1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003 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 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 -gdwarf-1 -gdwarf-1+ -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 -freduce-all-givs -fregmove -frename-registers @gol
285 -freorder-blocks -freorder-functions @gol
286 -frerun-cse-after-loop -frerun-loop-opt @gol
287 -frounding-math -fschedule-insns -fschedule-insns2 @gol
288 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
289 -fsched-spec-load-dangerous @gol
290 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
291 -fsched2-use-superblocks @gol
292 -fsched2-use-traces -fsignaling-nans @gol
293 -fsingle-precision-constant @gol
294 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
295 -funroll-all-loops -funroll-loops -fpeel-loops @gol
296 -funswitch-loops -fold-unroll-loops -fold-unroll-all-loops @gol
297 --param @var{name}=@var{value}
298 -O -O0 -O1 -O2 -O3 -Os}
300 @item Preprocessor Options
301 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
302 @gccoptlist{-A@var{question}=@var{answer} @gol
303 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
304 -C -dD -dI -dM -dN @gol
305 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
306 -idirafter @var{dir} @gol
307 -include @var{file} -imacros @var{file} @gol
308 -iprefix @var{file} -iwithprefix @var{dir} @gol
309 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
310 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
311 -P -fworking-directory -remap @gol
312 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
313 -Xpreprocessor @var{option}}
315 @item Assembler Option
316 @xref{Assembler Options,,Passing Options to the Assembler}.
317 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
320 @xref{Link Options,,Options for Linking}.
321 @gccoptlist{@var{object-file-name} -l@var{library} @gol
322 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
323 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
324 -Wl,@var{option} -Xlinker @var{option} @gol
327 @item Directory Options
328 @xref{Directory Options,,Options for Directory Search}.
329 @gccoptlist{-B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}}
332 @c I wrote this xref this way to avoid overfull hbox. -- rms
333 @xref{Target Options}.
334 @gccoptlist{-V @var{version} -b @var{machine}}
336 @item Machine Dependent Options
337 @xref{Submodel Options,,Hardware Models and Configurations}.
339 @emph{M680x0 Options}
340 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
341 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
342 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
343 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
344 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
346 @emph{M68hc1x Options}
347 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
348 -mauto-incdec -minmax -mlong-calls -mshort @gol
349 -msoft-reg-count=@var{count}}
352 @gccoptlist{-mg -mgnu -munix}
355 @gccoptlist{-mcpu=@var{cpu-type} @gol
356 -mtune=@var{cpu-type} @gol
357 -mcmodel=@var{code-model} @gol
359 -mapp-regs -mbroken-saverestore -mcypress @gol
360 -mfaster-structs -mflat @gol
361 -mfpu -mhard-float -mhard-quad-float @gol
362 -mimpure-text -mlittle-endian -mlive-g0 -mno-app-regs @gol
363 -mno-faster-structs -mno-flat -mno-fpu @gol
364 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles @gol
365 -msoft-float -msoft-quad-float -msparclite -mstack-bias @gol
366 -msupersparc -munaligned-doubles -mv8}
369 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
370 -mapcs-26 -mapcs-32 @gol
371 -mapcs-stack-check -mno-apcs-stack-check @gol
372 -mapcs-float -mno-apcs-float @gol
373 -mapcs-reentrant -mno-apcs-reentrant @gol
374 -msched-prolog -mno-sched-prolog @gol
375 -mlittle-endian -mbig-endian -mwords-little-endian @gol
376 -malignment-traps -mno-alignment-traps @gol
377 -msoft-float -mhard-float -mfpe @gol
378 -mthumb-interwork -mno-thumb-interwork @gol
379 -mcpu=@var{name} -march=@var{name} -mfpe=@var{name} @gol
380 -mstructure-size-boundary=@var{n} @gol
381 -mabort-on-noreturn @gol
382 -mlong-calls -mno-long-calls @gol
383 -msingle-pic-base -mno-single-pic-base @gol
384 -mpic-register=@var{reg} @gol
385 -mnop-fun-dllimport @gol
386 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
387 -mpoke-function-name @gol
389 -mtpcs-frame -mtpcs-leaf-frame @gol
390 -mcaller-super-interworking -mcallee-super-interworking}
392 @emph{MN10200 Options}
395 @emph{MN10300 Options}
396 @gccoptlist{-mmult-bug -mno-mult-bug @gol
397 -mam33 -mno-am33 @gol
398 -mam33-2 -mno-am33-2 @gol
401 @emph{M32R/D Options}
402 @gccoptlist{-m32r2 -m32rx -m32r -mcode-model=@var{model-type} @gol
403 -msdata=@var{sdata-type} -G @var{num}}
406 @gccoptlist{-m88000 -m88100 -m88110 -mbig-pic @gol
407 -mcheck-zero-division -mhandle-large-shift @gol
408 -midentify-revision -mno-check-zero-division @gol
409 -mno-ocs-debug-info -mno-ocs-frame-position @gol
410 -mno-optimize-arg-area -mno-serialize-volatile @gol
411 -mno-underscores -mocs-debug-info @gol
412 -mocs-frame-position -moptimize-arg-area @gol
413 -mserialize-volatile -mshort-data-@var{num} -msvr3 @gol
414 -msvr4 -mtrap-large-shift -muse-div-instruction @gol
415 -mversion-03.00 -mwarn-passed-structs}
417 @emph{RS/6000 and PowerPC Options}
418 @gccoptlist{-mcpu=@var{cpu-type} @gol
419 -mtune=@var{cpu-type} @gol
420 -mpower -mno-power -mpower2 -mno-power2 @gol
421 -mpowerpc -mpowerpc64 -mno-powerpc @gol
422 -maltivec -mno-altivec @gol
423 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
424 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
425 -mnew-mnemonics -mold-mnemonics @gol
426 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
427 -m64 -m32 -mxl-call -mno-xl-call -mpe @gol
428 -malign-power -malign-natural @gol
429 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
430 -mstring -mno-string -mupdate -mno-update @gol
431 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
432 -mstrict-align -mno-strict-align -mrelocatable @gol
433 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
434 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
435 -mdynamic-no-pic @gol
436 -mprioritize-restricted-insns=@var{priority} @gol
437 -msched-costly-dep=@var{dependence_type} @gol
438 -minsert-sched-nops=@var{scheme} @gol
439 -mcall-sysv -mcall-netbsd @gol
440 -maix-struct-return -msvr4-struct-return @gol
441 -mabi=altivec -mabi=no-altivec @gol
442 -mabi=spe -mabi=no-spe @gol
443 -misel=yes -misel=no @gol
444 -mspe=yes -mspe=no @gol
445 -mfloat-gprs=yes -mfloat-gprs=no @gol
446 -mprototype -mno-prototype @gol
447 -msim -mmvme -mads -myellowknife -memb -msdata @gol
448 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
450 @emph{Darwin Options}
451 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
452 -arch_only -bind_at_load -bundle -bundle_loader @gol
453 -client_name -compatibility_version -current_version @gol
454 -dependency-file -dylib_file -dylinker_install_name @gol
455 -dynamic -dynamiclib -exported_symbols_list @gol
456 -filelist -flat_namespace -force_cpusubtype_ALL @gol
457 -force_flat_namespace -headerpad_max_install_names @gol
458 -image_base -init -install_name -keep_private_externs @gol
459 -multi_module -multiply_defined -multiply_defined_unused @gol
460 -noall_load -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
461 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
462 -private_bundle -read_only_relocs -sectalign @gol
463 -sectobjectsymbols -whyload -seg1addr @gol
464 -sectcreate -sectobjectsymbols -sectorder @gol
465 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
466 -segprot -segs_read_only_addr -segs_read_write_addr @gol
467 -single_module -static -sub_library -sub_umbrella @gol
468 -twolevel_namespace -umbrella -undefined @gol
469 -unexported_symbols_list -weak_reference_mismatches @gol
473 @gccoptlist{-mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs @gol
474 -mfull-fp-blocks -mhc-struct-return -min-line-mul @gol
475 -mminimum-fp-blocks -mnohc-struct-return}
478 @gccoptlist{-mabicalls -march=@var{cpu-type} -mtune=@var{cpu=type} @gol
479 -mcpu=@var{cpu-type} -membedded-data -muninit-const-in-rodata @gol
480 -membedded-pic -mfp32 -mfp64 -mfused-madd -mno-fused-madd @gol
481 -mgas -mgp32 -mgp64 -mhard-float -mint64 -mips1 @gol
482 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
483 -mlong64 -mlong32 -mlong-calls -mmemcpy @gol
484 -mmips-as -mmips-tfile -mno-abicalls -mxgot @gol
485 -mno-embedded-data -mno-uninit-const-in-rodata @gol
486 -mno-embedded-pic -mno-long-calls @gol
487 -mno-memcpy -mno-mips-tfile -mno-rnames @gol
488 -mrnames -msoft-float @gol
489 -m4650 -msingle-float -mmad @gol
490 -EL -EB -G @var{num} -nocpp @gol
491 -mabi=32 -mabi=n32 -mabi=64 -mabi=eabi -mabi-fake-default @gol
492 -mfix7000 -mfix-sb1 -mno-fix-sb1 @gol
493 -mno-crt0 -mflush-func=@var{func} -mno-flush-func @gol
494 -mbranch-likely -mno-branch-likely}
496 @emph{i386 and x86-64 Options}
497 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
498 -mfpmath=@var{unit} @gol
499 -masm=@var{dialect} -mno-fancy-math-387 @gol
500 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
501 -mno-wide-multiply -mrtd -malign-double @gol
502 -mpreferred-stack-boundary=@var{num} @gol
503 -mmmx -msse -msse2 -mpni -m3dnow @gol
504 -mthreads -mno-align-stringops -minline-all-stringops @gol
505 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
506 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
507 -mno-red-zone -mno-tls-direct-seg-refs @gol
508 -mcmodel=@var{code-model} @gol
512 @gccoptlist{-march=@var{architecture-type} @gol
513 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
514 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
515 -mjump-in-delay -mlinker-opt -mlong-calls @gol
516 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
517 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
518 -mno-jump-in-delay -mno-long-load-store @gol
519 -mno-portable-runtime -mno-soft-float @gol
520 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
521 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
522 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
523 -nolibdld -static -threads}
525 @emph{Intel 960 Options}
526 @gccoptlist{-m@var{cpu-type} -masm-compat -mclean-linkage @gol
527 -mcode-align -mcomplex-addr -mleaf-procedures @gol
528 -mic-compat -mic2.0-compat -mic3.0-compat @gol
529 -mintel-asm -mno-clean-linkage -mno-code-align @gol
530 -mno-complex-addr -mno-leaf-procedures @gol
531 -mno-old-align -mno-strict-align -mno-tail-call @gol
532 -mnumerics -mold-align -msoft-float -mstrict-align @gol
535 @emph{DEC Alpha Options}
536 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
537 -mieee -mieee-with-inexact -mieee-conformant @gol
538 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
539 -mtrap-precision=@var{mode} -mbuild-constants @gol
540 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
541 -mbwx -mmax -mfix -mcix @gol
542 -mfloat-vax -mfloat-ieee @gol
543 -mexplicit-relocs -msmall-data -mlarge-data @gol
544 -msmall-text -mlarge-text @gol
545 -mmemory-latency=@var{time}}
547 @emph{DEC Alpha/VMS Options}
548 @gccoptlist{-mvms-return-codes}
550 @emph{H8/300 Options}
551 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
554 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
555 -m4-nofpu -m4-single-only -m4-single -m4 @gol
556 -m5-64media -m5-64media-nofpu @gol
557 -m5-32media -m5-32media-nofpu @gol
558 -m5-compact -m5-compact-nofpu @gol
559 -mb -ml -mdalign -mrelax @gol
560 -mbigtable -mfmovd -mhitachi -mnomacsave @gol
561 -mieee -misize -mpadstruct -mspace @gol
562 -mprefergot -musermode}
564 @emph{System V Options}
565 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
568 @gccoptlist{-EB -EL @gol
569 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
570 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
572 @emph{TMS320C3x/C4x Options}
573 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
574 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
575 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
576 -mparallel-insns -mparallel-mpy -mpreserve-float}
579 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
580 -mprolog-function -mno-prolog-function -mspace @gol
581 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
582 -mapp-regs -mno-app-regs @gol
583 -mdisable-callt -mno-disable-callt @gol
589 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
590 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
591 -mregparam -mnoregparam -msb -mnosb @gol
592 -mbitfield -mnobitfield -mhimem -mnohimem}
595 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
596 -mcall-prologues -mno-tablejump -mtiny-stack}
599 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
600 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
601 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
602 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
603 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
606 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
607 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
608 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
609 -mno-base-addresses -msingle-exit -mno-single-exit}
612 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
613 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
614 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
615 -minline-float-divide-max-throughput @gol
616 -minline-int-divide-min-latency @gol
617 -minline-int-divide-max-throughput -mno-dwarf2-asm @gol
618 -mfixed-range=@var{register-range}}
621 @gccoptlist{-mextmem -mextmemory -monchip -mno-asm-optimize @gol
622 -masm-optimize -mbranch-cost=@var{n} -mcond-exec=@var{n}}
624 @emph{S/390 and zSeries Options}
625 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
626 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
627 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
628 -m64 -m31 -mdebug -mno-debug -mesa -mzarch -mfused-madd -mno-fused-madd}
631 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
632 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
633 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
634 -mstack-align -mdata-align -mconst-align @gol
635 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
636 -melf -maout -melinux -mlinux -sim -sim2}
638 @emph{PDP-11 Options}
639 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
640 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
641 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
642 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
643 -mbranch-expensive -mbranch-cheap @gol
644 -msplit -mno-split -munix-asm -mdec-asm}
646 @emph{Xstormy16 Options}
649 @emph{Xtensa Options}
650 @gccoptlist{-mconst16 -mno-const16 @gol
651 -mfused-madd -mno-fused-madd @gol
652 -mtext-section-literals -mno-text-section-literals @gol
653 -mtarget-align -mno-target-align @gol
654 -mlongcalls -mno-longcalls}
657 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
658 -mhard-float -msoft-float @gol
659 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
660 -mdouble -mno-double @gol
661 -mmedia -mno-media -mmuladd -mno-muladd @gol
662 -mlibrary-pic -macc-4 -macc-8 @gol
663 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
664 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
665 -mvliw-branch -mno-vliw-branch @gol
666 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
667 -mno-nested-cond-exec -mtomcat-stats @gol
670 @item Code Generation Options
671 @xref{Code Gen Options,,Options for Code Generation Conventions}.
672 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
673 -ffixed-@var{reg} -fexceptions @gol
674 -fnon-call-exceptions -funwind-tables @gol
675 -fasynchronous-unwind-tables @gol
676 -finhibit-size-directive -finstrument-functions @gol
677 -fno-common -fno-ident @gol
678 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
679 -freg-struct-return -fshared-data -fshort-enums @gol
680 -fshort-double -fshort-wchar @gol
681 -fverbose-asm -fpack-struct -fstack-check @gol
682 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
683 -fargument-alias -fargument-noalias @gol
684 -fargument-noalias-global -fleading-underscore @gol
685 -ftls-model=@var{model} @gol
686 -ftrapv -fwrapv -fbounds-check}
690 * Overall Options:: Controlling the kind of output:
691 an executable, object files, assembler files,
692 or preprocessed source.
693 * C Dialect Options:: Controlling the variant of C language compiled.
694 * C++ Dialect Options:: Variations on C++.
695 * Objective-C Dialect Options:: Variations on Objective-C.
696 * Language Independent Options:: Controlling how diagnostics should be
698 * Warning Options:: How picky should the compiler be?
699 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
700 * Optimize Options:: How much optimization?
701 * Preprocessor Options:: Controlling header files and macro definitions.
702 Also, getting dependency information for Make.
703 * Assembler Options:: Passing options to the assembler.
704 * Link Options:: Specifying libraries and so on.
705 * Directory Options:: Where to find header files and libraries.
706 Where to find the compiler executable files.
707 * Spec Files:: How to pass switches to sub-processes.
708 * Target Options:: Running a cross-compiler, or an old version of GCC.
711 @node Overall Options
712 @section Options Controlling the Kind of Output
714 Compilation can involve up to four stages: preprocessing, compilation
715 proper, assembly and linking, always in that order. GCC is capable of
716 preprocessing and compiling several files either into several
717 assembler input files, or into one assembler input file; then each
718 assembler input file produces an object file, and linking combines all
719 the object files (those newly compiled, and those specified as input)
720 into an executable file.
722 @cindex file name suffix
723 For any given input file, the file name suffix determines what kind of
728 C source code which must be preprocessed.
731 C source code which should not be preprocessed.
734 C++ source code which should not be preprocessed.
737 Objective-C source code. Note that you must link with the library
738 @file{libobjc.a} to make an Objective-C program work.
741 Objective-C source code which should not be preprocessed.
744 C or C++ header file to be turned into a precompiled header.
748 @itemx @var{file}.cxx
749 @itemx @var{file}.cpp
750 @itemx @var{file}.CPP
751 @itemx @var{file}.c++
753 C++ source code which must be preprocessed. Note that in @samp{.cxx},
754 the last two letters must both be literally @samp{x}. Likewise,
755 @samp{.C} refers to a literal capital C@.
759 C++ header file to be turned into a precompiled header.
762 @itemx @var{file}.for
763 @itemx @var{file}.FOR
764 Fortran source code which should not be preprocessed.
767 @itemx @var{file}.fpp
768 @itemx @var{file}.FPP
769 Fortran source code which must be preprocessed (with the traditional
773 Fortran source code which must be preprocessed with a RATFOR
774 preprocessor (not included with GCC)@.
776 @xref{Overall Options,,Options Controlling the Kind of Output, g77,
777 Using and Porting GNU Fortran}, for more details of the handling of
780 @c FIXME: Descriptions of Java file types.
787 Ada source code file which contains a library unit declaration (a
788 declaration of a package, subprogram, or generic, or a generic
789 instantiation), or a library unit renaming declaration (a package,
790 generic, or subprogram renaming declaration). Such files are also
793 @itemx @var{file}.adb
794 Ada source code file containing a library unit body (a subprogram or
795 package body). Such files are also called @dfn{bodies}.
797 @c GCC also knows about some suffixes for languages not yet included:
806 Assembler code which must be preprocessed.
809 An object file to be fed straight into linking.
810 Any file name with no recognized suffix is treated this way.
814 You can specify the input language explicitly with the @option{-x} option:
817 @item -x @var{language}
818 Specify explicitly the @var{language} for the following input files
819 (rather than letting the compiler choose a default based on the file
820 name suffix). This option applies to all following input files until
821 the next @option{-x} option. Possible values for @var{language} are:
823 c c-header cpp-output
824 c++ c++-header c++-cpp-output
825 objective-c objective-c-header objc-cpp-output
826 assembler assembler-with-cpp
828 f77 f77-cpp-input ratfor
834 Turn off any specification of a language, so that subsequent files are
835 handled according to their file name suffixes (as they are if @option{-x}
836 has not been used at all).
838 @item -pass-exit-codes
839 @opindex pass-exit-codes
840 Normally the @command{gcc} program will exit with the code of 1 if any
841 phase of the compiler returns a non-success return code. If you specify
842 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
843 numerically highest error produced by any phase that returned an error
847 If you only want some of the stages of compilation, you can use
848 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
849 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
850 @command{gcc} is to stop. Note that some combinations (for example,
851 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
856 Compile or assemble the source files, but do not link. The linking
857 stage simply is not done. The ultimate output is in the form of an
858 object file for each source file.
860 By default, the object file name for a source file is made by replacing
861 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
863 Unrecognized input files, not requiring compilation or assembly, are
868 Stop after the stage of compilation proper; do not assemble. The output
869 is in the form of an assembler code file for each non-assembler input
872 By default, the assembler file name for a source file is made by
873 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
875 Input files that don't require compilation are ignored.
879 Stop after the preprocessing stage; do not run the compiler proper. The
880 output is in the form of preprocessed source code, which is sent to the
883 Input files which don't require preprocessing are ignored.
885 @cindex output file option
888 Place output in file @var{file}. This applies regardless to whatever
889 sort of output is being produced, whether it be an executable file,
890 an object file, an assembler file or preprocessed C code.
892 If you specify @option{-o} when compiling more than one input file, or
893 you are producing an executable file as output, all the source files
894 on the command line will be compiled at once.
896 If @option{-o} is not specified, the default is to put an executable file
897 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
898 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
899 all preprocessed C source on standard output.
903 Print (on standard error output) the commands executed to run the stages
904 of compilation. Also print the version number of the compiler driver
905 program and of the preprocessor and the compiler proper.
909 Like @option{-v} except the commands are not executed and all command
910 arguments are quoted. This is useful for shell scripts to capture the
911 driver-generated command lines.
915 Use pipes rather than temporary files for communication between the
916 various stages of compilation. This fails to work on some systems where
917 the assembler is unable to read from a pipe; but the GNU assembler has
922 Print (on the standard output) a description of the command line options
923 understood by @command{gcc}. If the @option{-v} option is also specified
924 then @option{--help} will also be passed on to the various processes
925 invoked by @command{gcc}, so that they can display the command line options
926 they accept. If the @option{-Wextra} option is also specified then command
927 line options which have no documentation associated with them will also
932 Print (on the standard output) a description of target specific command
933 line options for each tool.
937 Display the version number and copyrights of the invoked GCC.
941 @section Compiling C++ Programs
943 @cindex suffixes for C++ source
944 @cindex C++ source file suffixes
945 C++ source files conventionally use one of the suffixes @samp{.C},
946 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
947 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
948 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
949 files with these names and compiles them as C++ programs even if you
950 call the compiler the same way as for compiling C programs (usually
951 with the name @command{gcc}).
955 However, C++ programs often require class libraries as well as a
956 compiler that understands the C++ language---and under some
957 circumstances, you might want to compile programs or header files from
958 standard input, or otherwise without a suffix that flags them as C++
959 programs. You might also like to precompile a C header file with a
960 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
961 program that calls GCC with the default language set to C++, and
962 automatically specifies linking against the C++ library. On many
963 systems, @command{g++} is also installed with the name @command{c++}.
965 @cindex invoking @command{g++}
966 When you compile C++ programs, you may specify many of the same
967 command-line options that you use for compiling programs in any
968 language; or command-line options meaningful for C and related
969 languages; or options that are meaningful only for C++ programs.
970 @xref{C Dialect Options,,Options Controlling C Dialect}, for
971 explanations of options for languages related to C@.
972 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
973 explanations of options that are meaningful only for C++ programs.
975 @node C Dialect Options
976 @section Options Controlling C Dialect
977 @cindex dialect options
978 @cindex language dialect options
979 @cindex options, dialect
981 The following options control the dialect of C (or languages derived
982 from C, such as C++ and Objective-C) that the compiler accepts:
989 In C mode, support all ISO C90 programs. In C++ mode,
990 remove GNU extensions that conflict with ISO C++.
992 This turns off certain features of GCC that are incompatible with ISO
993 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
994 such as the @code{asm} and @code{typeof} keywords, and
995 predefined macros such as @code{unix} and @code{vax} that identify the
996 type of system you are using. It also enables the undesirable and
997 rarely used ISO trigraph feature. For the C compiler,
998 it disables recognition of C++ style @samp{//} comments as well as
999 the @code{inline} keyword.
1001 The alternate keywords @code{__asm__}, @code{__extension__},
1002 @code{__inline__} and @code{__typeof__} continue to work despite
1003 @option{-ansi}. You would not want to use them in an ISO C program, of
1004 course, but it is useful to put them in header files that might be included
1005 in compilations done with @option{-ansi}. Alternate predefined macros
1006 such as @code{__unix__} and @code{__vax__} are also available, with or
1007 without @option{-ansi}.
1009 The @option{-ansi} option does not cause non-ISO programs to be
1010 rejected gratuitously. For that, @option{-pedantic} is required in
1011 addition to @option{-ansi}. @xref{Warning Options}.
1013 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1014 option is used. Some header files may notice this macro and refrain
1015 from declaring certain functions or defining certain macros that the
1016 ISO standard doesn't call for; this is to avoid interfering with any
1017 programs that might use these names for other things.
1019 Functions which would normally be built in but do not have semantics
1020 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1021 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1022 built-in functions provided by GCC}, for details of the functions
1027 Determine the language standard. This option is currently only
1028 supported when compiling C or C++. A value for this option must be
1029 provided; possible values are
1034 ISO C90 (same as @option{-ansi}).
1036 @item iso9899:199409
1037 ISO C90 as modified in amendment 1.
1043 ISO C99. Note that this standard is not yet fully supported; see
1044 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1045 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1048 Default, ISO C90 plus GNU extensions (including some C99 features).
1052 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1053 this will become the default. The name @samp{gnu9x} is deprecated.
1056 The 1998 ISO C++ standard plus amendments.
1059 The same as @option{-std=c++98} plus GNU extensions. This is the
1060 default for C++ code.
1063 Even when this option is not specified, you can still use some of the
1064 features of newer standards in so far as they do not conflict with
1065 previous C standards. For example, you may use @code{__restrict__} even
1066 when @option{-std=c99} is not specified.
1068 The @option{-std} options specifying some version of ISO C have the same
1069 effects as @option{-ansi}, except that features that were not in ISO C90
1070 but are in the specified version (for example, @samp{//} comments and
1071 the @code{inline} keyword in ISO C99) are not disabled.
1073 @xref{Standards,,Language Standards Supported by GCC}, for details of
1074 these standard versions.
1076 @item -aux-info @var{filename}
1078 Output to the given filename prototyped declarations for all functions
1079 declared and/or defined in a translation unit, including those in header
1080 files. This option is silently ignored in any language other than C@.
1082 Besides declarations, the file indicates, in comments, the origin of
1083 each declaration (source file and line), whether the declaration was
1084 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1085 @samp{O} for old, respectively, in the first character after the line
1086 number and the colon), and whether it came from a declaration or a
1087 definition (@samp{C} or @samp{F}, respectively, in the following
1088 character). In the case of function definitions, a K&R-style list of
1089 arguments followed by their declarations is also provided, inside
1090 comments, after the declaration.
1094 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1095 keyword, so that code can use these words as identifiers. You can use
1096 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1097 instead. @option{-ansi} implies @option{-fno-asm}.
1099 In C++, this switch only affects the @code{typeof} keyword, since
1100 @code{asm} and @code{inline} are standard keywords. You may want to
1101 use the @option{-fno-gnu-keywords} flag instead, which has the same
1102 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1103 switch only affects the @code{asm} and @code{typeof} keywords, since
1104 @code{inline} is a standard keyword in ISO C99.
1107 @itemx -fno-builtin-@var{function}
1108 @opindex fno-builtin
1109 @cindex built-in functions
1110 Don't recognize built-in functions that do not begin with
1111 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1112 functions provided by GCC}, for details of the functions affected,
1113 including those which are not built-in functions when @option{-ansi} or
1114 @option{-std} options for strict ISO C conformance are used because they
1115 do not have an ISO standard meaning.
1117 GCC normally generates special code to handle certain built-in functions
1118 more efficiently; for instance, calls to @code{alloca} may become single
1119 instructions that adjust the stack directly, and calls to @code{memcpy}
1120 may become inline copy loops. The resulting code is often both smaller
1121 and faster, but since the function calls no longer appear as such, you
1122 cannot set a breakpoint on those calls, nor can you change the behavior
1123 of the functions by linking with a different library.
1125 With the @option{-fno-builtin-@var{function}} option
1126 only the built-in function @var{function} is
1127 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1128 function is named this is not built-in in this version of GCC, this
1129 option is ignored. There is no corresponding
1130 @option{-fbuiltin-@var{function}} option; if you wish to enable
1131 built-in functions selectively when using @option{-fno-builtin} or
1132 @option{-ffreestanding}, you may define macros such as:
1135 #define abs(n) __builtin_abs ((n))
1136 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1141 @cindex hosted environment
1143 Assert that compilation takes place in a hosted environment. This implies
1144 @option{-fbuiltin}. A hosted environment is one in which the
1145 entire standard library is available, and in which @code{main} has a return
1146 type of @code{int}. Examples are nearly everything except a kernel.
1147 This is equivalent to @option{-fno-freestanding}.
1149 @item -ffreestanding
1150 @opindex ffreestanding
1151 @cindex hosted environment
1153 Assert that compilation takes place in a freestanding environment. This
1154 implies @option{-fno-builtin}. A freestanding environment
1155 is one in which the standard library may not exist, and program startup may
1156 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1157 This is equivalent to @option{-fno-hosted}.
1159 @xref{Standards,,Language Standards Supported by GCC}, for details of
1160 freestanding and hosted environments.
1162 @item -fms-extensions
1163 @opindex fms-extensions
1164 Accept some non-standard constructs used in Microsoft header files.
1168 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1169 options for strict ISO C conformance) implies @option{-trigraphs}.
1171 @item -no-integrated-cpp
1172 @opindex no-integrated-cpp
1173 Performs a compilation in two passes: preprocessing and compiling. This
1174 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1175 @option{-B} option. The user supplied compilation step can then add in
1176 an additional preprocessing step after normal preprocessing but before
1177 compiling. The default is to use the integrated cpp (internal cpp)
1179 The semantics of this option will change if "cc1", "cc1plus", and
1180 "cc1obj" are merged.
1182 @cindex traditional C language
1183 @cindex C language, traditional
1185 @itemx -traditional-cpp
1186 @opindex traditional-cpp
1187 @opindex traditional
1188 Formerly, these options caused GCC to attempt to emulate a pre-standard
1189 C compiler. They are now only supported with the @option{-E} switch.
1190 The preprocessor continues to support a pre-standard mode. See the GNU
1191 CPP manual for details.
1193 @item -fcond-mismatch
1194 @opindex fcond-mismatch
1195 Allow conditional expressions with mismatched types in the second and
1196 third arguments. The value of such an expression is void. This option
1197 is not supported for C++.
1199 @item -funsigned-char
1200 @opindex funsigned-char
1201 Let the type @code{char} be unsigned, like @code{unsigned char}.
1203 Each kind of machine has a default for what @code{char} should
1204 be. It is either like @code{unsigned char} by default or like
1205 @code{signed char} by default.
1207 Ideally, a portable program should always use @code{signed char} or
1208 @code{unsigned char} when it depends on the signedness of an object.
1209 But many programs have been written to use plain @code{char} and
1210 expect it to be signed, or expect it to be unsigned, depending on the
1211 machines they were written for. This option, and its inverse, let you
1212 make such a program work with the opposite default.
1214 The type @code{char} is always a distinct type from each of
1215 @code{signed char} or @code{unsigned char}, even though its behavior
1216 is always just like one of those two.
1219 @opindex fsigned-char
1220 Let the type @code{char} be signed, like @code{signed char}.
1222 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1223 the negative form of @option{-funsigned-char}. Likewise, the option
1224 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1226 @item -fsigned-bitfields
1227 @itemx -funsigned-bitfields
1228 @itemx -fno-signed-bitfields
1229 @itemx -fno-unsigned-bitfields
1230 @opindex fsigned-bitfields
1231 @opindex funsigned-bitfields
1232 @opindex fno-signed-bitfields
1233 @opindex fno-unsigned-bitfields
1234 These options control whether a bit-field is signed or unsigned, when the
1235 declaration does not use either @code{signed} or @code{unsigned}. By
1236 default, such a bit-field is signed, because this is consistent: the
1237 basic integer types such as @code{int} are signed types.
1239 @item -fwritable-strings
1240 @opindex fwritable-strings
1241 Store string constants in the writable data segment and don't uniquize
1242 them. This is for compatibility with old programs which assume they can
1243 write into string constants.
1245 Writing into string constants is a very bad idea; ``constants'' should
1248 This option is deprecated.
1251 @node C++ Dialect Options
1252 @section Options Controlling C++ Dialect
1254 @cindex compiler options, C++
1255 @cindex C++ options, command line
1256 @cindex options, C++
1257 This section describes the command-line options that are only meaningful
1258 for C++ programs; but you can also use most of the GNU compiler options
1259 regardless of what language your program is in. For example, you
1260 might compile a file @code{firstClass.C} like this:
1263 g++ -g -frepo -O -c firstClass.C
1267 In this example, only @option{-frepo} is an option meant
1268 only for C++ programs; you can use the other options with any
1269 language supported by GCC@.
1271 Here is a list of options that are @emph{only} for compiling C++ programs:
1275 @item -fabi-version=@var{n}
1276 @opindex fabi-version
1277 Use version @var{n} of the C++ ABI. Version 2 is the version of the
1278 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1279 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1280 the version that conforms most closely to the C++ ABI specification.
1281 Therefore, the ABI obtained using version 0 will change as ABI bugs
1284 The default is version 1.
1286 @item -fno-access-control
1287 @opindex fno-access-control
1288 Turn off all access checking. This switch is mainly useful for working
1289 around bugs in the access control code.
1293 Check that the pointer returned by @code{operator new} is non-null
1294 before attempting to modify the storage allocated. This check is
1295 normally unnecessary because the C++ standard specifies that
1296 @code{operator new} will only return @code{0} if it is declared
1297 @samp{throw()}, in which case the compiler will always check the
1298 return value even without this option. In all other cases, when
1299 @code{operator new} has a non-empty exception specification, memory
1300 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1301 @samp{new (nothrow)}.
1303 @item -fconserve-space
1304 @opindex fconserve-space
1305 Put uninitialized or runtime-initialized global variables into the
1306 common segment, as C does. This saves space in the executable at the
1307 cost of not diagnosing duplicate definitions. If you compile with this
1308 flag and your program mysteriously crashes after @code{main()} has
1309 completed, you may have an object that is being destroyed twice because
1310 two definitions were merged.
1312 This option is no longer useful on most targets, now that support has
1313 been added for putting variables into BSS without making them common.
1315 @item -fno-const-strings
1316 @opindex fno-const-strings
1317 Give string constants type @code{char *} instead of type @code{const
1318 char *}. By default, G++ uses type @code{const char *} as required by
1319 the standard. Even if you use @option{-fno-const-strings}, you cannot
1320 actually modify the value of a string constant, unless you also use
1321 @option{-fwritable-strings}.
1323 This option might be removed in a future release of G++. For maximum
1324 portability, you should structure your code so that it works with
1325 string constants that have type @code{const char *}.
1327 @item -fno-elide-constructors
1328 @opindex fno-elide-constructors
1329 The C++ standard allows an implementation to omit creating a temporary
1330 which is only used to initialize another object of the same type.
1331 Specifying this option disables that optimization, and forces G++ to
1332 call the copy constructor in all cases.
1334 @item -fno-enforce-eh-specs
1335 @opindex fno-enforce-eh-specs
1336 Don't check for violation of exception specifications at runtime. This
1337 option violates the C++ standard, but may be useful for reducing code
1338 size in production builds, much like defining @samp{NDEBUG}. The compiler
1339 will still optimize based on the exception specifications.
1342 @itemx -fno-for-scope
1344 @opindex fno-for-scope
1345 If @option{-ffor-scope} is specified, the scope of variables declared in
1346 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1347 as specified by the C++ standard.
1348 If @option{-fno-for-scope} is specified, the scope of variables declared in
1349 a @i{for-init-statement} extends to the end of the enclosing scope,
1350 as was the case in old versions of G++, and other (traditional)
1351 implementations of C++.
1353 The default if neither flag is given to follow the standard,
1354 but to allow and give a warning for old-style code that would
1355 otherwise be invalid, or have different behavior.
1357 @item -fno-gnu-keywords
1358 @opindex fno-gnu-keywords
1359 Do not recognize @code{typeof} as a keyword, so that code can use this
1360 word as an identifier. You can use the keyword @code{__typeof__} instead.
1361 @option{-ansi} implies @option{-fno-gnu-keywords}.
1363 @item -fno-implicit-templates
1364 @opindex fno-implicit-templates
1365 Never emit code for non-inline templates which are instantiated
1366 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1367 @xref{Template Instantiation}, for more information.
1369 @item -fno-implicit-inline-templates
1370 @opindex fno-implicit-inline-templates
1371 Don't emit code for implicit instantiations of inline templates, either.
1372 The default is to handle inlines differently so that compiles with and
1373 without optimization will need the same set of explicit instantiations.
1375 @item -fno-implement-inlines
1376 @opindex fno-implement-inlines
1377 To save space, do not emit out-of-line copies of inline functions
1378 controlled by @samp{#pragma implementation}. This will cause linker
1379 errors if these functions are not inlined everywhere they are called.
1381 @item -fms-extensions
1382 @opindex fms-extensions
1383 Disable pedantic warnings about constructs used in MFC, such as implicit
1384 int and getting a pointer to member function via non-standard syntax.
1386 @item -fno-nonansi-builtins
1387 @opindex fno-nonansi-builtins
1388 Disable built-in declarations of functions that are not mandated by
1389 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1390 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1392 @item -fno-operator-names
1393 @opindex fno-operator-names
1394 Do not treat the operator name keywords @code{and}, @code{bitand},
1395 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1396 synonyms as keywords.
1398 @item -fno-optional-diags
1399 @opindex fno-optional-diags
1400 Disable diagnostics that the standard says a compiler does not need to
1401 issue. Currently, the only such diagnostic issued by G++ is the one for
1402 a name having multiple meanings within a class.
1405 @opindex fpermissive
1406 Downgrade some diagnostics about nonconformant code from errors to
1407 warnings. Thus, using @option{-fpermissive} will allow some
1408 nonconforming code to compile.
1412 Enable automatic template instantiation at link time. This option also
1413 implies @option{-fno-implicit-templates}. @xref{Template
1414 Instantiation}, for more information.
1418 Disable generation of information about every class with virtual
1419 functions for use by the C++ runtime type identification features
1420 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1421 of the language, you can save some space by using this flag. Note that
1422 exception handling uses the same information, but it will generate it as
1427 Emit statistics about front-end processing at the end of the compilation.
1428 This information is generally only useful to the G++ development team.
1430 @item -ftemplate-depth-@var{n}
1431 @opindex ftemplate-depth
1432 Set the maximum instantiation depth for template classes to @var{n}.
1433 A limit on the template instantiation depth is needed to detect
1434 endless recursions during template class instantiation. ANSI/ISO C++
1435 conforming programs must not rely on a maximum depth greater than 17.
1437 @item -fuse-cxa-atexit
1438 @opindex fuse-cxa-atexit
1439 Register destructors for objects with static storage duration with the
1440 @code{__cxa_atexit} function rather than the @code{atexit} function.
1441 This option is required for fully standards-compliant handling of static
1442 destructors, but will only work if your C library supports
1443 @code{__cxa_atexit}.
1447 Do not use weak symbol support, even if it is provided by the linker.
1448 By default, G++ will use weak symbols if they are available. This
1449 option exists only for testing, and should not be used by end-users;
1450 it will result in inferior code and has no benefits. This option may
1451 be removed in a future release of G++.
1455 Do not search for header files in the standard directories specific to
1456 C++, but do still search the other standard directories. (This option
1457 is used when building the C++ library.)
1460 In addition, these optimization, warning, and code generation options
1461 have meanings only for C++ programs:
1464 @item -fno-default-inline
1465 @opindex fno-default-inline
1466 Do not assume @samp{inline} for functions defined inside a class scope.
1467 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1468 functions will have linkage like inline functions; they just won't be
1471 @item -Wabi @r{(C++ only)}
1473 Warn when G++ generates code that is probably not compatible with the
1474 vendor-neutral C++ ABI. Although an effort has been made to warn about
1475 all such cases, there are probably some cases that are not warned about,
1476 even though G++ is generating incompatible code. There may also be
1477 cases where warnings are emitted even though the code that is generated
1480 You should rewrite your code to avoid these warnings if you are
1481 concerned about the fact that code generated by G++ may not be binary
1482 compatible with code generated by other compilers.
1484 The known incompatibilities at this point include:
1489 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1490 pack data into the same byte as a base class. For example:
1493 struct A @{ virtual void f(); int f1 : 1; @};
1494 struct B : public A @{ int f2 : 1; @};
1498 In this case, G++ will place @code{B::f2} into the same byte
1499 as@code{A::f1}; other compilers will not. You can avoid this problem
1500 by explicitly padding @code{A} so that its size is a multiple of the
1501 byte size on your platform; that will cause G++ and other compilers to
1502 layout @code{B} identically.
1505 Incorrect handling of tail-padding for virtual bases. G++ does not use
1506 tail padding when laying out virtual bases. For example:
1509 struct A @{ virtual void f(); char c1; @};
1510 struct B @{ B(); char c2; @};
1511 struct C : public A, public virtual B @{@};
1515 In this case, G++ will not place @code{B} into the tail-padding for
1516 @code{A}; other compilers will. You can avoid this problem by
1517 explicitly padding @code{A} so that its size is a multiple of its
1518 alignment (ignoring virtual base classes); that will cause G++ and other
1519 compilers to layout @code{C} identically.
1522 Incorrect handling of bit-fields with declared widths greater than that
1523 of their underlying types, when the bit-fields appear in a union. For
1527 union U @{ int i : 4096; @};
1531 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1532 union too small by the number of bits in an @code{int}.
1535 Empty classes can be placed at incorrect offsets. For example:
1545 struct C : public B, public A @{@};
1549 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1550 it should be placed at offset zero. G++ mistakenly believes that the
1551 @code{A} data member of @code{B} is already at offset zero.
1554 Names of template functions whose types involve @code{typename} or
1555 template template parameters can be mangled incorrectly.
1558 template <typename Q>
1559 void f(typename Q::X) @{@}
1561 template <template <typename> class Q>
1562 void f(typename Q<int>::X) @{@}
1566 Instantiations of these templates may be mangled incorrectly.
1570 @item -Wctor-dtor-privacy @r{(C++ only)}
1571 @opindex Wctor-dtor-privacy
1572 Warn when a class seems unusable because all the constructors or
1573 destructors in that class are private, and it has neither friends nor
1574 public static member functions.
1576 @item -Wnon-virtual-dtor @r{(C++ only)}
1577 @opindex Wnon-virtual-dtor
1578 Warn when a class appears to be polymorphic, thereby requiring a virtual
1579 destructor, yet it declares a non-virtual one.
1580 This warning is enabled by @option{-Wall}.
1582 @item -Wreorder @r{(C++ only)}
1584 @cindex reordering, warning
1585 @cindex warning for reordering of member initializers
1586 Warn when the order of member initializers given in the code does not
1587 match the order in which they must be executed. For instance:
1593 A(): j (0), i (1) @{ @}
1597 The compiler will rearrange the member initializers for @samp{i}
1598 and @samp{j} to match the declaration order of the members, emitting
1599 a warning to that effect. This warning is enabled by @option{-Wall}.
1602 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1605 @item -Weffc++ @r{(C++ only)}
1607 Warn about violations of the following style guidelines from Scott Meyers'
1608 @cite{Effective C++} book:
1612 Item 11: Define a copy constructor and an assignment operator for classes
1613 with dynamically allocated memory.
1616 Item 12: Prefer initialization to assignment in constructors.
1619 Item 14: Make destructors virtual in base classes.
1622 Item 15: Have @code{operator=} return a reference to @code{*this}.
1625 Item 23: Don't try to return a reference when you must return an object.
1629 Also warn about violations of the following style guidelines from
1630 Scott Meyers' @cite{More Effective C++} book:
1634 Item 6: Distinguish between prefix and postfix forms of increment and
1635 decrement operators.
1638 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1642 When selecting this option, be aware that the standard library
1643 headers do not obey all of these guidelines; use @samp{grep -v}
1644 to filter out those warnings.
1646 @item -Wno-deprecated @r{(C++ only)}
1647 @opindex Wno-deprecated
1648 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1650 @item -Wno-non-template-friend @r{(C++ only)}
1651 @opindex Wno-non-template-friend
1652 Disable warnings when non-templatized friend functions are declared
1653 within a template. Since the advent of explicit template specification
1654 support in G++, if the name of the friend is an unqualified-id (i.e.,
1655 @samp{friend foo(int)}), the C++ language specification demands that the
1656 friend declare or define an ordinary, nontemplate function. (Section
1657 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1658 could be interpreted as a particular specialization of a templatized
1659 function. Because this non-conforming behavior is no longer the default
1660 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1661 check existing code for potential trouble spots and is on by default.
1662 This new compiler behavior can be turned off with
1663 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1664 but disables the helpful warning.
1666 @item -Wold-style-cast @r{(C++ only)}
1667 @opindex Wold-style-cast
1668 Warn if an old-style (C-style) cast to a non-void type is used within
1669 a C++ program. The new-style casts (@samp{static_cast},
1670 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1671 unintended effects and much easier to search for.
1673 @item -Woverloaded-virtual @r{(C++ only)}
1674 @opindex Woverloaded-virtual
1675 @cindex overloaded virtual fn, warning
1676 @cindex warning for overloaded virtual fn
1677 Warn when a function declaration hides virtual functions from a
1678 base class. For example, in:
1685 struct B: public A @{
1690 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1698 will fail to compile.
