Continue
1878 debugging mode, where the object file in question may be recompiled and
1879 dynamically reloaded in the course of program execution, without the need
1880 to restart the program itself. Currently, Fix-and-Continue functionality
1881 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1885 @opindex -fzero-link
1886 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1887 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1888 compile time) with static class references that get initialized at load time,
1889 which improves run-time performance. Specifying the @option{-fzero-link} flag
1890 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1891 to be retained. This is useful in Zero-Link debugging mode, since it allows
1892 for individual class implementations to be modified during program execution.
1896 Dump interface declarations for all classes seen in the source file to a
1897 file named @file{@var{sourcename}.decl}.
1900 @opindex Wno-protocol
1901 If a class is declared to implement a protocol, a warning is issued for
1902 every method in the protocol that is not implemented by the class. The
1903 default behavior is to issue a warning for every method not explicitly
1904 implemented in the class, even if a method implementation is inherited
1905 from the superclass. If you use the @code{-Wno-protocol} option, then
1906 methods inherited from the superclass are considered to be implemented,
1907 and no warning is issued for them.
1911 Warn if multiple methods of different types for the same selector are
1912 found during compilation. The check is performed on the list of methods
1913 in the final stage of compilation. Additionally, a check is performed
1914 for each selector appearing in a @code{@@selector(@dots{})}
1915 expression, and a corresponding method for that selector has been found
1916 during compilation. Because these checks scan the method table only at
1917 the end of compilation, these warnings are not produced if the final
1918 stage of compilation is not reached, for example because an error is
1919 found during compilation, or because the @code{-fsyntax-only} option is
1922 @item -Wundeclared-selector
1923 @opindex Wundeclared-selector
1924 Warn if a @code{@@selector(@dots{})} expression referring to an
1925 undeclared selector is found. A selector is considered undeclared if no
1926 method with that name has been declared before the
1927 @code{@@selector(@dots{})} expression, either explicitly in an
1928 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
1929 an @code{@@implementation} section. This option always performs its
1930 checks as soon as a @code{@@selector(@dots{})} expression is found,
1931 while @code{-Wselector} only performs its checks in the final stage of
1932 compilation. This also enforces the coding style convention
1933 that methods and selectors must be declared before being used.
1935 @item -print-objc-runtime-info
1936 @opindex -print-objc-runtime-info
1937 Generate C header describing the largest structure that is passed by
1942 @node Language Independent Options
1943 @section Options to Control Diagnostic Messages Formatting
1944 @cindex options to control diagnostics formatting
1945 @cindex diagnostic messages
1946 @cindex message formatting
1948 Traditionally, diagnostic messages have been formatted irrespective of
1949 the output device's aspect (e.g.@: its width, @dots{}). The options described
1950 below can be used to control the diagnostic messages formatting
1951 algorithm, e.g.@: how many characters per line, how often source location
1952 information should be reported. Right now, only the C++ front end can
1953 honor these options. However it is expected, in the near future, that
1954 the remaining front ends would be able to digest them correctly.
1957 @item -fmessage-length=@var{n}
1958 @opindex fmessage-length
1959 Try to format error messages so that they fit on lines of about @var{n}
1960 characters. The default is 72 characters for @command{g++} and 0 for the rest of
1961 the front ends supported by GCC@. If @var{n} is zero, then no
1962 line-wrapping will be done; each error message will appear on a single
1965 @opindex fdiagnostics-show-location
1966 @item -fdiagnostics-show-location=once
1967 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1968 reporter to emit @emph{once} source location information; that is, in
1969 case the message is too long to fit on a single physical line and has to
1970 be wrapped, the source location won't be emitted (as prefix) again,
1971 over and over, in subsequent continuation lines. This is the default
1974 @item -fdiagnostics-show-location=every-line
1975 Only meaningful in line-wrapping mode. Instructs the diagnostic
1976 messages reporter to emit the same source location information (as
1977 prefix) for physical lines that result from the process of breaking
1978 a message which is too long to fit on a single line.
1982 @node Warning Options
1983 @section Options to Request or Suppress Warnings
1984 @cindex options to control warnings
1985 @cindex warning messages
1986 @cindex messages, warning
1987 @cindex suppressing warnings
1989 Warnings are diagnostic messages that report constructions which
1990 are not inherently erroneous but which are risky or suggest there
1991 may have been an error.
1993 You can request many specific warnings with options beginning @samp{-W},
1994 for example @option{-Wimplicit} to request warnings on implicit
1995 declarations. Each of these specific warning options also has a
1996 negative form beginning @samp{-Wno-} to turn off warnings;
1997 for example, @option{-Wno-implicit}. This manual lists only one of the
1998 two forms, whichever is not the default.
2000 The following options control the amount and kinds of warnings produced
2001 by GCC; for further, language-specific options also refer to
2002 @ref{C++ Dialect Options} and @ref{Objective-C Dialect Options}.
2005 @cindex syntax checking
2007 @opindex fsyntax-only
2008 Check the code for syntax errors, but don't do anything beyond that.
2012 Issue all the warnings demanded by strict ISO C and ISO C++;
2013 reject all programs that use forbidden extensions, and some other
2014 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2015 version of the ISO C standard specified by any @option{-std} option used.
2017 Valid ISO C and ISO C++ programs should compile properly with or without
2018 this option (though a rare few will require @option{-ansi} or a
2019 @option{-std} option specifying the required version of ISO C)@. However,
2020 without this option, certain GNU extensions and traditional C and C++
2021 features are supported as well. With this option, they are rejected.
2023 @option{-pedantic} does not cause warning messages for use of the
2024 alternate keywords whose names begin and end with @samp{__}. Pedantic
2025 warnings are also disabled in the expression that follows
2026 @code{__extension__}. However, only system header files should use
2027 these escape routes; application programs should avoid them.
2028 @xref{Alternate Keywords}.
2030 Some users try to use @option{-pedantic} to check programs for strict ISO
2031 C conformance. They soon find that it does not do quite what they want:
2032 it finds some non-ISO practices, but not all---only those for which
2033 ISO C @emph{requires} a diagnostic, and some others for which
2034 diagnostics have been added.
2036 A feature to report any failure to conform to ISO C might be useful in
2037 some instances, but would require considerable additional work and would
2038 be quite different from @option{-pedantic}. We don't have plans to
2039 support such a feature in the near future.
2041 Where the standard specified with @option{-std} represents a GNU
2042 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2043 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2044 extended dialect is based. Warnings from @option{-pedantic} are given
2045 where they are required by the base standard. (It would not make sense
2046 for such warnings to be given only for features not in the specified GNU
2047 C dialect, since by definition the GNU dialects of C include all
2048 features the compiler supports with the given option, and there would be
2049 nothing to warn about.)
2051 @item -pedantic-errors
2052 @opindex pedantic-errors
2053 Like @option{-pedantic}, except that errors are produced rather than
2058 Inhibit all warning messages.
2062 Inhibit warning messages about the use of @samp{#import}.
2064 @item -Wchar-subscripts
2065 @opindex Wchar-subscripts
2066 Warn if an array subscript has type @code{char}. This is a common cause
2067 of error, as programmers often forget that this type is signed on some
2072 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2073 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2077 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2078 the arguments supplied have types appropriate to the format string
2079 specified, and that the conversions specified in the format string make
2080 sense. This includes standard functions, and others specified by format
2081 attributes (@pxref{Function Attributes}), in the @code{printf},
2082 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2083 not in the C standard) families.
2085 The formats are checked against the format features supported by GNU
2086 libc version 2.2. These include all ISO C90 and C99 features, as well
2087 as features from the Single Unix Specification and some BSD and GNU
2088 extensions. Other library implementations may not support all these
2089 features; GCC does not support warning about features that go beyond a
2090 particular library's limitations. However, if @option{-pedantic} is used
2091 with @option{-Wformat}, warnings will be given about format features not
2092 in the selected standard version (but not for @code{strfmon} formats,
2093 since those are not in any version of the C standard). @xref{C Dialect
2094 Options,,Options Controlling C Dialect}.
2096 Since @option{-Wformat} also checks for null format arguments for
2097 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2099 @option{-Wformat} is included in @option{-Wall}. For more control over some
2100 aspects of format checking, the options @option{-Wformat-y2k},
2101 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2102 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2103 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2106 @opindex Wformat-y2k
2107 If @option{-Wformat} is specified, also warn about @code{strftime}
2108 formats which may yield only a two-digit year.
2110 @item -Wno-format-extra-args
2111 @opindex Wno-format-extra-args
2112 If @option{-Wformat} is specified, do not warn about excess arguments to a
2113 @code{printf} or @code{scanf} format function. The C standard specifies
2114 that such arguments are ignored.
2116 Where the unused arguments lie between used arguments that are
2117 specified with @samp{$} operand number specifications, normally
2118 warnings are still given, since the implementation could not know what
2119 type to pass to @code{va_arg} to skip the unused arguments. However,
2120 in the case of @code{scanf} formats, this option will suppress the
2121 warning if the unused arguments are all pointers, since the Single
2122 Unix Specification says that such unused arguments are allowed.
2124 @item -Wno-format-zero-length
2125 @opindex Wno-format-zero-length
2126 If @option{-Wformat} is specified, do not warn about zero-length formats.
2127 The C standard specifies that zero-length formats are allowed.
2129 @item -Wformat-nonliteral
2130 @opindex Wformat-nonliteral
2131 If @option{-Wformat} is specified, also warn if the format string is not a
2132 string literal and so cannot be checked, unless the format function
2133 takes its format arguments as a @code{va_list}.
2135 @item -Wformat-security
2136 @opindex Wformat-security
2137 If @option{-Wformat} is specified, also warn about uses of format
2138 functions that represent possible security problems. At present, this
2139 warns about calls to @code{printf} and @code{scanf} functions where the
2140 format string is not a string literal and there are no format arguments,
2141 as in @code{printf (foo);}. This may be a security hole if the format
2142 string came from untrusted input and contains @samp{%n}. (This is
2143 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2144 in future warnings may be added to @option{-Wformat-security} that are not
2145 included in @option{-Wformat-nonliteral}.)
2149 Enable @option{-Wformat} plus format checks not included in
2150 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2151 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2155 Warn about passing a null pointer for arguments marked as
2156 requiring a non-null value by the @code{nonnull} function attribute.
2158 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2159 can be disabled with the @option{-Wno-nonnull} option.
2161 @item -Winit-self @r{(C, C++, and Objective-C only)}
2163 Warn about uninitialized variables which are initialized with themselves.
2164 Note this option can only be used with the @option{-Wuninitialized} option,
2165 which in turn only works with @option{-O1} and above.
2167 For example, GCC will warn about @code{i} being uninitialized in the
2168 following snippet only when @option{-Winit-self} has been specified:
2179 @item -Wimplicit-int
2180 @opindex Wimplicit-int
2181 Warn when a declaration does not specify a type.
2183 @item -Wimplicit-function-declaration
2184 @itemx -Werror-implicit-function-declaration
2185 @opindex Wimplicit-function-declaration
2186 @opindex Werror-implicit-function-declaration
2187 Give a warning (or error) whenever a function is used before being
2192 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2196 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2197 function with external linkage, returning int, taking either zero
2198 arguments, two, or three arguments of appropriate types.
2200 @item -Wmissing-braces
2201 @opindex Wmissing-braces
2202 Warn if an aggregate or union initializer is not fully bracketed. In
2203 the following example, the initializer for @samp{a} is not fully
2204 bracketed, but that for @samp{b} is fully bracketed.
2207 int a[2][2] = @{ 0, 1, 2, 3 @};
2208 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2212 @opindex Wparentheses
2213 Warn if parentheses are omitted in certain contexts, such
2214 as when there is an assignment in a context where a truth value
2215 is expected, or when operators are nested whose precedence people
2216 often get confused about.
2218 Also warn about constructions where there may be confusion to which
2219 @code{if} statement an @code{else} branch belongs. Here is an example of
2234 In C, every @code{else} branch belongs to the innermost possible @code{if}
2235 statement, which in this example is @code{if (b)}. This is often not
2236 what the programmer expected, as illustrated in the above example by
2237 indentation the programmer chose. When there is the potential for this
2238 confusion, GCC will issue a warning when this flag is specified.
2239 To eliminate the warning, add explicit braces around the innermost
2240 @code{if} statement so there is no way the @code{else} could belong to
2241 the enclosing @code{if}. The resulting code would look like this:
2257 @item -Wsequence-point
2258 @opindex Wsequence-point
2259 Warn about code that may have undefined semantics because of violations
2260 of sequence point rules in the C standard.
2262 The C standard defines the order in which expressions in a C program are
2263 evaluated in terms of @dfn{sequence points}, which represent a partial
2264 ordering between the execution of parts of the program: those executed
2265 before the sequence point, and those executed after it. These occur
2266 after the evaluation of a full expression (one which is not part of a
2267 larger expression), after the evaluation of the first operand of a
2268 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2269 function is called (but after the evaluation of its arguments and the
2270 expression denoting the called function), and in certain other places.
2271 Other than as expressed by the sequence point rules, the order of
2272 evaluation of subexpressions of an expression is not specified. All
2273 these rules describe only a partial order rather than a total order,
2274 since, for example, if two functions are called within one expression
2275 with no sequence point between them, the order in which the functions
2276 are called is not specified. However, the standards committee have
2277 ruled that function calls do not overlap.
2279 It is not specified when between sequence points modifications to the
2280 values of objects take effect. Programs whose behavior depends on this
2281 have undefined behavior; the C standard specifies that ``Between the
2282 previous and next sequence point an object shall have its stored value
2283 modified at most once by the evaluation of an expression. Furthermore,
2284 the prior value shall be read only to determine the value to be
2285 stored.''. If a program breaks these rules, the results on any
2286 particular implementation are entirely unpredictable.
2288 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2289 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2290 diagnosed by this option, and it may give an occasional false positive
2291 result, but in general it has been found fairly effective at detecting
2292 this sort of problem in programs.
2294 The present implementation of this option only works for C programs. A
2295 future implementation may also work for C++ programs.
2297 The C standard is worded confusingly, therefore there is some debate
2298 over the precise meaning of the sequence point rules in subtle cases.
2299 Links to discussions of the problem, including proposed formal
2300 definitions, may be found on our readings page, at
2301 @w{@uref{http://gcc.gnu.org/readings.html}}.
2304 @opindex Wreturn-type
2305 Warn whenever a function is defined with a return-type that defaults to
2306 @code{int}. Also warn about any @code{return} statement with no
2307 return-value in a function whose return-type is not @code{void}.
2309 For C++, a function without return type always produces a diagnostic
2310 message, even when @option{-Wno-return-type} is specified. The only
2311 exceptions are @samp{main} and functions defined in system headers.
2315 Warn whenever a @code{switch} statement has an index of enumeral type
2316 and lacks a @code{case} for one or more of the named codes of that
2317 enumeration. (The presence of a @code{default} label prevents this
2318 warning.) @code{case} labels outside the enumeration range also
2319 provoke warnings when this option is used.
2321 @item -Wswitch-default
2322 @opindex Wswitch-switch
2323 Warn whenever a @code{switch} statement does not have a @code{default}
2327 @opindex Wswitch-enum
2328 Warn whenever a @code{switch} statement has an index of enumeral type
2329 and lacks a @code{case} for one or more of the named codes of that
2330 enumeration. @code{case} labels outside the enumeration range also
2331 provoke warnings when this option is used.
2335 Warn if any trigraphs are encountered that might change the meaning of
2336 the program (trigraphs within comments are not warned about).
2338 @item -Wunused-function
2339 @opindex Wunused-function
2340 Warn whenever a static function is declared but not defined or a
2341 non\-inline static function is unused.
2343 @item -Wunused-label
2344 @opindex Wunused-label
2345 Warn whenever a label is declared but not used.
2347 To suppress this warning use the @samp{unused} attribute
2348 (@pxref{Variable Attributes}).
2350 @item -Wunused-parameter
2351 @opindex Wunused-parameter
2352 Warn whenever a function parameter is unused aside from its declaration.
2354 To suppress this warning use the @samp{unused} attribute
2355 (@pxref{Variable Attributes}).
2357 @item -Wunused-variable
2358 @opindex Wunused-variable
2359 Warn whenever a local variable or non-constant static variable is unused
2360 aside from its declaration
2362 To suppress this warning use the @samp{unused} attribute
2363 (@pxref{Variable Attributes}).
2365 @item -Wunused-value
2366 @opindex Wunused-value
2367 Warn whenever a statement computes a result that is explicitly not used.
2369 To suppress this warning cast the expression to @samp{void}.
2373 All the above @option{-Wunused} options combined.
2375 In order to get a warning about an unused function parameter, you must
2376 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2377 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2379 @item -Wuninitialized
2380 @opindex Wuninitialized
2381 Warn if an automatic variable is used without first being initialized or
2382 if a variable may be clobbered by a @code{setjmp} call.
2384 These warnings are possible only in optimizing compilation,
2385 because they require data flow information that is computed only
2386 when optimizing. If you don't specify @option{-O}, you simply won't
2389 If you want to warn about code which uses the uninitialized value of the
2390 variable in its own initializer, use the @option{-Winit-self} option.
2392 These warnings occur only for variables that are candidates for
2393 register allocation. Therefore, they do not occur for a variable that
2394 is declared @code{volatile}, or whose address is taken, or whose size
2395 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2396 structures, unions or arrays, even when they are in registers.
2398 Note that there may be no warning about a variable that is used only
2399 to compute a value that itself is never used, because such
2400 computations may be deleted by data flow analysis before the warnings
2403 These warnings are made optional because GCC is not smart
2404 enough to see all the reasons why the code might be correct
2405 despite appearing to have an error. Here is one example of how
2426 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2427 always initialized, but GCC doesn't know this. Here is
2428 another common case:
2433 if (change_y) save_y = y, y = new_y;
2435 if (change_y) y = save_y;
2440 This has no bug because @code{save_y} is used only if it is set.
2442 @cindex @code{longjmp} warnings
2443 This option also warns when a non-volatile automatic variable might be
2444 changed by a call to @code{longjmp}. These warnings as well are possible
2445 only in optimizing compilation.
2447 The compiler sees only the calls to @code{setjmp}. It cannot know
2448 where @code{longjmp} will be called; in fact, a signal handler could
2449 call it at any point in the code. As a result, you may get a warning
2450 even when there is in fact no problem because @code{longjmp} cannot
2451 in fact be called at the place which would cause a problem.
2453 Some spurious warnings can be avoided if you declare all the functions
2454 you use that never return as @code{noreturn}. @xref{Function
2457 @item -Wunknown-pragmas
2458 @opindex Wunknown-pragmas
2459 @cindex warning for unknown pragmas
2460 @cindex unknown pragmas, warning
2461 @cindex pragmas, warning of unknown
2462 Warn when a #pragma directive is encountered which is not understood by
2463 GCC@. If this command line option is used, warnings will even be issued
2464 for unknown pragmas in system header files. This is not the case if
2465 the warnings were only enabled by the @option{-Wall} command line option.
2467 @item -Wstrict-aliasing
2468 @opindex Wstrict-aliasing
2469 This option is only active when @option{-fstrict-aliasing} is active.
2470 It warns about code which might break the strict aliasing rules that the
2471 compiler is using for optimization. The warning does not catch all
2472 cases, but does attempt to catch the more common pitfalls. It is
2473 included in @option{-Wall}.
2477 All of the above @samp{-W} options combined. This enables all the
2478 warnings about constructions that some users consider questionable, and
2479 that are easy to avoid (or modify to prevent the warning), even in
2480 conjunction with macros. This also enables some language-specific
2481 warnings described in @ref{C++ Dialect Options} and
2482 @ref{Objective-C Dialect Options}.
2485 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2486 Some of them warn about constructions that users generally do not
2487 consider questionable, but which occasionally you might wish to check
2488 for; others warn about constructions that are necessary or hard to avoid
2489 in some cases, and there is no simple way to modify the code to suppress
2496 (This option used to be called @option{-W}. The older name is still
2497 supported, but the newer name is more descriptive.) Print extra warning
2498 messages for these events:
2502 A function can return either with or without a value. (Falling
2503 off the end of the function body is considered returning without
2504 a value.) For example, this function would evoke such a
2518 An expression-statement or the left-hand side of a comma expression
2519 contains no side effects.
2520 To suppress the warning, cast the unused expression to void.
2521 For example, an expression such as @samp{x[i,j]} will cause a warning,
2522 but @samp{x[(void)i,j]} will not.
2525 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2528 A comparison like @samp{x<=y<=z} appears; this is equivalent to
2529 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
2530 that of ordinary mathematical notation.
2533 Storage-class specifiers like @code{static} are not the first things in
2534 a declaration. According to the C Standard, this usage is obsolescent.
2537 The return type of a function has a type qualifier such as @code{const}.
2538 Such a type qualifier has no effect, since the value returned by a
2539 function is not an lvalue. (But don't warn about the GNU extension of
2540 @code{volatile void} return types. That extension will be warned about
2541 if @option{-pedantic} is specified.)
2544 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2548 A comparison between signed and unsigned values could produce an
2549 incorrect result when the signed value is converted to unsigned.
2550 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2553 An aggregate has an initializer which does not initialize all members.
2554 For example, the following code would cause such a warning, because
2555 @code{x.h} would be implicitly initialized to zero:
2558 struct s @{ int f, g, h; @};
2559 struct s x = @{ 3, 4 @};
2563 A function parameter is declared without a type specifier in K&R-style
2571 An empty body occurs in an @samp{if} or @samp{else} statement.
2574 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2575 @samp{>}, or @samp{>=}.
2578 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2581 Any of several floating-point events that often indicate errors, such as
2582 overflow, underflow, loss of precision, etc.
2584 @item @r{(C++ only)}
2585 An enumerator and a non-enumerator both appear in a conditional expression.
2587 @item @r{(C++ only)}
2588 A non-static reference or non-static @samp{const} member appears in a
2589 class without constructors.
2591 @item @r{(C++ only)}
2592 Ambiguous virtual bases.
2594 @item @r{(C++ only)}
2595 Subscripting an array which has been declared @samp{register}.
2597 @item @r{(C++ only)}
2598 Taking the address of a variable which has been declared @samp{register}.
2600 @item @r{(C++ only)}
2601 A base class is not initialized in a derived class' copy constructor.
2604 @item -Wno-div-by-zero
2605 @opindex Wno-div-by-zero
2606 @opindex Wdiv-by-zero
2607 Do not warn about compile-time integer division by zero. Floating point
2608 division by zero is not warned about, as it can be a legitimate way of
2609 obtaining infinities and NaNs.
2611 @item -Wsystem-headers
2612 @opindex Wsystem-headers
2613 @cindex warnings from system headers
2614 @cindex system headers, warnings from
2615 Print warning messages for constructs found in system header files.
2616 Warnings from system headers are normally suppressed, on the assumption
2617 that they usually do not indicate real problems and would only make the
2618 compiler output harder to read. Using this command line option tells
2619 GCC to emit warnings from system headers as if they occurred in user
2620 code. However, note that using @option{-Wall} in conjunction with this
2621 option will @emph{not} warn about unknown pragmas in system
2622 headers---for that, @option{-Wunknown-pragmas} must also be used.
2625 @opindex Wfloat-equal
2626 Warn if floating point values are used in equality comparisons.
2628 The idea behind this is that sometimes it is convenient (for the
2629 programmer) to consider floating-point values as approximations to
2630 infinitely precise real numbers. If you are doing this, then you need
2631 to compute (by analyzing the code, or in some other way) the maximum or
2632 likely maximum error that the computation introduces, and allow for it
2633 when performing comparisons (and when producing output, but that's a
2634 different problem). In particular, instead of testing for equality, you
2635 would check to see whether the two values have ranges that overlap; and
2636 this is done with the relational operators, so equality comparisons are
2639 @item -Wtraditional @r{(C only)}
2640 @opindex Wtraditional
2641 Warn about certain constructs that behave differently in traditional and
2642 ISO C@. Also warn about ISO C constructs that have no traditional C
2643 equivalent, and/or problematic constructs which should be avoided.
2647 Macro parameters that appear within string literals in the macro body.
2648 In traditional C macro replacement takes place within string literals,
2649 but does not in ISO C@.
2652 In traditional C, some preprocessor directives did not exist.
2653 Traditional preprocessors would only consider a line to be a directive
2654 if the @samp{#} appeared in column 1 on the line. Therefore
2655 @option{-Wtraditional} warns about directives that traditional C
2656 understands but would ignore because the @samp{#} does not appear as the
2657 first character on the line. It also suggests you hide directives like
2658 @samp{#pragma} not understood by traditional C by indenting them. Some
2659 traditional implementations would not recognize @samp{#elif}, so it
2660 suggests avoiding it altogether.
2663 A function-like macro that appears without arguments.
2666 The unary plus operator.
2669 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2670 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2671 constants.) Note, these suffixes appear in macros defined in the system
2672 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2673 Use of these macros in user code might normally lead to spurious
2674 warnings, however gcc's integrated preprocessor has enough context to
2675 avoid warning in these cases.
2678 A function declared external in one block and then used after the end of
2682 A @code{switch} statement has an operand of type @code{long}.
2685 A non-@code{static} function declaration follows a @code{static} one.
2686 This construct is not accepted by some traditional C compilers.
2689 The ISO type of an integer constant has a different width or
2690 signedness from its traditional type. This warning is only issued if
2691 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2692 typically represent bit patterns, are not warned about.
2695 Usage of ISO string concatenation is detected.
2698 Initialization of automatic aggregates.
2701 Identifier conflicts with labels. Traditional C lacks a separate
2702 namespace for labels.
2705 Initialization of unions. If the initializer is zero, the warning is
2706 omitted. This is done under the assumption that the zero initializer in
2707 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2708 initializer warnings and relies on default initialization to zero in the
2712 Conversions by prototypes between fixed/floating point values and vice
2713 versa. The absence of these prototypes when compiling with traditional
2714 C would cause serious problems. This is a subset of the possible
2715 conversion warnings, for the full set use @option{-Wconversion}.
2718 Use of ISO C style function definitions. This warning intentionally is
2719 @emph{not} issued for prototype declarations or variadic functions
2720 because these ISO C features will appear in your code when using
2721 libiberty's traditional C compatibility macros, @code{PARAMS} and
2722 @code{VPARAMS}. This warning is also bypassed for nested functions
2723 because that feature is already a gcc extension and thus not relevant to
2724 traditional C compatibility.
2727 @item -Wdeclaration-after-statement @r{(C only)}
2728 @opindex Wdeclaration-after-statement
2729 Warn when a declaration is found after a statement in a block. This
2730 construct, known from C++, was introduced with ISO C99 and is by default
2731 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2732 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2736 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2738 @item -Wendif-labels
2739 @opindex Wendif-labels
2740 Warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2744 Warn whenever a local variable shadows another local variable, parameter or
2745 global variable or whenever a built-in function is shadowed.
2747 @item -Wlarger-than-@var{len}
2748 @opindex Wlarger-than
2749 Warn whenever an object of larger than @var{len} bytes is defined.
2751 @item -Wpointer-arith
2752 @opindex Wpointer-arith
2753 Warn about anything that depends on the ``size of'' a function type or
2754 of @code{void}. GNU C assigns these types a size of 1, for
2755 convenience in calculations with @code{void *} pointers and pointers
2758 @item -Wbad-function-cast @r{(C only)}
2759 @opindex Wbad-function-cast
2760 Warn whenever a function call is cast to a non-matching type.
2761 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2765 Warn whenever a pointer is cast so as to remove a type qualifier from
2766 the target type. For example, warn if a @code{const char *} is cast
2767 to an ordinary @code{char *}.
2770 @opindex Wcast-align
2771 Warn whenever a pointer is cast such that the required alignment of the
2772 target is increased. For example, warn if a @code{char *} is cast to
2773 an @code{int *} on machines where integers can only be accessed at
2774 two- or four-byte boundaries.
2776 @item -Wwrite-strings
2777 @opindex Wwrite-strings
2778 When compiling C, give string constants the type @code{const
2779 char[@var{length}]} so that
2780 copying the address of one into a non-@code{const} @code{char *}
2781 pointer will get a warning; when compiling C++, warn about the
2782 deprecated conversion from string constants to @code{char *}.
2783 These warnings will help you find at
2784 compile time code that can try to write into a string constant, but
2785 only if you have been very careful about using @code{const} in
2786 declarations and prototypes. Otherwise, it will just be a nuisance;
2787 this is why we did not make @option{-Wall} request these warnings.
2790 @opindex Wconversion
2791 Warn if a prototype causes a type conversion that is different from what
2792 would happen to the same argument in the absence of a prototype. This
2793 includes conversions of fixed point to floating and vice versa, and
2794 conversions changing the width or signedness of a fixed point argument
2795 except when the same as the default promotion.
2797 Also, warn if a negative integer constant expression is implicitly
2798 converted to an unsigned type. For example, warn about the assignment
2799 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2800 casts like @code{(unsigned) -1}.
2802 @item -Wsign-compare
2803 @opindex Wsign-compare
2804 @cindex warning for comparison of signed and unsigned values
2805 @cindex comparison of signed and unsigned values, warning
2806 @cindex signed and unsigned values, comparison warning
2807 Warn when a comparison between signed and unsigned values could produce
2808 an incorrect result when the signed value is converted to unsigned.
2809 This warning is also enabled by @option{-Wextra}; to get the other warnings
2810 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2812 @item -Waggregate-return
2813 @opindex Waggregate-return
2814 Warn if any functions that return structures or unions are defined or
2815 called. (In languages where you can return an array, this also elicits
2818 @item -Wstrict-prototypes @r{(C only)}
2819 @opindex Wstrict-prototypes
2820 Warn if a function is declared or defined without specifying the
2821 argument types. (An old-style function definition is permitted without
2822 a warning if preceded by a declaration which specifies the argument
2825 @item -Wold-style-definition @r{(C only)}
2826 @opindex Wold-style-definition
2827 Warn if an old-style function definition is used. A warning is given
2828 even if there is a previous prototype.
2830 @item -Wmissing-prototypes @r{(C only)}
2831 @opindex Wmissing-prototypes
2832 Warn if a global function is defined without a previous prototype
2833 declaration. This warning is issued even if the definition itself
2834 provides a prototype. The aim is to detect global functions that fail
2835 to be declared in header files.
2837 @item -Wmissing-declarations @r{(C only)}
2838 @opindex Wmissing-declarations
2839 Warn if a global function is defined without a previous declaration.
2840 Do so even if the definition itself provides a prototype.
2841 Use this option to detect global functions that are not declared in
2844 @item -Wmissing-noreturn
2845 @opindex Wmissing-noreturn
2846 Warn about functions which might be candidates for attribute @code{noreturn}.
2847 Note these are only possible candidates, not absolute ones. Care should
2848 be taken to manually verify functions actually do not ever return before
2849 adding the @code{noreturn} attribute, otherwise subtle code generation
2850 bugs could be introduced. You will not get a warning for @code{main} in
2851 hosted C environments.
2853 @item -Wmissing-format-attribute
2854 @opindex Wmissing-format-attribute
2856 If @option{-Wformat} is enabled, also warn about functions which might be
2857 candidates for @code{format} attributes. Note these are only possible
2858 candidates, not absolute ones. GCC will guess that @code{format}
2859 attributes might be appropriate for any function that calls a function
2860 like @code{vprintf} or @code{vscanf}, but this might not always be the
2861 case, and some functions for which @code{format} attributes are
2862 appropriate may not be detected. This option has no effect unless
2863 @option{-Wformat} is enabled (possibly by @option{-Wall}).
2865 @item -Wno-multichar
2866 @opindex Wno-multichar
2868 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
2869 Usually they indicate a typo in the user's code, as they have
2870 implementation-defined values, and should not be used in portable code.
2872 @item -Wno-deprecated-declarations
2873 @opindex Wno-deprecated-declarations
2874 Do not warn about uses of functions, variables, and types marked as
2875 deprecated by using the @code{deprecated} attribute.
2876 (@pxref{Function Attributes}, @pxref{Variable Attributes},
2877 @pxref{Type Attributes}.)
2881 Warn if a structure is given the packed attribute, but the packed
2882 attribute has no effect on the layout or size of the structure.
2883 Such structures may be mis-aligned for little benefit. For
2884 instance, in this code, the variable @code{f.x} in @code{struct bar}
2885 will be misaligned even though @code{struct bar} does not itself
2886 have the packed attribute:
2893 @} __attribute__((packed));
2903 Warn if padding is included in a structure, either to align an element
2904 of the structure or to align the whole structure. Sometimes when this
2905 happens it is possible to rearrange the fields of the structure to
2906 reduce the padding and so make the structure smaller.
2908 @item -Wredundant-decls
2909 @opindex Wredundant-decls
2910 Warn if anything is declared more than once in the same scope, even in
2911 cases where multiple declaration is valid and changes nothing.
2913 @item -Wnested-externs @r{(C only)}
2914 @opindex Wnested-externs
2915 Warn if an @code{extern} declaration is encountered within a function.
2917 @item -Wunreachable-code
2918 @opindex Wunreachable-code
2919 Warn if the compiler detects that code will never be executed.
2921 This option is intended to warn when the compiler detects that at
2922 least a whole line of source code will never be executed, because
2923 some condition is never satisfied or because it is after a
2924 procedure that never returns.
2926 It is possible for this option to produce a warning even though there
2927 are circumstances under which part of the affected line can be executed,
2928 so care should be taken when removing apparently-unreachable code.
2930 For instance, when a function is inlined, a warning may mean that the
2931 line is unreachable in only one inlined copy of the function.
2933 This option is not made part of @option{-Wall} because in a debugging
2934 version of a program there is often substantial code which checks
2935 correct functioning of the program and is, hopefully, unreachable
2936 because the program does work. Another common use of unreachable
2937 code is to provide behavior which is selectable at compile-time.
2941 Warn if a function can not be inlined and it was declared as inline.
2942 Even with this option, the compiler will not warn about failures to
2943 inline functions declared in system headers.
2945 The compiler uses a variety of heuristics to determine whether or not
2946 to inline a function. For example, the compiler takes into account
2947 the size of the function being inlined and the the amount of inlining
2948 that has already been done in the current function. Therefore,
2949 seemingly insignificant changes in the source program can cause the
2950 warnings produced by @option{-Winline} to appear or disappear.
2952 @item -Wno-invalid-offsetof @r{(C++ only)}
2953 @opindex Wno-invalid-offsetof
2954 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
2955 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
2956 to a non-POD type is undefined. In existing C++ implementations,
2957 however, @samp{offsetof} typically gives meaningful results even when
2958 applied to certain kinds of non-POD types. (Such as a simple
2959 @samp{struct} that fails to be a POD type only by virtue of having a
2960 constructor.) This flag is for users who are aware that they are
2961 writing nonportable code and who have deliberately chosen to ignore the
2964 The restrictions on @samp{offsetof} may be relaxed in a future version
2965 of the C++ standard.
2968 @opindex Winvalid-pch
2969 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
2970 the search path but can't be used.
2974 @opindex Wno-long-long
2975 Warn if @samp{long long} type is used. This is default. To inhibit
2976 the warning messages, use @option{-Wno-long-long}. Flags
2977 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
2978 only when @option{-pedantic} flag is used.
2980 @item -Wdisabled-optimization
2981 @opindex Wdisabled-optimization
2982 Warn if a requested optimization pass is disabled. This warning does
2983 not generally indicate that there is anything wrong with your code; it
2984 merely indicates that GCC's optimizers were unable to handle the code
2985 effectively. Often, the problem is that your code is too big or too
2986 complex; GCC will refuse to optimize programs when the optimization
2987 itself is likely to take inordinate amounts of time.
2991 Make all warnings into errors.
2994 @node Debugging Options
2995 @section Options for Debugging Your Program or GCC
2996 @cindex options, debugging
2997 @cindex debugging information options
2999 GCC has various special options that are used for debugging
3000 either your program or GCC:
3005 Produce debugging information in the operating system's native format
3006 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
3009 On most systems that use stabs format, @option{-g} enables use of extra
3010 debugging information that only GDB can use; this extra information
3011 makes debugging work better in GDB but will probably make other debuggers
3013 refuse to read the program. If you want to control for certain whether
3014 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3015 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3017 Unlike most other C compilers, GCC allows you to use @option{-g} with
3018 @option{-O}. The shortcuts taken by optimized code may occasionally
3019 produce surprising results: some variables you declared may not exist
3020 at all; flow of control may briefly move where you did not expect it;
3021 some statements may not be executed because they compute constant
3022 results or their values were already at hand; some statements may
3023 execute in different places because they were moved out of loops.
3025 Nevertheless it proves possible to debug optimized output. This makes
3026 it reasonable to use the optimizer for programs that might have bugs.
3028 The following options are useful when GCC is generated with the
3029 capability for more than one debugging format.
3033 Produce debugging information for use by GDB@. This means to use the
3034 most expressive format available (DWARF 2, stabs, or the native format
3035 if neither of those are supported), including GDB extensions if at all
3040 Produce debugging information in stabs format (if that is supported),
3041 without GDB extensions. This is the format used by DBX on most BSD
3042 systems. On MIPS, Alpha and System V Release 4 systems this option
3043 produces stabs debugging output which is not understood by DBX or SDB@.