1700 @item -Wno-pmf-conversions @r{(C++ only)}
1701 @opindex Wno-pmf-conversions
1702 Disable the diagnostic for converting a bound pointer to member function
1705 @item -Wsign-promo @r{(C++ only)}
1706 @opindex Wsign-promo
1707 Warn when overload resolution chooses a promotion from unsigned or
1708 enumeral type to a signed type, over a conversion to an unsigned type of
1709 the same size. Previous versions of G++ would try to preserve
1710 unsignedness, but the standard mandates the current behavior.
1712 @item -Wsynth @r{(C++ only)}
1714 @cindex warning for synthesized methods
1715 @cindex synthesized methods, warning
1716 Warn when G++'s synthesis behavior does not match that of cfront. For
1722 A& operator = (int);
1732 In this example, G++ will synthesize a default @samp{A& operator =
1733 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1736 @node Objective-C Dialect Options
1737 @section Options Controlling Objective-C Dialect
1739 @cindex compiler options, Objective-C
1740 @cindex Objective-C options, command line
1741 @cindex options, Objective-C
1742 (NOTE: This manual does not describe the Objective-C language itself. See
1743 @w{@uref{http://gcc.gnu.org/readings.html}} for references.)
1745 This section describes the command-line options that are only meaningful
1746 for Objective-C programs, but you can also use most of the GNU compiler
1747 options regardless of what language your program is in. For example,
1748 you might compile a file @code{some_class.m} like this:
1751 gcc -g -fgnu-runtime -O -c some_class.m
1755 In this example, @option{-fgnu-runtime} is an option meant only for
1756 Objective-C programs; you can use the other options with any language
1759 Here is a list of options that are @emph{only} for compiling Objective-C
1763 @item -fconstant-string-class=@var{class-name}
1764 @opindex fconstant-string-class
1765 Use @var{class-name} as the name of the class to instantiate for each
1766 literal string specified with the syntax @code{@@"@dots{}"}. The default
1767 class name is @code{NXConstantString} if the GNU runtime is being used, and
1768 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1769 @option{-fconstant-cfstrings} option, if also present, will override the
1770 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1771 to be laid out as constant CoreFoundation strings.
1774 @opindex fgnu-runtime
1775 Generate object code compatible with the standard GNU Objective-C
1776 runtime. This is the default for most types of systems.
1778 @item -fnext-runtime
1779 @opindex fnext-runtime
1780 Generate output compatible with the NeXT runtime. This is the default
1781 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1782 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1785 @item -fno-nil-receivers
1786 @opindex -fno-nil-receivers
1787 Assume that all Objective-C message dispatches (e.g.,
1788 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1789 is not @code{nil}. This allows for more efficient entry points in the runtime to be
1790 used. Currently, this option is only available in conjunction with
1791 the NeXT runtime on Mac OS X 10.3 and later.
1793 @item -fobjc-exceptions
1794 @opindex -fobjc-exceptions
1795 Enable syntactic support for structured exception handling in Objective-C,
1796 similar to what is offered by C++ and Java. Currently, this option is only
1797 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1805 @@catch (AnObjCClass *exc) @{
1812 @@catch (AnotherClass *exc) @{
1815 @@catch (id allOthers) @{
1825 The @code{@@throw} statement may appear anywhere in an Objective-C or
1826 Objective-C++ program; when used inside of a @code{@@catch} block, the
1827 @code{@@throw} may appear without an argument (as shown above), in which case
1828 the object caught by the @code{@@catch} will be rethrown.
1830 Note that only (pointers to) Objective-C objects may be thrown and
1831 caught using this scheme. When an object is thrown, it will be caught
1832 by the nearest @code{@@catch} clause capable of handling objects of that type,
1833 analogously to how @code{catch} blocks work in C++ and Java. A
1834 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1835 any and all Objective-C exceptions not caught by previous @code{@@catch}
1838 The @code{@@finally} clause, if present, will be executed upon exit from the
1839 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1840 regardless of whether any exceptions are thrown, caught or rethrown
1841 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1842 of the @code{finally} clause in Java.
1844 There are several caveats to using the new exception mechanism:
1848 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1849 idioms provided by the @code{NSException} class, the new
1850 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1851 systems, due to additional functionality needed in the (NeXT) Objective-C
1855 As mentioned above, the new exceptions do not support handling
1856 types other than Objective-C objects. Furthermore, when used from
1857 Objective-C++, the Objective-C exception model does not interoperate with C++
1858 exceptions at this time. This means you cannot @code{@@throw} an exception
1859 from Objective-C and @code{catch} it in C++, or vice versa
1860 (i.e., @code{throw @dots{} @@catch}).
1863 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1864 blocks for thread-safe execution:
1867 @@synchronized (ObjCClass *guard) @{
1872 Upon entering the @code{@@synchronized} block, a thread of execution shall
1873 first check whether a lock has been placed on the corresponding @code{guard}
1874 object by another thread. If it has, the current thread shall wait until
1875 the other thread relinquishes its lock. Once @code{guard} becomes available,
1876 the current thread will place its own lock on it, execute the code contained in
1877 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1878 making @code{guard} available to other threads).
1880 Unlike Java, Objective-C does not allow for entire methods to be marked
1881 @code{@@synchronized}. Note that throwing exceptions out of
1882 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1883 to be unlocked properly.
1885 @item -freplace-objc-classes
1886 @opindex -freplace-objc-classes
1887 Emit a special marker instructing @command{ld(1)} not to statically link in
1888 the resulting object file, and allow @command{dyld(1)} to load it in at
1889 run time instead. This is used in conjunction with the Fix-and-Continue
1890 debugging mode, where the object file in question may be recompiled and
1891 dynamically reloaded in the course of program execution, without the need
1892 to restart the program itself. Currently, Fix-and-Continue functionality
1893 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1897 @opindex -fzero-link
1898 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1899 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1900 compile time) with static class references that get initialized at load time,
1901 which improves run-time performance. Specifying the @option{-fzero-link} flag
1902 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1903 to be retained. This is useful in Zero-Link debugging mode, since it allows
1904 for individual class implementations to be modified during program execution.
1908 Dump interface declarations for all classes seen in the source file to a
1909 file named @file{@var{sourcename}.decl}.
1912 @opindex Wno-protocol
1913 If a class is declared to implement a protocol, a warning is issued for
1914 every method in the protocol that is not implemented by the class. The
1915 default behavior is to issue a warning for every method not explicitly
1916 implemented in the class, even if a method implementation is inherited
1917 from the superclass. If you use the @code{-Wno-protocol} option, then
1918 methods inherited from the superclass are considered to be implemented,
1919 and no warning is issued for them.
1923 Warn if multiple methods of different types for the same selector are
1924 found during compilation. The check is performed on the list of methods
1925 in the final stage of compilation. Additionally, a check is performed
1926 for each selector appearing in a @code{@@selector(@dots{})}
1927 expression, and a corresponding method for that selector has been found
1928 during compilation. Because these checks scan the method table only at
1929 the end of compilation, these warnings are not produced if the final
1930 stage of compilation is not reached, for example because an error is
1931 found during compilation, or because the @code{-fsyntax-only} option is
1934 @item -Wundeclared-selector
1935 @opindex Wundeclared-selector
1936 Warn if a @code{@@selector(@dots{})} expression referring to an
1937 undeclared selector is found. A selector is considered undeclared if no
1938 method with that name has been declared before the
1939 @code{@@selector(@dots{})} expression, either explicitly in an
1940 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
1941 an @code{@@implementation} section. This option always performs its
1942 checks as soon as a @code{@@selector(@dots{})} expression is found,
1943 while @code{-Wselector} only performs its checks in the final stage of
1944 compilation. This also enforces the coding style convention
1945 that methods and selectors must be declared before being used.
1947 @item -print-objc-runtime-info
1948 @opindex -print-objc-runtime-info
1949 Generate C header describing the largest structure that is passed by
1954 @node Language Independent Options
1955 @section Options to Control Diagnostic Messages Formatting
1956 @cindex options to control diagnostics formatting
1957 @cindex diagnostic messages
1958 @cindex message formatting
1960 Traditionally, diagnostic messages have been formatted irrespective of
1961 the output device's aspect (e.g.@: its width, @dots{}). The options described
1962 below can be used to control the diagnostic messages formatting
1963 algorithm, e.g.@: how many characters per line, how often source location
1964 information should be reported. Right now, only the C++ front end can
1965 honor these options. However it is expected, in the near future, that
1966 the remaining front ends would be able to digest them correctly.
1969 @item -fmessage-length=@var{n}
1970 @opindex fmessage-length
1971 Try to format error messages so that they fit on lines of about @var{n}
1972 characters. The default is 72 characters for @command{g++} and 0 for the rest of
1973 the front ends supported by GCC@. If @var{n} is zero, then no
1974 line-wrapping will be done; each error message will appear on a single
1977 @opindex fdiagnostics-show-location
1978 @item -fdiagnostics-show-location=once
1979 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1980 reporter to emit @emph{once} source location information; that is, in
1981 case the message is too long to fit on a single physical line and has to
1982 be wrapped, the source location won't be emitted (as prefix) again,
1983 over and over, in subsequent continuation lines. This is the default
1986 @item -fdiagnostics-show-location=every-line
1987 Only meaningful in line-wrapping mode. Instructs the diagnostic
1988 messages reporter to emit the same source location information (as
1989 prefix) for physical lines that result from the process of breaking
1990 a message which is too long to fit on a single line.
1994 @node Warning Options
1995 @section Options to Request or Suppress Warnings
1996 @cindex options to control warnings
1997 @cindex warning messages
1998 @cindex messages, warning
1999 @cindex suppressing warnings
2001 Warnings are diagnostic messages that report constructions which
2002 are not inherently erroneous but which are risky or suggest there
2003 may have been an error.
2005 You can request many specific warnings with options beginning @samp{-W},
2006 for example @option{-Wimplicit} to request warnings on implicit
2007 declarations. Each of these specific warning options also has a
2008 negative form beginning @samp{-Wno-} to turn off warnings;
2009 for example, @option{-Wno-implicit}. This manual lists only one of the
2010 two forms, whichever is not the default.
2012 The following options control the amount and kinds of warnings produced
2013 by GCC; for further, language-specific options also refer to
2014 @ref{C++ Dialect Options} and @ref{Objective-C Dialect Options}.
2017 @cindex syntax checking
2019 @opindex fsyntax-only
2020 Check the code for syntax errors, but don't do anything beyond that.
2024 Issue all the warnings demanded by strict ISO C and ISO C++;
2025 reject all programs that use forbidden extensions, and some other
2026 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2027 version of the ISO C standard specified by any @option{-std} option used.
2029 Valid ISO C and ISO C++ programs should compile properly with or without
2030 this option (though a rare few will require @option{-ansi} or a
2031 @option{-std} option specifying the required version of ISO C)@. However,
2032 without this option, certain GNU extensions and traditional C and C++
2033 features are supported as well. With this option, they are rejected.
2035 @option{-pedantic} does not cause warning messages for use of the
2036 alternate keywords whose names begin and end with @samp{__}. Pedantic
2037 warnings are also disabled in the expression that follows
2038 @code{__extension__}. However, only system header files should use
2039 these escape routes; application programs should avoid them.
2040 @xref{Alternate Keywords}.
2042 Some users try to use @option{-pedantic} to check programs for strict ISO
2043 C conformance. They soon find that it does not do quite what they want:
2044 it finds some non-ISO practices, but not all---only those for which
2045 ISO C @emph{requires} a diagnostic, and some others for which
2046 diagnostics have been added.
2048 A feature to report any failure to conform to ISO C might be useful in
2049 some instances, but would require considerable additional work and would
2050 be quite different from @option{-pedantic}. We don't have plans to
2051 support such a feature in the near future.
2053 Where the standard specified with @option{-std} represents a GNU
2054 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2055 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2056 extended dialect is based. Warnings from @option{-pedantic} are given
2057 where they are required by the base standard. (It would not make sense
2058 for such warnings to be given only for features not in the specified GNU
2059 C dialect, since by definition the GNU dialects of C include all
2060 features the compiler supports with the given option, and there would be
2061 nothing to warn about.)
2063 @item -pedantic-errors
2064 @opindex pedantic-errors
2065 Like @option{-pedantic}, except that errors are produced rather than
2070 Inhibit all warning messages.
2074 Inhibit warning messages about the use of @samp{#import}.
2076 @item -Wchar-subscripts
2077 @opindex Wchar-subscripts
2078 Warn if an array subscript has type @code{char}. This is a common cause
2079 of error, as programmers often forget that this type is signed on some
2084 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2085 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2089 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2090 the arguments supplied have types appropriate to the format string
2091 specified, and that the conversions specified in the format string make
2092 sense. This includes standard functions, and others specified by format
2093 attributes (@pxref{Function Attributes}), in the @code{printf},
2094 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2095 not in the C standard) families.
2097 The formats are checked against the format features supported by GNU
2098 libc version 2.2. These include all ISO C90 and C99 features, as well
2099 as features from the Single Unix Specification and some BSD and GNU
2100 extensions. Other library implementations may not support all these
2101 features; GCC does not support warning about features that go beyond a
2102 particular library's limitations. However, if @option{-pedantic} is used
2103 with @option{-Wformat}, warnings will be given about format features not
2104 in the selected standard version (but not for @code{strfmon} formats,
2105 since those are not in any version of the C standard). @xref{C Dialect
2106 Options,,Options Controlling C Dialect}.
2108 Since @option{-Wformat} also checks for null format arguments for
2109 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2111 @option{-Wformat} is included in @option{-Wall}. For more control over some
2112 aspects of format checking, the options @option{-Wformat-y2k},
2113 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2114 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2115 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2118 @opindex Wformat-y2k
2119 If @option{-Wformat} is specified, also warn about @code{strftime}
2120 formats which may yield only a two-digit year.
2122 @item -Wno-format-extra-args
2123 @opindex Wno-format-extra-args
2124 If @option{-Wformat} is specified, do not warn about excess arguments to a
2125 @code{printf} or @code{scanf} format function. The C standard specifies
2126 that such arguments are ignored.
2128 Where the unused arguments lie between used arguments that are
2129 specified with @samp{$} operand number specifications, normally
2130 warnings are still given, since the implementation could not know what
2131 type to pass to @code{va_arg} to skip the unused arguments. However,
2132 in the case of @code{scanf} formats, this option will suppress the
2133 warning if the unused arguments are all pointers, since the Single
2134 Unix Specification says that such unused arguments are allowed.
2136 @item -Wno-format-zero-length
2137 @opindex Wno-format-zero-length
2138 If @option{-Wformat} is specified, do not warn about zero-length formats.
2139 The C standard specifies that zero-length formats are allowed.
2141 @item -Wformat-nonliteral
2142 @opindex Wformat-nonliteral
2143 If @option{-Wformat} is specified, also warn if the format string is not a
2144 string literal and so cannot be checked, unless the format function
2145 takes its format arguments as a @code{va_list}.
2147 @item -Wformat-security
2148 @opindex Wformat-security
2149 If @option{-Wformat} is specified, also warn about uses of format
2150 functions that represent possible security problems. At present, this
2151 warns about calls to @code{printf} and @code{scanf} functions where the
2152 format string is not a string literal and there are no format arguments,
2153 as in @code{printf (foo);}. This may be a security hole if the format
2154 string came from untrusted input and contains @samp{%n}. (This is
2155 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2156 in future warnings may be added to @option{-Wformat-security} that are not
2157 included in @option{-Wformat-nonliteral}.)
2161 Enable @option{-Wformat} plus format checks not included in
2162 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2163 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2167 Warn about passing a null pointer for arguments marked as
2168 requiring a non-null value by the @code{nonnull} function attribute.
2170 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2171 can be disabled with the @option{-Wno-nonnull} option.
2173 @item -Winit-self @r{(C, C++, and Objective-C only)}
2175 Warn about uninitialized variables which are initialized with themselves.
2176 Note this option can only be used with the @option{-Wuninitialized} option,
2177 which in turn only works with @option{-O1} and above.
2179 For example, GCC will warn about @code{i} being uninitialized in the
2180 following snippet only when @option{-Winit-self} has been specified:
2191 @item -Wimplicit-int
2192 @opindex Wimplicit-int
2193 Warn when a declaration does not specify a type.
2195 @item -Wimplicit-function-declaration
2196 @itemx -Werror-implicit-function-declaration
2197 @opindex Wimplicit-function-declaration
2198 @opindex Werror-implicit-function-declaration
2199 Give a warning (or error) whenever a function is used before being
2204 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2208 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2209 function with external linkage, returning int, taking either zero
2210 arguments, two, or three arguments of appropriate types.
2212 @item -Wmissing-braces
2213 @opindex Wmissing-braces
2214 Warn if an aggregate or union initializer is not fully bracketed. In
2215 the following example, the initializer for @samp{a} is not fully
2216 bracketed, but that for @samp{b} is fully bracketed.
2219 int a[2][2] = @{ 0, 1, 2, 3 @};
2220 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2224 @opindex Wparentheses
2225 Warn if parentheses are omitted in certain contexts, such
2226 as when there is an assignment in a context where a truth value
2227 is expected, or when operators are nested whose precedence people
2228 often get confused about.
2230 Also warn about constructions where there may be confusion to which
2231 @code{if} statement an @code{else} branch belongs. Here is an example of
2246 In C, every @code{else} branch belongs to the innermost possible @code{if}
2247 statement, which in this example is @code{if (b)}. This is often not
2248 what the programmer expected, as illustrated in the above example by
2249 indentation the programmer chose. When there is the potential for this
2250 confusion, GCC will issue a warning when this flag is specified.
2251 To eliminate the warning, add explicit braces around the innermost
2252 @code{if} statement so there is no way the @code{else} could belong to
2253 the enclosing @code{if}. The resulting code would look like this:
2269 @item -Wsequence-point
2270 @opindex Wsequence-point
2271 Warn about code that may have undefined semantics because of violations
2272 of sequence point rules in the C standard.
2274 The C standard defines the order in which expressions in a C program are
2275 evaluated in terms of @dfn{sequence points}, which represent a partial
2276 ordering between the execution of parts of the program: those executed
2277 before the sequence point, and those executed after it. These occur
2278 after the evaluation of a full expression (one which is not part of a
2279 larger expression), after the evaluation of the first operand of a
2280 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2281 function is called (but after the evaluation of its arguments and the
2282 expression denoting the called function), and in certain other places.
2283 Other than as expressed by the sequence point rules, the order of
2284 evaluation of subexpressions of an expression is not specified. All
2285 these rules describe only a partial order rather than a total order,
2286 since, for example, if two functions are called within one expression
2287 with no sequence point between them, the order in which the functions
2288 are called is not specified. However, the standards committee have
2289 ruled that function calls do not overlap.
2291 It is not specified when between sequence points modifications to the
2292 values of objects take effect. Programs whose behavior depends on this
2293 have undefined behavior; the C standard specifies that ``Between the
2294 previous and next sequence point an object shall have its stored value
2295 modified at most once by the evaluation of an expression. Furthermore,
2296 the prior value shall be read only to determine the value to be
2297 stored.''. If a program breaks these rules, the results on any
2298 particular implementation are entirely unpredictable.
2300 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2301 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2302 diagnosed by this option, and it may give an occasional false positive
2303 result, but in general it has been found fairly effective at detecting
2304 this sort of problem in programs.
2306 The present implementation of this option only works for C programs. A
2307 future implementation may also work for C++ programs.
2309 The C standard is worded confusingly, therefore there is some debate
2310 over the precise meaning of the sequence point rules in subtle cases.
2311 Links to discussions of the problem, including proposed formal
2312 definitions, may be found on our readings page, at
2313 @w{@uref{http://gcc.gnu.org/readings.html}}.
2316 @opindex Wreturn-type
2317 Warn whenever a function is defined with a return-type that defaults to
2318 @code{int}. Also warn about any @code{return} statement with no
2319 return-value in a function whose return-type is not @code{void}.
2321 For C++, a function without return type always produces a diagnostic
2322 message, even when @option{-Wno-return-type} is specified. The only
2323 exceptions are @samp{main} and functions defined in system headers.
2327 Warn whenever a @code{switch} statement has an index of enumeral type
2328 and lacks a @code{case} for one or more of the named codes of that
2329 enumeration. (The presence of a @code{default} label prevents this
2330 warning.) @code{case} labels outside the enumeration range also
2331 provoke warnings when this option is used.
2333 @item -Wswitch-default
2334 @opindex Wswitch-switch
2335 Warn whenever a @code{switch} statement does not have a @code{default}
2339 @opindex Wswitch-enum
2340 Warn whenever a @code{switch} statement has an index of enumeral type
2341 and lacks a @code{case} for one or more of the named codes of that
2342 enumeration. @code{case} labels outside the enumeration range also
2343 provoke warnings when this option is used.
2347 Warn if any trigraphs are encountered that might change the meaning of
2348 the program (trigraphs within comments are not warned about).
2350 @item -Wunused-function
2351 @opindex Wunused-function
2352 Warn whenever a static function is declared but not defined or a
2353 non\-inline static function is unused.
2355 @item -Wunused-label
2356 @opindex Wunused-label
2357 Warn whenever a label is declared but not used.
2359 To suppress this warning use the @samp{unused} attribute
2360 (@pxref{Variable Attributes}).
2362 @item -Wunused-parameter
2363 @opindex Wunused-parameter
2364 Warn whenever a function parameter is unused aside from its declaration.
2366 To suppress this warning use the @samp{unused} attribute
2367 (@pxref{Variable Attributes}).
2369 @item -Wunused-variable
2370 @opindex Wunused-variable
2371 Warn whenever a local variable or non-constant static variable is unused
2372 aside from its declaration
2374 To suppress this warning use the @samp{unused} attribute
2375 (@pxref{Variable Attributes}).
2377 @item -Wunused-value
2378 @opindex Wunused-value
2379 Warn whenever a statement computes a result that is explicitly not used.
2381 To suppress this warning cast the expression to @samp{void}.
2385 All the above @option{-Wunused} options combined.
2387 In order to get a warning about an unused function parameter, you must
2388 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2389 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2391 @item -Wuninitialized
2392 @opindex Wuninitialized
2393 Warn if an automatic variable is used without first being initialized or
2394 if a variable may be clobbered by a @code{setjmp} call.
2396 These warnings are possible only in optimizing compilation,
2397 because they require data flow information that is computed only
2398 when optimizing. If you don't specify @option{-O}, you simply won't
2401 If you want to warn about code which uses the uninitialized value of the
2402 variable in its own initializer, use the @option{-Winit-self} option.
2404 These warnings occur only for variables that are candidates for
2405 register allocation. Therefore, they do not occur for a variable that
2406 is declared @code{volatile}, or whose address is taken, or whose size
2407 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2408 structures, unions or arrays, even when they are in registers.
2410 Note that there may be no warning about a variable that is used only
2411 to compute a value that itself is never used, because such
2412 computations may be deleted by data flow analysis before the warnings
2415 These warnings are made optional because GCC is not smart
2416 enough to see all the reasons why the code might be correct
2417 despite appearing to have an error. Here is one example of how
2438 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2439 always initialized, but GCC doesn't know this. Here is
2440 another common case:
2445 if (change_y) save_y = y, y = new_y;
2447 if (change_y) y = save_y;
2452 This has no bug because @code{save_y} is used only if it is set.
2454 @cindex @code{longjmp} warnings
2455 This option also warns when a non-volatile automatic variable might be
2456 changed by a call to @code{longjmp}. These warnings as well are possible
2457 only in optimizing compilation.
2459 The compiler sees only the calls to @code{setjmp}. It cannot know
2460 where @code{longjmp} will be called; in fact, a signal handler could
2461 call it at any point in the code. As a result, you may get a warning
2462 even when there is in fact no problem because @code{longjmp} cannot
2463 in fact be called at the place which would cause a problem.
2465 Some spurious warnings can be avoided if you declare all the functions
2466 you use that never return as @code{noreturn}. @xref{Function
2469 @item -Wunknown-pragmas
2470 @opindex Wunknown-pragmas
2471 @cindex warning for unknown pragmas
2472 @cindex unknown pragmas, warning
2473 @cindex pragmas, warning of unknown
2474 Warn when a #pragma directive is encountered which is not understood by
2475 GCC@. If this command line option is used, warnings will even be issued
2476 for unknown pragmas in system header files. This is not the case if
2477 the warnings were only enabled by the @option{-Wall} command line option.
2479 @item -Wstrict-aliasing
2480 @opindex Wstrict-aliasing
2481 This option is only active when @option{-fstrict-aliasing} is active.
2482 It warns about code which might break the strict aliasing rules that the
2483 compiler is using for optimization. The warning does not catch all
2484 cases, but does attempt to catch the more common pitfalls. It is
2485 included in @option{-Wall}.
2489 All of the above @samp{-W} options combined. This enables all the
2490 warnings about constructions that some users consider questionable, and
2491 that are easy to avoid (or modify to prevent the warning), even in
2492 conjunction with macros. This also enables some language-specific
2493 warnings described in @ref{C++ Dialect Options} and
2494 @ref{Objective-C Dialect Options}.
2497 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2498 Some of them warn about constructions that users generally do not
2499 consider questionable, but which occasionally you might wish to check
2500 for; others warn about constructions that are necessary or hard to avoid
2501 in some cases, and there is no simple way to modify the code to suppress
2508 (This option used to be called @option{-W}. The older name is still
2509 supported, but the newer name is more descriptive.) Print extra warning
2510 messages for these events:
2514 A function can return either with or without a value. (Falling
2515 off the end of the function body is considered returning without
2516 a value.) For example, this function would evoke such a
2530 An expression-statement or the left-hand side of a comma expression
2531 contains no side effects.
2532 To suppress the warning, cast the unused expression to void.
2533 For example, an expression such as @samp{x[i,j]} will cause a warning,
2534 but @samp{x[(void)i,j]} will not.
2537 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2540 A comparison like @samp{x<=y<=z} appears; this is equivalent to
2541 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
2542 that of ordinary mathematical notation.
2545 Storage-class specifiers like @code{static} are not the first things in
2546 a declaration. According to the C Standard, this usage is obsolescent.
2549 The return type of a function has a type qualifier such as @code{const}.
2550 Such a type qualifier has no effect, since the value returned by a
2551 function is not an lvalue. (But don't warn about the GNU extension of
2552 @code{volatile void} return types. That extension will be warned about
2553 if @option{-pedantic} is specified.)
2556 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2560 A comparison between signed and unsigned values could produce an
2561 incorrect result when the signed value is converted to unsigned.
2562 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2565 An aggregate has an initializer which does not initialize all members.
2566 For example, the following code would cause such a warning, because
2567 @code{x.h} would be implicitly initialized to zero:
2570 struct s @{ int f, g, h; @};
2571 struct s x = @{ 3, 4 @};
2575 A function parameter is declared without a type specifier in K&R-style
2583 An empty body occurs in an @samp{if} or @samp{else} statement.
2586 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2587 @samp{>}, or @samp{>=}.
2590 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2593 Any of several floating-point events that often indicate errors, such as
2594 overflow, underflow, loss of precision, etc.
2596 @item @r{(C++ only)}
2597 An enumerator and a non-enumerator both appear in a conditional expression.
2599 @item @r{(C++ only)}
2600 A non-static reference or non-static @samp{const} member appears in a
2601 class without constructors.
2603 @item @r{(C++ only)}
2604 Ambiguous virtual bases.
2606 @item @r{(C++ only)}
2607 Subscripting an array which has been declared @samp{register}.
2609 @item @r{(C++ only)}
2610 Taking the address of a variable which has been declared @samp{register}.
2612 @item @r{(C++ only)}
2613 A base class is not initialized in a derived class' copy constructor.
2616 @item -Wno-div-by-zero
2617 @opindex Wno-div-by-zero
2618 @opindex Wdiv-by-zero
2619 Do not warn about compile-time integer division by zero. Floating point
2620 division by zero is not warned about, as it can be a legitimate way of
2621 obtaining infinities and NaNs.
2623 @item -Wsystem-headers
2624 @opindex Wsystem-headers
2625 @cindex warnings from system headers
2626 @cindex system headers, warnings from
2627 Print warning messages for constructs found in system header files.
2628 Warnings from system headers are normally suppressed, on the assumption
2629 that they usually do not indicate real problems and would only make the
2630 compiler output harder to read. Using this command line option tells
2631 GCC to emit warnings from system headers as if they occurred in user
2632 code. However, note that using @option{-Wall} in conjunction with this
2633 option will @emph{not} warn about unknown pragmas in system
2634 headers---for that, @option{-Wunknown-pragmas} must also be used.
2637 @opindex Wfloat-equal
2638 Warn if floating point values are used in equality comparisons.
2640 The idea behind this is that sometimes it is convenient (for the
2641 programmer) to consider floating-point values as approximations to
2642 infinitely precise real numbers. If you are doing this, then you need
2643 to compute (by analyzing the code, or in some other way) the maximum or
2644 likely maximum error that the computation introduces, and allow for it
2645 when performing comparisons (and when producing output, but that's a
2646 different problem). In particular, instead of testing for equality, you
2647 would check to see whether the two values have ranges that overlap; and
2648 this is done with the relational operators, so equality comparisons are
2651 @item -Wtraditional @r{(C only)}
2652 @opindex Wtraditional
2653 Warn about certain constructs that behave differently in traditional and
2654 ISO C@. Also warn about ISO C constructs that have no traditional C
2655 equivalent, and/or problematic constructs which should be avoided.
2659 Macro parameters that appear within string literals in the macro body.
2660 In traditional C macro replacement takes place within string literals,
2661 but does not in ISO C@.
2664 In traditional C, some preprocessor directives did not exist.
2665 Traditional preprocessors would only consider a line to be a directive
2666 if the @samp{#} appeared in column 1 on the line. Therefore
2667 @option{-Wtraditional} warns about directives that traditional C
2668 understands but would ignore because the @samp{#} does not appear as the
2669 first character on the line. It also suggests you hide directives like
2670 @samp{#pragma} not understood by traditional C by indenting them. Some
2671 traditional implementations would not recognize @samp{#elif}, so it
2672 suggests avoiding it altogether.
2675 A function-like macro that appears without arguments.
2678 The unary plus operator.
2681 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2682 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2683 constants.) Note, these suffixes appear in macros defined in the system
2684 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2685 Use of these macros in user code might normally lead to spurious
2686 warnings, however gcc's integrated preprocessor has enough context to
2687 avoid warning in these cases.
2690 A function declared external in one block and then used after the end of
2694 A @code{switch} statement has an operand of type @code{long}.
2697 A non-@code{static} function declaration follows a @code{static} one.
2698 This construct is not accepted by some traditional C compilers.
2701 The ISO type of an integer constant has a different width or
2702 signedness from its traditional type. This warning is only issued if
2703 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2704 typically represent bit patterns, are not warned about.
2707 Usage of ISO string concatenation is detected.
2710 Initialization of automatic aggregates.
2713 Identifier conflicts with labels. Traditional C lacks a separate
2714 namespace for labels.
2717 Initialization of unions. If the initializer is zero, the warning is
2718 omitted. This is done under the assumption that the zero initializer in
2719 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2720 initializer warnings and relies on default initialization to zero in the
2724 Conversions by prototypes between fixed/floating point values and vice
2725 versa. The absence of these prototypes when compiling with traditional
2726 C would cause serious problems. This is a subset of the possible
2727 conversion warnings, for the full set use @option{-Wconversion}.
2730 Use of ISO C style function definitions. This warning intentionally is
2731 @emph{not} issued for prototype declarations or variadic functions
2732 because these ISO C features will appear in your code when using
2733 libiberty's traditional C compatibility macros, @code{PARAMS} and
2734 @code{VPARAMS}. This warning is also bypassed for nested functions
2735 because that feature is already a gcc extension and thus not relevant to
2736 traditional C compatibility.
2739 @item -Wdeclaration-after-statement @r{(C only)}
2740 @opindex Wdeclaration-after-statement
2741 Warn when a declaration is found after a statement in a block. This
2742 construct, known from C++, was introduced with ISO C99 and is by default
2743 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2744 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2748 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2750 @item -Wendif-labels
2751 @opindex Wendif-labels
2752 Warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2756 Warn whenever a local variable shadows another local variable, parameter or
2757 global variable or whenever a built-in function is shadowed.
2759 @item -Wlarger-than-@var{len}
2760 @opindex Wlarger-than
2761 Warn whenever an object of larger than @var{len} bytes is defined.
2763 @item -Wpointer-arith
2764 @opindex Wpointer-arith
2765 Warn about anything that depends on the ``size of'' a function type or
2766 of @code{void}. GNU C assigns these types a size of 1, for
2767 convenience in calculations with @code{void *} pointers and pointers
2770 @item -Wbad-function-cast @r{(C only)}
2771 @opindex Wbad-function-cast
2772 Warn whenever a function call is cast to a non-matching type.
2773 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2777 Warn whenever a pointer is cast so as to remove a type qualifier from
2778 the target type. For example, warn if a @code{const char *} is cast
2779 to an ordinary @code{char *}.
2782 @opindex Wcast-align
2783 Warn whenever a pointer is cast such that the required alignment of the
2784 target is increased. For example, warn if a @code{char *} is cast to
2785 an @code{int *} on machines where integers can only be accessed at
2786 two- or four-byte boundaries.
2788 @item -Wwrite-strings
2789 @opindex Wwrite-strings
2790 When compiling C, give string constants the type @code{const
2791 char[@var{length}]} so that
2792 copying the address of one into a non-@code{const} @code{char *}
2793 pointer will get a warning; when compiling C++, warn about the
2794 deprecated conversion from string constants to @code{char *}.
2795 These warnings will help you find at
2796 compile time code that can try to write into a string constant, but
2797 only if you have been very careful about using @code{const} in
2798 declarations and prototypes. Otherwise, it will just be a nuisance;
2799 this is why we did not make @option{-Wall} request these warnings.
2802 @opindex Wconversion
2803 Warn if a prototype causes a type conversion that is different from what
2804 would happen to the same argument in the absence of a prototype. This
2805 includes conversions of fixed point to floating and vice versa, and
2806 conversions changing the width or signedness of a fixed point argument
2807 except when the same as the default promotion.
2809 Also, warn if a negative integer constant expression is implicitly
2810 converted to an unsigned type. For example, warn about the assignment
2811 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2812 casts like @code{(unsigned) -1}.
2814 @item -Wsign-compare
2815 @opindex Wsign-compare
2816 @cindex warning for comparison of signed and unsigned values
2817 @cindex comparison of signed and unsigned values, warning
2818 @cindex signed and unsigned values, comparison warning
2819 Warn when a comparison between signed and unsigned values could produce
2820 an incorrect result when the signed value is converted to unsigned.
2821 This warning is also enabled by @option{-Wextra}; to get the other warnings
2822 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2824 @item -Waggregate-return
2825 @opindex Waggregate-return
2826 Warn if any functions that return structures or unions are defined or
2827 called. (In languages where you can return an array, this also elicits
2830 @item -Wstrict-prototypes @r{(C only)}
2831 @opindex Wstrict-prototypes
2832 Warn if a function is declared or defined without specifying the
2833 argument types. (An old-style function definition is permitted without
2834 a warning if preceded by a declaration which specifies the argument
2837 @item -Wold-style-definition @r{(C only)}
2838 @opindex Wold-style-definition
2839 Warn if an old-style function definition is used. A warning is given
2840 even if there is a previous prototype.
2842 @item -Wmissing-prototypes @r{(C only)}
2843 @opindex Wmissing-prototypes
2844 Warn if a global function is defined without a previous prototype
2845 declaration. This warning is issued even if the definition itself
2846 provides a prototype. The aim is to detect global functions that fail
2847 to be declared in header files.
2849 @item -Wmissing-declarations @r{(C only)}
2850 @opindex Wmissing-declarations
2851 Warn if a global function is defined without a previous declaration.
2852 Do so even if the definition itself provides a prototype.
2853 Use this option to detect global functions that are not declared in
2856 @item -Wmissing-noreturn
2857 @opindex Wmissing-noreturn
2858 Warn about functions which might be candidates for attribute @code{noreturn}.
2859 Note these are only possible candidates, not absolute ones. Care should
2860 be taken to manually verify functions actually do not ever return before
2861 adding the @code{noreturn} attribute, otherwise subtle code generation
2862 bugs could be introduced. You will not get a warning for @code{main} in
2863 hosted C environments.
2865 @item -Wmissing-format-attribute
2866 @opindex Wmissing-format-attribute
2868 If @option{-Wformat} is enabled, also warn about functions which might be
2869 candidates for @code{format} attributes. Note these are only possible
2870 candidates, not absolute ones. GCC will guess that @code{format}
2871 attributes might be appropriate for any function that calls a function
2872 like @code{vprintf} or @code{vscanf}, but this might not always be the
2873 case, and some functions for which @code{format} attributes are
2874 appropriate may not be detected. This option has no effect unless
2875 @option{-Wformat} is enabled (possibly by @option{-Wall}).
2877 @item -Wno-multichar
2878 @opindex Wno-multichar
2880 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
2881 Usually they indicate a typo in the user's code, as they have
2882 implementation-defined values, and should not be used in portable code.
2884 @item -Wno-deprecated-declarations
2885 @opindex Wno-deprecated-declarations
2886 Do not warn about uses of functions, variables, and types marked as
2887 deprecated by using the @code{deprecated} attribute.
2888 (@pxref{Function Attributes}, @pxref{Variable Attributes},
2889 @pxref{Type Attributes}.)
2893 Warn if a structure is given the packed attribute, but the packed
2894 attribute has no effect on the layout or size of the structure.
2895 Such structures may be mis-aligned for little benefit. For
2896 instance, in this code, the variable @code{f.x} in @code{struct bar}
2897 will be misaligned even though @code{struct bar} does not itself
2898 have the packed attribute:
2905 @} __attribute__((packed));
2915 Warn if padding is included in a structure, either to align an element
2916 of the structure or to align the whole structure. Sometimes when this
2917 happens it is possible to rearrange the fields of the structure to
2918 reduce the padding and so make the structure smaller.
2920 @item -Wredundant-decls
2921 @opindex Wredundant-decls
2922 Warn if anything is declared more than once in the same scope, even in
2923 cases where multiple declaration is valid and changes nothing.
2925 @item -Wnested-externs @r{(C only)}
2926 @opindex Wnested-externs
2927 Warn if an @code{extern} declaration is encountered within a function.
2929 @item -Wunreachable-code
2930 @opindex Wunreachable-code
2931 Warn if the compiler detects that code will never be executed.
2933 This option is intended to warn when the compiler detects that at
2934 least a whole line of source code will never be executed, because
2935 some condition is never satisfied or because it is after a
2936 procedure that never returns.
2938 It is possible for this option to produce a warning even though there
2939 are circumstances under which part of the affected line can be executed,
2940 so care should be taken when removing apparently-unreachable code.
2942 For instance, when a function is inlined, a warning may mean that the
2943 line is unreachable in only one inlined copy of the function.
2945 This option is not made part of @option{-Wall} because in a debugging
2946 version of a program there is often substantial code which checks
2947 correct functioning of the program and is, hopefully, unreachable
2948 because the program does work. Another common use of unreachable
2949 code is to provide behavior which is selectable at compile-time.
2953 Warn if a function can not be inlined and it was declared as inline.
2954 Even with this option, the compiler will not warn about failures to
2955 inline functions declared in system headers.
2957 The compiler uses a variety of heuristics to determine whether or not
2958 to inline a function. For example, the compiler takes into account
2959 the size of the function being inlined and the the amount of inlining
2960 that has already been done in the current function. Therefore,
2961 seemingly insignificant changes in the source program can cause the
2962 warnings produced by @option{-Winline} to appear or disappear.
2964 @item -Wno-invalid-offsetof @r{(C++ only)}
2965 @opindex Wno-invalid-offsetof
2966 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
2967 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
2968 to a non-POD type is undefined. In existing C++ implementations,
2969 however, @samp{offsetof} typically gives meaningful results even when
2970 applied to certain kinds of non-POD types. (Such as a simple
2971 @samp{struct} that fails to be a POD type only by virtue of having a
2972 constructor.) This flag is for users who are aware that they are
2973 writing nonportable code and who have deliberately chosen to ignore the
2976 The restrictions on @samp{offsetof} may be relaxed in a future version
2977 of the C++ standard.
2980 @opindex Winvalid-pch
2981 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
2982 the search path but can't be used.
2986 @opindex Wno-long-long
2987 Warn if @samp{long long} type is used. This is default. To inhibit
2988 the warning messages, use @option{-Wno-long-long}. Flags
2989 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
2990 only when @option{-pedantic} flag is used.
2992 @item -Wdisabled-optimization
2993 @opindex Wdisabled-optimization
2994 Warn if a requested optimization pass is disabled. This warning does
2995 not generally indicate that there is anything wrong with your code; it
2996 merely indicates that GCC's optimizers were unable to handle the code
2997 effectively. Often, the problem is that your code is too big or too
2998 complex; GCC will refuse to optimize programs when the optimization
2999 itself is likely to take inordinate amounts of time.
3003 Make all warnings into errors.
3006 @node Debugging Options
3007 @section Options for Debugging Your Program or GCC
3008 @cindex options, debugging
3009 @cindex debugging information options
3011 GCC has various special options that are used for debugging
3012 either your program or GCC:
3017 Produce debugging information in the operating system's native format
3018 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
3021 On most systems that use stabs format, @option{-g} enables use of extra
3022 debugging information that only GDB can use; this extra information
3023 makes debugging work better in GDB but will probably make other debuggers
3025 refuse to read the program. If you want to control for certain whether
3026 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3027 @option{-gxcoff+}, @option{-gxcoff}, @option{-gdwarf-1+}, @option{-gdwarf-1},
3028 or @option{-gvms} (see below).
3030 Unlike most other C compilers, GCC allows you to use @option{-g} with
3031 @option{-O}. The shortcuts taken by optimized code may occasionally
3032 produce surprising results: some variables you declared may not exist
3033 at all; flow of control may briefly move where you did not expect it;
3034 some statements may not be executed because they compute constant
3035 results or their values were already at hand; some statements may
3036 execute in different places because they were moved out of loops.
3038 Nevertheless it proves possible to debug optimized output. This makes
3039 it reasonable to use the optimizer for programs that might have bugs.
3041 The following options are useful when GCC is generated with the
3042 capability for more than one debugging format.
3046 Produce debugging information for use by GDB@. This means to use the
3047 most expressive format available (DWARF 2, stabs, or the native format
3048 if neither of those are supported), including GDB extensions if at all
3053 Produce debugging information in stabs format (if that is supported),
3054 without GDB extensions. This is the format used by DBX on most BSD
3055 systems. On MIPS, Alpha and System V Release 4 systems this option
3056 produces stabs debugging output which is not understood by DBX or SDB@.
3057 On System V Release 4 systems this option requires the GNU assembler.
3059 @item -feliminate-unused-debug-symbols
3060 @opindex feliminate-unused-debug-symbols
3061 Produce debugging information in stabs format (if that is supported),
3062 for only symbols that are actually used.
3066 Produce debugging information in stabs format (if that is supported),
3067 using GNU extensions understood only by the GNU debugger (GDB)@. The
3068 use of these extensions is likely to make other debuggers crash or
3069 refuse to read the program.
3073 Produce debugging information in COFF format (if that is supported).
3074 This is the format used by SDB on most System V systems prior to
3079 Produce debugging information in XCOFF format (if that is supported).
3080 This is the format used by the DBX debugger on IBM RS/6000 systems.
3084 Produce debugging information in XCOFF format (if that is supported),
3085 using GNU extensions understood only by the GNU debugger (GDB)@. The
3086 use of these extensions is likely to make other debuggers crash or
3087 refuse to read the program, and may cause assemblers other than the GNU
3088 assembler (GAS) to fail with an error.
3092 Produce debugging information in DWARF version 1 format (if that is
3093 supported). This is the format used by SDB on most System V Release 4
3096 This option is deprecated.
3100 Produce debugging information in DWARF version 1 format (if that is
3101 supported), using GNU extensions understood only by the GNU debugger
3102 (GDB)@. The use of these extensions is likely to make other debuggers
3103 crash or refuse to read the program.
3105 This option is deprecated.
3109 Produce debugging information in DWARF version 2 format (if that is
3110 supported). This is the format used by DBX on IRIX 6.
3114 Produce debugging information in VMS debug format (if that is
3115 supported). This is the format used by DEBUG on VMS systems.
3118 @itemx -ggdb@var{level}
3119 @itemx -gstabs@var{level}
3120 @itemx -gcoff@var{level}
3121 @itemx -gxcoff@var{level}
3122 @itemx -gvms@var{level}
3123 Request debugging information and also use @var{level} to specify how
3124 much information. The default level is 2.
3126 Level 1 produces minimal information, enough for making backtraces in
3127 parts of the program that you don't plan to debug. This includes
3128 descriptions of functions and external variables, but no information
3129 about local variables and no line numbers.
3131 Level 3 includes extra information, such as all the macro definitions
3132 present in the program. Some debuggers support macro expansion when
3133 you use @option{-g3}.
3135 Note that in order to avoid confusion between DWARF1 debug level 2,
3136 and DWARF2, neither @option{-gdwarf} nor @option{-gdwarf-2} accept
3137 a concatenated debug level. Instead use an additional @option{-g@var{level}}
3138 option to change the debug level for DWARF1 or DWARF2.
3140 @item -feliminate-dwarf2-dups
3141 @opindex feliminate-dwarf2-dups
3142 Compress DWARF2 debugging information by eliminating duplicated
3143 information about each symbol. This option only makes sense when
3144 generating DWARF2 debugging information with @option{-gdwarf-2}.
3146 @cindex @command{prof}
3149 Generate extra code to write profile information suitable for the
3150 analysis program @command{prof}. You must use this option when compiling
3151 the source files you want data about, and you must also use it when
3154 @cindex @command{gprof}
3157 Generate extra code to write profile information suitable for the
3158 analysis program @command{gprof}. You must use this option when compiling
3159 the source files you want data about, and you must also use it when
3164 Makes the compiler print out each function name as it is compiled, and
3165 print some statistics about each pass when it finishes.
3168 @opindex ftime-report
3169 Makes the compiler print some statistics about the time consumed by each
3170 pass when it finishes.
3173 @opindex fmem-report
3174 Makes the compiler print some statistics about permanent memory
3175 allocation when it finishes.
3177 @item -fprofile-arcs
3178 @opindex fprofile-arcs
3179 Add code so that program flow @dfn{arcs} are instrumented. During
3180 execution the program records how many times each branch and call is
3181 executed and how many times it is taken or returns. When the compiled
3182 program exits it saves this data to a file called
3183 @file{@var{auxname}.gcda} for each source file. The data may be used for
3184 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3185 test coverage analysis (@option{-ftest-coverage}). Each object file's
3186 @var{auxname} is generated from the name of the output file, if
3187 explicitly specified and it is not the final executable, otherwise it is
3188 the basename of the source file. In both cases any suffix is removed
3189 (e.g. @file{foo.gcda} for input file @file{dir/foo.c}, or
3190 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3195 Compile the source files with @option{-fprofile-arcs} plus optimization
3196 and code generation options. For test coverage analysis, use the
3197 additional @option{-ftest-coverage} option. You do not need to profile
3198 every source file in a program.
3201 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3202 (the latter implies the former).
3205 Run the program on a representative workload to generate the arc profile
3206 information. This may be repeated any number of times. You can run
3207 concurrent instances of your program, and provided that the file system
3208 supports locking, the data files will be correctly updated. Also
3209 @code{fork} calls are detected and correctly handled (double counting
3213 For profile-directed optimizations, compile the source files again with
3214 the same optimization and code generation options plus
3215 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3216 Control Optimization}).
3219 For test coverage analysis, use @command{gcov} to produce human readable
3220 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3221 @command{gcov} documentation for further information.
3225 With @option{-fprofile-arcs}, for each function of your program GCC
3226 creates a program flow graph, then finds a spanning tree for the graph.
3227 Only arcs that are not on the spanning tree have to be instrumented: the
3228 compiler adds code to count the number of times that these arcs are
3229 executed. When an arc is the only exit or only entrance to a block, the
3230 instrumentation code can be added to the block; otherwise, a new basic
3231 block must be created to hold the instrumentation code.
3234 @item -ftest-coverage
3235 @opindex ftest-coverage
3236 Produce a notes file that the @command{gcov} code-coverage utility
3237 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3238 show program coverage. Each source file's note file is called
3239 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3240 above for a description of @var{auxname} and instructions on how to
3241 generate test coverage data. Coverage data will match the source files
3242 more closely, if you do not optimize.
3244 @item -d@var{letters}
3246 Says to make debugging dumps during compilation at times specified by
3247 @var{letters}. This is used for debugging the compiler. The file names
3248 for most of the dumps are made by appending a pass number and a word to
3249 the @var{dumpname}. @var{dumpname} is generated from the name of the
3250 output file, if explicitly specified and it is not an executable,
3251 otherwise it is the basename of the source file. In both cases any
3252 suffix is removed (e.g. @file{foo.01.rtl} or @file{foo.02.sibling}).
3253 Here are the possible letters for use in @var{letters}, and their
3259 Annotate the assembler output with miscellaneous debugging information.
3262 Dump after computing branch probabilities, to @file{@var{file}.12.bp}.
3265 Dump after block reordering, to @file{@var{file}.30.bbro}.
3268 Dump after instruction combination, to the file @file{@var{file}.20.combine}.
3271 Dump after the first if conversion, to the file @file{@var{file}.14.ce1}.
3272 Also dump after the second if conversion, to the file @file{@var{file}.21.ce2}.
3275 Dump after branch target load optimization, to to @file{@var{file}.32.btl}.
3276 Also dump after delayed branch scheduling, to @file{@var{file}.36.dbr}.
3279 Dump all macro definitions, at the end of preprocessing, in addition to
3283 Dump after the second if conversion, to @file{@var{file}.31.ce3}.
3286 Dump after control and data flow analysis, to @file{@var{file}.11.cfg}.
3287 Also dump after life analysis, to @file{@var{file}.19.life}.
3290 Dump after purging @code{ADDRESSOF} codes, to @file{@var{file}.07.addressof}.
3293 Dump after global register allocation, to @file{@var{file}.25.greg}.
3296 Dump after GCSE, to @file{@var{file}.08.gcse}.
3297 Also dump after jump bypassing and control flow optimizations, to
3298 @file{@var{file}.10.bypass}.
3301 Dump after finalization of EH handling code, to @file{@var{file}.03.eh}.
3304 Dump after sibling call optimizations, to @file{@var{file}.02.sibling}.
3307 Dump after the first jump optimization, to @file{@var{file}.04.jump}.
3310 Dump after conversion from registers to stack, to @file{@var{file}.34.stack}.
3313 Dump after local register allocation, to @file{@var{file}.24.lreg}.
3316 Dump after loop optimization passes, to @file{@var{file}.09.loop} and
3317 @file{@var{file}.16.loop2}.
3320 Dump after performing the machine dependent reorganization pass, to
3321 @file{@var{file}.35.mach}.
3324 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3327 Dump after the register move pass, to @file{@var{file}.22.regmove}.
3330 Dump after post-reload optimizations, to @file{@var{file}.26.postreload}.
3333 Dump after RTL generation, to @file{@var{file}.01.rtl}.
3336 Dump after the second scheduling pass, to @file{@var{file}.33.sched2}.
3339 Dump after CSE (including the jump optimization that sometimes follows
3340 CSE), to @file{@var{file}.06.cse}.
3343 Dump after the first scheduling pass, to @file{@var{file}.23.sched}.
3346 Dump after the second CSE pass (including the jump optimization that
3347 sometimes follows CSE), to @file{@var{file}.18.cse2}.
3350 Dump after running tracer, to @file{@var{file}.15.tracer}.
3353 Dump after null pointer elimination pass to @file{@var{file}.05.null}.
3356 Dump callgraph and unit-at-a-time optimization @file{@var{file}.00.unit}.
3359 Dump after the value profile transformations, to @file{@var{file}.13.vpt}.
3362 Dump after the second flow pass, to @file{@var{file}.27.flow2}.
3365 Dump after the peephole pass, to @file{@var{file}.28.peephole2}.
3368 Dump after constructing the web, to @file{@var{file}.17.web}.
3371 Produce all the dumps listed above.
3374 Produce a core dump whenever an error occurs.
3377 Print statistics on memory usage, at the end of the run, to
3381 Annotate the assembler output with a comment indicating which
3382 pattern and alternative was used. The length of each instruction is
3386 Dump the RTL in the assembler output as a comment before each instruction.
3387 Also turns on @option{-dp} annotation.
3390 For each of the other indicated dump files (except for
3391 @file{@var{file}.01.rtl}), dump a representation of the control flow graph
3392 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3395 Just generate RTL for a function instead of compiling it. Usually used
3399 Dump debugging information during parsing, to standard error.
3402 @item -fdump-unnumbered
3403 @opindex fdump-unnumbered
3404 When doing debugging dumps (see @option{-d} option above), suppress instruction
3405 numbers and line number note output. This makes it more feasible to
3406 use diff on debugging dumps for compiler invocations with different
3407 options, in particular with and without @option{-g}.
3409 @item -fdump-translation-unit @r{(C and C++ only)}
3410 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3411 @opindex fdump-translation-unit
3412 Dump a representation of the tree structure for the entire translation
3413 unit to a file. The file name is made by appending @file{.tu} to the
3414 source file name. If the @samp{-@var{options}} form is used, @var{options}
3415 controls the details of the dump as described for the
3416 @option{-fdump-tree} options.
3418 @item -fdump-class-hierarchy @r{(C++ only)}
3419 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3420 @opindex fdump-class-hierarchy
3421 Dump a representation of each class's hierarchy and virtual function
3422 table layout to a file. The file name is made by appending @file{.class}
3423 to the source file name. If the @samp{-@var{options}} form is used,
3424 @var{options} controls the details of the dump as described for the
3425 @option{-fdump-tree} options.
3427 @item -fdump-tree-@var{switch} @r{(C++ only)}
3428 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C++ only)}
3430 Control the dumping at various stages of processing the intermediate
3431 language tree to a file. The file name is generated by appending a switch
3432 specific suffix to the source file name. If the @samp{-@var{options}}
3433 form is used, @var{options} is a list of @samp{-} separated options that
3434 control the details of the dump. Not all options are applicable to all
3435 dumps, those which are not meaningful will be ignored. The following
3436 options are available
3440 Print the address of each node. Usually this is not meaningful as it
3441 changes according to the environment and source file. Its primary use
3442 is for tying up a dump file with a debug environment.
3444 Inhibit dumping of members of a scope or body of a function merely
3445 because that scope has been reached. Only dump such items when they
3446 are directly reachable by some other path.
3448 Turn on all options.
3451 The following tree dumps are possible:
3454 Dump before any tree based optimization, to @file{@var{file}.original}.
3456 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3458 Dump after function inlining, to @file{@var{file}.inlined}.
3461 @item -frandom-seed=@var{string}
3462 @opindex frandom-string
3463 This option provides a seed that GCC uses when it would otherwise use
3464 random numbers. It is used to generate certain symbol names
3465 that have to be different in every compiled file. It is also used to
3466 place unique stamps in coverage data files and the object files that
3467 produce them. You can use the @option{-frandom-seed} option to produce
3468 reproducibly identical object files.
3470 The @var{string} should be different for every file you compile.
3472 @item -fsched-verbose=@var{n}
3473 @opindex fsched-verbose
3474 On targets that use instruction scheduling, this option controls the
3475 amount of debugging output the scheduler prints. This information is
3476 written to standard error, unless @option{-dS} or @option{-dR} is
3477 specified, in which case it is output to the usual dump
3478 listing file, @file{.sched} or @file{.sched2} respectively. However
3479 for @var{n} greater than nine, the output is always printed to standard
3482 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3483 same information as @option{-dRS}. For @var{n} greater than one, it
3484 also output basic block probabilities, detailed ready list information
3485 and unit/insn info. For @var{n} greater than two, it includes RTL
3486 at abort point, control-flow and regions info. And for @var{n} over
3487 four, @option{-fsched-verbose} also includes dependence info.
3491 Store the usual ``temporary'' intermediate files permanently; place them
3492 in the current directory and name them based on the source file. Thus,
3493 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3494 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3495 preprocessed @file{foo.i} output file even though the compiler now
3496 normally uses an integrated preprocessor.
3500 Report the CPU time taken by each subprocess in the compilation
3501 sequence. For C source files, this is the compiler proper and assembler
3502 (plus the linker if linking is done). The output looks like this:
3509 The first number on each line is the ``user time,'' that is time spent
3510 executing the program itself. The second number is ``system time,''
3511 time spent executing operating system routines on behalf of the program.
3512 Both numbers are in seconds.
3514 @item -print-file-name=@var{library}
3515 @opindex print-file-name
3516 Print the full absolute name of the library file @var{library} that
3517 would be used when linking---and don't do anything else. With this
3518 option, GCC does not compile or link anything; it just prints the
3521 @item -print-multi-directory
3522 @opindex print-multi-directory
3523 Print the directory name corresponding to the multilib selected by any
3524 other switches present in the command line. This directory is supposed
3525 to exist in @env{GCC_EXEC_PREFIX}.
3527 @item -print-multi-lib
3528 @opindex print-multi-lib
3529 Print the mapping from multilib directory names to compiler switches
3530 that enable them. The directory name is separated from the switches by
3531 @samp{;}, and each switch starts with an @samp{@@} instead of the
3532 @samp{-}, without spaces between multiple switches. This is supposed to
3533 ease shell-processing.
3535 @item -print-prog-name=@var{program}
3536 @opindex print-prog-name
3537 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3539 @item -print-libgcc-file-name
3540 @opindex print-libgcc-file-name
3541 Same as @option{-print-file-name=libgcc.a}.
3543 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3544 but you do want to link with @file{libgcc.a}. You can do
3547 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3550 @item -print-search-dirs
3551 @opindex print-search-dirs
3552 Print the name of the configured installation directory and a list of
3553 program and library directories gcc will search---and don't do anything else.
3555 This is useful when gcc prints the error message
3556 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3557 To resolve this you either need to put @file{cpp0} and the other compiler
3558 components where gcc expects to find them, or you can set the environment
3559 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3560 Don't forget the trailing '/'.
3561 @xref{Environment Variables}.
3564 @opindex dumpmachine
3565 Print the compiler's target machine (for example,
3566 @samp{i686-pc-linux-gnu})---and don't do anything else.
3569 @opindex dumpversion
3570 Print the compiler version (for example, @samp{3.0})---and don't do
3575 Print the compiler's built-in specs---and don't do anything else. (This
3576 is used when GCC itself is being built.) @xref{Spec Files}.
3578 @item -feliminate-unused-debug-types
3579 @opindex feliminate-unused-debug-types
3580 Normally, when producing DWARF2 output, GCC will emit debugging
3581 information for all types declared in a compilation
3582 unit, regardless of whether or not they are actually used
3583 in that compilation unit. Sometimes this is useful, such as
3584 if, in the debugger, you want to cast a value to a type that is
3585 not actually used in your program (but is declared). More often,
3586 however, this results in a significant amount of wasted space.
3587 With this option, GCC will avoid producing debug symbol output
3588 for types that are nowhere used in the source file being compiled.
3591 @node Optimize Options
3592 @section Options That Control Optimization
3593 @cindex optimize options
3594 @cindex options, optimization
3596 These options control various sorts of optimizations.
3598 Without any optimization option, the compiler's goal is to reduce the
3599 cost of compilation and to make debugging produce the expected
3600 results. Statements are independent: if you stop the program with a
3601 breakpoint between statements, you can then assign a new value to any
3602 variable or change the program counter to any other statement in the
3603 function and get exactly the results you would expect from the source
3606 Turning on optimization flags makes the compiler attempt to improve
3607 the performance and/or code size at the expense of compilation time
3608 and possibly the ability to debug the program.
3610 The compiler performs optimisation based on the knowledge it has of
3611 the program. Using the @option{-funit-at-a-time} flag will allow the
3612 compiler to consider information gained from later functions in the
3613 file when compiling a function. Compiling multiple files at once to a
3614 single output file (and using @option{-funit-at-a-time}) will allow
3615 the compiler to use information gained from all of the files when
3616 compiling each of them.
3618 Not all optimizations are controlled directly by a flag. Only
3619 optimizations that have a flag are listed.
3626 Optimize. Optimizing compilation takes somewhat more time, and a lot
3627 more memory for a large function.
3629 With @option{-O}, the compiler tries to reduce code size and execution
3630 time, without performing any optimizations that take a great deal of
3633 @option{-O} turns on the following optimization flags:
3634 @gccoptlist{-fdefer-pop @gol
3635 -fmerge-constants @gol
3637 -floop-optimize @gol
3639 -fif-conversion @gol
3640 -fif-conversion2 @gol
3641 -fdelayed-branch @gol
3642 -fguess-branch-probability @gol
3645 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
3646 where doing so does not interfere with debugging.
3650 Optimize even more. GCC performs nearly all supported optimizations
3651 that do not involve a space-speed tradeoff. The compiler does not
3652 perform loop unrolling or function inlining when you specify @option{-O2}.
3653 As compared to @option{-O}, this option increases both compilation time
3654 and the performance of the generated code.
3656 @option{-O2} turns on all optimization flags specified by @option{-O}. It
3657 also turns on the following optimization flags:
3658 @gccoptlist{-fforce-mem @gol
3659 -foptimize-sibling-calls @gol
3660 -fstrength-reduce @gol
3661 -fcse-follow-jumps -fcse-skip-blocks @gol
3662 -frerun-cse-after-loop -frerun-loop-opt @gol
3663 -fgcse -fgcse-lm -fgcse-sm -fgcse-las @gol
3664 -fdelete-null-pointer-checks @gol
3665 -fexpensive-optimizations @gol
3667 -fschedule-insns -fschedule-insns2 @gol
3668 -fsched-interblock -fsched-spec @gol
3671 -freorder-blocks -freorder-functions @gol
3672 -fstrict-aliasing @gol
3673 -funit-at-a-time @gol
3674 -falign-functions -falign-jumps @gol
3675 -falign-loops -falign-labels}
3677 Please note the warning under @option{-fgcse} about
3678 invoking @option{-O2} on programs that use computed gotos.
3682 Optimize yet more. @option{-O3} turns on all optimizations specified by
3683 @option{-O2} and also turns on the @option{-finline-functions},
3684 @option{-fweb} and @option{-frename-registers} options.
3688 Do not optimize. This is the default.
3692 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
3693 do not typically increase code size. It also performs further
3694 optimizations designed to reduce code size.
3696 @option{-Os} disables the following optimization flags:
3697 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
3698 -falign-labels -freorder-blocks -fprefetch-loop-arrays}
3700 If you use multiple @option{-O} options, with or without level numbers,
3701 the last such option is the one that is effective.
3704 Options of the form @option{-f@var{flag}} specify machine-independent
3705 flags. Most flags have both positive and negative forms; the negative
3706 form of @option{-ffoo} would be @option{-fno-foo}. In the table
3707 below, only one of the forms is listed---the one you typically will
3708 use. You can figure out the other form by either removing @samp{no-}
3711 The following options control specific optimizations. They are either
3712 activated by @option{-O} options or are related to ones that are. You
3713 can use the following flags in the rare cases when ``fine-tuning'' of
3714 optimizations to be performed is desired.
3717 @item -fno-default-inline
3718 @opindex fno-default-inline
3719 Do not make member functions inline by default merely because they are
3720 defined inside the class scope (C++ only). Otherwise, when you specify
3721 @w{@option{-O}}, member functions defined inside class scope are compiled
3722 inline by default; i.e., you don't need to add @samp{inline} in front of
3723 the member function name.
3725 @item -fno-defer-pop
3726 @opindex fno-defer-pop
3727 Always pop the arguments to each function call as soon as that function
3728 returns. For machines which must pop arguments after a function call,
3729 the compiler normally lets arguments accumulate on the stack for several
3730 function calls and pops them all at once.
3732 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3736 Force memory operands to be copied into registers before doing
3737 arithmetic on them. This produces better code by making all memory
3738 references potential common subexpressions. When they are not common
3739 subexpressions, instruction combination should eliminate the separate
3742 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3745 @opindex fforce-addr
3746 Force memory address constants to be copied into registers before
3747 doing arithmetic on them. This may produce better code just as
3748 @option{-fforce-mem} may.
3750 @item -fomit-frame-pointer
3751 @opindex fomit-frame-pointer
3752 Don't keep the frame pointer in a register for functions that
3753 don't need one. This avoids the instructions to save, set up and
3754 restore frame pointers; it also makes an extra register available
3755 in many functions. @strong{It also makes debugging impossible on
3758 On some machines, such as the VAX, this flag has no effect, because
3759 the standard calling sequence automatically handles the frame pointer
3760 and nothing is saved by pretending it doesn't exist. The
3761 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3762 whether a target machine supports this flag. @xref{Registers,,Register
3763 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
3765 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3767 @item -foptimize-sibling-calls
3768 @opindex foptimize-sibling-calls
3769 Optimize sibling and tail recursive calls.
3771 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3775 Don't pay attention to the @code{inline} keyword. Normally this option
3776 is used to keep the compiler from expanding any functions inline.
3777 Note that if you are not optimizing, no functions can be expanded inline.
3779 @item -finline-functions
3780 @opindex finline-functions
3781 Integrate all simple functions into their callers. The compiler
3782 heuristically decides which functions are simple enough to be worth
3783 integrating in this way.
3785 If all calls to a given function are integrated, and the function is
3786 declared @code{static}, then the function is normally not output as
3787 assembler code in its own right.
3789 Enabled at level @option{-O3}.
3791 @item -finline-limit=@var{n}
3792 @opindex finline-limit
3793 By default, gcc limits the size of functions that can be inlined. This flag
3794 allows the control of this limit for functions that are explicitly marked as
3795 inline (i.e., marked with the inline keyword or defined within the class
3796 definition in c++). @var{n} is the size of functions that can be inlined in
3797 number of pseudo instructions (not counting parameter handling). The default
3798 value of @var{n} is 600.
3799 Increasing this value can result in more inlined code at
3800 the cost of compilation time and memory consumption. Decreasing usually makes
3801 the compilation faster and less code will be inlined (which presumably
3802 means slower programs). This option is particularly useful for programs that
3803 use inlining heavily such as those based on recursive templates with C++.
3805 Inlining is actually controlled by a number of parameters, which may be
3806 specified individually by using @option{--param @var{name}=@var{value}}.
3807 The @option{-finline-limit=@var{n}} option sets some of these parameters
3811 @item max-inline-insns-single
3812 is set to @var{n}/2.
3813 @item max-inline-insns-auto
3814 is set to @var{n}/2.
3815 @item min-inline-insns
3816 is set to 130 or @var{n}/4, whichever is smaller.
3817 @item max-inline-insns-rtl
3821 See below for a documentation of the individual
3822 parameters controlling inlining.
3824 @emph{Note:} pseudo instruction represents, in this particular context, an
3825 abstract measurement of function's size. In no way, it represents a count
3826 of assembly instructions and as such its exact meaning might change from one
3827 release to an another.
3829 @item -fkeep-inline-functions
3830 @opindex fkeep-inline-functions
3831 Even if all calls to a given function are integrated, and the function
3832 is declared @code{static}, nevertheless output a separate run-time
3833 callable version of the function. This switch does not affect
3834 @code{extern inline} functions.
3836 @item -fkeep-static-consts
3837 @opindex fkeep-static-consts
3838 Emit variables declared @code{static const} when optimization isn't turned
3839 on, even if the variables aren't referenced.
3841 GCC enables this option by default. If you want to force the compiler to
3842 check if the variable was referenced, regardless of whether or not
3843 optimization is turned on, use the @option{-fno-keep-static-consts} option.
3845 @item -fmerge-constants
3846 Attempt to merge identical constants (string constants and floating point
3847 constants) across compilation units.
3849 This option is the default for optimized compilation if the assembler and
3850 linker support it. Use @option{-fno-merge-constants} to inhibit this
3853 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3855 @item -fmerge-all-constants
3856 Attempt to merge identical constants and identical variables.
3858 This option implies @option{-fmerge-constants}. In addition to
3859 @option{-fmerge-constants} this considers e.g. even constant initialized
3860 arrays or initialized constant variables with integral or floating point
3861 types. Languages like C or C++ require each non-automatic variable to
3862 have distinct location, so using this option will result in non-conforming
3867 Use a graph coloring register allocator. Currently this option is meant
3868 for testing, so we are interested to hear about miscompilations with
3871 @item -fno-branch-count-reg
3872 @opindex fno-branch-count-reg
3873 Do not use ``decrement and branch'' instructions on a count register,
3874 but instead generate a sequence of instructions that decrement a
3875 register, compare it against zero, then branch based upon the result.
3876 This option is only meaningful on architectures that support such
3877 instructions, which include x86, PowerPC, IA-64 and S/390.
3879 The default is @option{-fbranch-count-reg}, enabled when
3880 @option{-fstrength-reduce} is enabled.
3882 @item -fno-function-cse
3883 @opindex fno-function-cse
3884 Do not put function addresses in registers; make each instruction that
3885 calls a constant function contain the function's address explicitly.
3887 This option results in less efficient code, but some strange hacks
3888 that alter the assembler output may be confused by the optimizations
3889 performed when this option is not used.
3891 The default is @option{-ffunction-cse}
3893 @item -fno-zero-initialized-in-bss
3894 @opindex fno-zero-initialized-in-bss
3895 If the target supports a BSS section, GCC by default puts variables that
3896 are initialized to zero into BSS@. This can save space in the resulting
3899 This option turns off this behavior because some programs explicitly
3900 rely on variables going to the data section. E.g., so that the
3901 resulting executable can find the beginning of that section and/or make
3902 assumptions based on that.
3904 The default is @option{-fzero-initialized-in-bss}.
3906 @item -fstrength-reduce
3907 @opindex fstrength-reduce
3908 Perform the optimizations of loop strength reduction and
3909 elimination of iteration variables.
3911 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3913 @item -fthread-jumps
3914 @opindex fthread-jumps
3915 Perform optimizations where we check to see if a jump branches to a
3916 location where another comparison subsumed by the first is found. If
3917 so, the first branch is redirected to either the destination of the
3918 second branch or a point immediately following it, depending on whether
3919 the condition is known to be true or false.
3921 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3923 @item -fcse-follow-jumps
3924 @opindex fcse-follow-jumps
3925 In common subexpression elimination, scan through jump instructions
3926 when the target of the jump is not reached by any other path. For
3927 example, when CSE encounters an @code{if} statement with an
3928 @code{else} clause, CSE will follow the jump when the condition
3931 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3933 @item -fcse-skip-blocks
3934 @opindex fcse-skip-blocks
3935 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
3936 follow jumps which conditionally skip over blocks. When CSE
3937 encounters a simple @code{if} statement with no else clause,
3938 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
3939 body of the @code{if}.
3941 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3943 @item -frerun-cse-after-loop
3944 @opindex frerun-cse-after-loop
3945 Re-run common subexpression elimination after loop optimizations has been
3948 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3950 @item -frerun-loop-opt
3951 @opindex frerun-loop-opt
3952 Run the loop optimizer twice.
3954 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3958 Perform a global common subexpression elimination pass.
3959 This pass also performs global constant and copy propagation.
3961 @emph{Note:} When compiling a program using computed gotos, a GCC
3962 extension, you may get better runtime performance if you disable
3963 the global common subexpression elimination pass by adding
3964 @option{-fno-gcse} to the command line.
3966 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3970 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
3971 attempt to move loads which are only killed by stores into themselves. This
3972 allows a loop containing a load/store sequence to be changed to a load outside
3973 the loop, and a copy/store within the loop.
3975 Enabled by default when gcse is enabled.
3979 When @option{-fgcse-sm} is enabled, a store motion pass is run after
3980 global common subexpression elimination. This pass will attempt to move
3981 stores out of loops. When used in conjunction with @option{-fgcse-lm},
3982 loops containing a load/store sequence can be changed to a load before
3983 the loop and a store after the loop.
3985 Enabled by default when gcse is enabled.
3989 When @option{-fgcse-las} is enabled, the global common subexpression
3990 elimination pass eliminates redundant loads that come after stores to the
3991 same memory location (both partial and full redundancies).
3993 Enabled by default when gcse is enabled.
3995 @item -floop-optimize
3996 @opindex floop-optimize
3997 Perform loop optimizations: move constant expressions out of loops, simplify
3998 exit test conditions and optionally do strength-reduction and loop unrolling as
4001 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4003 @item -fcrossjumping
4004 @opindex crossjumping
4005 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4006 resulting code may or may not perform better than without cross-jumping.
4008 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4010 @item -fif-conversion
4011 @opindex if-conversion
4012 Attempt to transform conditional jumps into branch-less equivalents. This
4013 include use of conditional moves, min, max, set flags and abs instructions, and
4014 some tricks doable by standard arithmetics. The use of conditional execution
4015 on chips where it is available is controlled by @code{if-conversion2}.
4017 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4019 @item -fif-conversion2
4020 @opindex if-conversion2
4021 Use conditional execution (where available) to transform conditional jumps into
4022 branch-less equivalents.
4024 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4026 @item -fdelete-null-pointer-checks
4027 @opindex fdelete-null-pointer-checks
4028 Use global dataflow analysis to identify and eliminate useless checks
4029 for null pointers. The compiler assumes that dereferencing a null
4030 pointer would have halted the program. If a pointer is checked after
4031 it has already been dereferenced, it cannot be null.
4033 In some environments, this assumption is not true, and programs can
4034 safely dereference null pointers. Use
4035 @option{-fno-delete-null-pointer-checks} to disable this optimization
4036 for programs which depend on that behavior.
4038 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4040 @item -fexpensive-optimizations
4041 @opindex fexpensive-optimizations
4042 Perform a number of minor optimizations that are relatively expensive.
4044 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4046 @item -foptimize-register-move
4048 @opindex foptimize-register-move
4050 Attempt to reassign register numbers in move instructions and as
4051 operands of other simple instructions in order to maximize the amount of
4052 register tying. This is especially helpful on machines with two-operand
4055 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4058 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4060 @item -fdelayed-branch
4061 @opindex fdelayed-branch
4062 If supported for the target machine, attempt to reorder instructions
4063 to exploit instruction slots available after delayed branch
4066 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4068 @item -fschedule-insns
4069 @opindex fschedule-insns
4070 If supported for the target machine, attempt to reorder instructions to
4071 eliminate execution stalls due to required data being unavailable. This
4072 helps machines that have slow floating point or memory load instructions
4073 by allowing other instructions to be issued until the result of the load
4074 or floating point instruction is required.
4076 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4078 @item -fschedule-insns2
4079 @opindex fschedule-insns2
4080 Similar to @option{-fschedule-insns}, but requests an additional pass of
4081 instruction scheduling after register allocation has been done. This is
4082 especially useful on machines with a relatively small number of
4083 registers and where memory load instructions take more than one cycle.
4085 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4087 @item -fno-sched-interblock
4088 @opindex fno-sched-interblock
4089 Don't schedule instructions across basic blocks. This is normally
4090 enabled by default when scheduling before register allocation, i.e.@:
4091 with @option{-fschedule-insns} or at @option{-O2} or higher.
4093 @item -fno-sched-spec
4094 @opindex fno-sched-spec
4095 Don't allow speculative motion of non-load instructions. This is normally
4096 enabled by default when scheduling before register allocation, i.e.@:
4097 with @option{-fschedule-insns} or at @option{-O2} or higher.
4099 @item -fsched-spec-load
4100 @opindex fsched-spec-load
4101 Allow speculative motion of some load instructions. This only makes
4102 sense when scheduling before register allocation, i.e.@: with
4103 @option{-fschedule-insns} or at @option{-O2} or higher.
4105 @item -fsched-spec-load-dangerous
4106 @opindex fsched-spec-load-dangerous
4107 Allow speculative motion of more load instructions. This only makes
4108 sense when scheduling before register allocation, i.e.@: with
4109 @option{-fschedule-insns} or at @option{-O2} or higher.
4111 @item -fsched-stalled-insns=@var{n}
4112 @opindex fsched-stalled-insns
4113 Define how many insns (if any) can be moved prematurely from the queue
4114 of stalled insns into the ready list, during the second scheduling pass.
4116 @item -fsched-stalled-insns-dep=@var{n}
4117 @opindex fsched-stalled-insns-dep
4118 Define how many insn groups (cycles) will be examined for a dependency
4119 on a stalled insn that is candidate for premature removal from the queue
4120 of stalled insns. Has an effect only during the second scheduling pass,
4121 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4123 @item -fsched2-use-superblocks
4124 @opindex fsched2-use-superblocks
4125 When scheduling after register allocation, do use superblock scheduling
4126 algorithm. Superblock scheduling allows motion across basic block boundaries
4127 resulting on faster schedules. This option is experimental, as not all machine
4128 descriptions used by GCC model the CPU closely enough to avoid unreliable
4129 results from the algorithm.
4131 This only makes sense when scheduling after register allocation, i.e.@: with
4132 @option{-fschedule-insns2} or at @option{-O2} or higher.
4134 @item -fsched2-use-traces
4135 @opindex fsched2-use-traces
4136 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4137 allocation and additionally perform code duplication in order to increase the
4138 size of superblocks using tracer pass. See @option{-ftracer} for details on
4141 This mode should produce faster but significantly longer programs. Also
4142 without @code{-fbranch-probabilities} the traces constructed may not match the
4143 reality and hurt the performance. This only makes
4144 sense when scheduling after register allocation, i.e.@: with
4145 @option{-fschedule-insns2} or at @option{-O2} or higher.
4147 @item -fcaller-saves
4148 @opindex fcaller-saves
4149 Enable values to be allocated in registers that will be clobbered by
4150 function calls, by emitting extra instructions to save and restore the
4151 registers around such calls. Such allocation is done only when it
4152 seems to result in better code than would otherwise be produced.
4154 This option is always enabled by default on certain machines, usually
4155 those which have no call-preserved registers to use instead.
4157 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4159 @item -fmove-all-movables
4160 @opindex fmove-all-movables
4161 Forces all invariant computations in loops to be moved
4164 @item -freduce-all-givs
4165 @opindex freduce-all-givs
4166 Forces all general-induction variables in loops to be
4169 @emph{Note:} When compiling programs written in Fortran,
4170 @option{-fmove-all-movables} and @option{-freduce-all-givs} are enabled
4171 by default when you use the optimizer.
4173 These options may generate better or worse code; results are highly
4174 dependent on the structure of loops within the source code.
4176 These two options are intended to be removed someday, once
4177 they have helped determine the efficacy of various
4178 approaches to improving loop optimizations.
4180 Please let us (@w{@email{gcc@@gcc.gnu.org}} and @w{@email{fortran@@gnu.org}})
4181 know how use of these options affects
4182 the performance of your production code.
4183 We're very interested in code that runs @emph{slower}
4184 when these options are @emph{enabled}.
4187 @itemx -fno-peephole2
4188 @opindex fno-peephole
4189 @opindex fno-peephole2
4190 Disable any machine-specific peephole optimizations. The difference
4191 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4192 are implemented in the compiler; some targets use one, some use the
4193 other, a few use both.
4195 @option{-fpeephole} is enabled by default.
4196 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4198 @item -fno-guess-branch-probability
4199 @opindex fno-guess-branch-probability
4200 Do not guess branch probabilities using a randomized model.
4202 Sometimes gcc will opt to use a randomized model to guess branch
4203 probabilities, when none are available from either profiling feedback
4204 (@option{-fprofile-arcs}) or @samp{__builtin_expect}. This means that
4205 different runs of the compiler on the same program may produce different
4208 In a hard real-time system, people don't want different runs of the
4209 compiler to produce code that has different behavior; minimizing
4210 non-determinism is of paramount import. This switch allows users to
4211 reduce non-determinism, possibly at the expense of inferior
4214 The default is @option{-fguess-branch-probability} at levels
4215 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4217 @item -freorder-blocks
4218 @opindex freorder-blocks
4219 Reorder basic blocks in the compiled function in order to reduce number of
4220 taken branches and improve code locality.
4222 Enabled at levels @option{-O2}, @option{-O3}.
4224 @item -freorder-functions
4225 @opindex freorder-functions
4226 Reorder basic blocks in the compiled function in order to reduce number of
4227 taken branches and improve code locality. This is implemented by using special
4228 subsections @code{text.hot} for most frequently executed functions and
4229 @code{text.unlikely} for unlikely executed functions. Reordering is done by
4230 the linker so object file format must support named sections and linker must
4231 place them in a reasonable way.
4233 Also profile feedback must be available in to make this option effective. See
4234 @option{-fprofile-arcs} for details.
4236 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4238 @item -fstrict-aliasing
4239 @opindex fstrict-aliasing
4240 Allows the compiler to assume the strictest aliasing rules applicable to
4241 the language being compiled. For C (and C++), this activates
4242 optimizations based on the type of expressions. In particular, an
4243 object of one type is assumed never to reside at the same address as an
4244 object of a different type, unless the types are almost the same. For
4245 example, an @code{unsigned int} can alias an @code{int}, but not a
4246 @code{void*} or a @code{double}. A character type may alias any other
4249 Pay special attention to code like this:
4262 The practice of reading from a different union member than the one most
4263 recently written to (called ``type-punning'') is common. Even with
4264 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4265 is accessed through the union type. So, the code above will work as
4266 expected. However, this code might not:
4277 Every language that wishes to perform language-specific alias analysis
4278 should define a function that computes, given an @code{tree}
4279 node, an alias set for the node. Nodes in different alias sets are not
4280 allowed to alias. For an example, see the C front-end function
4281 @code{c_get_alias_set}.
4283 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4285 @item -falign-functions
4286 @itemx -falign-functions=@var{n}
4287 @opindex falign-functions
4288 Align the start of functions to the next power-of-two greater than
4289 @var{n}, skipping up to @var{n} bytes. For instance,
4290 @option{-falign-functions=32} aligns functions to the next 32-byte
4291 boundary, but @option{-falign-functions=24} would align to the next
4292 32-byte boundary only if this can be done by skipping 23 bytes or less.
4294 @option{-fno-align-functions} and @option{-falign-functions=1} are
4295 equivalent and mean that functions will not be aligned.
4297 Some assemblers only support this flag when @var{n} is a power of two;
4298 in that case, it is rounded up.
4300 If @var{n} is not specified or is zero, use a machine-dependent default.
4302 Enabled at levels @option{-O2}, @option{-O3}.
4304 @item -falign-labels
4305 @itemx -falign-labels=@var{n}
4306 @opindex falign-labels
4307 Align all branch targets to a power-of-two boundary, skipping up to
4308 @var{n} bytes like @option{-falign-functions}. This option can easily
4309 make code slower, because it must insert dummy operations for when the
4310 branch target is reached in the usual flow of the code.
4312 @option{-fno-align-labels} and @option{-falign-labels=1} are
4313 equivalent and mean that labels will not be aligned.
4315 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
4316 are greater than this value, then their values are used instead.
4318 If @var{n} is not specified or is zero, use a machine-dependent default
4319 which is very likely to be @samp{1}, meaning no alignment.
4321 Enabled at levels @option{-O2}, @option{-O3}.
4324 @itemx -falign-loops=@var{n}
4325 @opindex falign-loops
4326 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
4327 like @option{-falign-functions}. The hope is that the loop will be
4328 executed many times, which will make up for any execution of the dummy
4331 @option{-fno-align-loops} and @option{-falign-loops=1} are
4332 equivalent and mean that loops will not be aligned.
4334 If @var{n} is not specified or is zero, use a machine-dependent default.
4336 Enabled at levels @option{-O2}, @option{-O3}.
4339 @itemx -falign-jumps=@var{n}
4340 @opindex falign-jumps
4341 Align branch targets to a power-of-two boundary, for branch targets
4342 where the targets can only be reached by jumping, skipping up to @var{n}
4343 bytes like @option{-falign-functions}. In this case, no dummy operations
4346 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
4347 equivalent and mean that loops will not be aligned.
4349 If @var{n} is not specified or is zero, use a machine-dependent default.
4351 Enabled at levels @option{-O2}, @option{-O3}.
4353 @item -frename-registers
4354 @opindex frename-registers
4355 Attempt to avoid false dependencies in scheduled code by making use
4356 of registers left over after register allocation. This optimization
4357 will most benefit processors with lots of registers. It can, however,
4358 make debugging impossible, since variables will no longer stay in
4359 a ``home register''.
4363 Constructs webs as commonly used for register allocation purposes and assign
4364 each web individual pseudo register. This allows our register allocation pass
4365 to operate on pseudos directly, but also strengthens several other optimization
4366 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
4367 however, make debugging impossible, since variables will no longer stay in a
4370 Enabled at levels @option{-O3}.
4372 @item -fno-cprop-registers
4373 @opindex fno-cprop-registers
4374 After register allocation and post-register allocation instruction splitting,
4375 we perform a copy-propagation pass to try to reduce scheduling dependencies
4376 and occasionally eliminate the copy.
4378 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4382 The following options control compiler behavior regarding floating
4383 point arithmetic. These options trade off between speed and
4384 correctness. All must be specifically enabled.
4388 @opindex ffloat-store
4389 Do not store floating point variables in registers, and inhibit other
4390 options that might change whether a floating point value is taken from a
4393 @cindex floating point precision
4394 This option prevents undesirable excess precision on machines such as
4395 the 68000 where the floating registers (of the 68881) keep more
4396 precision than a @code{double} is supposed to have. Similarly for the
4397 x86 architecture. For most programs, the excess precision does only
4398 good, but a few programs rely on the precise definition of IEEE floating
4399 point. Use @option{-ffloat-store} for such programs, after modifying
4400 them to store all pertinent intermediate computations into variables.
4404 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
4405 @option{-fno-trapping-math}, @option{-ffinite-math-only},
4406 @option{-fno-rounding-math} and @option{-fno-signaling-nans}.
4408 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
4410 This option should never be turned on by any @option{-O} option since
4411 it can result in incorrect output for programs which depend on
4412 an exact implementation of IEEE or ISO rules/specifications for
4415 @item -fno-math-errno
4416 @opindex fno-math-errno
4417 Do not set ERRNO after calling math functions that are executed
4418 with a single instruction, e.g., sqrt. A program that relies on
4419 IEEE exceptions for math error handling may want to use this flag
4420 for speed while maintaining IEEE arithmetic compatibility.
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{-fmath-errno}.
4429 @item -funsafe-math-optimizations
4430 @opindex funsafe-math-optimizations
4431 Allow optimizations for floating-point arithmetic that (a) assume
4432 that arguments and results are valid and (b) may violate IEEE or
4433 ANSI standards. When used at link-time, it may include libraries
4434 or startup files that change the default FPU control word or other
4435 similar optimizations.
4437 This option should never be turned on by any @option{-O} option since
4438 it can result in incorrect output for programs which depend on
4439 an exact implementation of IEEE or ISO rules/specifications for
4442 The default is @option{-fno-unsafe-math-optimizations}.
4444 @item -ffinite-math-only
4445 @opindex ffinite-math-only
4446 Allow optimizations for floating-point arithmetic that assume
4447 that arguments and results are not NaNs or +-Infs.
4449 This option should never be turned on by any @option{-O} option since
4450 it can result in incorrect output for programs which depend on
4451 an exact implementation of IEEE or ISO rules/specifications.
4453 The default is @option{-fno-finite-math-only}.
4455 @item -fno-trapping-math
4456 @opindex fno-trapping-math
4457 Compile code assuming that floating-point operations cannot generate
4458 user-visible traps. These traps include division by zero, overflow,
4459 underflow, inexact result and invalid operation. This option implies
4460 @option{-fno-signaling-nans}. Setting this option may allow faster
4461 code if one relies on ``non-stop'' IEEE arithmetic, for example.
4463 This option should never be turned on by any @option{-O} option since
4464 it can result in incorrect output for programs which depend on
4465 an exact implementation of IEEE or ISO rules/specifications for
4468 The default is @option{-ftrapping-math}.
4470 @item -frounding-math
4471 @opindex frounding-math
4472 Disable transformations and optimizations that assume default floating
4473 point rounding behavior. This is round-to-zero for all floating point
4474 to integer conversions, and round-to-nearest for all other arithmetic
4475 truncations. This option should be specified for programs that change
4476 the FP rounding mode dynamically, or that may be executed with a
4477 non-default rounding mode. This option disables constant folding of
4478 floating point expressions at compile-time (which may be affected by
4479 rounding mode) and arithmetic transformations that are unsafe in the
4480 presence of sign-dependent rounding modes.
4482 The default is @option{-fno-rounding-math}.
4484 This option is experimental and does not currently guarantee to
4485 disable all GCC optimizations that are affected by rounding mode.
4486 Future versions of gcc may provide finer control of this setting
4487 using C99's @code{FENV_ACCESS} pragma. This command line option
4488 will be used to specify the default state for @code{FENV_ACCESS}.
4490 @item -fsignaling-nans
4491 @opindex fsignaling-nans
4492 Compile code assuming that IEEE signaling NaNs may generate user-visible
4493 traps during floating-point operations. Setting this option disables
4494 optimizations that may change the number of exceptions visible with
4495 signaling NaNs. This option implies @option{-ftrapping-math}.
4497 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
4500 The default is @option{-fno-signaling-nans}.
4502 This option is experimental and does not currently guarantee to
4503 disable all GCC optimizations that affect signaling NaN behavior.
4505 @item -fsingle-precision-constant
4506 @opindex fsingle-precision-constant
4507 Treat floating point constant as single precision constant instead of
4508 implicitly converting it to double precision constant.