3044 On System V Release 4 systems this option requires the GNU assembler.
3046 @item -feliminate-unused-debug-symbols
3047 @opindex feliminate-unused-debug-symbols
3048 Produce debugging information in stabs format (if that is supported),
3049 for only symbols that are actually used.
3053 Produce debugging information in stabs format (if that is supported),
3054 using GNU extensions understood only by the GNU debugger (GDB)@. The
3055 use of these extensions is likely to make other debuggers crash or
3056 refuse to read the program.
3060 Produce debugging information in COFF format (if that is supported).
3061 This is the format used by SDB on most System V systems prior to
3066 Produce debugging information in XCOFF format (if that is supported).
3067 This is the format used by the DBX debugger on IBM RS/6000 systems.
3071 Produce debugging information in XCOFF format (if that is supported),
3072 using GNU extensions understood only by the GNU debugger (GDB)@. The
3073 use of these extensions is likely to make other debuggers crash or
3074 refuse to read the program, and may cause assemblers other than the GNU
3075 assembler (GAS) to fail with an error.
3079 Produce debugging information in DWARF version 2 format (if that is
3080 supported). This is the format used by DBX on IRIX 6.
3084 Produce debugging information in VMS debug format (if that is
3085 supported). This is the format used by DEBUG on VMS systems.
3088 @itemx -ggdb@var{level}
3089 @itemx -gstabs@var{level}
3090 @itemx -gcoff@var{level}
3091 @itemx -gxcoff@var{level}
3092 @itemx -gvms@var{level}
3093 Request debugging information and also use @var{level} to specify how
3094 much information. The default level is 2.
3096 Level 1 produces minimal information, enough for making backtraces in
3097 parts of the program that you don't plan to debug. This includes
3098 descriptions of functions and external variables, but no information
3099 about local variables and no line numbers.
3101 Level 3 includes extra information, such as all the macro definitions
3102 present in the program. Some debuggers support macro expansion when
3103 you use @option{-g3}.
3105 Note that in order to avoid confusion between DWARF1 debug level 2,
3106 and DWARF2 @option{-gdwarf-2} does not accept a concatenated debug
3107 level. Instead use an additional @option{-g@var{level}} option to
3108 change the debug level for DWARF2.
3110 @item -feliminate-dwarf2-dups
3111 @opindex feliminate-dwarf2-dups
3112 Compress DWARF2 debugging information by eliminating duplicated
3113 information about each symbol. This option only makes sense when
3114 generating DWARF2 debugging information with @option{-gdwarf-2}.
3116 @cindex @command{prof}
3119 Generate extra code to write profile information suitable for the
3120 analysis program @command{prof}. You must use this option when compiling
3121 the source files you want data about, and you must also use it when
3124 @cindex @command{gprof}
3127 Generate extra code to write profile information suitable for the
3128 analysis program @command{gprof}. You must use this option when compiling
3129 the source files you want data about, and you must also use it when
3134 Makes the compiler print out each function name as it is compiled, and
3135 print some statistics about each pass when it finishes.
3138 @opindex ftime-report
3139 Makes the compiler print some statistics about the time consumed by each
3140 pass when it finishes.
3143 @opindex fmem-report
3144 Makes the compiler print some statistics about permanent memory
3145 allocation when it finishes.
3147 @item -fprofile-arcs
3148 @opindex fprofile-arcs
3149 Add code so that program flow @dfn{arcs} are instrumented. During
3150 execution the program records how many times each branch and call is
3151 executed and how many times it is taken or returns. When the compiled
3152 program exits it saves this data to a file called
3153 @file{@var{auxname}.gcda} for each source file. The data may be used for
3154 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3155 test coverage analysis (@option{-ftest-coverage}). Each object file's
3156 @var{auxname} is generated from the name of the output file, if
3157 explicitly specified and it is not the final executable, otherwise it is
3158 the basename of the source file. In both cases any suffix is removed
3159 (e.g. @file{foo.gcda} for input file @file{dir/foo.c}, or
3160 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3165 Compile the source files with @option{-fprofile-arcs} plus optimization
3166 and code generation options. For test coverage analysis, use the
3167 additional @option{-ftest-coverage} option. You do not need to profile
3168 every source file in a program.
3171 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3172 (the latter implies the former).
3175 Run the program on a representative workload to generate the arc profile
3176 information. This may be repeated any number of times. You can run
3177 concurrent instances of your program, and provided that the file system
3178 supports locking, the data files will be correctly updated. Also
3179 @code{fork} calls are detected and correctly handled (double counting
3183 For profile-directed optimizations, compile the source files again with
3184 the same optimization and code generation options plus
3185 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3186 Control Optimization}).
3189 For test coverage analysis, use @command{gcov} to produce human readable
3190 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3191 @command{gcov} documentation for further information.
3195 With @option{-fprofile-arcs}, for each function of your program GCC
3196 creates a program flow graph, then finds a spanning tree for the graph.
3197 Only arcs that are not on the spanning tree have to be instrumented: the
3198 compiler adds code to count the number of times that these arcs are
3199 executed. When an arc is the only exit or only entrance to a block, the
3200 instrumentation code can be added to the block; otherwise, a new basic
3201 block must be created to hold the instrumentation code.
3204 @item -ftest-coverage
3205 @opindex ftest-coverage
3206 Produce a notes file that the @command{gcov} code-coverage utility
3207 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3208 show program coverage. Each source file's note file is called
3209 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3210 above for a description of @var{auxname} and instructions on how to
3211 generate test coverage data. Coverage data will match the source files
3212 more closely, if you do not optimize.
3214 @item -d@var{letters}
3216 Says to make debugging dumps during compilation at times specified by
3217 @var{letters}. This is used for debugging the compiler. The file names
3218 for most of the dumps are made by appending a pass number and a word to
3219 the @var{dumpname}. @var{dumpname} is generated from the name of the
3220 output file, if explicitly specified and it is not an executable,
3221 otherwise it is the basename of the source file. In both cases any
3222 suffix is removed (e.g. @file{foo.01.rtl} or @file{foo.02.sibling}).
3223 Here are the possible letters for use in @var{letters}, and their
3229 Annotate the assembler output with miscellaneous debugging information.
3232 Dump after computing branch probabilities, to @file{@var{file}.12.bp}.
3235 Dump after block reordering, to @file{@var{file}.30.bbro}.
3238 Dump after instruction combination, to the file @file{@var{file}.20.combine}.
3241 Dump after the first if conversion, to the file @file{@var{file}.14.ce1}.
3242 Also dump after the second if conversion, to the file @file{@var{file}.21.ce2}.
3245 Dump after branch target load optimization, to to @file{@var{file}.32.btl}.
3246 Also dump after delayed branch scheduling, to @file{@var{file}.36.dbr}.
3249 Dump all macro definitions, at the end of preprocessing, in addition to
3253 Dump after the second if conversion, to @file{@var{file}.31.ce3}.
3256 Dump after control and data flow analysis, to @file{@var{file}.11.cfg}.
3257 Also dump after life analysis, to @file{@var{file}.19.life}.
3260 Dump after purging @code{ADDRESSOF} codes, to @file{@var{file}.07.addressof}.
3263 Dump after global register allocation, to @file{@var{file}.25.greg}.
3266 Dump after GCSE, to @file{@var{file}.08.gcse}.
3267 Also dump after jump bypassing and control flow optimizations, to
3268 @file{@var{file}.10.bypass}.
3271 Dump after finalization of EH handling code, to @file{@var{file}.03.eh}.
3274 Dump after sibling call optimizations, to @file{@var{file}.02.sibling}.
3277 Dump after the first jump optimization, to @file{@var{file}.04.jump}.
3280 Dump after conversion from registers to stack, to @file{@var{file}.34.stack}.
3283 Dump after local register allocation, to @file{@var{file}.24.lreg}.
3286 Dump after loop optimization passes, to @file{@var{file}.09.loop} and
3287 @file{@var{file}.16.loop2}.
3290 Dump after performing the machine dependent reorganization pass, to
3291 @file{@var{file}.35.mach}.
3294 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3297 Dump after the register move pass, to @file{@var{file}.22.regmove}.
3300 Dump after post-reload optimizations, to @file{@var{file}.26.postreload}.
3303 Dump after RTL generation, to @file{@var{file}.01.rtl}.
3306 Dump after the second scheduling pass, to @file{@var{file}.33.sched2}.
3309 Dump after CSE (including the jump optimization that sometimes follows
3310 CSE), to @file{@var{file}.06.cse}.
3313 Dump after the first scheduling pass, to @file{@var{file}.23.sched}.
3316 Dump after the second CSE pass (including the jump optimization that
3317 sometimes follows CSE), to @file{@var{file}.18.cse2}.
3320 Dump after running tracer, to @file{@var{file}.15.tracer}.
3323 Dump after null pointer elimination pass to @file{@var{file}.05.null}.
3326 Dump callgraph and unit-at-a-time optimization @file{@var{file}.00.unit}.
3329 Dump after the value profile transformations, to @file{@var{file}.13.vpt}.
3332 Dump after the second flow pass, to @file{@var{file}.27.flow2}.
3335 Dump after the peephole pass, to @file{@var{file}.28.peephole2}.
3338 Dump after constructing the web, to @file{@var{file}.17.web}.
3341 Produce all the dumps listed above.
3344 Produce a core dump whenever an error occurs.
3347 Print statistics on memory usage, at the end of the run, to
3351 Annotate the assembler output with a comment indicating which
3352 pattern and alternative was used. The length of each instruction is
3356 Dump the RTL in the assembler output as a comment before each instruction.
3357 Also turns on @option{-dp} annotation.
3360 For each of the other indicated dump files (except for
3361 @file{@var{file}.01.rtl}), dump a representation of the control flow graph
3362 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3365 Just generate RTL for a function instead of compiling it. Usually used
3369 Dump debugging information during parsing, to standard error.
3372 @item -fdump-unnumbered
3373 @opindex fdump-unnumbered
3374 When doing debugging dumps (see @option{-d} option above), suppress instruction
3375 numbers and line number note output. This makes it more feasible to
3376 use diff on debugging dumps for compiler invocations with different
3377 options, in particular with and without @option{-g}.
3379 @item -fdump-translation-unit @r{(C and C++ only)}
3380 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3381 @opindex fdump-translation-unit
3382 Dump a representation of the tree structure for the entire translation
3383 unit to a file. The file name is made by appending @file{.tu} to the
3384 source file name. If the @samp{-@var{options}} form is used, @var{options}
3385 controls the details of the dump as described for the
3386 @option{-fdump-tree} options.
3388 @item -fdump-class-hierarchy @r{(C++ only)}
3389 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3390 @opindex fdump-class-hierarchy
3391 Dump a representation of each class's hierarchy and virtual function
3392 table layout to a file. The file name is made by appending @file{.class}
3393 to the source file name. If the @samp{-@var{options}} form is used,
3394 @var{options} controls the details of the dump as described for the
3395 @option{-fdump-tree} options.
3397 @item -fdump-tree-@var{switch} @r{(C++ only)}
3398 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C++ only)}
3400 Control the dumping at various stages of processing the intermediate
3401 language tree to a file. The file name is generated by appending a switch
3402 specific suffix to the source file name. If the @samp{-@var{options}}
3403 form is used, @var{options} is a list of @samp{-} separated options that
3404 control the details of the dump. Not all options are applicable to all
3405 dumps, those which are not meaningful will be ignored. The following
3406 options are available
3410 Print the address of each node. Usually this is not meaningful as it
3411 changes according to the environment and source file. Its primary use
3412 is for tying up a dump file with a debug environment.
3414 Inhibit dumping of members of a scope or body of a function merely
3415 because that scope has been reached. Only dump such items when they
3416 are directly reachable by some other path.
3418 Turn on all options.
3421 The following tree dumps are possible:
3424 Dump before any tree based optimization, to @file{@var{file}.original}.
3426 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3428 Dump after function inlining, to @file{@var{file}.inlined}.
3431 @item -frandom-seed=@var{string}
3432 @opindex frandom-string
3433 This option provides a seed that GCC uses when it would otherwise use
3434 random numbers. It is used to generate certain symbol names
3435 that have to be different in every compiled file. It is also used to
3436 place unique stamps in coverage data files and the object files that
3437 produce them. You can use the @option{-frandom-seed} option to produce
3438 reproducibly identical object files.
3440 The @var{string} should be different for every file you compile.
3442 @item -fsched-verbose=@var{n}
3443 @opindex fsched-verbose
3444 On targets that use instruction scheduling, this option controls the
3445 amount of debugging output the scheduler prints. This information is
3446 written to standard error, unless @option{-dS} or @option{-dR} is
3447 specified, in which case it is output to the usual dump
3448 listing file, @file{.sched} or @file{.sched2} respectively. However
3449 for @var{n} greater than nine, the output is always printed to standard
3452 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3453 same information as @option{-dRS}. For @var{n} greater than one, it
3454 also output basic block probabilities, detailed ready list information
3455 and unit/insn info. For @var{n} greater than two, it includes RTL
3456 at abort point, control-flow and regions info. And for @var{n} over
3457 four, @option{-fsched-verbose} also includes dependence info.
3461 Store the usual ``temporary'' intermediate files permanently; place them
3462 in the current directory and name them based on the source file. Thus,
3463 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3464 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3465 preprocessed @file{foo.i} output file even though the compiler now
3466 normally uses an integrated preprocessor.
3470 Report the CPU time taken by each subprocess in the compilation
3471 sequence. For C source files, this is the compiler proper and assembler
3472 (plus the linker if linking is done). The output looks like this:
3479 The first number on each line is the ``user time,'' that is time spent
3480 executing the program itself. The second number is ``system time,''
3481 time spent executing operating system routines on behalf of the program.
3482 Both numbers are in seconds.
3484 @item -print-file-name=@var{library}
3485 @opindex print-file-name
3486 Print the full absolute name of the library file @var{library} that
3487 would be used when linking---and don't do anything else. With this
3488 option, GCC does not compile or link anything; it just prints the
3491 @item -print-multi-directory
3492 @opindex print-multi-directory
3493 Print the directory name corresponding to the multilib selected by any
3494 other switches present in the command line. This directory is supposed
3495 to exist in @env{GCC_EXEC_PREFIX}.
3497 @item -print-multi-lib
3498 @opindex print-multi-lib
3499 Print the mapping from multilib directory names to compiler switches
3500 that enable them. The directory name is separated from the switches by
3501 @samp{;}, and each switch starts with an @samp{@@} instead of the
3502 @samp{-}, without spaces between multiple switches. This is supposed to
3503 ease shell-processing.
3505 @item -print-prog-name=@var{program}
3506 @opindex print-prog-name
3507 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3509 @item -print-libgcc-file-name
3510 @opindex print-libgcc-file-name
3511 Same as @option{-print-file-name=libgcc.a}.
3513 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3514 but you do want to link with @file{libgcc.a}. You can do
3517 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3520 @item -print-search-dirs
3521 @opindex print-search-dirs
3522 Print the name of the configured installation directory and a list of
3523 program and library directories gcc will search---and don't do anything else.
3525 This is useful when gcc prints the error message
3526 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3527 To resolve this you either need to put @file{cpp0} and the other compiler
3528 components where gcc expects to find them, or you can set the environment
3529 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3530 Don't forget the trailing '/'.
3531 @xref{Environment Variables}.
3534 @opindex dumpmachine
3535 Print the compiler's target machine (for example,
3536 @samp{i686-pc-linux-gnu})---and don't do anything else.
3539 @opindex dumpversion
3540 Print the compiler version (for example, @samp{3.0})---and don't do
3545 Print the compiler's built-in specs---and don't do anything else. (This
3546 is used when GCC itself is being built.) @xref{Spec Files}.
3548 @item -feliminate-unused-debug-types
3549 @opindex feliminate-unused-debug-types
3550 Normally, when producing DWARF2 output, GCC will emit debugging
3551 information for all types declared in a compilation
3552 unit, regardless of whether or not they are actually used
3553 in that compilation unit. Sometimes this is useful, such as
3554 if, in the debugger, you want to cast a value to a type that is
3555 not actually used in your program (but is declared). More often,
3556 however, this results in a significant amount of wasted space.
3557 With this option, GCC will avoid producing debug symbol output
3558 for types that are nowhere used in the source file being compiled.
3561 @node Optimize Options
3562 @section Options That Control Optimization
3563 @cindex optimize options
3564 @cindex options, optimization
3566 These options control various sorts of optimizations.
3568 Without any optimization option, the compiler's goal is to reduce the
3569 cost of compilation and to make debugging produce the expected
3570 results. Statements are independent: if you stop the program with a
3571 breakpoint between statements, you can then assign a new value to any
3572 variable or change the program counter to any other statement in the
3573 function and get exactly the results you would expect from the source
3576 Turning on optimization flags makes the compiler attempt to improve
3577 the performance and/or code size at the expense of compilation time
3578 and possibly the ability to debug the program.
3580 The compiler performs optimisation based on the knowledge it has of
3581 the program. Using the @option{-funit-at-a-time} flag will allow the
3582 compiler to consider information gained from later functions in the
3583 file when compiling a function. Compiling multiple files at once to a
3584 single output file (and using @option{-funit-at-a-time}) will allow
3585 the compiler to use information gained from all of the files when
3586 compiling each of them.
3588 Not all optimizations are controlled directly by a flag. Only
3589 optimizations that have a flag are listed.
3596 Optimize. Optimizing compilation takes somewhat more time, and a lot
3597 more memory for a large function.
3599 With @option{-O}, the compiler tries to reduce code size and execution
3600 time, without performing any optimizations that take a great deal of
3603 @option{-O} turns on the following optimization flags:
3604 @gccoptlist{-fdefer-pop @gol
3605 -fmerge-constants @gol
3607 -floop-optimize @gol
3608 -fif-conversion @gol
3609 -fif-conversion2 @gol
3610 -fdelayed-branch @gol
3611 -fguess-branch-probability @gol
3614 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
3615 where doing so does not interfere with debugging.
3619 Optimize even more. GCC performs nearly all supported optimizations
3620 that do not involve a space-speed tradeoff. The compiler does not
3621 perform loop unrolling or function inlining when you specify @option{-O2}.
3622 As compared to @option{-O}, this option increases both compilation time
3623 and the performance of the generated code.
3625 @option{-O2} turns on all optimization flags specified by @option{-O}. It
3626 also turns on the following optimization flags:
3627 @gccoptlist{-fforce-mem @gol
3628 -foptimize-sibling-calls @gol
3629 -fstrength-reduce @gol
3630 -fcse-follow-jumps -fcse-skip-blocks @gol
3631 -frerun-cse-after-loop -frerun-loop-opt @gol
3632 -fgcse -fgcse-lm -fgcse-sm -fgcse-las @gol
3633 -fdelete-null-pointer-checks @gol
3634 -fexpensive-optimizations @gol
3636 -fschedule-insns -fschedule-insns2 @gol
3637 -fsched-interblock -fsched-spec @gol
3640 -freorder-blocks -freorder-functions @gol
3641 -fstrict-aliasing @gol
3642 -funit-at-a-time @gol
3643 -falign-functions -falign-jumps @gol
3644 -falign-loops -falign-labels @gol
3647 Please note the warning under @option{-fgcse} about
3648 invoking @option{-O2} on programs that use computed gotos.
3652 Optimize yet more. @option{-O3} turns on all optimizations specified by
3653 @option{-O2} and also turns on the @option{-finline-functions},
3654 @option{-fweb} and @option{-frename-registers} options.
3658 Do not optimize. This is the default.
3662 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
3663 do not typically increase code size. It also performs further
3664 optimizations designed to reduce code size.
3666 @option{-Os} disables the following optimization flags:
3667 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
3668 -falign-labels -freorder-blocks -fprefetch-loop-arrays}
3670 If you use multiple @option{-O} options, with or without level numbers,
3671 the last such option is the one that is effective.
3674 Options of the form @option{-f@var{flag}} specify machine-independent
3675 flags. Most flags have both positive and negative forms; the negative
3676 form of @option{-ffoo} would be @option{-fno-foo}. In the table
3677 below, only one of the forms is listed---the one you typically will
3678 use. You can figure out the other form by either removing @samp{no-}
3681 The following options control specific optimizations. They are either
3682 activated by @option{-O} options or are related to ones that are. You
3683 can use the following flags in the rare cases when ``fine-tuning'' of
3684 optimizations to be performed is desired.
3687 @item -fno-default-inline
3688 @opindex fno-default-inline
3689 Do not make member functions inline by default merely because they are
3690 defined inside the class scope (C++ only). Otherwise, when you specify
3691 @w{@option{-O}}, member functions defined inside class scope are compiled
3692 inline by default; i.e., you don't need to add @samp{inline} in front of
3693 the member function name.
3695 @item -fno-defer-pop
3696 @opindex fno-defer-pop
3697 Always pop the arguments to each function call as soon as that function
3698 returns. For machines which must pop arguments after a function call,
3699 the compiler normally lets arguments accumulate on the stack for several
3700 function calls and pops them all at once.
3702 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3706 Force memory operands to be copied into registers before doing
3707 arithmetic on them. This produces better code by making all memory
3708 references potential common subexpressions. When they are not common
3709 subexpressions, instruction combination should eliminate the separate
3712 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3715 @opindex fforce-addr
3716 Force memory address constants to be copied into registers before
3717 doing arithmetic on them. This may produce better code just as
3718 @option{-fforce-mem} may.
3720 @item -fomit-frame-pointer
3721 @opindex fomit-frame-pointer
3722 Don't keep the frame pointer in a register for functions that
3723 don't need one. This avoids the instructions to save, set up and
3724 restore frame pointers; it also makes an extra register available
3725 in many functions. @strong{It also makes debugging impossible on
3728 On some machines, such as the VAX, this flag has no effect, because
3729 the standard calling sequence automatically handles the frame pointer
3730 and nothing is saved by pretending it doesn't exist. The
3731 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3732 whether a target machine supports this flag. @xref{Registers,,Register
3733 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
3735 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3737 @item -foptimize-sibling-calls
3738 @opindex foptimize-sibling-calls
3739 Optimize sibling and tail recursive calls.
3741 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3745 Don't pay attention to the @code{inline} keyword. Normally this option
3746 is used to keep the compiler from expanding any functions inline.
3747 Note that if you are not optimizing, no functions can be expanded inline.
3749 @item -finline-functions
3750 @opindex finline-functions
3751 Integrate all simple functions into their callers. The compiler
3752 heuristically decides which functions are simple enough to be worth
3753 integrating in this way.
3755 If all calls to a given function are integrated, and the function is
3756 declared @code{static}, then the function is normally not output as
3757 assembler code in its own right.
3759 Enabled at level @option{-O3}.
3761 @item -finline-limit=@var{n}
3762 @opindex finline-limit
3763 By default, gcc limits the size of functions that can be inlined. This flag
3764 allows the control of this limit for functions that are explicitly marked as
3765 inline (i.e., marked with the inline keyword or defined within the class
3766 definition in c++). @var{n} is the size of functions that can be inlined in
3767 number of pseudo instructions (not counting parameter handling). The default
3768 value of @var{n} is 600.
3769 Increasing this value can result in more inlined code at
3770 the cost of compilation time and memory consumption. Decreasing usually makes
3771 the compilation faster and less code will be inlined (which presumably
3772 means slower programs). This option is particularly useful for programs that
3773 use inlining heavily such as those based on recursive templates with C++.
3775 Inlining is actually controlled by a number of parameters, which may be
3776 specified individually by using @option{--param @var{name}=@var{value}}.
3777 The @option{-finline-limit=@var{n}} option sets some of these parameters
3781 @item max-inline-insns-single
3782 is set to @var{n}/2.
3783 @item max-inline-insns-auto
3784 is set to @var{n}/2.
3785 @item min-inline-insns
3786 is set to 130 or @var{n}/4, whichever is smaller.
3787 @item max-inline-insns-rtl
3791 See below for a documentation of the individual
3792 parameters controlling inlining.
3794 @emph{Note:} pseudo instruction represents, in this particular context, an
3795 abstract measurement of function's size. In no way, it represents a count
3796 of assembly instructions and as such its exact meaning might change from one
3797 release to an another.
3799 @item -fkeep-inline-functions
3800 @opindex fkeep-inline-functions
3801 Even if all calls to a given function are integrated, and the function
3802 is declared @code{static}, nevertheless output a separate run-time
3803 callable version of the function. This switch does not affect
3804 @code{extern inline} functions.
3806 @item -fkeep-static-consts
3807 @opindex fkeep-static-consts
3808 Emit variables declared @code{static const} when optimization isn't turned
3809 on, even if the variables aren't referenced.
3811 GCC enables this option by default. If you want to force the compiler to
3812 check if the variable was referenced, regardless of whether or not
3813 optimization is turned on, use the @option{-fno-keep-static-consts} option.
3815 @item -fmerge-constants
3816 Attempt to merge identical constants (string constants and floating point
3817 constants) across compilation units.
3819 This option is the default for optimized compilation if the assembler and
3820 linker support it. Use @option{-fno-merge-constants} to inhibit this
3823 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3825 @item -fmerge-all-constants
3826 Attempt to merge identical constants and identical variables.
3828 This option implies @option{-fmerge-constants}. In addition to
3829 @option{-fmerge-constants} this considers e.g. even constant initialized
3830 arrays or initialized constant variables with integral or floating point
3831 types. Languages like C or C++ require each non-automatic variable to
3832 have distinct location, so using this option will result in non-conforming
3837 Use a graph coloring register allocator. Currently this option is meant
3838 for testing, so we are interested to hear about miscompilations with
3841 @item -fno-branch-count-reg
3842 @opindex fno-branch-count-reg
3843 Do not use ``decrement and branch'' instructions on a count register,
3844 but instead generate a sequence of instructions that decrement a
3845 register, compare it against zero, then branch based upon the result.
3846 This option is only meaningful on architectures that support such
3847 instructions, which include x86, PowerPC, IA-64 and S/390.
3849 The default is @option{-fbranch-count-reg}, enabled when
3850 @option{-fstrength-reduce} is enabled.
3852 @item -fno-function-cse
3853 @opindex fno-function-cse
3854 Do not put function addresses in registers; make each instruction that
3855 calls a constant function contain the function's address explicitly.
3857 This option results in less efficient code, but some strange hacks
3858 that alter the assembler output may be confused by the optimizations
3859 performed when this option is not used.
3861 The default is @option{-ffunction-cse}
3863 @item -fno-zero-initialized-in-bss
3864 @opindex fno-zero-initialized-in-bss
3865 If the target supports a BSS section, GCC by default puts variables that
3866 are initialized to zero into BSS@. This can save space in the resulting
3869 This option turns off this behavior because some programs explicitly
3870 rely on variables going to the data section. E.g., so that the
3871 resulting executable can find the beginning of that section and/or make
3872 assumptions based on that.
3874 The default is @option{-fzero-initialized-in-bss}.
3876 @item -fstrength-reduce
3877 @opindex fstrength-reduce
3878 Perform the optimizations of loop strength reduction and
3879 elimination of iteration variables.
3881 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3883 @item -fthread-jumps
3884 @opindex fthread-jumps
3885 Perform optimizations where we check to see if a jump branches to a
3886 location where another comparison subsumed by the first is found. If
3887 so, the first branch is redirected to either the destination of the
3888 second branch or a point immediately following it, depending on whether
3889 the condition is known to be true or false.
3891 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3893 @item -fcse-follow-jumps
3894 @opindex fcse-follow-jumps
3895 In common subexpression elimination, scan through jump instructions
3896 when the target of the jump is not reached by any other path. For
3897 example, when CSE encounters an @code{if} statement with an
3898 @code{else} clause, CSE will follow the jump when the condition
3901 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3903 @item -fcse-skip-blocks
3904 @opindex fcse-skip-blocks
3905 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
3906 follow jumps which conditionally skip over blocks. When CSE
3907 encounters a simple @code{if} statement with no else clause,
3908 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
3909 body of the @code{if}.
3911 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3913 @item -frerun-cse-after-loop
3914 @opindex frerun-cse-after-loop
3915 Re-run common subexpression elimination after loop optimizations has been
3918 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3920 @item -frerun-loop-opt
3921 @opindex frerun-loop-opt
3922 Run the loop optimizer twice.
3924 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3928 Perform a global common subexpression elimination pass.
3929 This pass also performs global constant and copy propagation.
3931 @emph{Note:} When compiling a program using computed gotos, a GCC
3932 extension, you may get better runtime performance if you disable
3933 the global common subexpression elimination pass by adding
3934 @option{-fno-gcse} to the command line.
3936 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3940 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
3941 attempt to move loads which are only killed by stores into themselves. This
3942 allows a loop containing a load/store sequence to be changed to a load outside
3943 the loop, and a copy/store within the loop.
3945 Enabled by default when gcse is enabled.
3949 When @option{-fgcse-sm} is enabled, a store motion pass is run after
3950 global common subexpression elimination. This pass will attempt to move
3951 stores out of loops. When used in conjunction with @option{-fgcse-lm},
3952 loops containing a load/store sequence can be changed to a load before
3953 the loop and a store after the loop.
3955 Enabled by default when gcse is enabled.
3959 When @option{-fgcse-las} is enabled, the global common subexpression
3960 elimination pass eliminates redundant loads that come after stores to the
3961 same memory location (both partial and full redundancies).
3963 Enabled by default when gcse is enabled.
3965 @item -floop-optimize
3966 @opindex floop-optimize
3967 Perform loop optimizations: move constant expressions out of loops, simplify
3968 exit test conditions and optionally do strength-reduction and loop unrolling as
3971 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3973 @item -fcrossjumping
3974 @opindex crossjumping
3975 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
3976 resulting code may or may not perform better than without cross-jumping.
3978 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3980 @item -fif-conversion
3981 @opindex if-conversion
3982 Attempt to transform conditional jumps into branch-less equivalents. This
3983 include use of conditional moves, min, max, set flags and abs instructions, and
3984 some tricks doable by standard arithmetics. The use of conditional execution
3985 on chips where it is available is controlled by @code{if-conversion2}.
3987 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3989 @item -fif-conversion2
3990 @opindex if-conversion2
3991 Use conditional execution (where available) to transform conditional jumps into
3992 branch-less equivalents.
3994 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3996 @item -fdelete-null-pointer-checks
3997 @opindex fdelete-null-pointer-checks
3998 Use global dataflow analysis to identify and eliminate useless checks
3999 for null pointers. The compiler assumes that dereferencing a null
4000 pointer would have halted the program. If a pointer is checked after
4001 it has already been dereferenced, it cannot be null.
4003 In some environments, this assumption is not true, and programs can
4004 safely dereference null pointers. Use
4005 @option{-fno-delete-null-pointer-checks} to disable this optimization
4006 for programs which depend on that behavior.
4008 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4010 @item -fexpensive-optimizations
4011 @opindex fexpensive-optimizations
4012 Perform a number of minor optimizations that are relatively expensive.
4014 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4016 @item -foptimize-register-move
4018 @opindex foptimize-register-move
4020 Attempt to reassign register numbers in move instructions and as
4021 operands of other simple instructions in order to maximize the amount of
4022 register tying. This is especially helpful on machines with two-operand
4025 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4028 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4030 @item -fdelayed-branch
4031 @opindex fdelayed-branch
4032 If supported for the target machine, attempt to reorder instructions
4033 to exploit instruction slots available after delayed branch
4036 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4038 @item -fschedule-insns
4039 @opindex fschedule-insns
4040 If supported for the target machine, attempt to reorder instructions to
4041 eliminate execution stalls due to required data being unavailable. This
4042 helps machines that have slow floating point or memory load instructions
4043 by allowing other instructions to be issued until the result of the load
4044 or floating point instruction is required.
4046 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4048 @item -fschedule-insns2
4049 @opindex fschedule-insns2
4050 Similar to @option{-fschedule-insns}, but requests an additional pass of
4051 instruction scheduling after register allocation has been done. This is
4052 especially useful on machines with a relatively small number of
4053 registers and where memory load instructions take more than one cycle.
4055 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4057 @item -fno-sched-interblock
4058 @opindex fno-sched-interblock
4059 Don't schedule instructions across basic blocks. This is normally
4060 enabled by default when scheduling before register allocation, i.e.@:
4061 with @option{-fschedule-insns} or at @option{-O2} or higher.
4063 @item -fno-sched-spec
4064 @opindex fno-sched-spec
4065 Don't allow speculative motion of non-load instructions. This is normally
4066 enabled by default when scheduling before register allocation, i.e.@:
4067 with @option{-fschedule-insns} or at @option{-O2} or higher.
4069 @item -fsched-spec-load
4070 @opindex fsched-spec-load
4071 Allow speculative motion of some load instructions. This only makes
4072 sense when scheduling before register allocation, i.e.@: with
4073 @option{-fschedule-insns} or at @option{-O2} or higher.
4075 @item -fsched-spec-load-dangerous
4076 @opindex fsched-spec-load-dangerous
4077 Allow speculative motion of more load instructions. This only makes
4078 sense when scheduling before register allocation, i.e.@: with
4079 @option{-fschedule-insns} or at @option{-O2} or higher.
4081 @item -fsched-stalled-insns=@var{n}
4082 @opindex fsched-stalled-insns
4083 Define how many insns (if any) can be moved prematurely from the queue
4084 of stalled insns into the ready list, during the second scheduling pass.
4086 @item -fsched-stalled-insns-dep=@var{n}
4087 @opindex fsched-stalled-insns-dep
4088 Define how many insn groups (cycles) will be examined for a dependency
4089 on a stalled insn that is candidate for premature removal from the queue
4090 of stalled insns. Has an effect only during the second scheduling pass,
4091 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4093 @item -fsched2-use-superblocks
4094 @opindex fsched2-use-superblocks
4095 When scheduling after register allocation, do use superblock scheduling
4096 algorithm. Superblock scheduling allows motion across basic block boundaries
4097 resulting on faster schedules. This option is experimental, as not all machine
4098 descriptions used by GCC model the CPU closely enough to avoid unreliable
4099 results from the algorithm.
4101 This only makes sense when scheduling after register allocation, i.e.@: with
4102 @option{-fschedule-insns2} or at @option{-O2} or higher.
4104 @item -fsched2-use-traces
4105 @opindex fsched2-use-traces
4106 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4107 allocation and additionally perform code duplication in order to increase the
4108 size of superblocks using tracer pass. See @option{-ftracer} for details on
4111 This mode should produce faster but significantly longer programs. Also
4112 without @code{-fbranch-probabilities} the traces constructed may not match the
4113 reality and hurt the performance. This only makes
4114 sense when scheduling after register allocation, i.e.@: with
4115 @option{-fschedule-insns2} or at @option{-O2} or higher.
4117 @item -fcaller-saves
4118 @opindex fcaller-saves
4119 Enable values to be allocated in registers that will be clobbered by
4120 function calls, by emitting extra instructions to save and restore the
4121 registers around such calls. Such allocation is done only when it
4122 seems to result in better code than would otherwise be produced.
4124 This option is always enabled by default on certain machines, usually
4125 those which have no call-preserved registers to use instead.
4127 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4129 @item -fmove-all-movables
4130 @opindex fmove-all-movables
4131 Forces all invariant computations in loops to be moved
4134 @item -freduce-all-givs
4135 @opindex freduce-all-givs
4136 Forces all general-induction variables in loops to be
4139 @emph{Note:} When compiling programs written in Fortran,
4140 @option{-fmove-all-movables} and @option{-freduce-all-givs} are enabled
4141 by default when you use the optimizer.
4143 These options may generate better or worse code; results are highly
4144 dependent on the structure of loops within the source code.
4146 These two options are intended to be removed someday, once
4147 they have helped determine the efficacy of various
4148 approaches to improving loop optimizations.
4150 Please let us (@w{@email{gcc@@gcc.gnu.org}} and @w{@email{fortran@@gnu.org}})
4151 know how use of these options affects
4152 the performance of your production code.
4153 We're very interested in code that runs @emph{slower}
4154 when these options are @emph{enabled}.
4157 @itemx -fno-peephole2
4158 @opindex fno-peephole
4159 @opindex fno-peephole2
4160 Disable any machine-specific peephole optimizations. The difference
4161 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4162 are implemented in the compiler; some targets use one, some use the
4163 other, a few use both.
4165 @option{-fpeephole} is enabled by default.
4166 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4168 @item -fno-guess-branch-probability
4169 @opindex fno-guess-branch-probability
4170 Do not guess branch probabilities using a randomized model.
4172 Sometimes gcc will opt to use a randomized model to guess branch
4173 probabilities, when none are available from either profiling feedback
4174 (@option{-fprofile-arcs}) or @samp{__builtin_expect}. This means that
4175 different runs of the compiler on the same program may produce different
4178 In a hard real-time system, people don't want different runs of the
4179 compiler to produce code that has different behavior; minimizing
4180 non-determinism is of paramount import. This switch allows users to
4181 reduce non-determinism, possibly at the expense of inferior
4184 The default is @option{-fguess-branch-probability} at levels
4185 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4187 @item -freorder-blocks
4188 @opindex freorder-blocks
4189 Reorder basic blocks in the compiled function in order to reduce number of
4190 taken branches and improve code locality.