4513 The following options control optimizations that may improve
4514 performance, but are not enabled by any @option{-O} options. This
4515 section includes experimental options that may produce broken code.
4518 @item -fbranch-probabilities
4519 @opindex fbranch-probabilities
4520 After running a program compiled with @option{-fprofile-arcs}
4521 (@pxref{Debugging Options,, Options for Debugging Your Program or
4522 @command{gcc}}), you can compile it a second time using
4523 @option{-fbranch-probabilities}, to improve optimizations based on
4524 the number of times each branch was taken. When the program
4525 compiled with @option{-fprofile-arcs} exits it saves arc execution
4526 counts to a file called @file{@var{sourcename}.gcda} for each source
4527 file The information in this data file is very dependent on the
4528 structure of the generated code, so you must use the same source code
4529 and the same optimization options for both compilations.
4531 With @option{-fbranch-probabilities}, GCC puts a
4532 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
4533 These can be used to improve optimization. Currently, they are only
4534 used in one place: in @file{reorg.c}, instead of guessing which path a
4535 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
4536 exactly determine which path is taken more often.
4538 @item -fprofile-values
4539 @opindex fprofile-values
4540 If combined with @option{-fprofile-arcs}, it adds code so that some
4541 data about values of expressions in the program is gathered.
4543 With @option{-fbranch-probabilities}, it reads back the data gathered
4544 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
4545 notes to instructions for their later usage in optimizations.
4549 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
4550 a code to gather information about values of expressions.
4552 With @option{-fbranch-probabilities}, it reads back the data gathered
4553 and actually performs the optimizations based on them.
4554 Currently the optimizations include specialization of division operation
4555 using the knowledge about the value of the denominator.
4559 Use a graph coloring register allocator. Currently this option is meant
4560 for testing, so we are interested to hear about miscompilations with
4565 Perform tail duplication to enlarge superblock size. This transformation
4566 simplifies the control flow of the function allowing other optimizations to do
4569 @item -funit-at-a-time
4570 @opindex funit-at-a-time
4571 Parse the whole compilation unit before starting to produce code.
4572 This allows some extra optimizations to take place but consumes more
4575 @item -funroll-loops
4576 @opindex funroll-loops
4577 Unroll loops whose number of iterations can be determined at compile time or
4578 upon entry to the loop. @option{-funroll-loops} implies
4579 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
4580 (i.e. complete removal of loops with small constant number of iterations).
4581 This option makes code larger, and may or may not make it run faster.
4583 @item -funroll-all-loops
4584 @opindex funroll-all-loops
4585 Unroll all loops, even if their number of iterations is uncertain when
4586 the loop is entered. This usually makes programs run more slowly.
4587 @option{-funroll-all-loops} implies the same options as
4588 @option{-funroll-loops}.
4591 @opindex fpeel-loops
4592 Peels the loops for that there is enough information that they do not
4593 roll much (from profile feedback). It also turns on complete loop peeling
4594 (i.e. complete removal of loops with small constant number of iterations).
4596 @item -funswitch-loops
4597 @opindex funswitch-loops
4598 Move branches with loop invariant conditions out of the loop, with duplicates
4599 of the loop on both branches (modified according to result of the condition).
4601 @item -fold-unroll-loops
4602 @opindex fold-unroll-loops
4603 Unroll loops whose number of iterations can be determined at compile
4604 time or upon entry to the loop, using the old loop unroller whose loop
4605 recognition is based on notes from frontend. @option{-fold-unroll-loops} implies
4606 both @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4607 option makes code larger, and may or may not make it run faster.
4609 @item -fold-unroll-all-loops
4610 @opindex fold-unroll-all-loops
4611 Unroll all loops, even if their number of iterations is uncertain when
4612 the loop is entered. This is done using the old loop unroller whose loop
4613 recognition is based on notes from frontend. This usually makes programs run more slowly.
4614 @option{-fold-unroll-all-loops} implies the same options as
4615 @option{-fold-unroll-loops}.
4617 @item -funswitch-loops
4618 @opindex funswitch-loops
4619 Move branches with loop invariant conditions out of the loop, with duplicates
4620 of the loop on both branches (modified according to result of the condition).
4622 @item -funswitch-loops
4623 @opindex funswitch-loops
4624 Move branches with loop invariant conditions out of the loop, with duplicates
4625 of the loop on both branches (modified according to result of the condition).
4627 @item -fprefetch-loop-arrays
4628 @opindex fprefetch-loop-arrays
4629 If supported by the target machine, generate instructions to prefetch
4630 memory to improve the performance of loops that access large arrays.
4632 Disabled at level @option{-Os}.
4634 @item -ffunction-sections
4635 @itemx -fdata-sections
4636 @opindex ffunction-sections
4637 @opindex fdata-sections
4638 Place each function or data item into its own section in the output
4639 file if the target supports arbitrary sections. The name of the
4640 function or the name of the data item determines the section's name
4643 Use these options on systems where the linker can perform optimizations
4644 to improve locality of reference in the instruction space. Most systems
4645 using the ELF object format and SPARC processors running Solaris 2 have
4646 linkers with such optimizations. AIX may have these optimizations in
4649 Only use these options when there are significant benefits from doing
4650 so. When you specify these options, the assembler and linker will
4651 create larger object and executable files and will also be slower.
4652 You will not be able to use @code{gprof} on all systems if you
4653 specify this option and you may have problems with debugging if
4654 you specify both this option and @option{-g}.
4656 @item -fbranch-target-load-optimize
4657 @opindex fbranch-target-load-optimize
4658 Perform branch target register load optimization before prologue / epilogue
4660 The use of target registers can typically be exposed only during reload,
4661 thus hoisting loads out of loops and doing inter-block scheduling needs
4662 a separate optimization pass.
4664 @item -fbranch-target-load-optimize2
4665 @opindex fbranch-target-load-optimize2
4666 Perform branch target register load optimization after prologue / epilogue
4669 @item --param @var{name}=@var{value}
4671 In some places, GCC uses various constants to control the amount of
4672 optimization that is done. For example, GCC will not inline functions
4673 that contain more that a certain number of instructions. You can
4674 control some of these constants on the command-line using the
4675 @option{--param} option.
4677 In each case, the @var{value} is an integer. The allowable choices for
4678 @var{name} are given in the following table:
4681 @item max-crossjump-edges
4682 The maximum number of incoming edges to consider for crossjumping.
4683 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
4684 the number of edges incoming to each block. Increasing values mean
4685 more aggressive optimization, making the compile time increase with
4686 probably small improvement in executable size.
4688 @item max-delay-slot-insn-search
4689 The maximum number of instructions to consider when looking for an
4690 instruction to fill a delay slot. If more than this arbitrary number of
4691 instructions is searched, the time savings from filling the delay slot
4692 will be minimal so stop searching. Increasing values mean more
4693 aggressive optimization, making the compile time increase with probably
4694 small improvement in executable run time.
4696 @item max-delay-slot-live-search
4697 When trying to fill delay slots, the maximum number of instructions to
4698 consider when searching for a block with valid live register
4699 information. Increasing this arbitrarily chosen value means more
4700 aggressive optimization, increasing the compile time. This parameter
4701 should be removed when the delay slot code is rewritten to maintain the
4704 @item max-gcse-memory
4705 The approximate maximum amount of memory that will be allocated in
4706 order to perform the global common subexpression elimination
4707 optimization. If more memory than specified is required, the
4708 optimization will not be done.
4710 @item max-gcse-passes
4711 The maximum number of passes of GCSE to run.
4713 @item max-pending-list-length
4714 The maximum number of pending dependencies scheduling will allow
4715 before flushing the current state and starting over. Large functions
4716 with few branches or calls can create excessively large lists which
4717 needlessly consume memory and resources.
4719 @item max-inline-insns-single
4720 Several parameters control the tree inliner used in gcc.
4721 This number sets the maximum number of instructions (counted in gcc's
4722 internal representation) in a single function that the tree inliner
4723 will consider for inlining. This only affects functions declared
4724 inline and methods implemented in a class declaration (C++).
4725 The default value is 500.
4727 @item max-inline-insns-auto
4728 When you use @option{-finline-functions} (included in @option{-O3}),
4729 a lot of functions that would otherwise not be considered for inlining
4730 by the compiler will be investigated. To those functions, a different
4731 (more restrictive) limit compared to functions declared inline can
4733 The default value is 150.
4735 @item large-function-insns
4736 The limit specifying really large functions. For functions greater than this
4737 limit inlining is constrained by @option{--param large-function-growth}.
4738 This parameter is useful primarily to avoid extreme compilation time caused by non-linear
4739 algorithms used by the backend.
4740 This parameter is ignored when @option{-funit-at-a-time} is not used.
4741 The default value is 30000.
4743 @item large-function-growth
4744 Specifies maximal growth of large functtion caused by inlining in percents.
4745 This parameter is ignored when @option{-funit-at-a-time} is not used.
4746 The default value is 200.
4748 @item inline-unit-growth
4749 Specifies maximal overall growth of the compilation unit caused by inlining.
4750 This parameter is ignored when @option{-funit-at-a-time} is not used.
4751 The default value is 150.
4753 @item max-inline-insns-rtl
4754 For languages that use the RTL inliner (this happens at a later stage
4755 than tree inlining), you can set the maximum allowable size (counted
4756 in RTL instructions) for the RTL inliner with this parameter.
4757 The default value is 600.
4759 @item max-unrolled-insns
4760 The maximum number of instructions that a loop should have if that loop
4761 is unrolled, and if the loop is unrolled, it determines how many times
4762 the loop code is unrolled.
4764 @item max-average-unrolled-insns
4765 The maximum number of instructions biased by probabilities of their execution
4766 that a loop should have if that loop is unrolled, and if the loop is unrolled,
4767 it determines how many times the loop code is unrolled.
4769 @item max-unroll-times
4770 The maximum number of unrollings of a single loop.
4772 @item max-peeled-insns
4773 The maximum number of instructions that a loop should have if that loop
4774 is peeled, and if the loop is peeled, it determines how many times
4775 the loop code is peeled.
4777 @item max-peel-times
4778 The maximum number of peelings of a single loop.
4780 @item max-completely-peeled-insns
4781 The maximum number of insns of a completely peeled loop.
4783 @item max-completely-peel-times
4784 The maximum number of iterations of a loop to be suitable for complete peeling.
4786 @item max-unswitch-insns
4787 The maximum number of insns of an unswitched loop.
4789 @item max-unswitch-level
4790 The maximum number of branches unswitched in a single loop.
4792 @item hot-bb-count-fraction
4793 Select fraction of the maximal count of repetitions of basic block in program
4794 given basic block needs to have to be considered hot.
4796 @item hot-bb-frequency-fraction
4797 Select fraction of the maximal frequency of executions of basic block in
4798 function given basic block needs to have to be considered hot
4800 @item tracer-dynamic-coverage
4801 @itemx tracer-dynamic-coverage-feedback
4803 This value is used to limit superblock formation once the given percentage of
4804 executed instructions is covered. This limits unnecessary code size
4807 The @option{tracer-dynamic-coverage-feedback} is used only when profile
4808 feedback is available. The real profiles (as opposed to statically estimated
4809 ones) are much less balanced allowing the threshold to be larger value.
4811 @item tracer-max-code-growth
4812 Stop tail duplication once code growth has reached given percentage. This is
4813 rather hokey argument, as most of the duplicates will be eliminated later in
4814 cross jumping, so it may be set to much higher values than is the desired code
4817 @item tracer-min-branch-ratio
4819 Stop reverse growth when the reverse probability of best edge is less than this
4820 threshold (in percent).
4822 @item tracer-min-branch-ratio
4823 @itemx tracer-min-branch-ratio-feedback
4825 Stop forward growth if the best edge do have probability lower than this
4828 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
4829 compilation for profile feedback and one for compilation without. The value
4830 for compilation with profile feedback needs to be more conservative (higher) in
4831 order to make tracer effective.
4833 @item max-cse-path-length
4835 Maximum number of basic blocks on path that cse considers.
4837 @item ggc-min-expand
4839 GCC uses a garbage collector to manage its own memory allocation. This
4840 parameter specifies the minimum percentage by which the garbage
4841 collector's heap should be allowed to expand between collections.
4842 Tuning this may improve compilation speed; it has no effect on code
4845 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
4846 RAM >= 1GB. If @code{getrlimit} is available, the notion of "RAM" is
4847 the smallest of actual RAM, RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If
4848 GCC is not able to calculate RAM on a particular platform, the lower
4849 bound of 30% is used. Setting this parameter and
4850 @option{ggc-min-heapsize} to zero causes a full collection to occur at
4851 every opportunity. This is extremely slow, but can be useful for
4854 @item ggc-min-heapsize
4856 Minimum size of the garbage collector's heap before it begins bothering
4857 to collect garbage. The first collection occurs after the heap expands
4858 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
4859 tuning this may improve compilation speed, and has no effect on code
4862 The default is RAM/8, with a lower bound of 4096 (four megabytes) and an
4863 upper bound of 131072 (128 megabytes). If @code{getrlimit} is
4864 available, the notion of "RAM" is the smallest of actual RAM,
4865 RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If GCC is not able to calculate
4866 RAM on a particular platform, the lower bound is used. Setting this
4867 parameter very large effectively disables garbage collection. Setting
4868 this parameter and @option{ggc-min-expand} to zero causes a full
4869 collection to occur at every opportunity.
4871 @item reorder-blocks-duplicate
4872 @itemx reorder-blocks-duplicate-feedback
4874 Used by basic block reordering pass to decide whether to use unconditional
4875 branch or duplicate the code on its destination. Code is duplicated when its
4876 estimated size is smaller than this value multiplied by the estimated size of
4877 unconditional jump in the hot spots of the program.
4879 The @option{reorder-block-duplicate-feedback} is used only when profile
4880 feedback is available and may be set to higher values than
4881 @option{reorder-block-duplicate} since information about the hot spots is more
4886 @node Preprocessor Options
4887 @section Options Controlling the Preprocessor
4888 @cindex preprocessor options
4889 @cindex options, preprocessor
4891 These options control the C preprocessor, which is run on each C source
4892 file before actual compilation.
4894 If you use the @option{-E} option, nothing is done except preprocessing.
4895 Some of these options make sense only together with @option{-E} because
4896 they cause the preprocessor output to be unsuitable for actual
4901 You can use @option{-Wp,@var{option}} to bypass the compiler driver
4902 and pass @var{option} directly through to the preprocessor. If
4903 @var{option} contains commas, it is split into multiple options at the
4904 commas. However, many options are modified, translated or interpreted
4905 by the compiler driver before being passed to the preprocessor, and
4906 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
4907 interface is undocumented and subject to change, so whenever possible
4908 you should avoid using @option{-Wp} and let the driver handle the
4911 @item -Xpreprocessor @var{option}
4912 @opindex preprocessor
4913 Pass @var{option} as an option to the preprocessor. You can use this to
4914 supply system-specific preprocessor options which GCC does not know how to
4917 If you want to pass an option that takes an argument, you must use
4918 @option{-Xpreprocessor} twice, once for the option and once for the argument.
4921 @include cppopts.texi
4923 @node Assembler Options
4924 @section Passing Options to the Assembler
4926 @c prevent bad page break with this line
4927 You can pass options to the assembler.
4930 @item -Wa,@var{option}
4932 Pass @var{option} as an option to the assembler. If @var{option}
4933 contains commas, it is split into multiple options at the commas.
4935 @item -Xassembler @var{option}
4937 Pass @var{option} as an option to the assembler. You can use this to
4938 supply system-specific assembler options which GCC does not know how to
4941 If you want to pass an option that takes an argument, you must use
4942 @option{-Xassembler} twice, once for the option and once for the argument.
4947 @section Options for Linking
4948 @cindex link options
4949 @cindex options, linking
4951 These options come into play when the compiler links object files into
4952 an executable output file. They are meaningless if the compiler is
4953 not doing a link step.
4957 @item @var{object-file-name}
4958 A file name that does not end in a special recognized suffix is
4959 considered to name an object file or library. (Object files are
4960 distinguished from libraries by the linker according to the file
4961 contents.) If linking is done, these object files are used as input
4970 If any of these options is used, then the linker is not run, and
4971 object file names should not be used as arguments. @xref{Overall
4975 @item -l@var{library}
4976 @itemx -l @var{library}
4978 Search the library named @var{library} when linking. (The second
4979 alternative with the library as a separate argument is only for
4980 POSIX compliance and is not recommended.)
4982 It makes a difference where in the command you write this option; the
4983 linker searches and processes libraries and object files in the order they
4984 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
4985 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
4986 to functions in @samp{z}, those functions may not be loaded.
4988 The linker searches a standard list of directories for the library,
4989 which is actually a file named @file{lib@var{library}.a}. The linker
4990 then uses this file as if it had been specified precisely by name.
4992 The directories searched include several standard system directories
4993 plus any that you specify with @option{-L}.
4995 Normally the files found this way are library files---archive files
4996 whose members are object files. The linker handles an archive file by
4997 scanning through it for members which define symbols that have so far
4998 been referenced but not defined. But if the file that is found is an
4999 ordinary object file, it is linked in the usual fashion. The only
5000 difference between using an @option{-l} option and specifying a file name
5001 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5002 and searches several directories.
5006 You need this special case of the @option{-l} option in order to
5007 link an Objective-C program.
5010 @opindex nostartfiles
5011 Do not use the standard system startup files when linking.
5012 The standard system libraries are used normally, unless @option{-nostdlib}
5013 or @option{-nodefaultlibs} is used.
5015 @item -nodefaultlibs
5016 @opindex nodefaultlibs
5017 Do not use the standard system libraries when linking.
5018 Only the libraries you specify will be passed to the linker.
5019 The standard startup files are used normally, unless @option{-nostartfiles}
5020 is used. The compiler may generate calls to memcmp, memset, and memcpy
5021 for System V (and ISO C) environments or to bcopy and bzero for
5022 BSD environments. These entries are usually resolved by entries in
5023 libc. These entry points should be supplied through some other
5024 mechanism when this option is specified.
5028 Do not use the standard system startup files or libraries when linking.
5029 No startup files and only the libraries you specify will be passed to
5030 the linker. The compiler may generate calls to memcmp, memset, and memcpy
5031 for System V (and ISO C) environments or to bcopy and bzero for
5032 BSD environments. These entries are usually resolved by entries in
5033 libc. These entry points should be supplied through some other
5034 mechanism when this option is specified.
5036 @cindex @option{-lgcc}, use with @option{-nostdlib}
5037 @cindex @option{-nostdlib} and unresolved references
5038 @cindex unresolved references and @option{-nostdlib}
5039 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5040 @cindex @option{-nodefaultlibs} and unresolved references
5041 @cindex unresolved references and @option{-nodefaultlibs}
5042 One of the standard libraries bypassed by @option{-nostdlib} and
5043 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5044 that GCC uses to overcome shortcomings of particular machines, or special
5045 needs for some languages.
5046 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5047 Collection (GCC) Internals},
5048 for more discussion of @file{libgcc.a}.)
5049 In most cases, you need @file{libgcc.a} even when you want to avoid
5050 other standard libraries. In other words, when you specify @option{-nostdlib}
5051 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5052 This ensures that you have no unresolved references to internal GCC
5053 library subroutines. (For example, @samp{__main}, used to ensure C++
5054 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5055 GNU Compiler Collection (GCC) Internals}.)
5059 Produce a position independent executable on targets which support it.
5060 For predictable results, you must also specify the same set of options
5061 that were used to generate code (@option{-fpie}, @option{-fPIE},
5062 or model suboptions) when you specify this option.
5066 Remove all symbol table and relocation information from the executable.
5070 On systems that support dynamic linking, this prevents linking with the shared
5071 libraries. On other systems, this option has no effect.
5075 Produce a shared object which can then be linked with other objects to
5076 form an executable. Not all systems support this option. For predictable
5077 results, you must also specify the same set of options that were used to
5078 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5079 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5080 needs to build supplementary stub code for constructors to work. On
5081 multi-libbed systems, @samp{gcc -shared} must select the correct support
5082 libraries to link against. Failing to supply the correct flags may lead
5083 to subtle defects. Supplying them in cases where they are not necessary
5086 @item -shared-libgcc
5087 @itemx -static-libgcc
5088 @opindex shared-libgcc
5089 @opindex static-libgcc
5090 On systems that provide @file{libgcc} as a shared library, these options
5091 force the use of either the shared or static version respectively.
5092 If no shared version of @file{libgcc} was built when the compiler was
5093 configured, these options have no effect.
5095 There are several situations in which an application should use the
5096 shared @file{libgcc} instead of the static version. The most common
5097 of these is when the application wishes to throw and catch exceptions
5098 across different shared libraries. In that case, each of the libraries
5099 as well as the application itself should use the shared @file{libgcc}.
5101 Therefore, the G++ and GCJ drivers automatically add
5102 @option{-shared-libgcc} whenever you build a shared library or a main
5103 executable, because C++ and Java programs typically use exceptions, so
5104 this is the right thing to do.
5106 If, instead, you use the GCC driver to create shared libraries, you may
5107 find that they will not always be linked with the shared @file{libgcc}.
5108 If GCC finds, at its configuration time, that you have a GNU linker that
5109 does not support option @option{--eh-frame-hdr}, it will link the shared
5110 version of @file{libgcc} into shared libraries by default. Otherwise,
5111 it will take advantage of the linker and optimize away the linking with
5112 the shared version of @file{libgcc}, linking with the static version of
5113 libgcc by default. This allows exceptions to propagate through such
5114 shared libraries, without incurring relocation costs at library load
5117 However, if a library or main executable is supposed to throw or catch
5118 exceptions, you must link it using the G++ or GCJ driver, as appropriate
5119 for the languages used in the program, or using the option
5120 @option{-shared-libgcc}, such that it is linked with the shared
5125 Bind references to global symbols when building a shared object. Warn
5126 about any unresolved references (unless overridden by the link editor
5127 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
5130 @item -Xlinker @var{option}
5132 Pass @var{option} as an option to the linker. You can use this to
5133 supply system-specific linker options which GCC does not know how to
5136 If you want to pass an option that takes an argument, you must use
5137 @option{-Xlinker} twice, once for the option and once for the argument.
5138 For example, to pass @option{-assert definitions}, you must write
5139 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
5140 @option{-Xlinker "-assert definitions"}, because this passes the entire
5141 string as a single argument, which is not what the linker expects.
5143 @item -Wl,@var{option}
5145 Pass @var{option} as an option to the linker. If @var{option} contains
5146 commas, it is split into multiple options at the commas.
5148 @item -u @var{symbol}
5150 Pretend the symbol @var{symbol} is undefined, to force linking of
5151 library modules to define it. You can use @option{-u} multiple times with
5152 different symbols to force loading of additional library modules.
5155 @node Directory Options
5156 @section Options for Directory Search
5157 @cindex directory options
5158 @cindex options, directory search
5161 These options specify directories to search for header files, for
5162 libraries and for parts of the compiler:
5167 Add the directory @var{dir} to the head of the list of directories to be
5168 searched for header files. This can be used to override a system header
5169 file, substituting your own version, since these directories are
5170 searched before the system header file directories. However, you should
5171 not use this option to add directories that contain vendor-supplied
5172 system header files (use @option{-isystem} for that). If you use more than
5173 one @option{-I} option, the directories are scanned in left-to-right
5174 order; the standard system directories come after.
5176 If a standard system include directory, or a directory specified with
5177 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
5178 option will be ignored. The directory will still be searched but as a
5179 system directory at its normal position in the system include chain.
5180 This is to ensure that GCC's procedure to fix buggy system headers and
5181 the ordering for the include_next directive are not inadvertently changed.
5182 If you really need to change the search order for system directories,
5183 use the @option{-nostdinc} and/or @option{-isystem} options.
5187 Any directories you specify with @option{-I} options before the @option{-I-}
5188 option are searched only for the case of @samp{#include "@var{file}"};
5189 they are not searched for @samp{#include <@var{file}>}.
5191 If additional directories are specified with @option{-I} options after
5192 the @option{-I-}, these directories are searched for all @samp{#include}
5193 directives. (Ordinarily @emph{all} @option{-I} directories are used
5196 In addition, the @option{-I-} option inhibits the use of the current
5197 directory (where the current input file came from) as the first search
5198 directory for @samp{#include "@var{file}"}. There is no way to
5199 override this effect of @option{-I-}. With @option{-I.} you can specify
5200 searching the directory which was current when the compiler was
5201 invoked. That is not exactly the same as what the preprocessor does
5202 by default, but it is often satisfactory.
5204 @option{-I-} does not inhibit the use of the standard system directories
5205 for header files. Thus, @option{-I-} and @option{-nostdinc} are
5210 Add directory @var{dir} to the list of directories to be searched
5213 @item -B@var{prefix}
5215 This option specifies where to find the executables, libraries,
5216 include files, and data files of the compiler itself.
5218 The compiler driver program runs one or more of the subprograms
5219 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
5220 @var{prefix} as a prefix for each program it tries to run, both with and
5221 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
5223 For each subprogram to be run, the compiler driver first tries the
5224 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
5225 was not specified, the driver tries two standard prefixes, which are
5226 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
5227 those results in a file name that is found, the unmodified program
5228 name is searched for using the directories specified in your
5229 @env{PATH} environment variable.
5231 The compiler will check to see if the path provided by the @option{-B}
5232 refers to a directory, and if necessary it will add a directory
5233 separator character at the end of the path.
5235 @option{-B} prefixes that effectively specify directory names also apply
5236 to libraries in the linker, because the compiler translates these
5237 options into @option{-L} options for the linker. They also apply to
5238 includes files in the preprocessor, because the compiler translates these
5239 options into @option{-isystem} options for the preprocessor. In this case,
5240 the compiler appends @samp{include} to the prefix.
5242 The run-time support file @file{libgcc.a} can also be searched for using
5243 the @option{-B} prefix, if needed. If it is not found there, the two
5244 standard prefixes above are tried, and that is all. The file is left
5245 out of the link if it is not found by those means.
5247 Another way to specify a prefix much like the @option{-B} prefix is to use
5248 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
5251 As a special kludge, if the path provided by @option{-B} is
5252 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
5253 9, then it will be replaced by @file{[dir/]include}. This is to help
5254 with boot-strapping the compiler.
5256 @item -specs=@var{file}
5258 Process @var{file} after the compiler reads in the standard @file{specs}
5259 file, in order to override the defaults that the @file{gcc} driver
5260 program uses when determining what switches to pass to @file{cc1},
5261 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
5262 @option{-specs=@var{file}} can be specified on the command line, and they
5263 are processed in order, from left to right.
5269 @section Specifying subprocesses and the switches to pass to them
5272 @command{gcc} is a driver program. It performs its job by invoking a
5273 sequence of other programs to do the work of compiling, assembling and
5274 linking. GCC interprets its command-line parameters and uses these to
5275 deduce which programs it should invoke, and which command-line options
5276 it ought to place on their command lines. This behavior is controlled
5277 by @dfn{spec strings}. In most cases there is one spec string for each
5278 program that GCC can invoke, but a few programs have multiple spec
5279 strings to control their behavior. The spec strings built into GCC can
5280 be overridden by using the @option{-specs=} command-line switch to specify
5283 @dfn{Spec files} are plaintext files that are used to construct spec
5284 strings. They consist of a sequence of directives separated by blank
5285 lines. The type of directive is determined by the first non-whitespace
5286 character on the line and it can be one of the following:
5289 @item %@var{command}
5290 Issues a @var{command} to the spec file processor. The commands that can
5294 @item %include <@var{file}>
5296 Search for @var{file} and insert its text at the current point in the
5299 @item %include_noerr <@var{file}>
5300 @cindex %include_noerr
5301 Just like @samp{%include}, but do not generate an error message if the include
5302 file cannot be found.
5304 @item %rename @var{old_name} @var{new_name}
5306 Rename the spec string @var{old_name} to @var{new_name}.
5310 @item *[@var{spec_name}]:
5311 This tells the compiler to create, override or delete the named spec
5312 string. All lines after this directive up to the next directive or
5313 blank line are considered to be the text for the spec string. If this
5314 results in an empty string then the spec will be deleted. (Or, if the
5315 spec did not exist, then nothing will happened.) Otherwise, if the spec
5316 does not currently exist a new spec will be created. If the spec does
5317 exist then its contents will be overridden by the text of this
5318 directive, unless the first character of that text is the @samp{+}
5319 character, in which case the text will be appended to the spec.
5321 @item [@var{suffix}]:
5322 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
5323 and up to the next directive or blank line are considered to make up the
5324 spec string for the indicated suffix. When the compiler encounters an
5325 input file with the named suffix, it will processes the spec string in
5326 order to work out how to compile that file. For example:
5333 This says that any input file whose name ends in @samp{.ZZ} should be
5334 passed to the program @samp{z-compile}, which should be invoked with the
5335 command-line switch @option{-input} and with the result of performing the
5336 @samp{%i} substitution. (See below.)
5338 As an alternative to providing a spec string, the text that follows a
5339 suffix directive can be one of the following:
5342 @item @@@var{language}
5343 This says that the suffix is an alias for a known @var{language}. This is
5344 similar to using the @option{-x} command-line switch to GCC to specify a
5345 language explicitly. For example:
5352 Says that .ZZ files are, in fact, C++ source files.
5355 This causes an error messages saying:
5358 @var{name} compiler not installed on this system.
5362 GCC already has an extensive list of suffixes built into it.
5363 This directive will add an entry to the end of the list of suffixes, but
5364 since the list is searched from the end backwards, it is effectively
5365 possible to override earlier entries using this technique.
5369 GCC has the following spec strings built into it. Spec files can
5370 override these strings or create their own. Note that individual
5371 targets can also add their own spec strings to this list.
5374 asm Options to pass to the assembler
5375 asm_final Options to pass to the assembler post-processor
5376 cpp Options to pass to the C preprocessor
5377 cc1 Options to pass to the C compiler
5378 cc1plus Options to pass to the C++ compiler
5379 endfile Object files to include at the end of the link
5380 link Options to pass to the linker
5381 lib Libraries to include on the command line to the linker
5382 libgcc Decides which GCC support library to pass to the linker
5383 linker Sets the name of the linker
5384 predefines Defines to be passed to the C preprocessor
5385 signed_char Defines to pass to CPP to say whether @code{char} is signed
5387 startfile Object files to include at the start of the link
5390 Here is a small example of a spec file:
5396 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
5399 This example renames the spec called @samp{lib} to @samp{old_lib} and
5400 then overrides the previous definition of @samp{lib} with a new one.
5401 The new definition adds in some extra command-line options before
5402 including the text of the old definition.
5404 @dfn{Spec strings} are a list of command-line options to be passed to their
5405 corresponding program. In addition, the spec strings can contain
5406 @samp{%}-prefixed sequences to substitute variable text or to
5407 conditionally insert text into the command line. Using these constructs
5408 it is possible to generate quite complex command lines.
5410 Here is a table of all defined @samp{%}-sequences for spec
5411 strings. Note that spaces are not generated automatically around the
5412 results of expanding these sequences. Therefore you can concatenate them
5413 together or combine them with constant text in a single argument.
5417 Substitute one @samp{%} into the program name or argument.
5420 Substitute the name of the input file being processed.
5423 Substitute the basename of the input file being processed.
5424 This is the substring up to (and not including) the last period
5425 and not including the directory.
5428 This is the same as @samp{%b}, but include the file suffix (text after
5432 Marks the argument containing or following the @samp{%d} as a
5433 temporary file name, so that that file will be deleted if GCC exits
5434 successfully. Unlike @samp{%g}, this contributes no text to the
5437 @item %g@var{suffix}
5438 Substitute a file name that has suffix @var{suffix} and is chosen
5439 once per compilation, and mark the argument in the same way as
5440 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
5441 name is now chosen in a way that is hard to predict even when previously
5442 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
5443 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
5444 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
5445 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
5446 was simply substituted with a file name chosen once per compilation,
5447 without regard to any appended suffix (which was therefore treated
5448 just like ordinary text), making such attacks more likely to succeed.
5450 @item %u@var{suffix}
5451 Like @samp{%g}, but generates a new temporary file name even if
5452 @samp{%u@var{suffix}} was already seen.
5454 @item %U@var{suffix}
5455 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
5456 new one if there is no such last file name. In the absence of any
5457 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
5458 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
5459 would involve the generation of two distinct file names, one
5460 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
5461 simply substituted with a file name chosen for the previous @samp{%u},
5462 without regard to any appended suffix.
5464 @item %j@var{suffix}
5465 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
5466 writable, and if save-temps is off; otherwise, substitute the name
5467 of a temporary file, just like @samp{%u}. This temporary file is not
5468 meant for communication between processes, but rather as a junk
5471 @item %|@var{suffix}
5472 @itemx %m@var{suffix}
5473 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
5474 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
5475 all. These are the two most common ways to instruct a program that it
5476 should read from standard input or write to standard output. If you
5477 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
5478 construct: see for example @file{f/lang-specs.h}.
5480 @item %.@var{SUFFIX}
5481 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
5482 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
5483 terminated by the next space or %.
5486 Marks the argument containing or following the @samp{%w} as the
5487 designated output file of this compilation. This puts the argument
5488 into the sequence of arguments that @samp{%o} will substitute later.
5491 Substitutes the names of all the output files, with spaces
5492 automatically placed around them. You should write spaces
5493 around the @samp{%o} as well or the results are undefined.
5494 @samp{%o} is for use in the specs for running the linker.
5495 Input files whose names have no recognized suffix are not compiled
5496 at all, but they are included among the output files, so they will
5500 Substitutes the suffix for object files. Note that this is
5501 handled specially when it immediately follows @samp{%g, %u, or %U},
5502 because of the need for those to form complete file names. The
5503 handling is such that @samp{%O} is treated exactly as if it had already
5504 been substituted, except that @samp{%g, %u, and %U} do not currently
5505 support additional @var{suffix} characters following @samp{%O} as they would
5506 following, for example, @samp{.o}.
5509 Substitutes the standard macro predefinitions for the
5510 current target machine. Use this when running @code{cpp}.
5513 Like @samp{%p}, but puts @samp{__} before and after the name of each
5514 predefined macro, except for macros that start with @samp{__} or with
5515 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
5519 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
5520 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
5521 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
5525 Current argument is the name of a library or startup file of some sort.
5526 Search for that file in a standard list of directories and substitute
5527 the full name found.
5530 Print @var{str} as an error message. @var{str} is terminated by a newline.
5531 Use this when inconsistent options are detected.
5534 Substitute the contents of spec string @var{name} at this point.
5537 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
5539 @item %x@{@var{option}@}
5540 Accumulate an option for @samp{%X}.
5543 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
5547 Output the accumulated assembler options specified by @option{-Wa}.
5550 Output the accumulated preprocessor options specified by @option{-Wp}.
5553 Process the @code{asm} spec. This is used to compute the
5554 switches to be passed to the assembler.
5557 Process the @code{asm_final} spec. This is a spec string for
5558 passing switches to an assembler post-processor, if such a program is
5562 Process the @code{link} spec. This is the spec for computing the
5563 command line passed to the linker. Typically it will make use of the
5564 @samp{%L %G %S %D and %E} sequences.
5567 Dump out a @option{-L} option for each directory that GCC believes might
5568 contain startup files. If the target supports multilibs then the
5569 current multilib directory will be prepended to each of these paths.
5572 Output the multilib directory with directory separators replaced with
5573 @samp{_}. If multilib directories are not set, or the multilib directory is
5574 @file{.} then this option emits nothing.
5577 Process the @code{lib} spec. This is a spec string for deciding which
5578 libraries should be included on the command line to the linker.
5581 Process the @code{libgcc} spec. This is a spec string for deciding
5582 which GCC support library should be included on the command line to the linker.
5585 Process the @code{startfile} spec. This is a spec for deciding which
5586 object files should be the first ones passed to the linker. Typically
5587 this might be a file named @file{crt0.o}.
5590 Process the @code{endfile} spec. This is a spec string that specifies
5591 the last object files that will be passed to the linker.
5594 Process the @code{cpp} spec. This is used to construct the arguments
5595 to be passed to the C preprocessor.
5598 Process the @code{signed_char} spec. This is intended to be used
5599 to tell cpp whether a char is signed. It typically has the definition:
5601 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
5605 Process the @code{cc1} spec. This is used to construct the options to be
5606 passed to the actual C compiler (@samp{cc1}).
5609 Process the @code{cc1plus} spec. This is used to construct the options to be
5610 passed to the actual C++ compiler (@samp{cc1plus}).
5613 Substitute the variable part of a matched option. See below.
5614 Note that each comma in the substituted string is replaced by
5618 Remove all occurrences of @code{-S} from the command line. Note---this
5619 command is position dependent. @samp{%} commands in the spec string
5620 before this one will see @code{-S}, @samp{%} commands in the spec string
5621 after this one will not.
5623 @item %:@var{function}(@var{args})
5624 Call the named function @var{function}, passing it @var{args}.
5625 @var{args} is first processed as a nested spec string, then split
5626 into an argument vector in the usual fashion. The function returns
5627 a string which is processed as if it had appeared literally as part
5628 of the current spec.
5630 The following built-in spec functions are provided:
5633 @item @code{if-exists}
5634 The @code{if-exists} spec function takes one argument, an absolute
5635 pathname to a file. If the file exists, @code{if-exists} returns the
5636 pathname. Here is a small example of its usage:
5640 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
5643 @item @code{if-exists-else}
5644 The @code{if-exists-else} spec function is similar to the @code{if-exists}
5645 spec function, except that it takes two arguments. The first argument is
5646 an absolute pathname to a file. If the file exists, @code{if-exists-else}
5647 returns the pathname. If it does not exist, it returns the second argument.
5648 This way, @code{if-exists-else} can be used to select one file or another,
5649 based on the existence of the first. Here is a small example of its usage:
5653 crt0%O%s %:if-exists(crti%O%s) \
5654 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
5659 Substitutes the @code{-S} switch, if that switch was given to GCC@.
5660 If that switch was not specified, this substitutes nothing. Note that
5661 the leading dash is omitted when specifying this option, and it is
5662 automatically inserted if the substitution is performed. Thus the spec
5663 string @samp{%@{foo@}} would match the command-line option @option{-foo}
5664 and would output the command line option @option{-foo}.
5666 @item %W@{@code{S}@}
5667 Like %@{@code{S}@} but mark last argument supplied within as a file to be
5670 @item %@{@code{S}*@}
5671 Substitutes all the switches specified to GCC whose names start
5672 with @code{-S}, but which also take an argument. This is used for
5673 switches like @option{-o}, @option{-D}, @option{-I}, etc.
5674 GCC considers @option{-o foo} as being
5675 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
5676 text, including the space. Thus two arguments would be generated.
5678 @item %@{@code{S}*&@code{T}*@}
5679 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
5680 (the order of @code{S} and @code{T} in the spec is not significant).
5681 There can be any number of ampersand-separated variables; for each the
5682 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
5684 @item %@{@code{S}:@code{X}@}
5685 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
5687 @item %@{!@code{S}:@code{X}@}
5688 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
5690 @item %@{@code{S}*:@code{X}@}
5691 Substitutes @code{X} if one or more switches whose names start with
5692 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
5693 once, no matter how many such switches appeared. However, if @code{%*}
5694 appears somewhere in @code{X}, then @code{X} will be substituted once
5695 for each matching switch, with the @code{%*} replaced by the part of
5696 that switch that matched the @code{*}.
5698 @item %@{.@code{S}:@code{X}@}
5699 Substitutes @code{X}, if processing a file with suffix @code{S}.
5701 @item %@{!.@code{S}:@code{X}@}
5702 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
5704 @item %@{@code{S}|@code{P}:@code{X}@}
5705 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
5706 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
5707 although they have a stronger binding than the @samp{|}. If @code{%*}
5708 appears in @code{X}, all of the alternatives must be starred, and only
5709 the first matching alternative is substituted.
5711 For example, a spec string like this:
5714 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
5717 will output the following command-line options from the following input
5718 command-line options:
5723 -d fred.c -foo -baz -boggle
5724 -d jim.d -bar -baz -boggle
5727 @item %@{S:X; T:Y; :D@}
5729 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
5730 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
5731 be as many clauses as you need. This may be combined with @code{.},
5732 @code{!}, @code{|}, and @code{*} as needed.
5737 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
5738 construct may contain other nested @samp{%} constructs or spaces, or
5739 even newlines. They are processed as usual, as described above.
5740 Trailing white space in @code{X} is ignored. White space may also
5741 appear anywhere on the left side of the colon in these constructs,
5742 except between @code{.} or @code{*} and the corresponding word.
5744 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
5745 handled specifically in these constructs. If another value of
5746 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
5747 @option{-W} switch is found later in the command line, the earlier
5748 switch value is ignored, except with @{@code{S}*@} where @code{S} is
5749 just one letter, which passes all matching options.
5751 The character @samp{|} at the beginning of the predicate text is used to
5752 indicate that a command should be piped to the following command, but
5753 only if @option{-pipe} is specified.
5755 It is built into GCC which switches take arguments and which do not.
5756 (You might think it would be useful to generalize this to allow each
5757 compiler's spec to say which switches take arguments. But this cannot
5758 be done in a consistent fashion. GCC cannot even decide which input
5759 files have been specified without knowing which switches take arguments,
5760 and it must know which input files to compile in order to tell which
5763 GCC also knows implicitly that arguments starting in @option{-l} are to be
5764 treated as compiler output files, and passed to the linker in their
5765 proper position among the other output files.
5767 @c man begin OPTIONS
5769 @node Target Options
5770 @section Specifying Target Machine and Compiler Version
5771 @cindex target options
5772 @cindex cross compiling
5773 @cindex specifying machine version
5774 @cindex specifying compiler version and target machine
5775 @cindex compiler version, specifying
5776 @cindex target machine, specifying
5778 The usual way to run GCC is to run the executable called @file{gcc}, or
5779 @file{<machine>-gcc} when cross-compiling, or
5780 @file{<machine>-gcc-<version>} to run a version other than the one that
5781 was installed last. Sometimes this is inconvenient, so GCC provides
5782 options that will switch to another cross-compiler or version.
5785 @item -b @var{machine}
5787 The argument @var{machine} specifies the target machine for compilation.
5789 The value to use for @var{machine} is the same as was specified as the
5790 machine type when configuring GCC as a cross-compiler. For
5791 example, if a cross-compiler was configured with @samp{configure
5792 i386v}, meaning to compile for an 80386 running System V, then you
5793 would specify @option{-b i386v} to run that cross compiler.
5795 @item -V @var{version}
5797 The argument @var{version} specifies which version of GCC to run.
5798 This is useful when multiple versions are installed. For example,
5799 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
5802 The @option{-V} and @option{-b} options work by running the
5803 @file{<machine>-gcc-<version>} executable, so there's no real reason to
5804 use them if you can just run that directly.
5806 @node Submodel Options
5807 @section Hardware Models and Configurations
5808 @cindex submodel options
5809 @cindex specifying hardware config
5810 @cindex hardware models and configurations, specifying
5811 @cindex machine dependent options
5813 Earlier we discussed the standard option @option{-b} which chooses among
5814 different installed compilers for completely different target
5815 machines, such as VAX vs.@: 68000 vs.@: 80386.
5817 In addition, each of these target machine types can have its own
5818 special options, starting with @samp{-m}, to choose among various
5819 hardware models or configurations---for example, 68010 vs 68020,
5820 floating coprocessor or none. A single installed version of the
5821 compiler can compile for any model or configuration, according to the
5824 Some configurations of the compiler also support additional special
5825 options, usually for compatibility with other compilers on the same
5828 These options are defined by the macro @code{TARGET_SWITCHES} in the
5829 machine description. The default for the options is also defined by
5830 that macro, which enables you to change the defaults.
5842 * RS/6000 and PowerPC Options::
5846 * i386 and x86-64 Options::
5848 * Intel 960 Options::
5849 * DEC Alpha Options::
5850 * DEC Alpha/VMS Options::
5853 * System V Options::
5854 * TMS320C3x/C4x Options::
5862 * S/390 and zSeries Options::
5866 * Xstormy16 Options::
5871 @node M680x0 Options
5872 @subsection M680x0 Options
5873 @cindex M680x0 options
5875 These are the @samp{-m} options defined for the 68000 series. The default
5876 values for these options depends on which style of 68000 was selected when
5877 the compiler was configured; the defaults for the most common choices are
5885 Generate output for a 68000. This is the default
5886 when the compiler is configured for 68000-based systems.
5888 Use this option for microcontrollers with a 68000 or EC000 core,
5889 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
5895 Generate output for a 68020. This is the default
5896 when the compiler is configured for 68020-based systems.
5900 Generate output containing 68881 instructions for floating point.
5901 This is the default for most 68020 systems unless @option{--nfp} was
5902 specified when the compiler was configured.
5906 Generate output for a 68030. This is the default when the compiler is
5907 configured for 68030-based systems.
5911 Generate output for a 68040. This is the default when the compiler is
5912 configured for 68040-based systems.
5914 This option inhibits the use of 68881/68882 instructions that have to be
5915 emulated by software on the 68040. Use this option if your 68040 does not
5916 have code to emulate those instructions.
5920 Generate output for a 68060. This is the default when the compiler is
5921 configured for 68060-based systems.
5923 This option inhibits the use of 68020 and 68881/68882 instructions that
5924 have to be emulated by software on the 68060. Use this option if your 68060
5925 does not have code to emulate those instructions.
5929 Generate output for a CPU32. This is the default
5930 when the compiler is configured for CPU32-based systems.
5932 Use this option for microcontrollers with a
5933 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
5934 68336, 68340, 68341, 68349 and 68360.
5938 Generate output for a 520X ``coldfire'' family cpu. This is the default
5939 when the compiler is configured for 520X-based systems.
5941 Use this option for microcontroller with a 5200 core, including
5942 the MCF5202, MCF5203, MCF5204 and MCF5202.
5947 Generate output for a 68040, without using any of the new instructions.
5948 This results in code which can run relatively efficiently on either a
5949 68020/68881 or a 68030 or a 68040. The generated code does use the
5950 68881 instructions that are emulated on the 68040.
5954 Generate output for a 68060, without using any of the new instructions.
5955 This results in code which can run relatively efficiently on either a
5956 68020/68881 or a 68030 or a 68040. The generated code does use the
5957 68881 instructions that are emulated on the 68060.
5960 @opindex msoft-float
5961 Generate output containing library calls for floating point.
5962 @strong{Warning:} the requisite libraries are not available for all m68k
5963 targets. Normally the facilities of the machine's usual C compiler are
5964 used, but this can't be done directly in cross-compilation. You must
5965 make your own arrangements to provide suitable library functions for
5966 cross-compilation. The embedded targets @samp{m68k-*-aout} and
5967 @samp{m68k-*-coff} do provide software floating point support.
5971 Consider type @code{int} to be 16 bits wide, like @code{short int}.
5974 @opindex mnobitfield
5975 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
5976 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
5980 Do use the bit-field instructions. The @option{-m68020} option implies
5981 @option{-mbitfield}. This is the default if you use a configuration
5982 designed for a 68020.
5986 Use a different function-calling convention, in which functions
5987 that take a fixed number of arguments return with the @code{rtd}
5988 instruction, which pops their arguments while returning. This
5989 saves one instruction in the caller since there is no need to pop
5990 the arguments there.
5992 This calling convention is incompatible with the one normally
5993 used on Unix, so you cannot use it if you need to call libraries
5994 compiled with the Unix compiler.
5996 Also, you must provide function prototypes for all functions that
5997 take variable numbers of arguments (including @code{printf});
5998 otherwise incorrect code will be generated for calls to those
6001 In addition, seriously incorrect code will result if you call a
6002 function with too many arguments. (Normally, extra arguments are
6003 harmlessly ignored.)
6005 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
6006 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
6009 @itemx -mno-align-int
6011 @opindex mno-align-int
6012 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
6013 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
6014 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
6015 Aligning variables on 32-bit boundaries produces code that runs somewhat
6016 faster on processors with 32-bit busses at the expense of more memory.
6018 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
6019 align structures containing the above types differently than
6020 most published application binary interface specifications for the m68k.
6024 Use the pc-relative addressing mode of the 68000 directly, instead of
6025 using a global offset table. At present, this option implies @option{-fpic},
6026 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
6027 not presently supported with @option{-mpcrel}, though this could be supported for
6028 68020 and higher processors.
6030 @item -mno-strict-align
6031 @itemx -mstrict-align
6032 @opindex mno-strict-align
6033 @opindex mstrict-align
6034 Do not (do) assume that unaligned memory references will be handled by
6038 Generate code that allows the data segment to be located in a different
6039 area of memory from the text segment. This allows for execute in place in
6040 an environment without virtual memory management. This option implies -fPIC.
6043 Generate code that assumes that the data segment follows the text segment.
6044 This is the default.
6046 @item -mid-shared-library
6047 Generate code that supports shared libraries via the library ID method.
6048 This allows for execute in place and shared libraries in an environment
6049 without virtual memory management. This option implies -fPIC.
6051 @item -mno-id-shared-library
6052 Generate code that doesn't assume ID based shared libraries are being used.
6053 This is the default.
6055 @item -mshared-library-id=n
6056 Specified the identification number of the ID based shared library being
6057 compiled. Specifying a value of 0 will generate more compact code, specifying
6058 other values will force the allocation of that number to the current
6059 library but is no more space or time efficient than omitting this option.
6063 @node M68hc1x Options
6064 @subsection M68hc1x Options
6065 @cindex M68hc1x options
6067 These are the @samp{-m} options defined for the 68hc11 and 68hc12
6068 microcontrollers. The default values for these options depends on
6069 which style of microcontroller was selected when the compiler was configured;
6070 the defaults for the most common choices are given below.
6077 Generate output for a 68HC11. This is the default
6078 when the compiler is configured for 68HC11-based systems.
6084 Generate output for a 68HC12. This is the default
6085 when the compiler is configured for 68HC12-based systems.
6091 Generate output for a 68HCS12.
6094 @opindex mauto-incdec
6095 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
6102 Enable the use of 68HC12 min and max instructions.
6105 @itemx -mno-long-calls
6106 @opindex mlong-calls
6107 @opindex mno-long-calls
6108 Treat all calls as being far away (near). If calls are assumed to be
6109 far away, the compiler will use the @code{call} instruction to
6110 call a function and the @code{rtc} instruction for returning.
6114 Consider type @code{int} to be 16 bits wide, like @code{short int}.
6116 @item -msoft-reg-count=@var{count}
6117 @opindex msoft-reg-count
6118 Specify the number of pseudo-soft registers which are used for the
6119 code generation. The maximum number is 32. Using more pseudo-soft
6120 register may or may not result in better code depending on the program.
6121 The default is 4 for 68HC11 and 2 for 68HC12.
6126 @subsection VAX Options
6129 These @samp{-m} options are defined for the VAX:
6134 Do not output certain jump instructions (@code{aobleq} and so on)
6135 that the Unix assembler for the VAX cannot handle across long
6140 Do output those jump instructions, on the assumption that you
6141 will assemble with the GNU assembler.
6145 Output code for g-format floating point numbers instead of d-format.
6149 @subsection SPARC Options
6150 @cindex SPARC options
6152 These @samp{-m} switches are supported on the SPARC:
6157 @opindex mno-app-regs
6159 Specify @option{-mapp-regs} to generate output using the global registers
6160 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
6163 To be fully SVR4 ABI compliant at the cost of some performance loss,
6164 specify @option{-mno-app-regs}. You should compile libraries and system
6165 software with this option.
6170 @opindex mhard-float
6171 Generate output containing floating point instructions. This is the
6177 @opindex msoft-float
6178 Generate output containing library calls for floating point.
6179 @strong{Warning:} the requisite libraries are not available for all SPARC
6180 targets. Normally the facilities of the machine's usual C compiler are
6181 used, but this cannot be done directly in cross-compilation. You must make
6182 your own arrangements to provide suitable library functions for
6183 cross-compilation. The embedded targets @samp{sparc-*-aout} and
6184 @samp{sparclite-*-*} do provide software floating point support.
6186 @option{-msoft-float} changes the calling convention in the output file;
6187 therefore, it is only useful if you compile @emph{all} of a program with
6188 this option. In particular, you need to compile @file{libgcc.a}, the
6189 library that comes with GCC, with @option{-msoft-float} in order for
6192 @item -mhard-quad-float
6193 @opindex mhard-quad-float
6194 Generate output containing quad-word (long double) floating point
6197 @item -msoft-quad-float
6198 @opindex msoft-quad-float
6199 Generate output containing library calls for quad-word (long double)
6200 floating point instructions. The functions called are those specified
6201 in the SPARC ABI@. This is the default.
6203 As of this writing, there are no sparc implementations that have hardware
6204 support for the quad-word floating point instructions. They all invoke
6205 a trap handler for one of these instructions, and then the trap handler
6206 emulates the effect of the instruction. Because of the trap handler overhead,
6207 this is much slower than calling the ABI library routines. Thus the
6208 @option{-msoft-quad-float} option is the default.
6214 With @option{-mflat}, the compiler does not generate save/restore instructions
6215 and will use a ``flat'' or single register window calling convention.
6216 This model uses %i7 as the frame pointer and is compatible with the normal
6217 register window model. Code from either may be intermixed.
6218 The local registers and the input registers (0--5) are still treated as
6219 ``call saved'' registers and will be saved on the stack as necessary.
6221 With @option{-mno-flat} (the default), the compiler emits save/restore
6222 instructions (except for leaf functions) and is the normal mode of operation.
6224 @item -mno-unaligned-doubles
6225 @itemx -munaligned-doubles
6226 @opindex mno-unaligned-doubles
6227 @opindex munaligned-doubles
6228 Assume that doubles have 8 byte alignment. This is the default.
6230 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
6231 alignment only if they are contained in another type, or if they have an
6232 absolute address. Otherwise, it assumes they have 4 byte alignment.
6233 Specifying this option avoids some rare compatibility problems with code
6234 generated by other compilers. It is not the default because it results
6235 in a performance loss, especially for floating point code.
6237 @item -mno-faster-structs
6238 @itemx -mfaster-structs
6239 @opindex mno-faster-structs
6240 @opindex mfaster-structs
6241 With @option{-mfaster-structs}, the compiler assumes that structures
6242 should have 8 byte alignment. This enables the use of pairs of
6243 @code{ldd} and @code{std} instructions for copies in structure
6244 assignment, in place of twice as many @code{ld} and @code{st} pairs.
6245 However, the use of this changed alignment directly violates the SPARC
6246 ABI@. Thus, it's intended only for use on targets where the developer
6247 acknowledges that their resulting code will not be directly in line with
6248 the rules of the ABI@.
6251 @opindex mimpure-text
6252 @option{-mimpure-text}, used in addition to @option{-shared}, tells
6253 the compiler to not pass @option{-z text} to the linker when linking a
6254 shared object. Using this option, you can link position-dependent
6255 code into a shared object.
6257 @option{-mimpure-text} suppresses the ``relocations remain against
6258 allocatable but non-writable sections'' linker error message.
6259 However, the necessary relocations will trigger copy-on-write, and the
6260 shared object is not actually shared across processes. Instead of
6261 using @option{-mimpure-text}, you should compile all source code with
6262 @option{-fpic} or @option{-fPIC}.
6264 This option is only available on SunOS and Solaris.
6270 These two options select variations on the SPARC architecture.
6272 By default (unless specifically configured for the Fujitsu SPARClite),
6273 GCC generates code for the v7 variant of the SPARC architecture.
6275 @option{-mv8} will give you SPARC v8 code. The only difference from v7
6276 code is that the compiler emits the integer multiply and integer
6277 divide instructions which exist in SPARC v8 but not in SPARC v7.
6279 @option{-msparclite} will give you SPARClite code. This adds the integer
6280 multiply, integer divide step and scan (@code{ffs}) instructions which
6281 exist in SPARClite but not in SPARC v7.
6283 These options are deprecated and will be deleted in a future GCC release.
6284 They have been replaced with @option{-mcpu=xxx}.
6289 @opindex msupersparc
6290 These two options select the processor for which the code is optimized.
6292 With @option{-mcypress} (the default), the compiler optimizes code for the
6293 Cypress CY7C602 chip, as used in the SPARCStation/SPARCServer 3xx series.
6294 This is also appropriate for the older SPARCStation 1, 2, IPX etc.
6296 With @option{-msupersparc} the compiler optimizes code for the SuperSPARC cpu, as
6297 used in the SPARCStation 10, 1000 and 2000 series. This flag also enables use
6298 of the full SPARC v8 instruction set.
6300 These options are deprecated and will be deleted in a future GCC release.
6301 They have been replaced with @option{-mcpu=xxx}.
6303 @item -mcpu=@var{cpu_type}
6305 Set the instruction set, register set, and instruction scheduling parameters
6306 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
6307 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
6308 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
6309 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
6312 Default instruction scheduling parameters are used for values that select
6313 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
6314 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
6316 Here is a list of each supported architecture and their supported
6321 v8: supersparc, hypersparc
6322 sparclite: f930, f934, sparclite86x
6324 v9: ultrasparc, ultrasparc3
6327 @item -mtune=@var{cpu_type}
6329 Set the instruction scheduling parameters for machine type
6330 @var{cpu_type}, but do not set the instruction set or register set that the
6331 option @option{-mcpu=@var{cpu_type}} would.
6333 The same values for @option{-mcpu=@var{cpu_type}} can be used for
6334 @option{-mtune=@var{cpu_type}}, but the only useful values are those
6335 that select a particular cpu implementation. Those are @samp{cypress},
6336 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
6337 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
6342 These @samp{-m} switches are supported in addition to the above
6343 on the SPARCLET processor.
6346 @item -mlittle-endian
6347 @opindex mlittle-endian
6348 Generate code for a processor running in little-endian mode.
6352 Treat register @code{%g0} as a normal register.
6353 GCC will continue to clobber it as necessary but will not assume
6354 it always reads as 0.
6356 @item -mbroken-saverestore
6357 @opindex mbroken-saverestore
6358 Generate code that does not use non-trivial forms of the @code{save} and
6359 @code{restore} instructions. Early versions of the SPARCLET processor do
6360 not correctly handle @code{save} and @code{restore} instructions used with
6361 arguments. They correctly handle them used without arguments. A @code{save}
6362 instruction used without arguments increments the current window pointer
6363 but does not allocate a new stack frame. It is assumed that the window
6364 overflow trap handler will properly handle this case as will interrupt
6368 These @samp{-m} switches are supported in addition to the above
6369 on SPARC V9 processors in 64-bit environments.
6372 @item -mlittle-endian
6373 @opindex mlittle-endian
6374 Generate code for a processor running in little-endian mode. It is only
6375 available for a few configurations and most notably not on Solaris.
6381 Generate code for a 32-bit or 64-bit environment.
6382 The 32-bit environment sets int, long and pointer to 32 bits.
6383 The 64-bit environment sets int to 32 bits and long and pointer
6386 @item -mcmodel=medlow
6387 @opindex mcmodel=medlow
6388 Generate code for the Medium/Low code model: the program must be linked
6389 in the low 32 bits of the address space. Pointers are 64 bits.
6390 Programs can be statically or dynamically linked.
6392 @item -mcmodel=medmid
6393 @opindex mcmodel=medmid
6394 Generate code for the Medium/Middle code model: the program must be linked
6395 in the low 44 bits of the address space, the text segment must be less than
6396 2G bytes, and data segment must be within 2G of the text segment.
6397 Pointers are 64 bits.
6399 @item -mcmodel=medany
6400 @opindex mcmodel=medany
6401 Generate code for the Medium/Anywhere code model: the program may be linked
6402 anywhere in the address space, the text segment must be less than
6403 2G bytes, and data segment must be within 2G of the text segment.
6404 Pointers are 64 bits.
6406 @item -mcmodel=embmedany
6407 @opindex mcmodel=embmedany
6408 Generate code for the Medium/Anywhere code model for embedded systems:
6409 assume a 32-bit text and a 32-bit data segment, both starting anywhere
6410 (determined at link time). Register %g4 points to the base of the
6411 data segment. Pointers are still 64 bits.
6412 Programs are statically linked, PIC is not supported.
6415 @itemx -mno-stack-bias
6416 @opindex mstack-bias
6417 @opindex mno-stack-bias
6418 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
6419 frame pointer if present, are offset by @minus{}2047 which must be added back
6420 when making stack frame references.
6421 Otherwise, assume no such offset is present.
6425 @subsection ARM Options
6428 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6433 @opindex mapcs-frame
6434 Generate a stack frame that is compliant with the ARM Procedure Call
6435 Standard for all functions, even if this is not strictly necessary for
6436 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6437 with this option will cause the stack frames not to be generated for
6438 leaf functions. The default is @option{-mno-apcs-frame}.
6442 This is a synonym for @option{-mapcs-frame}.
6446 Generate code for a processor running with a 26-bit program counter,
6447 and conforming to the function calling standards for the APCS 26-bit
6448 option. This option replaces the @option{-m2} and @option{-m3} options
6449 of previous releases of the compiler.
6453 Generate code for a processor running with a 32-bit program counter,
6454 and conforming to the function calling standards for the APCS 32-bit
6455 option. This option replaces the @option{-m6} option of previous releases
6459 @c not currently implemented
6460 @item -mapcs-stack-check
6461 @opindex mapcs-stack-check
6462 Generate code to check the amount of stack space available upon entry to
6463 every function (that actually uses some stack space). If there is
6464 insufficient space available then either the function
6465 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6466 called, depending upon the amount of stack space required. The run time
6467 system is required to provide these functions. The default is
6468 @option{-mno-apcs-stack-check}, since this produces smaller code.
6470 @c not currently implemented
6472 @opindex mapcs-float
6473 Pass floating point arguments using the float point registers. This is
6474 one of the variants of the APCS@. This option is recommended if the
6475 target hardware has a floating point unit or if a lot of floating point
6476 arithmetic is going to be performed by the code. The default is
6477 @option{-mno-apcs-float}, since integer only code is slightly increased in
6478 size if @option{-mapcs-float} is used.
6480 @c not currently implemented
6481 @item -mapcs-reentrant
6482 @opindex mapcs-reentrant
6483 Generate reentrant, position independent code. The default is
6484 @option{-mno-apcs-reentrant}.
6487 @item -mthumb-interwork
6488 @opindex mthumb-interwork
6489 Generate code which supports calling between the ARM and Thumb
6490 instruction sets. Without this option the two instruction sets cannot
6491 be reliably used inside one program. The default is
6492 @option{-mno-thumb-interwork}, since slightly larger code is generated
6493 when @option{-mthumb-interwork} is specified.
6495 @item -mno-sched-prolog
6496 @opindex mno-sched-prolog
6497 Prevent the reordering of instructions in the function prolog, or the
6498 merging of those instruction with the instructions in the function's
6499 body. This means that all functions will start with a recognizable set
6500 of instructions (or in fact one of a choice from a small set of
6501 different function prologues), and this information can be used to
6502 locate the start if functions inside an executable piece of code. The
6503 default is @option{-msched-prolog}.
6506 @opindex mhard-float
6507 Generate output containing floating point instructions. This is the
6511 @opindex msoft-float
6512 Generate output containing library calls for floating point.
6513 @strong{Warning:} the requisite libraries are not available for all ARM
6514 targets. Normally the facilities of the machine's usual C compiler are
6515 used, but this cannot be done directly in cross-compilation. You must make
6516 your own arrangements to provide suitable library functions for
6519 @option{-msoft-float} changes the calling convention in the output file;
6520 therefore, it is only useful if you compile @emph{all} of a program with
6521 this option. In particular, you need to compile @file{libgcc.a}, the
6522 library that comes with GCC, with @option{-msoft-float} in order for
6525 @item -mlittle-endian
6526 @opindex mlittle-endian
6527 Generate code for a processor running in little-endian mode. This is
6528 the default for all standard configurations.
6531 @opindex mbig-endian
6532 Generate code for a processor running in big-endian mode; the default is
6533 to compile code for a little-endian processor.
6535 @item -mwords-little-endian
6536 @opindex mwords-little-endian
6537 This option only applies when generating code for big-endian processors.
6538 Generate code for a little-endian word order but a big-endian byte
6539 order. That is, a byte order of the form @samp{32107654}. Note: this
6540 option should only be used if you require compatibility with code for
6541 big-endian ARM processors generated by versions of the compiler prior to
6544 @item -malignment-traps
6545 @opindex malignment-traps
6546 Generate code that will not trap if the MMU has alignment traps enabled.
6547 On ARM architectures prior to ARMv4, there were no instructions to
6548 access half-word objects stored in memory. However, when reading from
6549 memory a feature of the ARM architecture allows a word load to be used,
6550 even if the address is unaligned, and the processor core will rotate the
6551 data as it is being loaded. This option tells the compiler that such
6552 misaligned accesses will cause a MMU trap and that it should instead
6553 synthesize the access as a series of byte accesses. The compiler can
6554 still use word accesses to load half-word data if it knows that the
6555 address is aligned to a word boundary.
6557 This option is ignored when compiling for ARM architecture 4 or later,
6558 since these processors have instructions to directly access half-word
6561 @item -mno-alignment-traps
6562 @opindex mno-alignment-traps
6563 Generate code that assumes that the MMU will not trap unaligned
6564 accesses. This produces better code when the target instruction set
6565 does not have half-word memory operations (i.e.@: implementations prior to
6568 Note that you cannot use this option to access unaligned word objects,
6569 since the processor will only fetch one 32-bit aligned object from
6572 The default setting for most targets is @option{-mno-alignment-traps}, since
6573 this produces better code when there are no half-word memory
6574 instructions available.
6576 @item -mshort-load-bytes
6577 @itemx -mno-short-load-words
6578 @opindex mshort-load-bytes
6579 @opindex mno-short-load-words
6580 These are deprecated aliases for @option{-malignment-traps}.
6582 @item -mno-short-load-bytes
6583 @itemx -mshort-load-words
6584 @opindex mno-short-load-bytes
6585 @opindex mshort-load-words
6586 This are deprecated aliases for @option{-mno-alignment-traps}.
6588 @item -mcpu=@var{name}
6590 This specifies the name of the target ARM processor. GCC uses this name
6591 to determine what kind of instructions it can emit when generating
6592 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6593 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6594 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6595 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6596 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6597 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm8},
6598 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6599 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6600 @samp{arm920t}, @samp{arm926ejs}, @samp{arm940t}, @samp{arm9tdmi},
6601 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ejs},
6602 @samp{arm1136js}, @samp{arm1136jfs} ,@samp{xscale}, @samp{iwmmxt},
6605 @itemx -mtune=@var{name}
6607 This option is very similar to the @option{-mcpu=} option, except that
6608 instead of specifying the actual target processor type, and hence
6609 restricting which instructions can be used, it specifies that GCC should
6610 tune the performance of the code as if the target were of the type
6611 specified in this option, but still choosing the instructions that it
6612 will generate based on the cpu specified by a @option{-mcpu=} option.
6613 For some ARM implementations better performance can be obtained by using
6616 @item -march=@var{name}
6618 This specifies the name of the target ARM architecture. GCC uses this
6619 name to determine what kind of instructions it can emit when generating
6620 assembly code. This option can be used in conjunction with or instead
6621 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6622 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6623 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6j},
6624 @samp{iwmmxt}, @samp{ep9312}.
6626 @item -mfpe=@var{number}
6627 @itemx -mfp=@var{number}
6630 This specifies the version of the floating point emulation available on
6631 the target. Permissible values are 2 and 3. @option{-mfp=} is a synonym
6632 for @option{-mfpe=}, for compatibility with older versions of GCC@.
6634 @item -mstructure-size-boundary=@var{n}
6635 @opindex mstructure-size-boundary
6636 The size of all structures and unions will be rounded up to a multiple
6637 of the number of bits set by this option. Permissible values are 8 and
6638 32. The default value varies for different toolchains. For the COFF
6639 targeted toolchain the default value is 8. Specifying the larger number
6640 can produce faster, more efficient code, but can also increase the size
6641 of the program. The two values are potentially incompatible. Code
6642 compiled with one value cannot necessarily expect to work with code or
6643 libraries compiled with the other value, if they exchange information
6644 using structures or unions.
6646 @item -mabort-on-noreturn
6647 @opindex mabort-on-noreturn
6648 Generate a call to the function @code{abort} at the end of a
6649 @code{noreturn} function. It will be executed if the function tries to
6653 @itemx -mno-long-calls
6654 @opindex mlong-calls
6655 @opindex mno-long-calls
6656 Tells the compiler to perform function calls by first loading the
6657 address of the function into a register and then performing a subroutine
6658 call on this register. This switch is needed if the target function
6659 will lie outside of the 64 megabyte addressing range of the offset based
6660 version of subroutine call instruction.
6662 Even if this switch is enabled, not all function calls will be turned
6663 into long calls. The heuristic is that static functions, functions
6664 which have the @samp{short-call} attribute, functions that are inside
6665 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6666 definitions have already been compiled within the current compilation
6667 unit, will not be turned into long calls. The exception to this rule is
6668 that weak function definitions, functions with the @samp{long-call}
6669 attribute or the @samp{section} attribute, and functions that are within
6670 the scope of a @samp{#pragma long_calls} directive, will always be
6671 turned into long calls.
6673 This feature is not enabled by default. Specifying
6674 @option{-mno-long-calls} will restore the default behavior, as will
6675 placing the function calls within the scope of a @samp{#pragma
6676 long_calls_off} directive. Note these switches have no effect on how
6677 the compiler generates code to handle function calls via function
6680 @item -mnop-fun-dllimport
6681 @opindex mnop-fun-dllimport
6682 Disable support for the @code{dllimport} attribute.
6684 @item -msingle-pic-base
6685 @opindex msingle-pic-base
6686 Treat the register used for PIC addressing as read-only, rather than
6687 loading it in the prologue for each function. The run-time system is
6688 responsible for initializing this register with an appropriate value
6689 before execution begins.
6691 @item -mpic-register=@var{reg}
6692 @opindex mpic-register
6693 Specify the register to be used for PIC addressing. The default is R10
6694 unless stack-checking is enabled, when R9 is used.
6696 @item -mcirrus-fix-invalid-insns
6697 @opindex mcirrus-fix-invalid-insns
6698 @opindex mno-cirrus-fix-invalid-insns
6699 Insert NOPs into the instruction stream to in order to work around
6700 problems with invalid Maverick instruction combinations. This option
6701 is only valid if the @option{-mcpu=ep9312} option has been used to
6702 enable generation of instructions for the Cirrus Maverick floating
6703 point co-processor. This option is not enabled by default, since the
6704 problem is only present in older Maverick implementations. The default
6705 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
6708 @item -mpoke-function-name
6709 @opindex mpoke-function-name
6710 Write the name of each function into the text section, directly
6711 preceding the function prologue. The generated code is similar to this:
6715 .ascii "arm_poke_function_name", 0
6718 .word 0xff000000 + (t1 - t0)
6719 arm_poke_function_name
6721 stmfd sp!, @{fp, ip, lr, pc@}
6725 When performing a stack backtrace, code can inspect the value of
6726 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6727 location @code{pc - 12} and the top 8 bits are set, then we know that
6728 there is a function name embedded immediately preceding this location
6729 and has length @code{((pc[-3]) & 0xff000000)}.
6733 Generate code for the 16-bit Thumb instruction set. The default is to
6734 use the 32-bit ARM instruction set.
6737 @opindex mtpcs-frame
6738 Generate a stack frame that is compliant with the Thumb Procedure Call
6739 Standard for all non-leaf functions. (A leaf function is one that does
6740 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6742 @item -mtpcs-leaf-frame
6743 @opindex mtpcs-leaf-frame
6744 Generate a stack frame that is compliant with the Thumb Procedure Call
6745 Standard for all leaf functions. (A leaf function is one that does
6746 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6748 @item -mcallee-super-interworking
6749 @opindex mcallee-super-interworking
6750 Gives all externally visible functions in the file being compiled an ARM
6751 instruction set header which switches to Thumb mode before executing the
6752 rest of the function. This allows these functions to be called from
6753 non-interworking code.
6755 @item -mcaller-super-interworking
6756 @opindex mcaller-super-interworking
6757 Allows calls via function pointers (including virtual functions) to
6758 execute correctly regardless of whether the target code has been
6759 compiled for interworking or not. There is a small overhead in the cost
6760 of executing a function pointer if this option is enabled.
6764 @node MN10200 Options
6765 @subsection MN10200 Options
6766 @cindex MN10200 options
6768 These @option{-m} options are defined for Matsushita MN10200 architectures:
6773 Indicate to the linker that it should perform a relaxation optimization pass
6774 to shorten branches, calls and absolute memory addresses. This option only
6775 has an effect when used on the command line for the final link step.
6777 This option makes symbolic debugging impossible.
6780 @node MN10300 Options
6781 @subsection MN10300 Options
6782 @cindex MN10300 options
6784 These @option{-m} options are defined for Matsushita MN10300 architectures:
6789 Generate code to avoid bugs in the multiply instructions for the MN10300
6790 processors. This is the default.
6793 @opindex mno-mult-bug
6794 Do not generate code to avoid bugs in the multiply instructions for the
6799 Generate code which uses features specific to the AM33 processor.
6803 Do not generate code which uses features specific to the AM33 processor. This
6808 Do not link in the C run-time initialization object file.
6812 Indicate to the linker that it should perform a relaxation optimization pass
6813 to shorten branches, calls and absolute memory addresses. This option only
6814 has an effect when used on the command line for the final link step.
6816 This option makes symbolic debugging impossible.
6820 @node M32R/D Options
6821 @subsection M32R/D Options
6822 @cindex M32R/D options
6824 These @option{-m} options are defined for Renesas M32R/D architectures:
6829 Generate code for the M32R/2@.
6833 Generate code for the M32R/X@.
6837 Generate code for the M32R@. This is the default.
6839 @item -mcode-model=small
6840 @opindex mcode-model=small
6841 Assume all objects live in the lower 16MB of memory (so that their addresses
6842 can be loaded with the @code{ld24} instruction), and assume all subroutines
6843 are reachable with the @code{bl} instruction.
6844 This is the default.
6846 The addressability of a particular object can be set with the
6847 @code{model} attribute.
6849 @item -mcode-model=medium
6850 @opindex mcode-model=medium
6851 Assume objects may be anywhere in the 32-bit address space (the compiler
6852 will generate @code{seth/add3} instructions to load their addresses), and
6853 assume all subroutines are reachable with the @code{bl} instruction.
6855 @item -mcode-model=large
6856 @opindex mcode-model=large
6857 Assume objects may be anywhere in the 32-bit address space (the compiler
6858 will generate @code{seth/add3} instructions to load their addresses), and
6859 assume subroutines may not be reachable with the @code{bl} instruction
6860 (the compiler will generate the much slower @code{seth/add3/jl}
6861 instruction sequence).
6864 @opindex msdata=none
6865 Disable use of the small data area. Variables will be put into
6866 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
6867 @code{section} attribute has been specified).
6868 This is the default.
6870 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
6871 Objects may be explicitly put in the small data area with the
6872 @code{section} attribute using one of these sections.
6875 @opindex msdata=sdata
6876 Put small global and static data in the small data area, but do not
6877 generate special code to reference them.
6881 Put small global and static data in the small data area, and generate
6882 special instructions to reference them.
6886 @cindex smaller data references
6887 Put global and static objects less than or equal to @var{num} bytes
6888 into the small data or bss sections instead of the normal data or bss
6889 sections. The default value of @var{num} is 8.
6890 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
6891 for this option to have any effect.
6893 All modules should be compiled with the same @option{-G @var{num}} value.
6894 Compiling with different values of @var{num} may or may not work; if it
6895 doesn't the linker will give an error message---incorrect code will not be
6901 @subsection M88K Options
6902 @cindex M88k options
6904 These @samp{-m} options are defined for Motorola 88k architectures:
6909 Generate code that works well on both the m88100 and the
6914 Generate code that works best for the m88100, but that also
6919 Generate code that works best for the m88110, and may not run
6924 Obsolete option to be removed from the next revision.
6927 @item -midentify-revision
6928 @opindex midentify-revision
6929 @cindex identifying source, compiler (88k)
6930 Include an @code{ident} directive in the assembler output recording the
6931 source file name, compiler name and version, timestamp, and compilation
6934 @item -mno-underscores
6935 @opindex mno-underscores
6936 @cindex underscores, avoiding (88k)
6937 In assembler output, emit symbol names without adding an underscore
6938 character at the beginning of each name. The default is to use an
6939 underscore as prefix on each name.
6941 @item -mocs-debug-info
6942 @itemx -mno-ocs-debug-info
6943 @opindex mocs-debug-info
6944 @opindex mno-ocs-debug-info
6946 @cindex debugging, 88k OCS
6947 Include (or omit) additional debugging information (about registers used
6948 in each stack frame) as specified in the 88open Object Compatibility
6949 Standard, ``OCS''@. This extra information allows debugging of code that
6950 has had the frame pointer eliminated. The default for SVr4 and Delta 88
6951 SVr3.2 is to include this information; other 88k configurations omit this
6952 information by default.
6954 @item -mocs-frame-position
6955 @opindex mocs-frame-position
6956 @cindex register positions in frame (88k)
6957 When emitting COFF debugging information for automatic variables and
6958 parameters stored on the stack, use the offset from the canonical frame
6959 address, which is the stack pointer (register 31) on entry to the
6960 function. The SVr4 and Delta88 SVr3.2, and BCS configurations use
6961 @option{-mocs-frame-position}; other 88k configurations have the default
6962 @option{-mno-ocs-frame-position}.
6964 @item -mno-ocs-frame-position
6965 @opindex mno-ocs-frame-position
6966 @cindex register positions in frame (88k)
6967 When emitting COFF debugging information for automatic variables and
6968 parameters stored on the stack, use the offset from the frame pointer
6969 register (register 30). When this option is in effect, the frame
6970 pointer is not eliminated when debugging information is selected by the
6973 @item -moptimize-arg-area
6974 @opindex moptimize-arg-area
6975 @cindex arguments in frame (88k)
6976 Save space by reorganizing the stack frame. This option generates code
6977 that does not agree with the 88open specifications, but uses less
6980 @itemx -mno-optimize-arg-area
6981 @opindex mno-optimize-arg-area
6982 Do not reorganize the stack frame to save space. This is the default.
6983 The generated conforms to the specification, but uses more memory.
6985 @item -mshort-data-@var{num}
6986 @opindex mshort-data
6987 @cindex smaller data references (88k)
6988 @cindex r0-relative references (88k)
6989 Generate smaller data references by making them relative to @code{r0},
6990 which allows loading a value using a single instruction (rather than the
6991 usual two). You control which data references are affected by
6992 specifying @var{num} with this option. For example, if you specify
6993 @option{-mshort-data-512}, then the data references affected are those
6994 involving displacements of less than 512 bytes.
6995 @option{-mshort-data-@var{num}} is not effective for @var{num} greater
6998 @item -mserialize-volatile
6999 @opindex mserialize-volatile
7000 @itemx -mno-serialize-volatile
7001 @opindex mno-serialize-volatile
7002 @cindex sequential consistency on 88k
7003 Do, or don't, generate code to guarantee sequential consistency
7004 of volatile memory references. By default, consistency is
7007 The order of memory references made by the MC88110 processor does
7008 not always match the order of the instructions requesting those
7009 references. In particular, a load instruction may execute before
7010 a preceding store instruction. Such reordering violates
7011 sequential consistency of volatile memory references, when there
7012 are multiple processors. When consistency must be guaranteed,
7013 GCC generates special instructions, as needed, to force
7014 execution in the proper order.
7016 The MC88100 processor does not reorder memory references and so
7017 always provides sequential consistency. However, by default, GCC
7018 generates the special instructions to guarantee consistency
7019 even when you use @option{-m88100}, so that the code may be run on an
7020 MC88110 processor. If you intend to run your code only on the
7021 MC88100 processor, you may use @option{-mno-serialize-volatile}.
7023 The extra code generated to guarantee consistency may affect the
7024 performance of your application. If you know that you can safely
7025 forgo this guarantee, you may use @option{-mno-serialize-volatile}.
7031 @cindex assembler syntax, 88k
7033 Turn on (@option{-msvr4}) or off (@option{-msvr3}) compiler extensions
7034 related to System V release 4 (SVr4). This controls the following:
7038 Which variant of the assembler syntax to emit.
7040 @option{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
7041 that is used on System V release 4.
7043 @option{-msvr4} makes GCC issue additional declaration directives used in
7047 @option{-msvr4} is the default for the m88k-motorola-sysv4 configuration.
7048 @option{-msvr3} is the default for all other m88k configurations.
7050 @item -mversion-03.00
7051 @opindex mversion-03.00
7052 This option is obsolete, and is ignored.
7053 @c ??? which asm syntax better for GAS? option there too?
7055 @item -mno-check-zero-division
7056 @itemx -mcheck-zero-division
7057 @opindex mno-check-zero-division
7058 @opindex mcheck-zero-division
7059 @cindex zero division on 88k
7060 Do, or don't, generate code to guarantee that integer division by
7061 zero will be detected. By default, detection is guaranteed.
7063 Some models of the MC88100 processor fail to trap upon integer
7064 division by zero under certain conditions. By default, when
7065 compiling code that might be run on such a processor, GCC
7066 generates code that explicitly checks for zero-valued divisors
7067 and traps with exception number 503 when one is detected. Use of
7068 @option{-mno-check-zero-division} suppresses such checking for code
7069 generated to run on an MC88100 processor.
7071 GCC assumes that the MC88110 processor correctly detects all instances
7072 of integer division by zero. When @option{-m88110} is specified, no
7073 explicit checks for zero-valued divisors are generated, and both
7074 @option{-mcheck-zero-division} and @option{-mno-check-zero-division} are
7077 @item -muse-div-instruction
7078 @opindex muse-div-instruction
7079 @cindex divide instruction, 88k
7080 Use the div instruction for signed integer division on the
7081 MC88100 processor. By default, the div instruction is not used.
7083 On the MC88100 processor the signed integer division instruction
7084 div) traps to the operating system on a negative operand. The
7085 operating system transparently completes the operation, but at a
7086 large cost in execution time. By default, when compiling code
7087 that might be run on an MC88100 processor, GCC emulates signed
7088 integer division using the unsigned integer division instruction
7089 divu), thereby avoiding the large penalty of a trap to the
7090 operating system. Such emulation has its own, smaller, execution
7091 cost in both time and space. To the extent that your code's
7092 important signed integer division operations are performed on two
7093 nonnegative operands, it may be desirable to use the div
7094 instruction directly.
7096 On the MC88110 processor the div instruction (also known as the
7097 divs instruction) processes negative operands without trapping to
7098 the operating system. When @option{-m88110} is specified,
7099 @option{-muse-div-instruction} is ignored, and the div instruction is used
7100 for signed integer division.
7102 Note that the result of dividing @code{INT_MIN} by @minus{}1 is undefined. In
7103 particular, the behavior of such a division with and without
7104 @option{-muse-div-instruction} may differ.
7106 @item -mtrap-large-shift
7107 @itemx -mhandle-large-shift
7108 @opindex mtrap-large-shift
7109 @opindex mhandle-large-shift
7110 @cindex bit shift overflow (88k)
7111 @cindex large bit shifts (88k)
7112 Include code to detect bit-shifts of more than 31 bits; respectively,
7113 trap such shifts or emit code to handle them properly. By default GCC
7114 makes no special provision for large bit shifts.
7116 @item -mwarn-passed-structs
7117 @opindex mwarn-passed-structs
7118 @cindex structure passing (88k)
7119 Warn when a function passes a struct as an argument or result.
7120 Structure-passing conventions have changed during the evolution of the C
7121 language, and are often the source of portability problems. By default,
7122 GCC issues no such warning.
7125 @c break page here to avoid unsightly interparagraph stretch.
7129 @node RS/6000 and PowerPC Options
7130 @subsection IBM RS/6000 and PowerPC Options
7131 @cindex RS/6000 and PowerPC Options
7132 @cindex IBM RS/6000 and PowerPC Options
7134 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
7142 @itemx -mpowerpc-gpopt
7143 @itemx -mno-powerpc-gpopt
7144 @itemx -mpowerpc-gfxopt
7145 @itemx -mno-powerpc-gfxopt
7147 @itemx -mno-powerpc64
7153 @opindex mno-powerpc
7154 @opindex mpowerpc-gpopt
7155 @opindex mno-powerpc-gpopt
7156 @opindex mpowerpc-gfxopt
7157 @opindex mno-powerpc-gfxopt
7159 @opindex mno-powerpc64
7160 GCC supports two related instruction set architectures for the
7161 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
7162 instructions supported by the @samp{rios} chip set used in the original
7163 RS/6000 systems and the @dfn{PowerPC} instruction set is the
7164 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
7165 the IBM 4xx microprocessors.
7167 Neither architecture is a subset of the other. However there is a
7168 large common subset of instructions supported by both. An MQ
7169 register is included in processors supporting the POWER architecture.
7171 You use these options to specify which instructions are available on the
7172 processor you are using. The default value of these options is
7173 determined when configuring GCC@. Specifying the
7174 @option{-mcpu=@var{cpu_type}} overrides the specification of these
7175 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
7176 rather than the options listed above.
7178 The @option{-mpower} option allows GCC to generate instructions that
7179 are found only in the POWER architecture and to use the MQ register.
7180 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
7181 to generate instructions that are present in the POWER2 architecture but
7182 not the original POWER architecture.
7184 The @option{-mpowerpc} option allows GCC to generate instructions that
7185 are found only in the 32-bit subset of the PowerPC architecture.
7186 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
7187 GCC to use the optional PowerPC architecture instructions in the
7188 General Purpose group, including floating-point square root. Specifying
7189 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
7190 use the optional PowerPC architecture instructions in the Graphics
7191 group, including floating-point select.
7193 The @option{-mpowerpc64} option allows GCC to generate the additional
7194 64-bit instructions that are found in the full PowerPC64 architecture
7195 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
7196 @option{-mno-powerpc64}.
7198 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
7199 will use only the instructions in the common subset of both
7200 architectures plus some special AIX common-mode calls, and will not use
7201 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
7202 permits GCC to use any instruction from either architecture and to
7203 allow use of the MQ register; specify this for the Motorola MPC601.
7205 @item -mnew-mnemonics
7206 @itemx -mold-mnemonics
7207 @opindex mnew-mnemonics
7208 @opindex mold-mnemonics
7209 Select which mnemonics to use in the generated assembler code. With
7210 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
7211 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
7212 assembler mnemonics defined for the POWER architecture. Instructions
7213 defined in only one architecture have only one mnemonic; GCC uses that
7214 mnemonic irrespective of which of these options is specified.
7216 GCC defaults to the mnemonics appropriate for the architecture in
7217 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
7218 value of these option. Unless you are building a cross-compiler, you
7219 should normally not specify either @option{-mnew-mnemonics} or
7220 @option{-mold-mnemonics}, but should instead accept the default.
7222 @item -mcpu=@var{cpu_type}
7224 Set architecture type, register usage, choice of mnemonics, and
7225 instruction scheduling parameters for machine type @var{cpu_type}.
7226 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
7227 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
7228 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
7229 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
7230 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
7231 @samp{860}, @samp{970}, @samp{common}, @samp{ec603e}, @samp{G3},
7232 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
7233 @samp{power4}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
7234 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
7236 @option{-mcpu=common} selects a completely generic processor. Code
7237 generated under this option will run on any POWER or PowerPC processor.
7238 GCC will use only the instructions in the common subset of both
7239 architectures, and will not use the MQ register. GCC assumes a generic
7240 processor model for scheduling purposes.
7242 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
7243 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
7244 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
7245 types, with an appropriate, generic processor model assumed for
7246 scheduling purposes.
7248 The other options specify a specific processor. Code generated under
7249 those options will run best on that processor, and may not run at all on
7252 The @option{-mcpu} options automatically enable or disable the
7253 following options: @option{-maltivec}, @option{-mhard-float},
7254 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
7255 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
7256 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
7257 @option{-mstring}. The particular options set for any particular CPU
7258 will vary between compiler versions, depending on what setting seems
7259 to produce optimal code for that CPU; it doesn't necessarily reflect
7260 the actual hardware's capabilities. If you wish to set an individual
7261 option to a particular value, you may specify it after the
7262 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
7264 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
7265 not enabled or disabled by the @option{-mcpu} option at present, since
7266 AIX does not have full support for these options. You may still
7267 enable or disable them individually if you're sure it'll work in your
7270 @item -mtune=@var{cpu_type}
7272 Set the instruction scheduling parameters for machine type
7273 @var{cpu_type}, but do not set the architecture type, register usage, or
7274 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
7275 values for @var{cpu_type} are used for @option{-mtune} as for
7276 @option{-mcpu}. If both are specified, the code generated will use the
7277 architecture, registers, and mnemonics set by @option{-mcpu}, but the
7278 scheduling parameters set by @option{-mtune}.
7283 @opindex mno-altivec
7284 These switches enable or disable the use of built-in functions that
7285 allow access to the AltiVec instruction set. You may also need to set
7286 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
7291 Extend the current ABI with SPE ABI extensions. This does not change
7292 the default ABI, instead it adds the SPE ABI extensions to the current
7296 @opindex mabi=no-spe
7297 Disable Booke SPE ABI extensions for the current ABI.
7299 @item -misel=@var{yes/no}
7302 This switch enables or disables the generation of ISEL instructions.
7304 @item -mspe=@var{yes/no}
7307 This switch enables or disables the generation of SPE simd
7310 @item -mfloat-gprs=@var{yes/no}
7312 @opindex mfloat-gprs
7313 This switch enables or disables the generation of floating point
7314 operations on the general purpose registers for architectures that
7315 support it. This option is currently only available on the MPC8540.