4192 Enabled at levels @option{-O2}, @option{-O3}.
4194 @item -freorder-functions
4195 @opindex freorder-functions
4196 Reorder basic blocks in the compiled function in order to reduce number of
4197 taken branches and improve code locality. This is implemented by using special
4198 subsections @code{text.hot} for most frequently executed functions and
4199 @code{text.unlikely} for unlikely executed functions. Reordering is done by
4200 the linker so object file format must support named sections and linker must
4201 place them in a reasonable way.
4203 Also profile feedback must be available in to make this option effective. See
4204 @option{-fprofile-arcs} for details.
4206 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4208 @item -fstrict-aliasing
4209 @opindex fstrict-aliasing
4210 Allows the compiler to assume the strictest aliasing rules applicable to
4211 the language being compiled. For C (and C++), this activates
4212 optimizations based on the type of expressions. In particular, an
4213 object of one type is assumed never to reside at the same address as an
4214 object of a different type, unless the types are almost the same. For
4215 example, an @code{unsigned int} can alias an @code{int}, but not a
4216 @code{void*} or a @code{double}. A character type may alias any other
4219 Pay special attention to code like this:
4232 The practice of reading from a different union member than the one most
4233 recently written to (called ``type-punning'') is common. Even with
4234 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4235 is accessed through the union type. So, the code above will work as
4236 expected. However, this code might not:
4247 Every language that wishes to perform language-specific alias analysis
4248 should define a function that computes, given an @code{tree}
4249 node, an alias set for the node. Nodes in different alias sets are not
4250 allowed to alias. For an example, see the C front-end function
4251 @code{c_get_alias_set}.
4253 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4255 @item -falign-functions
4256 @itemx -falign-functions=@var{n}
4257 @opindex falign-functions
4258 Align the start of functions to the next power-of-two greater than
4259 @var{n}, skipping up to @var{n} bytes. For instance,
4260 @option{-falign-functions=32} aligns functions to the next 32-byte
4261 boundary, but @option{-falign-functions=24} would align to the next
4262 32-byte boundary only if this can be done by skipping 23 bytes or less.
4264 @option{-fno-align-functions} and @option{-falign-functions=1} are
4265 equivalent and mean that functions will not be aligned.
4267 Some assemblers only support this flag when @var{n} is a power of two;
4268 in that case, it is rounded up.
4270 If @var{n} is not specified or is zero, use a machine-dependent default.
4272 Enabled at levels @option{-O2}, @option{-O3}.
4274 @item -falign-labels
4275 @itemx -falign-labels=@var{n}
4276 @opindex falign-labels
4277 Align all branch targets to a power-of-two boundary, skipping up to
4278 @var{n} bytes like @option{-falign-functions}. This option can easily
4279 make code slower, because it must insert dummy operations for when the
4280 branch target is reached in the usual flow of the code.
4282 @option{-fno-align-labels} and @option{-falign-labels=1} are
4283 equivalent and mean that labels will not be aligned.
4285 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
4286 are greater than this value, then their values are used instead.
4288 If @var{n} is not specified or is zero, use a machine-dependent default
4289 which is very likely to be @samp{1}, meaning no alignment.
4291 Enabled at levels @option{-O2}, @option{-O3}.
4294 @itemx -falign-loops=@var{n}
4295 @opindex falign-loops
4296 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
4297 like @option{-falign-functions}. The hope is that the loop will be
4298 executed many times, which will make up for any execution of the dummy
4301 @option{-fno-align-loops} and @option{-falign-loops=1} are
4302 equivalent and mean that loops will not be aligned.
4304 If @var{n} is not specified or is zero, use a machine-dependent default.
4306 Enabled at levels @option{-O2}, @option{-O3}.
4309 @itemx -falign-jumps=@var{n}
4310 @opindex falign-jumps
4311 Align branch targets to a power-of-two boundary, for branch targets
4312 where the targets can only be reached by jumping, skipping up to @var{n}
4313 bytes like @option{-falign-functions}. In this case, no dummy operations
4316 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
4317 equivalent and mean that loops will not be aligned.
4319 If @var{n} is not specified or is zero, use a machine-dependent default.
4321 Enabled at levels @option{-O2}, @option{-O3}.
4323 @item -frename-registers
4324 @opindex frename-registers
4325 Attempt to avoid false dependencies in scheduled code by making use
4326 of registers left over after register allocation. This optimization
4327 will most benefit processors with lots of registers. It can, however,
4328 make debugging impossible, since variables will no longer stay in
4329 a ``home register''.
4333 Constructs webs as commonly used for register allocation purposes and assign
4334 each web individual pseudo register. This allows our register allocation pass
4335 to operate on pseudos directly, but also strengthens several other optimization
4336 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
4337 however, make debugging impossible, since variables will no longer stay in a
4340 Enabled at levels @option{-O3}.
4342 @item -fno-cprop-registers
4343 @opindex fno-cprop-registers
4344 After register allocation and post-register allocation instruction splitting,
4345 we perform a copy-propagation pass to try to reduce scheduling dependencies
4346 and occasionally eliminate the copy.
4348 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4350 @item -fprofile-generate
4351 @opindex fprofile-generate
4352 Enable options usually used for instrumenting application to produce profile usefull
4353 for later recompilation profile feedback based optimization.
4355 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}
4358 @opindex fprofile-use
4359 Enable profile feedback directed optimizations, and optimizations
4360 generally profitable only with profile feedback available.
4362 The following options are enabled: @code{-fbranch-probabilities},
4363 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
4367 The following options control compiler behavior regarding floating
4368 point arithmetic. These options trade off between speed and
4369 correctness. All must be specifically enabled.
4373 @opindex ffloat-store
4374 Do not store floating point variables in registers, and inhibit other
4375 options that might change whether a floating point value is taken from a
4378 @cindex floating point precision
4379 This option prevents undesirable excess precision on machines such as
4380 the 68000 where the floating registers (of the 68881) keep more
4381 precision than a @code{double} is supposed to have. Similarly for the
4382 x86 architecture. For most programs, the excess precision does only
4383 good, but a few programs rely on the precise definition of IEEE floating
4384 point. Use @option{-ffloat-store} for such programs, after modifying
4385 them to store all pertinent intermediate computations into variables.
4389 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
4390 @option{-fno-trapping-math}, @option{-ffinite-math-only},
4391 @option{-fno-rounding-math} and @option{-fno-signaling-nans}.
4393 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
4395 This option should never be turned on by any @option{-O} option since
4396 it can result in incorrect output for programs which depend on
4397 an exact implementation of IEEE or ISO rules/specifications for
4400 @item -fno-math-errno
4401 @opindex fno-math-errno
4402 Do not set ERRNO after calling math functions that are executed
4403 with a single instruction, e.g., sqrt. A program that relies on
4404 IEEE exceptions for math error handling may want to use this flag
4405 for speed while maintaining IEEE arithmetic compatibility.
4407 This option should never be turned on by any @option{-O} option since
4408 it can result in incorrect output for programs which depend on
4409 an exact implementation of IEEE or ISO rules/specifications for
4412 The default is @option{-fmath-errno}.
4414 @item -funsafe-math-optimizations
4415 @opindex funsafe-math-optimizations
4416 Allow optimizations for floating-point arithmetic that (a) assume
4417 that arguments and results are valid and (b) may violate IEEE or
4418 ANSI standards. When used at link-time, it may include libraries
4419 or startup files that change the default FPU control word or other
4420 similar optimizations.
4422 This option should never be turned on by any @option{-O} option since
4423 it can result in incorrect output for programs which depend on
4424 an exact implementation of IEEE or ISO rules/specifications for
4427 The default is @option{-fno-unsafe-math-optimizations}.
4429 @item -ffinite-math-only
4430 @opindex ffinite-math-only
4431 Allow optimizations for floating-point arithmetic that assume
4432 that arguments and results are not NaNs or +-Infs.
4434 This option should never be turned on by any @option{-O} option since
4435 it can result in incorrect output for programs which depend on
4436 an exact implementation of IEEE or ISO rules/specifications.
4438 The default is @option{-fno-finite-math-only}.
4440 @item -fno-trapping-math
4441 @opindex fno-trapping-math
4442 Compile code assuming that floating-point operations cannot generate
4443 user-visible traps. These traps include division by zero, overflow,
4444 underflow, inexact result and invalid operation. This option implies
4445 @option{-fno-signaling-nans}. Setting this option may allow faster
4446 code if one relies on ``non-stop'' IEEE arithmetic, for example.
4448 This option should never be turned on by any @option{-O} option since
4449 it can result in incorrect output for programs which depend on
4450 an exact implementation of IEEE or ISO rules/specifications for
4453 The default is @option{-ftrapping-math}.
4455 @item -frounding-math
4456 @opindex frounding-math
4457 Disable transformations and optimizations that assume default floating
4458 point rounding behavior. This is round-to-zero for all floating point
4459 to integer conversions, and round-to-nearest for all other arithmetic
4460 truncations. This option should be specified for programs that change
4461 the FP rounding mode dynamically, or that may be executed with a
4462 non-default rounding mode. This option disables constant folding of
4463 floating point expressions at compile-time (which may be affected by
4464 rounding mode) and arithmetic transformations that are unsafe in the
4465 presence of sign-dependent rounding modes.
4467 The default is @option{-fno-rounding-math}.
4469 This option is experimental and does not currently guarantee to
4470 disable all GCC optimizations that are affected by rounding mode.
4471 Future versions of gcc may provide finer control of this setting
4472 using C99's @code{FENV_ACCESS} pragma. This command line option
4473 will be used to specify the default state for @code{FENV_ACCESS}.
4475 @item -fsignaling-nans
4476 @opindex fsignaling-nans
4477 Compile code assuming that IEEE signaling NaNs may generate user-visible
4478 traps during floating-point operations. Setting this option disables
4479 optimizations that may change the number of exceptions visible with
4480 signaling NaNs. This option implies @option{-ftrapping-math}.
4482 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
4485 The default is @option{-fno-signaling-nans}.
4487 This option is experimental and does not currently guarantee to
4488 disable all GCC optimizations that affect signaling NaN behavior.
4490 @item -fsingle-precision-constant
4491 @opindex fsingle-precision-constant
4492 Treat floating point constant as single precision constant instead of
4493 implicitly converting it to double precision constant.
4498 The following options control optimizations that may improve
4499 performance, but are not enabled by any @option{-O} options. This
4500 section includes experimental options that may produce broken code.
4503 @item -fbranch-probabilities
4504 @opindex fbranch-probabilities
4505 After running a program compiled with @option{-fprofile-arcs}
4506 (@pxref{Debugging Options,, Options for Debugging Your Program or
4507 @command{gcc}}), you can compile it a second time using
4508 @option{-fbranch-probabilities}, to improve optimizations based on
4509 the number of times each branch was taken. When the program
4510 compiled with @option{-fprofile-arcs} exits it saves arc execution
4511 counts to a file called @file{@var{sourcename}.gcda} for each source
4512 file The information in this data file is very dependent on the
4513 structure of the generated code, so you must use the same source code
4514 and the same optimization options for both compilations.
4516 With @option{-fbranch-probabilities}, GCC puts a
4517 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
4518 These can be used to improve optimization. Currently, they are only
4519 used in one place: in @file{reorg.c}, instead of guessing which path a
4520 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
4521 exactly determine which path is taken more often.
4523 @item -fprofile-values
4524 @opindex fprofile-values
4525 If combined with @option{-fprofile-arcs}, it adds code so that some
4526 data about values of expressions in the program is gathered.
4528 With @option{-fbranch-probabilities}, it reads back the data gathered
4529 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
4530 notes to instructions for their later usage in optimizations.
4534 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
4535 a code to gather information about values of expressions.
4537 With @option{-fbranch-probabilities}, it reads back the data gathered
4538 and actually performs the optimizations based on them.
4539 Currently the optimizations include specialization of division operation
4540 using the knowledge about the value of the denominator.
4544 Use a graph coloring register allocator. Currently this option is meant
4545 for testing, so we are interested to hear about miscompilations with
4550 Perform tail duplication to enlarge superblock size. This transformation
4551 simplifies the control flow of the function allowing other optimizations to do
4554 @item -funit-at-a-time
4555 @opindex funit-at-a-time
4556 Parse the whole compilation unit before starting to produce code.
4557 This allows some extra optimizations to take place but consumes more
4560 @item -funroll-loops
4561 @opindex funroll-loops
4562 Unroll loops whose number of iterations can be determined at compile time or
4563 upon entry to the loop. @option{-funroll-loops} implies
4564 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
4565 (i.e. complete removal of loops with small constant number of iterations).
4566 This option makes code larger, and may or may not make it run faster.
4568 @item -funroll-all-loops
4569 @opindex funroll-all-loops
4570 Unroll all loops, even if their number of iterations is uncertain when
4571 the loop is entered. This usually makes programs run more slowly.
4572 @option{-funroll-all-loops} implies the same options as
4573 @option{-funroll-loops}.
4576 @opindex fpeel-loops
4577 Peels the loops for that there is enough information that they do not
4578 roll much (from profile feedback). It also turns on complete loop peeling
4579 (i.e. complete removal of loops with small constant number of iterations).
4581 @item -funswitch-loops
4582 @opindex funswitch-loops
4583 Move branches with loop invariant conditions out of the loop, with duplicates
4584 of the loop on both branches (modified according to result of the condition).
4586 @item -fold-unroll-loops
4587 @opindex fold-unroll-loops
4588 Unroll loops whose number of iterations can be determined at compile
4589 time or upon entry to the loop, using the old loop unroller whose loop
4590 recognition is based on notes from frontend. @option{-fold-unroll-loops} implies
4591 both @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4592 option makes code larger, and may or may not make it run faster.
4594 @item -fold-unroll-all-loops
4595 @opindex fold-unroll-all-loops
4596 Unroll all loops, even if their number of iterations is uncertain when
4597 the loop is entered. This is done using the old loop unroller whose loop
4598 recognition is based on notes from frontend. This usually makes programs run more slowly.
4599 @option{-fold-unroll-all-loops} implies the same options as
4600 @option{-fold-unroll-loops}.
4602 @item -funswitch-loops
4603 @opindex funswitch-loops
4604 Move branches with loop invariant conditions out of the loop, with duplicates
4605 of the loop on both branches (modified according to result of the condition).
4607 @item -funswitch-loops
4608 @opindex funswitch-loops
4609 Move branches with loop invariant conditions out of the loop, with duplicates
4610 of the loop on both branches (modified according to result of the condition).
4612 @item -fprefetch-loop-arrays
4613 @opindex fprefetch-loop-arrays
4614 If supported by the target machine, generate instructions to prefetch
4615 memory to improve the performance of loops that access large arrays.
4617 Disabled at level @option{-Os}.
4619 @item -ffunction-sections
4620 @itemx -fdata-sections
4621 @opindex ffunction-sections
4622 @opindex fdata-sections
4623 Place each function or data item into its own section in the output
4624 file if the target supports arbitrary sections. The name of the
4625 function or the name of the data item determines the section's name
4628 Use these options on systems where the linker can perform optimizations
4629 to improve locality of reference in the instruction space. Most systems
4630 using the ELF object format and SPARC processors running Solaris 2 have
4631 linkers with such optimizations. AIX may have these optimizations in
4634 Only use these options when there are significant benefits from doing
4635 so. When you specify these options, the assembler and linker will
4636 create larger object and executable files and will also be slower.
4637 You will not be able to use @code{gprof} on all systems if you
4638 specify this option and you may have problems with debugging if
4639 you specify both this option and @option{-g}.
4641 @item -fbranch-target-load-optimize
4642 @opindex fbranch-target-load-optimize
4643 Perform branch target register load optimization before prologue / epilogue
4645 The use of target registers can typically be exposed only during reload,
4646 thus hoisting loads out of loops and doing inter-block scheduling needs
4647 a separate optimization pass.
4649 @item -fbranch-target-load-optimize2
4650 @opindex fbranch-target-load-optimize2
4651 Perform branch target register load optimization after prologue / epilogue
4654 @item --param @var{name}=@var{value}
4656 In some places, GCC uses various constants to control the amount of
4657 optimization that is done. For example, GCC will not inline functions
4658 that contain more that a certain number of instructions. You can
4659 control some of these constants on the command-line using the
4660 @option{--param} option.
4662 In each case, the @var{value} is an integer. The allowable choices for
4663 @var{name} are given in the following table:
4666 @item max-crossjump-edges
4667 The maximum number of incoming edges to consider for crossjumping.
4668 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
4669 the number of edges incoming to each block. Increasing values mean
4670 more aggressive optimization, making the compile time increase with
4671 probably small improvement in executable size.
4673 @item max-delay-slot-insn-search
4674 The maximum number of instructions to consider when looking for an
4675 instruction to fill a delay slot. If more than this arbitrary number of
4676 instructions is searched, the time savings from filling the delay slot
4677 will be minimal so stop searching. Increasing values mean more
4678 aggressive optimization, making the compile time increase with probably
4679 small improvement in executable run time.
4681 @item max-delay-slot-live-search
4682 When trying to fill delay slots, the maximum number of instructions to
4683 consider when searching for a block with valid live register
4684 information. Increasing this arbitrarily chosen value means more
4685 aggressive optimization, increasing the compile time. This parameter
4686 should be removed when the delay slot code is rewritten to maintain the
4689 @item max-gcse-memory
4690 The approximate maximum amount of memory that will be allocated in
4691 order to perform the global common subexpression elimination
4692 optimization. If more memory than specified is required, the
4693 optimization will not be done.
4695 @item max-gcse-passes
4696 The maximum number of passes of GCSE to run.
4698 @item max-pending-list-length
4699 The maximum number of pending dependencies scheduling will allow
4700 before flushing the current state and starting over. Large functions
4701 with few branches or calls can create excessively large lists which
4702 needlessly consume memory and resources.
4704 @item max-inline-insns-single
4705 Several parameters control the tree inliner used in gcc.
4706 This number sets the maximum number of instructions (counted in gcc's
4707 internal representation) in a single function that the tree inliner
4708 will consider for inlining. This only affects functions declared
4709 inline and methods implemented in a class declaration (C++).
4710 The default value is 500.
4712 @item max-inline-insns-auto
4713 When you use @option{-finline-functions} (included in @option{-O3}),
4714 a lot of functions that would otherwise not be considered for inlining
4715 by the compiler will be investigated. To those functions, a different
4716 (more restrictive) limit compared to functions declared inline can
4718 The default value is 150.
4720 @item large-function-insns
4721 The limit specifying really large functions. For functions greater than this
4722 limit inlining is constrained by @option{--param large-function-growth}.
4723 This parameter is useful primarily to avoid extreme compilation time caused by non-linear
4724 algorithms used by the backend.
4725 This parameter is ignored when @option{-funit-at-a-time} is not used.
4726 The default value is 30000.
4728 @item large-function-growth
4729 Specifies maximal growth of large functtion caused by inlining in percents.
4730 This parameter is ignored when @option{-funit-at-a-time} is not used.
4731 The default value is 200.
4733 @item inline-unit-growth
4734 Specifies maximal overall growth of the compilation unit caused by inlining.
4735 This parameter is ignored when @option{-funit-at-a-time} is not used.
4736 The default value is 150.
4738 @item max-inline-insns-rtl
4739 For languages that use the RTL inliner (this happens at a later stage
4740 than tree inlining), you can set the maximum allowable size (counted
4741 in RTL instructions) for the RTL inliner with this parameter.
4742 The default value is 600.
4744 @item max-unrolled-insns
4745 The maximum number of instructions that a loop should have if that loop
4746 is unrolled, and if the loop is unrolled, it determines how many times
4747 the loop code is unrolled.
4749 @item max-average-unrolled-insns
4750 The maximum number of instructions biased by probabilities of their execution
4751 that a loop should have if that loop is unrolled, and if the loop is unrolled,
4752 it determines how many times the loop code is unrolled.
4754 @item max-unroll-times
4755 The maximum number of unrollings of a single loop.
4757 @item max-peeled-insns
4758 The maximum number of instructions that a loop should have if that loop
4759 is peeled, and if the loop is peeled, it determines how many times
4760 the loop code is peeled.
4762 @item max-peel-times
4763 The maximum number of peelings of a single loop.
4765 @item max-completely-peeled-insns
4766 The maximum number of insns of a completely peeled loop.
4768 @item max-completely-peel-times
4769 The maximum number of iterations of a loop to be suitable for complete peeling.
4771 @item max-unswitch-insns
4772 The maximum number of insns of an unswitched loop.
4774 @item max-unswitch-level
4775 The maximum number of branches unswitched in a single loop.
4777 @item hot-bb-count-fraction
4778 Select fraction of the maximal count of repetitions of basic block in program
4779 given basic block needs to have to be considered hot.
4781 @item hot-bb-frequency-fraction
4782 Select fraction of the maximal frequency of executions of basic block in
4783 function given basic block needs to have to be considered hot
4785 @item tracer-dynamic-coverage
4786 @itemx tracer-dynamic-coverage-feedback
4788 This value is used to limit superblock formation once the given percentage of
4789 executed instructions is covered. This limits unnecessary code size
4792 The @option{tracer-dynamic-coverage-feedback} is used only when profile
4793 feedback is available. The real profiles (as opposed to statically estimated
4794 ones) are much less balanced allowing the threshold to be larger value.
4796 @item tracer-max-code-growth
4797 Stop tail duplication once code growth has reached given percentage. This is
4798 rather hokey argument, as most of the duplicates will be eliminated later in
4799 cross jumping, so it may be set to much higher values than is the desired code
4802 @item tracer-min-branch-ratio
4804 Stop reverse growth when the reverse probability of best edge is less than this
4805 threshold (in percent).
4807 @item tracer-min-branch-ratio
4808 @itemx tracer-min-branch-ratio-feedback
4810 Stop forward growth if the best edge do have probability lower than this
4813 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
4814 compilation for profile feedback and one for compilation without. The value
4815 for compilation with profile feedback needs to be more conservative (higher) in
4816 order to make tracer effective.
4818 @item max-cse-path-length
4820 Maximum number of basic blocks on path that cse considers.
4822 @item ggc-min-expand
4824 GCC uses a garbage collector to manage its own memory allocation. This
4825 parameter specifies the minimum percentage by which the garbage
4826 collector's heap should be allowed to expand between collections.
4827 Tuning this may improve compilation speed; it has no effect on code
4830 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
4831 RAM >= 1GB. If @code{getrlimit} is available, the notion of "RAM" is
4832 the smallest of actual RAM, RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If
4833 GCC is not able to calculate RAM on a particular platform, the lower
4834 bound of 30% is used. Setting this parameter and
4835 @option{ggc-min-heapsize} to zero causes a full collection to occur at
4836 every opportunity. This is extremely slow, but can be useful for
4839 @item ggc-min-heapsize
4841 Minimum size of the garbage collector's heap before it begins bothering
4842 to collect garbage. The first collection occurs after the heap expands
4843 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
4844 tuning this may improve compilation speed, and has no effect on code
4847 The default is RAM/8, with a lower bound of 4096 (four megabytes) and an
4848 upper bound of 131072 (128 megabytes). If @code{getrlimit} is
4849 available, the notion of "RAM" is the smallest of actual RAM,
4850 RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If GCC is not able to calculate
4851 RAM on a particular platform, the lower bound is used. Setting this
4852 parameter very large effectively disables garbage collection. Setting
4853 this parameter and @option{ggc-min-expand} to zero causes a full
4854 collection to occur at every opportunity.
4857 @item max-reload-search-insns
4858 The maximum number of instruction reload should look backward for equivalent
4859 register. Increasing values mean more aggressive optimization, making the
4860 compile time increase with probably slightly better performance. The default
4864 @item reorder-blocks-duplicate
4865 @itemx reorder-blocks-duplicate-feedback
4867 Used by basic block reordering pass to decide whether to use unconditional
4868 branch or duplicate the code on its destination. Code is duplicated when its
4869 estimated size is smaller than this value multiplied by the estimated size of
4870 unconditional jump in the hot spots of the program.
4872 The @option{reorder-block-duplicate-feedback} is used only when profile
4873 feedback is available and may be set to higher values than
4874 @option{reorder-block-duplicate} since information about the hot spots is more
4879 @node Preprocessor Options
4880 @section Options Controlling the Preprocessor
4881 @cindex preprocessor options
4882 @cindex options, preprocessor
4884 These options control the C preprocessor, which is run on each C source
4885 file before actual compilation.
4887 If you use the @option{-E} option, nothing is done except preprocessing.
4888 Some of these options make sense only together with @option{-E} because
4889 they cause the preprocessor output to be unsuitable for actual
4894 You can use @option{-Wp,@var{option}} to bypass the compiler driver
4895 and pass @var{option} directly through to the preprocessor. If
4896 @var{option} contains commas, it is split into multiple options at the
4897 commas. However, many options are modified, translated or interpreted
4898 by the compiler driver before being passed to the preprocessor, and
4899 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
4900 interface is undocumented and subject to change, so whenever possible
4901 you should avoid using @option{-Wp} and let the driver handle the
4904 @item -Xpreprocessor @var{option}
4905 @opindex preprocessor
4906 Pass @var{option} as an option to the preprocessor. You can use this to
4907 supply system-specific preprocessor options which GCC does not know how to
4910 If you want to pass an option that takes an argument, you must use
4911 @option{-Xpreprocessor} twice, once for the option and once for the argument.
4914 @include cppopts.texi
4916 @node Assembler Options
4917 @section Passing Options to the Assembler
4919 @c prevent bad page break with this line
4920 You can pass options to the assembler.
4923 @item -Wa,@var{option}
4925 Pass @var{option} as an option to the assembler. If @var{option}
4926 contains commas, it is split into multiple options at the commas.
4928 @item -Xassembler @var{option}
4930 Pass @var{option} as an option to the assembler. You can use this to
4931 supply system-specific assembler options which GCC does not know how to
4934 If you want to pass an option that takes an argument, you must use
4935 @option{-Xassembler} twice, once for the option and once for the argument.
4940 @section Options for Linking
4941 @cindex link options
4942 @cindex options, linking
4944 These options come into play when the compiler links object files into
4945 an executable output file. They are meaningless if the compiler is
4946 not doing a link step.
4950 @item @var{object-file-name}
4951 A file name that does not end in a special recognized suffix is
4952 considered to name an object file or library. (Object files are
4953 distinguished from libraries by the linker according to the file
4954 contents.) If linking is done, these object files are used as input
4963 If any of these options is used, then the linker is not run, and
4964 object file names should not be used as arguments. @xref{Overall
4968 @item -l@var{library}
4969 @itemx -l @var{library}
4971 Search the library named @var{library} when linking. (The second
4972 alternative with the library as a separate argument is only for
4973 POSIX compliance and is not recommended.)
4975 It makes a difference where in the command you write this option; the
4976 linker searches and processes libraries and object files in the order they
4977 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
4978 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
4979 to functions in @samp{z}, those functions may not be loaded.
4981 The linker searches a standard list of directories for the library,
4982 which is actually a file named @file{lib@var{library}.a}. The linker
4983 then uses this file as if it had been specified precisely by name.
4985 The directories searched include several standard system directories
4986 plus any that you specify with @option{-L}.
4988 Normally the files found this way are library files---archive files
4989 whose members are object files. The linker handles an archive file by
4990 scanning through it for members which define symbols that have so far
4991 been referenced but not defined. But if the file that is found is an
4992 ordinary object file, it is linked in the usual fashion. The only
4993 difference between using an @option{-l} option and specifying a file name
4994 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
4995 and searches several directories.
4999 You need this special case of the @option{-l} option in order to
5000 link an Objective-C program.
5003 @opindex nostartfiles
5004 Do not use the standard system startup files when linking.
5005 The standard system libraries are used normally, unless @option{-nostdlib}
5006 or @option{-nodefaultlibs} is used.
5008 @item -nodefaultlibs
5009 @opindex nodefaultlibs
5010 Do not use the standard system libraries when linking.
5011 Only the libraries you specify will be passed to the linker.
5012 The standard startup files are used normally, unless @option{-nostartfiles}
5013 is used. The compiler may generate calls to memcmp, memset, and memcpy
5014 for System V (and ISO C) environments or to bcopy and bzero for
5015 BSD environments. These entries are usually resolved by entries in
5016 libc. These entry points should be supplied through some other
5017 mechanism when this option is specified.
5021 Do not use the standard system startup files or libraries when linking.
5022 No startup files and only the libraries you specify will be passed to
5023 the linker. The compiler may generate calls to memcmp, memset, and memcpy
5024 for System V (and ISO C) environments or to bcopy and bzero for
5025 BSD environments. These entries are usually resolved by entries in
5026 libc. These entry points should be supplied through some other
5027 mechanism when this option is specified.
5029 @cindex @option{-lgcc}, use with @option{-nostdlib}
5030 @cindex @option{-nostdlib} and unresolved references
5031 @cindex unresolved references and @option{-nostdlib}
5032 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5033 @cindex @option{-nodefaultlibs} and unresolved references
5034 @cindex unresolved references and @option{-nodefaultlibs}
5035 One of the standard libraries bypassed by @option{-nostdlib} and
5036 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5037 that GCC uses to overcome shortcomings of particular machines, or special
5038 needs for some languages.
5039 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5040 Collection (GCC) Internals},
5041 for more discussion of @file{libgcc.a}.)
5042 In most cases, you need @file{libgcc.a} even when you want to avoid
5043 other standard libraries. In other words, when you specify @option{-nostdlib}
5044 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5045 This ensures that you have no unresolved references to internal GCC
5046 library subroutines. (For example, @samp{__main}, used to ensure C++
5047 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5048 GNU Compiler Collection (GCC) Internals}.)
5052 Produce a position independent executable on targets which support it.
5053 For predictable results, you must also specify the same set of options
5054 that were used to generate code (@option{-fpie}, @option{-fPIE},
5055 or model suboptions) when you specify this option.
5059 Remove all symbol table and relocation information from the executable.
5063 On systems that support dynamic linking, this prevents linking with the shared
5064 libraries. On other systems, this option has no effect.
5068 Produce a shared object which can then be linked with other objects to
5069 form an executable. Not all systems support this option. For predictable
5070 results, you must also specify the same set of options that were used to
5071 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5072 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5073 needs to build supplementary stub code for constructors to work. On
5074 multi-libbed systems, @samp{gcc -shared} must select the correct support
5075 libraries to link against. Failing to supply the correct flags may lead
5076 to subtle defects. Supplying them in cases where they are not necessary
5079 @item -shared-libgcc
5080 @itemx -static-libgcc
5081 @opindex shared-libgcc
5082 @opindex static-libgcc
5083 On systems that provide @file{libgcc} as a shared library, these options
5084 force the use of either the shared or static version respectively.
5085 If no shared version of @file{libgcc} was built when the compiler was
5086 configured, these options have no effect.
5088 There are several situations in which an application should use the
5089 shared @file{libgcc} instead of the static version. The most common
5090 of these is when the application wishes to throw and catch exceptions
5091 across different shared libraries. In that case, each of the libraries
5092 as well as the application itself should use the shared @file{libgcc}.
5094 Therefore, the G++ and GCJ drivers automatically add
5095 @option{-shared-libgcc} whenever you build a shared library or a main
5096 executable, because C++ and Java programs typically use exceptions, so
5097 this is the right thing to do.
5099 If, instead, you use the GCC driver to create shared libraries, you may
5100 find that they will not always be linked with the shared @file{libgcc}.
5101 If GCC finds, at its configuration time, that you have a GNU linker that
5102 does not support option @option{--eh-frame-hdr}, it will link the shared
5103 version of @file{libgcc} into shared libraries by default. Otherwise,
5104 it will take advantage of the linker and optimize away the linking with
5105 the shared version of @file{libgcc}, linking with the static version of
5106 libgcc by default. This allows exceptions to propagate through such
5107 shared libraries, without incurring relocation costs at library load
5110 However, if a library or main executable is supposed to throw or catch
5111 exceptions, you must link it using the G++ or GCJ driver, as appropriate
5112 for the languages used in the program, or using the option
5113 @option{-shared-libgcc}, such that it is linked with the shared
5118 Bind references to global symbols when building a shared object. Warn
5119 about any unresolved references (unless overridden by the link editor
5120 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
5123 @item -Xlinker @var{option}
5125 Pass @var{option} as an option to the linker. You can use this to
5126 supply system-specific linker options which GCC does not know how to
5129 If you want to pass an option that takes an argument, you must use
5130 @option{-Xlinker} twice, once for the option and once for the argument.
5131 For example, to pass @option{-assert definitions}, you must write
5132 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
5133 @option{-Xlinker "-assert definitions"}, because this passes the entire
5134 string as a single argument, which is not what the linker expects.
5136 @item -Wl,@var{option}
5138 Pass @var{option} as an option to the linker. If @var{option} contains
5139 commas, it is split into multiple options at the commas.
5141 @item -u @var{symbol}
5143 Pretend the symbol @var{symbol} is undefined, to force linking of
5144 library modules to define it. You can use @option{-u} multiple times with
5145 different symbols to force loading of additional library modules.
5148 @node Directory Options
5149 @section Options for Directory Search
5150 @cindex directory options
5151 @cindex options, directory search
5154 These options specify directories to search for header files, for
5155 libraries and for parts of the compiler:
5160 Add the directory @var{dir} to the head of the list of directories to be
5161 searched for header files. This can be used to override a system header
5162 file, substituting your own version, since these directories are
5163 searched before the system header file directories. However, you should
5164 not use this option to add directories that contain vendor-supplied
5165 system header files (use @option{-isystem} for that). If you use more than
5166 one @option{-I} option, the directories are scanned in left-to-right
5167 order; the standard system directories come after.
5169 If a standard system include directory, or a directory specified with
5170 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
5171 option will be ignored. The directory will still be searched but as a
5172 system directory at its normal position in the system include chain.
5173 This is to ensure that GCC's procedure to fix buggy system headers and
5174 the ordering for the include_next directive are not inadvertently changed.
5175 If you really need to change the search order for system directories,
5176 use the @option{-nostdinc} and/or @option{-isystem} options.
5180 Any directories you specify with @option{-I} options before the @option{-I-}
5181 option are searched only for the case of @samp{#include "@var{file}"};
5182 they are not searched for @samp{#include <@var{file}>}.
5184 If additional directories are specified with @option{-I} options after
5185 the @option{-I-}, these directories are searched for all @samp{#include}
5186 directives. (Ordinarily @emph{all} @option{-I} directories are used
5189 In addition, the @option{-I-} option inhibits the use of the current
5190 directory (where the current input file came from) as the first search
5191 directory for @samp{#include "@var{file}"}. There is no way to
5192 override this effect of @option{-I-}. With @option{-I.} you can specify
5193 searching the directory which was current when the compiler was
5194 invoked. That is not exactly the same as what the preprocessor does
5195 by default, but it is often satisfactory.
5197 @option{-I-} does not inhibit the use of the standard system directories
5198 for header files. Thus, @option{-I-} and @option{-nostdinc} are
5203 Add directory @var{dir} to the list of directories to be searched
5206 @item -B@var{prefix}
5208 This option specifies where to find the executables, libraries,
5209 include files, and data files of the compiler itself.
5211 The compiler driver program runs one or more of the subprograms
5212 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
5213 @var{prefix} as a prefix for each program it tries to run, both with and
5214 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
5216 For each subprogram to be run, the compiler driver first tries the
5217 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
5218 was not specified, the driver tries two standard prefixes, which are
5219 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
5220 those results in a file name that is found, the unmodified program
5221 name is searched for using the directories specified in your
5222 @env{PATH} environment variable.
5224 The compiler will check to see if the path provided by the @option{-B}
5225 refers to a directory, and if necessary it will add a directory
5226 separator character at the end of the path.
5228 @option{-B} prefixes that effectively specify directory names also apply
5229 to libraries in the linker, because the compiler translates these
5230 options into @option{-L} options for the linker. They also apply to
5231 includes files in the preprocessor, because the compiler translates these
5232 options into @option{-isystem} options for the preprocessor. In this case,
5233 the compiler appends @samp{include} to the prefix.
5235 The run-time support file @file{libgcc.a} can also be searched for using
5236 the @option{-B} prefix, if needed. If it is not found there, the two
5237 standard prefixes above are tried, and that is all. The file is left
5238 out of the link if it is not found by those means.
5240 Another way to specify a prefix much like the @option{-B} prefix is to use
5241 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
5244 As a special kludge, if the path provided by @option{-B} is
5245 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
5246 9, then it will be replaced by @file{[dir/]include}. This is to help
5247 with boot-strapping the compiler.
5249 @item -specs=@var{file}
5251 Process @var{file} after the compiler reads in the standard @file{specs}
5252 file, in order to override the defaults that the @file{gcc} driver
5253 program uses when determining what switches to pass to @file{cc1},
5254 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
5255 @option{-specs=@var{file}} can be specified on the command line, and they
5256 are processed in order, from left to right.
5262 @section Specifying subprocesses and the switches to pass to them
5265 @command{gcc} is a driver program. It performs its job by invoking a
5266 sequence of other programs to do the work of compiling, assembling and
5267 linking. GCC interprets its command-line parameters and uses these to
5268 deduce which programs it should invoke, and which command-line options
5269 it ought to place on their command lines. This behavior is controlled
5270 by @dfn{spec strings}. In most cases there is one spec string for each
5271 program that GCC can invoke, but a few programs have multiple spec
5272 strings to control their behavior. The spec strings built into GCC can
5273 be overridden by using the @option{-specs=} command-line switch to specify
5276 @dfn{Spec files} are plaintext files that are used to construct spec
5277 strings. They consist of a sequence of directives separated by blank
5278 lines. The type of directive is determined by the first non-whitespace
5279 character on the line and it can be one of the following:
5282 @item %@var{command}
5283 Issues a @var{command} to the spec file processor. The commands that can
5287 @item %include <@var{file}>
5289 Search for @var{file} and insert its text at the current point in the
5292 @item %include_noerr <@var{file}>
5293 @cindex %include_noerr
5294 Just like @samp{%include}, but do not generate an error message if the include
5295 file cannot be found.
5297 @item %rename @var{old_name} @var{new_name}
5299 Rename the spec string @var{old_name} to @var{new_name}.
5303 @item *[@var{spec_name}]:
5304 This tells the compiler to create, override or delete the named spec
5305 string. All lines after this directive up to the next directive or
5306 blank line are considered to be the text for the spec string. If this
5307 results in an empty string then the spec will be deleted. (Or, if the
5308 spec did not exist, then nothing will happened.) Otherwise, if the spec
5309 does not currently exist a new spec will be created. If the spec does
5310 exist then its contents will be overridden by the text of this
5311 directive, unless the first character of that text is the @samp{+}
5312 character, in which case the text will be appended to the spec.
5314 @item [@var{suffix}]:
5315 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
5316 and up to the next directive or blank line are considered to make up the
5317 spec string for the indicated suffix. When the compiler encounters an
5318 input file with the named suffix, it will processes the spec string in
5319 order to work out how to compile that file. For example:
5326 This says that any input file whose name ends in @samp{.ZZ} should be
5327 passed to the program @samp{z-compile}, which should be invoked with the
5328 command-line switch @option{-input} and with the result of performing the
5329 @samp{%i} substitution. (See below.)
5331 As an alternative to providing a spec string, the text that follows a
5332 suffix directive can be one of the following:
5335 @item @@@var{language}
5336 This says that the suffix is an alias for a known @var{language}. This is
5337 similar to using the @option{-x} command-line switch to GCC to specify a
5338 language explicitly. For example:
5345 Says that .ZZ files are, in fact, C++ source files.
5348 This causes an error messages saying:
5351 @var{name} compiler not installed on this system.
5355 GCC already has an extensive list of suffixes built into it.
5356 This directive will add an entry to the end of the list of suffixes, but
5357 since the list is searched from the end backwards, it is effectively
5358 possible to override earlier entries using this technique.
5362 GCC has the following spec strings built into it. Spec files can
5363 override these strings or create their own. Note that individual
5364 targets can also add their own spec strings to this list.
5367 asm Options to pass to the assembler
5368 asm_final Options to pass to the assembler post-processor
5369 cpp Options to pass to the C preprocessor
5370 cc1 Options to pass to the C compiler
5371 cc1plus Options to pass to the C++ compiler
5372 endfile Object files to include at the end of the link
5373 link Options to pass to the linker
5374 lib Libraries to include on the command line to the linker
5375 libgcc Decides which GCC support library to pass to the linker
5376 linker Sets the name of the linker
5377 predefines Defines to be passed to the C preprocessor
5378 signed_char Defines to pass to CPP to say whether @code{char} is signed
5380 startfile Object files to include at the start of the link
5383 Here is a small example of a spec file:
5389 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
5392 This example renames the spec called @samp{lib} to @samp{old_lib} and
5393 then overrides the previous definition of @samp{lib} with a new one.
5394 The new definition adds in some extra command-line options before
5395 including the text of the old definition.
5397 @dfn{Spec strings} are a list of command-line options to be passed to their
5398 corresponding program. In addition, the spec strings can contain
5399 @samp{%}-prefixed sequences to substitute variable text or to
5400 conditionally insert text into the command line. Using these constructs
5401 it is possible to generate quite complex command lines.
5403 Here is a table of all defined @samp{%}-sequences for spec
5404 strings. Note that spaces are not generated automatically around the
5405 results of expanding these sequences. Therefore you can concatenate them
5406 together or combine them with constant text in a single argument.
5410 Substitute one @samp{%} into the program name or argument.
5413 Substitute the name of the input file being processed.
5416 Substitute the basename of the input file being processed.
5417 This is the substring up to (and not including) the last period
5418 and not including the directory.
5421 This is the same as @samp{%b}, but include the file suffix (text after
5425 Marks the argument containing or following the @samp{%d} as a
5426 temporary file name, so that that file will be deleted if GCC exits
5427 successfully. Unlike @samp{%g}, this contributes no text to the
5430 @item %g@var{suffix}
5431 Substitute a file name that has suffix @var{suffix} and is chosen
5432 once per compilation, and mark the argument in the same way as
5433 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
5434 name is now chosen in a way that is hard to predict even when previously
5435 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
5436 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
5437 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
5438 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
5439 was simply substituted with a file name chosen once per compilation,
5440 without regard to any appended suffix (which was therefore treated
5441 just like ordinary text), making such attacks more likely to succeed.
5443 @item %u@var{suffix}
5444 Like @samp{%g}, but generates a new temporary file name even if
5445 @samp{%u@var{suffix}} was already seen.
5447 @item %U@var{suffix}
5448 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
5449 new one if there is no such last file name. In the absence of any
5450 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
5451 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
5452 would involve the generation of two distinct file names, one
5453 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
5454 simply substituted with a file name chosen for the previous @samp{%u},
5455 without regard to any appended suffix.
5457 @item %j@var{suffix}
5458 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
5459 writable, and if save-temps is off; otherwise, substitute the name
5460 of a temporary file, just like @samp{%u}. This temporary file is not
5461 meant for communication between processes, but rather as a junk
5464 @item %|@var{suffix}
5465 @itemx %m@var{suffix}
5466 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
5467 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
5468 all. These are the two most common ways to instruct a program that it
5469 should read from standard input or write to standard output. If you
5470 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
5471 construct: see for example @file{f/lang-specs.h}.
5473 @item %.@var{SUFFIX}
5474 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
5475 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
5476 terminated by the next space or %.
5479 Marks the argument containing or following the @samp{%w} as the
5480 designated output file of this compilation. This puts the argument
5481 into the sequence of arguments that @samp{%o} will substitute later.
5484 Substitutes the names of all the output files, with spaces
5485 automatically placed around them. You should write spaces
5486 around the @samp{%o} as well or the results are undefined.
5487 @samp{%o} is for use in the specs for running the linker.
5488 Input files whose names have no recognized suffix are not compiled
5489 at all, but they are included among the output files, so they will
5493 Substitutes the suffix for object files. Note that this is
5494 handled specially when it immediately follows @samp{%g, %u, or %U},
5495 because of the need for those to form complete file names. The
5496 handling is such that @samp{%O} is treated exactly as if it had already
5497 been substituted, except that @samp{%g, %u, and %U} do not currently
5498 support additional @var{suffix} characters following @samp{%O} as they would
5499 following, for example, @samp{.o}.
5502 Substitutes the standard macro predefinitions for the
5503 current target machine. Use this when running @code{cpp}.
5506 Like @samp{%p}, but puts @samp{__} before and after the name of each
5507 predefined macro, except for macros that start with @samp{__} or with
5508 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
5512 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
5513 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
5514 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
5518 Current argument is the name of a library or startup file of some sort.
5519 Search for that file in a standard list of directories and substitute
5520 the full name found.
5523 Print @var{str} as an error message. @var{str} is terminated by a newline.
5524 Use this when inconsistent options are detected.
5527 Substitute the contents of spec string @var{name} at this point.
5530 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
5532 @item %x@{@var{option}@}
5533 Accumulate an option for @samp{%X}.
5536 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
5540 Output the accumulated assembler options specified by @option{-Wa}.
5543 Output the accumulated preprocessor options specified by @option{-Wp}.
5546 Process the @code{asm} spec. This is used to compute the
5547 switches to be passed to the assembler.
5550 Process the @code{asm_final} spec. This is a spec string for
5551 passing switches to an assembler post-processor, if such a program is
5555 Process the @code{link} spec. This is the spec for computing the
5556 command line passed to the linker. Typically it will make use of the
5557 @samp{%L %G %S %D and %E} sequences.
5560 Dump out a @option{-L} option for each directory that GCC believes might
5561 contain startup files. If the target supports multilibs then the
5562 current multilib directory will be prepended to each of these paths.
5565 Output the multilib directory with directory separators replaced with
5566 @samp{_}. If multilib directories are not set, or the multilib directory is
5567 @file{.} then this option emits nothing.
5570 Process the @code{lib} spec. This is a spec string for deciding which
5571 libraries should be included on the command line to the linker.
5574 Process the @code{libgcc} spec. This is a spec string for deciding
5575 which GCC support library should be included on the command line to the linker.
5578 Process the @code{startfile} spec. This is a spec for deciding which
5579 object files should be the first ones passed to the linker. Typically
5580 this might be a file named @file{crt0.o}.
5583 Process the @code{endfile} spec. This is a spec string that specifies
5584 the last object files that will be passed to the linker.
5587 Process the @code{cpp} spec. This is used to construct the arguments
5588 to be passed to the C preprocessor.
5591 Process the @code{signed_char} spec. This is intended to be used
5592 to tell cpp whether a char is signed. It typically has the definition:
5594 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
5598 Process the @code{cc1} spec. This is used to construct the options to be
5599 passed to the actual C compiler (@samp{cc1}).
5602 Process the @code{cc1plus} spec. This is used to construct the options to be
5603 passed to the actual C++ compiler (@samp{cc1plus}).
5606 Substitute the variable part of a matched option. See below.
5607 Note that each comma in the substituted string is replaced by
5611 Remove all occurrences of @code{-S} from the command line. Note---this
5612 command is position dependent. @samp{%} commands in the spec string
5613 before this one will see @code{-S}, @samp{%} commands in the spec string
5614 after this one will not.
5616 @item %:@var{function}(@var{args})
5617 Call the named function @var{function}, passing it @var{args}.
5618 @var{args} is first processed as a nested spec string, then split
5619 into an argument vector in the usual fashion. The function returns
5620 a string which is processed as if it had appeared literally as part
5621 of the current spec.
5623 The following built-in spec functions are provided:
5626 @item @code{if-exists}
5627 The @code{if-exists} spec function takes one argument, an absolute
5628 pathname to a file. If the file exists, @code{if-exists} returns the
5629 pathname. Here is a small example of its usage:
5633 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
5636 @item @code{if-exists-else}
5637 The @code{if-exists-else} spec function is similar to the @code{if-exists}
5638 spec function, except that it takes two arguments. The first argument is
5639 an absolute pathname to a file. If the file exists, @code{if-exists-else}
5640 returns the pathname. If it does not exist, it returns the second argument.
5641 This way, @code{if-exists-else} can be used to select one file or another,
5642 based on the existence of the first. Here is a small example of its usage:
5646 crt0%O%s %:if-exists(crti%O%s) \
5647 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
5652 Substitutes the @code{-S} switch, if that switch was given to GCC@.
5653 If that switch was not specified, this substitutes nothing. Note that
5654 the leading dash is omitted when specifying this option, and it is
5655 automatically inserted if the substitution is performed. Thus the spec
5656 string @samp{%@{foo@}} would match the command-line option @option{-foo}
5657 and would output the command line option @option{-foo}.
5659 @item %W@{@code{S}@}
5660 Like %@{@code{S}@} but mark last argument supplied within as a file to be
5663 @item %@{@code{S}*@}
5664 Substitutes all the switches specified to GCC whose names start
5665 with @code{-S}, but which also take an argument. This is used for
5666 switches like @option{-o}, @option{-D}, @option{-I}, etc.
5667 GCC considers @option{-o foo} as being
5668 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
5669 text, including the space. Thus two arguments would be generated.
5671 @item %@{@code{S}*&@code{T}*@}
5672 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
5673 (the order of @code{S} and @code{T} in the spec is not significant).
5674 There can be any number of ampersand-separated variables; for each the
5675 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
5677 @item %@{@code{S}:@code{X}@}
5678 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
5680 @item %@{!@code{S}:@code{X}@}
5681 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
5683 @item %@{@code{S}*:@code{X}@}
5684 Substitutes @code{X} if one or more switches whose names start with
5685 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
5686 once, no matter how many such switches appeared. However, if @code{%*}
5687 appears somewhere in @code{X}, then @code{X} will be substituted once
5688 for each matching switch, with the @code{%*} replaced by the part of
5689 that switch that matched the @code{*}.
5691 @item %@{.@code{S}:@code{X}@}
5692 Substitutes @code{X}, if processing a file with suffix @code{S}.
5694 @item %@{!.@code{S}:@code{X}@}
5695 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
5697 @item %@{@code{S}|@code{P}:@code{X}@}
5698 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
5699 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
5700 although they have a stronger binding than the @samp{|}. If @code{%*}
5701 appears in @code{X}, all of the alternatives must be starred, and only
5702 the first matching alternative is substituted.
5704 For example, a spec string like this:
5707 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
5710 will output the following command-line options from the following input
5711 command-line options:
5716 -d fred.c -foo -baz -boggle
5717 -d jim.d -bar -baz -boggle
5720 @item %@{S:X; T:Y; :D@}
5722 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
5723 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
5724 be as many clauses as you need. This may be combined with @code{.},
5725 @code{!}, @code{|}, and @code{*} as needed.
5730 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
5731 construct may contain other nested @samp{%} constructs or spaces, or
5732 even newlines. They are processed as usual, as described above.
5733 Trailing white space in @code{X} is ignored. White space may also
5734 appear anywhere on the left side of the colon in these constructs,
5735 except between @code{.} or @code{*} and the corresponding word.
5737 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
5738 handled specifically in these constructs. If another value of
5739 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
5740 @option{-W} switch is found later in the command line, the earlier
5741 switch value is ignored, except with @{@code{S}*@} where @code{S} is
5742 just one letter, which passes all matching options.
5744 The character @samp{|} at the beginning of the predicate text is used to
5745 indicate that a command should be piped to the following command, but
5746 only if @option{-pipe} is specified.
5748 It is built into GCC which switches take arguments and which do not.
5749 (You might think it would be useful to generalize this to allow each
5750 compiler's spec to say which switches take arguments. But this cannot
5751 be done in a consistent fashion. GCC cannot even decide which input
5752 files have been specified without knowing which switches take arguments,
5753 and it must know which input files to compile in order to tell which
5756 GCC also knows implicitly that arguments starting in @option{-l} are to be
5757 treated as compiler output files, and passed to the linker in their
5758 proper position among the other output files.
5760 @c man begin OPTIONS
5762 @node Target Options
5763 @section Specifying Target Machine and Compiler Version
5764 @cindex target options
5765 @cindex cross compiling
5766 @cindex specifying machine version
5767 @cindex specifying compiler version and target machine
5768 @cindex compiler version, specifying
5769 @cindex target machine, specifying
5771 The usual way to run GCC is to run the executable called @file{gcc}, or
5772 @file{<machine>-gcc} when cross-compiling, or
5773 @file{<machine>-gcc-<version>} to run a version other than the one that
5774 was installed last. Sometimes this is inconvenient, so GCC provides
5775 options that will switch to another cross-compiler or version.
5778 @item -b @var{machine}
5780 The argument @var{machine} specifies the target machine for compilation.
5782 The value to use for @var{machine} is the same as was specified as the
5783 machine type when configuring GCC as a cross-compiler. For
5784 example, if a cross-compiler was configured with @samp{configure
5785 i386v}, meaning to compile for an 80386 running System V, then you
5786 would specify @option{-b i386v} to run that cross compiler.
5788 @item -V @var{version}
5790 The argument @var{version} specifies which version of GCC to run.
5791 This is useful when multiple versions are installed. For example,
5792 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
5795 The @option{-V} and @option{-b} options work by running the
5796 @file{<machine>-gcc-<version>} executable, so there's no real reason to
5797 use them if you can just run that directly.
5799 @node Submodel Options
5800 @section Hardware Models and Configurations
5801 @cindex submodel options
5802 @cindex specifying hardware config
5803 @cindex hardware models and configurations, specifying
5804 @cindex machine dependent options
5806 Earlier we discussed the standard option @option{-b} which chooses among
5807 different installed compilers for completely different target
5808 machines, such as VAX vs.@: 68000 vs.@: 80386.
5810 In addition, each of these target machine types can have its own
5811 special options, starting with @samp{-m}, to choose among various
5812 hardware models or configurations---for example, 68010 vs 68020,
5813 floating coprocessor or none. A single installed version of the
5814 compiler can compile for any model or configuration, according to the
5817 Some configurations of the compiler also support additional special
5818 options, usually for compatibility with other compilers on the same
5821 These options are defined by the macro @code{TARGET_SWITCHES} in the
5822 machine description. The default for the options is also defined by
5823 that macro, which enables you to change the defaults.
5833 * RS/6000 and PowerPC Options::
5836 * i386 and x86-64 Options::
5838 * Intel 960 Options::
5839 * DEC Alpha Options::
5840 * DEC Alpha/VMS Options::
5843 * System V Options::
5844 * TMS320C3x/C4x Options::
5852 * S/390 and zSeries Options::
5856 * Xstormy16 Options::
5861 @node M680x0 Options
5862 @subsection M680x0 Options
5863 @cindex M680x0 options
5865 These are the @samp{-m} options defined for the 68000 series. The default
5866 values for these options depends on which style of 68000 was selected when
5867 the compiler was configured; the defaults for the most common choices are
5875 Generate output for a 68000. This is the default
5876 when the compiler is configured for 68000-based systems.
5878 Use this option for microcontrollers with a 68000 or EC000 core,
5879 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
5885 Generate output for a 68020. This is the default
5886 when the compiler is configured for 68020-based systems.
5890 Generate output containing 68881 instructions for floating point.
5891 This is the default for most 68020 systems unless @option{--nfp} was
5892 specified when the compiler was configured.
5896 Generate output for a 68030. This is the default when the compiler is
5897 configured for 68030-based systems.
5901 Generate output for a 68040. This is the default when the compiler is
5902 configured for 68040-based systems.
5904 This option inhibits the use of 68881/68882 instructions that have to be
5905 emulated by software on the 68040. Use this option if your 68040 does not
5906 have code to emulate those instructions.
5910 Generate output for a 68060. This is the default when the compiler is
5911 configured for 68060-based systems.
5913 This option inhibits the use of 68020 and 68881/68882 instructions that
5914 have to be emulated by software on the 68060. Use this option if your 68060
5915 does not have code to emulate those instructions.
5919 Generate output for a CPU32. This is the default
5920 when the compiler is configured for CPU32-based systems.
5922 Use this option for microcontrollers with a
5923 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
5924 68336, 68340, 68341, 68349 and 68360.
5928 Generate output for a 520X ``coldfire'' family cpu. This is the default
5929 when the compiler is configured for 520X-based systems.
5931 Use this option for microcontroller with a 5200 core, including
5932 the MCF5202, MCF5203, MCF5204 and MCF5202.
5937 Generate output for a 68040, without using any of the new instructions.
5938 This results in code which can run relatively efficiently on either a
5939 68020/68881 or a 68030 or a 68040. The generated code does use the
5940 68881 instructions that are emulated on the 68040.
5944 Generate output for a 68060, without using any of the new instructions.
5945 This results in code which can run relatively efficiently on either a
5946 68020/68881 or a 68030 or a 68040. The generated code does use the
5947 68881 instructions that are emulated on the 68060.
5950 @opindex msoft-float
5951 Generate output containing library calls for floating point.
5952 @strong{Warning:} the requisite libraries are not available for all m68k
5953 targets. Normally the facilities of the machine's usual C compiler are
5954 used, but this can't be done directly in cross-compilation. You must
5955 make your own arrangements to provide suitable library functions for
5956 cross-compilation. The embedded targets @samp{m68k-*-aout} and
5957 @samp{m68k-*-coff} do provide software floating point support.
5961 Consider type @code{int} to be 16 bits wide, like @code{short int}.
5964 @opindex mnobitfield
5965 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
5966 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
5970 Do use the bit-field instructions. The @option{-m68020} option implies
5971 @option{-mbitfield}. This is the default if you use a configuration
5972 designed for a 68020.
5976 Use a different function-calling convention, in which functions
5977 that take a fixed number of arguments return with the @code{rtd}
5978 instruction, which pops their arguments while returning. This
5979 saves one instruction in the caller since there is no need to pop
5980 the arguments there.
5982 This calling convention is incompatible with the one normally
5983 used on Unix, so you cannot use it if you need to call libraries
5984 compiled with the Unix compiler.
5986 Also, you must provide function prototypes for all functions that
5987 take variable numbers of arguments (including @code{printf});
5988 otherwise incorrect code will be generated for calls to those
5991 In addition, seriously incorrect code will result if you call a
5992 function with too many arguments. (Normally, extra arguments are
5993 harmlessly ignored.)
5995 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
5996 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
5999 @itemx -mno-align-int
6001 @opindex mno-align-int
6002 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
6003 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
6004 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
6005 Aligning variables on 32-bit boundaries produces code that runs somewhat
6006 faster on processors with 32-bit busses at the expense of more memory.
6008 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
6009 align structures containing the above types differently than
6010 most published application binary interface specifications for the m68k.
6014 Use the pc-relative addressing mode of the 68000 directly, instead of
6015 using a global offset table. At present, this option implies @option{-fpic},
6016 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
6017 not presently supported with @option{-mpcrel}, though this could be supported for
6018 68020 and higher processors.
6020 @item -mno-strict-align
6021 @itemx -mstrict-align
6022 @opindex mno-strict-align
6023 @opindex mstrict-align
6024 Do not (do) assume that unaligned memory references will be handled by
6028 Generate code that allows the data segment to be located in a different
6029 area of memory from the text segment. This allows for execute in place in
6030 an environment without virtual memory management. This option implies -fPIC.
6033 Generate code that assumes that the data segment follows the text segment.
6034 This is the default.
6036 @item -mid-shared-library
6037 Generate code that supports shared libraries via the library ID method.
6038 This allows for execute in place and shared libraries in an environment
6039 without virtual memory management. This option implies -fPIC.
6041 @item -mno-id-shared-library
6042 Generate code that doesn't assume ID based shared libraries are being used.
6043 This is the default.
6045 @item -mshared-library-id=n
6046 Specified the identification number of the ID based shared library being
6047 compiled. Specifying a value of 0 will generate more compact code, specifying
6048 other values will force the allocation of that number to the current
6049 library but is no more space or time efficient than omitting this option.
6053 @node M68hc1x Options
6054 @subsection M68hc1x Options
6055 @cindex M68hc1x options
6057 These are the @samp{-m} options defined for the 68hc11 and 68hc12
6058 microcontrollers. The default values for these options depends on
6059 which style of microcontroller was selected when the compiler was configured;
6060 the defaults for the most common choices are given below.
6067 Generate output for a 68HC11. This is the default
6068 when the compiler is configured for 68HC11-based systems.
6074 Generate output for a 68HC12. This is the default
6075 when the compiler is configured for 68HC12-based systems.
6081 Generate output for a 68HCS12.
6084 @opindex mauto-incdec
6085 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
6092 Enable the use of 68HC12 min and max instructions.
6095 @itemx -mno-long-calls
6096 @opindex mlong-calls
6097 @opindex mno-long-calls
6098 Treat all calls as being far away (near). If calls are assumed to be
6099 far away, the compiler will use the @code{call} instruction to
6100 call a function and the @code{rtc} instruction for returning.
6104 Consider type @code{int} to be 16 bits wide, like @code{short int}.
6106 @item -msoft-reg-count=@var{count}
6107 @opindex msoft-reg-count
6108 Specify the number of pseudo-soft registers which are used for the
6109 code generation. The maximum number is 32. Using more pseudo-soft
6110 register may or may not result in better code depending on the program.
6111 The default is 4 for 68HC11 and 2 for 68HC12.
6116 @subsection VAX Options
6119 These @samp{-m} options are defined for the VAX:
6124 Do not output certain jump instructions (@code{aobleq} and so on)
6125 that the Unix assembler for the VAX cannot handle across long
6130 Do output those jump instructions, on the assumption that you
6131 will assemble with the GNU assembler.
6135 Output code for g-format floating point numbers instead of d-format.
6139 @subsection SPARC Options
6140 @cindex SPARC options
6142 These @samp{-m} switches are supported on the SPARC:
6147 @opindex mno-app-regs
6149 Specify @option{-mapp-regs} to generate output using the global registers
6150 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
6153 To be fully SVR4 ABI compliant at the cost of some performance loss,
6154 specify @option{-mno-app-regs}. You should compile libraries and system
6155 software with this option.
6160 @opindex mhard-float
6161 Generate output containing floating point instructions. This is the
6167 @opindex msoft-float
6168 Generate output containing library calls for floating point.
6169 @strong{Warning:} the requisite libraries are not available for all SPARC
6170 targets. Normally the facilities of the machine's usual C compiler are
6171 used, but this cannot be done directly in cross-compilation. You must make
6172 your own arrangements to provide suitable library functions for
6173 cross-compilation. The embedded targets @samp{sparc-*-aout} and
6174 @samp{sparclite-*-*} do provide software floating point support.
6176 @option{-msoft-float} changes the calling convention in the output file;
6177 therefore, it is only useful if you compile @emph{all} of a program with
6178 this option. In particular, you need to compile @file{libgcc.a}, the
6179 library that comes with GCC, with @option{-msoft-float} in order for
6182 @item -mhard-quad-float
6183 @opindex mhard-quad-float
6184 Generate output containing quad-word (long double) floating point
6187 @item -msoft-quad-float
6188 @opindex msoft-quad-float
6189 Generate output containing library calls for quad-word (long double)
6190 floating point instructions. The functions called are those specified
6191 in the SPARC ABI@. This is the default.
6193 As of this writing, there are no sparc implementations that have hardware
6194 support for the quad-word floating point instructions. They all invoke
6195 a trap handler for one of these instructions, and then the trap handler
6196 emulates the effect of the instruction. Because of the trap handler overhead,
6197 this is much slower than calling the ABI library routines. Thus the
6198 @option{-msoft-quad-float} option is the default.
6204 With @option{-mflat}, the compiler does not generate save/restore instructions
6205 and will use a ``flat'' or single register window calling convention.
6206 This model uses %i7 as the frame pointer and is compatible with the normal
6207 register window model. Code from either may be intermixed.
6208 The local registers and the input registers (0--5) are still treated as
6209 ``call saved'' registers and will be saved on the stack as necessary.
6211 With @option{-mno-flat} (the default), the compiler emits save/restore
6212 instructions (except for leaf functions) and is the normal mode of operation.
6214 @item -mno-unaligned-doubles
6215 @itemx -munaligned-doubles
6216 @opindex mno-unaligned-doubles
6217 @opindex munaligned-doubles
6218 Assume that doubles have 8 byte alignment. This is the default.
6220 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
6221 alignment only if they are contained in another type, or if they have an
6222 absolute address. Otherwise, it assumes they have 4 byte alignment.
6223 Specifying this option avoids some rare compatibility problems with code
6224 generated by other compilers. It is not the default because it results
6225 in a performance loss, especially for floating point code.
6227 @item -mno-faster-structs
6228 @itemx -mfaster-structs
6229 @opindex mno-faster-structs
6230 @opindex mfaster-structs
6231 With @option{-mfaster-structs}, the compiler assumes that structures
6232 should have 8 byte alignment. This enables the use of pairs of
6233 @code{ldd} and @code{std} instructions for copies in structure
6234 assignment, in place of twice as many @code{ld} and @code{st} pairs.
6235 However, the use of this changed alignment directly violates the SPARC
6236 ABI@. Thus, it's intended only for use on targets where the developer
6237 acknowledges that their resulting code will not be directly in line with
6238 the rules of the ABI@.
6241 @opindex mimpure-text
6242 @option{-mimpure-text}, used in addition to @option{-shared}, tells
6243 the compiler to not pass @option{-z text} to the linker when linking a
6244 shared object. Using this option, you can link position-dependent
6245 code into a shared object.
6247 @option{-mimpure-text} suppresses the ``relocations remain against
6248 allocatable but non-writable sections'' linker error message.
6249 However, the necessary relocations will trigger copy-on-write, and the
6250 shared object is not actually shared across processes. Instead of
6251 using @option{-mimpure-text}, you should compile all source code with
6252 @option{-fpic} or @option{-fPIC}.
6254 This option is only available on SunOS and Solaris.
6260 These two options select variations on the SPARC architecture.
6262 By default (unless specifically configured for the Fujitsu SPARClite),
6263 GCC generates code for the v7 variant of the SPARC architecture.
6265 @option{-mv8} will give you SPARC v8 code. The only difference from v7
6266 code is that the compiler emits the integer multiply and integer
6267 divide instructions which exist in SPARC v8 but not in SPARC v7.
6269 @option{-msparclite} will give you SPARClite code. This adds the integer
6270 multiply, integer divide step and scan (@code{ffs}) instructions which
6271 exist in SPARClite but not in SPARC v7.
6273 These options are deprecated and will be deleted in a future GCC release.
6274 They have been replaced with @option{-mcpu=xxx}.
6279 @opindex msupersparc
6280 These two options select the processor for which the code is optimized.
6282 With @option{-mcypress} (the default), the compiler optimizes code for the
6283 Cypress CY7C602 chip, as used in the SPARCStation/SPARCServer 3xx series.
6284 This is also appropriate for the older SPARCStation 1, 2, IPX etc.
6286 With @option{-msupersparc} the compiler optimizes code for the SuperSPARC cpu, as
6287 used in the SPARCStation 10, 1000 and 2000 series. This flag also enables use
6288 of the full SPARC v8 instruction set.
6290 These options are deprecated and will be deleted in a future GCC release.
6291 They have been replaced with @option{-mcpu=xxx}.
6293 @item -mcpu=@var{cpu_type}
6295 Set the instruction set, register set, and instruction scheduling parameters
6296 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
6297 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
6298 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
6299 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
6302 Default instruction scheduling parameters are used for values that select
6303 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
6304 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
6306 Here is a list of each supported architecture and their supported
6311 v8: supersparc, hypersparc
6312 sparclite: f930, f934, sparclite86x
6314 v9: ultrasparc, ultrasparc3
6317 @item -mtune=@var{cpu_type}
6319 Set the instruction scheduling parameters for machine type
6320 @var{cpu_type}, but do not set the instruction set or register set that the
6321 option @option{-mcpu=@var{cpu_type}} would.
6323 The same values for @option{-mcpu=@var{cpu_type}} can be used for
6324 @option{-mtune=@var{cpu_type}}, but the only useful values are those
6325 that select a particular cpu implementation. Those are @samp{cypress},
6326 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
6327 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
6332 These @samp{-m} switches are supported in addition to the above
6333 on the SPARCLET processor.
6336 @item -mlittle-endian
6337 @opindex mlittle-endian
6338 Generate code for a processor running in little-endian mode.
6342 Treat register @code{%g0} as a normal register.
6343 GCC will continue to clobber it as necessary but will not assume
6344 it always reads as 0.
6346 @item -mbroken-saverestore
6347 @opindex mbroken-saverestore
6348 Generate code that does not use non-trivial forms of the @code{save} and
6349 @code{restore} instructions. Early versions of the SPARCLET processor do
6350 not correctly handle @code{save} and @code{restore} instructions used with
6351 arguments. They correctly handle them used without arguments. A @code{save}
6352 instruction used without arguments increments the current window pointer
6353 but does not allocate a new stack frame. It is assumed that the window
6354 overflow trap handler will properly handle this case as will interrupt
6358 These @samp{-m} switches are supported in addition to the above
6359 on SPARC V9 processors in 64-bit environments.
6362 @item -mlittle-endian
6363 @opindex mlittle-endian
6364 Generate code for a processor running in little-endian mode. It is only
6365 available for a few configurations and most notably not on Solaris.
6371 Generate code for a 32-bit or 64-bit environment.
6372 The 32-bit environment sets int, long and pointer to 32 bits.
6373 The 64-bit environment sets int to 32 bits and long and pointer
6376 @item -mcmodel=medlow
6377 @opindex mcmodel=medlow
6378 Generate code for the Medium/Low code model: the program must be linked
6379 in the low 32 bits of the address space. Pointers are 64 bits.
6380 Programs can be statically or dynamically linked.
6382 @item -mcmodel=medmid
6383 @opindex mcmodel=medmid
6384 Generate code for the Medium/Middle code model: the program must be linked
6385 in the low 44 bits of the address space, the text segment must be less than
6386 2G bytes, and data segment must be within 2G of the text segment.
6387 Pointers are 64 bits.
6389 @item -mcmodel=medany
6390 @opindex mcmodel=medany
6391 Generate code for the Medium/Anywhere code model: the program may be linked
6392 anywhere in the address space, the text segment must be less than
6393 2G bytes, and data segment must be within 2G of the text segment.
6394 Pointers are 64 bits.
6396 @item -mcmodel=embmedany
6397 @opindex mcmodel=embmedany
6398 Generate code for the Medium/Anywhere code model for embedded systems:
6399 assume a 32-bit text and a 32-bit data segment, both starting anywhere
6400 (determined at link time). Register %g4 points to the base of the
6401 data segment. Pointers are still 64 bits.
6402 Programs are statically linked, PIC is not supported.
6405 @itemx -mno-stack-bias
6406 @opindex mstack-bias
6407 @opindex mno-stack-bias
6408 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
6409 frame pointer if present, are offset by @minus{}2047 which must be added back
6410 when making stack frame references.
6411 Otherwise, assume no such offset is present.
6415 @subsection ARM Options
6418 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6423 @opindex mapcs-frame
6424 Generate a stack frame that is compliant with the ARM Procedure Call
6425 Standard for all functions, even if this is not strictly necessary for
6426 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6427 with this option will cause the stack frames not to be generated for
6428 leaf functions. The default is @option{-mno-apcs-frame}.
6432 This is a synonym for @option{-mapcs-frame}.
6436 Generate code for a processor running with a 26-bit program counter,
6437 and conforming to the function calling standards for the APCS 26-bit
6438 option. This option replaces the @option{-m2} and @option{-m3} options
6439 of previous releases of the compiler.
6443 Generate code for a processor running with a 32-bit program counter,
6444 and conforming to the function calling standards for the APCS 32-bit
6445 option. This option replaces the @option{-m6} option of previous releases
6449 @c not currently implemented
6450 @item -mapcs-stack-check
6451 @opindex mapcs-stack-check
6452 Generate code to check the amount of stack space available upon entry to
6453 every function (that actually uses some stack space). If there is
6454 insufficient space available then either the function
6455 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6456 called, depending upon the amount of stack space required. The run time
6457 system is required to provide these functions. The default is
6458 @option{-mno-apcs-stack-check}, since this produces smaller code.
6460 @c not currently implemented
6462 @opindex mapcs-float
6463 Pass floating point arguments using the float point registers. This is
6464 one of the variants of the APCS@. This option is recommended if the
6465 target hardware has a floating point unit or if a lot of floating point
6466 arithmetic is going to be performed by the code. The default is
6467 @option{-mno-apcs-float}, since integer only code is slightly increased in
6468 size if @option{-mapcs-float} is used.
6470 @c not currently implemented
6471 @item -mapcs-reentrant
6472 @opindex mapcs-reentrant
6473 Generate reentrant, position independent code. The default is
6474 @option{-mno-apcs-reentrant}.
6477 @item -mthumb-interwork
6478 @opindex mthumb-interwork
6479 Generate code which supports calling between the ARM and Thumb
6480 instruction sets. Without this option the two instruction sets cannot
6481 be reliably used inside one program. The default is
6482 @option{-mno-thumb-interwork}, since slightly larger code is generated
6483 when @option{-mthumb-interwork} is specified.
6485 @item -mno-sched-prolog
6486 @opindex mno-sched-prolog
6487 Prevent the reordering of instructions in the function prolog, or the
6488 merging of those instruction with the instructions in the function's
6489 body. This means that all functions will start with a recognizable set
6490 of instructions (or in fact one of a choice from a small set of
6491 different function prologues), and this information can be used to
6492 locate the start if functions inside an executable piece of code. The
6493 default is @option{-msched-prolog}.
6496 @opindex mhard-float
6497 Generate output containing floating point instructions. This is the
6501 @opindex msoft-float
6502 Generate output containing library calls for floating point.
6503 @strong{Warning:} the requisite libraries are not available for all ARM
6504 targets. Normally the facilities of the machine's usual C compiler are
6505 used, but this cannot be done directly in cross-compilation. You must make
6506 your own arrangements to provide suitable library functions for
6509 @option{-msoft-float} changes the calling convention in the output file;
6510 therefore, it is only useful if you compile @emph{all} of a program with
6511 this option. In particular, you need to compile @file{libgcc.a}, the
6512 library that comes with GCC, with @option{-msoft-float} in order for
6515 @item -mlittle-endian
6516 @opindex mlittle-endian
6517 Generate code for a processor running in little-endian mode. This is
6518 the default for all standard configurations.