7318 @itemx -mno-fp-in-toc
7319 @itemx -mno-sum-in-toc
7320 @itemx -mminimal-toc
7322 @opindex mno-fp-in-toc
7323 @opindex mno-sum-in-toc
7324 @opindex mminimal-toc
7325 Modify generation of the TOC (Table Of Contents), which is created for
7326 every executable file. The @option{-mfull-toc} option is selected by
7327 default. In that case, GCC will allocate at least one TOC entry for
7328 each unique non-automatic variable reference in your program. GCC
7329 will also place floating-point constants in the TOC@. However, only
7330 16,384 entries are available in the TOC@.
7332 If you receive a linker error message that saying you have overflowed
7333 the available TOC space, you can reduce the amount of TOC space used
7334 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
7335 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
7336 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
7337 generate code to calculate the sum of an address and a constant at
7338 run-time instead of putting that sum into the TOC@. You may specify one
7339 or both of these options. Each causes GCC to produce very slightly
7340 slower and larger code at the expense of conserving TOC space.
7342 If you still run out of space in the TOC even when you specify both of
7343 these options, specify @option{-mminimal-toc} instead. This option causes
7344 GCC to make only one TOC entry for every file. When you specify this
7345 option, GCC will produce code that is slower and larger but which
7346 uses extremely little TOC space. You may wish to use this option
7347 only on files that contain less frequently executed code.
7353 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
7354 @code{long} type, and the infrastructure needed to support them.
7355 Specifying @option{-maix64} implies @option{-mpowerpc64} and
7356 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
7357 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
7362 @opindex mno-xl-call
7363 On AIX, pass floating-point arguments to prototyped functions beyond the
7364 register save area (RSA) on the stack in addition to argument FPRs. The
7365 AIX calling convention was extended but not initially documented to
7366 handle an obscure K&R C case of calling a function that takes the
7367 address of its arguments with fewer arguments than declared. AIX XL
7368 compilers access floating point arguments which do not fit in the
7369 RSA from the stack when a subroutine is compiled without
7370 optimization. Because always storing floating-point arguments on the
7371 stack is inefficient and rarely needed, this option is not enabled by
7372 default and only is necessary when calling subroutines compiled by AIX
7373 XL compilers without optimization.
7377 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
7378 application written to use message passing with special startup code to
7379 enable the application to run. The system must have PE installed in the
7380 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
7381 must be overridden with the @option{-specs=} option to specify the
7382 appropriate directory location. The Parallel Environment does not
7383 support threads, so the @option{-mpe} option and the @option{-pthread}
7384 option are incompatible.
7386 @item -malign-natural
7387 @itemx -malign-power
7388 @opindex malign-natural
7389 @opindex malign-power
7390 On AIX, Darwin, and 64-bit PowerPC Linux, the option
7391 @option{-malign-natural} overrides the ABI-defined alignment of larger
7392 types, such as floating-point doubles, on their natural size-based boundary.
7393 The option @option{-malign-power} instructs GCC to follow the ABI-specified
7394 alignment rules. GCC defaults to the standard alignment defined in the ABI.
7398 @opindex msoft-float
7399 @opindex mhard-float
7400 Generate code that does not use (uses) the floating-point register set.
7401 Software floating point emulation is provided if you use the
7402 @option{-msoft-float} option, and pass the option to GCC when linking.
7405 @itemx -mno-multiple
7407 @opindex mno-multiple
7408 Generate code that uses (does not use) the load multiple word
7409 instructions and the store multiple word instructions. These
7410 instructions are generated by default on POWER systems, and not
7411 generated on PowerPC systems. Do not use @option{-mmultiple} on little
7412 endian PowerPC systems, since those instructions do not work when the
7413 processor is in little endian mode. The exceptions are PPC740 and
7414 PPC750 which permit the instructions usage in little endian mode.
7420 Generate code that uses (does not use) the load string instructions
7421 and the store string word instructions to save multiple registers and
7422 do small block moves. These instructions are generated by default on
7423 POWER systems, and not generated on PowerPC systems. Do not use
7424 @option{-mstring} on little endian PowerPC systems, since those
7425 instructions do not work when the processor is in little endian mode.
7426 The exceptions are PPC740 and PPC750 which permit the instructions
7427 usage in little endian mode.
7433 Generate code that uses (does not use) the load or store instructions
7434 that update the base register to the address of the calculated memory
7435 location. These instructions are generated by default. If you use
7436 @option{-mno-update}, there is a small window between the time that the
7437 stack pointer is updated and the address of the previous frame is
7438 stored, which means code that walks the stack frame across interrupts or
7439 signals may get corrupted data.
7442 @itemx -mno-fused-madd
7443 @opindex mfused-madd
7444 @opindex mno-fused-madd
7445 Generate code that uses (does not use) the floating point multiply and
7446 accumulate instructions. These instructions are generated by default if
7447 hardware floating is used.
7449 @item -mno-bit-align
7451 @opindex mno-bit-align
7453 On System V.4 and embedded PowerPC systems do not (do) force structures
7454 and unions that contain bit-fields to be aligned to the base type of the
7457 For example, by default a structure containing nothing but 8
7458 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
7459 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
7460 the structure would be aligned to a 1 byte boundary and be one byte in
7463 @item -mno-strict-align
7464 @itemx -mstrict-align
7465 @opindex mno-strict-align
7466 @opindex mstrict-align
7467 On System V.4 and embedded PowerPC systems do not (do) assume that
7468 unaligned memory references will be handled by the system.
7471 @itemx -mno-relocatable
7472 @opindex mrelocatable
7473 @opindex mno-relocatable
7474 On embedded PowerPC systems generate code that allows (does not allow)
7475 the program to be relocated to a different address at runtime. If you
7476 use @option{-mrelocatable} on any module, all objects linked together must
7477 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
7479 @item -mrelocatable-lib
7480 @itemx -mno-relocatable-lib
7481 @opindex mrelocatable-lib
7482 @opindex mno-relocatable-lib
7483 On embedded PowerPC systems generate code that allows (does not allow)
7484 the program to be relocated to a different address at runtime. Modules
7485 compiled with @option{-mrelocatable-lib} can be linked with either modules
7486 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
7487 with modules compiled with the @option{-mrelocatable} options.
7493 On System V.4 and embedded PowerPC systems do not (do) assume that
7494 register 2 contains a pointer to a global area pointing to the addresses
7495 used in the program.
7498 @itemx -mlittle-endian
7500 @opindex mlittle-endian
7501 On System V.4 and embedded PowerPC systems compile code for the
7502 processor in little endian mode. The @option{-mlittle-endian} option is
7503 the same as @option{-mlittle}.
7508 @opindex mbig-endian
7509 On System V.4 and embedded PowerPC systems compile code for the
7510 processor in big endian mode. The @option{-mbig-endian} option is
7511 the same as @option{-mbig}.
7513 @item -mdynamic-no-pic
7514 @opindex mdynamic-no-pic
7515 On Darwin and Mac OS X systems, compile code so that it is not
7516 relocatable, but that its external references are relocatable. The
7517 resulting code is suitable for applications, but not shared
7520 @item -mprioritize-restricted-insns=@var{priority}
7521 @opindex mprioritize-restricted-insns
7522 This option controls the priority that is assigned to
7523 dispatch-slot restricted instructions during the second scheduling
7524 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
7525 @var{no/highest/second-highest} priority to dispatch slot restricted
7528 @item -msched-costly-dep=@var{dependence_type}
7529 @opindex msched-costly-dep
7530 This option controls which dependences are considered costly
7531 by the target during instruction scheduling. The argument
7532 @var{dependence_type} takes one of the following values:
7533 @var{no}: no dependence is costly,
7534 @var{all}: all dependences are costly,
7535 @var{true_store_to_load}: a true dependence from store to load is costly,
7536 @var{store_to_load}: any dependence from store to load is costly,
7537 @var{number}: any dependence which latency >= @var{number} is costly.
7539 @item -minsert-sched-nops=@var{scheme}
7540 @opindex minsert-sched-nops
7541 This option controls which nop insertion scheme will be used during
7542 the second scheduling pass. The argument @var{scheme} takes one of the
7544 @var{no}: Don't insert nops.
7545 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
7546 according to the scheduler's grouping.
7547 @var{regroup_exact}: Insert nops to force costly dependent insns into
7548 separate groups. Insert exactly as many nops as needed to force an insn
7549 to a new group, according to the estimatied processor grouping.
7550 @var{number}: Insert nops to force costly dependent insns into
7551 separate groups. Insert @var{number} nops to force an insn to a new group.
7555 On System V.4 and embedded PowerPC systems compile code using calling
7556 conventions that adheres to the March 1995 draft of the System V
7557 Application Binary Interface, PowerPC processor supplement. This is the
7558 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
7560 @item -mcall-sysv-eabi
7561 @opindex mcall-sysv-eabi
7562 Specify both @option{-mcall-sysv} and @option{-meabi} options.
7564 @item -mcall-sysv-noeabi
7565 @opindex mcall-sysv-noeabi
7566 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
7568 @item -mcall-solaris
7569 @opindex mcall-solaris
7570 On System V.4 and embedded PowerPC systems compile code for the Solaris
7574 @opindex mcall-linux
7575 On System V.4 and embedded PowerPC systems compile code for the
7576 Linux-based GNU system.
7580 On System V.4 and embedded PowerPC systems compile code for the
7581 Hurd-based GNU system.
7584 @opindex mcall-netbsd
7585 On System V.4 and embedded PowerPC systems compile code for the
7586 NetBSD operating system.
7588 @item -maix-struct-return
7589 @opindex maix-struct-return
7590 Return all structures in memory (as specified by the AIX ABI)@.
7592 @item -msvr4-struct-return
7593 @opindex msvr4-struct-return
7594 Return structures smaller than 8 bytes in registers (as specified by the
7598 @opindex mabi=altivec
7599 Extend the current ABI with AltiVec ABI extensions. This does not
7600 change the default ABI, instead it adds the AltiVec ABI extensions to
7603 @item -mabi=no-altivec
7604 @opindex mabi=no-altivec
7605 Disable AltiVec ABI extensions for the current ABI.
7608 @itemx -mno-prototype
7610 @opindex mno-prototype
7611 On System V.4 and embedded PowerPC systems assume that all calls to
7612 variable argument functions are properly prototyped. Otherwise, the
7613 compiler must insert an instruction before every non prototyped call to
7614 set or clear bit 6 of the condition code register (@var{CR}) to
7615 indicate whether floating point values were passed in the floating point
7616 registers in case the function takes a variable arguments. With
7617 @option{-mprototype}, only calls to prototyped variable argument functions
7618 will set or clear the bit.
7622 On embedded PowerPC systems, assume that the startup module is called
7623 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
7624 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
7629 On embedded PowerPC systems, assume that the startup module is called
7630 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
7635 On embedded PowerPC systems, assume that the startup module is called
7636 @file{crt0.o} and the standard C libraries are @file{libads.a} and
7640 @opindex myellowknife
7641 On embedded PowerPC systems, assume that the startup module is called
7642 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
7647 On System V.4 and embedded PowerPC systems, specify that you are
7648 compiling for a VxWorks system.
7652 Specify that you are compiling for the WindISS simulation environment.
7656 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
7657 header to indicate that @samp{eabi} extended relocations are used.
7663 On System V.4 and embedded PowerPC systems do (do not) adhere to the
7664 Embedded Applications Binary Interface (eabi) which is a set of
7665 modifications to the System V.4 specifications. Selecting @option{-meabi}
7666 means that the stack is aligned to an 8 byte boundary, a function
7667 @code{__eabi} is called to from @code{main} to set up the eabi
7668 environment, and the @option{-msdata} option can use both @code{r2} and
7669 @code{r13} to point to two separate small data areas. Selecting
7670 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
7671 do not call an initialization function from @code{main}, and the
7672 @option{-msdata} option will only use @code{r13} to point to a single
7673 small data area. The @option{-meabi} option is on by default if you
7674 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
7677 @opindex msdata=eabi
7678 On System V.4 and embedded PowerPC systems, put small initialized
7679 @code{const} global and static data in the @samp{.sdata2} section, which
7680 is pointed to by register @code{r2}. Put small initialized
7681 non-@code{const} global and static data in the @samp{.sdata} section,
7682 which is pointed to by register @code{r13}. Put small uninitialized
7683 global and static data in the @samp{.sbss} section, which is adjacent to
7684 the @samp{.sdata} section. The @option{-msdata=eabi} option is
7685 incompatible with the @option{-mrelocatable} option. The
7686 @option{-msdata=eabi} option also sets the @option{-memb} option.
7689 @opindex msdata=sysv
7690 On System V.4 and embedded PowerPC systems, put small global and static
7691 data in the @samp{.sdata} section, which is pointed to by register
7692 @code{r13}. Put small uninitialized global and static data in the
7693 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
7694 The @option{-msdata=sysv} option is incompatible with the
7695 @option{-mrelocatable} option.
7697 @item -msdata=default
7699 @opindex msdata=default
7701 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
7702 compile code the same as @option{-msdata=eabi}, otherwise compile code the
7703 same as @option{-msdata=sysv}.
7706 @opindex msdata-data
7707 On System V.4 and embedded PowerPC systems, put small global and static
7708 data in the @samp{.sdata} section. Put small uninitialized global and
7709 static data in the @samp{.sbss} section. Do not use register @code{r13}
7710 to address small data however. This is the default behavior unless
7711 other @option{-msdata} options are used.
7715 @opindex msdata=none
7717 On embedded PowerPC systems, put all initialized global and static data
7718 in the @samp{.data} section, and all uninitialized data in the
7719 @samp{.bss} section.
7723 @cindex smaller data references (PowerPC)
7724 @cindex .sdata/.sdata2 references (PowerPC)
7725 On embedded PowerPC systems, put global and static items less than or
7726 equal to @var{num} bytes into the small data or bss sections instead of
7727 the normal data or bss section. By default, @var{num} is 8. The
7728 @option{-G @var{num}} switch is also passed to the linker.
7729 All modules should be compiled with the same @option{-G @var{num}} value.
7732 @itemx -mno-regnames
7734 @opindex mno-regnames
7735 On System V.4 and embedded PowerPC systems do (do not) emit register
7736 names in the assembly language output using symbolic forms.
7739 @itemx -mno-longcall
7741 @opindex mno-longcall
7742 Default to making all function calls via pointers, so that functions
7743 which reside further than 64 megabytes (67,108,864 bytes) from the
7744 current location can be called. This setting can be overridden by the
7745 @code{shortcall} function attribute, or by @code{#pragma longcall(0)}.
7747 Some linkers are capable of detecting out-of-range calls and generating
7748 glue code on the fly. On these systems, long calls are unnecessary and
7749 generate slower code. As of this writing, the AIX linker can do this,
7750 as can the GNU linker for PowerPC/64. It is planned to add this feature
7751 to the GNU linker for 32-bit PowerPC systems as well.
7753 On Mach-O (Darwin) systems, this option directs the compiler emit to
7754 the glue for every direct call, and the Darwin linker decides whether
7755 to use or discard it.
7757 In the future, we may cause GCC to ignore all longcall specifications
7758 when the linker is known to generate glue.
7762 Adds support for multithreading with the @dfn{pthreads} library.
7763 This option sets flags for both the preprocessor and linker.
7767 @node Darwin Options
7768 @subsection Darwin Options
7769 @cindex Darwin options
7771 These options are defined for all architectures running the Darwin operating
7772 system. They are useful for compatibility with other Mac OS compilers.
7777 Loads all members of static archive libraries.
7778 See man ld(1) for more information.
7780 @item -arch_errors_fatal
7781 @opindex arch_errors_fatal
7782 Cause the errors having to do with files that have the wrong architecture
7786 @opindex bind_at_load
7787 Causes the output file to be marked such that the dynamic linker will
7788 bind all undefined references when the file is loaded or launched.
7792 Produce a Mach-o bundle format file.
7793 See man ld(1) for more information.
7795 @item -bundle_loader @var{executable}
7796 @opindex bundle_loader
7797 This specifies the @var{executable} that will be loading the build
7798 output file being linked. See man ld(1) for more information.
7800 @item -allowable_client @var{client_name}
7804 @item -compatibility_version
7805 @item -current_version
7806 @item -dependency-file
7808 @item -dylinker_install_name
7811 @item -exported_symbols_list
7813 @item -flat_namespace
7814 @item -force_cpusubtype_ALL
7815 @item -force_flat_namespace
7816 @item -headerpad_max_install_names
7820 @item -keep_private_externs
7822 @item -multiply_defined
7823 @item -multiply_defined_unused
7825 @item -nofixprebinding
7828 @item -noseglinkedit
7829 @item -pagezero_size
7831 @item -prebind_all_twolevel_modules
7832 @item -private_bundle
7833 @item -read_only_relocs
7835 @item -sectobjectsymbols
7839 @item -sectobjectsymbols
7841 @item -seg_addr_table
7842 @item -seg_addr_table_filename
7845 @item -segs_read_only_addr
7846 @item -segs_read_write_addr
7847 @item -single_module
7851 @item -twolevel_namespace
7854 @item -unexported_symbols_list
7855 @item -weak_reference_mismatches
7858 @opindex allowable_client
7860 @opindex client_name
7861 @opindex compatibility_version
7862 @opindex current_version
7863 @opindex dependency-file
7865 @opindex dylinker_install_name
7868 @opindex exported_symbols_list
7870 @opindex flat_namespace
7871 @opindex force_cpusubtype_ALL
7872 @opindex force_flat_namespace
7873 @opindex headerpad_max_install_names
7876 @opindex install_name
7877 @opindex keep_private_externs
7878 @opindex multi_module
7879 @opindex multiply_defined
7880 @opindex multiply_defined_unused
7882 @opindex nofixprebinding
7883 @opindex nomultidefs
7885 @opindex noseglinkedit
7886 @opindex pagezero_size
7888 @opindex prebind_all_twolevel_modules
7889 @opindex private_bundle
7890 @opindex read_only_relocs
7892 @opindex sectobjectsymbols
7896 @opindex sectobjectsymbols
7898 @opindex seg_addr_table
7899 @opindex seg_addr_table_filename
7900 @opindex seglinkedit
7902 @opindex segs_read_only_addr
7903 @opindex segs_read_write_addr
7904 @opindex single_module
7906 @opindex sub_library
7907 @opindex sub_umbrella
7908 @opindex twolevel_namespace
7911 @opindex unexported_symbols_list
7912 @opindex weak_reference_mismatches
7913 @opindex whatsloaded
7915 This options are available for Darwin linker. Darwin linker man page
7916 describes them in detail.
7921 @subsection IBM RT Options
7923 @cindex IBM RT options
7925 These @samp{-m} options are defined for the IBM RT PC:
7929 @opindex min-line-mul
7930 Use an in-line code sequence for integer multiplies. This is the
7933 @item -mcall-lib-mul
7934 @opindex mcall-lib-mul
7935 Call @code{lmul$$} for integer multiples.
7937 @item -mfull-fp-blocks
7938 @opindex mfull-fp-blocks
7939 Generate full-size floating point data blocks, including the minimum
7940 amount of scratch space recommended by IBM@. This is the default.
7942 @item -mminimum-fp-blocks
7943 @opindex mminimum-fp-blocks
7944 Do not include extra scratch space in floating point data blocks. This
7945 results in smaller code, but slower execution, since scratch space must
7946 be allocated dynamically.
7948 @cindex @file{stdarg.h} and RT PC
7949 @item -mfp-arg-in-fpregs
7950 @opindex mfp-arg-in-fpregs
7951 Use a calling sequence incompatible with the IBM calling convention in
7952 which floating point arguments are passed in floating point registers.
7953 Note that @code{stdarg.h} will not work with floating point operands
7954 if this option is specified.
7956 @item -mfp-arg-in-gregs
7957 @opindex mfp-arg-in-gregs
7958 Use the normal calling convention for floating point arguments. This is
7961 @item -mhc-struct-return
7962 @opindex mhc-struct-return
7963 Return structures of more than one word in memory, rather than in a
7964 register. This provides compatibility with the MetaWare HighC (hc)
7965 compiler. Use the option @option{-fpcc-struct-return} for compatibility
7966 with the Portable C Compiler (pcc).
7968 @item -mnohc-struct-return
7969 @opindex mnohc-struct-return
7970 Return some structures of more than one word in registers, when
7971 convenient. This is the default. For compatibility with the
7972 IBM-supplied compilers, use the option @option{-fpcc-struct-return} or the
7973 option @option{-mhc-struct-return}.
7977 @subsection MIPS Options
7978 @cindex MIPS options
7980 These @samp{-m} options are defined for the MIPS family of computers:
7984 @item -march=@var{arch}
7986 Generate code that will run on @var{arch}, which can be the name of a
7987 generic MIPS ISA, or the name of a particular processor.
7989 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
7990 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
7991 The processor names are:
7992 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
7994 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
7995 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
7999 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4300},
8000 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
8001 The special value @samp{from-abi} selects the
8002 most compatible architecture for the selected ABI (that is,
8003 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
8005 In processor names, a final @samp{000} can be abbreviated as @samp{k}
8006 (for example, @samp{-march=r2k}). Prefixes are optional, and
8007 @samp{vr} may be written @samp{r}.
8009 GCC defines two macros based on the value of this option. The first
8010 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
8011 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
8012 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
8013 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
8014 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
8016 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
8017 above. In other words, it will have the full prefix and will not
8018 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
8019 the macro names the resolved architecture (either @samp{"mips1"} or
8020 @samp{"mips3"}). It names the default architecture when no
8021 @option{-march} option is given.
8023 @item -mtune=@var{arch}
8025 Optimize for @var{arch}. Among other things, this option controls
8026 the way instructions are scheduled, and the perceived cost of arithmetic
8027 operations. The list of @var{arch} values is the same as for
8030 When this option is not used, GCC will optimize for the processor
8031 specified by @option{-march}. By using @option{-march} and
8032 @option{-mtune} together, it is possible to generate code that will
8033 run on a family of processors, but optimize the code for one
8034 particular member of that family.
8036 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
8037 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
8038 @samp{-march} ones described above.
8042 Equivalent to @samp{-march=mips1}.
8046 Equivalent to @samp{-march=mips2}.
8050 Equivalent to @samp{-march=mips3}.
8054 Equivalent to @samp{-march=mips4}.
8058 Equivalent to @samp{-march=mips32}.
8062 Equivalent to @samp{-march=mips32r2}.
8066 Equivalent to @samp{-march=mips64}.
8069 @itemx -mno-fused-madd
8070 @opindex mfused-madd
8071 @opindex mno-fused-madd
8072 Generate code that uses (does not use) the floating point multiply and
8073 accumulate instructions, when they are available. These instructions
8074 are generated by default if they are available, but this may be
8075 undesirable if the extra precision causes problems or on certain chips
8076 in the mode where denormals are rounded to zero where denormals
8077 generated by multiply and accumulate instructions cause exceptions
8082 Assume that floating point registers are 32 bits wide.
8086 Assume that floating point registers are 64 bits wide.
8090 Assume that general purpose registers are 32 bits wide.
8094 Assume that general purpose registers are 64 bits wide.
8098 Force int and long types to be 64 bits wide. See @option{-mlong32} for an
8099 explanation of the default, and the width of pointers.
8103 Force long types to be 64 bits wide. See @option{-mlong32} for an
8104 explanation of the default, and the width of pointers.
8108 Force long, int, and pointer types to be 32 bits wide.
8110 The default size of ints, longs and pointers depends on the ABI@. All
8111 the supported ABIs use 32-bit ints. The n64 ABI uses 64-bit longs, as
8112 does the 64-bit Cygnus EABI; the others use 32-bit longs. Pointers
8113 are the same size as longs, or the same size as integer registers,
8114 whichever is smaller.
8128 Generate code for the given ABI@.
8130 Note that there are two embedded ABIs: @option{-mabi=eabi}
8131 selects the one defined by Cygnus while @option{-meabi=meabi}
8132 selects the one defined by MIPS@. Both these ABIs have
8133 32-bit and 64-bit variants. Normally, GCC will generate
8134 64-bit code when you select a 64-bit architecture, but you
8135 can use @option{-mgp32} to get 32-bit code instead.
8137 @item -mabi-fake-default
8138 @opindex mabi-fake-default
8139 You don't want to know what this option does. No, really. I mean
8140 it. Move on to the next option.
8142 What? You're still here? Oh, well@enddots{} Ok, here's the deal. GCC
8143 wants the default set of options to get the root of the multilib tree,
8144 and the shared library SONAMEs without any multilib-indicating
8145 suffixes. This is not convenience for @samp{mips64-linux-gnu}, since
8146 we want to default to the N32 ABI, while still being binary-compatible
8147 with @samp{mips-linux-gnu} if you stick to the O32 ABI@. Being
8148 binary-compatible means shared libraries should have the same SONAMEs,
8149 and libraries should live in the same location. Having O32 libraries
8150 in a sub-directory named say @file{o32} is not acceptable.
8152 So we trick GCC into believing that O32 is the default ABI, except
8153 that we override the default with some internal command-line
8154 processing magic. Problem is, if we stopped at that, and you then
8155 created a multilib-aware package that used the output of @command{gcc
8156 -print-multi-lib} to decide which multilibs to build, and how, and
8157 you'd find yourself in an awkward situation when you found out that
8158 some of the options listed ended up mapping to the same multilib, and
8159 none of your libraries was actually built for the multilib that
8160 @option{-print-multi-lib} claims to be the default. So we added this
8161 option that disables the default switcher, falling back to GCC's
8162 original notion of the default library. Confused yet?
8164 For short: don't ever use this option, unless you find it in the list
8165 of additional options to be used when building for multilibs, in the
8166 output of @option{gcc -print-multi-lib}.
8170 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
8171 add normal debug information. This is the default for all
8172 platforms except for the OSF/1 reference platform, using the OSF/rose
8173 object format. If the either of the @option{-gstabs} or @option{-gstabs+}
8174 switches are used, the @file{mips-tfile} program will encapsulate the
8175 stabs within MIPS ECOFF@.
8179 Generate code for the GNU assembler. This is the default on the OSF/1
8180 reference platform, using the OSF/rose object format. Also, this is
8181 the default if the configure option @option{--with-gnu-as} is used.
8183 @item -msplit-addresses
8184 @itemx -mno-split-addresses
8185 @opindex msplit-addresses
8186 @opindex mno-split-addresses
8187 Generate code to load the high and low parts of address constants separately.
8188 This allows GCC to optimize away redundant loads of the high order
8189 bits of addresses. This optimization requires GNU as and GNU ld.
8190 This optimization is enabled by default for some embedded targets where
8191 GNU as and GNU ld are standard.
8197 The @option{-mrnames} switch says to output code using the MIPS software
8198 names for the registers, instead of the hardware names (ie, @var{a0}
8199 instead of @var{$4}). The only known assembler that supports this option
8200 is the Algorithmics assembler.
8206 The @option{-mmemcpy} switch makes all block moves call the appropriate
8207 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
8208 generating inline code.
8211 @itemx -mno-mips-tfile
8212 @opindex mmips-tfile
8213 @opindex mno-mips-tfile
8214 The @option{-mno-mips-tfile} switch causes the compiler not
8215 postprocess the object file with the @file{mips-tfile} program,
8216 after the MIPS assembler has generated it to add debug support. If
8217 @file{mips-tfile} is not run, then no local variables will be
8218 available to the debugger. In addition, @file{stage2} and
8219 @file{stage3} objects will have the temporary file names passed to the
8220 assembler embedded in the object file, which means the objects will
8221 not compare the same. The @option{-mno-mips-tfile} switch should only
8222 be used when there are bugs in the @file{mips-tfile} program that
8223 prevents compilation.
8226 @opindex msoft-float
8227 Generate output containing library calls for floating point.
8228 @strong{Warning:} the requisite libraries are not part of GCC@.
8229 Normally the facilities of the machine's usual C compiler are used, but
8230 this can't be done directly in cross-compilation. You must make your
8231 own arrangements to provide suitable library functions for
8235 @opindex mhard-float
8236 Generate output containing floating point instructions. This is the
8237 default if you use the unmodified sources.
8240 @itemx -mno-abicalls
8242 @opindex mno-abicalls
8243 Emit (or do not emit) the pseudo operations @samp{.abicalls},
8244 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
8245 position independent code.
8251 Lift (or do not lift) the usual restrictions on the size of the global
8254 GCC normally uses a single instruction to load values from the GOT.
8255 While this is relatively efficient, it will only work if the GOT
8256 is smaller than about 64k. Anything larger will cause the linker
8257 to report an error such as:
8259 @cindex relocation truncated to fit (MIPS)
8261 relocation truncated to fit: R_MIPS_GOT16 foobar
8264 If this happens, you should recompile your code with @option{-mxgot}.
8265 It should then work with very large GOTs, although it will also be
8266 less efficient, since it will take three instructions to fetch the
8267 value of a global symbol.
8269 Note that some linkers can create multiple GOTs. If you have such a
8270 linker, you should only need to use @option{-mxgot} when a single object
8271 file accesses more than 64k's worth of GOT entries. Very few do.
8273 These options have no effect unless GCC is generating position
8277 @itemx -mno-long-calls
8278 @opindex mlong-calls
8279 @opindex mno-long-calls
8280 Do all calls with the @samp{JALR} instruction, which requires
8281 loading up a function's address into a register before the call.
8282 You need to use this switch, if you call outside of the current
8283 512 megabyte segment to functions that are not through pointers.
8285 @item -membedded-pic
8286 @itemx -mno-embedded-pic
8287 @opindex membedded-pic
8288 @opindex mno-embedded-pic
8289 Generate PIC code suitable for some embedded systems. All calls are
8290 made using PC relative address, and all data is addressed using the $gp
8291 register. No more than 65536 bytes of global data may be used. This
8292 requires GNU as and GNU ld which do most of the work. This currently
8293 only works on targets which use ECOFF; it does not work with ELF@.
8295 @item -membedded-data
8296 @itemx -mno-embedded-data
8297 @opindex membedded-data
8298 @opindex mno-embedded-data
8299 Allocate variables to the read-only data section first if possible, then
8300 next in the small data section if possible, otherwise in data. This gives
8301 slightly slower code than the default, but reduces the amount of RAM required
8302 when executing, and thus may be preferred for some embedded systems.
8304 @item -muninit-const-in-rodata
8305 @itemx -mno-uninit-const-in-rodata
8306 @opindex muninit-const-in-rodata
8307 @opindex mno-uninit-const-in-rodata
8308 When used together with @option{-membedded-data}, it will always store uninitialized
8309 const variables in the read-only data section.
8311 @item -msingle-float
8312 @itemx -mdouble-float
8313 @opindex msingle-float
8314 @opindex mdouble-float
8315 The @option{-msingle-float} switch tells gcc to assume that the floating
8316 point coprocessor only supports single precision operations, as on the
8317 @samp{r4650} chip. The @option{-mdouble-float} switch permits gcc to use
8318 double precision operations. This is the default.
8324 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
8325 as on the @samp{r4650} chip.
8329 Turns on @option{-msingle-float}, @option{-mmad}, and, at least for now,
8330 @option{-mcpu=r4650}.
8336 Enable 16-bit instructions.
8340 Compile code for the processor in little endian mode.
8341 The requisite libraries are assumed to exist.
8345 Compile code for the processor in big endian mode.
8346 The requisite libraries are assumed to exist.
8350 @cindex smaller data references (MIPS)
8351 @cindex gp-relative references (MIPS)
8352 Put global and static items less than or equal to @var{num} bytes into
8353 the small data or bss sections instead of the normal data or bss
8354 section. This allows the assembler to emit one word memory reference
8355 instructions based on the global pointer (@var{gp} or @var{$28}),
8356 instead of the normal two words used. By default, @var{num} is 8 when
8357 the MIPS assembler is used, and 0 when the GNU assembler is used. The
8358 @option{-G @var{num}} switch is also passed to the assembler and linker.
8359 All modules should be compiled with the same @option{-G @var{num}}
8364 Tell the MIPS assembler to not run its preprocessor over user
8365 assembler files (with a @samp{.s} suffix) when assembling them.
8369 Pass an option to gas which will cause nops to be inserted if
8370 the read of the destination register of an mfhi or mflo instruction
8371 occurs in the following two instructions.
8376 Work around certain SB-1 CPU core errata.
8377 (This flag currently works around the SB-1 revision 2
8378 ``F1'' and ``F2'' floating point errata.)
8382 Do not include the default crt0.
8384 @item -mflush-func=@var{func}
8385 @itemx -mno-flush-func
8386 @opindex mflush-func
8387 Specifies the function to call to flush the I and D caches, or to not
8388 call any such function. If called, the function must take the same
8389 arguments as the common @code{_flush_func()}, that is, the address of the
8390 memory range for which the cache is being flushed, the size of the
8391 memory range, and the number 3 (to flush both caches). The default
8392 depends on the target gcc was configured for, but commonly is either
8393 @samp{_flush_func} or @samp{__cpu_flush}.
8395 @item -mbranch-likely
8396 @itemx -mno-branch-likely
8397 @opindex mbranch-likely
8398 @opindex mno-branch-likely
8399 Enable or disable use of Branch Likely instructions, regardless of the
8400 default for the selected architecture. By default, Branch Likely
8401 instructions may be generated if they are supported by the selected
8402 architecture. An exception is for the MIPS32 and MIPS64 architectures
8403 and processors which implement those architectures; for those, Branch
8404 Likely instructions will not be generated by default because the MIPS32
8405 and MIPS64 architectures specifically deprecate their use.
8408 @node i386 and x86-64 Options
8409 @subsection Intel 386 and AMD x86-64 Options
8410 @cindex i386 Options
8411 @cindex x86-64 Options
8412 @cindex Intel 386 Options
8413 @cindex AMD x86-64 Options
8415 These @samp{-m} options are defined for the i386 and x86-64 family of
8419 @item -mtune=@var{cpu-type}
8421 Tune to @var{cpu-type} everything applicable about the generated code, except
8422 for the ABI and the set of available instructions. The choices for
8423 @var{cpu-type} are @samp{i386}, @samp{i486}, @samp{i586}, @samp{i686},
8424 @samp{pentium}, @samp{pentium-mmx}, @samp{pentiumpro}, @samp{pentium2},
8425 @samp{pentium3}, @samp{pentium4}, @samp{k6}, @samp{k6-2}, @samp{k6-3},
8426 @samp{athlon}, @samp{athlon-tbird}, @samp{athlon-4}, @samp{athlon-xp},
8427 @samp{athlon-mp}, @samp{winchip-c6}, @samp{winchip2}, @samp{k8}, @samp{c3}
8430 While picking a specific @var{cpu-type} will schedule things appropriately
8431 for that particular chip, the compiler will not generate any code that
8432 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8433 being used. @samp{i586} is equivalent to @samp{pentium} and @samp{i686}
8434 is equivalent to @samp{pentiumpro}. @samp{k6} and @samp{athlon} are the
8435 AMD chips as opposed to the Intel ones.
8437 @item -march=@var{cpu-type}
8439 Generate instructions for the machine type @var{cpu-type}. The choices
8440 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8441 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8443 @item -mcpu=@var{cpu-type}
8445 A deprecated synonym for @option{-mtune}.
8454 @opindex mpentiumpro
8455 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8456 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8457 These synonyms are deprecated.
8459 @item -mfpmath=@var{unit}
8461 generate floating point arithmetics for selected unit @var{unit}. the choices
8466 Use the standard 387 floating point coprocessor present majority of chips and
8467 emulated otherwise. Code compiled with this option will run almost everywhere.
8468 The temporary results are computed in 80bit precision instead of precision
8469 specified by the type resulting in slightly different results compared to most
8470 of other chips. See @option{-ffloat-store} for more detailed description.
8472 This is the default choice for i386 compiler.
8475 Use scalar floating point instructions present in the SSE instruction set.
8476 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8477 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8478 instruction set supports only single precision arithmetics, thus the double and
8479 extended precision arithmetics is still done using 387. Later version, present
8480 only in Pentium4 and the future AMD x86-64 chips supports double precision
8483 For i387 you need to use @option{-march=@var{cpu-type}}, @option{-msse} or
8484 @option{-msse2} switches to enable SSE extensions and make this option
8485 effective. For x86-64 compiler, these extensions are enabled by default.
8487 The resulting code should be considerably faster in majority of cases and avoid
8488 the numerical instability problems of 387 code, but may break some existing
8489 code that expects temporaries to be 80bit.
8491 This is the default choice for x86-64 compiler.
8494 Use all SSE extensions enabled by @option{-msse2} as well as the new
8495 SSE extensions in Prescott New Instructions. @option{-mpni} also
8496 enables 2 builtin functions, @code{__builtin_ia32_monitor} and
8497 @code{__builtin_ia32_mwait}, for new instructions @code{monitor} and
8501 Attempt to utilize both instruction sets at once. This effectively double the
8502 amount of available registers and on chips with separate execution units for
8503 387 and SSE the execution resources too. Use this option with care, as it is
8504 still experimental, because gcc register allocator does not model separate
8505 functional units well resulting in instable performance.
8508 @item -masm=@var{dialect}
8509 @opindex masm=@var{dialect}
8510 Output asm instructions using selected @var{dialect}. Supported choices are
8511 @samp{intel} or @samp{att} (the default one).
8516 @opindex mno-ieee-fp
8517 Control whether or not the compiler uses IEEE floating point
8518 comparisons. These handle correctly the case where the result of a
8519 comparison is unordered.
8522 @opindex msoft-float
8523 Generate output containing library calls for floating point.
8524 @strong{Warning:} the requisite libraries are not part of GCC@.
8525 Normally the facilities of the machine's usual C compiler are used, but
8526 this can't be done directly in cross-compilation. You must make your
8527 own arrangements to provide suitable library functions for
8530 On machines where a function returns floating point results in the 80387
8531 register stack, some floating point opcodes may be emitted even if
8532 @option{-msoft-float} is used.
8534 @item -mno-fp-ret-in-387
8535 @opindex mno-fp-ret-in-387
8536 Do not use the FPU registers for return values of functions.
8538 The usual calling convention has functions return values of types
8539 @code{float} and @code{double} in an FPU register, even if there
8540 is no FPU@. The idea is that the operating system should emulate
8543 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8544 in ordinary CPU registers instead.
8546 @item -mno-fancy-math-387
8547 @opindex mno-fancy-math-387
8548 Some 387 emulators do not support the @code{sin}, @code{cos} and
8549 @code{sqrt} instructions for the 387. Specify this option to avoid
8550 generating those instructions. This option is the default on FreeBSD,
8551 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8552 indicates that the target cpu will always have an FPU and so the
8553 instruction will not need emulation. As of revision 2.6.1, these
8554 instructions are not generated unless you also use the
8555 @option{-funsafe-math-optimizations} switch.
8557 @item -malign-double
8558 @itemx -mno-align-double
8559 @opindex malign-double
8560 @opindex mno-align-double
8561 Control whether GCC aligns @code{double}, @code{long double}, and
8562 @code{long long} variables on a two word boundary or a one word
8563 boundary. Aligning @code{double} variables on a two word boundary will
8564 produce code that runs somewhat faster on a @samp{Pentium} at the
8565 expense of more memory.
8567 @strong{Warning:} if you use the @option{-malign-double} switch,
8568 structures containing the above types will be aligned differently than
8569 the published application binary interface specifications for the 386
8570 and will not be binary compatible with structures in code compiled
8571 without that switch.
8573 @item -m96bit-long-double
8574 @item -m128bit-long-double
8575 @opindex m96bit-long-double
8576 @opindex m128bit-long-double
8577 These switches control the size of @code{long double} type. The i386
8578 application binary interface specifies the size to be 96 bits,
8579 so @option{-m96bit-long-double} is the default in 32 bit mode.
8581 Modern architectures (Pentium and newer) would prefer @code{long double}
8582 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8583 conforming to the ABI, this would not be possible. So specifying a
8584 @option{-m128bit-long-double} will align @code{long double}
8585 to a 16 byte boundary by padding the @code{long double} with an additional
8588 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8589 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8591 Notice that neither of these options enable any extra precision over the x87
8592 standard of 80 bits for a @code{long double}.
8594 @strong{Warning:} if you override the default value for your target ABI, the
8595 structures and arrays containing @code{long double} will change their size as
8596 well as function calling convention for function taking @code{long double}
8597 will be modified. Hence they will not be binary compatible with arrays or
8598 structures in code compiled without that switch.
8602 @itemx -mno-svr3-shlib
8603 @opindex msvr3-shlib
8604 @opindex mno-svr3-shlib
8605 Control whether GCC places uninitialized local variables into the
8606 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8607 into @code{bss}. These options are meaningful only on System V Release 3.
8611 Use a different function-calling convention, in which functions that
8612 take a fixed number of arguments return with the @code{ret} @var{num}
8613 instruction, which pops their arguments while returning. This saves one
8614 instruction in the caller since there is no need to pop the arguments
8617 You can specify that an individual function is called with this calling
8618 sequence with the function attribute @samp{stdcall}. You can also
8619 override the @option{-mrtd} option by using the function attribute
8620 @samp{cdecl}. @xref{Function Attributes}.
8622 @strong{Warning:} this calling convention is incompatible with the one
8623 normally used on Unix, so you cannot use it if you need to call
8624 libraries compiled with the Unix compiler.
8626 Also, you must provide function prototypes for all functions that
8627 take variable numbers of arguments (including @code{printf});
8628 otherwise incorrect code will be generated for calls to those
8631 In addition, seriously incorrect code will result if you call a
8632 function with too many arguments. (Normally, extra arguments are
8633 harmlessly ignored.)
8635 @item -mregparm=@var{num}
8637 Control how many registers are used to pass integer arguments. By
8638 default, no registers are used to pass arguments, and at most 3
8639 registers can be used. You can control this behavior for a specific
8640 function by using the function attribute @samp{regparm}.
8641 @xref{Function Attributes}.
8643 @strong{Warning:} if you use this switch, and
8644 @var{num} is nonzero, then you must build all modules with the same
8645 value, including any libraries. This includes the system libraries and
8648 @item -mpreferred-stack-boundary=@var{num}
8649 @opindex mpreferred-stack-boundary
8650 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8651 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8652 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8653 size (@option{-Os}), in which case the default is the minimum correct
8654 alignment (4 bytes for x86, and 8 bytes for x86-64).
8656 On Pentium and PentiumPro, @code{double} and @code{long double} values
8657 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8658 suffer significant run time performance penalties. On Pentium III, the
8659 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8660 penalties if it is not 16 byte aligned.
8662 To ensure proper alignment of this values on the stack, the stack boundary
8663 must be as aligned as that required by any value stored on the stack.
8664 Further, every function must be generated such that it keeps the stack
8665 aligned. Thus calling a function compiled with a higher preferred
8666 stack boundary from a function compiled with a lower preferred stack
8667 boundary will most likely misalign the stack. It is recommended that
8668 libraries that use callbacks always use the default setting.
8670 This extra alignment does consume extra stack space, and generally
8671 increases code size. Code that is sensitive to stack space usage, such
8672 as embedded systems and operating system kernels, may want to reduce the
8673 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8691 These switches enable or disable the use of built-in functions that allow
8692 direct access to the MMX, SSE and 3Dnow extensions of the instruction set.
8694 @xref{X86 Built-in Functions}, for details of the functions enabled
8695 and disabled by these switches.
8697 To have SSE/SSE2 instructions generated automatically from floating-point
8698 code, see @option{-mfpmath=sse}.
8701 @itemx -mno-push-args
8703 @opindex mno-push-args
8704 Use PUSH operations to store outgoing parameters. This method is shorter
8705 and usually equally fast as method using SUB/MOV operations and is enabled
8706 by default. In some cases disabling it may improve performance because of
8707 improved scheduling and reduced dependencies.
8709 @item -maccumulate-outgoing-args
8710 @opindex maccumulate-outgoing-args
8711 If enabled, the maximum amount of space required for outgoing arguments will be
8712 computed in the function prologue. This is faster on most modern CPUs
8713 because of reduced dependencies, improved scheduling and reduced stack usage
8714 when preferred stack boundary is not equal to 2. The drawback is a notable
8715 increase in code size. This switch implies @option{-mno-push-args}.
8719 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8720 on thread-safe exception handling must compile and link all code with the
8721 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8722 @option{-D_MT}; when linking, it links in a special thread helper library
8723 @option{-lmingwthrd} which cleans up per thread exception handling data.
8725 @item -mno-align-stringops
8726 @opindex mno-align-stringops
8727 Do not align destination of inlined string operations. This switch reduces
8728 code size and improves performance in case the destination is already aligned,
8729 but gcc don't know about it.
8731 @item -minline-all-stringops
8732 @opindex minline-all-stringops
8733 By default GCC inlines string operations only when destination is known to be
8734 aligned at least to 4 byte boundary. This enables more inlining, increase code
8735 size, but may improve performance of code that depends on fast memcpy, strlen
8736 and memset for short lengths.