6521 @opindex mbig-endian
6522 Generate code for a processor running in big-endian mode; the default is
6523 to compile code for a little-endian processor.
6525 @item -mwords-little-endian
6526 @opindex mwords-little-endian
6527 This option only applies when generating code for big-endian processors.
6528 Generate code for a little-endian word order but a big-endian byte
6529 order. That is, a byte order of the form @samp{32107654}. Note: this
6530 option should only be used if you require compatibility with code for
6531 big-endian ARM processors generated by versions of the compiler prior to
6534 @item -malignment-traps
6535 @opindex malignment-traps
6536 Generate code that will not trap if the MMU has alignment traps enabled.
6537 On ARM architectures prior to ARMv4, there were no instructions to
6538 access half-word objects stored in memory. However, when reading from
6539 memory a feature of the ARM architecture allows a word load to be used,
6540 even if the address is unaligned, and the processor core will rotate the
6541 data as it is being loaded. This option tells the compiler that such
6542 misaligned accesses will cause a MMU trap and that it should instead
6543 synthesize the access as a series of byte accesses. The compiler can
6544 still use word accesses to load half-word data if it knows that the
6545 address is aligned to a word boundary.
6547 This option is ignored when compiling for ARM architecture 4 or later,
6548 since these processors have instructions to directly access half-word
6551 @item -mno-alignment-traps
6552 @opindex mno-alignment-traps
6553 Generate code that assumes that the MMU will not trap unaligned
6554 accesses. This produces better code when the target instruction set
6555 does not have half-word memory operations (i.e.@: implementations prior to
6558 Note that you cannot use this option to access unaligned word objects,
6559 since the processor will only fetch one 32-bit aligned object from
6562 The default setting for most targets is @option{-mno-alignment-traps}, since
6563 this produces better code when there are no half-word memory
6564 instructions available.
6566 @item -mshort-load-bytes
6567 @itemx -mno-short-load-words
6568 @opindex mshort-load-bytes
6569 @opindex mno-short-load-words
6570 These are deprecated aliases for @option{-malignment-traps}.
6572 @item -mno-short-load-bytes
6573 @itemx -mshort-load-words
6574 @opindex mno-short-load-bytes
6575 @opindex mshort-load-words
6576 This are deprecated aliases for @option{-mno-alignment-traps}.
6578 @item -mcpu=@var{name}
6580 This specifies the name of the target ARM processor. GCC uses this name
6581 to determine what kind of instructions it can emit when generating
6582 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6583 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6584 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6585 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6586 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6587 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm8},
6588 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6589 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6590 @samp{arm920t}, @samp{arm926ejs}, @samp{arm940t}, @samp{arm9tdmi},
6591 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ejs},
6592 @samp{arm1136js}, @samp{arm1136jfs} ,@samp{xscale}, @samp{iwmmxt},
6595 @itemx -mtune=@var{name}
6597 This option is very similar to the @option{-mcpu=} option, except that
6598 instead of specifying the actual target processor type, and hence
6599 restricting which instructions can be used, it specifies that GCC should
6600 tune the performance of the code as if the target were of the type
6601 specified in this option, but still choosing the instructions that it
6602 will generate based on the cpu specified by a @option{-mcpu=} option.
6603 For some ARM implementations better performance can be obtained by using
6606 @item -march=@var{name}
6608 This specifies the name of the target ARM architecture. GCC uses this
6609 name to determine what kind of instructions it can emit when generating
6610 assembly code. This option can be used in conjunction with or instead
6611 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6612 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6613 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6j},
6614 @samp{iwmmxt}, @samp{ep9312}.
6616 @item -mfpe=@var{number}
6617 @itemx -mfp=@var{number}
6620 This specifies the version of the floating point emulation available on
6621 the target. Permissible values are 2 and 3. @option{-mfp=} is a synonym
6622 for @option{-mfpe=}, for compatibility with older versions of GCC@.
6624 @item -mstructure-size-boundary=@var{n}
6625 @opindex mstructure-size-boundary
6626 The size of all structures and unions will be rounded up to a multiple
6627 of the number of bits set by this option. Permissible values are 8 and
6628 32. The default value varies for different toolchains. For the COFF
6629 targeted toolchain the default value is 8. Specifying the larger number
6630 can produce faster, more efficient code, but can also increase the size
6631 of the program. The two values are potentially incompatible. Code
6632 compiled with one value cannot necessarily expect to work with code or
6633 libraries compiled with the other value, if they exchange information
6634 using structures or unions.
6636 @item -mabort-on-noreturn
6637 @opindex mabort-on-noreturn
6638 Generate a call to the function @code{abort} at the end of a
6639 @code{noreturn} function. It will be executed if the function tries to
6643 @itemx -mno-long-calls
6644 @opindex mlong-calls
6645 @opindex mno-long-calls
6646 Tells the compiler to perform function calls by first loading the
6647 address of the function into a register and then performing a subroutine
6648 call on this register. This switch is needed if the target function
6649 will lie outside of the 64 megabyte addressing range of the offset based
6650 version of subroutine call instruction.
6652 Even if this switch is enabled, not all function calls will be turned
6653 into long calls. The heuristic is that static functions, functions
6654 which have the @samp{short-call} attribute, functions that are inside
6655 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6656 definitions have already been compiled within the current compilation
6657 unit, will not be turned into long calls. The exception to this rule is
6658 that weak function definitions, functions with the @samp{long-call}
6659 attribute or the @samp{section} attribute, and functions that are within
6660 the scope of a @samp{#pragma long_calls} directive, will always be
6661 turned into long calls.
6663 This feature is not enabled by default. Specifying
6664 @option{-mno-long-calls} will restore the default behavior, as will
6665 placing the function calls within the scope of a @samp{#pragma
6666 long_calls_off} directive. Note these switches have no effect on how
6667 the compiler generates code to handle function calls via function
6670 @item -mnop-fun-dllimport
6671 @opindex mnop-fun-dllimport
6672 Disable support for the @code{dllimport} attribute.
6674 @item -msingle-pic-base
6675 @opindex msingle-pic-base
6676 Treat the register used for PIC addressing as read-only, rather than
6677 loading it in the prologue for each function. The run-time system is
6678 responsible for initializing this register with an appropriate value
6679 before execution begins.
6681 @item -mpic-register=@var{reg}
6682 @opindex mpic-register
6683 Specify the register to be used for PIC addressing. The default is R10
6684 unless stack-checking is enabled, when R9 is used.
6686 @item -mcirrus-fix-invalid-insns
6687 @opindex mcirrus-fix-invalid-insns
6688 @opindex mno-cirrus-fix-invalid-insns
6689 Insert NOPs into the instruction stream to in order to work around
6690 problems with invalid Maverick instruction combinations. This option
6691 is only valid if the @option{-mcpu=ep9312} option has been used to
6692 enable generation of instructions for the Cirrus Maverick floating
6693 point co-processor. This option is not enabled by default, since the
6694 problem is only present in older Maverick implementations. The default
6695 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
6698 @item -mpoke-function-name
6699 @opindex mpoke-function-name
6700 Write the name of each function into the text section, directly
6701 preceding the function prologue. The generated code is similar to this:
6705 .ascii "arm_poke_function_name", 0
6708 .word 0xff000000 + (t1 - t0)
6709 arm_poke_function_name
6711 stmfd sp!, @{fp, ip, lr, pc@}
6715 When performing a stack backtrace, code can inspect the value of
6716 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6717 location @code{pc - 12} and the top 8 bits are set, then we know that
6718 there is a function name embedded immediately preceding this location
6719 and has length @code{((pc[-3]) & 0xff000000)}.
6723 Generate code for the 16-bit Thumb instruction set. The default is to
6724 use the 32-bit ARM instruction set.
6727 @opindex mtpcs-frame
6728 Generate a stack frame that is compliant with the Thumb Procedure Call
6729 Standard for all non-leaf functions. (A leaf function is one that does
6730 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6732 @item -mtpcs-leaf-frame
6733 @opindex mtpcs-leaf-frame
6734 Generate a stack frame that is compliant with the Thumb Procedure Call
6735 Standard for all leaf functions. (A leaf function is one that does
6736 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6738 @item -mcallee-super-interworking
6739 @opindex mcallee-super-interworking
6740 Gives all externally visible functions in the file being compiled an ARM
6741 instruction set header which switches to Thumb mode before executing the
6742 rest of the function. This allows these functions to be called from
6743 non-interworking code.
6745 @item -mcaller-super-interworking
6746 @opindex mcaller-super-interworking
6747 Allows calls via function pointers (including virtual functions) to
6748 execute correctly regardless of whether the target code has been
6749 compiled for interworking or not. There is a small overhead in the cost
6750 of executing a function pointer if this option is enabled.
6754 @node MN10300 Options
6755 @subsection MN10300 Options
6756 @cindex MN10300 options
6758 These @option{-m} options are defined for Matsushita MN10300 architectures:
6763 Generate code to avoid bugs in the multiply instructions for the MN10300
6764 processors. This is the default.
6767 @opindex mno-mult-bug
6768 Do not generate code to avoid bugs in the multiply instructions for the
6773 Generate code which uses features specific to the AM33 processor.
6777 Do not generate code which uses features specific to the AM33 processor. This
6782 Do not link in the C run-time initialization object file.
6786 Indicate to the linker that it should perform a relaxation optimization pass
6787 to shorten branches, calls and absolute memory addresses. This option only
6788 has an effect when used on the command line for the final link step.
6790 This option makes symbolic debugging impossible.
6794 @node M32R/D Options
6795 @subsection M32R/D Options
6796 @cindex M32R/D options
6798 These @option{-m} options are defined for Renesas M32R/D architectures:
6803 Generate code for the M32R/2@.
6807 Generate code for the M32R/X@.
6811 Generate code for the M32R@. This is the default.
6814 @opindex mmodel=small
6815 Assume all objects live in the lower 16MB of memory (so that their addresses
6816 can be loaded with the @code{ld24} instruction), and assume all subroutines
6817 are reachable with the @code{bl} instruction.
6818 This is the default.
6820 The addressability of a particular object can be set with the
6821 @code{model} attribute.
6823 @item -mmodel=medium
6824 @opindex mmodel=medium
6825 Assume objects may be anywhere in the 32-bit address space (the compiler
6826 will generate @code{seth/add3} instructions to load their addresses), and
6827 assume all subroutines are reachable with the @code{bl} instruction.
6830 @opindex mmodel=large
6831 Assume objects may be anywhere in the 32-bit address space (the compiler
6832 will generate @code{seth/add3} instructions to load their addresses), and
6833 assume subroutines may not be reachable with the @code{bl} instruction
6834 (the compiler will generate the much slower @code{seth/add3/jl}
6835 instruction sequence).
6838 @opindex msdata=none
6839 Disable use of the small data area. Variables will be put into
6840 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
6841 @code{section} attribute has been specified).
6842 This is the default.
6844 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
6845 Objects may be explicitly put in the small data area with the
6846 @code{section} attribute using one of these sections.
6849 @opindex msdata=sdata
6850 Put small global and static data in the small data area, but do not
6851 generate special code to reference them.
6855 Put small global and static data in the small data area, and generate
6856 special instructions to reference them.
6860 @cindex smaller data references
6861 Put global and static objects less than or equal to @var{num} bytes
6862 into the small data or bss sections instead of the normal data or bss
6863 sections. The default value of @var{num} is 8.
6864 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
6865 for this option to have any effect.
6867 All modules should be compiled with the same @option{-G @var{num}} value.
6868 Compiling with different values of @var{num} may or may not work; if it
6869 doesn't the linker will give an error message---incorrect code will not be
6874 Makes the M32R specific code in the compiler display some statistics
6875 that might help in debugging programs.
6878 @opindex malign-loops
6879 Align all loops to a 32-byte boundary.
6881 @item -mno-align-loops
6882 @opindex mno-align-loops
6883 Do not enforce a 32-byte alignment for loops. This is the default.
6885 @item -missue-rate=@var{number}
6886 @opindex missue-rate=@var{number}
6887 Issue @var{number} instructions per cycle. @var{number} can only be 1
6890 @item -mbranch-cost=@var{number}
6891 @opindex mbranch-cost=@var{number}
6892 @var{number} can only be 1 or 2. If it is 1 then branches will be
6893 preferred over conditional code, if it is 2, then the opposite will
6896 @item -mflush-trap=@var{number}
6897 @opindex mflush-trap=@var{number}
6898 Specifies the trap number to use to flush the cache. The default is
6899 12. Valid numbers are between 0 and 15 inclusive.
6901 @item -mno-flush-trap
6902 @opindex mno-flush-trap
6903 Specifies that the cache cannot be flushed by using a trap.
6905 @item -mflush-func=@var{name}
6906 @opindex mflush-func=@var{name}
6907 Specifies the name of the operating system function to call to flush
6908 the cache. The default is @emph{_flush_cache}, but a function call
6909 will only be used if a trap is not available.
6911 @item -mno-flush-func
6912 @opindex mno-flush-func
6913 Indicates that there is no OS function for flushing the cache.
6917 @node RS/6000 and PowerPC Options
6918 @subsection IBM RS/6000 and PowerPC Options
6919 @cindex RS/6000 and PowerPC Options
6920 @cindex IBM RS/6000 and PowerPC Options
6922 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
6930 @itemx -mpowerpc-gpopt
6931 @itemx -mno-powerpc-gpopt
6932 @itemx -mpowerpc-gfxopt
6933 @itemx -mno-powerpc-gfxopt
6935 @itemx -mno-powerpc64
6941 @opindex mno-powerpc
6942 @opindex mpowerpc-gpopt
6943 @opindex mno-powerpc-gpopt
6944 @opindex mpowerpc-gfxopt
6945 @opindex mno-powerpc-gfxopt
6947 @opindex mno-powerpc64
6948 GCC supports two related instruction set architectures for the
6949 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
6950 instructions supported by the @samp{rios} chip set used in the original
6951 RS/6000 systems and the @dfn{PowerPC} instruction set is the
6952 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
6953 the IBM 4xx microprocessors.
6955 Neither architecture is a subset of the other. However there is a
6956 large common subset of instructions supported by both. An MQ
6957 register is included in processors supporting the POWER architecture.
6959 You use these options to specify which instructions are available on the
6960 processor you are using. The default value of these options is
6961 determined when configuring GCC@. Specifying the
6962 @option{-mcpu=@var{cpu_type}} overrides the specification of these
6963 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
6964 rather than the options listed above.
6966 The @option{-mpower} option allows GCC to generate instructions that
6967 are found only in the POWER architecture and to use the MQ register.
6968 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
6969 to generate instructions that are present in the POWER2 architecture but
6970 not the original POWER architecture.
6972 The @option{-mpowerpc} option allows GCC to generate instructions that
6973 are found only in the 32-bit subset of the PowerPC architecture.
6974 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
6975 GCC to use the optional PowerPC architecture instructions in the
6976 General Purpose group, including floating-point square root. Specifying
6977 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
6978 use the optional PowerPC architecture instructions in the Graphics
6979 group, including floating-point select.
6981 The @option{-mpowerpc64} option allows GCC to generate the additional
6982 64-bit instructions that are found in the full PowerPC64 architecture
6983 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
6984 @option{-mno-powerpc64}.
6986 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
6987 will use only the instructions in the common subset of both
6988 architectures plus some special AIX common-mode calls, and will not use
6989 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
6990 permits GCC to use any instruction from either architecture and to
6991 allow use of the MQ register; specify this for the Motorola MPC601.
6993 @item -mnew-mnemonics
6994 @itemx -mold-mnemonics
6995 @opindex mnew-mnemonics
6996 @opindex mold-mnemonics
6997 Select which mnemonics to use in the generated assembler code. With
6998 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
6999 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
7000 assembler mnemonics defined for the POWER architecture. Instructions
7001 defined in only one architecture have only one mnemonic; GCC uses that
7002 mnemonic irrespective of which of these options is specified.
7004 GCC defaults to the mnemonics appropriate for the architecture in
7005 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
7006 value of these option. Unless you are building a cross-compiler, you
7007 should normally not specify either @option{-mnew-mnemonics} or
7008 @option{-mold-mnemonics}, but should instead accept the default.
7010 @item -mcpu=@var{cpu_type}
7012 Set architecture type, register usage, choice of mnemonics, and
7013 instruction scheduling parameters for machine type @var{cpu_type}.
7014 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
7015 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
7016 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
7017 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
7018 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
7019 @samp{860}, @samp{970}, @samp{common}, @samp{ec603e}, @samp{G3},
7020 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
7021 @samp{power4}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
7022 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
7024 @option{-mcpu=common} selects a completely generic processor. Code
7025 generated under this option will run on any POWER or PowerPC processor.
7026 GCC will use only the instructions in the common subset of both
7027 architectures, and will not use the MQ register. GCC assumes a generic
7028 processor model for scheduling purposes.
7030 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
7031 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
7032 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
7033 types, with an appropriate, generic processor model assumed for
7034 scheduling purposes.
7036 The other options specify a specific processor. Code generated under
7037 those options will run best on that processor, and may not run at all on
7040 The @option{-mcpu} options automatically enable or disable the
7041 following options: @option{-maltivec}, @option{-mhard-float},
7042 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
7043 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
7044 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
7045 @option{-mstring}. The particular options set for any particular CPU
7046 will vary between compiler versions, depending on what setting seems
7047 to produce optimal code for that CPU; it doesn't necessarily reflect
7048 the actual hardware's capabilities. If you wish to set an individual
7049 option to a particular value, you may specify it after the
7050 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
7052 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
7053 not enabled or disabled by the @option{-mcpu} option at present, since
7054 AIX does not have full support for these options. You may still
7055 enable or disable them individually if you're sure it'll work in your
7058 @item -mtune=@var{cpu_type}
7060 Set the instruction scheduling parameters for machine type
7061 @var{cpu_type}, but do not set the architecture type, register usage, or
7062 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
7063 values for @var{cpu_type} are used for @option{-mtune} as for
7064 @option{-mcpu}. If both are specified, the code generated will use the
7065 architecture, registers, and mnemonics set by @option{-mcpu}, but the
7066 scheduling parameters set by @option{-mtune}.
7071 @opindex mno-altivec
7072 These switches enable or disable the use of built-in functions that
7073 allow access to the AltiVec instruction set. You may also need to set
7074 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
7079 Extend the current ABI with SPE ABI extensions. This does not change
7080 the default ABI, instead it adds the SPE ABI extensions to the current
7084 @opindex mabi=no-spe
7085 Disable Booke SPE ABI extensions for the current ABI.
7087 @item -misel=@var{yes/no}
7090 This switch enables or disables the generation of ISEL instructions.
7092 @item -mspe=@var{yes/no}
7095 This switch enables or disables the generation of SPE simd
7098 @item -mfloat-gprs=@var{yes/no}
7100 @opindex mfloat-gprs
7101 This switch enables or disables the generation of floating point
7102 operations on the general purpose registers for architectures that
7103 support it. This option is currently only available on the MPC8540.
7106 @itemx -mno-fp-in-toc
7107 @itemx -mno-sum-in-toc
7108 @itemx -mminimal-toc
7110 @opindex mno-fp-in-toc
7111 @opindex mno-sum-in-toc
7112 @opindex mminimal-toc
7113 Modify generation of the TOC (Table Of Contents), which is created for
7114 every executable file. The @option{-mfull-toc} option is selected by
7115 default. In that case, GCC will allocate at least one TOC entry for
7116 each unique non-automatic variable reference in your program. GCC
7117 will also place floating-point constants in the TOC@. However, only
7118 16,384 entries are available in the TOC@.
7120 If you receive a linker error message that saying you have overflowed
7121 the available TOC space, you can reduce the amount of TOC space used
7122 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
7123 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
7124 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
7125 generate code to calculate the sum of an address and a constant at
7126 run-time instead of putting that sum into the TOC@. You may specify one
7127 or both of these options. Each causes GCC to produce very slightly
7128 slower and larger code at the expense of conserving TOC space.
7130 If you still run out of space in the TOC even when you specify both of
7131 these options, specify @option{-mminimal-toc} instead. This option causes
7132 GCC to make only one TOC entry for every file. When you specify this
7133 option, GCC will produce code that is slower and larger but which
7134 uses extremely little TOC space. You may wish to use this option
7135 only on files that contain less frequently executed code.
7141 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
7142 @code{long} type, and the infrastructure needed to support them.
7143 Specifying @option{-maix64} implies @option{-mpowerpc64} and
7144 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
7145 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
7150 @opindex mno-xl-call
7151 On AIX, pass floating-point arguments to prototyped functions beyond the
7152 register save area (RSA) on the stack in addition to argument FPRs. The
7153 AIX calling convention was extended but not initially documented to
7154 handle an obscure K&R C case of calling a function that takes the
7155 address of its arguments with fewer arguments than declared. AIX XL
7156 compilers access floating point arguments which do not fit in the
7157 RSA from the stack when a subroutine is compiled without
7158 optimization. Because always storing floating-point arguments on the
7159 stack is inefficient and rarely needed, this option is not enabled by
7160 default and only is necessary when calling subroutines compiled by AIX
7161 XL compilers without optimization.
7165 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
7166 application written to use message passing with special startup code to
7167 enable the application to run. The system must have PE installed in the
7168 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
7169 must be overridden with the @option{-specs=} option to specify the
7170 appropriate directory location. The Parallel Environment does not
7171 support threads, so the @option{-mpe} option and the @option{-pthread}
7172 option are incompatible.
7174 @item -malign-natural
7175 @itemx -malign-power
7176 @opindex malign-natural
7177 @opindex malign-power
7178 On AIX, Darwin, and 64-bit PowerPC Linux, the option
7179 @option{-malign-natural} overrides the ABI-defined alignment of larger
7180 types, such as floating-point doubles, on their natural size-based boundary.
7181 The option @option{-malign-power} instructs GCC to follow the ABI-specified
7182 alignment rules. GCC defaults to the standard alignment defined in the ABI.
7186 @opindex msoft-float
7187 @opindex mhard-float
7188 Generate code that does not use (uses) the floating-point register set.
7189 Software floating point emulation is provided if you use the
7190 @option{-msoft-float} option, and pass the option to GCC when linking.
7193 @itemx -mno-multiple
7195 @opindex mno-multiple
7196 Generate code that uses (does not use) the load multiple word
7197 instructions and the store multiple word instructions. These
7198 instructions are generated by default on POWER systems, and not
7199 generated on PowerPC systems. Do not use @option{-mmultiple} on little
7200 endian PowerPC systems, since those instructions do not work when the
7201 processor is in little endian mode. The exceptions are PPC740 and
7202 PPC750 which permit the instructions usage in little endian mode.
7208 Generate code that uses (does not use) the load string instructions
7209 and the store string word instructions to save multiple registers and
7210 do small block moves. These instructions are generated by default on
7211 POWER systems, and not generated on PowerPC systems. Do not use
7212 @option{-mstring} on little endian PowerPC systems, since those
7213 instructions do not work when the processor is in little endian mode.
7214 The exceptions are PPC740 and PPC750 which permit the instructions
7215 usage in little endian mode.
7221 Generate code that uses (does not use) the load or store instructions
7222 that update the base register to the address of the calculated memory
7223 location. These instructions are generated by default. If you use
7224 @option{-mno-update}, there is a small window between the time that the
7225 stack pointer is updated and the address of the previous frame is
7226 stored, which means code that walks the stack frame across interrupts or
7227 signals may get corrupted data.
7230 @itemx -mno-fused-madd
7231 @opindex mfused-madd
7232 @opindex mno-fused-madd
7233 Generate code that uses (does not use) the floating point multiply and
7234 accumulate instructions. These instructions are generated by default if
7235 hardware floating is used.
7237 @item -mno-bit-align
7239 @opindex mno-bit-align
7241 On System V.4 and embedded PowerPC systems do not (do) force structures
7242 and unions that contain bit-fields to be aligned to the base type of the
7245 For example, by default a structure containing nothing but 8
7246 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
7247 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
7248 the structure would be aligned to a 1 byte boundary and be one byte in
7251 @item -mno-strict-align
7252 @itemx -mstrict-align
7253 @opindex mno-strict-align
7254 @opindex mstrict-align
7255 On System V.4 and embedded PowerPC systems do not (do) assume that
7256 unaligned memory references will be handled by the system.
7259 @itemx -mno-relocatable
7260 @opindex mrelocatable
7261 @opindex mno-relocatable
7262 On embedded PowerPC systems generate code that allows (does not allow)
7263 the program to be relocated to a different address at runtime. If you
7264 use @option{-mrelocatable} on any module, all objects linked together must
7265 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
7267 @item -mrelocatable-lib
7268 @itemx -mno-relocatable-lib
7269 @opindex mrelocatable-lib
7270 @opindex mno-relocatable-lib
7271 On embedded PowerPC systems generate code that allows (does not allow)
7272 the program to be relocated to a different address at runtime. Modules
7273 compiled with @option{-mrelocatable-lib} can be linked with either modules
7274 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
7275 with modules compiled with the @option{-mrelocatable} options.
7281 On System V.4 and embedded PowerPC systems do not (do) assume that
7282 register 2 contains a pointer to a global area pointing to the addresses
7283 used in the program.
7286 @itemx -mlittle-endian
7288 @opindex mlittle-endian
7289 On System V.4 and embedded PowerPC systems compile code for the
7290 processor in little endian mode. The @option{-mlittle-endian} option is
7291 the same as @option{-mlittle}.
7296 @opindex mbig-endian
7297 On System V.4 and embedded PowerPC systems compile code for the
7298 processor in big endian mode. The @option{-mbig-endian} option is
7299 the same as @option{-mbig}.
7301 @item -mdynamic-no-pic
7302 @opindex mdynamic-no-pic
7303 On Darwin and Mac OS X systems, compile code so that it is not
7304 relocatable, but that its external references are relocatable. The
7305 resulting code is suitable for applications, but not shared
7308 @item -mprioritize-restricted-insns=@var{priority}
7309 @opindex mprioritize-restricted-insns
7310 This option controls the priority that is assigned to
7311 dispatch-slot restricted instructions during the second scheduling
7312 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
7313 @var{no/highest/second-highest} priority to dispatch slot restricted
7316 @item -msched-costly-dep=@var{dependence_type}
7317 @opindex msched-costly-dep
7318 This option controls which dependences are considered costly
7319 by the target during instruction scheduling. The argument
7320 @var{dependence_type} takes one of the following values:
7321 @var{no}: no dependence is costly,
7322 @var{all}: all dependences are costly,
7323 @var{true_store_to_load}: a true dependence from store to load is costly,
7324 @var{store_to_load}: any dependence from store to load is costly,
7325 @var{number}: any dependence which latency >= @var{number} is costly.
7327 @item -minsert-sched-nops=@var{scheme}
7328 @opindex minsert-sched-nops
7329 This option controls which nop insertion scheme will be used during
7330 the second scheduling pass. The argument @var{scheme} takes one of the
7332 @var{no}: Don't insert nops.
7333 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
7334 according to the scheduler's grouping.
7335 @var{regroup_exact}: Insert nops to force costly dependent insns into
7336 separate groups. Insert exactly as many nops as needed to force an insn
7337 to a new group, according to the estimated processor grouping.
7338 @var{number}: Insert nops to force costly dependent insns into
7339 separate groups. Insert @var{number} nops to force an insn to a new group.
7343 On System V.4 and embedded PowerPC systems compile code using calling
7344 conventions that adheres to the March 1995 draft of the System V
7345 Application Binary Interface, PowerPC processor supplement. This is the
7346 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
7348 @item -mcall-sysv-eabi
7349 @opindex mcall-sysv-eabi
7350 Specify both @option{-mcall-sysv} and @option{-meabi} options.
7352 @item -mcall-sysv-noeabi
7353 @opindex mcall-sysv-noeabi
7354 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
7356 @item -mcall-solaris
7357 @opindex mcall-solaris
7358 On System V.4 and embedded PowerPC systems compile code for the Solaris
7362 @opindex mcall-linux
7363 On System V.4 and embedded PowerPC systems compile code for the
7364 Linux-based GNU system.
7368 On System V.4 and embedded PowerPC systems compile code for the
7369 Hurd-based GNU system.
7372 @opindex mcall-netbsd
7373 On System V.4 and embedded PowerPC systems compile code for the
7374 NetBSD operating system.
7376 @item -maix-struct-return
7377 @opindex maix-struct-return
7378 Return all structures in memory (as specified by the AIX ABI)@.
7380 @item -msvr4-struct-return
7381 @opindex msvr4-struct-return
7382 Return structures smaller than 8 bytes in registers (as specified by the
7386 @opindex mabi=altivec
7387 Extend the current ABI with AltiVec ABI extensions. This does not
7388 change the default ABI, instead it adds the AltiVec ABI extensions to
7391 @item -mabi=no-altivec
7392 @opindex mabi=no-altivec
7393 Disable AltiVec ABI extensions for the current ABI.
7396 @itemx -mno-prototype
7398 @opindex mno-prototype
7399 On System V.4 and embedded PowerPC systems assume that all calls to
7400 variable argument functions are properly prototyped. Otherwise, the
7401 compiler must insert an instruction before every non prototyped call to
7402 set or clear bit 6 of the condition code register (@var{CR}) to
7403 indicate whether floating point values were passed in the floating point
7404 registers in case the function takes a variable arguments. With
7405 @option{-mprototype}, only calls to prototyped variable argument functions
7406 will set or clear the bit.
7410 On embedded PowerPC systems, assume that the startup module is called
7411 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
7412 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
7417 On embedded PowerPC systems, assume that the startup module is called
7418 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
7423 On embedded PowerPC systems, assume that the startup module is called
7424 @file{crt0.o} and the standard C libraries are @file{libads.a} and
7428 @opindex myellowknife
7429 On embedded PowerPC systems, assume that the startup module is called
7430 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
7435 On System V.4 and embedded PowerPC systems, specify that you are
7436 compiling for a VxWorks system.
7440 Specify that you are compiling for the WindISS simulation environment.
7444 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
7445 header to indicate that @samp{eabi} extended relocations are used.
7451 On System V.4 and embedded PowerPC systems do (do not) adhere to the
7452 Embedded Applications Binary Interface (eabi) which is a set of
7453 modifications to the System V.4 specifications. Selecting @option{-meabi}
7454 means that the stack is aligned to an 8 byte boundary, a function
7455 @code{__eabi} is called to from @code{main} to set up the eabi
7456 environment, and the @option{-msdata} option can use both @code{r2} and
7457 @code{r13} to point to two separate small data areas. Selecting
7458 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
7459 do not call an initialization function from @code{main}, and the
7460 @option{-msdata} option will only use @code{r13} to point to a single
7461 small data area. The @option{-meabi} option is on by default if you
7462 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
7465 @opindex msdata=eabi
7466 On System V.4 and embedded PowerPC systems, put small initialized
7467 @code{const} global and static data in the @samp{.sdata2} section, which
7468 is pointed to by register @code{r2}. Put small initialized
7469 non-@code{const} global and static data in the @samp{.sdata} section,
7470 which is pointed to by register @code{r13}. Put small uninitialized
7471 global and static data in the @samp{.sbss} section, which is adjacent to
7472 the @samp{.sdata} section. The @option{-msdata=eabi} option is
7473 incompatible with the @option{-mrelocatable} option. The
7474 @option{-msdata=eabi} option also sets the @option{-memb} option.
7477 @opindex msdata=sysv
7478 On System V.4 and embedded PowerPC systems, put small global and static
7479 data in the @samp{.sdata} section, which is pointed to by register
7480 @code{r13}. Put small uninitialized global and static data in the
7481 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
7482 The @option{-msdata=sysv} option is incompatible with the
7483 @option{-mrelocatable} option.
7485 @item -msdata=default
7487 @opindex msdata=default
7489 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
7490 compile code the same as @option{-msdata=eabi}, otherwise compile code the
7491 same as @option{-msdata=sysv}.
7494 @opindex msdata-data
7495 On System V.4 and embedded PowerPC systems, put small global and static
7496 data in the @samp{.sdata} section. Put small uninitialized global and
7497 static data in the @samp{.sbss} section. Do not use register @code{r13}
7498 to address small data however. This is the default behavior unless
7499 other @option{-msdata} options are used.
7503 @opindex msdata=none
7505 On embedded PowerPC systems, put all initialized global and static data
7506 in the @samp{.data} section, and all uninitialized data in the
7507 @samp{.bss} section.
7511 @cindex smaller data references (PowerPC)
7512 @cindex .sdata/.sdata2 references (PowerPC)
7513 On embedded PowerPC systems, put global and static items less than or
7514 equal to @var{num} bytes into the small data or bss sections instead of
7515 the normal data or bss section. By default, @var{num} is 8. The
7516 @option{-G @var{num}} switch is also passed to the linker.
7517 All modules should be compiled with the same @option{-G @var{num}} value.
7520 @itemx -mno-regnames
7522 @opindex mno-regnames
7523 On System V.4 and embedded PowerPC systems do (do not) emit register
7524 names in the assembly language output using symbolic forms.
7527 @itemx -mno-longcall
7529 @opindex mno-longcall
7530 Default to making all function calls via pointers, so that functions
7531 which reside further than 64 megabytes (67,108,864 bytes) from the
7532 current location can be called. This setting can be overridden by the
7533 @code{shortcall} function attribute, or by @code{#pragma longcall(0)}.
7535 Some linkers are capable of detecting out-of-range calls and generating
7536 glue code on the fly. On these systems, long calls are unnecessary and
7537 generate slower code. As of this writing, the AIX linker can do this,
7538 as can the GNU linker for PowerPC/64. It is planned to add this feature
7539 to the GNU linker for 32-bit PowerPC systems as well.
7541 On Mach-O (Darwin) systems, this option directs the compiler emit to
7542 the glue for every direct call, and the Darwin linker decides whether
7543 to use or discard it.
7545 In the future, we may cause GCC to ignore all longcall specifications
7546 when the linker is known to generate glue.
7550 Adds support for multithreading with the @dfn{pthreads} library.
7551 This option sets flags for both the preprocessor and linker.
7555 @node Darwin Options
7556 @subsection Darwin Options
7557 @cindex Darwin options
7559 These options are defined for all architectures running the Darwin operating
7560 system. They are useful for compatibility with other Mac OS compilers.
7565 Loads all members of static archive libraries.
7566 See man ld(1) for more information.
7568 @item -arch_errors_fatal
7569 @opindex arch_errors_fatal
7570 Cause the errors having to do with files that have the wrong architecture
7574 @opindex bind_at_load
7575 Causes the output file to be marked such that the dynamic linker will
7576 bind all undefined references when the file is loaded or launched.
7580 Produce a Mach-o bundle format file.
7581 See man ld(1) for more information.
7583 @item -bundle_loader @var{executable}
7584 @opindex bundle_loader
7585 This specifies the @var{executable} that will be loading the build
7586 output file being linked. See man ld(1) for more information.
7588 @item -allowable_client @var{client_name}
7592 @itemx -compatibility_version
7593 @itemx -current_version
7594 @itemx -dependency-file
7596 @itemx -dylinker_install_name
7599 @itemx -exported_symbols_list
7601 @itemx -flat_namespace
7602 @itemx -force_cpusubtype_ALL
7603 @itemx -force_flat_namespace
7604 @itemx -headerpad_max_install_names
7607 @itemx -install_name
7608 @itemx -keep_private_externs
7609 @itemx -multi_module
7610 @itemx -multiply_defined
7611 @itemx -multiply_defined_unused
7613 @itemx -nofixprebinding
7616 @itemx -noseglinkedit
7617 @itemx -pagezero_size
7619 @itemx -prebind_all_twolevel_modules
7620 @itemx -private_bundle
7621 @itemx -read_only_relocs
7623 @itemx -sectobjectsymbols
7627 @itemx -sectobjectsymbols
7629 @itemx -seg_addr_table
7630 @itemx -seg_addr_table_filename
7633 @itemx -segs_read_only_addr
7634 @itemx -segs_read_write_addr
7635 @itemx -single_module
7638 @itemx -sub_umbrella
7639 @itemx -twolevel_namespace
7642 @itemx -unexported_symbols_list
7643 @itemx -weak_reference_mismatches
7646 @opindex allowable_client
7648 @opindex client_name
7649 @opindex compatibility_version
7650 @opindex current_version
7651 @opindex dependency-file
7653 @opindex dylinker_install_name
7656 @opindex exported_symbols_list
7658 @opindex flat_namespace
7659 @opindex force_cpusubtype_ALL
7660 @opindex force_flat_namespace
7661 @opindex headerpad_max_install_names
7664 @opindex install_name
7665 @opindex keep_private_externs
7666 @opindex multi_module
7667 @opindex multiply_defined
7668 @opindex multiply_defined_unused
7670 @opindex nofixprebinding
7671 @opindex nomultidefs
7673 @opindex noseglinkedit
7674 @opindex pagezero_size
7676 @opindex prebind_all_twolevel_modules
7677 @opindex private_bundle
7678 @opindex read_only_relocs
7680 @opindex sectobjectsymbols
7684 @opindex sectobjectsymbols
7686 @opindex seg_addr_table
7687 @opindex seg_addr_table_filename
7688 @opindex seglinkedit
7690 @opindex segs_read_only_addr
7691 @opindex segs_read_write_addr
7692 @opindex single_module
7694 @opindex sub_library
7695 @opindex sub_umbrella
7696 @opindex twolevel_namespace
7699 @opindex unexported_symbols_list
7700 @opindex weak_reference_mismatches
7701 @opindex whatsloaded
7703 These options are available for Darwin linker. Darwin linker man page
7704 describes them in detail.
7709 @subsection MIPS Options
7710 @cindex MIPS options
7716 Generate big-endian code.
7720 Generate little-endian code. This is the default for @samp{mips*el-*-*}
7723 @item -march=@var{arch}
7725 Generate code that will run on @var{arch}, which can be the name of a
7726 generic MIPS ISA, or the name of a particular processor.
7728 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
7729 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
7730 The processor names are:
7731 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
7733 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
7734 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
7738 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4300},
7739 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
7740 The special value @samp{from-abi} selects the
7741 most compatible architecture for the selected ABI (that is,
7742 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
7744 In processor names, a final @samp{000} can be abbreviated as @samp{k}
7745 (for example, @samp{-march=r2k}). Prefixes are optional, and
7746 @samp{vr} may be written @samp{r}.
7748 GCC defines two macros based on the value of this option. The first
7749 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
7750 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
7751 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
7752 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
7753 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
7755 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
7756 above. In other words, it will have the full prefix and will not
7757 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
7758 the macro names the resolved architecture (either @samp{"mips1"} or
7759 @samp{"mips3"}). It names the default architecture when no
7760 @option{-march} option is given.
7762 @item -mtune=@var{arch}
7764 Optimize for @var{arch}. Among other things, this option controls
7765 the way instructions are scheduled, and the perceived cost of arithmetic
7766 operations. The list of @var{arch} values is the same as for
7769 When this option is not used, GCC will optimize for the processor
7770 specified by @option{-march}. By using @option{-march} and
7771 @option{-mtune} together, it is possible to generate code that will
7772 run on a family of processors, but optimize the code for one
7773 particular member of that family.
7775 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
7776 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
7777 @samp{-march} ones described above.
7781 Equivalent to @samp{-march=mips1}.
7785 Equivalent to @samp{-march=mips2}.
7789 Equivalent to @samp{-march=mips3}.
7793 Equivalent to @samp{-march=mips4}.
7797 Equivalent to @samp{-march=mips32}.
7801 Equivalent to @samp{-march=mips32r2}.
7805 Equivalent to @samp{-march=mips64}.
7811 Use (do not use) the MIPS16 ISA.
7823 Generate code for the given ABI@.
7825 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
7826 generates 64-bit code when you select a 64-bit architecture, but you
7827 can use @option{-mgp32} to get 32-bit code instead.
7830 @itemx -mno-abicalls
7832 @opindex mno-abicalls
7833 Generate (do not generate) SVR4-style position-independent code.
7834 @option{-mabicalls} is the default for SVR4-based systems.
7840 Lift (do not lift) the usual restrictions on the size of the global
7843 GCC normally uses a single instruction to load values from the GOT.
7844 While this is relatively efficient, it will only work if the GOT
7845 is smaller than about 64k. Anything larger will cause the linker
7846 to report an error such as:
7848 @cindex relocation truncated to fit (MIPS)
7850 relocation truncated to fit: R_MIPS_GOT16 foobar
7853 If this happens, you should recompile your code with @option{-mxgot}.
7854 It should then work with very large GOTs, although it will also be
7855 less efficient, since it will take three instructions to fetch the
7856 value of a global symbol.
7858 Note that some linkers can create multiple GOTs. If you have such a
7859 linker, you should only need to use @option{-mxgot} when a single object
7860 file accesses more than 64k's worth of GOT entries. Very few do.
7862 These options have no effect unless GCC is generating position
7865 @item -membedded-pic
7866 @itemx -mno-embedded-pic
7867 @opindex membedded-pic
7868 @opindex mno-embedded-pic
7869 Generate (do not generate) position-independent code suitable for some
7870 embedded systems. All calls are made using PC relative addresses, and
7871 all data is addressed using the $gp register. No more than 65536
7872 bytes of global data may be used. This requires GNU as and GNU ld,
7873 which do most of the work.
7877 Assume that general-purpose registers are 32 bits wide.
7881 Assume that general-purpose registers are 64 bits wide.
7885 Assume that floating-point registers are 32 bits wide.
7889 Assume that floating-point registers are 64 bits wide.
7892 @opindex mhard-float
7893 Use floating-point coprocessor instructions.
7896 @opindex msoft-float
7897 Do not use floating-point coprocessor instructions. Implement
7898 floating-point calculations using library calls instead.
7900 @item -msingle-float
7901 @opindex msingle-float
7902 Assume that the floating-point coprocessor only supports single-precision
7905 @itemx -mdouble-float
7906 @opindex mdouble-float
7907 Assume that the floating-point coprocessor supports double-precision
7908 operations. This is the default.
7912 Force @code{int} and @code{long} types to be 64 bits wide. See
7913 @option{-mlong32} for an explanation of the default and the way
7914 that the pointer size is determined.
7918 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
7919 an explanation of the default and the way that the pointer size is
7924 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
7926 The default size of @code{int}s, @code{long}s and pointers depends on
7927 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
7928 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
7929 32-bit @code{long}s. Pointers are the same size as @code{long}s,
7930 or the same size as integer registers, whichever is smaller.
7934 @cindex smaller data references (MIPS)
7935 @cindex gp-relative references (MIPS)
7936 Put global and static items less than or equal to @var{num} bytes into
7937 the small data or bss section instead of the normal data or bss section.
7938 This allows the data to be accessed using a single instruction.
7940 All modules should be compiled with the same @option{-G @var{num}}
7943 @item -membedded-data
7944 @itemx -mno-embedded-data
7945 @opindex membedded-data
7946 @opindex mno-embedded-data
7947 Allocate variables to the read-only data section first if possible, then
7948 next in the small data section if possible, otherwise in data. This gives
7949 slightly slower code than the default, but reduces the amount of RAM required
7950 when executing, and thus may be preferred for some embedded systems.
7952 @item -muninit-const-in-rodata
7953 @itemx -mno-uninit-const-in-rodata
7954 @opindex muninit-const-in-rodata
7955 @opindex mno-uninit-const-in-rodata
7956 Put uninitialized @code{const} variables in the read-only data section.
7957 This option is only meaningful in conjunction with @option{-membedded-data}.
7959 @item -msplit-addresses
7960 @itemx -mno-split-addresses
7961 @opindex msplit-addresses
7962 @opindex mno-split-addresses
7963 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
7964 relocation operators. This option has been superceded by
7965 @option{-mexplicit-relocs} but is retained for backwards compatibility.
7967 @item -mexplicit-relocs
7968 @itemx -mno-explicit-relocs
7969 @opindex mexplicit-relocs
7970 @opindex mno-explicit-relocs
7971 Use (do not use) assembler relocation operators when dealing with symbolic
7972 addresses. The alternative, selected by @option{-mno-explicit-relocs},
7973 is to use assembler macros instead.
7975 @option{-mexplicit-relocs} is usually the default if GCC was
7976 configured to use an assembler that supports relocation operators.
7977 However, there are two exceptions:
7981 GCC is not yet able to generate explicit relocations for the combination
7982 of @option{-mabi=64} and @option{-mno-abicalls}. This will be addressed
7983 in a future release.
7986 The combination of @option{-mabicalls} and @option{-fno-unit-at-a-time}
7987 implies @option{-mno-explicit-relocs} unless explicitly overridden.
7988 This is because, when generating abicalls, the choice of relocation
7989 depends on whether a symbol is local or global. In some rare cases,
7990 GCC will not be able to decide this until the whole compilation unit
7998 Generate (do not generate) code that refers to registers using their
7999 software names. The default is @option{-mno-rnames}, which tells GCC
8000 to use hardware names like @samp{$4} instead of software names like
8001 @samp{a0}. The only assembler known to support @option{-rnames} is
8002 the Algorithmics assembler.
8004 @item -mcheck-zero-division
8005 @itemx -mno-check-zero-division
8006 @opindex mcheck-zero-division
8007 @opindex mno-check-zero-division
8008 Trap (do not trap) on integer division by zero. The default is
8009 @option{-mcheck-zero-division}.
8015 Force (do not force) the use of @code{memcpy()} for non-trivial block
8016 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
8017 most constant-sized copies.
8020 @itemx -mno-long-calls
8021 @opindex mlong-calls
8022 @opindex mno-long-calls
8023 Disable (do not disable) use of the @code{jal} instruction. Calling
8024 functions using @code{jal} is more efficient but requires the caller
8025 and callee to be in the same 256 megabyte segment.
8027 This option has no effect on abicalls code. The default is
8028 @option{-mno-long-calls}.
8034 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
8035 instructions, as provided by the R4650 ISA.
8038 @itemx -mno-fused-madd
8039 @opindex mfused-madd
8040 @opindex mno-fused-madd
8041 Enable (disable) use of the floating point multiply-accumulate
8042 instructions, when they are available. The default is
8043 @option{-mfused-madd}.
8045 When multiply-accumulate instructions are used, the intermediate
8046 product is calculated to infinite precision and is not subject to
8047 the FCSR Flush to Zero bit. This may be undesirable in some
8052 Tell the MIPS assembler to not run its preprocessor over user
8053 assembler files (with a @samp{.s} suffix) when assembling them.
8058 Work around certain SB-1 CPU core errata.
8059 (This flag currently works around the SB-1 revision 2
8060 ``F1'' and ``F2'' floating point errata.)
8062 @item -mflush-func=@var{func}
8063 @itemx -mno-flush-func
8064 @opindex mflush-func
8065 Specifies the function to call to flush the I and D caches, or to not
8066 call any such function. If called, the function must take the same
8067 arguments as the common @code{_flush_func()}, that is, the address of the
8068 memory range for which the cache is being flushed, the size of the
8069 memory range, and the number 3 (to flush both caches). The default
8070 depends on the target gcc was configured for, but commonly is either
8071 @samp{_flush_func} or @samp{__cpu_flush}.
8073 @item -mbranch-likely
8074 @itemx -mno-branch-likely
8075 @opindex mbranch-likely
8076 @opindex mno-branch-likely
8077 Enable or disable use of Branch Likely instructions, regardless of the
8078 default for the selected architecture. By default, Branch Likely
8079 instructions may be generated if they are supported by the selected
8080 architecture. An exception is for the MIPS32 and MIPS64 architectures
8081 and processors which implement those architectures; for those, Branch
8082 Likely instructions will not be generated by default because the MIPS32
8083 and MIPS64 architectures specifically deprecate their use.
8086 @node i386 and x86-64 Options
8087 @subsection Intel 386 and AMD x86-64 Options
8088 @cindex i386 Options
8089 @cindex x86-64 Options
8090 @cindex Intel 386 Options
8091 @cindex AMD x86-64 Options
8093 These @samp{-m} options are defined for the i386 and x86-64 family of
8097 @item -mtune=@var{cpu-type}
8099 Tune to @var{cpu-type} everything applicable about the generated code, except
8100 for the ABI and the set of available instructions. The choices for
8104 Original Intel's i386 CPU.
8106 Intel's i486 CPU. (No scheduling is implemented for this chip.)
8108 Intel Pentium CPU with no MMX support.
8110 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8111 @item i686, pentiumpro
8112 Intel PentiumPro CPU.
8114 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8116 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8119 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8121 AMD K6 CPU with MMX instruction set support.
8123 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8124 @item athlon, athlon-tbird
8125 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8127 @item athlon-4, athlon-xp, athlon-mp
8128 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8129 instruction set support.
8130 @item k8, opteron, athlon64, athlon-fx
8131 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8132 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8134 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8137 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8138 instruction set support.
8140 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8141 implemented for this chip.)
8143 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8144 implemented for this chip.)
8147 While picking a specific @var{cpu-type} will schedule things appropriately
8148 for that particular chip, the compiler will not generate any code that
8149 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8152 @item -march=@var{cpu-type}
8154 Generate instructions for the machine type @var{cpu-type}. The choices
8155 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8156 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8158 @item -mcpu=@var{cpu-type}
8160 A deprecated synonym for @option{-mtune}.
8169 @opindex mpentiumpro
8170 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8171 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8172 These synonyms are deprecated.
8174 @item -mfpmath=@var{unit}
8176 generate floating point arithmetics for selected unit @var{unit}. the choices
8181 Use the standard 387 floating point coprocessor present majority of chips and
8182 emulated otherwise. Code compiled with this option will run almost everywhere.
8183 The temporary results are computed in 80bit precision instead of precision
8184 specified by the type resulting in slightly different results compared to most
8185 of other chips. See @option{-ffloat-store} for more detailed description.
8187 This is the default choice for i386 compiler.
8190 Use scalar floating point instructions present in the SSE instruction set.
8191 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8192 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8193 instruction set supports only single precision arithmetics, thus the double and
8194 extended precision arithmetics is still done using 387. Later version, present
8195 only in Pentium4 and the future AMD x86-64 chips supports double precision
8198 For i387 you need to use @option{-march=@var{cpu-type}}, @option{-msse} or
8199 @option{-msse2} switches to enable SSE extensions and make this option
8200 effective. For x86-64 compiler, these extensions are enabled by default.
8202 The resulting code should be considerably faster in majority of cases and avoid
8203 the numerical instability problems of 387 code, but may break some existing
8204 code that expects temporaries to be 80bit.
8206 This is the default choice for x86-64 compiler.
8209 Use all SSE extensions enabled by @option{-msse2} as well as the new
8210 SSE extensions in Prescott New Instructions. @option{-mpni} also
8211 enables 2 builtin functions, @code{__builtin_ia32_monitor} and
8212 @code{__builtin_ia32_mwait}, for new instructions @code{monitor} and
8216 Attempt to utilize both instruction sets at once. This effectively double the
8217 amount of available registers and on chips with separate execution units for
8218 387 and SSE the execution resources too. Use this option with care, as it is
8219 still experimental, because gcc register allocator does not model separate
8220 functional units well resulting in instable performance.
8223 @item -masm=@var{dialect}
8224 @opindex masm=@var{dialect}
8225 Output asm instructions using selected @var{dialect}. Supported choices are
8226 @samp{intel} or @samp{att} (the default one).
8231 @opindex mno-ieee-fp
8232 Control whether or not the compiler uses IEEE floating point
8233 comparisons. These handle correctly the case where the result of a
8234 comparison is unordered.
8237 @opindex msoft-float
8238 Generate output containing library calls for floating point.
8239 @strong{Warning:} the requisite libraries are not part of GCC@.
8240 Normally the facilities of the machine's usual C compiler are used, but
8241 this can't be done directly in cross-compilation. You must make your
8242 own arrangements to provide suitable library functions for
8245 On machines where a function returns floating point results in the 80387
8246 register stack, some floating point opcodes may be emitted even if
8247 @option{-msoft-float} is used.
8249 @item -mno-fp-ret-in-387
8250 @opindex mno-fp-ret-in-387
8251 Do not use the FPU registers for return values of functions.
8253 The usual calling convention has functions return values of types
8254 @code{float} and @code{double} in an FPU register, even if there
8255 is no FPU@. The idea is that the operating system should emulate
8258 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8259 in ordinary CPU registers instead.
8261 @item -mno-fancy-math-387
8262 @opindex mno-fancy-math-387
8263 Some 387 emulators do not support the @code{sin}, @code{cos} and
8264 @code{sqrt} instructions for the 387. Specify this option to avoid
8265 generating those instructions. This option is the default on FreeBSD,
8266 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8267 indicates that the target cpu will always have an FPU and so the
8268 instruction will not need emulation. As of revision 2.6.1, these
8269 instructions are not generated unless you also use the
8270 @option{-funsafe-math-optimizations} switch.
8272 @item -malign-double
8273 @itemx -mno-align-double
8274 @opindex malign-double
8275 @opindex mno-align-double
8276 Control whether GCC aligns @code{double}, @code{long double}, and
8277 @code{long long} variables on a two word boundary or a one word
8278 boundary. Aligning @code{double} variables on a two word boundary will
8279 produce code that runs somewhat faster on a @samp{Pentium} at the
8280 expense of more memory.
8282 @strong{Warning:} if you use the @option{-malign-double} switch,
8283 structures containing the above types will be aligned differently than
8284 the published application binary interface specifications for the 386
8285 and will not be binary compatible with structures in code compiled
8286 without that switch.
8288 @item -m96bit-long-double
8289 @itemx -m128bit-long-double
8290 @opindex m96bit-long-double
8291 @opindex m128bit-long-double
8292 These switches control the size of @code{long double} type. The i386
8293 application binary interface specifies the size to be 96 bits,
8294 so @option{-m96bit-long-double} is the default in 32 bit mode.
8296 Modern architectures (Pentium and newer) would prefer @code{long double}
8297 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8298 conforming to the ABI, this would not be possible. So specifying a
8299 @option{-m128bit-long-double} will align @code{long double}
8300 to a 16 byte boundary by padding the @code{long double} with an additional
8303 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8304 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8306 Notice that neither of these options enable any extra precision over the x87
8307 standard of 80 bits for a @code{long double}.
8309 @strong{Warning:} if you override the default value for your target ABI, the
8310 structures and arrays containing @code{long double} will change their size as
8311 well as function calling convention for function taking @code{long double}
8312 will be modified. Hence they will not be binary compatible with arrays or
8313 structures in code compiled without that switch.
8317 @itemx -mno-svr3-shlib
8318 @opindex msvr3-shlib
8319 @opindex mno-svr3-shlib
8320 Control whether GCC places uninitialized local variables into the
8321 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8322 into @code{bss}. These options are meaningful only on System V Release 3.
8326 Use a different function-calling convention, in which functions that
8327 take a fixed number of arguments return with the @code{ret} @var{num}
8328 instruction, which pops their arguments while returning. This saves one
8329 instruction in the caller since there is no need to pop the arguments
8332 You can specify that an individual function is called with this calling
8333 sequence with the function attribute @samp{stdcall}. You can also
8334 override the @option{-mrtd} option by using the function attribute
8335 @samp{cdecl}. @xref{Function Attributes}.
8337 @strong{Warning:} this calling convention is incompatible with the one
8338 normally used on Unix, so you cannot use it if you need to call
8339 libraries compiled with the Unix compiler.
8341 Also, you must provide function prototypes for all functions that
8342 take variable numbers of arguments (including @code{printf});
8343 otherwise incorrect code will be generated for calls to those
8346 In addition, seriously incorrect code will result if you call a
8347 function with too many arguments. (Normally, extra arguments are
8348 harmlessly ignored.)
8350 @item -mregparm=@var{num}
8352 Control how many registers are used to pass integer arguments. By
8353 default, no registers are used to pass arguments, and at most 3
8354 registers can be used. You can control this behavior for a specific
8355 function by using the function attribute @samp{regparm}.
8356 @xref{Function Attributes}.
8358 @strong{Warning:} if you use this switch, and
8359 @var{num} is nonzero, then you must build all modules with the same
8360 value, including any libraries. This includes the system libraries and
8363 @item -mpreferred-stack-boundary=@var{num}
8364 @opindex mpreferred-stack-boundary
8365 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8366 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8367 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8368 size (@option{-Os}), in which case the default is the minimum correct
8369 alignment (4 bytes for x86, and 8 bytes for x86-64).
8371 On Pentium and PentiumPro, @code{double} and @code{long double} values
8372 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8373 suffer significant run time performance penalties. On Pentium III, the
8374 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8375 penalties if it is not 16 byte aligned.
8377 To ensure proper alignment of this values on the stack, the stack boundary
8378 must be as aligned as that required by any value stored on the stack.
8379 Further, every function must be generated such that it keeps the stack
8380 aligned. Thus calling a function compiled with a higher preferred
8381 stack boundary from a function compiled with a lower preferred stack
8382 boundary will most likely misalign the stack. It is recommended that
8383 libraries that use callbacks always use the default setting.
8385 This extra alignment does consume extra stack space, and generally
8386 increases code size. Code that is sensitive to stack space usage, such
8387 as embedded systems and operating system kernels, may want to reduce the
8388 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8406 These switches enable or disable the use of built-in functions that allow
8407 direct access to the MMX, SSE and 3Dnow extensions of the instruction set.
8409 @xref{X86 Built-in Functions}, for details of the functions enabled
8410 and disabled by these switches.
8412 To have SSE/SSE2 instructions generated automatically from floating-point
8413 code, see @option{-mfpmath=sse}.
8416 @itemx -mno-push-args
8418 @opindex mno-push-args
8419 Use PUSH operations to store outgoing parameters. This method is shorter
8420 and usually equally fast as method using SUB/MOV operations and is enabled
8421 by default. In some cases disabling it may improve performance because of
8422 improved scheduling and reduced dependencies.
8424 @item -maccumulate-outgoing-args
8425 @opindex maccumulate-outgoing-args
8426 If enabled, the maximum amount of space required for outgoing arguments will be
8427 computed in the function prologue. This is faster on most modern CPUs
8428 because of reduced dependencies, improved scheduling and reduced stack usage
8429 when preferred stack boundary is not equal to 2. The drawback is a notable
8430 increase in code size. This switch implies @option{-mno-push-args}.
8434 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8435 on thread-safe exception handling must compile and link all code with the
8436 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8437 @option{-D_MT}; when linking, it links in a special thread helper library
8438 @option{-lmingwthrd} which cleans up per thread exception handling data.
8440 @item -mno-align-stringops
8441 @opindex mno-align-stringops
8442 Do not align destination of inlined string operations. This switch reduces
8443 code size and improves performance in case the destination is already aligned,
8444 but gcc don't know about it.
8446 @item -minline-all-stringops
8447 @opindex minline-all-stringops
8448 By default GCC inlines string operations only when destination is known to be
8449 aligned at least to 4 byte boundary. This enables more inlining, increase code
8450 size, but may improve performance of code that depends on fast memcpy, strlen
8451 and memset for short lengths.
8453 @item -momit-leaf-frame-pointer
8454 @opindex momit-leaf-frame-pointer
8455 Don't keep the frame pointer in a register for leaf functions. This
8456 avoids the instructions to save, set up and restore frame pointers and
8457 makes an extra register available in leaf functions. The option
8458 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8459 which might make debugging harder.
8461 @item -mtls-direct-seg-refs
8462 @itemx -mno-tls-direct-seg-refs
8463 @opindex mtls-direct-seg-refs
8464 Controls whether TLS variables may be accessed with offsets from the
8465 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8466 or whether the thread base pointer must be added. Whether or not this
8467 is legal depends on the operating system, and whether it maps the
8468 segment to cover the entire TLS area.
8470 For systems that use GNU libc, the default is on.
8473 These @samp{-m} switches are supported in addition to the above
8474 on AMD x86-64 processors in 64-bit environments.
8481 Generate code for a 32-bit or 64-bit environment.
8482 The 32-bit environment sets int, long and pointer to 32 bits and
8483 generates code that runs on any i386 system.
8484 The 64-bit environment sets int to 32 bits and long and pointer
8485 to 64 bits and generates code for AMD's x86-64 architecture.
8488 @opindex no-red-zone
8489 Do not use a so called red zone for x86-64 code. The red zone is mandated
8490 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8491 stack pointer that will not be modified by signal or interrupt handlers
8492 and therefore can be used for temporary data without adjusting the stack
8493 pointer. The flag @option{-mno-red-zone} disables this red zone.
8495 @item -mcmodel=small
8496 @opindex mcmodel=small
8497 Generate code for the small code model: the program and its symbols must
8498 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8499 Programs can be statically or dynamically linked. This is the default
8502 @item -mcmodel=kernel
8503 @opindex mcmodel=kernel
8504 Generate code for the kernel code model. The kernel runs in the
8505 negative 2 GB of the address space.
8506 This model has to be used for Linux kernel code.
8508 @item -mcmodel=medium
8509 @opindex mcmodel=medium
8510 Generate code for the medium model: The program is linked in the lower 2
8511 GB of the address space but symbols can be located anywhere in the
8512 address space. Programs can be statically or dynamically linked, but
8513 building of shared libraries are not supported with the medium model.
8515 @item -mcmodel=large
8516 @opindex mcmodel=large
8517 Generate code for the large model: This model makes no assumptions
8518 about addresses and sizes of sections. Currently GCC does not implement
8523 @subsection HPPA Options
8524 @cindex HPPA Options
8526 These @samp{-m} options are defined for the HPPA family of computers:
8529 @item -march=@var{architecture-type}
8531 Generate code for the specified architecture. The choices for
8532 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8533 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8534 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8535 architecture option for your machine. Code compiled for lower numbered
8536 architectures will run on higher numbered architectures, but not the
8539 PA 2.0 support currently requires gas snapshot 19990413 or later. The
8540 next release of binutils (current is 2.9.1) will probably contain PA 2.0
8544 @itemx -mpa-risc-1-1
8545 @itemx -mpa-risc-2-0
8546 @opindex mpa-risc-1-0
8547 @opindex mpa-risc-1-1
8548 @opindex mpa-risc-2-0
8549 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8552 @opindex mbig-switch
8553 Generate code suitable for big switch tables. Use this option only if
8554 the assembler/linker complain about out of range branches within a switch
8557 @item -mjump-in-delay
8558 @opindex mjump-in-delay
8559 Fill delay slots of function calls with unconditional jump instructions
8560 by modifying the return pointer for the function call to be the target
8561 of the conditional jump.
8563 @item -mdisable-fpregs
8564 @opindex mdisable-fpregs
8565 Prevent floating point registers from being used in any manner. This is
8566 necessary for compiling kernels which perform lazy context switching of
8567 floating point registers. If you use this option and attempt to perform
8568 floating point operations, the compiler will abort.
8570 @item -mdisable-indexing
8571 @opindex mdisable-indexing
8572 Prevent the compiler from using indexing address modes. This avoids some
8573 rather obscure problems when compiling MIG generated code under MACH@.
8575 @item -mno-space-regs
8576 @opindex mno-space-regs
8577 Generate code that assumes the target has no space registers. This allows
8578 GCC to generate faster indirect calls and use unscaled index address modes.
8580 Such code is suitable for level 0 PA systems and kernels.
8582 @item -mfast-indirect-calls
8583 @opindex mfast-indirect-calls
8584 Generate code that assumes calls never cross space boundaries. This
8585 allows GCC to emit code which performs faster indirect calls.
8587 This option will not work in the presence of shared libraries or nested
8590 @item -mlong-load-store
8591 @opindex mlong-load-store
8592 Generate 3-instruction load and store sequences as sometimes required by
8593 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8596 @item -mportable-runtime
8597 @opindex mportable-runtime
8598 Use the portable calling conventions proposed by HP for ELF systems.
8602 Enable the use of assembler directives only GAS understands.
8604 @item -mschedule=@var{cpu-type}
8606 Schedule code according to the constraints for the machine type
8607 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8608 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8609 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8610 proper scheduling option for your machine. The default scheduling is
8614 @opindex mlinker-opt
8615 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8616 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8617 linkers in which they give bogus error messages when linking some programs.
8620 @opindex msoft-float
8621 Generate output containing library calls for floating point.
8622 @strong{Warning:} the requisite libraries are not available for all HPPA
8623 targets. Normally the facilities of the machine's usual C compiler are
8624 used, but this cannot be done directly in cross-compilation. You must make
8625 your own arrangements to provide suitable library functions for
8626 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8627 does provide software floating point support.
8629 @option{-msoft-float} changes the calling convention in the output file;
8630 therefore, it is only useful if you compile @emph{all} of a program with
8631 this option. In particular, you need to compile @file{libgcc.a}, the
8632 library that comes with GCC, with @option{-msoft-float} in order for
8637 Generate the predefine, @code{_SIO}, for server IO. The default is
8638 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8639 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO. These
8640 options are available under HP-UX and HI-UX.
8644 Use GNU ld specific options. This passes @option{-shared} to ld when
8645 building a shared library. It is the default when GCC is configured,
8646 explicitly or implicitly, with the GNU linker. This option does not
8647 have any affect on which ld is called, it only changes what parameters
8648 are passed to that ld. The ld that is called is determined by the
8649 @option{--with-ld} configure option, gcc's program search path, and
8650 finally by the user's @env{PATH}. The linker used by GCC can be printed
8651 using @samp{which `gcc -print-prog-name=ld`}.
8655 Use HP ld specific options. This passes @option{-b} to ld when building
8656 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8657 links. It is the default when GCC is configured, explicitly or
8658 implicitly, with the HP linker. This option does not have any affect on
8659 which ld is called, it only changes what parameters are passed to that
8660 ld. The ld that is called is determined by the @option{--with-ld}
8661 configure option, gcc's program search path, and finally by the user's
8662 @env{PATH}. The linker used by GCC can be printed using @samp{which
8663 `gcc -print-prog-name=ld`}.
8666 @opindex mno-long-calls
8667 Generate code that uses long call sequences. This ensures that a call
8668 is always able to reach linker generated stubs. The default is to generate
8669 long calls only when the distance from the call site to the beginning
8670 of the function or translation unit, as the case may be, exceeds a
8671 predefined limit set by the branch type being used. The limits for
8672 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8673 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8676 Distances are measured from the beginning of functions when using the
8677 @option{-ffunction-sections} option, or when using the @option{-mgas}
8678 and @option{-mno-portable-runtime} options together under HP-UX with
8681 It is normally not desirable to use this option as it will degrade
8682 performance. However, it may be useful in large applications,
8683 particularly when partial linking is used to build the application.
8685 The types of long calls used depends on the capabilities of the
8686 assembler and linker, and the type of code being generated. The
8687 impact on systems that support long absolute calls, and long pic
8688 symbol-difference or pc-relative calls should be relatively small.
8689 However, an indirect call is used on 32-bit ELF systems in pic code
8690 and it is quite long.
8694 Suppress the generation of link options to search libdld.sl when the
8695 @option{-static} option is specified on HP-UX 10 and later.
8699 The HP-UX implementation of setlocale in libc has a dependency on
8700 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8701 when the @option{-static} option is specified, special link options
8702 are needed to resolve this dependency.
8704 On HP-UX 10 and later, the GCC driver adds the necessary options to
8705 link with libdld.sl when the @option{-static} option is specified.
8706 This causes the resulting binary to be dynamic. On the 64-bit port,
8707 the linkers generate dynamic binaries by default in any case. The
8708 @option{-nolibdld} option can be used to prevent the GCC driver from
8709 adding these link options.
8713 Add support for multithreading with the @dfn{dce thread} library
8714 under HP-UX. This option sets flags for both the preprocessor and
8718 @node Intel 960 Options
8719 @subsection Intel 960 Options
8721 These @samp{-m} options are defined for the Intel 960 implementations:
8724 @item -m@var{cpu-type}
8732 Assume the defaults for the machine type @var{cpu-type} for some of
8733 the other options, including instruction scheduling, floating point
8734 support, and addressing modes. The choices for @var{cpu-type} are
8735 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
8736 @samp{sa}, and @samp{sb}.
8743 @opindex msoft-float
8744 The @option{-mnumerics} option indicates that the processor does support
8745 floating-point instructions. The @option{-msoft-float} option indicates
8746 that floating-point support should not be assumed.
8748 @item -mleaf-procedures
8749 @itemx -mno-leaf-procedures
8750 @opindex mleaf-procedures
8751 @opindex mno-leaf-procedures
8752 Do (or do not) attempt to alter leaf procedures to be callable with the
8753 @code{bal} instruction as well as @code{call}. This will result in more
8754 efficient code for explicit calls when the @code{bal} instruction can be
8755 substituted by the assembler or linker, but less efficient code in other
8756 cases, such as calls via function pointers, or using a linker that doesn't
8757 support this optimization.
8760 @itemx -mno-tail-call
8762 @opindex mno-tail-call
8763 Do (or do not) make additional attempts (beyond those of the
8764 machine-independent portions of the compiler) to optimize tail-recursive
8765 calls into branches. You may not want to do this because the detection of
8766 cases where this is not valid is not totally complete. The default is
8767 @option{-mno-tail-call}.
8769 @item -mcomplex-addr
8770 @itemx -mno-complex-addr
8771 @opindex mcomplex-addr
8772 @opindex mno-complex-addr
8773 Assume (or do not assume) that the use of a complex addressing mode is a
8774 win on this implementation of the i960. Complex addressing modes may not
8775 be worthwhile on the K-series, but they definitely are on the C-series.
8776 The default is currently @option{-mcomplex-addr} for all processors except
8780 @itemx -mno-code-align
8781 @opindex mcode-align
8782 @opindex mno-code-align
8783 Align code to 8-byte boundaries for faster fetching (or don't bother).
8784 Currently turned on by default for C-series implementations only.
8787 @item -mclean-linkage
8788 @itemx -mno-clean-linkage
8789 @opindex mclean-linkage
8790 @opindex mno-clean-linkage
8791 These options are not fully implemented.
8795 @itemx -mic2.0-compat
8796 @itemx -mic3.0-compat
8798 @opindex mic2.0-compat
8799 @opindex mic3.0-compat
8800 Enable compatibility with iC960 v2.0 or v3.0.
8804 @opindex masm-compat
8806 Enable compatibility with the iC960 assembler.
8808 @item -mstrict-align
8809 @itemx -mno-strict-align
8810 @opindex mstrict-align
8811 @opindex mno-strict-align
8812 Do not permit (do permit) unaligned accesses.
8816 Enable structure-alignment compatibility with Intel's gcc release version
8817 1.3 (based on gcc 1.37). This option implies @option{-mstrict-align}.
8819 @item -mlong-double-64
8820 @opindex mlong-double-64
8821 Implement type @samp{long double} as 64-bit floating point numbers.
8822 Without the option @samp{long double} is implemented by 80-bit
8823 floating point numbers. The only reason we have it because there is
8824 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
8825 is only useful for people using soft-float targets. Otherwise, we
8826 should recommend against use of it.
8830 @node DEC Alpha Options
8831 @subsection DEC Alpha Options
8833 These @samp{-m} options are defined for the DEC Alpha implementations:
8836 @item -mno-soft-float
8838 @opindex mno-soft-float
8839 @opindex msoft-float
8840 Use (do not use) the hardware floating-point instructions for
8841 floating-point operations. When @option{-msoft-float} is specified,
8842 functions in @file{libgcc.a} will be used to perform floating-point
8843 operations. Unless they are replaced by routines that emulate the
8844 floating-point operations, or compiled in such a way as to call such
8845 emulations routines, these routines will issue floating-point
8846 operations. If you are compiling for an Alpha without floating-point
8847 operations, you must ensure that the library is built so as not to call
8850 Note that Alpha implementations without floating-point operations are
8851 required to have floating-point registers.
8856 @opindex mno-fp-regs
8857 Generate code that uses (does not use) the floating-point register set.
8858 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8859 register set is not used, floating point operands are passed in integer
8860 registers as if they were integers and floating-point results are passed
8861 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8862 so any function with a floating-point argument or return value called by code
8863 compiled with @option{-mno-fp-regs} must also be compiled with that
8866 A typical use of this option is building a kernel that does not use,
8867 and hence need not save and restore, any floating-point registers.
8871 The Alpha architecture implements floating-point hardware optimized for
8872 maximum performance. It is mostly compliant with the IEEE floating
8873 point standard. However, for full compliance, software assistance is
8874 required. This option generates code fully IEEE compliant code
8875 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8876 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8877 defined during compilation. The resulting code is less efficient but is
8878 able to correctly support denormalized numbers and exceptional IEEE
8879 values such as not-a-number and plus/minus infinity. Other Alpha
8880 compilers call this option @option{-ieee_with_no_inexact}.
8882 @item -mieee-with-inexact
8883 @opindex mieee-with-inexact
8884 This is like @option{-mieee} except the generated code also maintains
8885 the IEEE @var{inexact-flag}. Turning on this option causes the
8886 generated code to implement fully-compliant IEEE math. In addition to
8887 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8888 macro. On some Alpha implementations the resulting code may execute
8889 significantly slower than the code generated by default. Since there is
8890 very little code that depends on the @var{inexact-flag}, you should
8891 normally not specify this option. Other Alpha compilers call this
8892 option @option{-ieee_with_inexact}.
8894 @item -mfp-trap-mode=@var{trap-mode}
8895 @opindex mfp-trap-mode
8896 This option controls what floating-point related traps are enabled.
8897 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8898 The trap mode can be set to one of four values:
8902 This is the default (normal) setting. The only traps that are enabled
8903 are the ones that cannot be disabled in software (e.g., division by zero
8907 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8911 Like @samp{su}, but the instructions are marked to be safe for software
8912 completion (see Alpha architecture manual for details).
8915 Like @samp{su}, but inexact traps are enabled as well.
8918 @item -mfp-rounding-mode=@var{rounding-mode}
8919 @opindex mfp-rounding-mode
8920 Selects the IEEE rounding mode. Other Alpha compilers call this option
8921 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8926 Normal IEEE rounding mode. Floating point numbers are rounded towards
8927 the nearest machine number or towards the even machine number in case
8931 Round towards minus infinity.
8934 Chopped rounding mode. Floating point numbers are rounded towards zero.
8937 Dynamic rounding mode. A field in the floating point control register
8938 (@var{fpcr}, see Alpha architecture reference manual) controls the
8939 rounding mode in effect. The C library initializes this register for
8940 rounding towards plus infinity. Thus, unless your program modifies the
8941 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8944 @item -mtrap-precision=@var{trap-precision}
8945 @opindex mtrap-precision
8946 In the Alpha architecture, floating point traps are imprecise. This
8947 means without software assistance it is impossible to recover from a
8948 floating trap and program execution normally needs to be terminated.