8738 @item -momit-leaf-frame-pointer
8739 @opindex momit-leaf-frame-pointer
8740 Don't keep the frame pointer in a register for leaf functions. This
8741 avoids the instructions to save, set up and restore frame pointers and
8742 makes an extra register available in leaf functions. The option
8743 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8744 which might make debugging harder.
8746 @item -mtls-direct-seg-refs
8747 @itemx -mno-tls-direct-seg-refs
8748 @opindex mtls-direct-seg-refs
8749 Controls whether TLS variables may be accessed with offsets from the
8750 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8751 or whether the thread base pointer must be added. Whether or not this
8752 is legal depends on the operating system, and whether it maps the
8753 segment to cover the entire TLS area.
8755 For systems that use GNU libc, the default is on.
8758 These @samp{-m} switches are supported in addition to the above
8759 on AMD x86-64 processors in 64-bit environments.
8766 Generate code for a 32-bit or 64-bit environment.
8767 The 32-bit environment sets int, long and pointer to 32 bits and
8768 generates code that runs on any i386 system.
8769 The 64-bit environment sets int to 32 bits and long and pointer
8770 to 64 bits and generates code for AMD's x86-64 architecture.
8773 @opindex no-red-zone
8774 Do not use a so called red zone for x86-64 code. The red zone is mandated
8775 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8776 stack pointer that will not be modified by signal or interrupt handlers
8777 and therefore can be used for temporary data without adjusting the stack
8778 pointer. The flag @option{-mno-red-zone} disables this red zone.
8780 @item -mcmodel=small
8781 @opindex mcmodel=small
8782 Generate code for the small code model: the program and its symbols must
8783 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8784 Programs can be statically or dynamically linked. This is the default
8787 @item -mcmodel=kernel
8788 @opindex mcmodel=kernel
8789 Generate code for the kernel code model. The kernel runs in the
8790 negative 2 GB of the address space.
8791 This model has to be used for Linux kernel code.
8793 @item -mcmodel=medium
8794 @opindex mcmodel=medium
8795 Generate code for the medium model: The program is linked in the lower 2
8796 GB of the address space but symbols can be located anywhere in the
8797 address space. Programs can be statically or dynamically linked, but
8798 building of shared libraries are not supported with the medium model.
8800 @item -mcmodel=large
8801 @opindex mcmodel=large
8802 Generate code for the large model: This model makes no assumptions
8803 about addresses and sizes of sections. Currently GCC does not implement
8808 @subsection HPPA Options
8809 @cindex HPPA Options
8811 These @samp{-m} options are defined for the HPPA family of computers:
8814 @item -march=@var{architecture-type}
8816 Generate code for the specified architecture. The choices for
8817 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8818 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8819 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8820 architecture option for your machine. Code compiled for lower numbered
8821 architectures will run on higher numbered architectures, but not the
8824 PA 2.0 support currently requires gas snapshot 19990413 or later. The
8825 next release of binutils (current is 2.9.1) will probably contain PA 2.0
8829 @itemx -mpa-risc-1-1
8830 @itemx -mpa-risc-2-0
8831 @opindex mpa-risc-1-0
8832 @opindex mpa-risc-1-1
8833 @opindex mpa-risc-2-0
8834 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8837 @opindex mbig-switch
8838 Generate code suitable for big switch tables. Use this option only if
8839 the assembler/linker complain about out of range branches within a switch
8842 @item -mjump-in-delay
8843 @opindex mjump-in-delay
8844 Fill delay slots of function calls with unconditional jump instructions
8845 by modifying the return pointer for the function call to be the target
8846 of the conditional jump.
8848 @item -mdisable-fpregs
8849 @opindex mdisable-fpregs
8850 Prevent floating point registers from being used in any manner. This is
8851 necessary for compiling kernels which perform lazy context switching of
8852 floating point registers. If you use this option and attempt to perform
8853 floating point operations, the compiler will abort.
8855 @item -mdisable-indexing
8856 @opindex mdisable-indexing
8857 Prevent the compiler from using indexing address modes. This avoids some
8858 rather obscure problems when compiling MIG generated code under MACH@.
8860 @item -mno-space-regs
8861 @opindex mno-space-regs
8862 Generate code that assumes the target has no space registers. This allows
8863 GCC to generate faster indirect calls and use unscaled index address modes.
8865 Such code is suitable for level 0 PA systems and kernels.
8867 @item -mfast-indirect-calls
8868 @opindex mfast-indirect-calls
8869 Generate code that assumes calls never cross space boundaries. This
8870 allows GCC to emit code which performs faster indirect calls.
8872 This option will not work in the presence of shared libraries or nested
8875 @item -mlong-load-store
8876 @opindex mlong-load-store
8877 Generate 3-instruction load and store sequences as sometimes required by
8878 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8881 @item -mportable-runtime
8882 @opindex mportable-runtime
8883 Use the portable calling conventions proposed by HP for ELF systems.
8887 Enable the use of assembler directives only GAS understands.
8889 @item -mschedule=@var{cpu-type}
8891 Schedule code according to the constraints for the machine type
8892 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8893 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8894 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8895 proper scheduling option for your machine. The default scheduling is
8899 @opindex mlinker-opt
8900 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8901 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8902 linkers in which they give bogus error messages when linking some programs.
8905 @opindex msoft-float
8906 Generate output containing library calls for floating point.
8907 @strong{Warning:} the requisite libraries are not available for all HPPA
8908 targets. Normally the facilities of the machine's usual C compiler are
8909 used, but this cannot be done directly in cross-compilation. You must make
8910 your own arrangements to provide suitable library functions for
8911 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8912 does provide software floating point support.
8914 @option{-msoft-float} changes the calling convention in the output file;
8915 therefore, it is only useful if you compile @emph{all} of a program with
8916 this option. In particular, you need to compile @file{libgcc.a}, the
8917 library that comes with GCC, with @option{-msoft-float} in order for
8922 Generate the predefine, @code{_SIO}, for server IO. The default is
8923 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8924 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO. These
8925 options are available under HP-UX and HI-UX.
8929 Use GNU ld specific options. This passes @option{-shared} to ld when
8930 building a shared library. It is the default when GCC is configured,
8931 explicitly or implicitly, with the GNU linker. This option does not
8932 have any affect on which ld is called, it only changes what parameters
8933 are passed to that ld. The ld that is called is determined by the
8934 @option{--with-ld} configure option, gcc's program search path, and
8935 finally by the user's @env{PATH}. The linker used by GCC can be printed
8936 using @samp{which `gcc -print-prog-name=ld`}.
8940 Use HP ld specific options. This passes @option{-b} to ld when building
8941 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8942 links. It is the default when GCC is configured, explicitly or
8943 implicitly, with the HP linker. This option does not have any affect on
8944 which ld is called, it only changes what parameters are passed to that
8945 ld. The ld that is called is determined by the @option{--with-ld}
8946 configure option, gcc's program search path, and finally by the user's
8947 @env{PATH}. The linker used by GCC can be printed using @samp{which
8948 `gcc -print-prog-name=ld`}.
8951 @opindex mno-long-calls
8952 Generate code that uses long call sequences. This ensures that a call
8953 is always able to reach linker generated stubs. The default is to generate
8954 long calls only when the distance from the call site to the beginning
8955 of the function or translation unit, as the case may be, exceeds a
8956 predefined limit set by the branch type being used. The limits for
8957 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8958 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8961 Distances are measured from the beginning of functions when using the
8962 @option{-ffunction-sections} option, or when using the @option{-mgas}
8963 and @option{-mno-portable-runtime} options together under HP-UX with
8966 It is normally not desirable to use this option as it will degrade
8967 performance. However, it may be useful in large applications,
8968 particularly when partial linking is used to build the application.
8970 The types of long calls used depends on the capabilities of the
8971 assembler and linker, and the type of code being generated. The
8972 impact on systems that support long absolute calls, and long pic
8973 symbol-difference or pc-relative calls should be relatively small.
8974 However, an indirect call is used on 32-bit ELF systems in pic code
8975 and it is quite long.
8979 Suppress the generation of link options to search libdld.sl when the
8980 @option{-static} option is specified on HP-UX 10 and later.
8984 The HP-UX implementation of setlocale in libc has a dependency on
8985 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8986 when the @option{-static} option is specified, special link options
8987 are needed to resolve this dependency.
8989 On HP-UX 10 and later, the GCC driver adds the necessary options to
8990 link with libdld.sl when the @option{-static} option is specified.
8991 This causes the resulting binary to be dynamic. On the 64-bit port,
8992 the linkers generate dynamic binaries by default in any case. The
8993 @option{-nolibdld} option can be used to prevent the GCC driver from
8994 adding these link options.
8998 Add support for multithreading with the @dfn{dce thread} library
8999 under HP-UX. This option sets flags for both the preprocessor and
9003 @node Intel 960 Options
9004 @subsection Intel 960 Options
9006 These @samp{-m} options are defined for the Intel 960 implementations:
9009 @item -m@var{cpu-type}
9017 Assume the defaults for the machine type @var{cpu-type} for some of
9018 the other options, including instruction scheduling, floating point
9019 support, and addressing modes. The choices for @var{cpu-type} are
9020 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
9021 @samp{sa}, and @samp{sb}.
9028 @opindex msoft-float
9029 The @option{-mnumerics} option indicates that the processor does support
9030 floating-point instructions. The @option{-msoft-float} option indicates
9031 that floating-point support should not be assumed.
9033 @item -mleaf-procedures
9034 @itemx -mno-leaf-procedures
9035 @opindex mleaf-procedures
9036 @opindex mno-leaf-procedures
9037 Do (or do not) attempt to alter leaf procedures to be callable with the
9038 @code{bal} instruction as well as @code{call}. This will result in more
9039 efficient code for explicit calls when the @code{bal} instruction can be
9040 substituted by the assembler or linker, but less efficient code in other
9041 cases, such as calls via function pointers, or using a linker that doesn't
9042 support this optimization.
9045 @itemx -mno-tail-call
9047 @opindex mno-tail-call
9048 Do (or do not) make additional attempts (beyond those of the
9049 machine-independent portions of the compiler) to optimize tail-recursive
9050 calls into branches. You may not want to do this because the detection of
9051 cases where this is not valid is not totally complete. The default is
9052 @option{-mno-tail-call}.
9054 @item -mcomplex-addr
9055 @itemx -mno-complex-addr
9056 @opindex mcomplex-addr
9057 @opindex mno-complex-addr
9058 Assume (or do not assume) that the use of a complex addressing mode is a
9059 win on this implementation of the i960. Complex addressing modes may not
9060 be worthwhile on the K-series, but they definitely are on the C-series.
9061 The default is currently @option{-mcomplex-addr} for all processors except
9065 @itemx -mno-code-align
9066 @opindex mcode-align
9067 @opindex mno-code-align
9068 Align code to 8-byte boundaries for faster fetching (or don't bother).
9069 Currently turned on by default for C-series implementations only.
9072 @item -mclean-linkage
9073 @itemx -mno-clean-linkage
9074 @opindex mclean-linkage
9075 @opindex mno-clean-linkage
9076 These options are not fully implemented.
9080 @itemx -mic2.0-compat
9081 @itemx -mic3.0-compat
9083 @opindex mic2.0-compat
9084 @opindex mic3.0-compat
9085 Enable compatibility with iC960 v2.0 or v3.0.
9089 @opindex masm-compat
9091 Enable compatibility with the iC960 assembler.
9093 @item -mstrict-align
9094 @itemx -mno-strict-align
9095 @opindex mstrict-align
9096 @opindex mno-strict-align
9097 Do not permit (do permit) unaligned accesses.
9101 Enable structure-alignment compatibility with Intel's gcc release version
9102 1.3 (based on gcc 1.37). This option implies @option{-mstrict-align}.
9104 @item -mlong-double-64
9105 @opindex mlong-double-64
9106 Implement type @samp{long double} as 64-bit floating point numbers.
9107 Without the option @samp{long double} is implemented by 80-bit
9108 floating point numbers. The only reason we have it because there is
9109 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
9110 is only useful for people using soft-float targets. Otherwise, we
9111 should recommend against use of it.
9115 @node DEC Alpha Options
9116 @subsection DEC Alpha Options
9118 These @samp{-m} options are defined for the DEC Alpha implementations:
9121 @item -mno-soft-float
9123 @opindex mno-soft-float
9124 @opindex msoft-float
9125 Use (do not use) the hardware floating-point instructions for
9126 floating-point operations. When @option{-msoft-float} is specified,
9127 functions in @file{libgcc.a} will be used to perform floating-point
9128 operations. Unless they are replaced by routines that emulate the
9129 floating-point operations, or compiled in such a way as to call such
9130 emulations routines, these routines will issue floating-point
9131 operations. If you are compiling for an Alpha without floating-point
9132 operations, you must ensure that the library is built so as not to call
9135 Note that Alpha implementations without floating-point operations are
9136 required to have floating-point registers.
9141 @opindex mno-fp-regs
9142 Generate code that uses (does not use) the floating-point register set.
9143 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9144 register set is not used, floating point operands are passed in integer
9145 registers as if they were integers and floating-point results are passed
9146 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9147 so any function with a floating-point argument or return value called by code
9148 compiled with @option{-mno-fp-regs} must also be compiled with that
9151 A typical use of this option is building a kernel that does not use,
9152 and hence need not save and restore, any floating-point registers.
9156 The Alpha architecture implements floating-point hardware optimized for
9157 maximum performance. It is mostly compliant with the IEEE floating
9158 point standard. However, for full compliance, software assistance is
9159 required. This option generates code fully IEEE compliant code
9160 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9161 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9162 defined during compilation. The resulting code is less efficient but is
9163 able to correctly support denormalized numbers and exceptional IEEE
9164 values such as not-a-number and plus/minus infinity. Other Alpha
9165 compilers call this option @option{-ieee_with_no_inexact}.
9167 @item -mieee-with-inexact
9168 @opindex mieee-with-inexact
9169 This is like @option{-mieee} except the generated code also maintains
9170 the IEEE @var{inexact-flag}. Turning on this option causes the
9171 generated code to implement fully-compliant IEEE math. In addition to
9172 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9173 macro. On some Alpha implementations the resulting code may execute
9174 significantly slower than the code generated by default. Since there is
9175 very little code that depends on the @var{inexact-flag}, you should
9176 normally not specify this option. Other Alpha compilers call this
9177 option @option{-ieee_with_inexact}.
9179 @item -mfp-trap-mode=@var{trap-mode}
9180 @opindex mfp-trap-mode
9181 This option controls what floating-point related traps are enabled.
9182 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9183 The trap mode can be set to one of four values:
9187 This is the default (normal) setting. The only traps that are enabled
9188 are the ones that cannot be disabled in software (e.g., division by zero
9192 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9196 Like @samp{su}, but the instructions are marked to be safe for software
9197 completion (see Alpha architecture manual for details).
9200 Like @samp{su}, but inexact traps are enabled as well.
9203 @item -mfp-rounding-mode=@var{rounding-mode}
9204 @opindex mfp-rounding-mode
9205 Selects the IEEE rounding mode. Other Alpha compilers call this option
9206 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9211 Normal IEEE rounding mode. Floating point numbers are rounded towards
9212 the nearest machine number or towards the even machine number in case
9216 Round towards minus infinity.
9219 Chopped rounding mode. Floating point numbers are rounded towards zero.
9222 Dynamic rounding mode. A field in the floating point control register
9223 (@var{fpcr}, see Alpha architecture reference manual) controls the
9224 rounding mode in effect. The C library initializes this register for
9225 rounding towards plus infinity. Thus, unless your program modifies the
9226 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9229 @item -mtrap-precision=@var{trap-precision}
9230 @opindex mtrap-precision
9231 In the Alpha architecture, floating point traps are imprecise. This
9232 means without software assistance it is impossible to recover from a
9233 floating trap and program execution normally needs to be terminated.
9234 GCC can generate code that can assist operating system trap handlers
9235 in determining the exact location that caused a floating point trap.
9236 Depending on the requirements of an application, different levels of
9237 precisions can be selected:
9241 Program precision. This option is the default and means a trap handler
9242 can only identify which program caused a floating point exception.
9245 Function precision. The trap handler can determine the function that
9246 caused a floating point exception.
9249 Instruction precision. The trap handler can determine the exact
9250 instruction that caused a floating point exception.
9253 Other Alpha compilers provide the equivalent options called
9254 @option{-scope_safe} and @option{-resumption_safe}.
9256 @item -mieee-conformant
9257 @opindex mieee-conformant
9258 This option marks the generated code as IEEE conformant. You must not
9259 use this option unless you also specify @option{-mtrap-precision=i} and either
9260 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9261 is to emit the line @samp{.eflag 48} in the function prologue of the
9262 generated assembly file. Under DEC Unix, this has the effect that
9263 IEEE-conformant math library routines will be linked in.
9265 @item -mbuild-constants
9266 @opindex mbuild-constants
9267 Normally GCC examines a 32- or 64-bit integer constant to
9268 see if it can construct it from smaller constants in two or three
9269 instructions. If it cannot, it will output the constant as a literal and
9270 generate code to load it from the data segment at runtime.
9272 Use this option to require GCC to construct @emph{all} integer constants
9273 using code, even if it takes more instructions (the maximum is six).
9275 You would typically use this option to build a shared library dynamic
9276 loader. Itself a shared library, it must relocate itself in memory
9277 before it can find the variables and constants in its own data segment.
9283 Select whether to generate code to be assembled by the vendor-supplied
9284 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9302 Indicate whether GCC should generate code to use the optional BWX,
9303 CIX, FIX and MAX instruction sets. The default is to use the instruction
9304 sets supported by the CPU type specified via @option{-mcpu=} option or that
9305 of the CPU on which GCC was built if none was specified.
9310 @opindex mfloat-ieee
9311 Generate code that uses (does not use) VAX F and G floating point
9312 arithmetic instead of IEEE single and double precision.
9314 @item -mexplicit-relocs
9315 @itemx -mno-explicit-relocs
9316 @opindex mexplicit-relocs
9317 @opindex mno-explicit-relocs
9318 Older Alpha assemblers provided no way to generate symbol relocations
9319 except via assembler macros. Use of these macros does not allow
9320 optimal instruction scheduling. GNU binutils as of version 2.12
9321 supports a new syntax that allows the compiler to explicitly mark
9322 which relocations should apply to which instructions. This option
9323 is mostly useful for debugging, as GCC detects the capabilities of
9324 the assembler when it is built and sets the default accordingly.
9328 @opindex msmall-data
9329 @opindex mlarge-data
9330 When @option{-mexplicit-relocs} is in effect, static data is
9331 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9332 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9333 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9334 16-bit relocations off of the @code{$gp} register. This limits the
9335 size of the small data area to 64KB, but allows the variables to be
9336 directly accessed via a single instruction.
9338 The default is @option{-mlarge-data}. With this option the data area
9339 is limited to just below 2GB. Programs that require more than 2GB of
9340 data must use @code{malloc} or @code{mmap} to allocate the data in the
9341 heap instead of in the program's data segment.
9343 When generating code for shared libraries, @option{-fpic} implies
9344 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9348 @opindex msmall-text
9349 @opindex mlarge-text
9350 When @option{-msmall-text} is used, the compiler assumes that the
9351 code of the entire program (or shared library) fits in 4MB, and is
9352 thus reachable with a branch instruction. When @option{-msmall-data}
9353 is used, the compiler can assume that all local symbols share the
9354 same @code{$gp} value, and thus reduce the number of instructions
9355 required for a function call from 4 to 1.
9357 The default is @option{-mlarge-text}.
9359 @item -mcpu=@var{cpu_type}
9361 Set the instruction set and instruction scheduling parameters for
9362 machine type @var{cpu_type}. You can specify either the @samp{EV}
9363 style name or the corresponding chip number. GCC supports scheduling
9364 parameters for the EV4, EV5 and EV6 family of processors and will
9365 choose the default values for the instruction set from the processor
9366 you specify. If you do not specify a processor type, GCC will default
9367 to the processor on which the compiler was built.
9369 Supported values for @var{cpu_type} are
9375 Schedules as an EV4 and has no instruction set extensions.
9379 Schedules as an EV5 and has no instruction set extensions.
9383 Schedules as an EV5 and supports the BWX extension.
9388 Schedules as an EV5 and supports the BWX and MAX extensions.
9392 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9396 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9399 @item -mtune=@var{cpu_type}
9401 Set only the instruction scheduling parameters for machine type
9402 @var{cpu_type}. The instruction set is not changed.
9404 @item -mmemory-latency=@var{time}
9405 @opindex mmemory-latency
9406 Sets the latency the scheduler should assume for typical memory
9407 references as seen by the application. This number is highly
9408 dependent on the memory access patterns used by the application
9409 and the size of the external cache on the machine.
9411 Valid options for @var{time} are
9415 A decimal number representing clock cycles.
9421 The compiler contains estimates of the number of clock cycles for
9422 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9423 (also called Dcache, Scache, and Bcache), as well as to main memory.
9424 Note that L3 is only valid for EV5.
9429 @node DEC Alpha/VMS Options
9430 @subsection DEC Alpha/VMS Options
9432 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9435 @item -mvms-return-codes
9436 @opindex mvms-return-codes
9437 Return VMS condition codes from main. The default is to return POSIX
9438 style condition (e.g.@ error) codes.
9441 @node H8/300 Options
9442 @subsection H8/300 Options
9444 These @samp{-m} options are defined for the H8/300 implementations:
9449 Shorten some address references at link time, when possible; uses the
9450 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9451 ld, Using ld}, for a fuller description.
9455 Generate code for the H8/300H@.
9459 Generate code for the H8S@.
9463 Generate code for the H8S and H8/300H in the normal mode. This switch
9464 must be used either with -mh or -ms.
9468 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9472 Make @code{int} data 32 bits by default.
9476 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9477 The default for the H8/300H and H8S is to align longs and floats on 4
9479 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9480 This option has no effect on the H8/300.
9484 @subsection SH Options
9486 These @samp{-m} options are defined for the SH implementations:
9491 Generate code for the SH1.
9495 Generate code for the SH2.
9498 Generate code for the SH2e.
9502 Generate code for the SH3.
9506 Generate code for the SH3e.
9510 Generate code for the SH4 without a floating-point unit.
9512 @item -m4-single-only
9513 @opindex m4-single-only
9514 Generate code for the SH4 with a floating-point unit that only
9515 supports single-precision arithmetic.
9519 Generate code for the SH4 assuming the floating-point unit is in
9520 single-precision mode by default.
9524 Generate code for the SH4.
9528 Compile code for the processor in big endian mode.
9532 Compile code for the processor in little endian mode.
9536 Align doubles at 64-bit boundaries. Note that this changes the calling
9537 conventions, and thus some functions from the standard C library will
9538 not work unless you recompile it first with @option{-mdalign}.
9542 Shorten some address references at link time, when possible; uses the
9543 linker option @option{-relax}.
9547 Use 32-bit offsets in @code{switch} tables. The default is to use
9552 Enable the use of the instruction @code{fmovd}.
9556 Comply with the calling conventions defined by Renesas.
9560 Mark the @code{MAC} register as call-clobbered, even if
9561 @option{-mhitachi} is given.
9565 Increase IEEE-compliance of floating-point code.
9569 Dump instruction size and location in the assembly code.
9573 This option is deprecated. It pads structures to multiple of 4 bytes,
9574 which is incompatible with the SH ABI@.
9578 Optimize for space instead of speed. Implied by @option{-Os}.
9582 When generating position-independent code, emit function calls using
9583 the Global Offset Table instead of the Procedure Linkage Table.
9587 Generate a library function call to invalidate instruction cache
9588 entries, after fixing up a trampoline. This library function call
9589 doesn't assume it can write to the whole memory address space. This
9590 is the default when the target is @code{sh-*-linux*}.
9593 @node System V Options
9594 @subsection Options for System V
9596 These additional options are available on System V Release 4 for
9597 compatibility with other compilers on those systems:
9602 Create a shared object.
9603 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
9607 Identify the versions of each tool used by the compiler, in a
9608 @code{.ident} assembler directive in the output.
9612 Refrain from adding @code{.ident} directives to the output file (this is
9615 @item -YP,@var{dirs}
9617 Search the directories @var{dirs}, and no others, for libraries
9618 specified with @option{-l}.
9622 Look in the directory @var{dir} to find the M4 preprocessor.
9623 The assembler uses this option.
9624 @c This is supposed to go with a -Yd for predefined M4 macro files, but
9625 @c the generic assembler that comes with Solaris takes just -Ym.
9628 @node TMS320C3x/C4x Options
9629 @subsection TMS320C3x/C4x Options
9630 @cindex TMS320C3x/C4x Options
9632 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
9636 @item -mcpu=@var{cpu_type}
9638 Set the instruction set, register set, and instruction scheduling
9639 parameters for machine type @var{cpu_type}. Supported values for
9640 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
9641 @samp{c44}. The default is @samp{c40} to generate code for the
9646 @itemx -msmall-memory
9648 @opindex mbig-memory
9650 @opindex msmall-memory
9652 Generates code for the big or small memory model. The small memory
9653 model assumed that all data fits into one 64K word page. At run-time
9654 the data page (DP) register must be set to point to the 64K page
9655 containing the .bss and .data program sections. The big memory model is
9656 the default and requires reloading of the DP register for every direct
9663 Allow (disallow) allocation of general integer operands into the block
9670 Enable (disable) generation of code using decrement and branch,
9671 DBcond(D), instructions. This is enabled by default for the C4x. To be
9672 on the safe side, this is disabled for the C3x, since the maximum
9673 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
9674 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
9675 that it can utilize the decrement and branch instruction, but will give
9676 up if there is more than one memory reference in the loop. Thus a loop
9677 where the loop counter is decremented can generate slightly more
9678 efficient code, in cases where the RPTB instruction cannot be utilized.
9680 @item -mdp-isr-reload
9682 @opindex mdp-isr-reload
9684 Force the DP register to be saved on entry to an interrupt service
9685 routine (ISR), reloaded to point to the data section, and restored on
9686 exit from the ISR@. This should not be required unless someone has
9687 violated the small memory model by modifying the DP register, say within
9694 For the C3x use the 24-bit MPYI instruction for integer multiplies
9695 instead of a library call to guarantee 32-bit results. Note that if one
9696 of the operands is a constant, then the multiplication will be performed
9697 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
9698 then squaring operations are performed inline instead of a library call.
9701 @itemx -mno-fast-fix
9703 @opindex mno-fast-fix
9704 The C3x/C4x FIX instruction to convert a floating point value to an
9705 integer value chooses the nearest integer less than or equal to the
9706 floating point value rather than to the nearest integer. Thus if the
9707 floating point number is negative, the result will be incorrectly
9708 truncated an additional code is necessary to detect and correct this
9709 case. This option can be used to disable generation of the additional
9710 code required to correct the result.
9716 Enable (disable) generation of repeat block sequences using the RPTB
9717 instruction for zero overhead looping. The RPTB construct is only used
9718 for innermost loops that do not call functions or jump across the loop
9719 boundaries. There is no advantage having nested RPTB loops due to the
9720 overhead required to save and restore the RC, RS, and RE registers.
9721 This is enabled by default with @option{-O2}.
9723 @item -mrpts=@var{count}
9727 Enable (disable) the use of the single instruction repeat instruction
9728 RPTS@. If a repeat block contains a single instruction, and the loop
9729 count can be guaranteed to be less than the value @var{count}, GCC will
9730 emit a RPTS instruction instead of a RPTB@. If no value is specified,
9731 then a RPTS will be emitted even if the loop count cannot be determined
9732 at compile time. Note that the repeated instruction following RPTS does
9733 not have to be reloaded from memory each iteration, thus freeing up the
9734 CPU buses for operands. However, since interrupts are blocked by this
9735 instruction, it is disabled by default.
9737 @item -mloop-unsigned
9738 @itemx -mno-loop-unsigned
9739 @opindex mloop-unsigned
9740 @opindex mno-loop-unsigned
9741 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
9742 is @math{2^{31} + 1} since these instructions test if the iteration count is
9743 negative to terminate the loop. If the iteration count is unsigned
9744 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
9745 exceeded. This switch allows an unsigned iteration count.
9749 Try to emit an assembler syntax that the TI assembler (asm30) is happy
9750 with. This also enforces compatibility with the API employed by the TI
9751 C3x C compiler. For example, long doubles are passed as structures
9752 rather than in floating point registers.
9758 Generate code that uses registers (stack) for passing arguments to functions.
9759 By default, arguments are passed in registers where possible rather
9760 than by pushing arguments on to the stack.
9762 @item -mparallel-insns
9763 @itemx -mno-parallel-insns
9764 @opindex mparallel-insns
9765 @opindex mno-parallel-insns
9766 Allow the generation of parallel instructions. This is enabled by
9767 default with @option{-O2}.
9769 @item -mparallel-mpy
9770 @itemx -mno-parallel-mpy
9771 @opindex mparallel-mpy
9772 @opindex mno-parallel-mpy
9773 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
9774 provided @option{-mparallel-insns} is also specified. These instructions have
9775 tight register constraints which can pessimize the code generation
9781 @subsection V850 Options
9782 @cindex V850 Options
9784 These @samp{-m} options are defined for V850 implementations:
9788 @itemx -mno-long-calls
9789 @opindex mlong-calls
9790 @opindex mno-long-calls
9791 Treat all calls as being far away (near). If calls are assumed to be
9792 far away, the compiler will always load the functions address up into a
9793 register, and call indirect through the pointer.
9799 Do not optimize (do optimize) basic blocks that use the same index
9800 pointer 4 or more times to copy pointer into the @code{ep} register, and
9801 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
9802 option is on by default if you optimize.
9804 @item -mno-prolog-function
9805 @itemx -mprolog-function
9806 @opindex mno-prolog-function
9807 @opindex mprolog-function
9808 Do not use (do use) external functions to save and restore registers
9809 at the prologue and epilogue of a function. The external functions
9810 are slower, but use less code space if more than one function saves
9811 the same number of registers. The @option{-mprolog-function} option
9812 is on by default if you optimize.
9816 Try to make the code as small as possible. At present, this just turns
9817 on the @option{-mep} and @option{-mprolog-function} options.
9821 Put static or global variables whose size is @var{n} bytes or less into
9822 the tiny data area that register @code{ep} points to. The tiny data
9823 area can hold up to 256 bytes in total (128 bytes for byte references).
9827 Put static or global variables whose size is @var{n} bytes or less into
9828 the small data area that register @code{gp} points to. The small data
9829 area can hold up to 64 kilobytes.
9833 Put static or global variables whose size is @var{n} bytes or less into
9834 the first 32 kilobytes of memory.
9838 Specify that the target processor is the V850.
9841 @opindex mbig-switch
9842 Generate code suitable for big switch tables. Use this option only if
9843 the assembler/linker complain about out of range branches within a switch
9848 This option will cause r2 and r5 to be used in the code generated by
9849 the compiler. This setting is the default.
9852 @opindex mno-app-regs
9853 This option will cause r2 and r5 to be treated as fixed registers.
9857 Specify that the target processor is the V850E1. The preprocessor
9858 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
9859 this option is used.
9863 Specify that the target processor is the V850E. The preprocessor
9864 constant @samp{__v850e__} will be defined if this option is used.
9866 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
9867 are defined then a default target processor will be chosen and the
9868 relevant @samp{__v850*__} preprocessor constant will be defined.
9870 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
9871 defined, regardless of which processor variant is the target.
9873 @item -mdisable-callt
9874 @opindex mdisable-callt
9875 This option will suppress generation of the CALLT instruction for the
9876 v850e and v850e1 flavors of the v850 architecture. The default is
9877 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
9882 @subsection ARC Options
9885 These options are defined for ARC implementations:
9890 Compile code for little endian mode. This is the default.
9894 Compile code for big endian mode.
9897 @opindex mmangle-cpu
9898 Prepend the name of the cpu to all public symbol names.
9899 In multiple-processor systems, there are many ARC variants with different
9900 instruction and register set characteristics. This flag prevents code
9901 compiled for one cpu to be linked with code compiled for another.
9902 No facility exists for handling variants that are ``almost identical''.
9903 This is an all or nothing option.
9905 @item -mcpu=@var{cpu}
9907 Compile code for ARC variant @var{cpu}.
9908 Which variants are supported depend on the configuration.
9909 All variants support @option{-mcpu=base}, this is the default.
9911 @item -mtext=@var{text-section}
9912 @itemx -mdata=@var{data-section}
9913 @itemx -mrodata=@var{readonly-data-section}
9917 Put functions, data, and readonly data in @var{text-section},
9918 @var{data-section}, and @var{readonly-data-section} respectively
9919 by default. This can be overridden with the @code{section} attribute.
9920 @xref{Variable Attributes}.
9925 @subsection NS32K Options
9926 @cindex NS32K options
9928 These are the @samp{-m} options defined for the 32000 series. The default
9929 values for these options depends on which style of 32000 was selected when
9930 the compiler was configured; the defaults for the most common choices are
9938 Generate output for a 32032. This is the default
9939 when the compiler is configured for 32032 and 32016 based systems.
9945 Generate output for a 32332. This is the default
9946 when the compiler is configured for 32332-based systems.
9952 Generate output for a 32532. This is the default
9953 when the compiler is configured for 32532-based systems.
9957 Generate output containing 32081 instructions for floating point.
9958 This is the default for all systems.
9962 Generate output containing 32381 instructions for floating point. This
9963 also implies @option{-m32081}. The 32381 is only compatible with the 32332
9964 and 32532 cpus. This is the default for the pc532-netbsd configuration.
9968 Try and generate multiply-add floating point instructions @code{polyF}
9969 and @code{dotF}. This option is only available if the @option{-m32381}
9970 option is in effect. Using these instructions requires changes to
9971 register allocation which generally has a negative impact on
9972 performance. This option should only be enabled when compiling code
9973 particularly likely to make heavy use of multiply-add instructions.
9976 @opindex mnomulti-add
9977 Do not try and generate multiply-add floating point instructions
9978 @code{polyF} and @code{dotF}. This is the default on all platforms.
9981 @opindex msoft-float
9982 Generate output containing library calls for floating point.
9983 @strong{Warning:} the requisite libraries may not be available.
9985 @item -mieee-compare
9986 @itemx -mno-ieee-compare
9987 @opindex mieee-compare
9988 @opindex mno-ieee-compare
9989 Control whether or not the compiler uses IEEE floating point
9990 comparisons. These handle correctly the case where the result of a
9991 comparison is unordered.
9992 @strong{Warning:} the requisite kernel support may not be available.
9995 @opindex mnobitfield
9996 Do not use the bit-field instructions. On some machines it is faster to
9997 use shifting and masking operations. This is the default for the pc532.
10001 Do use the bit-field instructions. This is the default for all platforms
10006 Use a different function-calling convention, in which functions
10007 that take a fixed number of arguments return pop their
10008 arguments on return with the @code{ret} instruction.
10010 This calling convention is incompatible with the one normally
10011 used on Unix, so you cannot use it if you need to call libraries
10012 compiled with the Unix compiler.
10014 Also, you must provide function prototypes for all functions that
10015 take variable numbers of arguments (including @code{printf});
10016 otherwise incorrect code will be generated for calls to those
10019 In addition, seriously incorrect code will result if you call a
10020 function with too many arguments. (Normally, extra arguments are
10021 harmlessly ignored.)
10023 This option takes its name from the 680x0 @code{rtd} instruction.
10028 Use a different function-calling convention where the first two arguments
10029 are passed in registers.
10031 This calling convention is incompatible with the one normally
10032 used on Unix, so you cannot use it if you need to call libraries
10033 compiled with the Unix compiler.
10036 @opindex mnoregparam
10037 Do not pass any arguments in registers. This is the default for all
10042 It is OK to use the sb as an index register which is always loaded with
10043 zero. This is the default for the pc532-netbsd target.
10047 The sb register is not available for use or has not been initialized to
10048 zero by the run time system. This is the default for all targets except
10049 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10050 @option{-fpic} is set.
10054 Many ns32000 series addressing modes use displacements of up to 512MB@.
10055 If an address is above 512MB then displacements from zero can not be used.
10056 This option causes code to be generated which can be loaded above 512MB@.
10057 This may be useful for operating systems or ROM code.
10061 Assume code will be loaded in the first 512MB of virtual address space.
10062 This is the default for all platforms.
10068 @subsection AVR Options
10069 @cindex AVR Options
10071 These options are defined for AVR implementations:
10074 @item -mmcu=@var{mcu}
10076 Specify ATMEL AVR instruction set or MCU type.
10078 Instruction set avr1 is for the minimal AVR core, not supported by the C
10079 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10080 attiny11, attiny12, attiny15, attiny28).
10082 Instruction set avr2 (default) is for the classic AVR core with up to
10083 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10084 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10085 at90c8534, at90s8535).
10087 Instruction set avr3 is for the classic AVR core with up to 128K program
10088 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10090 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10091 memory space (MCU types: atmega8, atmega83, atmega85).
10093 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10094 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10095 atmega64, atmega128, at43usb355, at94k).
10099 Output instruction sizes to the asm file.
10101 @item -minit-stack=@var{N}
10102 @opindex minit-stack
10103 Specify the initial stack address, which may be a symbol or numeric value,
10104 @samp{__stack} is the default.
10106 @item -mno-interrupts
10107 @opindex mno-interrupts
10108 Generated code is not compatible with hardware interrupts.
10109 Code size will be smaller.
10111 @item -mcall-prologues
10112 @opindex mcall-prologues
10113 Functions prologues/epilogues expanded as call to appropriate
10114 subroutines. Code size will be smaller.
10116 @item -mno-tablejump
10117 @opindex mno-tablejump
10118 Do not generate tablejump insns which sometimes increase code size.
10121 @opindex mtiny-stack
10122 Change only the low 8 bits of the stack pointer.
10125 @node MCore Options
10126 @subsection MCore Options
10127 @cindex MCore options
10129 These are the @samp{-m} options defined for the Motorola M*Core
10135 @itemx -mno-hardlit
10137 @opindex mno-hardlit
10138 Inline constants into the code stream if it can be done in two
10139 instructions or less.
10145 Use the divide instruction. (Enabled by default).
10147 @item -mrelax-immediate
10148 @itemx -mno-relax-immediate
10149 @opindex mrelax-immediate
10150 @opindex mno-relax-immediate
10151 Allow arbitrary sized immediates in bit operations.
10153 @item -mwide-bitfields
10154 @itemx -mno-wide-bitfields
10155 @opindex mwide-bitfields
10156 @opindex mno-wide-bitfields
10157 Always treat bit-fields as int-sized.
10159 @item -m4byte-functions
10160 @itemx -mno-4byte-functions
10161 @opindex m4byte-functions
10162 @opindex mno-4byte-functions
10163 Force all functions to be aligned to a four byte boundary.
10165 @item -mcallgraph-data
10166 @itemx -mno-callgraph-data
10167 @opindex mcallgraph-data
10168 @opindex mno-callgraph-data
10169 Emit callgraph information.
10172 @itemx -mno-slow-bytes
10173 @opindex mslow-bytes
10174 @opindex mno-slow-bytes
10175 Prefer word access when reading byte quantities.
10177 @item -mlittle-endian
10178 @itemx -mbig-endian
10179 @opindex mlittle-endian
10180 @opindex mbig-endian
10181 Generate code for a little endian target.
10187 Generate code for the 210 processor.
10190 @node IA-64 Options
10191 @subsection IA-64 Options
10192 @cindex IA-64 Options
10194 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10198 @opindex mbig-endian
10199 Generate code for a big endian target. This is the default for HP-UX@.
10201 @item -mlittle-endian
10202 @opindex mlittle-endian
10203 Generate code for a little endian target. This is the default for AIX5
10209 @opindex mno-gnu-as
10210 Generate (or don't) code for the GNU assembler. This is the default.
10211 @c Also, this is the default if the configure option @option{--with-gnu-as}
10217 @opindex mno-gnu-ld
10218 Generate (or don't) code for the GNU linker. This is the default.
10219 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10224 Generate code that does not use a global pointer register. The result
10225 is not position independent code, and violates the IA-64 ABI@.
10227 @item -mvolatile-asm-stop
10228 @itemx -mno-volatile-asm-stop
10229 @opindex mvolatile-asm-stop
10230 @opindex mno-volatile-asm-stop
10231 Generate (or don't) a stop bit immediately before and after volatile asm
10236 Generate code that works around Itanium B step errata.
10238 @item -mregister-names
10239 @itemx -mno-register-names
10240 @opindex mregister-names
10241 @opindex mno-register-names
10242 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10243 the stacked registers. This may make assembler output more readable.
10249 Disable (or enable) optimizations that use the small data section. This may
10250 be useful for working around optimizer bugs.
10252 @item -mconstant-gp
10253 @opindex mconstant-gp
10254 Generate code that uses a single constant global pointer value. This is
10255 useful when compiling kernel code.
10259 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10260 This is useful when compiling firmware code.
10262 @item -minline-float-divide-min-latency
10263 @opindex minline-float-divide-min-latency
10264 Generate code for inline divides of floating point values
10265 using the minimum latency algorithm.
10267 @item -minline-float-divide-max-throughput
10268 @opindex minline-float-divide-max-throughput
10269 Generate code for inline divides of floating point values
10270 using the maximum throughput algorithm.
10272 @item -minline-int-divide-min-latency
10273 @opindex minline-int-divide-min-latency
10274 Generate code for inline divides of integer values
10275 using the minimum latency algorithm.
10277 @item -minline-int-divide-max-throughput
10278 @opindex minline-int-divide-max-throughput
10279 Generate code for inline divides of integer values
10280 using the maximum throughput algorithm.
10282 @item -mno-dwarf2-asm
10283 @itemx -mdwarf2-asm
10284 @opindex mno-dwarf2-asm
10285 @opindex mdwarf2-asm
10286 Don't (or do) generate assembler code for the DWARF2 line number debugging
10287 info. This may be useful when not using the GNU assembler.
10289 @item -mfixed-range=@var{register-range}
10290 @opindex mfixed-range
10291 Generate code treating the given register range as fixed registers.
10292 A fixed register is one that the register allocator can not use. This is
10293 useful when compiling kernel code. A register range is specified as
10294 two registers separated by a dash. Multiple register ranges can be
10295 specified separated by a comma.
10297 @item -mearly-stop-bits
10298 @itemx -mno-early-stop-bits
10299 @opindex mearly-stop-bits
10300 @opindex mno-early-stop-bits
10301 Allow stop bits to be placed earlier than immediately preceding the
10302 instruction that triggered the stop bit. This can improve instruction
10303 scheduling, but does not always do so.
10307 @subsection D30V Options
10308 @cindex D30V Options
10310 These @samp{-m} options are defined for D30V implementations:
10315 Link the @samp{.text}, @samp{.data}, @samp{.bss}, @samp{.strings},
10316 @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections into external
10317 memory, which starts at location @code{0x80000000}.
10320 @opindex mextmemory
10321 Same as the @option{-mextmem} switch.
10325 Link the @samp{.text} section into onchip text memory, which starts at
10326 location @code{0x0}. Also link @samp{.data}, @samp{.bss},
10327 @samp{.strings}, @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections
10328 into onchip data memory, which starts at location @code{0x20000000}.
10330 @item -mno-asm-optimize
10331 @itemx -masm-optimize
10332 @opindex mno-asm-optimize
10333 @opindex masm-optimize
10334 Disable (enable) passing @option{-O} to the assembler when optimizing.
10335 The assembler uses the @option{-O} option to automatically parallelize
10336 adjacent short instructions where possible.
10338 @item -mbranch-cost=@var{n}
10339 @opindex mbranch-cost
10340 Increase the internal costs of branches to @var{n}. Higher costs means
10341 that the compiler will issue more instructions to avoid doing a branch.
10344 @item -mcond-exec=@var{n}
10345 @opindex mcond-exec
10346 Specify the maximum number of conditionally executed instructions that
10347 replace a branch. The default is 4.
10350 @node S/390 and zSeries Options
10351 @subsection S/390 and zSeries Options
10352 @cindex S/390 and zSeries Options
10354 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
10358 @itemx -msoft-float
10359 @opindex mhard-float
10360 @opindex msoft-float
10361 Use (do not use) the hardware floating-point instructions and registers
10362 for floating-point operations. When @option{-msoft-float} is specified,
10363 functions in @file{libgcc.a} will be used to perform floating-point
10364 operations. When @option{-mhard-float} is specified, the compiler
10365 generates IEEE floating-point instructions. This is the default.
10368 @itemx -mno-backchain
10369 @opindex mbackchain
10370 @opindex mno-backchain
10371 Generate (or do not generate) code which maintains an explicit
10372 backchain within the stack frame that points to the caller's frame.
10373 This is currently needed to allow debugging. The default is to
10374 generate the backchain.
10377 @itemx -mno-small-exec
10378 @opindex msmall-exec
10379 @opindex mno-small-exec
10380 Generate (or do not generate) code using the @code{bras} instruction
10381 to do subroutine calls.
10382 This only works reliably if the total executable size does not
10383 exceed 64k. The default is to use the @code{basr} instruction instead,
10384 which does not have this limitation.
10390 When @option{-m31} is specified, generate code compliant to the
10391 Linux for S/390 ABI@. When @option{-m64} is specified, generate
10392 code compliant to the Linux for zSeries ABI@. This allows GCC in
10393 particular to generate 64-bit instructions. For the @samp{s390}
10394 targets, the default is @option{-m31}, while the @samp{s390x}
10395 targets default to @option{-m64}.
10401 When @option{-mzarch} is specified, generate code using the
10402 instructions available on z/Architecture.
10403 When @option{-mesa} is specified, generate code using the
10404 instructions available on ESA/390. Note that @option{-mesa} is
10405 not possible with @option{-m64}.
10406 When generating code compliant to the Linux for S/390 ABI,
10407 the default is @option{-mesa}. When generating code compliant
10408 to the Linux for zSeries ABI, the default is @option{-mzarch}.
10414 Generate (or do not generate) code using the @code{mvcle} instruction
10415 to perform block moves. When @option{-mno-mvcle} is specified,
10416 use a @code{mvc} loop instead. This is the default.
10422 Print (or do not print) additional debug information when compiling.
10423 The default is to not print debug information.
10425 @item -march=@var{cpu-type}
10427 Generate code that will run on @var{cpu-type}, which is the name of a system
10428 representing a certain processor type. Possible values for
10429 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
10430 When generating code using the instructions available on z/Architecture,
10431 the default is @option{-march=z900}. Otherwise, the default is
10432 @option{-march=g5}.
10434 @item -mtune=@var{cpu-type}
10436 Tune to @var{cpu-type} everything applicable about the generated code,
10437 except for the ABI and the set of available instructions.
10438 The list of @var{cpu-type} values is the same as for @option{-march}.
10439 The default is the value used for @option{-march}.
10442 @itemx -mno-fused-madd
10443 @opindex mfused-madd
10444 @opindex mno-fused-madd
10445 Generate code that uses (does not use) the floating point multiply and
10446 accumulate instructions. These instructions are generated by default if
10447 hardware floating point is used.
10451 @subsection CRIS Options
10452 @cindex CRIS Options
10454 These options are defined specifically for the CRIS ports.
10457 @item -march=@var{architecture-type}
10458 @itemx -mcpu=@var{architecture-type}
10461 Generate code for the specified architecture. The choices for
10462 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10463 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX.
10464 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10467 @item -mtune=@var{architecture-type}
10469 Tune to @var{architecture-type} everything applicable about the generated
10470 code, except for the ABI and the set of available instructions. The
10471 choices for @var{architecture-type} are the same as for
10472 @option{-march=@var{architecture-type}}.
10474 @item -mmax-stack-frame=@var{n}
10475 @opindex mmax-stack-frame
10476 Warn when the stack frame of a function exceeds @var{n} bytes.
10478 @item -melinux-stacksize=@var{n}
10479 @opindex melinux-stacksize
10480 Only available with the @samp{cris-axis-aout} target. Arranges for
10481 indications in the program to the kernel loader that the stack of the
10482 program should be set to @var{n} bytes.
10488 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10489 @option{-march=v3} and @option{-march=v8} respectively.
10493 Enable CRIS-specific verbose debug-related information in the assembly
10494 code. This option also has the effect to turn off the @samp{#NO_APP}
10495 formatted-code indicator to the assembler at the beginning of the
10500 Do not use condition-code results from previous instruction; always emit
10501 compare and test instructions before use of condition codes.
10503 @item -mno-side-effects
10504 @opindex mno-side-effects
10505 Do not emit instructions with side-effects in addressing modes other than
10508 @item -mstack-align
10509 @itemx -mno-stack-align
10510 @itemx -mdata-align
10511 @itemx -mno-data-align
10512 @itemx -mconst-align
10513 @itemx -mno-const-align
10514 @opindex mstack-align
10515 @opindex mno-stack-align
10516 @opindex mdata-align
10517 @opindex mno-data-align
10518 @opindex mconst-align
10519 @opindex mno-const-align
10520 These options (no-options) arranges (eliminate arrangements) for the
10521 stack-frame, individual data and constants to be aligned for the maximum
10522 single data access size for the chosen CPU model. The default is to
10523 arrange for 32-bit alignment. ABI details such as structure layout are
10524 not affected by these options.
10532 Similar to the stack- data- and const-align options above, these options
10533 arrange for stack-frame, writable data and constants to all be 32-bit,
10534 16-bit or 8-bit aligned. The default is 32-bit alignment.
10536 @item -mno-prologue-epilogue
10537 @itemx -mprologue-epilogue
10538 @opindex mno-prologue-epilogue
10539 @opindex mprologue-epilogue
10540 With @option{-mno-prologue-epilogue}, the normal function prologue and
10541 epilogue that sets up the stack-frame are omitted and no return
10542 instructions or return sequences are generated in the code. Use this
10543 option only together with visual inspection of the compiled code: no
10544 warnings or errors are generated when call-saved registers must be saved,
10545 or storage for local variable needs to be allocated.
10549 @opindex mno-gotplt
10551 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10552 instruction sequences that load addresses for functions from the PLT part
10553 of the GOT rather than (traditional on other architectures) calls to the
10554 PLT. The default is @option{-mgotplt}.
10558 Legacy no-op option only recognized with the cris-axis-aout target.
10562 Legacy no-op option only recognized with the cris-axis-elf and
10563 cris-axis-linux-gnu targets.
10567 Only recognized with the cris-axis-aout target, where it selects a
10568 GNU/linux-like multilib, include files and instruction set for
10569 @option{-march=v8}.
10573 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10577 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
10578 to link with input-output functions from a simulator library. Code,
10579 initialized data and zero-initialized data are allocated consecutively.
10583 Like @option{-sim}, but pass linker options to locate initialized data at
10584 0x40000000 and zero-initialized data at 0x80000000.
10588 @subsection MMIX Options
10589 @cindex MMIX Options
10591 These options are defined for the MMIX:
10595 @itemx -mno-libfuncs
10597 @opindex mno-libfuncs
10598 Specify that intrinsic library functions are being compiled, passing all
10599 values in registers, no matter the size.
10602 @itemx -mno-epsilon
10604 @opindex mno-epsilon
10605 Generate floating-point comparison instructions that compare with respect
10606 to the @code{rE} epsilon register.
10608 @item -mabi=mmixware
10610 @opindex mabi-mmixware
10612 Generate code that passes function parameters and return values that (in
10613 the called function) are seen as registers @code{$0} and up, as opposed to
10614 the GNU ABI which uses global registers @code{$231} and up.
10616 @item -mzero-extend
10617 @itemx -mno-zero-extend
10618 @opindex mzero-extend
10619 @opindex mno-zero-extend
10620 When reading data from memory in sizes shorter than 64 bits, use (do not
10621 use) zero-extending load instructions by default, rather than
10622 sign-extending ones.
10625 @itemx -mno-knuthdiv
10627 @opindex mno-knuthdiv
10628 Make the result of a division yielding a remainder have the same sign as
10629 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10630 remainder follows the sign of the dividend. Both methods are
10631 arithmetically valid, the latter being almost exclusively used.
10633 @item -mtoplevel-symbols
10634 @itemx -mno-toplevel-symbols
10635 @opindex mtoplevel-symbols
10636 @opindex mno-toplevel-symbols
10637 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10638 code can be used with the @code{PREFIX} assembly directive.
10642 Generate an executable in the ELF format, rather than the default
10643 @samp{mmo} format used by the @command{mmix} simulator.
10645 @item -mbranch-predict
10646 @itemx -mno-branch-predict
10647 @opindex mbranch-predict
10648 @opindex mno-branch-predict
10649 Use (do not use) the probable-branch instructions, when static branch
10650 prediction indicates a probable branch.
10652 @item -mbase-addresses
10653 @itemx -mno-base-addresses
10654 @opindex mbase-addresses
10655 @opindex mno-base-addresses
10656 Generate (do not generate) code that uses @emph{base addresses}. Using a
10657 base address automatically generates a request (handled by the assembler
10658 and the linker) for a constant to be set up in a global register. The
10659 register is used for one or more base address requests within the range 0
10660 to 255 from the value held in the register. The generally leads to short
10661 and fast code, but the number of different data items that can be
10662 addressed is limited. This means that a program that uses lots of static
10663 data may require @option{-mno-base-addresses}.
10665 @item -msingle-exit
10666 @itemx -mno-single-exit
10667 @opindex msingle-exit
10668 @opindex mno-single-exit
10669 Force (do not force) generated code to have a single exit point in each
10673 @node PDP-11 Options
10674 @subsection PDP-11 Options
10675 @cindex PDP-11 Options
10677 These options are defined for the PDP-11:
10682 Use hardware FPP floating point. This is the default. (FIS floating
10683 point on the PDP-11/40 is not supported.)
10686 @opindex msoft-float
10687 Do not use hardware floating point.
10691 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10695 Return floating-point results in memory. This is the default.
10699 Generate code for a PDP-11/40.
10703 Generate code for a PDP-11/45. This is the default.
10707 Generate code for a PDP-11/10.
10709 @item -mbcopy-builtin
10710 @opindex bcopy-builtin
10711 Use inline @code{movstrhi} patterns for copying memory. This is the
10716 Do not use inline @code{movstrhi} patterns for copying memory.
10722 Use 16-bit @code{int}. This is the default.
10728 Use 32-bit @code{int}.
10731 @itemx -mno-float32
10733 @opindex mno-float32
10734 Use 64-bit @code{float}. This is the default.
10739 @opindex mno-float64
10740 Use 32-bit @code{float}.
10744 Use @code{abshi2} pattern. This is the default.
10748 Do not use @code{abshi2} pattern.
10750 @item -mbranch-expensive
10751 @opindex mbranch-expensive
10752 Pretend that branches are expensive. This is for experimenting with
10753 code generation only.
10755 @item -mbranch-cheap
10756 @opindex mbranch-cheap
10757 Do not pretend that branches are expensive. This is the default.
10761 Generate code for a system with split I&D.
10765 Generate code for a system without split I&D. This is the default.
10769 Use Unix assembler syntax. This is the default when configured for
10770 @samp{pdp11-*-bsd}.
10774 Use DEC assembler syntax. This is the default when configured for any
10775 PDP-11 target other than @samp{pdp11-*-bsd}.
10778 @node Xstormy16 Options
10779 @subsection Xstormy16 Options
10780 @cindex Xstormy16 Options
10782 These options are defined for Xstormy16:
10787 Choose startup files and linker script suitable for the simulator.
10791 @subsection FRV Options
10792 @cindex FRV Options
10798 Only use the first 32 general purpose registers.
10803 Use all 64 general purpose registers.
10808 Use only the first 32 floating point registers.
10813 Use all 64 floating point registers
10816 @opindex mhard-float
10818 Use hardware instructions for floating point operations.
10821 @opindex msoft-float
10823 Use library routines for floating point operations.
10828 Dynamically allocate condition code registers.
10833 Do not try to dynamically allocate condition code registers, only
10834 use @code{icc0} and @code{fcc0}.
10839 Change ABI to use double word insns.
10844 Do not use double word instructions.
10849 Use floating point double instructions.
10852 @opindex mno-double
10854 Do not use floating point double instructions.
10859 Use media instructions.
10864 Do not use media instructions.
10869 Use multiply and add/subtract instructions.
10872 @opindex mno-muladd
10874 Do not use multiply and add/subtract instructions.
10876 @item -mlibrary-pic
10877 @opindex mlibrary-pic
10879 Enable PIC support for building libraries
10884 Use only the first four media accumulator registers.
10889 Use all eight media accumulator registers.
10894 Pack VLIW instructions.
10899 Do not pack VLIW instructions.
10902 @opindex mno-eflags
10904 Do not mark ABI switches in e_flags.
10907 @opindex mcond-move
10909 Enable the use of conditional-move instructions (default).
10911 This switch is mainly for debugging the compiler and will likely be removed
10912 in a future version.
10914 @item -mno-cond-move
10915 @opindex mno-cond-move
10917 Disable the use of conditional-move instructions.
10919 This switch is mainly for debugging the compiler and will likely be removed
10920 in a future version.
10925 Enable the use of conditional set instructions (default).
10927 This switch is mainly for debugging the compiler and will likely be removed
10928 in a future version.
10933 Disable the use of conditional set instructions.
10935 This switch is mainly for debugging the compiler and will likely be removed
10936 in a future version.
10939 @opindex mcond-exec
10941 Enable the use of conditional execution (default).
10943 This switch is mainly for debugging the compiler and will likely be removed
10944 in a future version.
10946 @item -mno-cond-exec
10947 @opindex mno-cond-exec
10949 Disable the use of conditional execution.
10951 This switch is mainly for debugging the compiler and will likely be removed
10952 in a future version.
10954 @item -mvliw-branch
10955 @opindex mvliw-branch
10957 Run a pass to pack branches into VLIW instructions (default).
10959 This switch is mainly for debugging the compiler and will likely be removed
10960 in a future version.
10962 @item -mno-vliw-branch
10963 @opindex mno-vliw-branch
10965 Do not run a pass to pack branches into VLIW instructions.
10967 This switch is mainly for debugging the compiler and will likely be removed
10968 in a future version.
10970 @item -mmulti-cond-exec
10971 @opindex mmulti-cond-exec
10973 Enable optimization of @code{&&} and @code{||} in conditional execution
10976 This switch is mainly for debugging the compiler and will likely be removed
10977 in a future version.
10979 @item -mno-multi-cond-exec
10980 @opindex mno-multi-cond-exec
10982 Disable optimization of @code{&&} and @code{||} in conditional execution.
10984 This switch is mainly for debugging the compiler and will likely be removed
10985 in a future version.
10987 @item -mnested-cond-exec
10988 @opindex mnested-cond-exec
10990 Enable nested conditional execution optimizations (default).
10992 This switch is mainly for debugging the compiler and will likely be removed
10993 in a future version.
10995 @item -mno-nested-cond-exec
10996 @opindex mno-nested-cond-exec
10998 Disable nested conditional execution optimizations.
11000 This switch is mainly for debugging the compiler and will likely be removed
11001 in a future version.
11003 @item -mtomcat-stats
11004 @opindex mtomcat-stats
11006 Cause gas to print out tomcat statistics.
11008 @item -mcpu=@var{cpu}
11011 Select the processor type for which to generate code. Possible values are
11012 @samp{simple}, @samp{tomcat}, @samp{fr500}, @samp{fr400}, @samp{fr300},
11017 @node Xtensa Options
11018 @subsection Xtensa Options
11019 @cindex Xtensa Options
11021 These options are supported for Xtensa targets:
11025 @itemx -mno-const16
11027 @opindex mno-const16
11028 Enable or disable use of @code{CONST16} instructions for loading
11029 constant values. The @code{CONST16} instruction is currently not a
11030 standard option from Tensilica. When enabled, @code{CONST16}
11031 instructions are always used in place of the standard @code{L32R}
11032 instructions. The use of @code{CONST16} is enabled by default only if
11033 the @code{L32R} instruction is not available.
11036 @itemx -mno-fused-madd
11037 @opindex mfused-madd
11038 @opindex mno-fused-madd
11039 Enable or disable use of fused multiply/add and multiply/subtract
11040 instructions in the floating-point option. This has no effect if the
11041 floating-point option is not also enabled. Disabling fused multiply/add
11042 and multiply/subtract instructions forces the compiler to use separate
11043 instructions for the multiply and add/subtract operations. This may be
11044 desirable in some cases where strict IEEE 754-compliant results are
11045 required: the fused multiply add/subtract instructions do not round the
11046 intermediate result, thereby producing results with @emph{more} bits of
11047 precision than specified by the IEEE standard. Disabling fused multiply
11048 add/subtract instructions also ensures that the program output is not
11049 sensitive to the compiler's ability to combine multiply and add/subtract
11052 @item -mtext-section-literals
11053 @itemx -mno-text-section-literals
11054 @opindex mtext-section-literals
11055 @opindex mno-text-section-literals
11056 Control the treatment of literal pools. The default is
11057 @option{-mno-text-section-literals}, which places literals in a separate
11058 section in the output file. This allows the literal pool to be placed
11059 in a data RAM/ROM, and it also allows the linker to combine literal
11060 pools from separate object files to remove redundant literals and
11061 improve code size. With @option{-mtext-section-literals}, the literals
11062 are interspersed in the text section in order to keep them as close as
11063 possible to their references. This may be necessary for large assembly
11066 @item -mtarget-align
11067 @itemx -mno-target-align
11068 @opindex mtarget-align
11069 @opindex mno-target-align
11070 When this option is enabled, GCC instructs the assembler to
11071 automatically align instructions to reduce branch penalties at the
11072 expense of some code density. The assembler attempts to widen density
11073 instructions to align branch targets and the instructions following call
11074 instructions. If there are not enough preceding safe density
11075 instructions to align a target, no widening will be performed. The
11076 default is @option{-mtarget-align}. These options do not affect the
11077 treatment of auto-aligned instructions like @code{LOOP}, which the
11078 assembler will always align, either by widening density instructions or
11079 by inserting no-op instructions.
11082 @itemx -mno-longcalls
11083 @opindex mlongcalls
11084 @opindex mno-longcalls
11085 When this option is enabled, GCC instructs the assembler to translate
11086 direct calls to indirect calls unless it can determine that the target
11087 of a direct call is in the range allowed by the call instruction. This
11088 translation typically occurs for calls to functions in other source
11089 files. Specifically, the assembler translates a direct @code{CALL}
11090 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
11091 The default is @option{-mno-longcalls}. This option should be used in
11092 programs where the call target can potentially be out of range. This
11093 option is implemented in the assembler, not the compiler, so the
11094 assembly code generated by GCC will still show direct call
11095 instructions---look at the disassembled object code to see the actual
11096 instructions. Note that the assembler will use an indirect call for
11097 every cross-file call, not just those that really will be out of range.
11100 @node Code Gen Options
11101 @section Options for Code Generation Conventions
11102 @cindex code generation conventions
11103 @cindex options, code generation
11104 @cindex run-time options
11106 These machine-independent options control the interface conventions
11107 used in code generation.
11109 Most of them have both positive and negative forms; the negative form
11110 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
11111 one of the forms is listed---the one which is not the default. You
11112 can figure out the other form by either removing @samp{no-} or adding
11116 @item -fbounds-check
11117 @opindex fbounds-check
11118 For front-ends that support it, generate additional code to check that
11119 indices used to access arrays are within the declared range. This is
11120 currently only supported by the Java and Fortran 77 front-ends, where
11121 this option defaults to true and false respectively.
11125 This option generates traps for signed overflow on addition, subtraction,
11126 multiplication operations.
11130 This option instructs the compiler to assume that signed arithmetic
11131 overflow of addition, subtraction and multiplication wraps around
11132 using twos-complement representation. This flag enables some optimizations
11133 and disables other. This option is enabled by default for the Java
11134 front-end, as required by the Java language specification.
11137 @opindex fexceptions
11138 Enable exception handling. Generates extra code needed to propagate
11139 exceptions. For some targets, this implies GCC will generate frame
11140 unwind information for all functions, which can produce significant data
11141 size overhead, although it does not affect execution. If you do not
11142 specify this option, GCC will enable it by default for languages like
11143 C++ which normally require exception handling, and disable it for
11144 languages like C that do not normally require it. However, you may need
11145 to enable this option when compiling C code that needs to interoperate
11146 properly with exception handlers written in C++. You may also wish to
11147 disable this option if you are compiling older C++ programs that don't
11148 use exception handling.
11150 @item -fnon-call-exceptions
11151 @opindex fnon-call-exceptions
11152 Generate code that allows trapping instructions to throw exceptions.
11153 Note that this requires platform-specific runtime support that does
11154 not exist everywhere. Moreover, it only allows @emph{trapping}
11155 instructions to throw exceptions, i.e.@: memory references or floating
11156 point instructions. It does not allow exceptions to be thrown from
11157 arbitrary signal handlers such as @code{SIGALRM}.
11159 @item -funwind-tables
11160 @opindex funwind-tables
11161 Similar to @option{-fexceptions}, except that it will just generate any needed
11162 static data, but will not affect the generated code in any other way.
11163 You will normally not enable this option; instead, a language processor
11164 that needs this handling would enable it on your behalf.
11166 @item -fasynchronous-unwind-tables
11167 @opindex funwind-tables
11168 Generate unwind table in dwarf2 format, if supported by target machine. The
11169 table is exact at each instruction boundary, so it can be used for stack
11170 unwinding from asynchronous events (such as debugger or garbage collector).
11172 @item -fpcc-struct-return
11173 @opindex fpcc-struct-return
11174 Return ``short'' @code{struct} and @code{union} values in memory like
11175 longer ones, rather than in registers. This convention is less
11176 efficient, but it has the advantage of allowing intercallability between
11177 GCC-compiled files and files compiled with other compilers, particularly
11178 the Portable C Compiler (pcc).
11180 The precise convention for returning structures in memory depends
11181 on the target configuration macros.
11183 Short structures and unions are those whose size and alignment match
11184 that of some integer type.
11186 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11187 switch is not binary compatible with code compiled with the
11188 @option{-freg-struct-return} switch.
11189 Use it to conform to a non-default application binary interface.
11191 @item -freg-struct-return
11192 @opindex freg-struct-return
11193 Return @code{struct} and @code{union} values in registers when possible.
11194 This is more efficient for small structures than
11195 @option{-fpcc-struct-return}.
11197 If you specify neither @option{-fpcc-struct-return} nor
11198 @option{-freg-struct-return}, GCC defaults to whichever convention is
11199 standard for the target. If there is no standard convention, GCC
11200 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11201 the principal compiler. In those cases, we can choose the standard, and
11202 we chose the more efficient register return alternative.
11204 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11205 switch is not binary compatible with code compiled with the
11206 @option{-fpcc-struct-return} switch.
11207 Use it to conform to a non-default application binary interface.
11209 @item -fshort-enums
11210 @opindex fshort-enums
11211 Allocate to an @code{enum} type only as many bytes as it needs for the
11212 declared range of possible values. Specifically, the @code{enum} type
11213 will be equivalent to the smallest integer type which has enough room.
11215 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11216 code that is not binary compatible with code generated without that switch.
11217 Use it to conform to a non-default application binary interface.
11219 @item -fshort-double
11220 @opindex fshort-double
11221 Use the same size for @code{double} as for @code{float}.
11223 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
11224 code that is not binary compatible with code generated without that switch.
11225 Use it to conform to a non-default application binary interface.
11227 @item -fshort-wchar
11228 @opindex fshort-wchar
11229 Override the underlying type for @samp{wchar_t} to be @samp{short
11230 unsigned int} instead of the default for the target. This option is
11231 useful for building programs to run under WINE@.
11233 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11234 code that is not binary compatible with code generated without that switch.
11235 Use it to conform to a non-default application binary interface.
11237 @item -fshared-data
11238 @opindex fshared-data
11239 Requests that the data and non-@code{const} variables of this
11240 compilation be shared data rather than private data. The distinction
11241 makes sense only on certain operating systems, where shared data is
11242 shared between processes running the same program, while private data
11243 exists in one copy per process.
11246 @opindex fno-common
11247 In C, allocate even uninitialized global variables in the data section of the
11248 object file, rather than generating them as common blocks. This has the
11249 effect that if the same variable is declared (without @code{extern}) in
11250 two different compilations, you will get an error when you link them.
11251 The only reason this might be useful is if you wish to verify that the
11252 program will work on other systems which always work this way.
11256 Ignore the @samp{#ident} directive.
11258 @item -finhibit-size-directive
11259 @opindex finhibit-size-directive
11260 Don't output a @code{.size} assembler directive, or anything else that
11261 would cause trouble if the function is split in the middle, and the
11262 two halves are placed at locations far apart in memory. This option is
11263 used when compiling @file{crtstuff.c}; you should not need to use it
11266 @item -fverbose-asm
11267 @opindex fverbose-asm
11268 Put extra commentary information in the generated assembly code to
11269 make it more readable. This option is generally only of use to those
11270 who actually need to read the generated assembly code (perhaps while
11271 debugging the compiler itself).
11273 @option{-fno-verbose-asm}, the default, causes the
11274 extra information to be omitted and is useful when comparing two assembler
11279 @cindex global offset table
11281 Generate position-independent code (PIC) suitable for use in a shared
11282 library, if supported for the target machine. Such code accesses all
11283 constant addresses through a global offset table (GOT)@. The dynamic
11284 loader resolves the GOT entries when the program starts (the dynamic
11285 loader is not part of GCC; it is part of the operating system). If
11286 the GOT size for the linked executable exceeds a machine-specific
11287 maximum size, you get an error message from the linker indicating that
11288 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11289 instead. (These maximums are 16k on the m88k, 8k on the SPARC, and 32k
11290 on the m68k and RS/6000. The 386 has no such limit.)
11292 Position-independent code requires special support, and therefore works
11293 only on certain machines. For the 386, GCC supports PIC for System V
11294 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11295 position-independent.
11299 If supported for the target machine, emit position-independent code,
11300 suitable for dynamic linking and avoiding any limit on the size of the
11301 global offset table. This option makes a difference on the m68k, m88k,
11304 Position-independent code requires special support, and therefore works
11305 only on certain machines.
11311 These options are similar to @option{-fpic} and @option{-fPIC}, but
11312 generated position independent code can be only linked into executables.
11313 Usually these options are used when @option{-pie} GCC option will be
11314 used during linking.
11316 @item -ffixed-@var{reg}
11318 Treat the register named @var{reg} as a fixed register; generated code
11319 should never refer to it (except perhaps as a stack pointer, frame
11320 pointer or in some other fixed role).
11322 @var{reg} must be the name of a register. The register names accepted
11323 are machine-specific and are defined in the @code{REGISTER_NAMES}
11324 macro in the machine description macro file.
11326 This flag does not have a negative form, because it specifies a
11329 @item -fcall-used-@var{reg}
11330 @opindex fcall-used
11331 Treat the register named @var{reg} as an allocable register that is
11332 clobbered by function calls. It may be allocated for temporaries or
11333 variables that do not live across a call. Functions compiled this way
11334 will not save and restore the register @var{reg}.
11336 It is an error to used this flag with the frame pointer or stack pointer.
11337 Use of this flag for other registers that have fixed pervasive roles in
11338 the machine's execution model will produce disastrous results.
11340 This flag does not have a negative form, because it specifies a
11343 @item -fcall-saved-@var{reg}
11344 @opindex fcall-saved
11345 Treat the register named @var{reg} as an allocable register saved by
11346 functions. It may be allocated even for temporaries or variables that
11347 live across a call. Functions compiled this way will save and restore
11348 the register @var{reg} if they use it.
11350 It is an error to used this flag with the frame pointer or stack pointer.
11351 Use of this flag for other registers that have fixed pervasive roles in
11352 the machine's execution model will produce disastrous results.
11354 A different sort of disaster will result from the use of this flag for
11355 a register in which function values may be returned.
11357 This flag does not have a negative form, because it specifies a
11360 @item -fpack-struct
11361 @opindex fpack-struct
11362 Pack all structure members together without holes.
11364 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11365 code that is not binary compatible with code generated without that switch.
11366 Additionally, it makes the code suboptimal.
11367 Use it to conform to a non-default application binary interface.
11369 @item -finstrument-functions
11370 @opindex finstrument-functions
11371 Generate instrumentation calls for entry and exit to functions. Just
11372 after function entry and just before function exit, the following
11373 profiling functions will be called with the address of the current
11374 function and its call site. (On some platforms,
11375 @code{__builtin_return_address} does not work beyond the current
11376 function, so the call site information may not be available to the
11377 profiling functions otherwise.)
11380 void __cyg_profile_func_enter (void *this_fn,
11382 void __cyg_profile_func_exit (void *this_fn,
11386 The first argument is the address of the start of the current function,
11387 which may be looked up exactly in the symbol table.
11389 This instrumentation is also done for functions expanded inline in other
11390 functions. The profiling calls will indicate where, conceptually, the
11391 inline function is entered and exited. This means that addressable
11392 versions of such functions must be available. If all your uses of a
11393 function are expanded inline, this may mean an additional expansion of
11394 code size. If you use @samp{extern inline} in your C code, an
11395 addressable version of such functions must be provided. (This is
11396 normally the case anyways, but if you get lucky and the optimizer always
11397 expands the functions inline, you might have gotten away without
11398 providing static copies.)
11400 A function may be given the attribute @code{no_instrument_function}, in
11401 which case this instrumentation will not be done. This can be used, for
11402 example, for the profiling functions listed above, high-priority
11403 interrupt routines, and any functions from which the profiling functions
11404 cannot safely be called (perhaps signal handlers, if the profiling
11405 routines generate output or allocate memory).
11407 @item -fstack-check
11408 @opindex fstack-check
11409 Generate code to verify that you do not go beyond the boundary of the
11410 stack. You should specify this flag if you are running in an
11411 environment with multiple threads, but only rarely need to specify it in
11412 a single-threaded environment since stack overflow is automatically
11413 detected on nearly all systems if there is only one stack.
11415 Note that this switch does not actually cause checking to be done; the
11416 operating system must do that. The switch causes generation of code
11417 to ensure that the operating system sees the stack being extended.
11419 @item -fstack-limit-register=@var{reg}
11420 @itemx -fstack-limit-symbol=@var{sym}
11421 @itemx -fno-stack-limit
11422 @opindex fstack-limit-register
11423 @opindex fstack-limit-symbol
11424 @opindex fno-stack-limit
11425 Generate code to ensure that the stack does not grow beyond a certain value,
11426 either the value of a register or the address of a symbol. If the stack
11427 would grow beyond the value, a signal is raised. For most targets,
11428 the signal is raised before the stack overruns the boundary, so
11429 it is possible to catch the signal without taking special precautions.
11431 For instance, if the stack starts at absolute address @samp{0x80000000}
11432 and grows downwards, you can use the flags
11433 @option{-fstack-limit-symbol=__stack_limit} and
11434 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11435 of 128KB@. Note that this may only work with the GNU linker.
11437 @cindex aliasing of parameters
11438 @cindex parameters, aliased
11439 @item -fargument-alias
11440 @itemx -fargument-noalias
11441 @itemx -fargument-noalias-global
11442 @opindex fargument-alias
11443 @opindex fargument-noalias
11444 @opindex fargument-noalias-global
11445 Specify the possible relationships among parameters and between
11446 parameters and global data.
11448 @option{-fargument-alias} specifies that arguments (parameters) may
11449 alias each other and may alias global storage.@*
11450 @option{-fargument-noalias} specifies that arguments do not alias
11451 each other, but may alias global storage.@*
11452 @option{-fargument-noalias-global} specifies that arguments do not
11453 alias each other and do not alias global storage.
11455 Each language will automatically use whatever option is required by
11456 the language standard. You should not need to use these options yourself.
11458 @item -fleading-underscore
11459 @opindex fleading-underscore
11460 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11461 change the way C symbols are represented in the object file. One use
11462 is to help link with legacy assembly code.
11464 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11465 generate code that is not binary compatible with code generated without that
11466 switch. Use it to conform to a non-default application binary interface.
11467 Not all targets provide complete support for this switch.
11469 @item -ftls-model=@var{model}
11470 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11471 The @var{model} argument should be one of @code{global-dynamic},
11472 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
11474 The default without @option{-fpic} is @code{initial-exec}; with
11475 @option{-fpic} the default is @code{global-dynamic}.
11480 @node Environment Variables
11481 @section Environment Variables Affecting GCC
11482 @cindex environment variables
11484 @c man begin ENVIRONMENT
11485 This section describes several environment variables that affect how GCC
11486 operates. Some of them work by specifying directories or prefixes to use
11487 when searching for various kinds of files. Some are used to specify other
11488 aspects of the compilation environment.
11490 Note that you can also specify places to search using options such as
11491 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
11492 take precedence over places specified using environment variables, which
11493 in turn take precedence over those specified by the configuration of GCC@.
11494 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
11495 GNU Compiler Collection (GCC) Internals}.
11500 @c @itemx LC_COLLATE
11502 @c @itemx LC_MONETARY
11503 @c @itemx LC_NUMERIC
11508 @c @findex LC_COLLATE
11509 @findex LC_MESSAGES
11510 @c @findex LC_MONETARY
11511 @c @findex LC_NUMERIC
11515 These environment variables control the way that GCC uses
11516 localization information that allow GCC to work with different
11517 national conventions. GCC inspects the locale categories
11518 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
11519 so. These locale categories can be set to any value supported by your
11520 installation. A typical value is @samp{en_UK} for English in the United
11523 The @env{LC_CTYPE} environment variable specifies character
11524 classification. GCC uses it to determine the character boundaries in
11525 a string; this is needed for some multibyte encodings that contain quote
11526 and escape characters that would otherwise be interpreted as a string
11529 The @env{LC_MESSAGES} environment variable specifies the language to
11530 use in diagnostic messages.
11532 If the @env{LC_ALL} environment variable is set, it overrides the value
11533 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
11534 and @env{LC_MESSAGES} default to the value of the @env{LANG}
11535 environment variable. If none of these variables are set, GCC
11536 defaults to traditional C English behavior.
11540 If @env{TMPDIR} is set, it specifies the directory to use for temporary
11541 files. GCC uses temporary files to hold the output of one stage of
11542 compilation which is to be used as input to the next stage: for example,
11543 the output of the preprocessor, which is the input to the compiler
11546 @item GCC_EXEC_PREFIX
11547 @findex GCC_EXEC_PREFIX
11548 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
11549 names of the subprograms executed by the compiler. No slash is added
11550 when this prefix is combined with the name of a subprogram, but you can
11551 specify a prefix that ends with a slash if you wish.
11553 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
11554 an appropriate prefix to use based on the pathname it was invoked with.
11556 If GCC cannot find the subprogram using the specified prefix, it
11557 tries looking in the usual places for the subprogram.
11559 The default value of @env{GCC_EXEC_PREFIX} is
11560 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
11561 of @code{prefix} when you ran the @file{configure} script.
11563 Other prefixes specified with @option{-B} take precedence over this prefix.
11565 This prefix is also used for finding files such as @file{crt0.o} that are
11568 In addition, the prefix is used in an unusual way in finding the
11569 directories to search for header files. For each of the standard
11570 directories whose name normally begins with @samp{/usr/local/lib/gcc}
11571 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
11572 replacing that beginning with the specified prefix to produce an
11573 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
11574 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
11575 These alternate directories are searched first; the standard directories
11578 @item COMPILER_PATH
11579 @findex COMPILER_PATH
11580 The value of @env{COMPILER_PATH} is a colon-separated list of
11581 directories, much like @env{PATH}. GCC tries the directories thus
11582 specified when searching for subprograms, if it can't find the
11583 subprograms using @env{GCC_EXEC_PREFIX}.
11586 @findex LIBRARY_PATH
11587 The value of @env{LIBRARY_PATH} is a colon-separated list of
11588 directories, much like @env{PATH}. When configured as a native compiler,
11589 GCC tries the directories thus specified when searching for special
11590 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
11591 using GCC also uses these directories when searching for ordinary
11592 libraries for the @option{-l} option (but directories specified with
11593 @option{-L} come first).
11597 @cindex locale definition
11598 This variable is used to pass locale information to the compiler. One way in
11599 which this information is used is to determine the character set to be used
11600 when character literals, string literals and comments are parsed in C and C++.
11601 When the compiler is configured to allow multibyte characters,
11602 the following values for @env{LANG} are recognized:
11606 Recognize JIS characters.
11608 Recognize SJIS characters.
11610 Recognize EUCJP characters.
11613 If @env{LANG} is not defined, or if it has some other value, then the
11614 compiler will use mblen and mbtowc as defined by the default locale to
11615 recognize and translate multibyte characters.
11619 Some additional environments variables affect the behavior of the
11622 @include cppenv.texi
11626 @node Precompiled Headers
11627 @section Using Precompiled Headers
11628 @cindex precompiled headers
11629 @cindex speed of compilation
11631 Often large projects have many header files that are included in every
11632 source file. The time the compiler takes to process these header files
11633 over and over again can account for nearly all of the time required to
11634 build the project. To make builds faster, GCC allows users to
11635 `precompile' a header file; then, if builds can use the precompiled
11636 header file they will be much faster.
11638 To create a precompiled header file, simply compile it as you would any
11639 other file, if necessary using the @option{-x} option to make the driver
11640 treat it as a C or C++ header file. You will probably want to use a
11641 tool like @command{make} to keep the precompiled header up-to-date when
11642 the headers it contains change.
11644 A precompiled header file will be searched for when @code{#include} is
11645 seen in the compilation. As it searches for the included file
11646 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
11647 compiler looks for a precompiled header in each directory just before it
11648 looks for the include file in that directory. The name searched for is
11649 the name specified in the @code{#include} with @samp{.gch} appended. If
11650 the precompiled header file can't be used, it is ignored.
11652 For instance, if you have @code{#include "all.h"}, and you have
11653 @file{all.h.gch} in the same directory as @file{all.h}, then the
11654 precompiled header file will be used if possible, and the original
11655 header will be used otherwise.
11657 Alternatively, you might decide to put the precompiled header file in a
11658 directory and use @option{-I} to ensure that directory is searched
11659 before (or instead of) the directory containing the original header.
11660 Then, if you want to check that the precompiled header file is always
11661 used, you can put a file of the same name as the original header in this
11662 directory containing an @code{#error} command.
11664 This also works with @option{-include}. So yet another way to use
11665 precompiled headers, good for projects not designed with precompiled
11666 header files in mind, is to simply take most of the header files used by
11667 a project, include them from another header file, precompile that header
11668 file, and @option{-include} the precompiled header. If the header files
11669 have guards against multiple inclusion, they will be skipped because
11670 they've already been included (in the precompiled header).
11672 If you need to precompile the same header file for different
11673 languages, targets, or compiler options, you can instead make a
11674 @emph{directory} named like @file{all.h.gch}, and put each precompiled
11675 header in the directory. (It doesn't matter what you call the files
11676 in the directory, every precompiled header in the directory will be
11677 considered.) The first precompiled header encountered in the
11678 directory that is valid for this compilation will be used; they're
11679 searched in no particular order.
11681 There are many other possibilities, limited only by your imagination,
11682 good sense, and the constraints of your build system.
11684 A precompiled header file can be used only when these conditions apply:
11688 Only one precompiled header can be used in a particular compilation.
11690 A precompiled header can't be used once the first C token is seen. You
11691 can have preprocessor directives before a precompiled header; you can
11692 even include a precompiled header from inside another header, so long as
11693 there are no C tokens before the @code{#include}.
11695 The precompiled header file must be produced for the same language as
11696 the current compilation. You can't use a C precompiled header for a C++
11699 The precompiled header file must be produced by the same compiler
11700 version and configuration as the current compilation is using.
11701 The easiest way to guarantee this is to use the same compiler binary
11702 for creating and using precompiled headers.
11704 Any macros defined before the precompiled header (including with
11705 @option{-D}) must either be defined in the same way as when the
11706 precompiled header was generated, or must not affect the precompiled
11707 header, which usually means that the they don't appear in the
11708 precompiled header at all.
11710 Certain command-line options must be defined in the same way as when the
11711 precompiled header was generated. At present, it's not clear which
11712 options are safe to change and which are not; the safest choice is to
11713 use exactly the same options when generating and using the precompiled
11717 For all of these but the last, the compiler will automatically ignore
11718 the precompiled header if the conditions aren't met. For the last item,
11719 some option changes will cause the precompiled header to be rejected,
11720 but not all incompatible option combinations have yet been found. If
11721 you find a new incompatible combination, please consider filing a bug
11722 report, see @ref{Bugs}.
11724 @node Running Protoize
11725 @section Running Protoize
11727 The program @code{protoize} is an optional part of GCC@. You can use
11728 it to add prototypes to a program, thus converting the program to ISO
11729 C in one respect. The companion program @code{unprotoize} does the
11730 reverse: it removes argument types from any prototypes that are found.
11732 When you run these programs, you must specify a set of source files as
11733 command line arguments. The conversion programs start out by compiling
11734 these files to see what functions they define. The information gathered
11735 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
11737 After scanning comes actual conversion. The specified files are all
11738 eligible to be converted; any files they include (whether sources or
11739 just headers) are eligible as well.
11741 But not all the eligible files are converted. By default,
11742 @code{protoize} and @code{unprotoize} convert only source and header
11743 files in the current directory. You can specify additional directories
11744 whose files should be converted with the @option{-d @var{directory}}
11745 option. You can also specify particular files to exclude with the
11746 @option{-x @var{file}} option. A file is converted if it is eligible, its
11747 directory name matches one of the specified directory names, and its
11748 name within the directory has not been excluded.
11750 Basic conversion with @code{protoize} consists of rewriting most
11751 function definitions and function declarations to specify the types of
11752 the arguments. The only ones not rewritten are those for varargs
11755 @code{protoize} optionally inserts prototype declarations at the
11756 beginning of the source file, to make them available for any calls that
11757 precede the function's definition. Or it can insert prototype
11758 declarations with block scope in the blocks where undeclared functions
11761 Basic conversion with @code{unprotoize} consists of rewriting most
11762 function declarations to remove any argument types, and rewriting
11763 function definitions to the old-style pre-ISO form.
11765 Both conversion programs print a warning for any function declaration or
11766 definition that they can't convert. You can suppress these warnings
11769 The output from @code{protoize} or @code{unprotoize} replaces the
11770 original source file. The original file is renamed to a name ending
11771 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
11772 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
11773 for DOS) file already exists, then the source file is simply discarded.
11775 @code{protoize} and @code{unprotoize} both depend on GCC itself to
11776 scan the program and collect information about the functions it uses.
11777 So neither of these programs will work until GCC is installed.
11779 Here is a table of the options you can use with @code{protoize} and
11780 @code{unprotoize}. Each option works with both programs unless
11784 @item -B @var{directory}
11785 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
11786 usual directory (normally @file{/usr/local/lib}). This file contains
11787 prototype information about standard system functions. This option
11788 applies only to @code{protoize}.
11790 @item -c @var{compilation-options}
11791 Use @var{compilation-options} as the options when running @command{gcc} to
11792 produce the @samp{.X} files. The special option @option{-aux-info} is
11793 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
11795 Note that the compilation options must be given as a single argument to
11796 @code{protoize} or @code{unprotoize}. If you want to specify several
11797 @command{gcc} options, you must quote the entire set of compilation options
11798 to make them a single word in the shell.
11800 There are certain @command{gcc} arguments that you cannot use, because they
11801 would produce the wrong kind of output. These include @option{-g},
11802 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
11803 the @var{compilation-options}, they are ignored.
11806 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
11807 systems) instead of @samp{.c}. This is convenient if you are converting
11808 a C program to C++. This option applies only to @code{protoize}.
11811 Add explicit global declarations. This means inserting explicit
11812 declarations at the beginning of each source file for each function
11813 that is called in the file and was not declared. These declarations
11814 precede the first function definition that contains a call to an
11815 undeclared function. This option applies only to @code{protoize}.
11817 @item -i @var{string}
11818 Indent old-style parameter declarations with the string @var{string}.
11819 This option applies only to @code{protoize}.
11821 @code{unprotoize} converts prototyped function definitions to old-style
11822 function definitions, where the arguments are declared between the
11823 argument list and the initial @samp{@{}. By default, @code{unprotoize}
11824 uses five spaces as the indentation. If you want to indent with just
11825 one space instead, use @option{-i " "}.
11828 Keep the @samp{.X} files. Normally, they are deleted after conversion
11832 Add explicit local declarations. @code{protoize} with @option{-l} inserts
11833 a prototype declaration for each function in each block which calls the
11834 function without any declaration. This option applies only to
11838 Make no real changes. This mode just prints information about the conversions
11839 that would have been done without @option{-n}.
11842 Make no @samp{.save} files. The original files are simply deleted.
11843 Use this option with caution.
11845 @item -p @var{program}
11846 Use the program @var{program} as the compiler. Normally, the name
11847 @file{gcc} is used.
11850 Work quietly. Most warnings are suppressed.
11853 Print the version number, just like @option{-v} for @command{gcc}.
11856 If you need special compiler options to compile one of your program's
11857 source files, then you should generate that file's @samp{.X} file
11858 specially, by running @command{gcc} on that source file with the
11859 appropriate options and the option @option{-aux-info}. Then run
11860 @code{protoize} on the entire set of files. @code{protoize} will use
11861 the existing @samp{.X} file because it is newer than the source file.
11865 gcc -Dfoo=bar file1.c -aux-info file1.X
11870 You need to include the special files along with the rest in the
11871 @code{protoize} command, even though their @samp{.X} files already
11872 exist, because otherwise they won't get converted.
11874 @xref{Protoize Caveats}, for more information on how to use
11875 @code{protoize} successfully.