8949 GCC can generate code that can assist operating system trap handlers
8950 in determining the exact location that caused a floating point trap.
8951 Depending on the requirements of an application, different levels of
8952 precisions can be selected:
8956 Program precision. This option is the default and means a trap handler
8957 can only identify which program caused a floating point exception.
8960 Function precision. The trap handler can determine the function that
8961 caused a floating point exception.
8964 Instruction precision. The trap handler can determine the exact
8965 instruction that caused a floating point exception.
8968 Other Alpha compilers provide the equivalent options called
8969 @option{-scope_safe} and @option{-resumption_safe}.
8971 @item -mieee-conformant
8972 @opindex mieee-conformant
8973 This option marks the generated code as IEEE conformant. You must not
8974 use this option unless you also specify @option{-mtrap-precision=i} and either
8975 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8976 is to emit the line @samp{.eflag 48} in the function prologue of the
8977 generated assembly file. Under DEC Unix, this has the effect that
8978 IEEE-conformant math library routines will be linked in.
8980 @item -mbuild-constants
8981 @opindex mbuild-constants
8982 Normally GCC examines a 32- or 64-bit integer constant to
8983 see if it can construct it from smaller constants in two or three
8984 instructions. If it cannot, it will output the constant as a literal and
8985 generate code to load it from the data segment at runtime.
8987 Use this option to require GCC to construct @emph{all} integer constants
8988 using code, even if it takes more instructions (the maximum is six).
8990 You would typically use this option to build a shared library dynamic
8991 loader. Itself a shared library, it must relocate itself in memory
8992 before it can find the variables and constants in its own data segment.
8998 Select whether to generate code to be assembled by the vendor-supplied
8999 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9017 Indicate whether GCC should generate code to use the optional BWX,
9018 CIX, FIX and MAX instruction sets. The default is to use the instruction
9019 sets supported by the CPU type specified via @option{-mcpu=} option or that
9020 of the CPU on which GCC was built if none was specified.
9025 @opindex mfloat-ieee
9026 Generate code that uses (does not use) VAX F and G floating point
9027 arithmetic instead of IEEE single and double precision.
9029 @item -mexplicit-relocs
9030 @itemx -mno-explicit-relocs
9031 @opindex mexplicit-relocs
9032 @opindex mno-explicit-relocs
9033 Older Alpha assemblers provided no way to generate symbol relocations
9034 except via assembler macros. Use of these macros does not allow
9035 optimal instruction scheduling. GNU binutils as of version 2.12
9036 supports a new syntax that allows the compiler to explicitly mark
9037 which relocations should apply to which instructions. This option
9038 is mostly useful for debugging, as GCC detects the capabilities of
9039 the assembler when it is built and sets the default accordingly.
9043 @opindex msmall-data
9044 @opindex mlarge-data
9045 When @option{-mexplicit-relocs} is in effect, static data is
9046 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9047 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9048 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9049 16-bit relocations off of the @code{$gp} register. This limits the
9050 size of the small data area to 64KB, but allows the variables to be
9051 directly accessed via a single instruction.
9053 The default is @option{-mlarge-data}. With this option the data area
9054 is limited to just below 2GB. Programs that require more than 2GB of
9055 data must use @code{malloc} or @code{mmap} to allocate the data in the
9056 heap instead of in the program's data segment.
9058 When generating code for shared libraries, @option{-fpic} implies
9059 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9063 @opindex msmall-text
9064 @opindex mlarge-text
9065 When @option{-msmall-text} is used, the compiler assumes that the
9066 code of the entire program (or shared library) fits in 4MB, and is
9067 thus reachable with a branch instruction. When @option{-msmall-data}
9068 is used, the compiler can assume that all local symbols share the
9069 same @code{$gp} value, and thus reduce the number of instructions
9070 required for a function call from 4 to 1.
9072 The default is @option{-mlarge-text}.
9074 @item -mcpu=@var{cpu_type}
9076 Set the instruction set and instruction scheduling parameters for
9077 machine type @var{cpu_type}. You can specify either the @samp{EV}
9078 style name or the corresponding chip number. GCC supports scheduling
9079 parameters for the EV4, EV5 and EV6 family of processors and will
9080 choose the default values for the instruction set from the processor
9081 you specify. If you do not specify a processor type, GCC will default
9082 to the processor on which the compiler was built.
9084 Supported values for @var{cpu_type} are
9090 Schedules as an EV4 and has no instruction set extensions.
9094 Schedules as an EV5 and has no instruction set extensions.
9098 Schedules as an EV5 and supports the BWX extension.
9103 Schedules as an EV5 and supports the BWX and MAX extensions.
9107 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9111 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9114 @item -mtune=@var{cpu_type}
9116 Set only the instruction scheduling parameters for machine type
9117 @var{cpu_type}. The instruction set is not changed.
9119 @item -mmemory-latency=@var{time}
9120 @opindex mmemory-latency
9121 Sets the latency the scheduler should assume for typical memory
9122 references as seen by the application. This number is highly
9123 dependent on the memory access patterns used by the application
9124 and the size of the external cache on the machine.
9126 Valid options for @var{time} are
9130 A decimal number representing clock cycles.
9136 The compiler contains estimates of the number of clock cycles for
9137 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9138 (also called Dcache, Scache, and Bcache), as well as to main memory.
9139 Note that L3 is only valid for EV5.
9144 @node DEC Alpha/VMS Options
9145 @subsection DEC Alpha/VMS Options
9147 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9150 @item -mvms-return-codes
9151 @opindex mvms-return-codes
9152 Return VMS condition codes from main. The default is to return POSIX
9153 style condition (e.g.@ error) codes.
9156 @node H8/300 Options
9157 @subsection H8/300 Options
9159 These @samp{-m} options are defined for the H8/300 implementations:
9164 Shorten some address references at link time, when possible; uses the
9165 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9166 ld, Using ld}, for a fuller description.
9170 Generate code for the H8/300H@.
9174 Generate code for the H8S@.
9178 Generate code for the H8S and H8/300H in the normal mode. This switch
9179 must be used either with -mh or -ms.
9183 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9187 Make @code{int} data 32 bits by default.
9191 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9192 The default for the H8/300H and H8S is to align longs and floats on 4
9194 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9195 This option has no effect on the H8/300.
9199 @subsection SH Options
9201 These @samp{-m} options are defined for the SH implementations:
9206 Generate code for the SH1.
9210 Generate code for the SH2.
9213 Generate code for the SH2e.
9217 Generate code for the SH3.
9221 Generate code for the SH3e.
9225 Generate code for the SH4 without a floating-point unit.
9227 @item -m4-single-only
9228 @opindex m4-single-only
9229 Generate code for the SH4 with a floating-point unit that only
9230 supports single-precision arithmetic.
9234 Generate code for the SH4 assuming the floating-point unit is in
9235 single-precision mode by default.
9239 Generate code for the SH4.
9243 Compile code for the processor in big endian mode.
9247 Compile code for the processor in little endian mode.
9251 Align doubles at 64-bit boundaries. Note that this changes the calling
9252 conventions, and thus some functions from the standard C library will
9253 not work unless you recompile it first with @option{-mdalign}.
9257 Shorten some address references at link time, when possible; uses the
9258 linker option @option{-relax}.
9262 Use 32-bit offsets in @code{switch} tables. The default is to use
9267 Enable the use of the instruction @code{fmovd}.
9271 Comply with the calling conventions defined by Renesas.
9275 Mark the @code{MAC} register as call-clobbered, even if
9276 @option{-mhitachi} is given.
9280 Increase IEEE-compliance of floating-point code.
9284 Dump instruction size and location in the assembly code.
9288 This option is deprecated. It pads structures to multiple of 4 bytes,
9289 which is incompatible with the SH ABI@.
9293 Optimize for space instead of speed. Implied by @option{-Os}.
9297 When generating position-independent code, emit function calls using
9298 the Global Offset Table instead of the Procedure Linkage Table.
9302 Generate a library function call to invalidate instruction cache
9303 entries, after fixing up a trampoline. This library function call
9304 doesn't assume it can write to the whole memory address space. This
9305 is the default when the target is @code{sh-*-linux*}.
9308 @node System V Options
9309 @subsection Options for System V
9311 These additional options are available on System V Release 4 for
9312 compatibility with other compilers on those systems:
9317 Create a shared object.
9318 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
9322 Identify the versions of each tool used by the compiler, in a
9323 @code{.ident} assembler directive in the output.
9327 Refrain from adding @code{.ident} directives to the output file (this is
9330 @item -YP,@var{dirs}
9332 Search the directories @var{dirs}, and no others, for libraries
9333 specified with @option{-l}.
9337 Look in the directory @var{dir} to find the M4 preprocessor.
9338 The assembler uses this option.
9339 @c This is supposed to go with a -Yd for predefined M4 macro files, but
9340 @c the generic assembler that comes with Solaris takes just -Ym.
9343 @node TMS320C3x/C4x Options
9344 @subsection TMS320C3x/C4x Options
9345 @cindex TMS320C3x/C4x Options
9347 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
9351 @item -mcpu=@var{cpu_type}
9353 Set the instruction set, register set, and instruction scheduling
9354 parameters for machine type @var{cpu_type}. Supported values for
9355 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
9356 @samp{c44}. The default is @samp{c40} to generate code for the
9361 @itemx -msmall-memory
9363 @opindex mbig-memory
9365 @opindex msmall-memory
9367 Generates code for the big or small memory model. The small memory
9368 model assumed that all data fits into one 64K word page. At run-time
9369 the data page (DP) register must be set to point to the 64K page
9370 containing the .bss and .data program sections. The big memory model is
9371 the default and requires reloading of the DP register for every direct
9378 Allow (disallow) allocation of general integer operands into the block
9385 Enable (disable) generation of code using decrement and branch,
9386 DBcond(D), instructions. This is enabled by default for the C4x. To be
9387 on the safe side, this is disabled for the C3x, since the maximum
9388 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
9389 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
9390 that it can utilize the decrement and branch instruction, but will give
9391 up if there is more than one memory reference in the loop. Thus a loop
9392 where the loop counter is decremented can generate slightly more
9393 efficient code, in cases where the RPTB instruction cannot be utilized.
9395 @item -mdp-isr-reload
9397 @opindex mdp-isr-reload
9399 Force the DP register to be saved on entry to an interrupt service
9400 routine (ISR), reloaded to point to the data section, and restored on
9401 exit from the ISR@. This should not be required unless someone has
9402 violated the small memory model by modifying the DP register, say within
9409 For the C3x use the 24-bit MPYI instruction for integer multiplies
9410 instead of a library call to guarantee 32-bit results. Note that if one
9411 of the operands is a constant, then the multiplication will be performed
9412 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
9413 then squaring operations are performed inline instead of a library call.
9416 @itemx -mno-fast-fix
9418 @opindex mno-fast-fix
9419 The C3x/C4x FIX instruction to convert a floating point value to an
9420 integer value chooses the nearest integer less than or equal to the
9421 floating point value rather than to the nearest integer. Thus if the
9422 floating point number is negative, the result will be incorrectly
9423 truncated an additional code is necessary to detect and correct this
9424 case. This option can be used to disable generation of the additional
9425 code required to correct the result.
9431 Enable (disable) generation of repeat block sequences using the RPTB
9432 instruction for zero overhead looping. The RPTB construct is only used
9433 for innermost loops that do not call functions or jump across the loop
9434 boundaries. There is no advantage having nested RPTB loops due to the
9435 overhead required to save and restore the RC, RS, and RE registers.
9436 This is enabled by default with @option{-O2}.
9438 @item -mrpts=@var{count}
9442 Enable (disable) the use of the single instruction repeat instruction
9443 RPTS@. If a repeat block contains a single instruction, and the loop
9444 count can be guaranteed to be less than the value @var{count}, GCC will
9445 emit a RPTS instruction instead of a RPTB@. If no value is specified,
9446 then a RPTS will be emitted even if the loop count cannot be determined
9447 at compile time. Note that the repeated instruction following RPTS does
9448 not have to be reloaded from memory each iteration, thus freeing up the
9449 CPU buses for operands. However, since interrupts are blocked by this
9450 instruction, it is disabled by default.
9452 @item -mloop-unsigned
9453 @itemx -mno-loop-unsigned
9454 @opindex mloop-unsigned
9455 @opindex mno-loop-unsigned
9456 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
9457 is @math{2^{31} + 1} since these instructions test if the iteration count is
9458 negative to terminate the loop. If the iteration count is unsigned
9459 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
9460 exceeded. This switch allows an unsigned iteration count.
9464 Try to emit an assembler syntax that the TI assembler (asm30) is happy
9465 with. This also enforces compatibility with the API employed by the TI
9466 C3x C compiler. For example, long doubles are passed as structures
9467 rather than in floating point registers.
9473 Generate code that uses registers (stack) for passing arguments to functions.
9474 By default, arguments are passed in registers where possible rather
9475 than by pushing arguments on to the stack.
9477 @item -mparallel-insns
9478 @itemx -mno-parallel-insns
9479 @opindex mparallel-insns
9480 @opindex mno-parallel-insns
9481 Allow the generation of parallel instructions. This is enabled by
9482 default with @option{-O2}.
9484 @item -mparallel-mpy
9485 @itemx -mno-parallel-mpy
9486 @opindex mparallel-mpy
9487 @opindex mno-parallel-mpy
9488 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
9489 provided @option{-mparallel-insns} is also specified. These instructions have
9490 tight register constraints which can pessimize the code generation
9496 @subsection V850 Options
9497 @cindex V850 Options
9499 These @samp{-m} options are defined for V850 implementations:
9503 @itemx -mno-long-calls
9504 @opindex mlong-calls
9505 @opindex mno-long-calls
9506 Treat all calls as being far away (near). If calls are assumed to be
9507 far away, the compiler will always load the functions address up into a
9508 register, and call indirect through the pointer.
9514 Do not optimize (do optimize) basic blocks that use the same index
9515 pointer 4 or more times to copy pointer into the @code{ep} register, and
9516 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
9517 option is on by default if you optimize.
9519 @item -mno-prolog-function
9520 @itemx -mprolog-function
9521 @opindex mno-prolog-function
9522 @opindex mprolog-function
9523 Do not use (do use) external functions to save and restore registers
9524 at the prologue and epilogue of a function. The external functions
9525 are slower, but use less code space if more than one function saves
9526 the same number of registers. The @option{-mprolog-function} option
9527 is on by default if you optimize.
9531 Try to make the code as small as possible. At present, this just turns
9532 on the @option{-mep} and @option{-mprolog-function} options.
9536 Put static or global variables whose size is @var{n} bytes or less into
9537 the tiny data area that register @code{ep} points to. The tiny data
9538 area can hold up to 256 bytes in total (128 bytes for byte references).
9542 Put static or global variables whose size is @var{n} bytes or less into
9543 the small data area that register @code{gp} points to. The small data
9544 area can hold up to 64 kilobytes.
9548 Put static or global variables whose size is @var{n} bytes or less into
9549 the first 32 kilobytes of memory.
9553 Specify that the target processor is the V850.
9556 @opindex mbig-switch
9557 Generate code suitable for big switch tables. Use this option only if
9558 the assembler/linker complain about out of range branches within a switch
9563 This option will cause r2 and r5 to be used in the code generated by
9564 the compiler. This setting is the default.
9567 @opindex mno-app-regs
9568 This option will cause r2 and r5 to be treated as fixed registers.
9572 Specify that the target processor is the V850E1. The preprocessor
9573 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
9574 this option is used.
9578 Specify that the target processor is the V850E. The preprocessor
9579 constant @samp{__v850e__} will be defined if this option is used.
9581 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
9582 are defined then a default target processor will be chosen and the
9583 relevant @samp{__v850*__} preprocessor constant will be defined.
9585 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
9586 defined, regardless of which processor variant is the target.
9588 @item -mdisable-callt
9589 @opindex mdisable-callt
9590 This option will suppress generation of the CALLT instruction for the
9591 v850e and v850e1 flavors of the v850 architecture. The default is
9592 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
9597 @subsection ARC Options
9600 These options are defined for ARC implementations:
9605 Compile code for little endian mode. This is the default.
9609 Compile code for big endian mode.
9612 @opindex mmangle-cpu
9613 Prepend the name of the cpu to all public symbol names.
9614 In multiple-processor systems, there are many ARC variants with different
9615 instruction and register set characteristics. This flag prevents code
9616 compiled for one cpu to be linked with code compiled for another.
9617 No facility exists for handling variants that are ``almost identical''.
9618 This is an all or nothing option.
9620 @item -mcpu=@var{cpu}
9622 Compile code for ARC variant @var{cpu}.
9623 Which variants are supported depend on the configuration.
9624 All variants support @option{-mcpu=base}, this is the default.
9626 @item -mtext=@var{text-section}
9627 @itemx -mdata=@var{data-section}
9628 @itemx -mrodata=@var{readonly-data-section}
9632 Put functions, data, and readonly data in @var{text-section},
9633 @var{data-section}, and @var{readonly-data-section} respectively
9634 by default. This can be overridden with the @code{section} attribute.
9635 @xref{Variable Attributes}.
9640 @subsection NS32K Options
9641 @cindex NS32K options
9643 These are the @samp{-m} options defined for the 32000 series. The default
9644 values for these options depends on which style of 32000 was selected when
9645 the compiler was configured; the defaults for the most common choices are
9653 Generate output for a 32032. This is the default
9654 when the compiler is configured for 32032 and 32016 based systems.
9660 Generate output for a 32332. This is the default
9661 when the compiler is configured for 32332-based systems.
9667 Generate output for a 32532. This is the default
9668 when the compiler is configured for 32532-based systems.
9672 Generate output containing 32081 instructions for floating point.
9673 This is the default for all systems.
9677 Generate output containing 32381 instructions for floating point. This
9678 also implies @option{-m32081}. The 32381 is only compatible with the 32332
9679 and 32532 cpus. This is the default for the pc532-netbsd configuration.
9683 Try and generate multiply-add floating point instructions @code{polyF}
9684 and @code{dotF}. This option is only available if the @option{-m32381}
9685 option is in effect. Using these instructions requires changes to
9686 register allocation which generally has a negative impact on
9687 performance. This option should only be enabled when compiling code
9688 particularly likely to make heavy use of multiply-add instructions.
9691 @opindex mnomulti-add
9692 Do not try and generate multiply-add floating point instructions
9693 @code{polyF} and @code{dotF}. This is the default on all platforms.
9696 @opindex msoft-float
9697 Generate output containing library calls for floating point.
9698 @strong{Warning:} the requisite libraries may not be available.
9700 @item -mieee-compare
9701 @itemx -mno-ieee-compare
9702 @opindex mieee-compare
9703 @opindex mno-ieee-compare
9704 Control whether or not the compiler uses IEEE floating point
9705 comparisons. These handle correctly the case where the result of a
9706 comparison is unordered.
9707 @strong{Warning:} the requisite kernel support may not be available.
9710 @opindex mnobitfield
9711 Do not use the bit-field instructions. On some machines it is faster to
9712 use shifting and masking operations. This is the default for the pc532.
9716 Do use the bit-field instructions. This is the default for all platforms
9721 Use a different function-calling convention, in which functions
9722 that take a fixed number of arguments return pop their
9723 arguments on return with the @code{ret} instruction.
9725 This calling convention is incompatible with the one normally
9726 used on Unix, so you cannot use it if you need to call libraries
9727 compiled with the Unix compiler.
9729 Also, you must provide function prototypes for all functions that
9730 take variable numbers of arguments (including @code{printf});
9731 otherwise incorrect code will be generated for calls to those
9734 In addition, seriously incorrect code will result if you call a
9735 function with too many arguments. (Normally, extra arguments are
9736 harmlessly ignored.)
9738 This option takes its name from the 680x0 @code{rtd} instruction.
9743 Use a different function-calling convention where the first two arguments
9744 are passed in registers.
9746 This calling convention is incompatible with the one normally
9747 used on Unix, so you cannot use it if you need to call libraries
9748 compiled with the Unix compiler.
9751 @opindex mnoregparam
9752 Do not pass any arguments in registers. This is the default for all
9757 It is OK to use the sb as an index register which is always loaded with
9758 zero. This is the default for the pc532-netbsd target.
9762 The sb register is not available for use or has not been initialized to
9763 zero by the run time system. This is the default for all targets except
9764 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
9765 @option{-fpic} is set.
9769 Many ns32000 series addressing modes use displacements of up to 512MB@.
9770 If an address is above 512MB then displacements from zero can not be used.
9771 This option causes code to be generated which can be loaded above 512MB@.
9772 This may be useful for operating systems or ROM code.
9776 Assume code will be loaded in the first 512MB of virtual address space.
9777 This is the default for all platforms.
9783 @subsection AVR Options
9786 These options are defined for AVR implementations:
9789 @item -mmcu=@var{mcu}
9791 Specify ATMEL AVR instruction set or MCU type.
9793 Instruction set avr1 is for the minimal AVR core, not supported by the C
9794 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9795 attiny11, attiny12, attiny15, attiny28).
9797 Instruction set avr2 (default) is for the classic AVR core with up to
9798 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9799 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9800 at90c8534, at90s8535).
9802 Instruction set avr3 is for the classic AVR core with up to 128K program
9803 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9805 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9806 memory space (MCU types: atmega8, atmega83, atmega85).
9808 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9809 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9810 atmega64, atmega128, at43usb355, at94k).
9814 Output instruction sizes to the asm file.
9816 @item -minit-stack=@var{N}
9817 @opindex minit-stack
9818 Specify the initial stack address, which may be a symbol or numeric value,
9819 @samp{__stack} is the default.
9821 @item -mno-interrupts
9822 @opindex mno-interrupts
9823 Generated code is not compatible with hardware interrupts.
9824 Code size will be smaller.
9826 @item -mcall-prologues
9827 @opindex mcall-prologues
9828 Functions prologues/epilogues expanded as call to appropriate
9829 subroutines. Code size will be smaller.
9831 @item -mno-tablejump
9832 @opindex mno-tablejump
9833 Do not generate tablejump insns which sometimes increase code size.
9836 @opindex mtiny-stack
9837 Change only the low 8 bits of the stack pointer.
9841 @subsection MCore Options
9842 @cindex MCore options
9844 These are the @samp{-m} options defined for the Motorola M*Core
9852 @opindex mno-hardlit
9853 Inline constants into the code stream if it can be done in two
9854 instructions or less.
9860 Use the divide instruction. (Enabled by default).
9862 @item -mrelax-immediate
9863 @itemx -mno-relax-immediate
9864 @opindex mrelax-immediate
9865 @opindex mno-relax-immediate
9866 Allow arbitrary sized immediates in bit operations.
9868 @item -mwide-bitfields
9869 @itemx -mno-wide-bitfields
9870 @opindex mwide-bitfields
9871 @opindex mno-wide-bitfields
9872 Always treat bit-fields as int-sized.
9874 @item -m4byte-functions
9875 @itemx -mno-4byte-functions
9876 @opindex m4byte-functions
9877 @opindex mno-4byte-functions
9878 Force all functions to be aligned to a four byte boundary.
9880 @item -mcallgraph-data
9881 @itemx -mno-callgraph-data
9882 @opindex mcallgraph-data
9883 @opindex mno-callgraph-data
9884 Emit callgraph information.
9887 @itemx -mno-slow-bytes
9888 @opindex mslow-bytes
9889 @opindex mno-slow-bytes
9890 Prefer word access when reading byte quantities.
9892 @item -mlittle-endian
9894 @opindex mlittle-endian
9895 @opindex mbig-endian
9896 Generate code for a little endian target.
9902 Generate code for the 210 processor.
9906 @subsection IA-64 Options
9907 @cindex IA-64 Options
9909 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9913 @opindex mbig-endian
9914 Generate code for a big endian target. This is the default for HP-UX@.
9916 @item -mlittle-endian
9917 @opindex mlittle-endian
9918 Generate code for a little endian target. This is the default for AIX5
9925 Generate (or don't) code for the GNU assembler. This is the default.
9926 @c Also, this is the default if the configure option @option{--with-gnu-as}
9933 Generate (or don't) code for the GNU linker. This is the default.
9934 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9939 Generate code that does not use a global pointer register. The result
9940 is not position independent code, and violates the IA-64 ABI@.
9942 @item -mvolatile-asm-stop
9943 @itemx -mno-volatile-asm-stop
9944 @opindex mvolatile-asm-stop
9945 @opindex mno-volatile-asm-stop
9946 Generate (or don't) a stop bit immediately before and after volatile asm
9951 Generate code that works around Itanium B step errata.
9953 @item -mregister-names
9954 @itemx -mno-register-names
9955 @opindex mregister-names
9956 @opindex mno-register-names
9957 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9958 the stacked registers. This may make assembler output more readable.
9964 Disable (or enable) optimizations that use the small data section. This may
9965 be useful for working around optimizer bugs.
9968 @opindex mconstant-gp
9969 Generate code that uses a single constant global pointer value. This is
9970 useful when compiling kernel code.
9974 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9975 This is useful when compiling firmware code.
9977 @item -minline-float-divide-min-latency
9978 @opindex minline-float-divide-min-latency
9979 Generate code for inline divides of floating point values
9980 using the minimum latency algorithm.
9982 @item -minline-float-divide-max-throughput
9983 @opindex minline-float-divide-max-throughput
9984 Generate code for inline divides of floating point values
9985 using the maximum throughput algorithm.
9987 @item -minline-int-divide-min-latency
9988 @opindex minline-int-divide-min-latency
9989 Generate code for inline divides of integer values
9990 using the minimum latency algorithm.
9992 @item -minline-int-divide-max-throughput
9993 @opindex minline-int-divide-max-throughput
9994 Generate code for inline divides of integer values
9995 using the maximum throughput algorithm.
9997 @item -mno-dwarf2-asm
9999 @opindex mno-dwarf2-asm
10000 @opindex mdwarf2-asm
10001 Don't (or do) generate assembler code for the DWARF2 line number debugging
10002 info. This may be useful when not using the GNU assembler.
10004 @item -mfixed-range=@var{register-range}
10005 @opindex mfixed-range
10006 Generate code treating the given register range as fixed registers.
10007 A fixed register is one that the register allocator can not use. This is
10008 useful when compiling kernel code. A register range is specified as
10009 two registers separated by a dash. Multiple register ranges can be
10010 specified separated by a comma.
10012 @item -mearly-stop-bits
10013 @itemx -mno-early-stop-bits
10014 @opindex mearly-stop-bits
10015 @opindex mno-early-stop-bits
10016 Allow stop bits to be placed earlier than immediately preceding the
10017 instruction that triggered the stop bit. This can improve instruction
10018 scheduling, but does not always do so.
10022 @subsection D30V Options
10023 @cindex D30V Options
10025 These @samp{-m} options are defined for D30V implementations:
10030 Link the @samp{.text}, @samp{.data}, @samp{.bss}, @samp{.strings},
10031 @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections into external
10032 memory, which starts at location @code{0x80000000}.
10035 @opindex mextmemory
10036 Same as the @option{-mextmem} switch.
10040 Link the @samp{.text} section into onchip text memory, which starts at
10041 location @code{0x0}. Also link @samp{.data}, @samp{.bss},
10042 @samp{.strings}, @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections
10043 into onchip data memory, which starts at location @code{0x20000000}.
10045 @item -mno-asm-optimize
10046 @itemx -masm-optimize
10047 @opindex mno-asm-optimize
10048 @opindex masm-optimize
10049 Disable (enable) passing @option{-O} to the assembler when optimizing.
10050 The assembler uses the @option{-O} option to automatically parallelize
10051 adjacent short instructions where possible.
10053 @item -mbranch-cost=@var{n}
10054 @opindex mbranch-cost
10055 Increase the internal costs of branches to @var{n}. Higher costs means
10056 that the compiler will issue more instructions to avoid doing a branch.
10059 @item -mcond-exec=@var{n}
10060 @opindex mcond-exec
10061 Specify the maximum number of conditionally executed instructions that
10062 replace a branch. The default is 4.
10065 @node S/390 and zSeries Options
10066 @subsection S/390 and zSeries Options
10067 @cindex S/390 and zSeries Options
10069 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
10073 @itemx -msoft-float
10074 @opindex mhard-float
10075 @opindex msoft-float
10076 Use (do not use) the hardware floating-point instructions and registers
10077 for floating-point operations. When @option{-msoft-float} is specified,
10078 functions in @file{libgcc.a} will be used to perform floating-point
10079 operations. When @option{-mhard-float} is specified, the compiler
10080 generates IEEE floating-point instructions. This is the default.
10083 @itemx -mno-backchain
10084 @opindex mbackchain
10085 @opindex mno-backchain
10086 Generate (or do not generate) code which maintains an explicit
10087 backchain within the stack frame that points to the caller's frame.
10088 This is currently needed to allow debugging. The default is to
10089 generate the backchain.
10092 @itemx -mno-small-exec
10093 @opindex msmall-exec
10094 @opindex mno-small-exec
10095 Generate (or do not generate) code using the @code{bras} instruction
10096 to do subroutine calls.
10097 This only works reliably if the total executable size does not
10098 exceed 64k. The default is to use the @code{basr} instruction instead,
10099 which does not have this limitation.
10105 When @option{-m31} is specified, generate code compliant to the
10106 Linux for S/390 ABI@. When @option{-m64} is specified, generate
10107 code compliant to the Linux for zSeries ABI@. This allows GCC in
10108 particular to generate 64-bit instructions. For the @samp{s390}
10109 targets, the default is @option{-m31}, while the @samp{s390x}
10110 targets default to @option{-m64}.
10116 When @option{-mzarch} is specified, generate code using the
10117 instructions available on z/Architecture.
10118 When @option{-mesa} is specified, generate code using the
10119 instructions available on ESA/390. Note that @option{-mesa} is
10120 not possible with @option{-m64}.
10121 When generating code compliant to the Linux for S/390 ABI,
10122 the default is @option{-mesa}. When generating code compliant
10123 to the Linux for zSeries ABI, the default is @option{-mzarch}.
10129 Generate (or do not generate) code using the @code{mvcle} instruction
10130 to perform block moves. When @option{-mno-mvcle} is specified,
10131 use a @code{mvc} loop instead. This is the default.
10137 Print (or do not print) additional debug information when compiling.
10138 The default is to not print debug information.
10140 @item -march=@var{cpu-type}
10142 Generate code that will run on @var{cpu-type}, which is the name of a system
10143 representing a certain processor type. Possible values for
10144 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
10145 When generating code using the instructions available on z/Architecture,
10146 the default is @option{-march=z900}. Otherwise, the default is
10147 @option{-march=g5}.
10149 @item -mtune=@var{cpu-type}
10151 Tune to @var{cpu-type} everything applicable about the generated code,
10152 except for the ABI and the set of available instructions.
10153 The list of @var{cpu-type} values is the same as for @option{-march}.
10154 The default is the value used for @option{-march}.
10157 @itemx -mno-fused-madd
10158 @opindex mfused-madd
10159 @opindex mno-fused-madd
10160 Generate code that uses (does not use) the floating point multiply and
10161 accumulate instructions. These instructions are generated by default if
10162 hardware floating point is used.
10166 @subsection CRIS Options
10167 @cindex CRIS Options
10169 These options are defined specifically for the CRIS ports.
10172 @item -march=@var{architecture-type}
10173 @itemx -mcpu=@var{architecture-type}
10176 Generate code for the specified architecture. The choices for
10177 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10178 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX.
10179 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10182 @item -mtune=@var{architecture-type}
10184 Tune to @var{architecture-type} everything applicable about the generated
10185 code, except for the ABI and the set of available instructions. The
10186 choices for @var{architecture-type} are the same as for
10187 @option{-march=@var{architecture-type}}.
10189 @item -mmax-stack-frame=@var{n}
10190 @opindex mmax-stack-frame
10191 Warn when the stack frame of a function exceeds @var{n} bytes.
10193 @item -melinux-stacksize=@var{n}
10194 @opindex melinux-stacksize
10195 Only available with the @samp{cris-axis-aout} target. Arranges for
10196 indications in the program to the kernel loader that the stack of the
10197 program should be set to @var{n} bytes.
10203 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10204 @option{-march=v3} and @option{-march=v8} respectively.
10208 Enable CRIS-specific verbose debug-related information in the assembly
10209 code. This option also has the effect to turn off the @samp{#NO_APP}
10210 formatted-code indicator to the assembler at the beginning of the
10215 Do not use condition-code results from previous instruction; always emit
10216 compare and test instructions before use of condition codes.
10218 @item -mno-side-effects
10219 @opindex mno-side-effects
10220 Do not emit instructions with side-effects in addressing modes other than
10223 @item -mstack-align
10224 @itemx -mno-stack-align
10225 @itemx -mdata-align
10226 @itemx -mno-data-align
10227 @itemx -mconst-align
10228 @itemx -mno-const-align
10229 @opindex mstack-align
10230 @opindex mno-stack-align
10231 @opindex mdata-align
10232 @opindex mno-data-align
10233 @opindex mconst-align
10234 @opindex mno-const-align
10235 These options (no-options) arranges (eliminate arrangements) for the
10236 stack-frame, individual data and constants to be aligned for the maximum
10237 single data access size for the chosen CPU model. The default is to
10238 arrange for 32-bit alignment. ABI details such as structure layout are
10239 not affected by these options.
10247 Similar to the stack- data- and const-align options above, these options
10248 arrange for stack-frame, writable data and constants to all be 32-bit,
10249 16-bit or 8-bit aligned. The default is 32-bit alignment.
10251 @item -mno-prologue-epilogue
10252 @itemx -mprologue-epilogue
10253 @opindex mno-prologue-epilogue
10254 @opindex mprologue-epilogue
10255 With @option{-mno-prologue-epilogue}, the normal function prologue and
10256 epilogue that sets up the stack-frame are omitted and no return
10257 instructions or return sequences are generated in the code. Use this
10258 option only together with visual inspection of the compiled code: no
10259 warnings or errors are generated when call-saved registers must be saved,
10260 or storage for local variable needs to be allocated.
10264 @opindex mno-gotplt
10266 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10267 instruction sequences that load addresses for functions from the PLT part
10268 of the GOT rather than (traditional on other architectures) calls to the
10269 PLT. The default is @option{-mgotplt}.
10273 Legacy no-op option only recognized with the cris-axis-aout target.
10277 Legacy no-op option only recognized with the cris-axis-elf and
10278 cris-axis-linux-gnu targets.
10282 Only recognized with the cris-axis-aout target, where it selects a
10283 GNU/linux-like multilib, include files and instruction set for
10284 @option{-march=v8}.
10288 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10292 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
10293 to link with input-output functions from a simulator library. Code,
10294 initialized data and zero-initialized data are allocated consecutively.
10298 Like @option{-sim}, but pass linker options to locate initialized data at
10299 0x40000000 and zero-initialized data at 0x80000000.
10303 @subsection MMIX Options
10304 @cindex MMIX Options
10306 These options are defined for the MMIX:
10310 @itemx -mno-libfuncs
10312 @opindex mno-libfuncs
10313 Specify that intrinsic library functions are being compiled, passing all
10314 values in registers, no matter the size.
10317 @itemx -mno-epsilon
10319 @opindex mno-epsilon
10320 Generate floating-point comparison instructions that compare with respect
10321 to the @code{rE} epsilon register.
10323 @item -mabi=mmixware
10325 @opindex mabi-mmixware
10327 Generate code that passes function parameters and return values that (in
10328 the called function) are seen as registers @code{$0} and up, as opposed to
10329 the GNU ABI which uses global registers @code{$231} and up.
10331 @item -mzero-extend
10332 @itemx -mno-zero-extend
10333 @opindex mzero-extend
10334 @opindex mno-zero-extend
10335 When reading data from memory in sizes shorter than 64 bits, use (do not
10336 use) zero-extending load instructions by default, rather than
10337 sign-extending ones.
10340 @itemx -mno-knuthdiv
10342 @opindex mno-knuthdiv
10343 Make the result of a division yielding a remainder have the same sign as
10344 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10345 remainder follows the sign of the dividend. Both methods are
10346 arithmetically valid, the latter being almost exclusively used.
10348 @item -mtoplevel-symbols
10349 @itemx -mno-toplevel-symbols
10350 @opindex mtoplevel-symbols
10351 @opindex mno-toplevel-symbols
10352 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10353 code can be used with the @code{PREFIX} assembly directive.
10357 Generate an executable in the ELF format, rather than the default
10358 @samp{mmo} format used by the @command{mmix} simulator.
10360 @item -mbranch-predict
10361 @itemx -mno-branch-predict
10362 @opindex mbranch-predict
10363 @opindex mno-branch-predict
10364 Use (do not use) the probable-branch instructions, when static branch
10365 prediction indicates a probable branch.
10367 @item -mbase-addresses
10368 @itemx -mno-base-addresses
10369 @opindex mbase-addresses
10370 @opindex mno-base-addresses
10371 Generate (do not generate) code that uses @emph{base addresses}. Using a
10372 base address automatically generates a request (handled by the assembler
10373 and the linker) for a constant to be set up in a global register. The
10374 register is used for one or more base address requests within the range 0
10375 to 255 from the value held in the register. The generally leads to short
10376 and fast code, but the number of different data items that can be
10377 addressed is limited. This means that a program that uses lots of static
10378 data may require @option{-mno-base-addresses}.
10380 @item -msingle-exit
10381 @itemx -mno-single-exit
10382 @opindex msingle-exit
10383 @opindex mno-single-exit
10384 Force (do not force) generated code to have a single exit point in each
10388 @node PDP-11 Options
10389 @subsection PDP-11 Options
10390 @cindex PDP-11 Options
10392 These options are defined for the PDP-11:
10397 Use hardware FPP floating point. This is the default. (FIS floating
10398 point on the PDP-11/40 is not supported.)
10401 @opindex msoft-float
10402 Do not use hardware floating point.
10406 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10410 Return floating-point results in memory. This is the default.
10414 Generate code for a PDP-11/40.
10418 Generate code for a PDP-11/45. This is the default.
10422 Generate code for a PDP-11/10.
10424 @item -mbcopy-builtin
10425 @opindex bcopy-builtin
10426 Use inline @code{movstrhi} patterns for copying memory. This is the
10431 Do not use inline @code{movstrhi} patterns for copying memory.
10437 Use 16-bit @code{int}. This is the default.
10443 Use 32-bit @code{int}.
10446 @itemx -mno-float32
10448 @opindex mno-float32
10449 Use 64-bit @code{float}. This is the default.
10452 @itemx -mno-float64
10454 @opindex mno-float64
10455 Use 32-bit @code{float}.
10459 Use @code{abshi2} pattern. This is the default.
10463 Do not use @code{abshi2} pattern.
10465 @item -mbranch-expensive
10466 @opindex mbranch-expensive
10467 Pretend that branches are expensive. This is for experimenting with
10468 code generation only.
10470 @item -mbranch-cheap
10471 @opindex mbranch-cheap
10472 Do not pretend that branches are expensive. This is the default.
10476 Generate code for a system with split I&D.
10480 Generate code for a system without split I&D. This is the default.
10484 Use Unix assembler syntax. This is the default when configured for
10485 @samp{pdp11-*-bsd}.
10489 Use DEC assembler syntax. This is the default when configured for any
10490 PDP-11 target other than @samp{pdp11-*-bsd}.
10493 @node Xstormy16 Options
10494 @subsection Xstormy16 Options
10495 @cindex Xstormy16 Options
10497 These options are defined for Xstormy16:
10502 Choose startup files and linker script suitable for the simulator.
10506 @subsection FRV Options
10507 @cindex FRV Options
10513 Only use the first 32 general purpose registers.
10518 Use all 64 general purpose registers.
10523 Use only the first 32 floating point registers.
10528 Use all 64 floating point registers
10531 @opindex mhard-float
10533 Use hardware instructions for floating point operations.
10536 @opindex msoft-float
10538 Use library routines for floating point operations.
10543 Dynamically allocate condition code registers.
10548 Do not try to dynamically allocate condition code registers, only
10549 use @code{icc0} and @code{fcc0}.
10554 Change ABI to use double word insns.
10559 Do not use double word instructions.
10564 Use floating point double instructions.
10567 @opindex mno-double
10569 Do not use floating point double instructions.
10574 Use media instructions.
10579 Do not use media instructions.
10584 Use multiply and add/subtract instructions.
10587 @opindex mno-muladd
10589 Do not use multiply and add/subtract instructions.
10591 @item -mlibrary-pic
10592 @opindex mlibrary-pic
10594 Enable PIC support for building libraries
10599 Use only the first four media accumulator registers.
10604 Use all eight media accumulator registers.
10609 Pack VLIW instructions.
10614 Do not pack VLIW instructions.
10617 @opindex mno-eflags
10619 Do not mark ABI switches in e_flags.
10622 @opindex mcond-move
10624 Enable the use of conditional-move instructions (default).
10626 This switch is mainly for debugging the compiler and will likely be removed
10627 in a future version.
10629 @item -mno-cond-move
10630 @opindex mno-cond-move
10632 Disable the use of conditional-move instructions.
10634 This switch is mainly for debugging the compiler and will likely be removed
10635 in a future version.
10640 Enable the use of conditional set instructions (default).
10642 This switch is mainly for debugging the compiler and will likely be removed
10643 in a future version.
10648 Disable the use of conditional set instructions.
10650 This switch is mainly for debugging the compiler and will likely be removed
10651 in a future version.
10654 @opindex mcond-exec
10656 Enable the use of conditional execution (default).
10658 This switch is mainly for debugging the compiler and will likely be removed
10659 in a future version.
10661 @item -mno-cond-exec
10662 @opindex mno-cond-exec
10664 Disable the use of conditional execution.
10666 This switch is mainly for debugging the compiler and will likely be removed
10667 in a future version.
10669 @item -mvliw-branch
10670 @opindex mvliw-branch
10672 Run a pass to pack branches into VLIW instructions (default).
10674 This switch is mainly for debugging the compiler and will likely be removed
10675 in a future version.
10677 @item -mno-vliw-branch
10678 @opindex mno-vliw-branch
10680 Do not run a pass to pack branches into VLIW instructions.
10682 This switch is mainly for debugging the compiler and will likely be removed
10683 in a future version.
10685 @item -mmulti-cond-exec
10686 @opindex mmulti-cond-exec
10688 Enable optimization of @code{&&} and @code{||} in conditional execution
10691 This switch is mainly for debugging the compiler and will likely be removed
10692 in a future version.
10694 @item -mno-multi-cond-exec
10695 @opindex mno-multi-cond-exec
10697 Disable optimization of @code{&&} and @code{||} in conditional execution.
10699 This switch is mainly for debugging the compiler and will likely be removed
10700 in a future version.
10702 @item -mnested-cond-exec
10703 @opindex mnested-cond-exec
10705 Enable nested conditional execution optimizations (default).
10707 This switch is mainly for debugging the compiler and will likely be removed
10708 in a future version.
10710 @item -mno-nested-cond-exec
10711 @opindex mno-nested-cond-exec
10713 Disable nested conditional execution optimizations.
10715 This switch is mainly for debugging the compiler and will likely be removed
10716 in a future version.
10718 @item -mtomcat-stats
10719 @opindex mtomcat-stats
10721 Cause gas to print out tomcat statistics.
10723 @item -mcpu=@var{cpu}
10726 Select the processor type for which to generate code. Possible values are
10727 @samp{simple}, @samp{tomcat}, @samp{fr500}, @samp{fr400}, @samp{fr300},
10732 @node Xtensa Options
10733 @subsection Xtensa Options
10734 @cindex Xtensa Options
10736 These options are supported for Xtensa targets:
10740 @itemx -mno-const16
10742 @opindex mno-const16
10743 Enable or disable use of @code{CONST16} instructions for loading
10744 constant values. The @code{CONST16} instruction is currently not a
10745 standard option from Tensilica. When enabled, @code{CONST16}
10746 instructions are always used in place of the standard @code{L32R}
10747 instructions. The use of @code{CONST16} is enabled by default only if
10748 the @code{L32R} instruction is not available.
10751 @itemx -mno-fused-madd
10752 @opindex mfused-madd
10753 @opindex mno-fused-madd
10754 Enable or disable use of fused multiply/add and multiply/subtract
10755 instructions in the floating-point option. This has no effect if the
10756 floating-point option is not also enabled. Disabling fused multiply/add
10757 and multiply/subtract instructions forces the compiler to use separate
10758 instructions for the multiply and add/subtract operations. This may be
10759 desirable in some cases where strict IEEE 754-compliant results are
10760 required: the fused multiply add/subtract instructions do not round the
10761 intermediate result, thereby producing results with @emph{more} bits of
10762 precision than specified by the IEEE standard. Disabling fused multiply
10763 add/subtract instructions also ensures that the program output is not
10764 sensitive to the compiler's ability to combine multiply and add/subtract
10767 @item -mtext-section-literals
10768 @itemx -mno-text-section-literals
10769 @opindex mtext-section-literals
10770 @opindex mno-text-section-literals
10771 Control the treatment of literal pools. The default is
10772 @option{-mno-text-section-literals}, which places literals in a separate
10773 section in the output file. This allows the literal pool to be placed
10774 in a data RAM/ROM, and it also allows the linker to combine literal
10775 pools from separate object files to remove redundant literals and
10776 improve code size. With @option{-mtext-section-literals}, the literals
10777 are interspersed in the text section in order to keep them as close as
10778 possible to their references. This may be necessary for large assembly
10781 @item -mtarget-align
10782 @itemx -mno-target-align
10783 @opindex mtarget-align
10784 @opindex mno-target-align
10785 When this option is enabled, GCC instructs the assembler to
10786 automatically align instructions to reduce branch penalties at the
10787 expense of some code density. The assembler attempts to widen density
10788 instructions to align branch targets and the instructions following call
10789 instructions. If there are not enough preceding safe density
10790 instructions to align a target, no widening will be performed. The
10791 default is @option{-mtarget-align}. These options do not affect the
10792 treatment of auto-aligned instructions like @code{LOOP}, which the
10793 assembler will always align, either by widening density instructions or
10794 by inserting no-op instructions.
10797 @itemx -mno-longcalls
10798 @opindex mlongcalls
10799 @opindex mno-longcalls
10800 When this option is enabled, GCC instructs the assembler to translate
10801 direct calls to indirect calls unless it can determine that the target
10802 of a direct call is in the range allowed by the call instruction. This
10803 translation typically occurs for calls to functions in other source
10804 files. Specifically, the assembler translates a direct @code{CALL}
10805 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
10806 The default is @option{-mno-longcalls}. This option should be used in
10807 programs where the call target can potentially be out of range. This
10808 option is implemented in the assembler, not the compiler, so the
10809 assembly code generated by GCC will still show direct call
10810 instructions---look at the disassembled object code to see the actual
10811 instructions. Note that the assembler will use an indirect call for
10812 every cross-file call, not just those that really will be out of range.
10815 @node Code Gen Options
10816 @section Options for Code Generation Conventions
10817 @cindex code generation conventions
10818 @cindex options, code generation
10819 @cindex run-time options
10821 These machine-independent options control the interface conventions
10822 used in code generation.
10824 Most of them have both positive and negative forms; the negative form
10825 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
10826 one of the forms is listed---the one which is not the default. You
10827 can figure out the other form by either removing @samp{no-} or adding
10831 @item -fbounds-check
10832 @opindex fbounds-check
10833 For front-ends that support it, generate additional code to check that
10834 indices used to access arrays are within the declared range. This is
10835 currently only supported by the Java and Fortran 77 front-ends, where
10836 this option defaults to true and false respectively.
10840 This option generates traps for signed overflow on addition, subtraction,
10841 multiplication operations.
10845 This option instructs the compiler to assume that signed arithmetic
10846 overflow of addition, subtraction and multiplication wraps around
10847 using twos-complement representation. This flag enables some optimizations
10848 and disables other. This option is enabled by default for the Java
10849 front-end, as required by the Java language specification.
10852 @opindex fexceptions
10853 Enable exception handling. Generates extra code needed to propagate
10854 exceptions. For some targets, this implies GCC will generate frame
10855 unwind information for all functions, which can produce significant data
10856 size overhead, although it does not affect execution. If you do not
10857 specify this option, GCC will enable it by default for languages like
10858 C++ which normally require exception handling, and disable it for
10859 languages like C that do not normally require it. However, you may need
10860 to enable this option when compiling C code that needs to interoperate
10861 properly with exception handlers written in C++. You may also wish to
10862 disable this option if you are compiling older C++ programs that don't
10863 use exception handling.
10865 @item -fnon-call-exceptions
10866 @opindex fnon-call-exceptions
10867 Generate code that allows trapping instructions to throw exceptions.
10868 Note that this requires platform-specific runtime support that does
10869 not exist everywhere. Moreover, it only allows @emph{trapping}
10870 instructions to throw exceptions, i.e.@: memory references or floating
10871 point instructions. It does not allow exceptions to be thrown from
10872 arbitrary signal handlers such as @code{SIGALRM}.
10874 @item -funwind-tables
10875 @opindex funwind-tables
10876 Similar to @option{-fexceptions}, except that it will just generate any needed
10877 static data, but will not affect the generated code in any other way.
10878 You will normally not enable this option; instead, a language processor
10879 that needs this handling would enable it on your behalf.
10881 @item -fasynchronous-unwind-tables
10882 @opindex funwind-tables
10883 Generate unwind table in dwarf2 format, if supported by target machine. The
10884 table is exact at each instruction boundary, so it can be used for stack
10885 unwinding from asynchronous events (such as debugger or garbage collector).
10887 @item -fpcc-struct-return
10888 @opindex fpcc-struct-return
10889 Return ``short'' @code{struct} and @code{union} values in memory like
10890 longer ones, rather than in registers. This convention is less
10891 efficient, but it has the advantage of allowing intercallability between
10892 GCC-compiled files and files compiled with other compilers, particularly
10893 the Portable C Compiler (pcc).
10895 The precise convention for returning structures in memory depends
10896 on the target configuration macros.
10898 Short structures and unions are those whose size and alignment match
10899 that of some integer type.
10901 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
10902 switch is not binary compatible with code compiled with the
10903 @option{-freg-struct-return} switch.
10904 Use it to conform to a non-default application binary interface.
10906 @item -freg-struct-return
10907 @opindex freg-struct-return
10908 Return @code{struct} and @code{union} values in registers when possible.
10909 This is more efficient for small structures than
10910 @option{-fpcc-struct-return}.
10912 If you specify neither @option{-fpcc-struct-return} nor
10913 @option{-freg-struct-return}, GCC defaults to whichever convention is
10914 standard for the target. If there is no standard convention, GCC
10915 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
10916 the principal compiler. In those cases, we can choose the standard, and
10917 we chose the more efficient register return alternative.
10919 @strong{Warning:} code compiled with the @option{-freg-struct-return}
10920 switch is not binary compatible with code compiled with the
10921 @option{-fpcc-struct-return} switch.
10922 Use it to conform to a non-default application binary interface.
10924 @item -fshort-enums
10925 @opindex fshort-enums
10926 Allocate to an @code{enum} type only as many bytes as it needs for the
10927 declared range of possible values. Specifically, the @code{enum} type
10928 will be equivalent to the smallest integer type which has enough room.
10930 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
10931 code that is not binary compatible with code generated without that switch.
10932 Use it to conform to a non-default application binary interface.
10934 @item -fshort-double
10935 @opindex fshort-double
10936 Use the same size for @code{double} as for @code{float}.
10938 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
10939 code that is not binary compatible with code generated without that switch.
10940 Use it to conform to a non-default application binary interface.
10942 @item -fshort-wchar
10943 @opindex fshort-wchar
10944 Override the underlying type for @samp{wchar_t} to be @samp{short
10945 unsigned int} instead of the default for the target. This option is
10946 useful for building programs to run under WINE@.
10948 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
10949 code that is not binary compatible with code generated without that switch.
10950 Use it to conform to a non-default application binary interface.
10952 @item -fshared-data
10953 @opindex fshared-data
10954 Requests that the data and non-@code{const} variables of this
10955 compilation be shared data rather than private data. The distinction
10956 makes sense only on certain operating systems, where shared data is
10957 shared between processes running the same program, while private data
10958 exists in one copy per process.
10961 @opindex fno-common
10962 In C, allocate even uninitialized global variables in the data section of the
10963 object file, rather than generating them as common blocks. This has the
10964 effect that if the same variable is declared (without @code{extern}) in
10965 two different compilations, you will get an error when you link them.
10966 The only reason this might be useful is if you wish to verify that the
10967 program will work on other systems which always work this way.
10971 Ignore the @samp{#ident} directive.
10973 @item -finhibit-size-directive
10974 @opindex finhibit-size-directive
10975 Don't output a @code{.size} assembler directive, or anything else that
10976 would cause trouble if the function is split in the middle, and the
10977 two halves are placed at locations far apart in memory. This option is
10978 used when compiling @file{crtstuff.c}; you should not need to use it
10981 @item -fverbose-asm
10982 @opindex fverbose-asm
10983 Put extra commentary information in the generated assembly code to
10984 make it more readable. This option is generally only of use to those
10985 who actually need to read the generated assembly code (perhaps while
10986 debugging the compiler itself).
10988 @option{-fno-verbose-asm}, the default, causes the
10989 extra information to be omitted and is useful when comparing two assembler
10994 @cindex global offset table
10996 Generate position-independent code (PIC) suitable for use in a shared
10997 library, if supported for the target machine. Such code accesses all
10998 constant addresses through a global offset table (GOT)@. The dynamic
10999 loader resolves the GOT entries when the program starts (the dynamic
11000 loader is not part of GCC; it is part of the operating system). If
11001 the GOT size for the linked executable exceeds a machine-specific
11002 maximum size, you get an error message from the linker indicating that
11003 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11004 instead. (These maximums are 8k on the SPARC and 32k
11005 on the m68k and RS/6000. The 386 has no such limit.)
11007 Position-independent code requires special support, and therefore works
11008 only on certain machines. For the 386, GCC supports PIC for System V
11009 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11010 position-independent.
11014 If supported for the target machine, emit position-independent code,
11015 suitable for dynamic linking and avoiding any limit on the size of the
11016 global offset table. This option makes a difference on the m68k
11019 Position-independent code requires special support, and therefore works
11020 only on certain machines.
11026 These options are similar to @option{-fpic} and @option{-fPIC}, but
11027 generated position independent code can be only linked into executables.
11028 Usually these options are used when @option{-pie} GCC option will be
11029 used during linking.
11031 @item -ffixed-@var{reg}
11033 Treat the register named @var{reg} as a fixed register; generated code
11034 should never refer to it (except perhaps as a stack pointer, frame
11035 pointer or in some other fixed role).
11037 @var{reg} must be the name of a register. The register names accepted
11038 are machine-specific and are defined in the @code{REGISTER_NAMES}
11039 macro in the machine description macro file.
11041 This flag does not have a negative form, because it specifies a
11044 @item -fcall-used-@var{reg}
11045 @opindex fcall-used
11046 Treat the register named @var{reg} as an allocable register that is
11047 clobbered by function calls. It may be allocated for temporaries or
11048 variables that do not live across a call. Functions compiled this way
11049 will not save and restore the register @var{reg}.
11051 It is an error to used this flag with the frame pointer or stack pointer.
11052 Use of this flag for other registers that have fixed pervasive roles in
11053 the machine's execution model will produce disastrous results.
11055 This flag does not have a negative form, because it specifies a
11058 @item -fcall-saved-@var{reg}
11059 @opindex fcall-saved
11060 Treat the register named @var{reg} as an allocable register saved by
11061 functions. It may be allocated even for temporaries or variables that
11062 live across a call. Functions compiled this way will save and restore
11063 the register @var{reg} if they use it.
11065 It is an error to used this flag with the frame pointer or stack pointer.
11066 Use of this flag for other registers that have fixed pervasive roles in
11067 the machine's execution model will produce disastrous results.
11069 A different sort of disaster will result from the use of this flag for
11070 a register in which function values may be returned.
11072 This flag does not have a negative form, because it specifies a
11075 @item -fpack-struct
11076 @opindex fpack-struct
11077 Pack all structure members together without holes.
11079 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11080 code that is not binary compatible with code generated without that switch.
11081 Additionally, it makes the code suboptimal.
11082 Use it to conform to a non-default application binary interface.
11084 @item -finstrument-functions
11085 @opindex finstrument-functions
11086 Generate instrumentation calls for entry and exit to functions. Just
11087 after function entry and just before function exit, the following
11088 profiling functions will be called with the address of the current
11089 function and its call site. (On some platforms,
11090 @code{__builtin_return_address} does not work beyond the current
11091 function, so the call site information may not be available to the
11092 profiling functions otherwise.)
11095 void __cyg_profile_func_enter (void *this_fn,
11097 void __cyg_profile_func_exit (void *this_fn,
11101 The first argument is the address of the start of the current function,
11102 which may be looked up exactly in the symbol table.
11104 This instrumentation is also done for functions expanded inline in other
11105 functions. The profiling calls will indicate where, conceptually, the
11106 inline function is entered and exited. This means that addressable
11107 versions of such functions must be available. If all your uses of a
11108 function are expanded inline, this may mean an additional expansion of
11109 code size. If you use @samp{extern inline} in your C code, an
11110 addressable version of such functions must be provided. (This is
11111 normally the case anyways, but if you get lucky and the optimizer always
11112 expands the functions inline, you might have gotten away without
11113 providing static copies.)
11115 A function may be given the attribute @code{no_instrument_function}, in
11116 which case this instrumentation will not be done. This can be used, for
11117 example, for the profiling functions listed above, high-priority
11118 interrupt routines, and any functions from which the profiling functions
11119 cannot safely be called (perhaps signal handlers, if the profiling
11120 routines generate output or allocate memory).
11122 @item -fstack-check
11123 @opindex fstack-check
11124 Generate code to verify that you do not go beyond the boundary of the
11125 stack. You should specify this flag if you are running in an
11126 environment with multiple threads, but only rarely need to specify it in
11127 a single-threaded environment since stack overflow is automatically
11128 detected on nearly all systems if there is only one stack.
11130 Note that this switch does not actually cause checking to be done; the
11131 operating system must do that. The switch causes generation of code
11132 to ensure that the operating system sees the stack being extended.
11134 @item -fstack-limit-register=@var{reg}
11135 @itemx -fstack-limit-symbol=@var{sym}
11136 @itemx -fno-stack-limit
11137 @opindex fstack-limit-register
11138 @opindex fstack-limit-symbol
11139 @opindex fno-stack-limit
11140 Generate code to ensure that the stack does not grow beyond a certain value,
11141 either the value of a register or the address of a symbol. If the stack
11142 would grow beyond the value, a signal is raised. For most targets,
11143 the signal is raised before the stack overruns the boundary, so
11144 it is possible to catch the signal without taking special precautions.
11146 For instance, if the stack starts at absolute address @samp{0x80000000}
11147 and grows downwards, you can use the flags
11148 @option{-fstack-limit-symbol=__stack_limit} and
11149 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11150 of 128KB@. Note that this may only work with the GNU linker.
11152 @cindex aliasing of parameters
11153 @cindex parameters, aliased
11154 @item -fargument-alias
11155 @itemx -fargument-noalias
11156 @itemx -fargument-noalias-global
11157 @opindex fargument-alias
11158 @opindex fargument-noalias
11159 @opindex fargument-noalias-global
11160 Specify the possible relationships among parameters and between
11161 parameters and global data.
11163 @option{-fargument-alias} specifies that arguments (parameters) may
11164 alias each other and may alias global storage.@*
11165 @option{-fargument-noalias} specifies that arguments do not alias
11166 each other, but may alias global storage.@*
11167 @option{-fargument-noalias-global} specifies that arguments do not
11168 alias each other and do not alias global storage.
11170 Each language will automatically use whatever option is required by
11171 the language standard. You should not need to use these options yourself.
11173 @item -fleading-underscore
11174 @opindex fleading-underscore
11175 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11176 change the way C symbols are represented in the object file. One use
11177 is to help link with legacy assembly code.
11179 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11180 generate code that is not binary compatible with code generated without that
11181 switch. Use it to conform to a non-default application binary interface.
11182 Not all targets provide complete support for this switch.
11184 @item -ftls-model=@var{model}
11185 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11186 The @var{model} argument should be one of @code{global-dynamic},
11187 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
11189 The default without @option{-fpic} is @code{initial-exec}; with
11190 @option{-fpic} the default is @code{global-dynamic}.
11195 @node Environment Variables
11196 @section Environment Variables Affecting GCC
11197 @cindex environment variables
11199 @c man begin ENVIRONMENT
11200 This section describes several environment variables that affect how GCC
11201 operates. Some of them work by specifying directories or prefixes to use
11202 when searching for various kinds of files. Some are used to specify other
11203 aspects of the compilation environment.
11205 Note that you can also specify places to search using options such as
11206 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
11207 take precedence over places specified using environment variables, which
11208 in turn take precedence over those specified by the configuration of GCC@.
11209 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
11210 GNU Compiler Collection (GCC) Internals}.
11215 @c @itemx LC_COLLATE
11217 @c @itemx LC_MONETARY
11218 @c @itemx LC_NUMERIC
11223 @c @findex LC_COLLATE
11224 @findex LC_MESSAGES
11225 @c @findex LC_MONETARY
11226 @c @findex LC_NUMERIC
11230 These environment variables control the way that GCC uses
11231 localization information that allow GCC to work with different
11232 national conventions. GCC inspects the locale categories
11233 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
11234 so. These locale categories can be set to any value supported by your
11235 installation. A typical value is @samp{en_UK} for English in the United
11238 The @env{LC_CTYPE} environment variable specifies character
11239 classification. GCC uses it to determine the character boundaries in
11240 a string; this is needed for some multibyte encodings that contain quote
11241 and escape characters that would otherwise be interpreted as a string
11244 The @env{LC_MESSAGES} environment variable specifies the language to
11245 use in diagnostic messages.
11247 If the @env{LC_ALL} environment variable is set, it overrides the value
11248 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
11249 and @env{LC_MESSAGES} default to the value of the @env{LANG}
11250 environment variable. If none of these variables are set, GCC
11251 defaults to traditional C English behavior.
11255 If @env{TMPDIR} is set, it specifies the directory to use for temporary
11256 files. GCC uses temporary files to hold the output of one stage of
11257 compilation which is to be used as input to the next stage: for example,
11258 the output of the preprocessor, which is the input to the compiler
11261 @item GCC_EXEC_PREFIX
11262 @findex GCC_EXEC_PREFIX
11263 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
11264 names of the subprograms executed by the compiler. No slash is added
11265 when this prefix is combined with the name of a subprogram, but you can
11266 specify a prefix that ends with a slash if you wish.
11268 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
11269 an appropriate prefix to use based on the pathname it was invoked with.
11271 If GCC cannot find the subprogram using the specified prefix, it
11272 tries looking in the usual places for the subprogram.
11274 The default value of @env{GCC_EXEC_PREFIX} is
11275 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
11276 of @code{prefix} when you ran the @file{configure} script.
11278 Other prefixes specified with @option{-B} take precedence over this prefix.
11280 This prefix is also used for finding files such as @file{crt0.o} that are
11283 In addition, the prefix is used in an unusual way in finding the
11284 directories to search for header files. For each of the standard
11285 directories whose name normally begins with @samp{/usr/local/lib/gcc}
11286 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
11287 replacing that beginning with the specified prefix to produce an
11288 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
11289 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
11290 These alternate directories are searched first; the standard directories
11293 @item COMPILER_PATH
11294 @findex COMPILER_PATH
11295 The value of @env{COMPILER_PATH} is a colon-separated list of
11296 directories, much like @env{PATH}. GCC tries the directories thus
11297 specified when searching for subprograms, if it can't find the
11298 subprograms using @env{GCC_EXEC_PREFIX}.
11301 @findex LIBRARY_PATH
11302 The value of @env{LIBRARY_PATH} is a colon-separated list of
11303 directories, much like @env{PATH}. When configured as a native compiler,
11304 GCC tries the directories thus specified when searching for special
11305 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
11306 using GCC also uses these directories when searching for ordinary
11307 libraries for the @option{-l} option (but directories specified with
11308 @option{-L} come first).
11312 @cindex locale definition
11313 This variable is used to pass locale information to the compiler. One way in
11314 which this information is used is to determine the character set to be used
11315 when character literals, string literals and comments are parsed in C and C++.
11316 When the compiler is configured to allow multibyte characters,
11317 the following values for @env{LANG} are recognized:
11321 Recognize JIS characters.
11323 Recognize SJIS characters.
11325 Recognize EUCJP characters.
11328 If @env{LANG} is not defined, or if it has some other value, then the
11329 compiler will use mblen and mbtowc as defined by the default locale to
11330 recognize and translate multibyte characters.
11334 Some additional environments variables affect the behavior of the
11337 @include cppenv.texi
11341 @node Precompiled Headers
11342 @section Using Precompiled Headers
11343 @cindex precompiled headers
11344 @cindex speed of compilation
11346 Often large projects have many header files that are included in every
11347 source file. The time the compiler takes to process these header files
11348 over and over again can account for nearly all of the time required to
11349 build the project. To make builds faster, GCC allows users to
11350 `precompile' a header file; then, if builds can use the precompiled
11351 header file they will be much faster.
11353 To create a precompiled header file, simply compile it as you would any
11354 other file, if necessary using the @option{-x} option to make the driver
11355 treat it as a C or C++ header file. You will probably want to use a
11356 tool like @command{make} to keep the precompiled header up-to-date when
11357 the headers it contains change.
11359 A precompiled header file will be searched for when @code{#include} is
11360 seen in the compilation. As it searches for the included file
11361 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
11362 compiler looks for a precompiled header in each directory just before it
11363 looks for the include file in that directory. The name searched for is
11364 the name specified in the @code{#include} with @samp{.gch} appended. If
11365 the precompiled header file can't be used, it is ignored.
11367 For instance, if you have @code{#include "all.h"}, and you have
11368 @file{all.h.gch} in the same directory as @file{all.h}, then the
11369 precompiled header file will be used if possible, and the original
11370 header will be used otherwise.
11372 Alternatively, you might decide to put the precompiled header file in a
11373 directory and use @option{-I} to ensure that directory is searched
11374 before (or instead of) the directory containing the original header.
11375 Then, if you want to check that the precompiled header file is always
11376 used, you can put a file of the same name as the original header in this
11377 directory containing an @code{#error} command.
11379 This also works with @option{-include}. So yet another way to use
11380 precompiled headers, good for projects not designed with precompiled
11381 header files in mind, is to simply take most of the header files used by
11382 a project, include them from another header file, precompile that header
11383 file, and @option{-include} the precompiled header. If the header files
11384 have guards against multiple inclusion, they will be skipped because
11385 they've already been included (in the precompiled header).
11387 If you need to precompile the same header file for different
11388 languages, targets, or compiler options, you can instead make a
11389 @emph{directory} named like @file{all.h.gch}, and put each precompiled
11390 header in the directory. (It doesn't matter what you call the files
11391 in the directory, every precompiled header in the directory will be
11392 considered.) The first precompiled header encountered in the
11393 directory that is valid for this compilation will be used; they're
11394 searched in no particular order.
11396 There are many other possibilities, limited only by your imagination,
11397 good sense, and the constraints of your build system.
11399 A precompiled header file can be used only when these conditions apply:
11403 Only one precompiled header can be used in a particular compilation.
11405 A precompiled header can't be used once the first C token is seen. You
11406 can have preprocessor directives before a precompiled header; you can
11407 even include a precompiled header from inside another header, so long as
11408 there are no C tokens before the @code{#include}.
11410 The precompiled header file must be produced for the same language as
11411 the current compilation. You can't use a C precompiled header for a C++
11414 The precompiled header file must be produced by the same compiler
11415 version and configuration as the current compilation is using.
11416 The easiest way to guarantee this is to use the same compiler binary
11417 for creating and using precompiled headers.
11419 Any macros defined before the precompiled header (including with
11420 @option{-D}) must either be defined in the same way as when the
11421 precompiled header was generated, or must not affect the precompiled
11422 header, which usually means that the they don't appear in the
11423 precompiled header at all.
11425 Certain command-line options must be defined in the same way as when the
11426 precompiled header was generated. At present, it's not clear which
11427 options are safe to change and which are not; the safest choice is to
11428 use exactly the same options when generating and using the precompiled
11432 For all of these but the last, the compiler will automatically ignore
11433 the precompiled header if the conditions aren't met. For the last item,
11434 some option changes will cause the precompiled header to be rejected,
11435 but not all incompatible option combinations have yet been found. If
11436 you find a new incompatible combination, please consider filing a bug
11437 report, see @ref{Bugs}.
11439 @node Running Protoize
11440 @section Running Protoize
11442 The program @code{protoize} is an optional part of GCC@. You can use
11443 it to add prototypes to a program, thus converting the program to ISO
11444 C in one respect. The companion program @code{unprotoize} does the
11445 reverse: it removes argument types from any prototypes that are found.
11447 When you run these programs, you must specify a set of source files as
11448 command line arguments. The conversion programs start out by compiling
11449 these files to see what functions they define. The information gathered
11450 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
11452 After scanning comes actual conversion. The specified files are all
11453 eligible to be converted; any files they include (whether sources or
11454 just headers) are eligible as well.
11456 But not all the eligible files are converted. By default,
11457 @code{protoize} and @code{unprotoize} convert only source and header
11458 files in the current directory. You can specify additional directories
11459 whose files should be converted with the @option{-d @var{directory}}
11460 option. You can also specify particular files to exclude with the
11461 @option{-x @var{file}} option. A file is converted if it is eligible, its
11462 directory name matches one of the specified directory names, and its
11463 name within the directory has not been excluded.
11465 Basic conversion with @code{protoize} consists of rewriting most
11466 function definitions and function declarations to specify the types of
11467 the arguments. The only ones not rewritten are those for varargs
11470 @code{protoize} optionally inserts prototype declarations at the
11471 beginning of the source file, to make them available for any calls that
11472 precede the function's definition. Or it can insert prototype
11473 declarations with block scope in the blocks where undeclared functions
11476 Basic conversion with @code{unprotoize} consists of rewriting most
11477 function declarations to remove any argument types, and rewriting
11478 function definitions to the old-style pre-ISO form.
11480 Both conversion programs print a warning for any function declaration or
11481 definition that they can't convert. You can suppress these warnings
11484 The output from @code{protoize} or @code{unprotoize} replaces the
11485 original source file. The original file is renamed to a name ending
11486 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
11487 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
11488 for DOS) file already exists, then the source file is simply discarded.
11490 @code{protoize} and @code{unprotoize} both depend on GCC itself to
11491 scan the program and collect information about the functions it uses.
11492 So neither of these programs will work until GCC is installed.
11494 Here is a table of the options you can use with @code{protoize} and
11495 @code{unprotoize}. Each option works with both programs unless
11499 @item -B @var{directory}
11500 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
11501 usual directory (normally @file{/usr/local/lib}). This file contains
11502 prototype information about standard system functions. This option
11503 applies only to @code{protoize}.
11505 @item -c @var{compilation-options}
11506 Use @var{compilation-options} as the options when running @command{gcc} to
11507 produce the @samp{.X} files. The special option @option{-aux-info} is
11508 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
11510 Note that the compilation options must be given as a single argument to
11511 @code{protoize} or @code{unprotoize}. If you want to specify several
11512 @command{gcc} options, you must quote the entire set of compilation options
11513 to make them a single word in the shell.
11515 There are certain @command{gcc} arguments that you cannot use, because they
11516 would produce the wrong kind of output. These include @option{-g},
11517 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
11518 the @var{compilation-options}, they are ignored.
11521 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
11522 systems) instead of @samp{.c}. This is convenient if you are converting
11523 a C program to C++. This option applies only to @code{protoize}.
11526 Add explicit global declarations. This means inserting explicit
11527 declarations at the beginning of each source file for each function
11528 that is called in the file and was not declared. These declarations
11529 precede the first function definition that contains a call to an
11530 undeclared function. This option applies only to @code{protoize}.
11532 @item -i @var{string}
11533 Indent old-style parameter declarations with the string @var{string}.
11534 This option applies only to @code{protoize}.
11536 @code{unprotoize} converts prototyped function definitions to old-style
11537 function definitions, where the arguments are declared between the
11538 argument list and the initial @samp{@{}. By default, @code{unprotoize}
11539 uses five spaces as the indentation. If you want to indent with just
11540 one space instead, use @option{-i " "}.
11543 Keep the @samp{.X} files. Normally, they are deleted after conversion
11547 Add explicit local declarations. @code{protoize} with @option{-l} inserts
11548 a prototype declaration for each function in each block which calls the
11549 function without any declaration. This option applies only to
11553 Make no real changes. This mode just prints information about the conversions
11554 that would have been done without @option{-n}.
11557 Make no @samp{.save} files. The original files are simply deleted.
11558 Use this option with caution.
11560 @item -p @var{program}
11561 Use the program @var{program} as the compiler. Normally, the name
11562 @file{gcc} is used.
11565 Work quietly. Most warnings are suppressed.
11568 Print the version number, just like @option{-v} for @command{gcc}.
11571 If you need special compiler options to compile one of your program's
11572 source files, then you should generate that file's @samp{.X} file
11573 specially, by running @command{gcc} on that source file with the
11574 appropriate options and the option @option{-aux-info}. Then run
11575 @code{protoize} on the entire set of files. @code{protoize} will use
11576 the existing @samp{.X} file because it is newer than the source file.
11580 gcc -Dfoo=bar file1.c -aux-info file1.X
11585 You need to include the special files along with the rest in the
11586 @code{protoize} command, even though their @samp{.X} files already
11587 exist, because otherwise they won't get converted.
11589 @xref{Protoize Caveats}, for more information on how to use
11590 @code{protoize} successfully.